The Abundance and Organization of Salmonella Extracellular Polymeric Substances in Gallbladder-Mimicking Environments and In Vivo

ABSTRACT

Salmonella enterica serovar Typhi causes chronic infections by establishing biofilms on cholesterol gallstones. The production of extracellular polymeric substances (EPSs) is key to biofilm development, and biofilm architecture depends on which EPSs are made. The presence and spatial distribution of Salmonella EPSs produced in vitro and in vivo were investigated in Salmonella enterica serovar Typhimurium and S. Typhi biofilms by confocal microscopy. Comparisons between serovars and EPS-mutant bacteria were carried out by examining growth on cholesterol-coated surfaces, with human gallstones in ox or human bile, and in mice with gallstones. On cholesterol-coated surfaces, no major differences in EPS biomass were found between serovars. Cocultured biofilms containing wild-type (WT) and EPS-mutant bacteria demonstrated WT compensation for EPS mutations. Analysis of biofilm EPSs from gallbladder-mimicking conditions found that culture in human bile more consistently replicated the relative abundance and spatial organization of each EPS on gallstones from the chronic mouse model than culture in ox bile. S. Typhimurium biofilms cultured in vitro on gallstones in ox bile exhibited colocalized pairings of curli fimbriae/lipopolysaccharide and O-antigen capsule/cellulose, while these associations were not present in S. Typhi biofilms or in mouse gallstone biofilms. In general, the inclusion of human bile with gallstones in vitro replicated biofilm development on gallstones in vivo, demonstrating the strength of this model for studying biofilm parameters or EPS-directed therapeutic treatments.

INTRODUCTION

Salmonella enterica subsp. enterica serovar Typhi causes both acute and chronic disease. While acute typhoid fever is treatable with antibiotics (1), untreated, improperly managed, and asymptomatic infections allow for chronic infections (2, 3), which develop in 3 to 5% of cases (4, 5). Although symptoms often subside, chronically infected individuals are able to intermittently shed the pathogen and transmit it to others by the fecal-oral route (6–8). These individuals, termed carriers, are the only known biological reservoir of S. Typhi (9, 10). Therefore, eliminating chronic infections will be an essential step toward eradicating endemic typhoid fever, which is currently reported in 14.3 million individuals and to which 136,000 deaths each year are attributed (10). Treating chronic infections has proven to be challenging due to the high degree of antibiotic tolerance S. Typhi exhibits upon production of biofilms during chronic infection (3). This complication makes surgical resection of the infected site(s) (ordinarily the gallbladder and biliary tract) the most-effective method currently available for treating chronic carriers (3). However, the risks associated with surgery, the feasibility of conducting operations in areas where S. Typhi is endemic, and the biological consequences of cholecystectomy highlight major limitations to this approach (3, 11–14).

The production of biofilm extracellular polymeric substances (EPSs) by S. Typhi is fundamental to establishing and perpetuating chronic infection (15). The primary site of biofilm development is cholesterol gallstones, where the pathogen must manage multiple stressors such as bile salts and bile acids, which have potent detergent properties and can denature proteins, chelate calcium and iron, and damage DNA (16, 17). Additionally, during chronic infection, the gallbladder becomes highly infiltrated with innate and adaptive immune cells (18), and reactive oxidative species from macrophages and neutrophils present an additional antimicrobial defense that S. Typhi must tolerate (19). However, EPSs are responsible for biofilm tolerance to host immunity and antibiotics, making chronic disease eradication problematic. Salmonella has the ability to produce several EPSs, including carbohydrates, proteins, lipids, and extracellular DNA (20, 21), although the most prominent EPSs are the O-antigen capsule, curli fimbriae, colanic acid, cellulose, and, for S. Typhi, Vi antigen (22–26). A better understanding of these EPSs would enhance the rational drug design of therapeutics that facilitate biofilm clearance by specifically targeting EPSs essential for biofilm survival (27–31). However, studies that identify EPSs in vitro are significantly limited by the knowledge that EPS production—and therefore the ability to contribute to biofilm tolerance in vivo—is dependent on environmental conditions (24, 32, 33). Thus, laboratory models that do not attempt to replicate the gallbladder environment should be interpreted cautiously.

Here, we evaluate Salmonella biofilm growth models in order to determine the expression, magnitude, and localization of EPSs produced by both S. Typhimurium and S. Typhi in various gallbladder-mimicking environments. Using confocal laser scanning microscopy (CLSM), we have quantified the relative abundance and distribution of each EPS produced by Salmonella biofilms under different growth conditions in vitro as well as in vivo. Our data demonstrate that Salmonella biofilms cultured in vitro in human bile are similar to chronic S. Typhimurium biofilms in vivo in terms of the relative amount of EPS detected and EPS organization. Furthermore, we report how EPS relationships differ between experimental models as well as between the two serovars, and we discuss the possible implications of differences for targeting EPSs during chronic infection.

RESULTS

Evaluation of Salmonella biofilms on cholesterol-coated surfaces.

Salmonella biofilms were initially cultured on cholesterol-coated glass surfaces by using commercially available reagents and standard laboratory conditions, representing the simplest model for studying EPS development in biofilms. CLSM analysis was conducted by z-stack imaging of each EPS signal in the sample to generate 3-dimensional biofilm models (see Fig. S1 in the supplemental material). Analysis of wild-type (WT) Salmonella demonstrated that S. Typhimurium and S. Typhi produce O-antigen capsule, curli fimbriae, and cellulose (as well as Vi antigen for S. Typhi) with similar biofilm biomasses under these conditions (Fig. 1A) and that Vi antigen production accounts for the additional biomass observed in S. Typhi. Although there were considerably fewer live bacteria (63.5%) in S. Typhimurium biofilms than in S. Typhi biofilms (Fig. 1B), this result was not statistically significant. Strains with mutations in individual EPS genes or with combined EPS gene mutations were also examined. The biofilms formed by each EPS-deficient mutant had at least as many live cells as its WT background (Fig. 1A and B). Interestingly, the S. Typhimurium strain expressing Vi antigen produced O-antigen capsule, curli fimbriae, and cellulose in amounts similar to those of both serovars and Vi antigen in an amount proportional to that in S. Typhi, while the detection of live bacteria increased (Fig. 1A and B). Average and maximum thickness measurements in both WT strains and S. Typhimurium expressing Vi antigen (plasmid-borne expression) also demonstrate that all EPSs were distributed equally throughout the biofilms (Fig. 1C to F), which is also evident by maximum0intensity projections from z-stack orthogonal views of the biofilm EPSs (Fig. S1).

FIG 1.

Quantification of EPSs in Salmonella biofilms cultured on cholesterol-coated surfaces. Biofilms were cultured in a chambered coverglass treated first with a cholesterol coat and then with antibodies or stains specific for the detection of each EPS and for live bacteria. (An anti–Vi antigen antibody was applied only to biofilms produced by S. Typhimurium expressing Vi antigen [ViAg⁺], S. Typhi, or S. Typhi ΔtviB.) CLSM images were captured as z-stacks (n = 3) using a Zeiss LSM 800 confocal microscope at ×63 magnification. ImageJ software running Comstat2 was used to calculate the biomass of each EPS (A), the biomass of live cells in a biofilm (B), the average biomass thickness of each EPS (C and D), and the maximum biomass thickness of each EPS (E and F). Statistical analysis of biomass for each EPS (A) was conducted by comparing mutant with WT biofilms. Comparisons were made by two-way analysis of variance (ANOVA) with Dunnett’s correction for multiple comparisons. Significant differences in live cells (B) were tested for using one-way ANOVA with Tukey’s correction for multiple comparisons. Differences in average or maximum biomass thickness between EPSs within each strain (C to F) were identified by two-way ANOVA with Tukey’s correction for multiple comparisons. Error bars indicate standard deviations. Asterisks indicate significant differences (*, P < 0.05; **, P < 0.01; ***, P < 0.0005; ****, P < 0.0001).

Salmonella produces multiple EPSs. However, it is not known whether Salmonella compensates for the loss of an EPS by changing the expression of other EPSs and, if so, in which EPS-deficient mutant(s) this compensation can occur. As expected, CLSM detection of curli fimbriae, O-antigen capsule, cellulose, and Vi antigen was significantly reduced in strains with mutations to genes necessary for the production of each of these EPSs (the ΔcsgA, ΔyihO, ΔbcsE, and ΔtviB strains, respectively, as well as the ΔcsgA ΔyihO ΔwcaM ΔbcsE strain) (Fig. 1A). Notably, the mutant strain that is unable to produce curli fimbriae (ΔcsgA) exhibited a 1.4-fold increase in biofilm biomass due to significantly increased production of the O-antigen capsule and cellulose (Fig. 1A). The additional biomass was also associated with an increase in the average and maximum thicknesses of both EPSs; the O-antigen capsule was significantly thicker than cellulose (Fig. 1C and E). Regrettably, no reagents are available for the specific detection of colanic acid. However, mutant strains that are unable to produce colanic acid (ΔwcaM) unexpectedly showed a loss of the O-antigen capsule similar to that in the ΔyihO mutant. In both the ΔwcaM and ΔyihO mutants, elimination of the O-antigen capsule did not have a significant effect on the biomasses of curli fimbriae and cellulose, which resembled the corresponding WT EPSs in terms of biomass and coequal thickness distributions (Fig. 1A, C, and E). For the ΔwcaM mutant, the minimal amount of O-antigen capsule detected was distributed throughout the biofilm with other EPSs (Fig. 1C and E; Fig. S1). The elimination of cellulose (by the ΔbcsE mutation) from the biofilm did not change the remaining total biomass and did not cause significant differences in curli fimbriae or the O-antigen capsule (Fig. 1A; Fig. S1). Similarly, the elimination of the Vi antigen in biofilms produced by S. Typhi ΔtviB did not cause any significant changes to other biofilm EPSs, total biomass, or the distribution of other EPSs throughout the biofilm (Fig. 1A, D, and F; Fig. S1).

We demonstrated previously, by using the same biofilm culture techniques, that the O-antigen capsule, colanic acid, and the Vi antigen are required for Salmonella tolerance of hydrogen peroxide (H₂O₂) (34). Furthermore, when WT and mutant bacteria are cocultured in biofilm aggregates, WT-produced colanic acid and Vi antigen protect all biofilm-resident bacteria regardless of EPS-producing ability, but biofilms containing both the WT and mutants unable to produce the O-antigen capsule are not protected by the WT and remain susceptible (34). Given the novel community benefit for colanic acid and the Vi antigen (but not for the O-antigen capsule) and the discovery here that mutants unable to produce colanic acid also limit O-antigen capsule production, we investigated how EPS production in cocultured biofilms would compare to that in WT biofilms. Biofilms were cultured using a 1:1 input mixture of WT and EPS-deficient bacteria. We hypothesized that WT EPS production would compensate for the mutants, allowing WT-like biofilm levels for all WT-mutant combinations.

Prior to the completion of EPS quantification in coculture biofilms, the assumption that a 1:1 input of planktonic bacteria upon inoculation of biofilm cultures would lead to equal amounts of WT and EPS-deficient bacteria in mature biofilms was verified using bacteria constitutively expressing green fluorescent protein (GFP) or mCherry fluorescent protein (mCFP). First, WT salmonellae were made mCFP positive by transformation with pFPV-mCherry, and mutants were made GFP positive by transformation with pFPV-25.1; then each combination of the WT and a mutant was cocultured for CLSM analysis. This demonstrated that the amounts of WT and EPS-deficient bacteria in mature biofilms are approximately equal to the input ratio used in inoculating biofilms (Fig. 2A; Fig. S2C), with both the WT and the mutant mixed throughout the biofilm (Fig. 2B). The WT, on average, was responsible for 61.2% of the biofilm biomass and 47.5% of the average biomass thickness (Fig. 2A and B; Fig. S2C). The fluorescent-protein-expressing plasmids were then switched between the WT and EPS-deficient mutants to ensure that the initial results were not biased by differences in GFP and mCFP expression or fluorescent signal intensity. Similarly, the WT was responsible for 45.3% of the biomass in each biofilm or 52.5% of the average biomass thickness when WT strains expressing GFP were used (Fig. S2A and B). Although biofilms from S. Typhi ΔtviB had more biomass reported by mCFP (Fig. S2A), this was likely an artifact of the fluorescent reporter, given that the bacteria were evenly distributed in the biofilm (Fig. S2B), a pattern consistent with that of GFP-expressing mutants (Fig. 2A and B). Representative maximum-intensity projections from z-stack orthogonal views from each combination of mCFP- and GFP-expressing biofilms are provided in Fig. S2C; these show similar detection for WT and mutant strains.

FIG 2.

Proportions and distributions of WT and EPS-deficient mutant bacteria in cocultured biofilms on cholesterol-coated surfaces. Biofilms were cocultured using the WT strain expressing mCFP and EPS-deficient mutants expressing GFP. Unstained biofilms were fixed in 4% PFA, and CLSM images were then captured as z-stacks (n = 3) using a Zeiss LSM 800 confocal microscope at ×63 magnification. ImageJ software running Comstat2 was used to calculate the biomasses (A) or the average biomass thicknesses (B) of WT and mutant bacteria in each biofilm. Data are presented as relative abundances (determined by fluorescent-signal quantification), and error bars indicate standard errors of the means. Statistically significant differences in the proportions of WT versus mutant bacteria for each coculture were evaluated by a paired t test with the Benjamini, Krieger, and Yekutieli approach to determining the false discovery rate (43).

Quantification of EPSs in cocultured biofilms revealed biofilms containing WT-ΔcsgA, WT-ΔyihO, WT-ΔcsgA ΔyihO ΔwcaM ΔbcsE, and WT-ΔtviB cocultures closely resembled single-culture WT biofilms, with no significant differences in EPS biomass or total biofilm biomass (Fig. 3A; Fig. S3). While the level of detection of the O-antigen capsule was reduced in WT-ΔbcsE cocultures for unknown reasons, the production of cellulose provided by the WT was similar to that in WT single-culture biofilms. As observed in Fig. 1, addition of the Vi antigen to S. Typhimurium did not have obvious effects on other EPSs (Fig. 3A), and quantities of live cells in the biofilms were consistent with what was observed in single-culture biofilms containing either strain (Fig. 3B). Interestingly, coculture biofilms of WT and ΔwcaM bacteria did not restore the O-antigen capsule but did significantly increase total biomass by increasing the production of curli fimbriae and cellulose (Fig. 3A). Compared to that in single-strain biofilms, the spatial organization of EPSs in S. Typhimurium (but not S. Typhi) cocultured biofilms demonstrated several significant differences in average and maximum biomass thicknesses for certain EPSs (Fig. 3C to F). These results suggest that the WT strain is able to compensate for the lack of EPS production in cocultured mutants, but that the spatial distribution of the EPSs within the biofilms becomes dysregulated.

FIG 3.

Quantification of EPSs in cocultured Salmonella biofilms on cholesterol-coated surfaces. Biofilms were first cocultured in a chambered coverglass coated with cholesterol and then treated with antibodies or stains specific for the detection of each EPS and live bacteria. (An anti–Vi antigen antibody was applied only to cocultures containing S. Typhimurium expressing the Vi antigen, S. Typhi, or S. Typhi ΔtviB). CLSM images were captured as z-stacks (n = 3) by using a Zeiss LSM 800 confocal microscope at ×63 magnification. ImageJ software running Comstat2 was used to calculate the biomass of each EPS (A), the biomass of live cells in a biofilm (B), and the average (C and D) and maximum (E and F) biomass thicknesses of each EPS. (A and B) Bars for the WT (checkered) show reference values reported in Fig. 1. (A) Statistical analysis of biomass for each EPS was conducted by comparing the WT single culture with each WT-mutant coculture. Comparisons were made by two-way ANOVA with Dunnett’s correction for multiple comparisons. (B) No significant differences in live cells between different cocultures or between cocultures and WT single cultures were identified by one-way ANOVA (with Tukey’s correction). (C to F) Differences in average or maximum biomass thickness between EPSs within each coculture were identified by two-way ANOVA with Tukey’s correction for multiple comparisons. Error bars indicate standard deviations; asterisks indicate significant differences (*, P < 0.05; **, P < 0.01; ***, P < 0.0005; ****, P < 0.0001).

Biofilm growth on cholesterol gallstones with ox or human bile.

Cholesterol gallstones from patient donors with bile in the growth medium can be used to mimic in vivo gallbladder conditions. Ox bile, which is commercially available as a crude extract, and human bile were used to evaluate EPSs in biofilms cultured on human gallstones. In order to conduct CLSM analysis of EPSs that developed on gallstones, samples had to be prepared histologically for sectioning and mounted on slides. This process required biofilms to be fixed in neutral buffered formalin (NBF), which prevented the ability to stain for live bacteria using Syto9. Instead, anti-lipopolysaccharide (anti-LPS) antibodies specific for S. Typhimurium or S. Typhi were applied alongside EPS-specific antibodies and stains in order to quantify the cellular presence in biofilms.

When cultured in ox bile, S. Typhimurium and S. Typhi biofilms had similar amounts of total biomass development in the areas surveyed (Table 1), and production of each EPS was detected. As with cholesterol-coated surfaces without bile added (Fig. 4A), cell staining (Syto9 or LPS) accounted for similar shares of the biofilm in each serovar with ox bile (Fig. 4B and C). In ox bile-gallstone biofilms, the relative abundances of S. Typhimurium EPSs mirrored those on cholesterol-coated surfaces without bile; cellulose was the most detected EPS (36.2% and 37.1% of the biomass, respectively) (Fig. 4A and B). Despite the additional presence of the Vi antigen, S. Typhi gallstone biofilms grown in ox bile were also predominantly made up of cellulose (42.6% of biomass), in contrast with what was observed on cholesterol-coated surfaces without bile (27.3% of biomass) (Fig. 4A and C). For the two serovars, the levels of detection of other EPSs (the O-antigen capsule and curli fimbriae) were similar, and biofilm composition was consistent between peripheral- and center-cut biofilm sections (Fig. 4B and C).

TABLE 1.

Total biomass detected in S. Typhimurium and S. Typhi biofilms cultured on human gallstones with ox bile or human bile or in vivo

Growth condition	Biomass (μm3/μm)a
	S. Typhimurium			S. Typhib
				O-antigen capsule			Vi antigen
	Min	Max	Avg	Min	Max	Avg	Min	Max	Avg
3% Ox bile	21.1	40.3	30.8	11.6	22.3	17.2	21.5	36.2	27.1
2% Human bile	21.0	35.9	29.0	11.9	24.2	16.8	8.5	36.7	25.6
Mice	9.89	23.8	17.3	ND	ND	ND	ND	ND	ND

^aComstat2 software was used to calculate the biomass of each EPS in biofilm z-stacks, and values were added to determine the total. Minimum, maximum, and average values are derived from all z-stacks collected from each slide for the stated growth conditions. ND, not determined.

^bS. Typhi biofilm slides were stained with the O-antigen capsule or the Vi antigen.

FIG 4.

Quantification of EPSs and imaging of Salmonella biofilms cultured in 3% ox bile on human cholesterol gallstones. Gallstones were cut into sections and mounted on glass slides for EPS-specific stain and antibody treatment. CLSM images were captured as z-stacks (n = 3) using a Zeiss LSM 800 confocal microscope at ×63 magnification. DIC images were taken at ×10 magnification. (A) Relative abundances (determined by fluorescent-signal quantification) of WT EPSs on cholesterol-coated surfaces (reference data from Fig. 1). (B and C) ImageJ software running Comstat2 was used to calculate the biomasses of S. Typhimurium EPSs in peripheral, middle, or center cuts from two different gallstones (B) and S. Typhi EPSs in paired peripheral, middle, or center cuts from a single gallstone stained for the O-antigen capsule or the Vi antigen and all other detectable EPSs (C). EPS data are presented as relative abundances (determined by fluorescent-signal quantification), and error bars indicate standard errors of the means. (D and E) Representative images from each z-stack showing associations in EPS organization within biofilms. Bars, 100 μm in the top and bottom DIC images; 20 μm in the middle DIC image; 10 μm in all other images. Boxed areas in DIC panels are enlarged in the other panels. OAC, O-antigen capsule; CF, curli fimbriae; ViAg, Vi antigen.

Images of biofilm sections in ox bile showed that EPS relationships differ between S. Typhimurium and S. Typhi. For S. Typhimurium, LPS and curli fimbriae were frequently found to be in association; LPS nearly always colocalized with curli fimbriae (Fig. 4D). Furthermore, the O-antigen capsule and cellulose had a similar pairing and tended to occupy areas in which LPS was not present, although instances where all EPSs coexisted were observed (Fig. 4D). In contrast, S. Typhi biofilms did not have as rigid relationships; each EPS existed throughout the biofilm area and in association with all other EPSs (Fig. 4E).

Examination of S. Typhimurium and S. Typhi biofilms cultured on gallstones with human bile verified that each EPS tested for was produced by both serovars. Both S. Typhimurium and S. Typhi produced similar amounts of total biomass with human bile and ox bile (Table 1), although the composition of both biofilms skewed toward a prevalence of cellulose, which increased to 43.6% of biomass in S. Typhimurium biofilms and accounted for a majority of S. Typhi biofilm biomass, at 50.7% (Fig. 5A and B). However, the composition of each biofilm remained consistent in different biofilm sections (Fig. 5A and B). Despite differences in EPS detection, relationships between EPSs similar to those observed in ox bile were observed in human bile: S. Typhimurium LPS and curli fimbriae occupied similar areas of biofilms, and the O-antigen capsule and cellulose were found in association (Fig. 5C), whereas S. Typhi biofilm EPSs were detected throughout all biofilm areas and were able to associate with every other EPS (Fig. 5D).

FIG 5.

Quantification of EPSs and imaging of Salmonella biofilms cultured in 2% human bile on human cholesterol gallstones. Gallstones were cut into sections and mounted on glass slides for EPS-specific stain and antibody treatment. CLSM images were captured as z-stacks using a Zeiss LSM 800 confocal microscope at ×63 magnification. DIC images were taken at ×10 magnification. (A and B) ImageJ software running Comstat2 was used to calculate the biomasses of S. Typhimurium (A) and S. Typhi (B) EPSs in peripheral, middle, or center cuts of the gallstone. EPS data are presented as relative abundances (determined by fluorescent-signal quantification), and error bars indicate standard errors of the means. (C and D) Representative images from each z-stack showing associations in EPS organization within biofilms. Bars, 100 μm in DIC images and 10 μm in all other images. Boxed areas in DIC panels are enlarged in the other panels. OAC, O-antigen capsule; CF, curli fimbriae; ViAg, Vi antigen.

Biofilm development during chronic infections of mice with gallstones.

We have developed a mouse model of chronic Salmonella gallbladder infection using 129 × 1/SvJ mice fed a lithogenic diet. These NRAMP1⁺ (SLC11A1) mice with gallstones can become chronically infected in a process mirroring the human condition (4). We used this mouse model to examine EPS development on gallstones during early stages of chronic carriage. Upon sacrifice 14 to 21 days postinfection, the liver and spleen were removed from each mouse to confirm infection (Fig. S4). Gallbladders that were visibly enlarged or contained visible gallstones were selected for histological preparation and CLSM analysis as in the human gallstone in vitro assays. Microscopy confirmed that each EPS examined was produced in vivo, although the total biomass of S. Typhimurium biofilms was reduced from those in ox and human bile gallstones by 43.8 and 40.3% (respectively) (Table 1). Despite this difference, the detection of each EPS was consistent across mice and in different locations on each gallstone (Fig. 6A). Furthermore, cellulose remained the most-observed EPS, accounting for 43.0% of biofilm biomass on average (Fig. 6A), indicating that in vivo biofilms closely resemble S. Typhi and S. Typhimurium biofilms cultured in vitro in the presence of human bile.

FIG 6.

Quantification of EPSs and imaging of S. Typhimurium biofilms during chronic infection. Gallbladders containing gallstones were harvested from five chronically infected mice at 21 days postinfection and were prepared by histology for sectioning and mounting on glass slides. An EPS-specific stain and antibodies were applied to each section, and CLSM images were captured as z-stacks using a Zeiss LSM 800 confocal microscope at ×63 magnification. DIC images were taken at ×10 magnification. (A) ImageJ software running Comstat2 was used to calculate the biomass of S. Typhimurium EPSs in peripheral, middle, or center cuts from five gallstones of infected mice. EPS data are presented as relative abundances (determined by fluorescent-signal quantification), and error bars indicate standard errors of the means. (B to D) Representative images from z-stacks collected from separate mice showing associations in EPS organization within biofilms. Bars, 100 μm in DIC images and 10 μm in all other images. Boxed areas in DIC panels are enlarged in the other panels. OAC, O-antigen capsule; CF, curli fimbriae.

Image analysis of S. Typhimurium biofilms in mice yielded unique results, contrasting with those from previous S. Typhimurium biofilms cultured in bile (ox or human) (Fig. 4D and Fig. 5C, respectively). First, the colocalization of LPS with curli fimbriae and of the O-antigen capsule with cellulose was reduced (Fig. 6B to D). Furthermore, what appeared to be sections of single gallstones at low magnification using differential interference contrast (DIC) microscopy (Fig. 6C and D) were found at high magnification to actually be dense clusters of numerous, much smaller gallstones (Fig. 6C and D). Such gallstones have been observed previously in mice during dissection and have a fine, sandy characteristic (often called sludge) (46). Importantly, biofilms that developed on these sandy gallstones were frequently observed to be composed of each EPS, without any unique pairings observed between different EPSs. Although these results were unexpected based on observations made in human and ox bile, they demonstrate that the EPS relationships of S. Typhimurium biofilms that develop in vivo are similar to S. Typhi biofilms in vitro regardless of the bile source used.

DISCUSSION

Despite numerous differences between S. Typhi and S. Typhimurium (such as host range and specificity, primary virulence strategies, and clinical manifestations of disease), a striking number of similarities, at least at the genetic level, exist for their EPSs. Both serovars have the necessary genes for the O-antigen capsule, curli fimbriae, and cellulose (examined here), and both incorporate lipids and extracellular DNA in the assembly of biofilms (22–26). Two notable exclusive EPSs are the Vi antigen (which is specific to typhoidal serovars) and colanic acid (specific to nontyphoidal serovars). While the genetic capabilities of the two serovars are known, the expression and spatial distribution of the various EPSs, especially in vivo or under in vivo-like conditions, were not known. Because biofilm formation on gallstone surfaces is key to the development and maintenance of chronic gallbladder carriage, and because EPSs may be a therapeutic target, it is important to understand their architecture in biofilms.

Biofilms were first grown on cholesterol-coated glass surfaces in order to examine the EPS network under gallstone-mimicking conditions. Unfortunately, we were not able to measure colanic acid, because detection reagents are not available. This EPS would have been interesting to examine because typhoidal serovars do not produce colanic acid due to multiple conserved mutations in the wca operon (35, 36). However, using CLSM to detect the O-antigen capsule, curli fimbriae, cellulose, the Vi antigen, LPS, and live bacteria, we demonstrated that growth on cholesterol-coated surfaces supports equal biofilm development for the two serovars: the S. Typhimurium and S. Typhi WT biofilms produced similar amounts of the EPSs they have in common (Fig. 1). This approach also demonstrated that expression of the O-antigen capsule was altered by two other mutations (ΔcsgA, ΔwcaM). The elimination of curli fimbriae by mutation increased the average and maximum biofilm thicknesses of the O-antigen capsule in the biofilm, an outcome that is consistent with the finding by Tursi et al. that biofilms targeted with anti-amyloid antibodies or anti-CsgA serum have loose, noncompacted topography associated with increased distance between bacteria and growth surfaces (37). In that study, decreased density was also associated with barrier permeability and antibiotic sensitivity, suggesting that curli fimbriae are critical regulators of EPS architecture, as seen here. Although the O-antigen capsule was also altered in biofilms produced by ΔwcaM mutants, the presence of curli fimbriae kept biofilm density similar to that of WT biofilms. Other mutants (ΔyihO, ΔbcsE, ΔtviB) exhibited only EPS-specific reductions that were not associated with topographical changes. Overall, these findings validate the applicability of studies that screen for EPS-active compounds in vitro using S. Typhimurium or S. Typhi biofilms.

Prior analysis of biofilms produced by cocultures of EPS-deficient mutants and the WT demonstrated that tolerance to H₂O₂ is a complex process in which certain WT-produced EPSs protect the corresponding cocultured mutant strain, while others do not (34). Therefore, it was important to use CLSM to track biofilm EPS dynamics in cocultured adherent communities (Fig. 3). The single-strain biofilms discussed above are consistent with H₂O₂ tolerance phenotypes previously described for EPS-deficient mutants, in which biofilms produced by S. Typhimurium ΔyihO and S. Typhimurium ΔwcaM (which have reduced O-antigen capsule levels) are sensitive to H₂O₂, and biofilms produced by S. Typhimurium ΔcsgA (which show an increase in the O-antigen capsule) are not (34). Quantification of total EPSs in cocultured biofilms demonstrated that WT bacteria present in the biofilm are able to adjust for EPS deficiencies in curli fimbriae, the O-antigen capsule, cellulose, and the Vi antigen and to increase production to compensate for the specific mutation (Fig. 3A). In further support of coordination between WT and mutant bacteria is the fact that the O-antigen capsule and cellulose, which were stimulated in single-culture biofilms produced by the ΔcsgA mutant, remained present at WT levels in coculture biofilms containing WT and ΔcsgA bacteria, indicating that the ability to coordinate EPS production is bidirectional. However, consistently low O-antigen capsule levels in coculture biofilms containing WT and ΔwcaM bacteria demonstrate that off-target EPS changes are not necessarily corrected by the WT. These findings highlight the importance of using CLSM or other techniques to verify the actual EPS composition of biofilms whenever one is using EPS-deficient mutants to screen for potentially exploitable phenotypes.

After identifying how EPSs develop on cholesterol-coated surfaces with respect to WT and EPS-deficient mutant bacteria, we shifted our approach to document the changes that occur in WT Salmonella EPSs due to growth on human gallstones with the addition of ox or human bile so as to mimic the in vivo gallbladder environment in vitro as accurately as possible. This work demonstrated that both serovars increase cellulose production in human bile over that in ox bile without altering the total amounts of LPS, the O-antigen capsule, or curli fimbriae (Fig. 4 and 5). This trend remained true for S. Typhimurium during chronic infection of mice, since cellulose was detected at a level similar to that of biofilms cultured in human bile (Fig. 6). Along with the consistencies observed in the colocalization relationships of S. Typhi EPSs in vitro and S. Typhimurium EPSs in vivo, these results show why the chronic S. Typhimurium infection model works well as a representation of chronic S. Typhi infections, which naturally occur in the presence of human bile. Importantly, enhancement of cellulose in vivo did not prevent other EPSs from being produced, making comparisons between experiments conducted in vitro and in vivo more reliable. These two assays (S. Typhi in human bile and S. Typhimurium in vivo) are therefore the two experimental approaches that are expected to yield the most-reliable results for developing EPS-targeting agents. Best practice, of course, would be to demonstrate success in both models.

When analyzing mouse gallstone biofilms, we noted a propensity for biofilm development on the small, sandy-type gallstones. In the majority of gallbladders examined, the gallstones were not singular but a sludge that contained these numerous small cholesterol gallstone aggregates. Macroscopic aggregates of sandy gallstones would maximize the available surface area of cholesterol to which Salmonella can bind. It is unclear if these small aggregates coalesce into a solid stone, entrapping the Salmonella, or if the sludge aggregates represent the maximum extent of gallstone development in that mouse. This organization would contribute to the protected niche of Salmonella biofilms and to the extreme recalcitrance exhibited by Salmonella in vivo despite the strong antimicrobial activity of bile and the immune response.

Before this analysis, it was not known how production of the Vi antigen in S. Typhimurium might affect the production of other EPSs. However, CLSM analysis demonstrated that the S. Typhimurium strain expressing the Vi antigen mirrors both its parental WT strain in terms of O-antigen capsule, curli fimbriae, and cellulose produced and S. Typhi in terms of Vi antigen produced within the biofilm matrix (Fig. 1). As mentioned above, we could not examine colanic acid production in the matrix. It is intriguing to postulate that colanic acid and the Vi antigen may possess some similar properties/phenotypes, since each is serovar specific, so future efforts should develop methods to determine if the abnormal expression of colanic acid in S. Typhi affects the Vi antigen or if expression of the Vi antigen in S. Typhimurium affects colanic acid. However, because of the similarity between S. Typhimurium biofilms in vivo and S. Typhi biofilms in vitro in human bile (Fig. 5 and 6) and the lack of an effect on other EPSs of S. Typhimurium ectopically expressing the Vi antigen, the use of S. Typhimurium Vi-expressing strains in vivo is predicted to provide a powerful advantage in further studies developing anti-EPS compounds.

The number of EPSs that Salmonella produces poses challenges for further studies. While the use of cholesterol-coated surfaces to screen for phenotypes remains an efficient tool for studying Salmonella, this approach has its limitations, and users must take steps to validate the EPSs produced. Additional EPSs, such as extracellular DNA released upon cell death and lipids, were not included, and some, such as extracellular DNA, are unable to be investigated by mutation. The apparent ability, demonstrated here, of both WT and EPS-deficient mutant strains to adjust EPS expression in response to loss of the ability to produce one or more EPSs cautions against any assumption that therapeutic removal of a major EPS from a biofilm will not be circumvented by the upregulation of other EPSs that maintain biofilm recalcitrance. To progress in development, promising EPS-targeting compounds will need to be validated in multiple biofilm models, with care taken to document the effect, if any, on nontarget EPSs and its implications for potential efficacy; combination therapies may still be necessary. Determining the relative abundances of EPSs in donated patient samples will ultimately be a necessary step toward a complete understanding of chronic Salmonella gallstone biofilms. Through systematic evaluation of the ways in which EPSs develop in various infection models, this work takes an important step toward that goal by providing appropriate methods and context for comparison by donor studies.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

This study was conducted with the WT strains S. Typhimurium ATCC 14028 (JSG210) and S. Typhi rpoS⁺ Ty2 (JSG4383) as well as derivative strains with mutations to eliminate the production of one or more EPSs constructed in one of the WT backgrounds (Table 2). All planktonic cultures were grown for 16 h in tryptic soy broth (TSB). When appropriate, ampicillin (Amp) was added to the growth medium at 100 μg/ml.

TABLE 2.

WT and EPS mutant strains used for this study

Strain	Genotype	EPS deficiency	Reference or source
JSG210	WT S. Typhimurium	None	ATCC 14028
JSG3736	S. Typhimurium ΔcsgA	Curli fimbriae	26
JSG3742	S. Typhimurium ΔwcaM	Colanic acid	26
JSG3672	S. Typhimurium ΔyihO	O-antigen capsule	26
JSG3838	S. Typhimurium ΔbcsE	Cellulose	26
JSG3841	S. Typhimurium ΔcsgA ΔwcaM ΔyihO ΔbcsE	Curli fimbriae, colanic acid, O-antigen capsule, cellulose	26
JSG3738	S. Typhimurium/pTH170 (viaB)	Vi antigen positive	44
JSG4383	WT S. Typhi rpoS+	None	45
JSG4695	S. Typhi ΔtviB rpoS+	Vi antigen	34

Fluorescent markers to test mixed-community biofilms.

WT and EPS-deficient mutant strains were differentially marked with fluorescent proteins for use in coculture biofilms. The pFPV-25.1 (38) and pFPV-mCherry (39) plasmids (carrying genes for GFP or mCFP, respectively, constitutively expressed from the rpsM promoter) were isolated from overnight broth cultures of S. Typhimurium SL1344 or Escherichia coli DH10B (JSG1093 and JSG4704, respectively) using the QIAprep Spin Miniprep kit (Qiagen, Germantown, MD). Each strain (Table 2) was transformed with pFPV-25.1 or pFPV-mCherry, selected on Luria-Bertani (LB) medium supplemented with Amp, and incubated at 37°C. One resistant colony from each transformation was selected for further use in coculture experiments (see Table S1 in the supplemental material). All coculture experiments were conducted with an mCFP-expressing WT strain paired with a GFP-expressing mutant; then the pairings were reversed by testing GFP-expressing WT strains paired with mCFP-expressing mutants.

Biofilm growth and staining on a cholesterol-coated coverglass.

Planktonic bacteria from overnight broth cultures were used to initiate biofilm growth on an 8-well Nunc Lab-Tek chambered coverglass (Thermo Fisher Scientific, Waltham, MA) that had been coated with cholesterol (500 μg/well) as described previously (29). Planktonic bacteria were normalized to an optical density at 490 nm (OD₄₉₀) of 0.65 in TSB, diluted 1:6 in TSB, and incubated at 37°C for 3 h in a static 12-well polypropylene plate. When biofilms containing WT and mutant strains mixed together were started, normalized bacteria were diluted 1:12 into mixed culture so that the total bacteria were equivalent to those under single-strain conditions (diluted 1:6). After 3 h, static cultures were diluted 1:2,500 in TSB and transferred to a cholesterol-coated coverglass (200 μl/well). Biofilms were incubated at 30°C on a nutator for 96 h, and the supernatant was replaced with fresh medium (TSB) once every 24 h.

At the end of the 96-h growth period, biofilms were washed twice with a solution of Tris-buffered saline plus Tween 20 (TBST) to remove unattached and planktonic bacteria. All subsequent steps occurred at room temperature, and samples were shielded from light to prevent photobleaching. Coculture biofilms (containing bacteria expressing GFP and CFP) were fixed with a 4% (vol/vol) paraformaldehyde (PFA) solution in phosphate-buffered saline (PBS) (Affymetrix, Cleveland, OH) for 30 min, washed three times with TBST, and kept hydrated in TBST for imaging (described below). Single-strain biofilms (nonfluorescent bacteria) were used to label individual EPSs within the biofilm. Each stain/antibody was titrated to determine the optimal concentration/dilution for detecting each target accurately and specifically. Viable bacteria and cellulose were labeled with Syto9 (Molecular Probes, Eugene, OR) or Fluorescent Brightener 28 (Sigma-Aldrich, St. Louis, MO), respectively (Table S2), prior to fixation (4% PFA for 30 min). Additional EPSs (curli fimbriae, O-antigen capsule, or Vi antigen) were labeled by immunodetection (Table S2) after biofilms were fixed, washed three times with TBST, and blocked with 5% (wt/vol) bovine serum albumin (BSA) in TBST for 1 h. Each biofilm was washed three times with TBST after primary and secondary antibody incubation, and final samples were kept hydrated in TBST for imaging. All secondary antibodies targeted IgG (heavy plus light chain) for a particular primary-antibody species (Table S2) and were cross-adsorbed. Alexa Fluor antibodies were acquired from Invitrogen (Carlsbad, CA).

O-antigen capsule antiserum absorption.

An antiserum to S. Typhimurium raised in rabbit was provided by Aaron White (University of Saskatchewan). To generate an O-antigen capsule polyclonal antiserum, samples were adsorbed twice against an S. Typhimurium O-antigen capsule whole-operon deletion strain (JSG3675). For each round of adsorption, overnight cultures of JSG3675 were pelleted and fixed for 15 min on ice by resuspension in 1,000 μl 2% PFA prepared in PBS plus Ca/Mg. PFA was removed by washing fixed cells three times; then the sample was normalized to an OD₆₀₀ of 0.8, pelleted, and resuspended in 200 μl S. Typhimurium antiserum. Antiserum samples were incubated on an orbital shaker (200 rounds per min) for 1 h at room temperature before the removal of bacteria by centrifugation at 16,000 × g for 5 min.

Biofilm growth on human gallstones.

Human cholesterol gallstones were acquired from consenting patient donors undergoing cholecystectomy (IRB 2018H0104). Donated gallstones were cleaned and dried for storage. Prior to use, each gallstone was irradiated for 30 min by UV light (2.0 × 10⁴ μJ/cm²) in a Hoefer UVC 500 UV Crosslinker and was then aseptically transferred to TSB with bile and a normalized starting inoculum. All biofilms on human cholesterol gallstones were cultured in the presence of 3% (wt/vol) ox bile (MP Biomedicals, Solon, OH) or 2% (vol/vol) human bile (pooled from 5 human donors; IRB 2018H0104). When growth was initiated in the presence of ox bile, planktonic bacteria were first normalized to an OD₆₀₀ of 0.8 and then serially diluted 1:100 (approximately 2.0 × 10⁶ CFU/ml) in TSB–3% ox bile. For growth in the presence of human bile, bacteria were normalized to an OD₆₀₀ of 0.8 in TSB–2% human bile for a starting inoculum of approximately 2.0 × 10⁸ CFU/ml (the bile concentration and starting inoculum were amended to support growth, since human bile had more antibacterial activity against planktonic bacteria). Gallstone cultures were incubated on a rolling drum at 37°C for 6 days. The ox bile medium was replaced once every 24 h with fresh TSB–3% ox bile. The human bile medium was also replaced every 24 h with fresh TSB–2% human bile. After 6 days of growth, gallstone biofilms were washed twice with PBS and then immersed in 10% (vol/vol) neutral buffered formalin (NBF) for 24 h. Samples were then transferred to a 1:1 decalcification solution of 50% (vol/vol) aqueous formic acid and 20% (wt/vol) aqueous sodium citrate for an additional 16 h before transfer back to 10% NBF for histological preparation (described below).

Biofilm growth in a murine model of typhoid carriage.

Mice were housed and cared for in compliance with all rules set by the Abigail Wexner Research Institute (AWRI) Institutional Animal Care and Use Committee (IACUC). All experiments involving animals were authorized by AWRI IACUC protocol AR18-00080 and adhered to the statutes of the Animal Welfare Act and the guidelines of the Public Health Service as required in the Guide for the Care and Use of Laboratory Animals (40). The AWRI IACUC is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International.

Female 129 × 1/SvJ mice (The Jackson Laboratory, Bar Harbor, ME) were fed a lithogenic diet (mouse chow supplemented with 1% cholesterol and 0.5% cholic acid [Envigo, Indianapolis, IN]) for 8 weeks prior to infection. After the diet period, mice were infected with S. Typhimurium by intraperitoneal (i.p.) injection of 100 μl PBS containing 2.0 × 10⁴ CFU. Mice were sacrificed 21 days postinfection, and gallbladders, livers, and spleens were removed. Gallbladders were fixed in 10% NBF for 24 h before histological preparation (described below). Liver and spleen samples were homogenized in sterile PBS using a TissueLyser LT and sterile 5-mm stainless-steel beads (Qiagen, Germantown, MD) to verify infection and to determine the bacterial load of each organ by serial dilution on LB broth agar.

Histological preparation and staining of gallstone sections.

Human gallstones and mouse gallbladders (containing gallstones) fixed in 10% NBF were prepared for thin-section imaging by the AWRI Morphology Core. All solutions were supplemented with HistoGel (Thermo Fisher Scientific, Waltham, MA) due to gallstone friability. Samples were dehydrated at room temperature in an ethanol series (70% [overnight], 95% twice [20 min each time], 100% three times [30 min each time]) and were then transferred to Slide Brite (once for 30 min, once for 60 min). Samples were then infiltrated with molten paraffin wax ethanol (60°C, once for 30 min, once for 40 min, once for 60 min) before being embedded in molten paraffin wax and allowed to cool. Twenty-five sections were cut from each specimen using a Leica RM2255 microtome (Leica Biosystems, Buffalo Grove, IL) and were mounted on glass slides. The slides were air dried and then transferred to a 60°C oven for 30 min.

The 25 sections were cut beginning at the gallstone or gallbladder periphery (e.g., slide 1) and extending to the sample center (e.g., slide 25). From the available samples, three evenly spaced sections (e.g., slides 8, 16, and 24 representing peripheral, middle, or center cuts) were selected for EPS staining and immunodetection. When the O-antigen capsule and Vi antigen needed to be labeled in the same specimen, adjacent slides were used for alternate labeling of these two targets (e.g., slides 8, 16, and 24 for the O-antigen capsule and slides 9, 17, and 25 for the Vi antigen), because the primary antibodies for these two targets were raised in rabbit and thus were incompatible for secondary detection (all slides were always labeled for the other three compatible EPS targets). All slides were blocked for 60 min in 5% BSA-TBST and were then probed with a 5% BSA-TBST cocktail containing Fluorescent Brightener 28 and primary antibodies for the appropriate LPS, curli fimbriae, and either the O-antigen capsule or the Vi antigen (Table S2) for 2 h at room temperature, shielded from light. Slides were washed three times in TBST and were then incubated with a 5% BSA-TBST secondary-antibody cocktail for 1 h at room temperature. After a final wash (three times in TBST), a no. 1.5 coverslip was mounted with CoverGrip Coverslip Sealant (Biotium; Fremont, CA) and allowed to set overnight.

CLSM.

All imaging was conducted using an inverted Zeiss LSM 800 confocal laser scanning microscope (CLSM) controlled by ZEN 2.6 blue edition software and equipped with 10× (dry) and 63× (oil immersion; free working distance [WD], 1.4 [190 μm]) lenses. Biofilm and EPS quantification was conducted by z-stack imaging to generate 3-dimensional biofilm models. Fluorophore signals associated with each EPS were recorded separately for every slice within a z-stack according to the excitation and emission filter sets recommended by the manufacturer (Table S2). Three random z-stack images per sample were collected and analyzed by ImageJ software running Comstat2 to calculate values of biomass, average thickness, and maximum thickness for each EPS signal (41, 42).