A glycine betaine importer limits Salmonella stress resistance and tissue colonization by reducing trehalose production

Authors


E-mail piloniet@colorado.edu; Tel. (+1) 303 492 8555; Fax (+1) 303 492 7744.

Summary

Mechanisms by which Salmonella establish chronic infections are not well understood. Microbes respond to stress by importing or producing compatible solutes, small molecules that stabilize proteins and lipids. The Salmonella locus opuABCD (also called OpuC) encodes a predicted importer of the compatible solute glycine betaine. Under stress conditions, if glycine betaine cannot be imported, Salmonella enterica produce the disaccharide trehalose, a highly effective compatible solute. We demonstrate that strains lacking opuABCD accumulate more trehalose under stress conditions than wild-type strains. ΔopuABCD mutant strains are more resistant to high-salt, low-pH and -hydrogen peroxide, conditions that mimic aspects of innate immunity, in a trehalose-dependent manner. In addition, ΔopuABCD mutant strains require the trehalose production genes to out-compete wild-type strains in mice and macrophages. These data suggest that in the absence of opuABCD, trehalose accumulation increases bacterial resistance to stress in broth and mice. Thus, opuABCD reduces bacterial colonization via a mechanism that limits trehalose production. Mechanisms by which microbes limit disease may reveal novel pathways as therapeutic targets.

Introduction

The Gram-negative bacterium Salmonella enterica can be transmitted to naive hosts in contaminated water or food. To cause systemic infection, the bacteria cross the intestinal epithelium and reside in the liver, spleen and lymph nodes inside of professional phagocytes, particularly macrophages. S. enterica serovars Typhi and Paratyphi A–C cause human typhoid fever, a systemic acute infection that resolves into subacute chronic infection in approximately 4% of patients (Parry et al., 2002). Prior to the use of antibiotics, approximately 5–20% of typhoid patients died within 2–3 weeks of infection, but most people began to convalesce within 4 weeks (Osler, 1892). Today, the increasing prevalence of multi-drug-resistant Salmonellae is significant cause for concern (Cooke and Wain, 2004).

In the wild and in the laboratory, mice infected with S. enterica serovar Typhimurium develop a natural, systemic typhoid-like disease, in which acute infections become persistent and generally asymptomatic (Tsolis et al., 1999). S. Typhimurium reside within macrophages in mice, which contain the bacteria in vesicles and restrict S. Typhimurium replication (Fortier et al., 2005). Macrophage vesicles are a challenging microenvironment for S. Typhimurium because they are at low pH and contain oxidative radicals, antimicrobial peptides and limiting levels of nutrients (Knodler and Steele-Mortimer, 2003). For example, S. Typhimurium access to iron is restricted by Nramp1 (Slc11a1), which transports iron from pathogen-containing vesicles to the cytosol (Blackwell et al., 2003; Fritsche et al., 2008; Nairz et al., 2009).

Salmonella Typhimurium adapts to stresses encountered within macrophage vesicles using a variety of strategies. Resistance to antimicrobial peptides is mediated, for instance, by chemical modifications to lipopolysaccharide (Gunn, 2008). Saccharide deprivation stimulates utilization of host fatty acids as a carbon source, as inferred from the requirement for the glyoxylate bypass pathway during persistent infection (Fang et al., 2005). More general responses to stress include the import or production of compatible solutes, low-molecular-weight, highly soluble organic molecules that maintain fluid balance in the cell and protect proteins and lipids form denaturation (Luzardo et al., 2000; Hincha and Hagemann, 2004). Compatible solutes help cells acclimate to changes in osmolarity, pH and temperature, stressors encountered by S. Typhimurium during infection. Two major compatible solutes imported by S. Typhimurium are glycine betaine and proline, but only glycine betaine, an N-trimethylated amino acid, is present at physiological concentrations in host tissues (Lever et al., 1994). A third major compatible solute is trehalose, a disaccharide consisting of two glucose monomers produced by microbes, plants and invertebrates, but not mammals (Richards et al., 2002). Trehalose is a particularly effective stress protectant but is not needed for acute S. Typhimurium infection in mice lacking the Nramp1 cation transporter (Fang et al., 1996; Howells et al., 2002).

Systemic S. enterica has been described as a ‘stealth pathogen’, which remains under the radar of the vertebrate immune system (Tsolis et al., 2008). Only four to five bacterial rods are typically found within each macrophage in mouse tissues, suggesting the bacteria replicate only two to three times within a given macrophage (Sheppard et al., 2003; Nix et al., 2007). Thus, S. enterica may have evolved to limit replication within tissues to minimize host damage and/or the intensity of the inflammatory response. Consistent with this hypothesis, S. Typhimurium genes that limit bacterial pathogenicity have been discovered based on hypervirulence phenotypes of strains with loss-of-function mutations (Ho and Slauch, 2001). Phenotypes of hypervirulence include reduced lethal or infectious dose, colonization advantages and decreased survival time of the host (Foreman-Wykert and Miller, 2003; Gal-Mor et al., 2008; Baek et al., 2009). Multiple examples of microbial loci that limit virulence, also called ‘antivirulence’ loci, are known in S. Typhimurium and other persistent pathogens. However, direct mechanisms by which they curtail infection have not yet been described. How microbial genes limit disease is important to establish because they have the potential to suggest novel avenues for treatment. In this study, we demonstrate that the S. Typhimurium OpuABCD glycine betaine transport system limits virulence via a mechanism that reduces trehalose production.

Results

opuABCD is a putative S. enterica glycine-betaine importer

The RcsBCD (Regulation of capsule synthesis) sensor-kinase signalling system is important for virulence in murine models of acute and chronic infection (Detweiler et al., 2003; Domínguez-Bernal et al., 2004). A previous study used DNA microarrays to identify genes potentially regulated by RcsBCD, including STM1491 and STM1492 (Erickson and Detweiler, 2006). STM1491 and STM1492 are part of a four-gene putative operon that has significant homology to Osmoprotectant uptake locus C (OpuC), within Gram-negative and -positive bacteria. The four genes, STM1491–4, encode a putative ATPase, two permeases, and a periplasmic binding protein, each with 42–63% amino acid identity over the entire length of the corresponding protein to Pseudomonas syringae OpuABCD (Fig. 1A). The P. syringae and Bacillus subtilis OpuABCD loci are demonstrated high-affinity glycine betaine importers (Kappes et al., 1996; 1999; Chen and Beattie, 2007). In P. syringae, OpuABCD is the primary transporter for glycine betaine under high osmolarity (Chen and Beattie, 2007). In addition, the OpuA ATPases, including STM1491, are characterized by tandem cystathionine-β-synthase (CBS) domains, which are required for and predictive of functional osmoregulatory transport (Fig. 1B) (Biemans-Oldehinkel et al., 2006; Chen and Beattie, 2007). These observations support designation of STM1491–4 as opuA–D.

Figure 1.

STM1491–4 is a putative glycine betaine importer. A. The opu loci from Pseudomonas syringae (P.s. pathovar tomato strain DC3000), S. enterica STM1491–4 locus (S.e. serotype Typhimurium) and Bacillus subtilis (B.s. subspecies subtilis strain 168). Per cent values between arrows indicates shared amino acid identity. Predicted encoded proteins and COG families are indicated at the bottom. B. Alignment of OpuA predicted ATPases from S. Typhimurium (SeT; STM1491, NP_460451.1), P. syringae (Ps; PSTPO_4575, AAO58021.1) and B. subtilis (Bs; NP_391263.1). Conserved nucleotide-binding protein motifs are labelled, including the Walker A (P-loop), Q-loop, LSGGQ motif, Walker B, D-loop and H-loop. clustal w (1.81) multiple sequence alignment and consensus key: single, fully conserved residue (*); conservation of strong groups (:); conservation of weak groups (.).

ΔopuA–D mutants are resistant to salt stress

In E. coli and S. enterica, the compatible solute trehalose is produced under conditions of stress if glycine betaine is unavailable or cannot be imported (Larsen et al., 1987; Cayley et al., 1992). Since trehalose is a more effective compatible solute than glycine betaine (Hincha and Hagemann, 2004), strains lacking the opuA–D locus may have increased resistance to high salt. In M9 with 0.4 M sodium chloride, the ΔopuA–D mutant strain grew more rapidly than wild type and by 17 h, had attained an OD600 of 1.8 compared with only 1.5 (P < 0.01) respectively (Fig. 2A). Plating for colony-forming units (cfu) confirmed the observed differences in optical density, and examination of cell shape and size by microscopy revealed no distinctions (data not shown). In control experiments, there were no apparent differences in growth between wild type and ΔopuA–D mutants in nutrient rich (LB) (data not shown) and poor (M9 minimal) media at 37°C (Fig. 2B). To establish whether ΔopuA–D mutants are resistant to sucrose, another distinct osmotic stress, bacteria were grown in nutrient poor media with 0.5 M sucrose, an osmolarity equivalent to that of 0.4 M sodium chloride. No differences in growth between mutant and wild-type strains were observed (Fig. 2C). This observation indicates that salt and sucrose elicit distinct responses in not only wild type (Botsford et al., 1994) but also in ΔopuA–D mutant strains. S. Typhimurium responds differently to osmotic stress caused by sucrose as compared with sodium chloride in that sucrose exposure elicits glutamate accumulation at a lower osmolarity than salt (Botsford et al., 1994). Collectively the data suggest that S. enterica strains lacking opuA–D respond to salt stress in a specific manner that potentially reflects differential gene regulation.

Figure 2.

Growth advantage of ΔopuA–D strains in high salt requires otsBA. Strains (as noted) were grown in M9 with 0.4 M NaCl (A, D), 0.01 M NaCl (B) or 0.5 M sucrose (C). Optical density at 600 nm (OD600) was monitored for 17 h. Error bars are SD, n > 5 experiments. Asterisks indicate P < 0.01 (*) and 0.001 (**) compared with WT, as determined using an one-way anova with a Tukey post-test on the time required to grow from an OD600 of 0.1 to the OD600 reached by the strain with the lowest saturation level in the experiment (Smith and Bidochka, 1998).

Resistance of ΔopuA–D mutants to salt requires trehalose-production genes

Trehalose generation in S. Typhimurium and E. coli requires the osmoregulated trehalose synthesis (ots) operon encoding otsB and otsA, and involves the transfer of glucose from UDP-glucose to glucose-6-phosphate to yield trehalose-6-phosphate (Giaever et al., 1988; Cánovas et al., 2001). To establish whether the salt-resistance phenotype of ΔopuA–D mutants requires the otsBA trehalose production operon,ΔotsBA mutant strains were constructed. As expected, the trehalose production operon is required for normal growth in 0.4 M NaCl (Fig. 2A). In addition, deletion of otsBA eliminated the growth advantage of ΔopuA–D mutant strains and reduced growth rates to below that of wild type. Complementation of the ΔopuA–D, ΔotsBA double mutant with otsBA under the endogenous promoter on a mid-copy plasmid (potsBA) restored salt-resistance (Fig. 2D). In addition, the plasmid containing otsBA increased the growth rates of all strains, including wild type. These observations demonstrate that otsBA is necessary and sufficient for the phenotype of increased salt-resistance in ΔopuA–D mutant strains.

Under stress, ΔopuA–D mutants accumulate more trehalose than wild-type strains

Under stress, improved E. coli growth results from small increases, as little as 3%, in trehalose accumulation (McDougall et al., 1993; Mahmud et al., 2009). To establish whether trehalose accumulates in strains lacking the opu locus, bacteria were grown with 0.01 M or 0.4 M sodium chloride without glycine betaine. Standard methods for measuring trehalose accumulation involve quantifying glucose as a proxy for trehalose (Cánovas et al., 2001; Nigam, 2007). Saccharides were extracted under conditions in which trehalose is hydrolysed to glucose (Pazur and Forsberg, 1980), resolved by thin-layer chromatography (TLC) and stained with potassium permanganate (Figs 3A and 4A). Glucose levels were also analysed by densitometry (Figs 3B and 4B). Background levels of glucose were observed in all strains grown in 0.01 M sodium chloride (Fig. 3A and B, lanes 12–15) (Record et al., 1998). Wild-type strains had significant levels of glucose, as expected (Figs 3A and 4A, lane 8), but the ΔopuA–D mutant had 12% more glucose than the wild-type strain (Fig. 3B). Glucose accumulation in all strain backgrounds required an intact otsBA locus. Thus, ΔopuA–D mutant strains exposed to salt stress produce more trehalose than their wild-type counterparts.

Figure 3.

ΔopuA–D mutant strain produces excess trehalose. Bacteria were grown in M9 with 0.01 M NaCl or 0.4 M NaCl. A. Sugars were extracted from equal numbers of bacteria and analysed by thin-layer chromatography (TLC). Sugar standards, lanes 1–7, arabinose (10 µg), galactose (10 µg), sucrose (10 µg), maltose (10 µg), glycerol (80 µg), trehalose (20 µg), glucose (10 µg) respectively. Lanes 8–15 as noted. B. Integrated densitometric analysis of four TLCs was performed using Image J software. Arbitrary values correspond to the product of the area and mean grey values, 0.4 M NaCl (black) or 0.01 M NaCl (white). Error bars are SEM. C. Total glucose accumulation (mg ml−1) quantified by the glucose-hexokinase conversion assay. Strains as noted were grown in M9 in 0.4 M NaCl (black) or 0.01 M NaCl (white). Error bars are SEM, n > 3 experiments. D. Trehalose accumulated (ng per 1 × 106 bacteria) quantified as described in Experimental procedure from bacteria grown in M9 with 0.4 M NaCl. Error bars are SEM, n > 3 experiments. ns, not significant. Asterisks indicate P < 0.05 (*) and 0.01 (**).

Figure 4.

Complementation with potsBA restores trehalose accumulation. Bacteria were grown in M9 with 0.4 M NaCl. A. Sugars were extracted from equal numbers of bacteria and analysed by thin-layer chromatography (TLC). Sugar standards, lanes 1–7, arabinose (10 µg), galactose (10 µg), sucrose (10 µg), maltose (10 µg), glycerol (80 µg), trehalose (20 µg), glucose (10 µg) respectively. Lanes 8–15 as noted. B. Integrated densitometric analysis of the TLC presented in (B) was performed using Image J software. Arbitrary values correspond to the product of the area and mean grey values. C. Total glucose accumulated (mg ml−1) was measured from bacteria grown in M9 with 0.4 M NaCl and complemented with potsBA. Error bars are SEM, n > 3 experiments. D. Trehalose accumulated in the presence of potsBA. Error bars are SEM, n > 3 experiments. v: vector strains have pRB3-273c; ns, not significant. Asterisks indicate P < 0.01 (*) and 0.001 (**).

Quantification of glucose using a different method yielded similar results (Figs 3C and 4C). Briefly, glucose was converted to 6-phosphogluconate, the by-product NADH was quantified spectrophotometrically, and trehalose levels were calculated after subtracting background levels of glucose observed in the ΔotsBA strain (Cánovas et al., 2001). There were 1.7 × 10−15 moles of trehalose per wild-type cell, consistent with levels observed previously in bacteria exposed to salt stress (Table 1). The ΔopuA–D mutant strain accumulated 15% more trehalose per cell than the wild-type strain (Fig. 3D). An intact otsBA locus was required for trehalose accumulation in ΔopuA–D mutants and in all strains examined in both low and high salt. Complementation of ΔotsBA with otsBA expressed from a plasmid restored trehalose production (Fig. 4D), although the presence of the plasmid pRB3-273c reduced glucose accumulation in all strains by approximately one-third (Fig. 4). In conclusion, two different methods demonstrate otsBA-dependent accumulation of similar levels of trehalose in S. Typhimurium ΔopuA–D mutant strains exposed to stress.

Table 1.  Accumulated trehalose in E. coli and Salmonella enterica.
BacteriaNaCl concentration (molar)Trehalose concentration (moles trehalose per cell)Reference
E. coli– WT0.1Not detectable Record et al. (1998)
E. coli– WT1.013 × 10−15 Record et al. (1998)
S. enterica– WT0.01Not detectableThis study
S. enterica–ΔopuA–D0.01Not detectableThis study
S. enterica– WT0.41.7 × 10−15This study
S. enterica–ΔopuA–D0.42.0 × 10−15This study

ΔopuA–D mutants are resistant to low pH in an otsBA-dependent manner

Trehalose confers microbial resistance to low physiological pH (Carvalho et al., 2011). Upon oral inoculation, Salmonellae encounter low pH in the stomach and again in macrophage vesicles after traversing the gut lumen (Knodler and Steele-Mortimer, 2003; McConnell et al., 2008). To establish whether deletion of opuA–D increases S. Typhimurium resistance to low pH, equivalent numbers of wild-type and mutant bacteria were inoculated separately into pH 5.0 M9 and grown overnight at 37°C. As anticipated, the bacteria became saturated for growth at a lower OD600 (< 0.4) than at neutral pH (Stokes and Bayne, 1957; Huhtanen, 1975; Russell and Dombrowski, 1980). Nevertheless, the ΔopuA–D mutant strain grew faster than the wild-type strain and the ΔopuA–D, ΔotsBA double mutant lagged behind the wild-type strain (Fig. 5A). Thus, the phenotype of enhanced resistance to low pH observed in ΔopuA–D mutants requires the otsBA locus. Complementation of the ΔopuA–D, ΔotsBA double mutant with otsBA on a plasmid restored growth at low pH (Fig. 5B). Indeed, all strains carrying the complementing plasmid grew better than corresponding strains caring only the vector, suggesting that increased trehalose is sufficient to confer resistance to low pH. Overall, the data show that ΔopuA–D mutants have increased resistance to low pH that is dependent upon otsBA.

Figure 5.

Growth advantage of ΔopuA–D strains at low pH requires otsBA. Strains (as noted) were grown in M9 pH 5 (A–B) and OD600 nm was monitored for 17 h. Error bars are SD, n > 5 experiments. P < 0.01 from 6 to 8 h, for all strains compared with wild type, as determined using an one-way anova with a Tukey post-test (Smith and Bidochka, 1998).

ΔopuA–D mutants are resistant to hydrogen peroxide in an otsBA-dependent manner

Trehalose production protects microbes from oxidative stress, including exposure to hydrogen peroxide, which forms in macrophage phagosomes upon S. Typhimurium infection (Alvarez-Peral et al., 2002). To determine whether ΔopuA–D strains have increased resistance to hydrogen peroxide in liquid culture, equivalent numbers of mutant and wild-type bacteria were exposed to 10 mM hydrogen peroxide, a lethal concentration, in M9 broth for 90 min. Bacteria were periodically removed and plated on LB agar to allow for enumeration of cfu (Fig. 6A). Although viability decreased for all strains, ΔopuA–D had prolonged survival compared with wild type (P < 0.001). Survival of ΔopuA–D mutants in hydrogen peroxide also required an intact otsBA locus indicating a role for trehalose production in resistance to oxidative stress.

Figure 6.

OpuABCD confers resistance to hydrogen peroxide in an otsBA-dependent manner. A. Strains were grown in M9 with 10 mM hydrogen peroxide. Error bars are SD, n > 3 experiments. Asterisks indicate P < 0.05 (*) and 0.001 (***) compared with WT. B. Filter Disc inhibition assay. Error bars are SEM, n = 3. Asterisks indicate P < 0.01 (**) compared with WT.

To establish whether ΔopuA–D mutants grown on agar are similarly resistant to hydrogen peroxide in an otsBA-dependent manner, filter disc assays were performed. A bacterial lawn was spread onto M9 agar and filter discs treated with 150 mM hydrogen peroxide were overlaid. The zone of growth inhibition around each disc was measured after 18 h (Fig. 6B). Consistently smaller zones of inhibition were observed for the ΔopuA–D strains than wild type (P < 0.001), confirming the mutant is resistant to hydrogen peroxide. Hydrogen peroxide resistance required otsBA and was restored in ΔopuA–D, otsBA double mutants upon complementation with otsBA. These observations indicate that in the absence of OpuA–D, otsBA is necessary and sufficient for S. Typhimurium increased resistance to hydrogen peroxide.

ΔopuA–D mutants resist killing by macrophages in an otsBA-dependent manner

Upon phagocytosis by macrophages, bacteria are exposed to oxidative species generated by the respiratory burst including hydrogen peroxide and superoxide. S. Typhimurium has mechanisms by which it resists oxygen radicals, but approximately 90% of the bacteria are nevertheless killed within 2 h of phagocytosis by primary mouse macrophages (Vazquez-Torres et al., 2000). Reactive oxygen species limit bacterial growth by damaging, for instance, enzymes within the Krebs cycle (Richardson et al., 2011). To establish whether ΔopuA–D mutants have improved survival in macrophages relative to wild type, bone marrow derived mouse macrophages were isolated and activated with INF-γ and LPS. Mixed infection experiments were performed because they are highly sensitive based on the presence of an internal wild-type control (Beuzón and Holden, 2001). After infection with equivalent numbers of ΔopuA–D and wild-type bacteria, macrophages were lysed and cfu enumerated. Within 2 h, ΔopuA–D mutants out-competed the wild-type strain, a phenotype dependent upon otsBA (Fig. 7A). Treatment of cells with MDL 72527, an inhibitor of spermine oxidase that blocks the production of hydrogen peroxide (Chaturvedi et al., 2004), abrogated the increased survival of ΔopuA–D mutants. The spermine oxidase inhibitor had no significant effect on macrophage viability under these conditions (data not shown). As a control, equivalent numbers of ΔopuA–D mutant and wild-type strains were co-inoculated into M9 or LB and monitored by plating for cfu over a 24 h period. Under these circumstances, ΔopuA–D mutants did not out-compete wild-type strains (Fig. 7B). These results, in combination with hydrogen peroxide resistance observations, suggest that in the absence of opuA–D, S. Typhimurium have increased resistance to macrophage reactive oxygen species and possibly reactive nitrogen species that is dependent upon the otsBA trehalose production operon.

Figure 7.

OpuABCD confers resistance to the macrophage respiratory burst in an otsBA-dependent manner. A. Macrophages treated with and without ROS inhibitor (ROSi), MDL-72527, were infected with equivalent numbers of WT and ΔopuA–D (black) or ΔopuA–D,ΔotsBA (white) strains. Macrophages were lysed and CI values calculated. Error bars are SEM, n = 3. Asterisks indicate P < 0.01 (**) and 0.001 (***). B. Co-culture of a 1:1 mixture of WT and ΔopuA–D in M9 (grey) and LB broth (white). Each bar represents the average log CI; error bars are SD, n > 3. Asterisks indicate P < 0.05 (*) and 0.001 (***).

ΔopuA–D mutants out-compete wild-type strains in mice, a phenotype that requires otsBA

Data suggest that S. Typhimurium strains lacking the opu locus are resistant to several stressors, including high concentrations of salt, low pH and oxidative compounds. Each of these stressors may be encountered by bacteria during systemic infection of mammals, suggesting that ΔopuA–D mutants may have increased virulence in mice. To test this notion, mice were intragastrically inoculated with wild-type or an ΔopuA–D strain. A lethal dose50 (LD50), as defined by the dose at which 50% of the population dies within 28 days, was established. Nramp1−/− (BALB/C) mice were chosen for these studies because they develop fatal S. Typhimurium infections upon intragastric inoculation (Tsolis et al., 2011). The ΔopuA–D strain had only a threefold lower LD50 than wild type, a statistically insignificant difference (4.2 × 106 versus 1.3 × 107, for ΔopuA–D and wild type respectively; 95% confidence intervals: 4.2 × 105 to 4.2 × 107 and 3.4 × 106 to 4.6 × 107). These data indicate that the absence of opuA–D has little to no effect on the oral LD50 of mice. However, genes that limit virulence may have little effect on animal survival and yet impact pathogenesis as determined by more sensitive methods, such as tissue colonization (Gal-Mor et al., 2008). To establish whether ΔopuA–D mutants out-compete wild-type strains in tissue, animal mixed-infection experiments were performed. The mouse strain used was Sv129S6 (Nramp1+/+), which tolerates greater than 1010 orogastric cfu of S. Typhimurium. Mice fasted for 9–12 h were inoculated orogastrically with equivalent numbers of differentially marked mutant and wild-type bacteria. Three weeks post infection, tissues were harvested and cfu enumerated. In the caecum, spleen and liver, the mutant bacteria out-competed the wild type by 14- to 17-fold (Figs 8A–E and 9A). An opuA–D, otsBA double-mutant strain competed equivalently with the wild-type strain, suggesting the increased colonization phenotype of ΔopuA–D mutants requires the trehalose production operon. An otsBA mutant also competed equivalently with wild type for colonization in mice, indicating that trehalose production genes are not needed during systemic colonization. These data suggest that strains lacking opuA–D out-compete wild-type S. Typhimurium for tissue colonization in mice, and require otsBA to do so. Overall, the data indicate that the OpuA–D locus plays a role in limiting virulence.

Figure 8.

ΔopuA–D strains require the otsBA locus to out-compete wild-type strains in mice. Mice were orally infected with a 1:1 mixture of the strains indicated on the x-axes. After 3 weeks, mice were euthanized and their organs [caecum (C), Peyer's patches (PP), mesenteric lymph nodes (MLN), spleen (S) and liver (L)] immediately collected for analysis. Bacteria were enumerated after plating homogenized (A) caecum, (B) Peyer's patches, (C) mesenteric lymph nodes, (D) spleen and (E) liver. Each symbol represents one mouse, n = 6–23. P values are # < 0.01 compared with the null hypothesis and * < 0.05 compared with ΔopuA–D vs WT.

Figure 9.

Fasting time affects gastric pH and colonization of ΔopuA–D strains in mice. A. Mice were fasted for 9–12 h (closed circles) or 2–4 h (open circle) prior to oral inoculation with a 1:1 mixture of the ΔopuA–D and wild-type strains. After 3 weeks, mice were euthanized and their organs [caecum (C), Peyer's patches (PP), mesenteric lymph nodes (MLN), spleen (S) and liver (L)] immediately collected for analysis. Each symbol represents one mouse, n = 13–23. Asterisks indicate P < 0.01 (*), 0.001 (**) or 0.001 (***) compared with the null hypothesis. B. The pH of the gastric content was measured for mice fasted for 3 h (short) and 11 h (long); control (ctrl) mice were not fasted. Error bars are SD, n = 3–4 mice. Asterisks indicate P < 0.05 (*) and 0.01 (**) compared with control.

A long fast prior to inoculation is required for ΔopuA–D mutants out-compete wild-type strains in mice

While, a 9–12 h fast is standard in the field, members of the University of Colorado Institutional Animal Care and Use Committee encouraged the testing of shorter fasts to reduce animal stress. Therefore, another set of infections was performed in which food was withheld for only 2–4 h prior to infection. No differences were observed in tissue colonization between ΔopuA–D mutants and wild-type strains (Fig. 9A). These data suggest that a physiological consequence(s) of a long fast is needed for ΔopuA–D mutants to out-compete wild-type strains in mice.

The gastric pH of long-fasted mice is not lower than that of short-fasted mice

Given that ΔopuA–D mutant strains withstand low pH better than wild type, it was possible that the pH of the gastric contents of the mice was lower after a long versus a short fast. To test this hypothesis, gastric content pH was measured in mice after a 3 h and an 11 h fast. Control mice were not fasted and had a mean gastric pH of 3.9 (Fig. 9B). After a short fast the mean gastric pH was 2.5, greater than 10-fold more acidic than that of control mice. After a long fast the gastric pH was variable and ranged from 2.5 to 3.6. These data indicate that a long fast does not consistently increase gastric acidity, nor result in lower gastric pH than a short fast. Therefore, gastric acidity is not the dominant stress that results in increased tissue colonization by ΔopuA–D mutant strains relative to wild-type strains.

Discussion

Data within indicate that in the absence of the OpuABCD predicted glycine betaine importer, exposure to stress results in increased trehalose production via otsBA, and the bacteria develop enhanced resistance to multiple insults. The otsBA operon is activated by the stress–response sigma factor RpoS (Hengge-Aronis et al., 1991; Fang et al., 1996). RpoS is activated in M9 minimal media in high salt (Dong and Schellhorn, 2009), conditions under which significant trehalose production occurs. Upon infection of macrophages or mice, strains lacking opuA–D out-compete wild-type strains, a phenotype that requires otsBA. Pathogens do have access to glycine betaine in animals, which is present at approximately 35 µM in human blood and at higher concentrations in tissues (Lever and Slow, 2010). Only 20 µM glycine betaine is sufficient to restore the growth of P. syringae opuA–D mutants in the presence of 0.4 M sodium chloride (Chen and Beattie, 2007), suggesting that the concentration of glycine betaine in human blood is physiological with respect to bacterial stress responses. In contrast mammals do not make trehalose (Richards et al., 2002). These observations together suggest that during infection the S. Typhimurium OpuABCD predicted glycine betaine importer limits trehalose production, perhaps to maintain the balance between host and pathogen necessary for chronic infection.

Trehalose is produced by bacteria, archaea, fungi, plants and invertebrates in response to environmental stress and is hypothesized to stabilize proteins and lipids by replacing or trapping water at the surface of macromolecules (Fedorov et al., 2011). However, multiple laboratories have demonstrated that in S. Typhimurium, trehalose production is dispensable for pathogenicity. In BALB/C (Nramp1−/−) mice, which do not survive infection with S. Typhimurium, otsA mutants are fully virulent with regard to lethality and tissue colonization upon intraperitoneal and intragastric inoculation (Fang et al., 1996; Howells et al., 2002). In Sv129S6 (Nramp1+/+) mice, which survive acute infection and become chronically infected with S. Typhimurium, the otsBA operon plays no detectable role in tissue colonization. While apparently unnecessary for S. Typhimurium virulence, in other microbes trehalose is a demonstrated virulence determinant. In the fungal pathogen Candida albicans, trehalose production defends the wild-type yeast against reactive oxygen species produced by macrophages (Martínez-Esparza et al., 2007). In another fungal pathogen Cryptococcus gattii, trehalose production genes are important in broth for melanin synthesis, capsule formation and cell wall integrity. In mice, C. gattii requires trehalose production genes for colonization of the lung and brain (Ngamskulrungroj et al., 2009). Thus, trehalose accumulation can protect microbes from stress encountered in the host. The amount of trehalose produced by a pathogen may depend on the evolutionarily determined optimal tissue colonization level for that pathogen.

It was surprising that the colonization advantage of strains lacking the OpuABCD putative glycine betaine importer was a function of the period of time for which mice had fasted prior to oral gavage, as ΔopuA–D strains out-competed wild-type bacteria in mice fasted for 9–12, but not 2–4 h. One possibility was that gastric pH was lower after a long fast than after a short fast. However, this idea was not supported by data, suggesting that stomach pH is not the major contributing factor in the colonization advantage of ΔopuA–D. An alternative hypothesis is that long fasts may trigger inflammatory responses in mice. This notion has some support in the literature, as within 1 day of fasting, levels of the cytokine IL-12 increase in the liver and mononuclear cells increase expression of superoxide dismutase, which generates hydrogen peroxide (Shen et al., 2009). Inoculation after a 9–12 h fast may thus expose S. Typhimurium to macrophages that are more aggressively activated and/or have higher levels of oxygen radicals, resulting in trehalose production and increased survival in ΔopuA–D mutant strains. However, a considerable body of work will be required to test this and other hypotheses. Finally, long fasts may not be unnatural, as animals in the wild, including the house mouse, eat significantly less food over time than they prefer (Boutin, 1990). Thus, while members of an Institutional Animal Care and Use Committee may prefer the use of 2–4 h fasting periods, a 9–12 h fast may be a physiologically realistic time-frame.

Our data demonstrate that ΔopuA–D glycine betaine importer mutant strains out-compete wild-type strains in tissues by more than 10-fold, which is consistent with the role of OpuABCD as an antivirulence locus, as defined by the ability to limit one or more aspect of pathogenicity (Ho and Slauch, 2001). Previously described loci that limit virulence include ZirTS, a two-partner secretion system. The absence of functional ZirTS, like Opu, has a small and insignificant effect on the mouse survival. However, zirTS mutant strains out-compete wild-type strains in mixed-infection experiments by 3- to 13-fold (Gal-Mor et al., 2008). The mechanism(s) by which ZirTS limits virulence remains unknown. The S. Typhimurium global regulator of transcription leucine-responsive regulatory protein (LRP) also limits virulence. Strains lacking LRP out-compete wild type by three- to ninefold during mixed infection. It has been suggested that LRP functions indirectly to curtail colonization, as it negatively regulates type three secretion systems required for virulence (Baek et al., 2009). Thus, LRP mutants may have increased virulence due to increased or ectopic expression of virulence determinants.

Despite its potential role in limiting colonization in mice, an opu locus is present throughout the six subspecies of S. enterica at the orthologous position in the genome, including the human pathogens Typhi and Paratyphi A–C. However, other Enterobacteriaceae family members including Salmonella bongori , E. coli, Shigella and Klebsiella do not encode genes resembling an opu locus in the orthologous regions of their genomes. This suggests that an opu cassette was acquired by S. enterica after divergence from S. bongori. One possibility is that during systemic infection, which is generally associated with S. enterica but not S. bongori, there is selective pressure for Salmonellae to keep colonization of host tissues under a threshold defined by the host response to the pathogen. For instance, S. Typhimurium fitness may depend upon preventing excessive tissue damage that threatens survival of the host and thereby the bacteria. A smaller population of microbes may also be better able to establish and maintain chronic infection by eliciting a more modest host immune response. Selective pressures to retain the trehalose synthesizing genes otsA and otsB may reflect the importance of trehalose in withstanding diverse non-host environmental stress, such as desiccation and fluctuations in temperature. Overall, the data indicate that pathogens carefully regulate responses to stress encountered within the host environment.

Experimental procedures

Bacterial strains and culture conditions

Strains (Table 2) were derived from S. enterica serovar Typhimurium strain SL1344 (Smith et al., 1984). Bacteria were grown aerobically in Luria–Bertani (LB) medium or M9 minimal media (M9) [1 mM magnesium sulphate, 2% dextrose, 0.002% l-histidine, 0.05% casamino acids and 20% M9 5× salts (211.3 mM disodium hydrogen phosphate, 110 mM monopotassium phosphate, 93.4 mM ammonium chloride, 42.77 mM sodium chloride)], pH 7.4 at 37°C, with antibiotics as indicated: streptomycin, 30 µg ml−1; ampicillin, 50 µg ml−1; kanamycin 30 µg ml−1; and chloramphenicol 20 µg ml−1. For growth curves, overnight cultures were diluted to an OD600 of 0.01 in 200 µl of M9 at pH 7, pH 5, or with 0.4 M added sodium chloride. Bacteria were grown in a 96-well plate shaking in a Synergy2 plate reader (BioTek) at 37°C and the OD600 was recorded at 20 min intervals.

Table 2.  Bacterial strains used in this study related to experimental procedures.
StrainGenotype/plasmidReference
SL1344 hisG xyl rpsL (wild type) Smith et al. (1984)
14028 ATCC
csd22114028 s/pKD46 Detweiler et al. (2003)
kde444 STM1491–1494 (termed: ΔopuA–D)::cm Erickson and Detweiler (2006)
nds692 otsBA::kan (termed: ΔotsBA)This study
mcp783SL1344/pRB3-273cThis study
mcp786SL1344/potsBAThis study
mcp787 otsBA::kan/pRB3-273cThis study
mcp790 otsBA::kan/potsBAThis study
mcp809kde444/pRB3-273cThis study
mcp813kde444, otsBA::kanThis study
mcp854mcp813/pRB3-273cThis study
mcp855mcp813/potsBAThis study
mcp856kde444/potsBAThis study

Strain and plasmid construction

Deleted genes were replaced with kanamycin or chloramphenicol resistance markers (Datsenko and Wanner, 2000). Oligos used for generating otsBA::kan are otsBA-P1 (CGTTTGTGAGTCTCAATATGATGATAAGGAGGAGACCAGGGTGTAGGCTGGAGCTGCTTC), and otsBA-P2 (TGGTGCCCTTAGCGGGCGACTAGTCGCCGCTCGCGATATTCATATGAATATCCTCCTTA). Underline denotes P1 or P2 primer. Mutations were made in the 14028 background strain (source ATCC), verified by PCR [otsBA-5′ (ACTTACATGACTAATGAGAC) and otsBA-3′ (CAGCCAGGTAGATGTGTTGC)], and transduced into SL1344 using standard P22 phage transduction. Transductants were verified by growth on LB agar containing antibiotic and by PCR. Genomic DNA from strain SL1344 was used as a template to amplify otsB and otsA (−228 from otsB ORF to +1481 from otsA ORF) with 5′ primer #874 (TCTAAAGCTTAGCCAGGTAGATGTGTTGCT) and 3′ primer #875 (TTGGATCCGGCTGAATCCTTCTGACAAC). Underline denotes added sites for restriction endonucleases. The otsBA PCR product was digested with HindIII and BamHI and ligated into the same sites of pRB3-273c (Berggren et al., 1995) to construct potsBA. This plasmid was transformed into Salmonella by electroporation.

Carbohydrate extraction and quantification

Strains were grown overnight in 100 ml of M9 minimal media containing 0.01 M or 0.4 M NaCl. Bacteria were pelleted and resuspended in water to 4.6 × 108 bacteria per µl. Carbohydrates were extracted by incubation at 95°C for 20 h. Debris was pelleted and aqueous extracts (0.5 µl) were spotted onto a silica gel TLC plate and separated using an ethyl acetate–acetic acid–water–pyridine [26:14:7:2 (v/v/v/v)] solvent system. Resolved sugars were stained with 0.1% potassium permanganate in 1 M NaOH. Integrated densitometry analysis was performed with Image J software version 1.44 (NIH). Glucose content of aqueous extracts was additionally quantified with a Glucose (HK) Assay Kit (Sigma) according to the manufacturer's protocol. Trehalose levels were calculated by subtracting the glucose content of ΔotsBA mutant strains from wild-type and ΔopuA–D mutant strains and then dividing by two.

Hydrogen peroxide resistance

Overnight cultures were centrifuged at 4000 g for 10 min and cell pellets were resuspended in equal volume PBS. Cultures were diluted 1/10 in M9 broth and grown at 37°C for 1 h.

Broth resistance assay.  Cultures were further diluted to an OD600 of 0.05 in 200 µl of M9 broth with 10 mM hydrogen peroxide, corresponding to 1 nmol of hydrogen peroxide per bacterium and in accordance with previous studies (Fang et al., 1992; Tu et al., 2006). Bacteria were grown shaking in a 96-well plate at 37°C for 90 min. Bacterial survival was determined by plating on selective LB agar plates for cfu. Exposure of the strains to 0.1 mM hydrogen peroxide revealed no differences in strain survival.

Filter disc inhibition assay.  A total of 1 × 106 bacteria were spread on an LB streptomycin plate and four filter discs treated with either water or 150 mM H2O2 were overlayed. Zones of inhibition were measured after incubation at 37°C overnight.

Bacterial infection of cultured bone marrow-derived macrophages (BMDMs)

Bone marrow-derived macrophages (BMDMs) isolated as described (Silva-Herzog and Detweiler, 2010) were seeded at 1.5 × 105 cells per well in 24-well tissue culture plates for at 4 h before treatment with 2 ng ml−1 recombinant murine IFN-y (Peprotech, Rocky Hill, NJ), 20 µg ml−1 LPS from S. Typhimurium (Sigma). Where indicated, cells were treated with 25 µmol l−1 MDL 72527, a spermine oxidase inhibitor (Sigma). MDL 72527 had no significant affect on macrophage viability under these conditions as determined by trypan blue staining and cell counting. After 18–20 h at 37°C in 5% CO2, infection experiments were performed in triplicate. Bacteria were grown aerobically in LB medium overnight. The overnight cultures were diluted 1/5 in M9 minimal medium and exposed to 0.4 M sodium chloride for 1 h. The bacteria were then opsonized with 20% normal mouse serum (Sigma) and added to BMDMs at a multiplicity of infection (moi) of 10–13 for each strain (total moi of approximately 25), determined by plating for cfu. Infections were synchronized by centrifugation at 50 g for 5 min. After 30 min, gentamicin (100 µg ml−1) was added to kill extracellular bacteria, 90 min later, 10 µg ml−1 gentamicin was added to prevent extracellular replication of bacteria. At 1 and 2 h post infection, cells were washed twice with PBS, incubated with 0.1% Triton X-100 for 5 min, and lysed. Serial dilutions were plated to quantify cfu and the competitive indexes (CIs) were calculated: (cfumutant/cfuwild-type) output/(cfumutant/cfuwild-type) input (Beuzón and Holden, 2001). Levels of ROS were determined using H2DCFDA (Invitrogen, Carlsbad, CA) following the manufacturer's protocol (data not shown).

Determination of the LD50

Six- to eight-week-old, female, BALB/c mice were intragastrically inoculated with wild-type or ΔopuA–D strains at concentrations from 103 to 108 bacteria. Mice were monitored daily for 28 days and deaths were recorded as days post infection. The LD50 was calculated as previously described (Beuzón and Holden, 2001). The confidence interval was calculated by logit analysis (Reed and Muench, 1938).

Competition assays

Mice.  Seven-week-old SV129S6 mice (Taconic Laboratory, Hudson, NY) were orogastrically inoculated with equivalent numbers (1 × 109) of two bacterial strains (2 × 109 total), as verified by plating for cfu on selective LB agar. Three weeks after inoculation, tissues (caecum, liver, spleen, MLN and Peyer's patches) were collected in 1 ml of PBS, homogenized with a TissueMiser (Fisher Scientific, Pittsburgh, PA) and diluted in PBS for plating on selective LB agar plates. Colony-forming units were enumerated and CI values were calculated as described above. Research protocols were in accordance with the NIH Guide for the Care and Use of Laboratory Animals and approved by the University of Colorado Institutional Biosafety and Animal Care and Use Committees.

Broth.  The ΔopuA–D and wild-type strains were grown overnight in LB medium at 37°C. Experiments were performed in triplicate. Overnight bacterial cultures were centrifuged at 4000 g for 10 min and cell pellets were resuspended in equal volume phosphate-buffered saline (PBS). Equal numbers, 2 × 109 total bacteria (verified by plating for cfu) of the ΔopuA–D and wild-type strains were inoculated in LB broth or M9 minimal medium and incubated at 37°C for 24 h. At varying time points, cells were diluted and plated on selective LB agar plates. CIs were calculated as described above.

Gastric pH

After fasting for 3 or 11 h, or not fasting control mice, mice were euthanized and their stomachs dissected. The stomach was cut laterally and the stomach content was scraped into an eppendorf tube. The gastric content was centrifuged at 4000 g for 5 min and the supernatant was applied to a pH indicator strip pH 2.5–4.5 (EMD Chemicals, Gibbstown, NJ).

Statistics

Data were analysed using InStat version 3.1a (GraphPad Software, San Diego, CA). Two-way comparisons were analysed with a Student's t-test (parametric), Mann–Whitney (non-parametric), one-sample t-test (parametric, comparison with null hypothesis) or Wilcoxon rank sum test (non-parametric, comparison with null hypothesis). Multiple comparisons were examined by analysis of variance (anova) with a Tukey (parametric) or Dunn (non-parametric) post-test respectively (Smith and Bidochka, 1998).

Acknowledgements

We thank M.W. McCoy and N.R. Pace for discussions, K. Erickson and T. Fenn for the preliminary observation that ΔopuA–D mutants out-compete wild-type and the LD50 data. We thank N. Sanchez for help with ΔotsBA infection experiments and strain construction. This work was supported by AI072492 and AI095395 to C.D.

Ancillary