A tagged strain approach reveals formation of Salmonella subpopulations during persistent infection

To define the Salmonella population structure that arises during chronic infection, we employed a previously established tagged strain approach using a mixture of barcoded, phenotypically equivalent S. Typhimurium strains [23]. These Salmonella wild-type isogenic tagged strains (WITS) each carry a unique 40 base pair tag in between the malX and malY pseudogenes, are equally fit, and have been applied to studies of acute infection [23]. Utilizing these previously published sequence tags, we constructed 8 WITS strains in the S. Typhimurium SL1344 background (W1–W8; Table S1) and confirmed each strain to be equally fit when grown in broth culture (Figure S1A). 129X1/SvJ mice, which possess a wild-type Nramp1 allele and can be persistently colonized with S. Typhimurium [31]–[33], were orally inoculated with 10⁸ colony forming units (CFU) of an equal mixture of strains W1–W8 (Figure S1B–C). Total WITS CFU were enumerated by plating (Figure S1D) and qPCR was performed to determine the WITS abundances in systemic (spleen, liver, gallbladder) and intestinal (PP, small intestine, cecum, colon, feces) sites after 35 days of infection. Individual mice had WITS profiles that were distinct from other animals, with certain WITS comprising the majority of Salmonella found within infected tissues that varied on a mouse-by-mouse basis (Figure 1A). However, combined analysis of all infected mice revealed that all 8 WITS strains were represented in every tissue compartment (Figure 1A) and there was no statistically significant difference between the relative abundances of the WITS strains in each of the tissues, indicating all 8 WITS are equally represented in vivo (Table S2, one-way ANOVA and Kruskal-Wallis tests). A control experiment in which 4 of the 8 WITS were underrepresented in the inoculum resulted in their subsequent underrepresentation within infected tissues (Figure S2), indicating that these 4 WITS did not have any fitness advantage during infection.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Distinct Salmonella subpopulations arise in systemic and intestinal tissues during chronic infection.

Mice were oral gavaged with 10⁸ S. Typhimurium consisting of an equal mixture of 8 WITS strains. WITS profiles in various tissues were determined by qPCR. Data shown represent 3 independent experiments (n = 19). A) Spleen, liver, gallbladder (gb), PP (pp), small intestine (si), cecum, colon, and fecal samples were collected after 35 days of infection. Bars depict the relative abundance of each WITS strain in different tissue compartments, each box contains the WITS profile for an individual mouse (1–19). Systemic tissues are highlighted in gray. Inset: composition of the WITS inoculum. B) Bray-Curtis dissimilarity values of WITS relative abundances in organ A versus organ B. Each circle represents an individual mouse (n = 19), lines represent medians. The lowest median dissimilarity in the depicted comparisons was observed between the colon and cecum (black circles). Intergroup differences were evaluated by paired t-tests. ** p = 0.0024, *** p = 0.0002, **** p<0.0001.

https://doi.org/10.1371/journal.ppat.1004527.g001

The WITS compositions in systemic and intestinal tissues were compared in order to determine whether Salmonella subpopulations arose after 35 days of persistent infection, a time after which the bacteria have breached the intestinal epithelium and have spread systemically to the liver and spleen. The strain composition within individual mice varied depending on the site of infection (Figure 1A). In order to quantify potential differences in WITS abundances, we utilized a Bray-Curtis dissimilarity statistic, which has been commonly used in community abundance analyses in ecology and studies of the microbiota [34]–[36]. This calculation was applied to our model to obtain population-level distance values of WITS compositions in different sites. Bray-Curtis values were calculated between the WITS relative abundances of two tissues (see Materials and Methods), in which a score of 0 indicates an identical WITS profile in both organs and a score of 1 indicates completely dissimilar populations. A dissimilarity matrix was calculated for all tissue comparisons (Table 1). The subpopulation in the liver closely matched that of the spleen with a low mean dissimilarity score of 0.248 (Figure 1B, Table 1), which is consistent with these environments being highly connected by migration pathways through the bloodstream and/or lymphatics.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Bray-Curtis dissimilarity values of WITS relative abundances within murine tissues during persistent infection.

https://doi.org/10.1371/journal.ppat.1004527.t001

In addition to colonizing systemic sites, Salmonella persisted within intestinal tissues for 35 days. However, in contrast to the spleen and liver, which contained an average of 3–4 WITS, the intestinal tissues were colonized by 1–2 strains (Figure 1A). This suggested that while there was some bottlenecking in dissemination to systemic sites, stronger selection pressures likely existed within intestinal tissues. Further analysis of the WITS profiles indicated that the strain compositions in proximal gut tissues (PP and small intestine) were dissimilar from those present in distal gut tissues (cecum and colon, Figure 1A), with dissimilarity scores of 0.416–0.575 (Figure 1B, Table 1, Figure S3A). In contrast, the WITS composition in the cecum and colon were very similar with a score of 0.101 (Figure 1B, Table 1), which was significantly lower than the dissimilarity scores observed in the proximal gut (Figure S3A). Together, these data suggest that during persistent infection, different subpopulations of Salmonella form between proximal and distal gut tissues.

It is thought that Salmonella in the liver and gallbladder reseed the intestinal tract via bile, followed by subsequent shedding in the feces. If the bile ducts provided highly connected migration pathways between these sites, the WITS profiles should be similar between systemic and intestinal tissues. Although not all of the mice were colonized by Salmonella in the gallbladder (Figure 1A), the WITS profiles in the gallbladder were most similar to the compositions of the spleen and liver from these mice (Table 1, Figure S3B). In contrast, the WITS compositions in the gallbladder were very different from the composition within the intestinal tissues (Figure 1B, Table 1, Figure S3B). In addition, the WITS compositions in the distal gut were distinct from those in the systemic tissues with high dissimilarity scores >0.816 (Figure 1B, Table 1). Collectively, analysis of the WITS compositions in various compartments within each infected mouse demonstrate that spatially delimited Salmonella subpopulations form during persistent infection, with systemic organs containing populations that are distinct from those in intestinal tissues.

Increases in fecal shedding are attributed to clonal expansion of colonic Salmonella

Since host-to-host transmission requires high levels of Salmonella shed in the feces [4], [33], we wished to elucidate the kinetics and population dynamics of Salmonella shedding. Fecal samples were collected at various time points throughout the 35-day infection period (Figure 2A). An average of 6–7 WITS were present in feces after one day of infection, indicating some initial bottlenecking effects in the oral infection route may have occurred (Figure 2A–B). However, even greater dynamic changes in WITS compositions were observed at early time points in infection, with different strains shed at 7 and 14 days post-infection compared to day 1 (Figure 2A). Importantly, there was a dramatic decrease in the number of strains detected in the feces to an average of 1–2 WITS, which did not change during the 35-day infection (Figure 2A–B). Importantly, the sharp decrease in the number of strains shed in the feces on day 7 correlated with an increase in total fecal Salmonella CFU (Figure 2B), suggesting that clonal expansion of dominant WITS strains was responsible for increased fecal shedding.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Clonal expansion of cecal and colonic Salmonella is responsible for increased fecal shedding.

Fecal shedding of S. Typhimurium in mice orally infected with an equal mixture of 8 WITS strains over 35 days of infection. Data shown is representative of 3 independent experiments (n = 19). A) Profiles of WITS shed in feces at the indicated time points post-infection as determined by qPCR. Individual mice were tracked, and the identifying mouse number correlates to the same animal in Figure 1. B) Left y-axis: number of WITS strains present at the specified time points post-infection (black line: mean, SD). Right y-axis: total Salmonella CFU enumerated from fresh fecal pellets collected throughout infection (gray columns: mean, SD). C) Bray-Curtis dissimilarity values between WITS relative abundances in organ A and organ B (feces). Each circle represents an individual mouse (n = 19), lines represent medians. Intergroup differences were evaluated by paired t-tests. ns = not significant, ** p = 0.0041, *** p = 0.0010, **** p<0.0001.

https://doi.org/10.1371/journal.ppat.1004527.g002

To ascertain the tissue compartment that served as the source of clonal Salmonella expansion, we compared the WITS relative abundance profiles of the feces to both systemic and intestinal tissues to identify similarities. Although Salmonella initially invade the PP, the WITS compositions in the PP compared to the feces were significantly different at 35 days post-infection (Figures 2C, Table 1). In addition, the compositions of the Salmonella populations within systemic sites compared to the population composition in the feces were even more dissimilar (Table 1). This further corroborated our earlier finding that distinct Salmonella subpopulations arose between systemic and intestinal compartments. Furthermore, we did not observe an increase in the number of WITS strains present during increased fecal shedding (Figure 2B), which would be expected to occur if increased reseeding of systemic Salmonella was the source. Instead, these analyses revealed that the WITS profiles in both the cecum and colon very closely matched the composition of Salmonella shed in the feces (Figure 2B; Table 1). Importantly, the dissimilarity values between the distal gut sites and the feces were significantly lower than that of any other tissue compartment analyzed (Figure 2C, Table 1, Figure S3). Taken together, our results indicate that a clonal expansion of cecal and colonic Salmonella is responsible for the increases in fecal shedding.

Systemic Salmonella are excluded from the distal gut and subsequent fecal shedding

The results of our WITS experiment demonstrated that distinct subpopulations formed in systemic and intestinal tissues by 35 days post-infection (Figure 1). However, even though high Bray-Curtis scores were computed between systemic and distal gut tissues, values were <1 indicating there were small percentages of shared WITS in these sites. One limitation of our mixed inoculum approach was that we could not discern the directionality of dissemination. For example, it could not be determined whether WITS present in the distal gut were part of the initial population or if they arrived secondarily by seeding the intestinal tract from systemic sites. In order to determine the relative contribution of systemic and intestinal strains to fecal shedding, it required a strategy to mark Salmonella in these different sites within the host.

To address this, we developed a co-infection model that employed isogenic marked strains rapidly identifiable by differential plating on antibiotics. We used the parental streptomycin-resistant SL1344 strain that has a missense mutation in hisG, which is not required for virulence, and an isogenic SL1344-kan^R strain containing a kanamycin resistance cassette inserted at this site (hisG::aphT). These strains are equally fit in single and in mixed infections in mice inoculated by oral or IP routes [33]. In our co-infections, each mouse received 10⁸ of one strain by oral inoculation and 10³ of the isogenic strain by intraperitoneal (IP) injection. The IP route bypasses the gastrointestinal tract, such that Salmonella colonize systemic tissues first [32]. To confirm that successful reseeding occurs in our model, Salmonella shedding and tissue burdens were compared in control mice that received single IP infections or those that received single oral infections. Systemic IP-delivered Salmonella reseeded the small intestine, where they reached the same range of fecal shedding levels by 14 days post-infection as mice infected orally (Figure S4). However, the oral inoculation route resulted in >1,000-fold more Salmonella fecal CFUs 1 day post-infection compared to the IP route, and reached peak fecal shedding levels more rapidly (Figure S4A). Thus, in the co-infection model, the oral strain establishes an infection in the gut before the systemic strain reaches the intestine, allowing us to test the strength of priority effects in Salmonella population assembly.

Mice injected IP with a single Salmonella strain shed this strain in the feces as soon as 1 day post-infection (Figure S4A). This was in contrast to what occurred in mice that had been co-infected orally with an isogenic WT strain (Figure 3). The systemic strain was detected in the feces of only 5 of the 54 mice throughout the 30 days of infection (Figure 3A–C). Importantly, shedding of the systemic strain only occurred on a single day and did not persist. Since mice shed variable levels of Salmonella [33], [37], we wondered whether this would influence the ability of the IP strain to be shed. Surprisingly, the oral strain was exclusively shed in the feces of low (<10⁴ CFU/gram), moderate (<10⁸ CFU/gram), and super (≥10⁸ CFU/gram) shedder mice (Figure 3A–C). In addition, when the reciprocal combination of strains (oral: SL1344-kan^R, IP: SL1344) was used the same result was obtained throughout 60 days of infection (Figure 3, Figure S5A). Taken together, these results indicate that the established intestinal strain prevents colonization of the cecum and colon by Salmonella disseminating from systemic tissues.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Established gastrointestinal populations of Salmonella are exclusively shed in feces.

Differentially marked, isogenic Salmonella strains were used in co-infections. Mice were given 10⁸ SL1344 by drinking and injected intraperitoneally (IP) with 10³ SL1344-kan^R immediately afterwards. The reciprocal combination of strains was also tested and included. A–C) Salmonella CFU/gram feces were determined from fecal pellets collected over 30 days of infection. SL1344-kan^R was identified by differential plating or patching onto LB agar containing 40 µg/ml kanamycin. Oral strain CFU (black) are plotted on the top half of the graph and the IP strain CFU (gray) on the bottom. Limit of detection for a single fecal sample is 10 CFU/gram feces. Each plot represents Salmonella fecal CFU (median, range) for mice shedding at A) low (n = 8), B) moderate (n = 34), and C) super shedder (n = 12) levels. D) Oral (black) and IP (gray) strain composition of Salmonella shed in feces after 30 days of co-infection (mean, SD). Data represent four independent experiments (n = 54). **** p<0.0001, Wilcoxon matched-pairs signed rank test.

https://doi.org/10.1371/journal.ppat.1004527.g003

We next wondered what the composition of the Salmonella strains were within systemic tissues of mice that had been co-infected for 30 days. In contrast to the cecum and colon, the IP and oral strains were both present within systemic tissues after 30 days of co-infection. The spleen and liver were comprised of similar abundances of both strains (Figure 4), indicating that intestinal Salmonella effectively disseminated to systemic sites. Although the orally inoculated strain was present in the gallbladder, the IP strain comprised >80.44% of the total Salmonella population in this organ (Figure 4). In addition, the IP strain was present as a minority of the population present in the PP (19%), small intestine (30%), and mLN (38%) (Figure 4). The IP strain was not detected in the cecum and colon in 25 out of 28 mice, and comprised <8% in the remaining animals (Figure 4). Strikingly, the oral strain remained dominant in the cecum and feces during the 60-day infection (Figure S5). Thus, our results from the co-infection model and the WITS analyses suggest that Salmonella that are established in the cecum and colon prevent systemic subpopulations from colonizing important niches that are required for fecal shedding.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Systemic Salmonella are excluded from the distal gut during competition with an established gut strain.

Mice were co-infected with SL1344 orally and SL1344-kan^R IP. Animals were sacrificed after 30 days of infection. Specified organs were collected and plated for determination of oral (black) and IP (gray) strain abundance. The reciprocal combination of strains was also tested and included. Data represent three independent experiments (n = 28). Percent strain composition of Salmonella in tissues. (mean, SD). ns = non-significant, *p = 0.0159, ** p = 0.0087, ***p = 0.0002, **** p<0.0001, Wilcoxon matched-pairs signed rank tests.

https://doi.org/10.1371/journal.ppat.1004527.g004

Increased Salmonella levels in the gallbladder does not lead to fecal shedding of systemic bacteria

One possible explanation for the dominance of the oral strain in the distal gut and feces could be that there is insufficient reseeding of systemic Salmonella into the gastrointestinal tract. To test this possibility, we utilized an established gallstone model of infection, in which S. Typhimurium biofilm formation on gallstones increased reseeding and subsequent fecal shedding by 1,000-fold [38]. We fed mice a lithogenic diet for 10 weeks to induce gallstone formation that resulted in 1–9 stones/mouse as confirmed by ultrasound imaging (Figure S6). In contrast, mice on a standard diet never developed gallstones (Figure S6). As previously demonstrated, mice with gallstones that were infected with 10³ S. Typhimurium by IP injection shed >1,000-fold higher levels of bacteria 7 days post-infection compared to control mice (Figure S7).

To determine whether increased levels of S. Typhimurium in the gallbladder would allow systemic bacteria to colonize the cecum and/or colon, mice with diet-induced gallstones were co-infected orally with SL1344 and IP with SL1344-kan^R. By 14 days post-infection, mice with gallstones had a mean Salmonella gallbladder burden >10,000-fold higher than mice without gallstones (Figure 5A). This represented an increase in systemic Salmonella, as the gallbladders were exclusively colonized by the IP strain (Figure 5B). In addition, mice with diet-induced gallstones had significantly higher levels of S. Typhimurium in the small intestine, indicating that increased numbers of systemic bacteria had reseeded this site (Figure 5B). Despite this drastic increase in the levels of systemic Salmonella reseeding the intestine, the established intestinal strain remained dominant in the cecum, colon, and feces (Figure 5C). Taken together, our data suggest that in the presence of an established Salmonella strain, systemic Salmonella are excluded from colonizing crucial transmission niches in the distal gut.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Gallstone-induced increases in systemic reseeding Salmonella are insufficient to displace the established population in the cecum and colon.

Mice on an 11-week lithogenic diet developed cholesterol gallstones (green, n = 13), while mice on the normal base diet were devoid of gallstones (white, n = 10). All mice were co-infected with 10⁸ SL1344 orally and 10³ SL1344-kan^R by IP injection. Feces and specified tissues were collected 14 days post-infection to calculate Salmonella burden and strain abundance. A) Total Salmonella CFU per gram gallbladder tissue (mean, SD), B) percent abundance of the IP strain in the gallbladder and small intestine, and C) percent abundance of the oral strain in the cecum, colon, and feces were determined in control mice and mice with gallstones. Each circle represents an individual mouse, lines represent means. Data are representative of two independent experiments. ns = non-significant, *p = 0.0224, ***p = 0.0004, ****p<0.0001, unpaired Mann-Whitney tests.

https://doi.org/10.1371/journal.ppat.1004527.g005

The established intestinal strain is resistant to super-colonization regardless of infection route

Although the presence of gallstones increased the numbers of S. Typhimurium in the gallbladder to 10³–10⁷ CFU/organ as well as subsequent reseeding of the small intestine, it is possible that these levels were insufficient to compete with the established intestinal subpopulations (Tables S3, S4). Indeed, we have measured the levels of Salmonella in gastrointestinal sites and found that there is a range of 10¹–10⁸ total CFU (Table S3). To address this issue, we performed sequential infections in which resident intestinal Salmonella were challenged with a high oral dose of a second strain. First, mice were inoculated with 10³ SL1344 by IP injection to establish a systemic infection. This initial Salmonella strain was detected in the feces after 5–7 days and was persistently shed for 35 days (Figure 6A). These mice were then super-infected with 10⁸ SL1344-kan^R orally. Although, the orally inoculated strain was detected in the feces 1 day post-infection (dpi), it was not detected in the feces for the remaining 7 days post-oral inoculation (35–42 dpi, Figure 6A). The challenging oral strain was not detected in any systemic or intestinal tissues by 7 days post-challenge (42 dpi, Figure S8A). This demonstrates that super-infecting strains are excluded from colonizing the intestine in the presence of a resident, persistent intestinal Salmonella infection, regardless of the route of inoculation. Collectively, our results suggest that there is intraspecies competition for a transmission niche in the distal gut.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. Established strains are resistant to super-colonization regardless of infection route.

Sequential infections in mice were performed using SL1344 as the initial strain (Strain 1, black) and SL1344-kan^R as the competing strain (Strain 2, gray). Individual mice were monitored for fecal Salmonella shedding at the indicated time points. Pellets were plated to determine total Salmonella CFU and to discern strain abundances. Limit of detection for a single fecal sample is 10 CFU. A) Mice were first infected with 10³ SL1344 by IP injection. After 35 days, mice were challenged with 10⁸ SL1344-kan^R by drinking. Left: Fecal CFU (mean, SD) of established IP and competing (oral) strains. Right: Geometric mean of strain CFU in feces. Data are representative of two separate experiments (n = 10). B) Mice were first inoculated with 10⁸ SL1344 (black) by drinking, which established a persistent infection for 102 days. Animals were subsequently challenged with 10⁸ SL1344-kan^R (gray) by drinking. Left: Fecal CFU (mean, SD) of established strain SL1344 and challenge strain SL1344-kan^R in feces. Right: Geometric mean of strain CFU in feces. Data are representative of two separate experiments (n = 45). C) Mice were first infected orally with 10⁸ of either SL1344 or SL1344-kan^R 14 days prior to cohousing. Left: schematic of cohousing experiment; a SL1344-kan^R super shedder donor (gray) was cohoused with SL1344-infected recipient mice (black) and a naïve mouse (white) as a control. Right: total Salmonella CFU/gram feces in cohoused mice. Gray asterisks indicate shedding levels of the SL1344-kan^R super shedder donor, which was removed after 10 days of co-housing. Lines depict the geometric mean Salmonella CFU/gram feces of recipient mice shedding the established SL1344 strain (black) or challenging donor SL1344-kan^R strain. Data are representative of two independent experiments (n = 2 donors, 8 recipients).

https://doi.org/10.1371/journal.ppat.1004527.g006

Priority effects govern gut colonization and fecal shedding of Salmonella

Based on our evidence of intraspecies competition for a distal gut niche, we proposed that the dominance of established Salmonella in the cecum and colon is attributed to priority effects that govern distal gut colonization and subsequent fecal shedding. To test this notion, we performed sequential S. Typhimurium infections to evaluate the duration and strength of these competitive interactions. Mice were infected with 10⁸ SL1344 orally, and fecal shedding of Salmonella was monitored. All of the mice continued to shed Salmonella over the 102 days of infection (Figure 6B). After 102 days, the mice were inoculated orally with 10⁸ CFU of a second competing strain, SL1344-kan^R. The competing strain was detected in the feces during the first 3 days post-infection (Figure 6B). However, by 14 days post-challenge (116 dpi), 28 of the 45 mice were no longer shedding the competing Salmonella strain (Figure 6B). Finally, by 35 days post-challenge (137 dpi), the competing strain was not detected in the feces (Figure 6B), intestinal compartments, or systemic tissues of co-infected mice (Figure S8B). The reciprocal strain combinations were also tested: mice were first infected with 10⁸ SL1344-kan^R orally for 60 days before subsequent challenge with 10⁸ SL1344, in which the competing strain was cleared from the feces by 20 days post-challenge (Figure S9A). Thus, this colonization resistance against the same Salmonella species was maintained during the chronic stages of infection.

We next sought to determine whether the levels of the initial oral strain (SL1344) in the colon and in the feces would influence the clearance kinetics of the second competing oral strain (SL1344-kan^R). One day after the second oral inoculation, the percentage of the competing strain varied depending on the level of shedding of the resident strain. For example, in mice that were shedding >10⁸ CFU/g feces (super shedder mice), the competing SL1344-kan^R strain comprised 4.88% of the total population on the first day post-secondary inoculation (Figure S9B). In contrast, for low and moderate shedder mice, the competing strain comprised 42.02% and 35.58% of the total population, respectively (Figure S9B). These differences remained significant 5 days after infection with the second competing SL1344-kan^R strain. However, by days 10 and 14, the second SL1344-kan^R strain was no longer detected in the feces of any of the mice (Figure S9B). These results indicate that more robust and rapid priority effects are exhibited in mice that are colonized with higher colonic Salmonella loads.

Finally, to determine whether we would see the same intraspecies priority effects in the distal gut during host-to-host transmission, we utilized our previously established model of transmission from an infected, super shedder mouse to uninfected mice in the same cage [33]. In this experiment, the donor mouse was orally infected with SL1344-kan^R and was shedding >10⁸ CFU/g at 14 days post-inoculation (Figure 6C). As a positive control for host-to-host transmission, the donor mouse was co-housed with uninfected mice. Similar to our previous results, naïve mice began shedding SL1344-kan^R within 24 hours and continued to shed even after the donor was removed (Figure S10A). In contrast, recipient mice that had been infected for 14 days with SL1344 required 10 days of cohousing before low levels of the donor strain (<0.02% of all Salmonella) were detected in the feces (Figure 6C). In addition, shedding of the donor strain in the previously infected recipient mice was transient, and was not detected in the feces or tissues 10 days post-cohousing (Figure 6C, Figure S8C). Similar results were obtained when a SL1344 super shedder was cohoused with SL1344-kan^R infected mice. Cohousing for 12 days was required before the donor strain could be detected in the feces of recipient SL1344-kan^R mice (Figure S10B). The super shedder donor was left in the cage for an additional 6 days before removal, but consistent with previous findings, the donor strain was not detected in the feces of recipient mice by day 23 post-cohousing (Figure S10B). Together, these experiments show that priority effects determine Salmonella population assembly in intestinal transmission niches, where established subpopulations exert colonization resistance against incoming challengers.

Ablation of the established intestinal subpopulation permits distal gut colonization and fecal shedding of systemic Salmonella

Since the established subpopulation of Salmonella in the cecum and colon exerts colonization resistance, we proposed that their removal would allow challengers to occupy vital transmission niches. To test this idea, mice co-infected with 10⁸ SL1344 orally and 10³ SL1344-kan^R IP for 7 days were then treated with a single dose of kanamycin. Kanamycin is not absorbed systemically and thus was used to ablate the extracellular, kanamycin-sensitive bacteria in the gastrointestinal tract. Within 24 hours of antibiotic administration, fecal shedding of the established intestinal SL1344 decreased by ∼5 logs (Figure 7A, left). Concomitant with the decrease in the established strain, over 10⁷ CFU of the systemic SL1344-kan^R strain was shed per gram of feces (Figure 7A, left). By 4 days post-antibiotic treatment, the systemic strain was exclusively shed in the feces (Figure 7A, left) and was transmitted to naïve recipients (Figure 7A, right). Thus, priority effects arose during the first 7 days of infection, which coincided with the clonal expansion in the distal gut and feces observed in the WITS studies (Figure 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 7. Inhibitory priority effects arise early during infection and are facilitated by Salmonella hydrogenase.

A) Left: Mice were co-infected with 10⁸ SL1344 orally and 10³ SL1344-kan^R IP. A single dose of 20 mg kanamycin in 200 µl water was delivered by oral gavage after 7 days of co-infection. Geometric means of oral (black) and IP (gray) strain CFU shed per gram feces. Data are representative of two independent experiments (n = 10). Right: A kanamycin-induced supershedder (SS, gray asterisks) shed and transmitted the IP strain exclusively (SL1344-kan^R). Naïve uninfected recipients (open gray circles) began shedding SL1344-kan^R as soon as 1 day post-cohousing, and continued to shed the strain throughout 21 days even after removal of the SS donor. Data are representative of two independent experiments (n = 2 donors, 8 naïve recipients). B) All co-infected mice received 10³ WT SL1344 by IP injection (gray Salmonella). Mice in the control group received 10⁸ WT SL1344-kan^R orally (white Salmonella), and mice in the Δhyb group received 10⁸ ΔhybΔSPI1ΔSPI2 orally (red Salmonella). Data are representative of two independent experiments (control n = 10, Δhyb n = 7). Percent of fecal Salmonella comprised of the IP strain (WT SL1344) at the indicated days post-infection. Limit of percent IP strain detection was 0.0048% by differential plating, nd = not detected. ****p<0.0001, unpaired Mann-Whitney test. C) Salmonella strain composition in feces of Δhyb co-infected mice. Geometric means of ΔhybΔSPI1ΔSPI2 (oral) and WT SL1344 (IP) strains shed in feces over 15 days of co-infection. D) Percent of Salmonella comprised of ΔhybΔSPI1ΔSPI2 (oral) and WT SL1344 (IP) after 15 days of co-infection. gb = gallbladder, mln = mLN, si = small intestine. Each circle represents an individual animal, lines at medians. Limit of detection was 0.22% by differential plating. *p = 0.0156, ***p = 0.0006, Wilcoxon matched-pairs signed rank tests.

https://doi.org/10.1371/journal.ppat.1004527.g007

Based on these findings, we hypothesized that Salmonella strains were competing for limited nutrient or spatial resources within the cecum and colon, which inhibited the ability of systemic strains to colonize the distal gut. We tested this notion by gavaging co-infected mice (SL1344 oral, SL1344-kan^R IP) with 5 mg streptomycin in order to disrupt the microbiota and make more of these resources available [37], [39]–[41]. Both Salmonella strains are streptomycin-resistant, and previous studies have shown that streptomycin treatment of infected mice increases Salmonella fecal shedding to super shedder levels [33], [37]. We observed that all streptomycin-treated mice became super shedders, yet the increase in fecal Salmonella CFU reflected expansion of the oral SL1344 strain (Figure S11). This indicated that disrupting the microbiota with streptomycin treatment was insufficient to permit shedding of the systemic strain, as the newly available resources were likely immediately utilized by established intestinal Salmonella. Furthermore, since kanamycin does not enter mammalian cells, these results collectively indicate that established intestinal Salmonella occupy an extracellular transmission niche in the distal gut and exclude the bacteria that are reseeding the intestine from systemic sites.

Early inhibitory priority effects in the gastrointestinal tract are facilitated by intraspecies competition for a nutritional niche

Our data show that intraspecies priority effects govern Salmonella population assembly in the distal gut. Since our previous results demonstrated that clonal expansion and priority effects in the cecum and colon could occur by 7 days post-infection (Figure 2, 7A), we hypothesized that nutrient acquisition was very important during this stage of colonization. Indeed, ecological theory has implicated competition for nutrients as an important determinant in priority effects and community structure [28].

S. Typhimurium hydrogenase (hyb) is a key mediator of cecal ecosystem invasion and is required to consume a microbiota-derived metabolite [30]. In the un-inflamed gut of conventional mice with complex microbiota, hydrogenase enzymes facilitate consumption of hydrogen (H₂) intermediates in a SPI1- and SPI2-independent manner [30]. Similarly, we show here that Hyb is important for gut colonization and fecal shedding in 129SvJ mice with an intact conventional microbiota (Figure S12). To test the role of Hyb in intraspecies priority effects, co-infections were performed in which all mice were injected IP with 10³ wild-type (WT) SL1344 bacteria and one group of mice was co-inoculated orally with 10⁸ ΔhybΔSPI1ΔSPI2 isogenic mutant S. Typhimurium while control mice were co-inoculated orally with 10⁸ WT SL1344-kan^R. The hyb mutation was constructed in a ΔSPI1ΔSPI2 background to assess the need for hydrogenase in the context of a non-inflamed gut. The relative levels of each strain in the feces were monitored over 15 days of co-infection (Figure 7B–C). Importantly, the Salmonella shed in the feces at 4 and 7 days contained systemic WT bacteria and by 15 days post-infection were entirely comprised of the systemic WT strain in mice that received ΔhybΔSPI1ΔSPI2 orally (Figure 7B). Total levels of fecal Salmonella were significantly lower in the ΔhybΔSPI1ΔSPI2 co-infection group compared to controls (Figure S13A), which corresponds to the decreased fecal shedding of ΔhybΔSPI1ΔSPI2 mutants during single oral infections (Figure S12B). The strain compositions in the feces of these mice throughout infection indicated that the increase in total fecal CFU on day 15 reflected rapid reseeding and shedding of the systemic WT strain concomitant with declining levels of the oral ΔhybΔSPI1ΔSPI2 strain (Figure 7C). Taken together, we have demonstrated that the hydrogenase mutant was unable to effectively invade the cecal and colonic niche (Figure 7D), thereby nullifying any priority effects and allowing systemic Salmonella to colonize the distal gut with subsequent transmission in feces (Figure 7C–D).

S. Typhimurium SPI1 and SPI2 are required to maintain intraspecies colonization resistance during persistent infection

To determine whether intraspecies colonization resistance was still dependent on the maintenance of the extracellular intestinal niche during persistent infection, co-infected mice (oral: SL1344, IP: SL1344-kan^R) were treated with a single dose of kanamycin 42 days post-infection. Within 24 hours of antibiotic administration, fecal shedding of the established intestinal SL1344 was decreased by ∼6 logs concomitant with a ∼5 log increase in the systemic SL1344-kan^R strain (Figure 8A, left). The systemic strain was exclusively shed in the feces 2 days post-antibiotic treatment (Figure 8A, left), and comprised the entire Salmonella population in the cecum and colon after 7 days (Figure 8A, right). These data thus indicate that the extracellular niche in the cecum and colon is required to maintain intraspecies colonization resistance during persistent infection, which actively inhibits successful fecal shedding of systemic Salmonella.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 8. S. Typhimurium SPI1 and SPI2 are required to maintain intraspecies colonization resistance in the intestinal tract during persistent infection.

A) Mice were co-infected with 10⁸ SL1344 orally and 10³ SL1344-kan^R IP. A single dose of 20 mg kanamycin in 200 µl water was delivered by oral gavage 42 days post-infection. Data are representative of two independent experiments (n = 10). Left: Plot depicts geometric means of oral (black) and IP (gray) strains in feces. Right: Percent of Salmonella comprised of oral and IP strains in gut tissues 7 days after kanamycin treatment. Each circle represents an individual animal. ****p<0.0001, unpaired Mann-Whitney tests. B) All co-infected mice received 10³ WT SL1344 by IP injection (gray Salmonella). Mice in the control group received 10⁸ WT SL1344-kan^R orally (white Salmonella), while mice in the ΔSPI group received 10⁸ ΔSPI1ΔSPI2 orally (blue Salmonella). Data are representative of two independent experiments (n = 10/group). Percent of fecal Salmonella comprised of the IP strain (WT SL1344) at the indicated days post-infection. Limit of percent IP strain detection was 0.0011% by differential plating, nd = not detected. ns = not significant, **p = 0.0075, ****p<0.0001, unpaired Mann-Whitney tests. C) S. Typhimurium strain composition in feces of mice in the ΔSPI co-infection group. Geometric means of the ΔSPI1ΔSPI2 (oral) and WT SL1344 (IP) strains shed in feces over 70 days of co-infection. D) Percent of Salmonella comprised of ΔSPI1ΔSPI2 (oral) and WT SL1344 (IP) in intestinal tissues after 70 days of co-infection. mln = mLN, si = small intestine. Each circle represents an individual animal, lines at medians. Limit of detection was 0.1% by differential plating. *p = 0.0156, **p = 0.0020, Wilcoxon matched-pairs signed rank tests.

https://doi.org/10.1371/journal.ppat.1004527.g008

To gain more insight into how S. Typhimurium competitively excludes incoming challengers from colonizing the distal gut niche, we tested the potential role of the key virulence factors Salmonella Pathogenicity Islands SPI1 and SPI2, which encode type III secretion systems that deliver effector proteins required for persistence in host tissues [32], [42]–[44] and fecal transmission [33]. Co-infections were performed in which mice simultaneously received 10³ WT SL1344 by IP and 10⁸ isogenic ΔSPI1ΔSPI2 mutant bacteria orally. In the ΔSPI1ΔSPI2 co-infected mice, the IP-injected WT bacteria were not present in significant numbers at day 7 (Figure 8B). However, by day 25, 21.43% of all fecal Salmonella were WT bacteria, and by day 70, 98.86% were WT S. Typhimurium (Figure 8B). In addition, the total fecal Salmonella CFU in the control (SL1344 oral, SL1344-kan^R IP) and the ΔSPI1ΔSPI2 co-infected mice were similar (Figure S13B), which is consistent with our result that the systemic WT strain reseeded and replicated within the intestinal tract once the ΔSPI1ΔSPI2 mutant was cleared (Figure 8C). Indeed, examination of strain abundances in intestinal tissues after 70 days of co-infection confirmed that the systemic IP strain had predominantly colonized the mLN, small intestine, cecum, and colon while the initial ΔSPI1ΔSPI2 mutant was cleared from these sites (Figure 8D). These studies demonstrate that SPI1 and SPI2 are required for the established intestinal Salmonella population to maintain active colonization resistance against systemic reseeding bacteria.

Ethics statement

Experiments involving animals were performed in accordance with NIH guidelines, the Animal Welfare Act, and US federal law. All animal experiments were approved by the Stanford University Administrative Panel on Laboratory Animal Care (APLAC) and overseen by the Institutional Animal Care and Use Committee (IACUC) under Protocol ID 12826. Animals were housed in a centralized research animal facility certified by the Association of Assessment and Accreditation of Laboratory Animal Care (AAALAC) International.

Mouse strains and husbandry

129X1/SvJ and 129S1/SvImJ mice were obtained from Jackson Laboratories (Bar Harbor, ME). Male and female mice (5–7 weeks old) were housed under specific pathogen-free conditions in filter-top cages that were changed weekly by veterinary personnel. Sterile water and food were provided ad libitum. Mice were given 1 week to acclimate to the Stanford Research Animal Facility prior to experimentation.

Bacterial strains and growth conditions

The S. Typhimurium strains used in this study were derived from the streptomycin-resistant parental strain SL1344 [77]. A missense mutation (hisG46) in SL1344 results in histidine auxotrophy [78]. The isogenic SL1344-kan^R strain was created by replacing the hisG coding sequence with that of a kanamycin-resistance casette (hisG::aphT) using the methods of Datsenko and Wanner [33], [79]. Genetic manipulations were originally made in the S. Typhimurium LT2 background before being transferred to SL1344 by P22 transduction. This methodology was also used to construct wild-type isogenic tagged Salmonella (WITS strains: W1–W8), in which a unique 40-bp signature tag and the kanamycin-resistance cassette were inserted between the malX and malY pseudogenes. Grant et. al. previously established this approach and published the unique 40-bg sequence tags of 8 WITS strains [23], which were employed in this study (Table S1). Growth curves of W1–W8 in LB broth cultures were performed by optical density readings and plating for colony forming units (CFU) per milliliter (Figure S1A). ΔSPI1ΔSPI2 (orgA::tet, ssaV::kan) was generated previously for use in other studies [80]. The Δhyb (hypOhybABC::cm) deletion was constructed as described by Maier et. al., with P22 phage transduction to insert the deleted genomic region into the ΔSPI-1ΔSPI-2 strain ([30], Table S1). All constructs were verified by PCR.

All S. Typhimurium strains were grown at 37°C with aeration in Luria-Bertani (LB) medium containing the appropriate antibiotics: streptomycin (200 µg/ml), kanamycin (40 µg/ml), tetracycline (15 µg/ml) and chloramphenicol (8 µg/ml). For mouse inoculation, an overnight culture of bacteria was spun down and washed with phosphate-buffered saline (PBS) before resuspension to obtain the desired concentration.

Mouse infections

Food was removed 16 hours prior to all mouse infections. In WITS experiments, mice were inoculated with an equal mixture of strains W1–W8 via oral gavage of 10⁸ CFU in 100 µl PBS. For intraperitoneal (IP) infections, mice were injected with 10³ CFU in 100 µl PBS as previously described [32]. In the co-infection model, mice drank an oral dose of 10⁸ SL1344 in 20 µl PBS, then received an IP injection of 10³ SL1344-kan^R immediately afterwards. Co-infection experiments were repeated using the reciprocal combination of strains, SL1344-kan^R (oral) and SL1344 (IP), which had no effect on the trends observed.

Monitoring fecal shedding of S. Typhimurium

Individual mice were identified by distinct tail markings and tracked throughout the duration of infection. Between 2–3 fresh fecal pellets were collected directly into eppendorf tubes and weighed at the indicated time points. Pellets were resuspended in 500 µl PBS and CFU/gram feces were determined by plating serial dilutions on LB agar plates with the appropriate antibiotics. Low (<10⁴ CFU/gram), moderate (<10⁸ CFU/gram), and super shedder (≥10⁸ CFU/gram) mice were identified based on previously established criteria [33], [37].

S. Typhimurium burden in blood and tissues

Following collection of fresh fecal pellets, animals were sacrificed at the specified time points. Blood was collected by cardiac puncture and animals were euthanized by cervical dislocation. Sterile dissection tools were used to isolate individual organs, which were weighed prior to homogenization. The entire gastrointestinal tract was removed, and the small intestine was immediately separated from the distal gut and transferred to a new sterile petri dish. Visible PP (3–6/mouse) were isolated from the small intestine using sterile fine-tip straight tweezers and scalpels. PP, mLN, spleens, livers, and gall bladders were collected in 1 ml PBS. The small intestine, cecum, and colon were collected in 3 ml PBS. Homogenates were then serially diluted and plated onto LB agar containing the appropriate antibiotics to enumerate CFU/gram tissue. For co-infections with SL1344 and SL1344-kan^R, several dilutions were plated to ensure adequate colonies (>100 CFU per sample) for subsequent patch plating to determine strain abundance.

Genomic DNA extraction and WITS qPCR

For WITS experiments, 300 µl of tissue homogenate was inoculated into LB broth containing streptomycin (200 µg/ml) and kanamycin (40 µg/ml) as a recovery method to enrich for low abundance strains. An UltroSpec 2100pro spectrophotometer (Amersham Biosciences, Piscataway, NJ) was used to obtain optical density readings of the resulting bacterial cultures. Genomic DNA (gDNA) was extracted from 2×10⁹ S. Typhimurium from each sample in duplicate using a DNeasy blood and tissue kit (Qiagen, 69506) as per the manufacturer's protocol for Gram-negative bacteria.

All qPCRs were performed on an Applied Biosystems 7300 real-time PCR system. A 25 µl reaction contained 12.5 µl of FastStart SYBR Green Master Mix with Rox (Roche, 04913914001), 8 µl DNase/RNase-free water, 0.75 µl of forward and reverse (10 µM) primers (Table S1), and 3 µl of gDNA (1–10 ng). Standard curves were generated using gDNA from each W1–W8 strain. Reaction conditions were 50°C for 2 min; 95°C for 10 min; 40 cycles of 95°C for 15 s and 60°C for 1 min; followed by a dissociation stage of 95°C for 15 s, 60°C for 1 min, 95°C for 15 s, and 60°C for 15 s.

Determining relative abundance of strains

To determine presence of a WITS strain, the qPCR value had to be above a minimum threshold value. This measure of primer specificity was determined by a negative control matrix, in which a specific primer pair was tested on ∼11.25 ng of non-template gDNA from each of the other 7 WITS strains. To test primer sensitivity, detection limits were determined by test plates containing known CFU of each strain. Briefly, colonies were washed off the plates with PBS and gDNA was extracted from plates with varying abundances of WITS (i.e. 1 CFU Strain A with 10³–10⁵ CFU Strain B). qPCR was performed and revealed a detection limit of 1 CFU/strain amidst over 4800 CFU from non-target strains. To verify that our method of broth recovery and qPCR analyses accurately rendered WITS abundances, we compared relative abundances of an equal mixture of culture-grown W1–W8 as determined by our qPCR strategy versus plating CFU of individual dilutions of each strain (Fig. S1B).

Plating onto selective LB agar containing streptomycin (20 µg/ml) and kanamycin (40 µg/ml) was used to determine strain abundances in co-infections, sequential challenges, and transmission experiments. In addition to patch plating a minimum of 100 CFU per sample, undiluted samples were plated on selective plates to increase detection limits. For super shedder mice, this permitted detection of a strain comprising just 0.00000001% of the total S. Typhimurium population.

Determining equal fitness of WITS in vivo

The strain relative abundances were determined for each tissue in all of the 19 mice infected with the 10⁸ equal mixture of 8 WITS. The relative abundances of each WITS strain were analyzed by one-way ANOVA (parametric) and Kruskal-Wallis (non-parametric) tests in Prism statistical software. These analyses were performed for each tissue collected from infected mice. Non-significant P values indicated that a particular WITS was not under or over represented in any tissue type (Table S2). To further verify that certain WITS strains were not preferentially selected for, a control experiment was performed in which mice were orally infected with an inoculum comprised of a skewed WITS mixture (Figure S2A). Underrepresented strains: W2, W3, W5, W6 (4.17%–7.32% of inoculum), overrepresented strains: W1, W4, W7, W8 (17.39%–20.94% of inoculum). Relative abundances of WITS in infected tissues were determined by qPCR after 35 days of infection. For each of the 8 WITS, defined bins were constructed for a range of strain relative abundances, with which the observed frequencies were used to generate histograms (Figure S2B).

Bray-Curtis dissimilarity analyses of WITS relative abundances in mouse tissues

Bray-Curtis dissimilarity scores were computed to quantitatively compare Salmonella population compositions in different sites. The relative abundance (y) of each WITS (n) was compared between two tissue sites i and j. The Bray-Curtis dissimilarity (d^BCD) was calculated by:A value of 0 indicates an identical WITS composition between two sites, while a value of 1 signifies that two samples are completely dissimilar without any overlap in WITS representation.

Gallstone model

The lithogenic diet established by Crawford et. al. ([38]) was modified in our experiments to include the normal rodent diet (Harlan, Teklad 2018) supplied in the Stanford Research Animal Facility. Mice were fed normal base chow supplemented with 1% cholesterol and 0.5% cholic acid (Harlan, Teklad custom research diet) for 10 weeks to induce cholesterol gallstone formation. Mice on control and lithogenic diets were anesthetized with isoflurane and shaved in the abdominal area for ultrasound imaging. A Vevo 2100 system (VisualSonics) was used to confirm gallstone formation. Mice were given 1 week to recover prior to infection with S. Typhimurium.

Sequential infections with established and competing strains

For sequential infections in which the IP strain served as the initial strain, mice were first injected with 10³ SL1344 and the infection was allowed to establish for 35 days. Following that time period, mice were challenged with an oral dose of 10⁸ SL1344-kan^R. In experiments with sequential oral infections, mice first received 10⁸ SL1344 orally by drinking. A persistent infection was allowed to establish for 102 days before oral challenge with 10⁸ SL1344-kan^R. This sequential oral infection was performed with the reciprocal order of strains, in which SL1344-kan^R was given as the initial strain and SL1344 given as the challenge strain.

Cohousing experiments

Mice were infected orally with either 10⁸ SL1344 or SL1344-kan^R and fecal shedding of Salmonella was monitored over 14 days prior to the start of the experiment. A SL1344-kan^R super shedder donor was then cohoused with mice previously infected with SL1344, in addition to a naïve uninfected mouse as a control. Cohousing was continued for 10 days before the super shedder donor was removed. The reciprocal cohousing experiments were performed in which a SL1344 super shedder donor was cohoused with mice previously infected with SL1344-kan^R.

Kanamycin treatment

The aminoglycoside was administered orogastrically in a single dose of 20 mg (Sigma Aldrich, K4000) dissolved in 200 µl of water. Mice were transferred to new cages with autoclaved bedding, chow (Harlan, Teklad 2018S), and water at the time of administration.

Statistical analyses

Prism (GraphPad) was used to create all figures and perform all statistical analyses. Intergroup comparisons of Bray-Curtis dissimilarity values (e.g. spleen-cecum versus colon-cecum) were analyzed by paired t-tests. Comparisons of oral and IP strain abundances within the same group of mice were evaluated with Wilcoxon matched-pairs signed rank tests. Differences in CFUs and strain composition between groups were examined by unpaired nonparametric Mann-Whitney tests. Significance was defined by p≤0.05.