The intestinal fatty acid propionate inhibits Salmonella invasion through the post-translational control of HilD

Authors

  • Chien-Che Hung,

    1. Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA
    Search for more papers by this author
  • Cherilyn D. Garner,

    1. Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA
    Search for more papers by this author
  • James M. Slauch,

    1. Department of Microbiology, University of Illinois, Urbana, IL, USA
    2. College of Medicine, University of Illinois, Urbana, IL, USA
    Search for more papers by this author
  • Zachary W. Dwyer,

    1. Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA
    Search for more papers by this author
  • Sara D. Lawhon,

    1. Department of Veterinary Pathobiology, College of Veterinary Medicine and Biomedical Sciences, Texas A&M University, College Station, TX, USA
    Search for more papers by this author
  • Jonathan G. Frye,

    1. Bacterial Epidemiology and Antimicrobial Resistance Research Unit, U.S. Department of Agriculture, Agriculture Research Service, Richard B. Russell Research Center, Athens, GA, USA
    Search for more papers by this author
  • Michael McClelland,

    1. Department of Pathology and Laboratory Medicine, University of California, Irvine, Irvine, CA, USA
    Search for more papers by this author
  • Brian M. M. Ahmer,

    1. Department of Microbiology, The Ohio State University, Columbus, OH, USA
    2. Center for Microbial Interface Biology, The Ohio State University, Columbus, OH, USA
    Search for more papers by this author
  • Craig Altier

    Corresponding author
    • Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA
    Search for more papers by this author

For correspondence. E-mail altier@cornell.edu; Tel. (+1) 607 253 3926; Fax (+1) 607 253 3907.

Summary

To cause disease, Salmonella must invade the intestinal epithelium employing genes encoded within Salmonella Pathogenicity Island 1 (SPI1). We show here that propionate, a fatty acid abundant in the intestine of animals, repressed SPI1 at physiologically relevant concentration and pH, reducing expression of SPI1 transcriptional regulators and consequently decreasing expression and secretion of effector proteins, leading to reduced bacterial penetration of cultured epithelial cells. Essential to repression was hilD, which occupies the apex of the regulatory cascade within SPI1, as loss of only this gene among those of the regulon prevented repression of SPI1 transcription by propionate. Regulation through hilD, however, was achieved through the control of neither transcription nor translation. Instead, growth of Salmonella in propionate significantly reduced the stability of HilD. Extending protein half-life using a Lon protease mutant demonstrated that protein stability itself did not dictate the effects of propionate and suggested modification of HilD with subsequent degradation as the means of action. Furthermore, repression was significantly lessened in a mutant unable to produce propionyl-CoA, while further metabolism of propionyl-CoA appeared not to be required. These results suggest a mechanism of control of Salmonella virulence in which HilD is post-translationally modified using the high-energy intermediate propionyl-CoA.

Introduction

The complex chemical environment of the animal intestinal tract is created in large part by the vast number of bacteria that populate that organ. This environment is nearly devoid of oxygen, particularly in the large intestine, promoting the selective growth of bacterial species capable of the anaerobic metabolism of nutrients. In the absence of oxygen and other terminal electron acceptors, this is accomplished through fermentation, where energy is derived through substrate-level phosphorylation reactions, with reduced organic molecules being produced and excreted. Predominant among these molecules are the short-chain fatty acids acetate, butyrate and propionate, which can be present in high concentrations in the animal intestine (Argenzio et al., 1974; Argenzio and Southworth, 1975; Cummings et al., 1987; Macfarlane et al., 1992). These fatty acids are known to have numerous effects upon the host, and in high concentrations also inhibit the growth of some bacterial species. They have thus been investigated as possible preventatives or therapies for enteric bacterial infections (Levison, 1973; McHan and Shotts, 1993; Shin et al., 2002), either directly or through the manipulation of the intestinal microbiota to cause global changes in their production.

Salmonella is a ubiquitous pathogen that has evolved to survive and proliferate within this intestinal environment, infecting a wide variety of animal species and causing enteric disease. Serovars of Salmonella enterica are important sources of bacterial food-borne disease, causing illnesses ranging from self-limiting enteritis to life-threatening septicaemia, with the invasion of the epithelium that lines the intestine being an early required step for pathogenesis. Invasion of the intestinal tract is mediated by a type three secretion system (TTSS) that is encoded within Salmonella Pathogenicity Island 1 (SPI1) (Galan and Curtiss, 1989; Behlau and Miller, 1993; Groisman and Ochman, 1993). The TTSS forms a ‘needle complex’ (Kubori et al., 1998) that is used to inject effector proteins into host cells, causing rearrangement of the host cell cytoskeleton and leading to engulfment of the bacterium (reviewed by Galan, 2001). The SPI1 TTSS is used again once the bacteria are engulfed by macrophages, inducing pyroptosis, the proinflammatory cell death that characterizes salmonellosis (Fink and Cookson, 2007). Invasion, and hence the genes encoded in SPI1, are thus required for both the intestinal and septicaemic forms of disease (Galan and Curtiss, 1989; Jones and Falkow, 1994; Wallis and Galyov, 2000; Barthel et al., 2003).

Salmonella invasion is mediated by a number of environmental cues known or likely to be present within the intestine. Among these, short-chain fatty acids have been demonstrated to affect invasion in complex ways. Acetate and the weak acid formate induce the expression of SPI1 genes (Lawhon et al., 2002; Huang et al., 2008), while butyrate and propionate have been shown to have the opposite, repressive effect on the genes and functions of SPI1 (Lawhon et al., 2002; Gantois et al., 2006; Boyen et al., 2008). It has further been demonstrated that the concentrations and composition of fatty acids vary within regions of the intestine. Thus, the conditions in the distal small intestine, the area previously described as the location for invasion (Carter and Collins, 1974; Jones et al., 1994), reflect those known to induce invasion genes (Lawhon et al., 2002; Huang et al., 2008). Conversely, the high propionate and butyrate concentrations of the caecum and colon may be more likely to repress invasion (Lawhon et al., 2002). In fact, in vivo studies in mice have shown a correlation between a decrease in large intestinal short-chain fatty acid concentrations and an increased susceptibility to Salmonella infection (Voravuthikunchai and Lee, 1987; Garner et al., 2009).

In addition to the environmental regulators of invasion, an array of genetic factors has been shown to mediate invasion gene expression (Fig. 1). Invasion genes are controlled by a complex regulatory network of transcriptional and post-transcriptional regulators. Within SPI1, HilD, a transcriptional regulator of the AraC family, comprises a portion of an auto-regulatory circuit that includes RtsA and HilC (Ellermeier et al., 2005). Together, these induce expression of hilA, itself encoding a transcriptional activator of the inv/spa, prg/org and sic/sip operons that encode the secretion apparatus and secreted effector proteins required for cytoskeletal rearrangement (Bajaj et al., 1995; Darwin and Miller, 1999; Lostroh et al., 2000), but can also induce SPI1 expression independent of HilA (Altier et al., 2000a). Outside SPI1, Lon protease, HilE, FliZ and the BarA/SirA two-component regulator all control HilD through post-transcriptional mechanisms (Baxter et al., 2003; Boddicker and Jones, 2004; Chubiz et al., 2010; Martinez et al., 2011).

Figure 1.

A model for the control of SPI1 by propionate and genetic regulators. Arrows indicate positive control, while T-shaped bars indicate negative control.

To better understand the mechanism by which intestinal fatty acids repress Salmonella invasion, in this work we examined the effects of propionate on SPI1 gene expression and invasion, and investigated the metabolic and genetic pathways required for these effects. We found that at a physiologically relevant concentration, propionate repressed invasion and that pH was important for the repressive effect. Additionally, as the pathways for propionate metabolism have previously been characterized in Salmonella, we used genetic approaches to study the metabolic routes and products required for the repressive effect of this fatty acid. We show here that metabolism of propionic acid is necessary for its repressive effect and that in particular, the high-energy metabolic intermediate propionyl-CoA is likely required. We further demonstrate that, among the many regulators of invasion, the repressive effect of propionate functions solely through the central SPI1 regulator HilD, but that this control is evoked through neither transcription nor translation. Instead, growth of bacteria in propionate reduces the stability of HilD, implicating a post-translational method of control.

Results

Propionate inhibits Salmonella invasion through the repression of SPI1 gene expression

Fatty acids are the predominant metabolic product of the anaerobic intestinal microbiota and so largely define the environment of the mammalian intestinal tract. Previous work in our laboratory and those of others has shown that propionate, a major constituent of the large intestinal environment, reduces Salmonella SPI1 invasion gene expression in vitro when supplied alone or in combination with other fatty acids (Durant et al., 2000; Lawhon et al., 2002; Garner et al., 2009). To identify the propionate regulon more completely, we first performed microarray analyses, comparing gene expression of Salmonella serovar Typhimurium grown in buffered LB broth to that grown in buffered LB with the addition of propionate. We found that the expression of a majority of SPI1 genes was repressed by growth in the medium containing propionate. Of the 35 genes of SPI1, 22 were reduced in their expression by at least twofold with the addition of this fatty acid (Table S1). Among these were the central transcriptional regulators of SPI1 hilD and hilA, although the expression of hilC, an additional transcriptional regulator involved in the complex regulation of invasion, was not repressed. Propionate, being a weak acid, can enter the bacterial cytoplasm and can, at least in high concentration, reduce transmembrane potential and cytoplasmic pH (Repaske and Adler, 1981; Slonczewski et al., 1981). To determine whether the observed changes in gene expression might be due to this generic mechanism, we also employed microarrays to examine gene expression when bacteria were grown in medium containing acetate, a short-chain fatty acid with a pKa similar to that of propionate. In contrast to the effects of propionate, no SPI1 gene demonstrated reduced expression in the presence of acetate (Table S1). In fact, 16 of 35 SPI1 genes were induced at least twofold by acetate, supporting our previous finding that this fatty acid can induce SPI1 gene expression (Lawhon et al., 2002), and demonstrating that repression by propionate is not accomplished simply by the accumulation of weak acid within the bacterial cytoplasm. Additionally, we found that the most severe repressive effects of propionate were manifested within SPI1. Of the genes demonstrating repression by propionate but not by acetate, 20 of the 30 with the greatest repression were either encoded within SPI1 or were controlled by SPI1 regulators (Table 1). To confirm the microarray findings, we tested the effects of propionate on key SPI1 regulators and effectors by independent methods. Using media buffered to pH 6.7 and with 10 mM propionic acid, conditions designed to mimic those of the murine large intestine (Garner et al., 2009), we employed lacZY reporter fusions to the SPI1 invasion genes sipC, which encodes a secreted effector protein, and hilA and invF, both transcriptional regulators. We found there to be a significant decrease in gene expression for each of these fusions, more than eightfold for sipC and fivefold for each invF and hilA, when grown to stationary phase in the presence of propionic acid (Fig. 2A and B). As the oxygen tension is low within the intestinal lumen, we also tested the effect of propionate on these key invasion regulators when bacterial cultures were grown under strict anaerobic conditions. Expression of invF and hilA was lessened in anaerobically grown cultures in the absence of propionate, compared to microaerophically grown cultures, but expression of both remained significantly reduced by its addition (Fig. 2C). Weak acids, including propionic acid, are known to accumulate within the bacterial cytoplasm when cytoplasmic pH exceeds that of the external medium (Salmond et al., 1984). Additionally, previous work had shown that pH is important for the effects of fatty acids on invasion gene expression, suggesting that accumulation of the fatty acid in the bacterial cytoplasm is necessary for the observed effects (Lawhon et al., 2002; Huang et al., 2008). In contrast to the repression observed when the medium was maintained at pH 6.7, we found that at pH 8.0 propionate failed to repress sipC, invF or hilA. Instead, there was a slight increase in gene expression in the presence of propionate (Fig. 2A and data not shown). Combined, these results suggest that, although propionate does not exert its repressive effects by acidification of the bacterial cytoplasm, it must still enter the bacterium to repress the genes of SPI1.

Figure 2.

Propionate represses Salmonella invasion gene expression.

A. Propionate represses sipC expression. Cultures of a sipC::lacZY reporter fusion strain were grown as standing cultures without additive (black bars) or with 10 mM propionic acid (grey bars), with β-galactosidase assays used to assess sipC expression.

B. Propionate represses regulators of Salmonella SPI1. Cultures of strains with the hilA::lacZY or invF::lacZY reporter fusion were grown without additive (black bars) or with 10 mM propionic acid (grey bars). β-Galactosidase assays were used to assess hilA and invF expression.

C. Propionate represses sipC expression under anaerobic conditions. Cultures of a sipC::lacZY reporter fusion strain were grown anaerobically without additive (black bars) or with 10 mM propionic acid (grey bars), with β-galactosidase assays used to assess sipC expression. Error bars indicate standard deviation, and an asterisk (*) indicates a statistically significant difference due to the presence of propionic acid at P < 0.05.

Table 1. Salmonella genes most severely repressed by propionate but not repressed by acetate
Gene numberaGene symbolMedian expression with propionateb
  1. aGenes shown in bold are located within SPI1 or are controlled by SPI1 regulators.
  2. bProportion of expression for each gene is given in comparison to that produced during growth in medium without fatty acid added.
STM1091 sopB0.2
STM4315 rtsA0.2
STM2883 sipD0.2
STM2898 invG0.3
STM2900 invH0.3
STM2886 sicA0.3
STM2884 sipC0.3
STM2874 prgH0.3
STM2885 sipB0.3
STM2882 sipA0.3
STM1090 pipC0.3
STM3138 mcpA0.3
STM2873 prgI0.3
STM2871 prgK0.3
STM4464 0.4
STM4465 0.4
STM2872 prgJ0.4
STM3216 mcpC0.4
STM1690 pspA0.4
STM2153 yehE0.4
STM2315 yfbK0.4
STM2897 invE0.4
STM0271 0.4
STM4265 soxS0.4
STM1513 0.4
STM2879 sicP0.4
STM1774 sirC0.4
STM2880 0.4
STM1239 0.4
STM2875 hilD0.4

Previous studies have shown that propionate reduces Salmonella penetration of intestinal epithelial cells (Durant et al., 1999; Van Immerseel et al., 2004a; Boyen et al., 2008), and we found using a gentamicin-protection assay that overnight growth in propionic acid prior to infection significantly decreased invasion of cultured HEp-2 cells, by twofold (Fig. 3A). When grown in laboratory media, Salmonella secretes into the culture media the invasion proteins SipA, B, C and D, necessary for the penetration of epithelial cells (Hueck et al., 1995). Bacteria grown in medium with no additive produced four bands of apparent molecular weights equivalent to these proteins (Fig. 3B). Subsequent analysis by mass spectroscopy showed the four proteins to indeed be SipA, B, C and D (not shown). These four proteins were greatly diminished, however, in extracts from the same strain grown in the presence of propionic acid, reduced in their amounts by 12-, 12-, 4- and 5-fold, respectively, demonstrating that the repression of invasion genes is also manifested as a reduction of the proteins they encode. These results, taken together, show that propionate inhibits expression of invasion genes and the consequent production of secreted effector proteins, reducing invasion of epithelial cells.

Figure 3.

Propionate represses Salmonella invasion.

A. Propionate decreases Salmonella invasion of HEp-2 cells. The wild type strain was grown either with no additive or with 10 mM propionic acid. Invasion of HEp-2 cells was assessed using a gentamicin protection assay. Invasion is shown relative to the wild type without additive, which was set to 100%. Error bars indicate standard deviation, and an asterisk (*) indicates a statistically significant difference due to the presence of propionic acid at P < 0.05.

B. Propionate diminishes production of SPI1 secreted effector proteins. The wild type strain was grown with no additive (lane 1) or with 10 mM propionic acid (lane 2). Secreted proteins were then isolated and analysed by SDS-PAGE. Proteins with apparent molecular weights of 89, 67, 42 and 38 kDa that were reduced in cultures grown with propionic acid are designated with arrowheads. Molecular weights are shown on the right.

Metabolism of propionate is required for its repression of SPI1

As pH and thus the ability of propionate to enter the bacterial cytoplasm was important for its repressive effects on invasion genes, we next determined whether this fatty acid acted directly or whether it must first have been converted to a metabolic product to have its effect. As the pathways for propionate metabolism have been well characterized in Salmonella and Escherichia coli (Hesslinger et al., 1998; Horswill and Escalante-Semerena, 1999a,b; 2001; Palacios et al., 2003), we tested essential components of these pathways to determine whether they were important for the repressive effect of propionate. There are two characterized routes for the initial steps of propionate metabolism: propionate can be converted to propionyl phosphate by ackA or pduW and then metabolized to propionyl-CoA by a phosphotransacetylase, encoded by pta (Palacios et al., 2003). Alternatively, propionate can be metabolized directly to propionyl-CoA using the products of acs or prpE, which encode acyl-CoA synthetases (Horswill and Escalante-Semerena, 1999a). Propionyl-CoA is then converted to 2-methylcitrate by 2-methylcitrate synthase, encoded by prpC, and eventually to intermediates of the TCA cycle. A null mutant of prpE, acs and pta, eliminating all of the known routes of metabolism from propionate to propionyl-CoA, produced a significant, twofold increase in invasion gene expression in the presence of propionic acid when compared to that of the wild type grown with propionic acid, but the combined mutations failed to restore invasion gene expression to the level seen in the mutant without propionic acid added (Fig. 4A). It has been previously reported, however, that propionyl-CoA can be produced from endogenous 2-ketobutyrate through tdcE and pfl (Hesslinger et al., 1998). We thus created a mutant that deleted all of the known pathways for the production of propionyl-CoA from endogenous and exogenous sources (a prpE, acs, pta, pflB, tdcE mutant). We found that invasion gene expression was increased threefold from the wild type level in the presence of propionic acid in this mutant strain. We then examined the importance of the 2-methylcitrate pathway. We found a non-polar null mutant of prpC to have no effect on the repression of sipC::lacZY in the presence of propionate, suggesting that metabolic intermediates downstream from propionyl-CoA were unnecessary for this effect. To examine the specificity of these genetic pathways for the effects of propionate, we additionally tested a second short-chain fatty acid, butyrate, which is chemically similar to propionate and has also been shown to repress SPI1 genes (Durant et al., 1999; 2000; Lawhon et al., 2002; Gantois et al., 2006; Van Immerseel et al., 2006). For both the prpE, acs, pta mutant and the prpE, acs, pta, pflB, tdcE mutant, the increase in invasion gene expression seen in the presence of propionic acid was specific for this fatty acid, as butyric acid continued to fully repress sipC expression in these mutants, identically to its effect on the wild type strain (Fig. 4A). We further tested the importance of these metabolic pathways in bacteria grown under anaerobic conditions. The deletion of acs, prpE and pta in the absence of propionic acid had no effect on sipC expression, but the additional loss of endogenous propionyl-CoA production from 2-ketobutyrate, through mutation of pflB and tdcE, significantly increased that expression (Fig. 4B). Similarly, mutations predicted to prevent the production of propionyl phosphate from propionate, ackA and pduW, also increased this expression, further demonstrating the importance of propionyl-CoA to these effects. Unexpectedly, however, the additional of propionic acid to any mutant continued to repress sipC expression under anaerobic conditions (Fig. 4B). These results therefore implicate the production of propionyl-CoA as necessary for the negative effect of propionate on invasion and demonstrate its effect to be independent of generic effects that might conceivably be induced by fatty acids. Additionally, however, they suggest the presence of an uncharacterized anaerobically induced route for production of this molecule.

Figure 4.

Production of the intermediate propionyl-CoA is important for the repressive effect of propionate.

A. Wild type and mutant strains with the sipC::lacZY reporter fusion were grown under as standing cultures in media containing either no additive (black bars), 10 mM propionic acid (grey bars) or 10 mM butyric acid (white bars). β-Galactosidase assays were used to assess sipC expression. An asterisk (*) indicates a statistically significant difference for each respective strain grown with propionate or butyrate compared to no additive at P < 0.05. Two asterisks (**) indicates a statistically significant difference for a mutant strain grown with propionate compared to wild type grown with propionate at P < 0.05.

B. Wild type and mutant strains with the sipC::lacZY reporter fusion were grown under anaerobic conditions in media containing these same additives. An asterisk (*) indicates a statistically significant difference for mutants compared to the wild type. Error bars indicate standard deviation.

Repression by propionate functions through the central SPI1 regulator HilD

As we had shown that two important regulators of SPI1, invF and hilA, were repressed in the presence of fatty acids, we next determined whether other regulators of invasion were similarly affected. To address this, we examined three SPI1 regulators that occur further upstream in the invasion regulatory cascade: hilD and hilC, both encoded within SPI1, and rtsA, encoded outside the island, all of which are regulators of hilA (Fig. 1). As all of these are also known to regulate their own expression (Ellermeier et al., 2005), we used quantitative reverse transcription real-time PCR to determine relative gene expression without manipulation of the genes themselves. We found that messages of both hilD and rtsA were significantly decreased in the presence of propionic acid, while, consistent with our microarray data, that of hilC was not changed (Fig. 5A). These results show that propionate acts to decrease the expression of many regulators of invasion and acts either at the level of hilD or higher in the SPI1 regulatory cascade.

Figure 5.

Propionate represses SPI1 through the central regulator hilD.

A. Propionate reduces expression of SPI1 regulators. The wild type strain was grown with either no additive (black bars) or with 10 mM propionic acid (grey bars); total RNA was extracted and cDNA synthesis was performed. cDNA was then used as template for real-time PCR to measure relative expression of hilC, hilD and rtsA, with expression of these invasion genes normalized to that of 16S rRNA.

B. Activators of SPI1 other than hilD do not mediate the repressive effect of propionate. Wild type and mutant strains with the sipC::lacZY reporter fusion were grown in media containing either no additive (black bars) or 10 mM propionic acid (grey bars), and β-galactosidase assays were used to assess sipC expression. The wild type is shown using a separate axis to better discern the differences among mutant strains.

C. Inhibitors of SPI1 do not mediate the repressive effect of propionate. Wild type and mutant strains with the sipC::lacZY reporter fusion were grown in media containing either no additive (black bars) or 10 mM propionic acid (grey bars), and β-galactosidase assays were used to assess sipC expression. For all panels, error bars indicate standard deviation. An asterisk (*) indicates a statistically significant difference for a given strain with the addition of propionate as compared to no additive at P < 0.05.

HilD occupies a position at the apex of the regulatory cascade within SPI1, but can itself be controlled, both positively and negatively, by several genetic elements outside the island (Fig. 1). To determine the importance of HilD in repression by propionate, we examined the effects of mutations in genes known to encode regulators of HilD. Using the sipC::lacZY fusion, we first tested the positive regulators sirA and fliZ. sirA encodes a response regulator that induces expression of the regulatory RNAs CsrB and CsrC, which titrate the protein CsrA that can bind to hilD message and prevent translation (Altier et al., 2000a; Fortune et al., 2006; Martinez et al., 2011). FliZ has been shown to affect HilD by controlling its protein activity (Chubiz et al., 2010). Although expression of sipC was reduced, as expected, in mutants of either sirA or fliZ, culture of these strains in the presence of propionic acid continued to repress sipC to a degree proportionate to that of the wild type grown under the same conditions (Fig. 5B), indicating that propionate represses by a means independent of these two regulators. A mutant of ompR produced an identical result, also ruling out this known regulator of hilD as important to the effect of propionate (not shown). In contrast, the mutant of hilD itself became completely refractory to reduced sipC expression when propionic acid was present. We next tested two negative regulators of hilD: HilE is a well-characterized repressor of HilD, while Lon protease represses through its effects on either HilD itself or FliZ (Baxter et al., 2003; Boddicker et al., 2003; Boddicker and Jones, 2004; Chubiz et al., 2010). In both cases gene expression was increased when strains were grown without additive, compared to the wild type, but propionic acid continued to significantly reduce expression (Fig. 5C), indicating that neither of these repressors was important for HilD-mediated repression by propionate. Immediately downstream from HilD in the regulatory cascade lies HilA, itself a transcriptional activator (Fig. 1). Work presented here shows hilA to be one of the SPI1 genes repressed by propionate (Table S1 and Fig. 2B). To define the role of HilA in SPI1 repression, we further tested the effects of a hilA null mutant on sipC expression. We found that propionic acid continued to repress sipC in this mutant to a degree similar to that in the wild type (Fig. 5B). Thus, although propionate represses hilA through its control of hilD, control of downstream SPI1 genes by this fatty acid occurs, at least in part, by a HilA-independent mechanism. To verify the importance of HilD in repression of invasion by propionate, we additionally tested its effect on sopB, shown by microarray analysis to be strongly repressed by propionic acid (Table 1) and encoding a TTSS effector located outside SPI1 but controlled by SPI1 regulators (Ahmer et al., 1999; Pfeifer et al., 1999; Ehrbar et al., 2002). Using a plasmid-borne luxCDABE fusion to sopB and measuring light production over a time-course experiment, we observed that sopB exhibited the typical rise and fall in expression associated with growth phase that has been reported for genes controlled via SPI1 when grown in culture (Ernst et al., 1990; Lee and Falkow, 1990; Lundberg et al., 1999). The addition of propionic acid, however, significantly repressed this gene, greatly reducing its cumulative expression (Fig. 6A). Additionally, the loss of hilD severely reduced sopB expression and made sopB refractory to the effect of propionic acid (Fig. 6B). By contrast, sopB expression in a mutant of hilC, shown above to be unaffected by propionate, was indistinguishable from that of the wild type with or without the addition of propionic acid (Fig. 6C), and a hilC, hilD double mutant reproduced the phenotype of the hilD mutant alone (Fig. 6D). These results show that among known SPI1 regulators only hilD is required for the propionate-mediated repression of Salmonella genes essential for tissue invasion.

Figure 6.

hilD, but not hilC, is required for the repressive effect of propionate. (A) The wild type, (B) ΔhilD mutant, (C) ΔhilC mutant and (D) ΔhilC, ΔhilD mutant with the sopB::luxCDABE reporter fusion plasmid were grown in media containing either no additive (open symbols) or 10 mM propionic acid (closed symbols). Light production was used to assess sopB expression, and optical density at 600 nm (OD600) was used to measure growth. The measure of luminescence/OD600, in arbitrary units, was used to normalize for variation in growth rate. Error bars indicate standard deviation.

Although our results indicate that propionate functions to repress hilD by a means independent of the tested regulators, the control of this gene remains complex. HilD has been shown to regulate its own expression (Ellermeier et al., 2005), and thus mutants unable to produce the protein exhibit reduced expression of the gene. To determine specifically how propionate might affect hilD, we next examined its effects on transcription in the presence and absence of functional HilD. For this, we employed a transcriptional hilD::luxCDABE fusion carried on a plasmid in both wild type and ΔhilD strains. The fusion construct included the entire upstream region (to position −283 from the transcriptional start site) known to be required for maximal expression (Olekhnovich and Kadner, 2002), and so would be predicted to respond to the genetic and environmental regulators of hilD. As expected, we found that in the wild type strain propionic acid greatly affected hilD expression, reducing peak expression by twofold and total expression over the course of the experiment by 3.3-fold (Fig. 7). The loss of the chromosomal copy of hilD, however, both further reduced hilD::luxCDABE expression and eliminated the effect of propionic acid, demonstrating the necessity of intact HilD for these effects.

Figure 7.

Functional HilD is required for the repressive effect of propionate on hilD expression. Wild type (circles) and the ΔhilD mutant (squares) with the hilD::luxCDABE reporter fusion plasmid were grown in media containing either no additive (open symbols) or 10 mM propionic acid (closed symbols). Light production was used to assess hilD expression, and optical density at 600 nm (OD600) was used to measure growth. The measure of luminescence/OD600, in arbitrary units, was used to normalize for variation in growth rate. Error bars indicate standard deviation.

Propionate functions through the post-translational regulation of HilD

For genes that do not undergo auto-regulation, such control of a transcriptional fusion in a wild type strain would clearly indicate regulation at the level of transcription. The requirement here, however, that HilD be present allowed the possibility that propionate functions by controlling the chromosomal copy of hilD, which then affects the hilD::luxCDABE fusion, and that such control could be gained at any one of a number of levels. To examine the specific mechanism by which propionate exerted its control of SPI1 through hilD, we employed a transcriptional fusion of lacZ to hilA, the regulator immediately downstream from hilD in the regulatory cascade (Fig. 1). As the loss of hilD so profoundly reduced the expression of SPI1 genes, we induced the expression of hilA by including in the strain an allele of rtsA under the control of the tetRA promoter, allowing induction of rtsA with tetracycline, and thus of hilA as well (Golubeva et al., 2012). As shown in Fig. 8A, increasing concentrations of tetracycline commensurately increased hilA expression, but propionate significantly reduced that expression. This effect was dependent upon hilD, with a ΔhilD mutant being completely refractory to repression by propionic acid. We next used a similar system to determine whether control of hilD by propionate was achieved at the level of transcription. Here we again used the hilAlac transcriptional fusion, but in this case relieved hilD of its normal control by replacing its promoter with that of tetRA, providing inducible expression. We found, as expected, that increasing the concentration of tetracycline increased hilA expression, but also that the addition of propionic acid to the growth medium significantly reduced that expression in all cases (Fig. 8B). The effect of propionate was more pronounced at lower tetracycline concentrations (e.g. 0.1 μg ml−1 produced a greater than fivefold reduction, while 1 μg ml−1 had only a twofold effect), suggesting that larger amounts of HilD can blunt repression by propionate. Importantly, however, these results demonstrate that propionate does not function through the regulation of hilD transcription. To next test the effect of propionate on the translation of hilD, we constructed a chromosomally encoded, single-copy translational hilD’–‘lac fusion. We found that propionic acid did not reduce expression of this fusion, but in fact increased it slightly (Fig. 8C), demonstrating that propionate exhibited no repression of either transcription or translation of hilD. To ensure that the construct used here was adequate to detect such control, we also tested hilD expression in a double mutant of csrB and csrC. These mutations allow increased activity of the regulatory protein CsrA, which has been shown to bind to hilD message and repress translation (Martinez et al., 2011). As seen in Fig. 8C, hilD expression was reduced significantly in this mutant, demonstrating that translational control could be detected and therefore that propionate failed to affect hilD transcription or translation, and suggesting post-translational control of HilD as the means of invasion regulation by propionate.

Figure 8.

Propionate affects hilD through neither transcription nor translation.

A. Propionate regulates hilA through hilD. Strains carrying a single-copy chromosomal transcriptional hilAlac fusion and rtsA under the control of a tetracycline-inducible tetRA promoter were grown with or without propionic acid and with the concentrations of tetracycline shown. β-Galactosidase assays were used to assess hilA expression.

B. Propionate does not function through the regulation of hilD transcription. Strains carrying a single-copy chromosomal transcriptional hilAlac fusion and hilD under the control of a tetracycline-inducible tetRA promoter were grown with or without propionic acid and with the concentrations of tetracycline shown. β-Galactosidase assays were used to assess hilA expression.

C. Propionate does not affect the translation of hilD. Strains carrying a single-copy chromosomal translational hilD’–‘lac fusion and of the genotype shown were grown with 10 mM propionic acid or without additive and were assessed for β-galactosidase production. For all panels, error bars indicate standard deviation. An asterisk (*) indicates a statistically significant difference for a given strain with the addition of propionate as compared to no additive at P < 0.05. Two asterisks (**) indicates a statistically significant difference with mutation of csrBC at P < 0.05.

To determine how such post-translational control could be achieved, we examined whether the stability of HilD was affected by growth in the presence of propionic acid. We again used here a chromosomal copy of hilD under the control of the tetRA promoter, but one that carried a 3XFLAG tag (Chubiz et al., 2010). Cultures grown with or without propionic acid and induced with 0.8 μg ml−1 tetracycline were treated with the antibiotics rifampin, streptomycin and spectinomycin to block transcription and translation, and sampled at time points thereafter. Western blot analysis for HilD using an anti-FLAG antibody showed HilD degradation over time, with an apparent half-life of 150 min for protein obtained from bacteria grown without propionic acid (Fig. 9A). For bacteria grown with propionic acid, the amount of HilD was initially lower (∼60% of that from the untreated culture), as would be expected. Additionally, however, HilD half-life was reduced to less than one-third of that found in the untreated culture, to 42 min, due to propionic acid. Multiple repetitions of this experiment showed a reproducible reduction in HilD half-life of approximately threefold. Remarkably, however, β-galactosidase assays on the same cultures that were used for protein analysis at the initial time point showed a reduction of hilA expression of 10-fold due to growth in propionic acid (Fig. 9B), demonstrating the control of functional HilD by propionic acid disproportionate to the reduction in protein level. To investigate the importance of HilD level to repression by propionate, we next tested the effect of a lon mutation on HilD. Lon protease is known to negatively control HilD (Boddicker and Jones, 2004; Chubiz et al., 2010), and we observed an expected increase in HilD in the lon null mutant, particularly at later time points (Fig. 9A). The inclusion of propionic acid in the growth medium, however, did not reduce HilD half-life in this mutant as it did in the wild type strain. Importantly, propionic acid continued to severely repress hilA expression in this mutant, by 10-fold, identical to the wild type (Fig. 9B), and consistent with its effect on sipC (Fig. 5C). Thus, although the amount of HilD in the lon mutant grown with propionate exceeded that of the wild type grown without propionate, expression of the HilD-controlled gene hilA remained significantly reduced. These results, taken together, demonstrate that, although propionate reduces HilD stability, the amount of HilD does not itself dictate the repressive effect of this fatty acid, suggesting instead that propionate functions through a post-translational reduction of HilD activity, with consequent degradation.

Figure 9.

Propionate controls SPI1 through the post-translational regulation of HilD.

A. Propionate reduces the stability of HilD. A wild type or Δlon mutant strain carrying the hilAlac transcriptional fusion and a chromosomal HilD-3XFLAG construct under the control of the tetRA promoter was grown in the presence of 0.8 μg ml−1 tetracycline, with or without propionic acid. Transcription and translation were halted at time zero by the addition of an antibiotic cocktail, total protein was extracted, and HilD was quantified by Western blotting using an anti-FLAG antibody. Densitometric analysis of the blot was performed using ImageJ software, with all samples normalized to the wild type strain grown without propionic acid at time zero. ND, not determined.

B. Propionate exhibits regulation of hilA through the post-translational regulation of HilD. β-Galactosidase assays were used to assess hilA expression at time zero using the assay conditions described above.

Discussion

Propionate is a fatty acid produced as a metabolic by-product of bacterial fermentation and is thus found in high concentration within the intestinal lumen of humans and other animals. It has been previously demonstrated that this short-chain fatty acid can repress invasion (Durant et al., 1999; 2000; Lawhon et al., 2002; Van Immerseel et al., 2004b; 2006). The work presented here demonstrates that this repression occurs as a result of the post-translational control of the central SPI1 regulator HilD and requires the high-energy product of propionate metabolism, propionyl-CoA. Our studies show that propionate functions as a cytoplasmic signal to repress invasion. Although this and other fatty acids have been proposed to act as extracellular signals of the BarA/SirA two-component regulator of invasion (Chavez et al., 2010), we found this regulatory pathway to be dispensable to the effects of propionate.

Propionate and other short-chain fatty acids have long been known to affect the physiology and behaviour of bacteria. Propionate, along with acetate and butyrate, act as potent repellents of bacterial chemotaxis (Tsang et al., 1973). Short-chain fatty acids, being weak acids, can also enter bacteria and concentrate within the cytoplasm under the conditions of mildly acidic pH used in this study and present in the mammalian intestine. Cytoplasmic weak acids then reduce the internal pH, affecting transmembrane potential and reducing proton motive force (Repaske and Adler, 1981). In light of these generic effects, one might expect propionate to elicit global changes in Salmonella gene expression. The effects of propionate, however, showed a surprising specificity for the genes of SPI1 and those controlled by regulators within this island. A majority of the most highly repressed genes were those involved in invasion, while regulons known to be controlled in concert with invasion, such as the flagellar regulon, were not significantly affected (not shown). In addition, the failure of acetate to induce similar expression changes demonstrates that propionate affects invasion genes by a mechanism independent of the generic effects that might be caused by weak acid accumulation. As the experimental conditions chosen here were designed to approximate those of the mammalian intestinal tract in their propionate concentration and pH, these results suggest that under physiologically relevant conditions the primary effect of intestinal propionate on Salmonella is the repression of invasion.

The results of genetic studies conducted here strongly suggest that propionyl-CoA is the metabolite essential to the repression of invasion by propionate. Deletion of the genes necessary for production of propionyl-CoA from both endogenous and exogenous sources demonstrated a significant restoration in invasion gene expression in the presence of propionic acid. There was, however, not full restoration of invasion gene expression, suggesting that there may be other, uncharacterized routes of propionyl-CoA metabolism, thus preventing complete effects by the mutants tested. If further metabolism of propionyl-CoA is required for its repressive effect, it is not accomplished by the known route through the 2-methyl citrate cycle, as blocking this pathway failed to prevent repression of invasion gene expression.

It has previously been shown that butyrate, another intestinal short-chain fatty acid, similarly represses SPI1 genes (Durant et al., 1999; 2000; Lawhon et al., 2002; Gantois et al., 2006). Both of these fatty acids exist in high concentrations in the large intestine of mammals, and thus likely play important roles in the modulation of Salmonella virulence in the animal host. It is, however, apparent that these two closely related compounds (differing by only a single carbon atom) exert their effects by entirely different routes. It has been previously shown that butyrate affects multiple SPI1 promoters, with its control not limited to that of HilD (Golubeva et al., 2012), and data presented here show that the metabolic pathways required for the effects of propionate are not required for those of butyrate (Fig. 4A). Work in our laboratory also shows that butyrate does not exhibit the post-translational effects on HilD seen with propionate (C. Hung and C. Altier, unpubl. results). The specific means by which butyrate has its effects on invasion therefore remain to be discovered.

This work demonstrates that the repression of invasion genes by propionate functions specifically through HilD. A second, related SPI1 transcriptional regulator, HilC, was uninvolved in this repression, although hilC has been demonstrated to have only mild effects on Salmonella invasion (Schechter et al., 1999). By contrast, HilD occupies a position at the apex of the regulatory cascade within SPI1 (Fig. 1), and thus small changes in its concentration or function are likely to exert large effects on its target effector genes. Although much is known about the control of HilD itself, and a number of regulators have been identified, none was shown to be important for the effects of propionate. Significantly, control of HilD by propionate manifests itself post-translationally: this fatty acid demonstrated regulation through HilD even when hilD was expressed from exogenous promoters, and a translational reporter fusion to hilD failed to be repressed. Additionally, bacteria grown in the presence of propionate demonstrated a significantly reduced HilD half-life. Thus, protein stability may play some role in this control of HilD. Evidence presented here, however, demonstrates that the concentration of HilD present cannot be the defining factor in propionate control. The effects of propionate on SPI1 effectors downstream from HilD were disproportionate to that of HilD stability, with, for example, a 40% reduction in HilD leading to a 10-fold reduction in the expression of hilA (Fig. 9). Although it is common for the effects of bacterial regulators to be magnified within a regulatory cascade, such stringent control of hilA, a gene regulated directly by HilD, suggests an additional mechanism of control. Analysis of the Lon protease mutant further demonstrated this point: treatment of the Δlon mutant with propionate, in which HilD concentration remained similar to that of the untreated wild type, continued to produce profound repression of SPI1 genes. These cumulative results therefore strongly suggest that propionate functions to alter HilD activity, with changes in protein stability being only a sequela of its action.

One plausible means by which post-translational control of HilD might be achieved through the metabolic product propionyl-CoA is through protein modification and inactivation as the result of acylation. Propionyl-CoA could function by the inactivation of a protein or proteins required for invasion through the addition of a propionyl moiety donated by propionyl-CoA. Propionyl-CoA has previously been reported to inactivate PrpE, the propionyl-CoA synthetase, via N-Lysine propionylation (Garrity et al., 2007). For this to occur, protein acetyltransferase, encoded by Pat, transfers a propionyl group from propionyl-CoA to a lysine residue on PrpE, thus inactivating the protein (Starai et al., 2002; Garrity et al., 2007). Other examples exist in bacteria of N-lysine acylation through acyl-CoA intermediates as a means to alter protein structure and function. In Salmonella, Acs can be reversibly acetylated using acetyl-CoA as the acyl group donor, also through Pat (Starai et al., 2002). CheY of E. coli is also acetylated, and can additionally be auto-acetylated in vitro, in the absence of any catalysing enzyme (Barak et al., 1992; 2006; Barak and Eisenbach, 2001; Liarzi et al., 2010). If such a mechanism operates to control invasion genes, HilD remains the most likely direct target of regulation. Our genetic results demonstrate that the effects of propionate centre on HilD, and that they are post-translational, with an alteration of protein stability, as might be expected in the case of protein inactivation secondary to chemical modification. In addition, work in our laboratory suggests that a number of lysine residues of HilD are acetylated in vivo (C. Hung and C. Altier, unpubl. results), maintaining the possibility of protein propionylation as well. Work to further characterize this potential mechanism of control is ongoing.

The invasion of the intestinal epithelium in an animal host is clearly essential for productive infection by Salmonella. The fact then that propionate, a common constituent of that organ, exhibits such pronounced repression of invasion may at first seem incongruous. It is likely, however, that Salmonella uses this fatty acid, as well as others, as an environmental cue to differentiate regions of the intestinal tract. Although propionate can be present in the small intestine, its concentration is much higher in the colon and caecum, regions in which Salmonella invasion is repressed in the presence of the resident microbiota (Hapfelmeier and Hardt, 2005; Stecher et al., 2005). In humans, for example, the propionate concentration of the ileum is reported to be 1.5 mM, while within the colon its concentration ranges from 14 to 27 mM, varying by region (Cummings et al., 1987). Thus, as Salmonella passes into the large intestine, propionate may be one important signal to define for this pathogen that the possibility of productive infection has passed, and thus allow Salmonella to shift its energies to those required for survival within the intestinal lumen and passage to new hosts.

Experimental procedures

Bacterial strains, plasmids and growth conditions

Bacterial strains and plasmids used in this study are shown in Table S2. All strains are isogenic to Salmonella enterica serovar Typhimurium strain ATCC 14028s. All gene deletions were created using the previously described one-step inactivation method (Datsenko and Wanner, 2000). Briefly, PCR products were generated from the chloramphenicol or kanamycin resistance genes of pKD3 or pKD4, respectively, using primers carrying at their 5′ ends 40 bp of homology to the regions flanking the start and stop codons of the gene to be deleted. A Salmonella strain carrying pKD46, containing the λ Red recombinase for allelic exchange, was transformed with the resultant PCR products. All deletion mutants were checked for the loss of genes by PCR. Bacteriophage P22 transduction was used to transfer marked deletions and to create multiple mutations in strains (Sternberg and Maurer, 1991). To create unmarked deletions, the FLP recombinase was used to remove resistance markers (Datsenko and Wanner, 2000). The hilD’–‘lacZ chromosomal translational fusion was constructed using FLP recombinase recognition target-mediated integration as previously described (Ellermeier et al., 2002), and carried 261 bp of the hilD ORF. pBA427, carrying the hilD::luxCDABE fusion, was constructed by PCR amplifying hilD, subcloning into pCR-Blunt II-TOPO (Invitrogen), and then cloning the gene on an EcoRI fragment into pSB401 (Winson et al., 1998). The plasmid fusion of sopB to luxCDABE was described previously (Teplitski et al., 2003). Unless otherwise noted, for assays using media of an acidic pH, cultures were grown at 37°C in LB with 100 mM 3-(N-morpholino)propanesulfonic acid (MOPS) pH 6.7 containing either no additive, 10 mM propionic acid, or 10 mM butyric acid. For experiments where a pH of 8.0 was used, 100 mM 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES) pH 8.0 was used in place of MOPS. Anaerobic cultures were incubated standing overnight at 35°C in an anaerobic chamber.

β-Galactosidase assays

Cultures were grown overnight standing with either no additive, propionic acid, or butyric acid. All cultures were grown at least in triplicate, with all assays performed at least twice, and β-galactosidase activity was measured as previously described (Miller, 1992).

HEp-2 cell invasion assays

Cultures were grown overnight standing without additive or with propionic acid. The invasion assay was performed as previously described (Altier et al., 2000b), except that upon infection plates were centrifuged at 100 g. Quadruplicates were tested for each strain under each condition.

Reverse transcription real-time PCR

The wild type Salmonella strain was grown overnight with aeration in N-minimal media with 0.2% glucose to repress SPI1 gene expression. Aliquots were then subcultured into LB broth with 100 mM MOPS, pH 6.7 with either no additive or with propionic acid. Cultures were grown at 37°C with slow shaking (60 r.p.m.) for 4.5 h to reach an optical density at 600 nm (OD600) of 0.37–0.45. Three independent cultures were used for each condition. Total RNA was extracted and cDNA synthesis was performed as previously described (Huang et al., 2008). cDNA samples were diluted 2000-fold for detection of 16S rRNA as the control, and 20-fold for detection of all other gene products. cDNA was used as template for real-time PCR using B-R Syber green reagent (Quanta Biosciences) with cycling once at 95°C for 3 min followed by 40 cycles at 95°C for 20 s and 58°C for 1 min. Individual samples were each tested in triplicate. Primers for hilD and hilC expression were used as previously described (Huang et al., 2008). Other primers used were GGGAGTATATTACGGCATCAG and TTCATGAGTCTCTTCCATAGTG for rtsA, GTGCCAGCMGCCGCGGTAA and GACTACCAGGGTATCTAAT for 16S rRNA. The relative expression of invasion genes was normalized to that of 16S rRNA using iQ5 software (Bio-Rad).

Secreted protein isolation and analysis

The wild type Salmonella strain was grown without additive or with propionic acid at 37°C with shaking at 60 r.p.m. for 16 h. Proteins secreted into the culture supernatant were prepared and analysed as previously described (Altier et al., 2000b), and signal intensity was quantified using ImageJ software (Rasband, 1997–2012).

Microarrays

Experiments were performed as previously described (Lawhon et al., 2003; Frye et al., 2006). Briefly, bacteria were grown prior to RNA extraction in buffered LB medium (control) and either 15 mM propionic acid or acetic acid in buffered LB medium (experiments). cDNA probes from cells grown in LB and propionic acid or LB and acetic acid were hybridized to three arrays along with differentially labelled control probes from cells grown in LB. Dyes were switched and hybridizations were repeated to three additional arrays. The expression ratio of each gene was then calculated as the median of the six ratios from the six hybridizations. RNA measurements were analysed by calculating ratios and standard deviations between RNA from the two conditions. Genes with signals less than two standard deviations above background in both conditions were considered not detected and were removed prior to analysis.

Luciferase assays

Strains were grown overnight and then diluted 100-fold in the same medium with appropriate additives. Samples of 150 μl were inoculated into 96-well plates, and luminescence and OD600 were read every 20 min for 15 h using a Synergy 2 luminescence microplate reader (BioTek). Samples were tested in replicates of six or more.

HilD protein stability assays

A Western blot assay for detection of HilD-3XFLAG was performed as previously reported (Chubiz et al., 2010) with modification. Bacterial strains were grown overnight with aeration. One millilitre of each overnight culture was subcultured into 100 ml of the same media containing 0.8 μg ml−1 tetracycline with or without propionic acid and grown with aeration. After 2.5 h, 1 ml of culture was taken from each sample to assess the expression of hilAlacZ fusion present in this strain using a β-galactosidase assay, and the volume of each remaining culture was adjusted to reach a value of OD600 equal to 0.9–0.92 to ensure equivalent bacterial numbers. To prevent transcription and translation, an antibiotic cocktail containing rifampin, streptomycin and spectinomycin (final concentrations of 100 μg ml−1, 200 μg ml−1 and 50 μg ml−1 respectively) was added, and cultures were further incubated at 37°C. Every 20 min, 300 μl of each culture was removed and mixed with 100 μl of 4× SDS-PAGE sample buffer and immediately boiled for 7 min. Proteins from each sample were separated by 12.5% SDS-PAGE and transferred onto PVDF membrane. The membrane was blocked overnight with TBST buffer (50 mM Tris pH 7.6, 150 mM NaCl, 0.5% Tween 20) containing 5% skim milk, and then hybridized with monoclonal anti-FLAG M2 primary antibody (Sigma-Aldrich) and horseradish peroxidase-conjugated sheep anti-mouse IgG secondary antibody (GE Healthcare Life Sciences). HilD-3XFLAG was detected using Western Lightning enhanced chemiluminescence substrate (PerkinElmer), following the manufacturer's instructions. Signal intensity was quantified using ImageJ software (Rasband, 1997–2012).

Statistical analysis

Results from β-galactosidase assays, invasion assays and reverse transcription real-time PCR were analysed using a one-way analysis of variance to determine if the mean of at least one strain or condition differed from any of the others. The Tukey–Kramer Honestly Significant Difference multiple comparison test was then used to determine which means were statistically different. A P-value < 0.05 was considered significant. Statistical analysis was performed using Jmp 8.0 and 9.0 software (SAS).

Acknowledgements

We gratefully acknowledge the assistance of Jingwen Zhang in developing the protein stability assays. This project was supported by the National Research Initiative of the USDA Cooperative State Research, Education and Extension Service, award number 2005-35201-16270. MM was supported in part by NIH grants AI039557, AI052237, AI073971, AI075093, AI077645, AI083646, USDA grants 2009-03579-30127 and 2011-67017-30127, the Binational Agricultural Research and Development Fund, and CDMRP BCRP W81XWH-08-1-0720.

Ancillary