Fur regulates acid resistance in Shigella flexneri via RyhB and ydeP

Authors

  • Amanda G. Oglesby,

    1. Section of Molecular Genetics and Microbiology, The University of Texas at Austin, Austin, TX 78712, USA.
    Search for more papers by this author
  • Erin R. Murphy,

    1. Section of Molecular Genetics and Microbiology, The University of Texas at Austin, Austin, TX 78712, USA.
    Search for more papers by this author
  • Vishwanath R. Iyer,

    1. Section of Molecular Genetics and Microbiology, The University of Texas at Austin, Austin, TX 78712, USA.
    Search for more papers by this author
  • Shelley M. Payne

    Corresponding author
    1. Section of Molecular Genetics and Microbiology, The University of Texas at Austin, Austin, TX 78712, USA.
      E-mail payne@mail.utexas.edu; Tel. (+1) 512 471 9258; Fax (+1) 512 471 7088.
    Search for more papers by this author

E-mail payne@mail.utexas.edu; Tel. (+1) 512 471 9258; Fax (+1) 512 471 7088.

Summary

Shigella flexneri requires iron for survival, and the genes for iron uptake and homeostasis are regulated by the Fur protein. Microarrays were used to identify genes regulated by Fur and to study the physiological effects of iron availability in S. flexneri. These assays showed that the expression of genes involved in iron acquisition and acid response was induced by low-iron availability and by inactivation of fur. A fur null mutant was acid sensitive in media at pH 2.5, and acid sensitivity was also observed in the wild-type strain grown under iron-limiting conditions. Acid resistance of the fur mutant in minimal medium was restored by addition of glutamate during acid challenge, indicating that the glutamate-dependent acid resistance system was not defective. Inactivation of ryhB, a small regulatory RNA whose expression is repressed by Fur, restored acid resistance in the fur mutant, while overexpressing ryhB increased acid sensitivity in the wild-type strain. RyhB-regulated genes were identified by microarray analysis. The expression of one of the RyhB-repressed genes, ydeP, which encodes a putative oxidoreductase, suppressed acid sensitivity in the fur mutant. Furthermore, an S. flexneri ydeP mutant was defective for both glutamate-independent and glutamate-dependent acid resistance. The repression of ydeP by RyhB may be indirect, as real time polymerase chain reaction (PCR) experiments indicated that RyhB negatively regulates evgA, which encodes an activator of ydeP. These results demonstrate that the acid sensitivity defect of the S. flexneri fur mutant is due to repression of ydeP by RyhB, most likely via repression of evgA.

Introduction

Shigella flexneri, a causative agent of bacillary dysentery, initiates disease by invading colonic epithelial cells, multiplying intracellularly, and spreading to adjacent epithelial cells (Jennison and Verma, 2004). During infection, S. flexneri encounters different environmental conditions and must be able to respond efficiently to each of these in order to survive and cause disease. In particular, S. flexneri can survive the acidic conditions it encounters in the stomach, an ability that may contribute to its low infectious dose (10–500 organisms) (DuPont et al., 1989). While S. flexneri will not grow at a pH below 4.8 (Lin et al., 1995), this pathogen can survive for several hours at pH 2.5, a property termed acid resistance (Gorden and Small, 1993).

S. flexneri possesses two acid resistance systems. The first is dependent upon the presence of glutamate in the medium during acid challenge (Lin et al., 1995). This system requires the expression of glutamate decarboxylase, as well as an antiporter which imports glutamate and exports its decarboxylated product, gamma-aminobutyrate (Hersh et al., 1996; Waterman and Small, 1996; De Biase et al., 1999). The decarboxylation of glutamate is thought to protect bacteria by increasing the internal pH through the consumption of intracellular protons. In S. flexneri, gadA and gadB encode glutamate decarboxylase isozymes, and the glutamate/GABA antiporter is encoded by gadC (Waterman and Small, 1996). Two similar systems that use arginine or lysine instead of glutamate are also present in Escherichia coli (Park et al., 1996), but homologous genes are not present in the S. flexneri genome (Jin et al., 2002; Wei et al., 2003).

The regulation of glutamate-dependent acid resistance has been described in E. coli. This system is induced in logarithmically growing cells exposed to low pH, or in cells grown to stationary phase. The EvgA response regulator activates glutamate-dependent acid resistance during exponential growth at low pH (Masuda and Church, 2002; 2003; Nishino et al., 2003; Ma et al., 2004), whereas RpoS induces expression of this system during stationary phase (Castanie-Cornet et al., 1999; De Biase et al., 1999).

The second acid resistance system is glutamate-independent and is less well characterized than the glutamate-dependent acid resistance system. This glutamate-independent acid-resistance system protects cells in the absence of glutamate at pH 2.5 and is induced by RpoS in complex medium during stationary phase (Lin et al., 1995). Overexpression of evgA also induces this system in exponential phase cells (Masuda and Church, 2002). This system has also been called the oxidative acid resistance system, as its function is repressed by fermentative growth in media containing glucose and its expression requires the cAMP receptor protein (CRP) (Lin et al., 1995; Castanie-Cornet et al., 1999).

S. flexneri encounters iron-restricted, as well as acidic, environments in the host. Most iron in the host is sequestered in proteins such as lactoferrin, transferrin, and haemoglobin and is not readily available to bacterial pathogens. S. flexneri expresses several iron acquisition systems in iron-restricted environments in vivo and in vitro (Payne and Mey, 2004). Iron restriction limits bacterial growth, but high intracellular concentrations of iron can be harmful due to its ability to accelerate the formation of oxygen radicals in aerobically respiring bacteria. The acquisition of iron is therefore tightly regulated in response to the intracellular iron concentration. Iron homeostasis is largely achieved through the action of Fur (ferric uptake repressor), a 17 kDa iron-binding repressor protein (Hantke, 2001). Under iron-replete conditions, the Fur protein becomes ferrated and binds to a 19 bp sequence called the Fur box in the promoters of iron acquisition genes, thereby repressing their expression (Bagg and Neilands, 1987; Escolar et al., 1999; Baichoo and Helmann, 2002). In iron-depleted conditions, the apo-Fur protein is inactive, and transcription of iron acquisition genes proceeds.

In E. coli, Fur also represses the expression of ryhB, which encodes a small regulatory RNA that blocks gene expression by binding to and destabilizing target mRNAs (Masse and Gottesman, 2002; Masse et al., 2003). Degradation of the RyhB-mRNA complex is RNaseE-dependent (Masse et al., 2003). RyhB stability and mRNA binding are promoted by the small RNA chaperone protein Hfq (Masse et al., 2003). In E. coli, RyhB contributes to iron homeostasis by repressing the production of iron-containing proteins during growth in iron-depleted conditions. These RyhB-regulated genes include sdhCDAB (succinate dehydrogenase), sodB (Fe-superoxide dismutase), acnA (aconitase), fumA (fumarase) and ftn (ferritin) (Masse and Gottesman, 2002). The sequence of the S. flexneri ryhB gene is identical to that of E. coli K-12 (Wei et al., 2003), and the encoded small RNA likely serves similar regulatory functions.

In this study, we analysed the physiological effects of iron regulation by identifying targets of Fur regulation in S. flexneri using microarrays. As expected, the iron acquisition genes were induced during growth in iron-limiting conditions and in a fur mutant. Other genes not known to be regulated by iron were also identified in this analysis, several of which are members of the acid regulon in E. coli (Tucker et al., 2002). This report describes the role of Fur and RyhB in the regulation of acid resistance in S. flexneri.

Results

Acid response genes are induced in response to iron depletion and fur inactivation in S. flexneri

To study the physiological effects of decreased iron availability in S. flexneri, microarrays were used to identify iron- and Fur-regulated genes. We compared global gene expression of wild-type S. flexneri grown in high- and low-iron conditions, as well as the gene expression profiles of the wild type and fur mutant grown in high iron. Selected genes whose expression was induced in low iron and in the fur mutant are shown in Table 1.

Table 1.  Iron acquisition and stress response genes induced by iron depletion or fur inactivation during exponential growth.a
Gene nameUIDDescription of gene productFold inductionb in:
low iron (+ EDDA)fur mutant
Iron metabolism genes
 iutAIUTAAerobactin outer membrane receptord11.6 ± 3.020.6 ± 11.5
 iucAIUCAAerobactin synthesisd3.3 ± 0.94.6 ± 2.1
 iucBIUCBAerobactin synthesisd9.4 ± 1.924.3 ± 8.2
 iucCIUCCAerobactin synthesisd8.3 ± 2.428.6 ± 22.4
 fhuEB1102Coprogen outer membrane receptore,f5.9 ± 2.97.2 ± 5.4
 fhuFB4367Hydroxamate reductasef,g12.2 ± 3.313.0 ± 7.0
 exbBB3006TonB accessory proteinf,h6.0 ± 1.16.5 ± 3.9
 feoAB3408Ferrous iron uptakef,i2.4 ± 0.65.1 ± 3.8
 sitASITAFerrous iron uptakej10.9 ± 2.34.0 ± 3.3
 sitBSITBFerrous iron uptakej10.0 ± 3.33.7 ± 3.8
 sitCSITCFerrous iron uptakej5.8 ± 1.82.3 ± 1.6
 sitDSITDFerrous iron uptakej8.7 ± 1.72.3 ± 0.6
 ryhBN/AFur-repressed small RNAND68c
Acid response genes
 gadCB1492Glutamate/GABA antiporterk9.7 ± 10.07.0 ± 8.8
 gadBB1493Glutamate decarboxylasek10.3 ± 10.86.1 ± 7.3
 hdeAB3510Putative acid resistance chaperonel8.3 ± 8.86.3 ± 8.5
 gadEB3512Regulator of acid response genesm6.2 ± 6.13.4 ± 3.6
Oxidative stress genes
 dpsB0812DNA protection during starvationn4.4 ± 4.23.4 ± 3.9
 sufBB1683Fe-S cluster assemblyo2.7 ± 1.91.9 ± 0.8
 sufCB1682Fe-S cluster assemblyo2.2 ± 1.61.4 ± 0.8
 sufDB1681Fe-S cluster assemblyo2.4 ± 1.71.7 ± 0.8
 sufEB1679Fe-S cluster assemblyo2.9 ± 1.51.9 ± 0.9
 sufSB1680Fe-S cluster assemblyo3.0 ± 2.31.7 ± 0.8

As expected, known iron acquisition genes were induced in cells grown in the presence of the iron chelator EDDA compared with those grown in iron-replete conditions, and these genes were expressed at a higher level in the fur mutant compared with the wild-type strain. These iron acquisition genes have previously been shown to be repressed by iron via the Fur protein. Among these genes are the iucA-D and iutA genes required for aerobactin synthesis and uptake (Payne et al., 1983; de Lorenzo et al., 1987), the sit and feo genes involved in ferrous iron uptake (McHugh et al., 2003; Runyen-Janecky et al., 2003) and fhuE, fhuF and exbB (McHugh et al., 2003).

Several genes that are induced in E. coli by growth in acidic conditions (Tucker et al., 2002) also responded to iron limitation and fur inactivation in S. flexneri (Table 1). gadB and gadC, encoding glutamate decarboxylase and the glutamate/GABA antiporter respectively, were both induced upon iron depletion and fur inactivation, as was hdeA, encoding a putative periplasmic acid resistance chaperone. A gene encoding a central activator of acid resistance, gadE (yhiE ) (Ma et al., 2003) was also more highly expressed during growth in the presence of the iron chelator (low iron) than in high iron and in the fur mutant as compared with wild type. Although GadE induces expression gadA, as well as gadB and C in E. coli (Castanie-Cornet and Foster, 2001), gadA induction in response to low iron or loss of Fur was not observed in these experiments.

There were differences in the extent of induction of the acid response genes when comparing individual microarray experiments, and this variation was greater than that observed for the iron acquisition genes. In some experiments, the acid resistance genes were induced 10- to 20-fold while in other experiments, the induction of these genes was less than twofold (data not shown). In contrast, the genes encoding iron transport systems were consistently induced more than twofold in the low-iron cultures and in the fur mutant. These differences in induction of the acid resistance genes are therefore likely due to subtle variations in culture conditions unrelated to changes in iron availability. This same pattern of variation was seen with another group of iron and Fur-regulated genes, the suf and dps genes (Table 1 and Fig. 1). The suf genes are induced both by low iron and by oxidative stress (Outten et al., 2004). Thus the extent of induction of the acid resistance genes by low iron or in the fur mutant may depend on oxidative stress or other environmental conditions in addition to loss of iron or loss of Fur.

Figure 1.

Expression of Shigella genes in response to iron limitation and loss of Fur. Expression data from the microarray experiments described in Table 1 were used for hierarchical clustering as described in Experimental procedures and displayed in an expression map. Each column represents a single experiment and each row a single gene. The brightness of each square indicates the ratio of expression in either low-iron compared with high-iron conditions (–/+Fe) or in the fur mutant compared with wild type (–/+Fur). Red indicates induction while green indicates repression of that gene. Grey indicates that data for that gene were unavailable for analysis.

The fur mutant does not acidify the culture medium more than wild type

A possible explanation for the observed induction of acid response genes in the fur mutant is acidification of the medium caused by altered expression of metabolic pathways. To test this, the pH of the growth medium was determined for both the wild type and the fur mutant grown under the same conditions as used for the microarray experiments (Fig. 2). The pH of the culture medium for both strains remained above 6.5 throughout the growth period. As the induction of acid response genes during logarithmic growth requires an external pH of 5.5 or lower (Castanie-Cornet and Foster, 2001), it appears that induction of the acid response genes in the fur mutant was not a result of acidification of the culture medium.

Figure 2.

Change in culture pH of wild type and fur cultures. Wild type (SA101-diamonds) and fur (SA211-squares) were diluted 1:100 into L broth from overnight cultures in L broth, and were grown in batch culture at 37°C. pH and growth readings were taken every 30 min until strains reached late-logarithmic growth phase. Displayed are results of a representative experiment.

The fur mutant is defective for glutamate-independent acid resistance

The induction of the acid response genes in cells grown in low iron and in the fur mutant suggested that the cells were experiencing acid stress and that acid resistance could be affected in these cells. Therefore, we determined the acid resistance phenotype of the fur mutant. Following incubation in complex medium at pH 2.5, less than 1% of the fur mutant cells survived, while greater than 10% of wild-type cells were viable (data not shown). Because the concentration of glutamate in the complex medium (Luria broth [L broth]) was unknown, the assays were repeated in minimal medium with or without glutamate supplementation to determine if the acid resistance defect was specific to either the glutamate-dependent or glutamate-independent mechanism. As was observed in complex medium, only 1% of fur mutant cells survived after 2 h incubation in M9 minimal medium at pH 2.5 (Fig. 3A, open bars). This phenotype was suppressed by the addition of glutamate to the medium, indicating that the fur mutant is defective for glutamate-independent acid resistance (Fig. 3A, closed bars). As the induction of acid response genes was also observed in wild type S. flexneri grown in low iron, the effect of iron limitation on acid resistance in the wild type was tested. Wild-type S. flexneri grown prior to acid shock in medium containing the iron chelator EDDA to reduce iron availability displayed the same acid sensitivity phenotype as the fur mutant in both complex (data not shown) and minimal media (Fig. 3A). The acid-sensitive phenotype of the fur mutant was complemented by expression of fur from a plasmid (Fig. 3B). Overall, these results indicate that both iron and Fur are required for glutamate-independent acid resistance in S. flexneri.

Figure 3.

Iron and Fur are required for glutamate-independent acid resistance.
A. Wild type (SA100) and fur (SA1301), grown with and without EDDA as indicated, were incubated in M9, pH 2.5 without (open bars) or with (closed bars) 0.012% glutamate for 2 h.
B. Wild type (SM100) or the fur mutant (SM1301) carrying pACYC184 (Vector) or pFS2 (Fur+), as indicated, were incubated for 2 h in M9, pH 2.5 without (open bars) or with (closed bars) glutamate supplementation.
Per cent survival was determined as described in the Experimental procedures. Data shown are averages of three independent experiments. Error bars indicate one standard deviation.

Acid sensitivity in the fur mutant is due to induction of ryhB

Although the expression of the acid resistance genes was regulated by iron and Fur, we did not observe an apparent Fur box in the promoters of the acid-response genes or their identified regulators. This suggested that the effect of Fur on the acid-response genes was indirect. One mechanism for indirect regulation by Fur is via RyhB, a Fur-repressed small RNA that regulates a number of genes in E. coli (Masse and Gottesman, 2002). Real time polymerase chain reaction (PCR) confirmed that ryhB was induced under the same conditions that caused induction of the acid-resistance genes (Table 1) and in the growth conditions used for acid sensitivity assays (Fig. 4). To determine whether RyhB was influencing acid resistance, a ryhB deletion mutant and a fur ryhB double mutant were constructed. While the fur mutant was sensitive to acid challenge, the ryhB mutation suppressed acid sensitivity in the fur mutant, with nearly 100% of the fur ryhB cells surviving acid challenge in the absence of glutamate (Fig. 4A). The ryhB mutation alone had no effect on acid resistance (Fig. 4A).

Figure 4.

RyhB represses glutamate-independent acid resistance.
A. Wild type (SM100), fur (SM1301), fur ryhB (SM1302) and ryhB (SM100; ryhB::cam) were incubated in M9, pH 2.5 without (open bars) or with (closed bars) glutamate supplementation for 2 h.
B. Wild-type S. flexneri (SA100) carrying pQE-2 (Vector) or pERM111 (RyhB+) were incubated in M9, pH 2.5 without (open bars) or with (closed bars) glutamate supplementation for 2 h in the presence of the indicated concentration of IPTG.
Per cent survival was determined as described in the Experimental procedures. Data shown are averages of three independent experiments. Error bars indicate one standard deviation. Numbers below each graph represent the relative amount of ryhB expression as determined by real time PCR.

The data suggest that RyhB represses glutamate-independent acid resistance in S. flexneri. To test this directly, wild-type S. flexneri was transformed with a plasmid carrying ryhB under the control of an IPTG-inducible promoter. This strain was resistant to acid challenge in the presence of glutamate, but exhibited acid sensitivity in the absence of glutamate, proportional to the concentration of IPTG in the growth medium (Fig. 4B). As expected, ryhB expression was proportional to the concentrations of IPTG, as measured by real time PCR (Fig. 4B). Thus, overexpression of ryhB suppressed glutamate-independent acid resistance in S. flexneri. Taken together, these results suggest that RyhB represses one or more genes required for glutamate-independent acid resistance in S. flexneri.

Repression of sodB is not responsible for acid sensitivity in the S. flexneri fur mutant

At low pH, iron is more soluble, making it more bio-available and more likely to mediate iron-mediated oxygen toxicity. In E. coli, sodB, encoding a superoxide dismutase which helps protect cells against oxidative damage, is repressed RyhB (Masse and Gottesman, 2002) and is induced during growth at low pH (Stancik et al., 2002). As sodB was repressed in the S. flexneri fur mutant compared with wild type in microarray experiments (Fig. 1), the possibility that acid sensitivity in the fur mutant was due to repression of sodB and subsequent loss of protection against oxidative stress was tested. sodB and sodA single mutants and the sodB sodA double mutant were all resistant to acid challenge in the absence of glutamate (Fig. 5A), and overexpression of sodB did not restore acid resistance to the fur mutant (Fig. 5B). These data indicate that repression of sodB did not cause acid sensitivity in the S. flexneri fur mutant.

Figure 5.

Acid sensitivity in the S. flexneri fur mutant is not due to repression of sodB.
A. Wild type (SM100), fur (SM1301), sodA (SM1303), sodB (SM1304), and sodA sodB (SM1305) were incubated in M9, pH 2.5 without (open bars) or with (closed bars) glutamate supplementation for 2 h.
B. Wild type (SM100) or the fur mutant (SM1301) carrying pGEM-T easy (Vector) or pAGO-sodB1 (SodB+) as indicated, were incubated in M9, pH 2.5 without (open bars) or with (closed bars) glutamate supplementation for 2 h. Per cent survival was determined as described in the experimental procedures. Data shown are averages of three independent experiments. Error bars indicate one standard deviation.

Acid resistance genes are repressed by RyhB in the S. flexneri fur mutant

Increased acid sensitivity when ryhB is expressed and loss of acid sensitivity in the fur ryhB double mutant suggested that RyhB represses one or more genes needed for glutamate-independent acid resistance. Candidate genes were identified by growing the cells to stationary phase, as is done for assaying acid resistance, and then comparing gene expression profiles of the fur mutant with the wild type, as well as that of the fur ryhB double mutant compared with the fur mutant. This analysis revealed several genes with the expression pattern predicted for a RyhB-repressed gene required for glutamate-independent acid resistance, i.e. they are expressed at lower levels in the fur mutant relative to wild type and de-repressed in the fur ryhB double mutant compared with the fur single mutant (Table 2). Among these candidate genes were several acid-resistance genes, including gadB and gadC (glutamate-dependent acid resistance), ydeO (regulator of glutamate-dependent acid resistance), and ydeP (putative dehydrogenase and acid-resistance gene). The repression of the gad genes did not lead to loss of glutamate-dependent acid resistance in the fur mutant (Fig. 3), possibly because these genes are only repressed approximately twofold under these conditions.

Table 2.  RyhB-dependent Fur activation of acid response genes in stationary phase.a
GeneUIDDescription of gene productFold repression in fur versus wild typebFold induction in fur ryhB versus furc
  • a

    . Strains were grown for 20 h at 30°C in MOPS-buffered L broth at pH 7. Experiments were performed twice each using oligonucleotide microarrays as described in Experimental procedures.

  • b

    . Fold repression indicates the ratio of hybridized Cy3-labelled cDNA (wild type) to Cy5-labelled cDNA (fur) to the indicated spot. Average of two experiments.

  • c

    . Fold induction indicates the ratio of hybridized Cy5-labelled cDNA (fur ryhB) to Cy5-labelled cDNA (fur) to the indicated spot. Average of two experiments.

gadCB1492Glutamate/GABA antiporter2.03.9
gadBB1493Glutamate decarboxylase2.43.0
ydeOB1499Acid resistance regulator1.62.7
ydePB1501Putative oxidoreductase involved in acid resistance3.14.1

Repression of ydeP in the S. flexneri fur mutant causes acid sensitivity

With the exception of ydeP, all of the acid resistance genes shown in Table 2 are specifically involved in glutamate-dependent acid resistance in E. coli. The role of ydeP in the glutamate-independent acid sensitivity phenotype of the fur mutant was therefore investigated. The fur mutant was transformed with a plasmid carrying ydeP. Expression of ydeP from this construct completely suppressed the acid sensitivity phenotype of the fur mutant (Fig. 6). Suppression of acid sensitivity by ydeP expression was most likely not due to the indirect effect of titrating out available RyhB, because overexpression of sodB, whose mRNA is known to bind RyhB (Vecerek et al., 2003), did not affect acid sensitivity of the fur mutant (Fig. 5B). These results suggest that glutamate-independent acid sensitivity in the S. flexneri fur mutant is due to repression of ydeP by RyhB.

Figure 6.

Expression of ydeP suppresses acid sensitivity in the fur mutant. Wild type (SA100), fur (SA1301) and ydeP (SM1306) carrying pWKS30 (Vector) or pWKS-ydeP (YdeP+) as indicated, were incubated in M9, pH 2.5 without (open bars) or with (closed bars) glutamate supplementation for 2 h. Per cent survival was determined as described in the Experimental procedures. An asterisk (*) indicates that per cent survival at 2 h was below detectable levels (< 0.001%). Data shown are averages of three independent experiments. Error bars indicate one standard deviation.

To further test the role of ydeP in glutamate-independent acid resistance, a ydeP mutant was constructed by allelic exchange. The ydeP mutant was highly sensitive to acid challenge in the absence or presence of glutamate, and acid resistance was restored to the ydeP mutant by expression of ydeP from a plasmid (Fig. 6). These results show that ydeP is required for glutamate-independent as well as glutamate-dependent acid resistance in S. flexneri. Together, these data indicate that reduced ydeP expression, as was observed in the fur mutant, caused a defect in glutamate-independent acid resistance, whereas a complete loss of ydeP expression by gene inactivation led to defects in both glutamate-dependent and glutamate-independent acid resistance.

RyhB represses evgA and ydeP

These data do not distinguish between RyhB directly repressing ydeP or acting on an upstream activator. There is no sequence homology between RyhB and ydeP (data not shown) as might be expected if RyhB were directly interacting with the ydeP mRNA. All of the RyhB-repressed genes shown in Table 2, including ydeP, are members of the EvgA acid-induced regulon in E. coli. Analysis of the evgA DNA sequence indicated that there is a region of homology between evgA and RyhB. To test the hypothesis that RyhB represses ydeP via evgA, real time PCR was performed to determine the effect of ryhB overexpression on the relative levels of ydeP and evgA mRNA (Fig. 7). sodB, a known RyhB target, was included as a control for ryhB expression in this experiment. Both evgA and ydeP were repressed upon induction of ryhB expression (Fig. 7). These data are consistent with the model that repression of ydeP in S. flexneri is via repression of evgA by RyhB (Fig. 8).

Figure 7.

RyhB represses expression of evgA and ydeP. Wild type (SA100) carrying pERM111 (RyhB+) was grown for 6 h to late exponential/early stationary phase (OD650 ≈ 1) at 37°C in L broth. IPTG was added to half of the culture to a final concentration of 200 µM, and cultures were incubated for an additional 20 min. RNA isolated from each sample was used for real time PCR as described in the Experimental procedures to determine the relative quantities of sodB, evgA and ydeP mRNA. The relative expression of each gene upon IPTG induction of ryhB is reported as the percentage of expression of the gene in the absence of RyhB induction.

Figure 8.

Model of Fur regulation of ydeP. When iron is present, the ferrated Fur protein represses transcription of ryhB. In the absence of RyhB, EvgA is produced and activates transcription of ydeP, allowing for acid resistance. In the absence of iron or the Fur protein, RyhB is produced and represses evgA. In the absence of EvgA, ydeP is not expressed and cells are sensitive to acid.

Discussion

The role of Fur in many bacteria extends beyond regulating iron homeostasis. Fur represses several virulence factors, including Shiga toxin in Shiga-toxin producing strains of E. coli and Shigella dysenteriae (Calderwood and Mekalanos, 1987). In addition, Fur is required for protection against oxidative stress in E. coli (Touati et al., 1995), a trait that is most likely due to de-repression of iron and oxidative protection proteins such as SodB via regulation of the small RNA RyhB (Masse and Gottesman, 2002). Here we show that Fur also regulates acid resistance via RyhB in S. flexneri. Based on the data presented in this study, a model of regulation of acid resistance by Fur is shown in Fig. 8. When the Fur protein was inactivated by either mutation or iron depletion, ryhB is constitutively expressed. RyhB represses the expression of evgA, which is required for ydeP activation. As ydeP is required for acid resistance in S. flexneri (Fig. 6), these data indicate that repression of evgA by RyhB, causing reduced expression of ydeP, is the cause of acid sensitivity in the fur mutant.

Several links between Fur regulation and acid survival have been identified in organisms closely related to S. flexneri. A fur mutant in avian septicemic E. coli is defective for acid resistance (Zhu et al., 2002), and, in some strains of enterohemorrhagic E. coli, Fur positively regulates production of urease, a putative acid response protein (Heimer et al., 2002). Furthermore, the Salmonella enterica serovar Typhimurium Fur protein is required for survival in acidic conditions (Foster, 1991). Measurements of the internal pH of a S. typhimurium fur mutant, which was unable to survive acid exposure, demonstrated that the internal pH of the fur mutant was considerably lower than that of wild type in acidic environments, indicating a defect in pH homeostasis (Foster and Hall, 1992). These results suggested a relationship between pH homeostasis and acid resistance. There are indications that the fur mutant in S. flexneri is also defective for pH homeostasis and, in some conditions, maintains a lower internal pH than wild type. In the fur mutant, the expression of several acid-resistance genes was higher than in the wild type (Table 1), yet no acidification of the culture medium that could cause induction of these acid-resistance genes was observed (Fig. 2). These data suggest that that iron depletion or fur inactivation causes a lowering of the internal pH and thereby induces the expression of acid-resistance genes. Repression of ydeP, which we show to be the cause of acid sensitivity in the fur mutant, is also observed in our initial microarray experiments (Fig. 1). This data suggests that ydeP may be required for general pH homeostasis in addition to acid resistance.

Although Fur is required for acid survival of both S. typhimurium and S. flexneri, the pathways regulating the expression of the acid-response genes are quite different in these two organisms (Lin et al., 1995), and the Fur proteins of each organism may be activating acid-response genes in response to different environmental signals. The role of Fur in S. typhimurium acid survival is distinct from its role in iron regulation, as a mutation in fur that rendered the strain unable to regulate iron acquisition genes in response to iron availability had no effect on acid survival (Hall and Foster, 1996). Several genes in S. typhimurium are activated by Fur by an iron-independent mechanism, including several acid shock proteins (Foster and Hall, 1991; 1992; Campoy et al., 2002). This contrasts with what was observed here for S. flexneri, where the presence of iron, in addition to the Fur protein, is required for acid-resistance (Fig. 2), and iron-independent Fur regulation of acid-response genes was not evident (Table 1). There is a possibility of overlap in the regulatory cascade downstream of Fur in both of these organisms, however, because the ryhB coding sequence and its predicted Fur binding site are conserved in S. typhimurium (Masse and Gottesman, 2002). It will be of interest to see whether or not the role of RyhB in regulating acid response is shared between S. flexneri and S. typhimurium, or if the S. typhimurium Fur protein is regulating acid shock proteins in a more direct manner.

Masuda and Church (2002) previously observed that EvgA regulated both glutamate-dependent and glutamate-independent acid resistance in E. coli. They identified a number of genes regulated by EvgA, including ydeP. Inactivation of ydeP in E. coli led to acid sensitivity in exponentially growing cells, but not in cells grown to stationary phase (Masuda and Church, 2002). Activation of the glutamate-independent acid resistance system in E. coli is not induced until stationary phase (Castanie-Cornet et al., 1999), and it is likely that the acid sensitivity of the E. coli ydeP mutant is primarily due to a defect in the glutamate-dependent acid resistance system. In S. flexneri, mutation of ydeP did cause a loss of acid resistance in stationary phase, and both glutamate-dependent and glutamate-independent acid resistance are impaired in the S. flexneri ydeP mutant. Thus there may be differences in the two species in the role of YdeP in acid resistance.

Although the S. flexneri ydeP mutant was defective in both glutamate-dependent and glutamate-independent acid resistance, repression of ydeP in the S. flexneri fur mutant caused a defect only in the glutamate-independent system. The different phenotypes observed for the S. flexneri fur mutant and the ydeP mutant may be due to the different levels of ydeP that are expressed in each of these strains. The fur mutant, which produces less YdeP than wild type, may still make enough YdeP to maintain glutamate-dependent acid resistance. However, when expression of ydeP is completely abolished by gene inactivation, a more severe defect in acid resistance is observed, rendering cells unable to survive even in the presence of glutamate. These results suggest that ydeP plays roles in both the glutamate-dependent and glutamate-independent acid-resistance systems in S. flexneri.

How does ydeP contribute to acid resistance? The ydeP gene encodes a putative oxidoreductase with identity with fdhH, which encodes the alpha chain of formate dehydrogenase H. In E. coli, formate dehydrogenase H is part of the formate-hydrogen lyase (FHL) complex that oxidizes formate to carbon dioxide and dihydrogen (Sawers, 1994). The genes encoding components of the FHL complex are induced by anaerobiosis, formate and low pH (Rossmann et al., 1991). The advantage of induction of the FHL complex by acid is that the products of FHL oxidation of formate, carbon dioxide and dihydrogen, are less acidic than formate. This suggests that one of the mechanisms for acid survival employed by enteric bacteria during growth in acidic conditions is to alter metabolic pathways so that fewer acidic products are created, thus raising the internal pH of the cell. Although the enzymatic activity of YdeP has not been characterized, the conservation of the molybdopterin-binding domains that are commonly found in oxidative enzymes leads to the hypothesis that the protein encoded by ydeP may also play a role in oxidative growth. The identification of the substrate of this putative dehydrogenase and the biochemical pathway in which it participates will be important in elucidating the role that this protein plays in acid resistance in both E. coli and S. flexneri.

Experimental procedures

Bacterial strains, plasmids and growth conditions

Bacterial strains and plasmids used in this work are listed in Table 3. E. coli strains were routinely grown in L broth or Luria agar (L agar) (Gerhardt et al., 1994). L broth was prepared using 10% tryptone, 5% yeast extract and 10% NaCl. Where indicated L broth was buffered with 100 mM 4-morpholinepropanesulfonic acid (MOPS). S. flexneri strains were routinely grown in L broth or on tryptic soy broth agar (TSBA) plus 0.01% Congo red dye at 37°C. Ethylenediamine-N′N′-bis(2-hydroxyphenyl acetic acid) (EDDA) was deferrated as previously described (Rogers, 1973) and added to L broth at a concentration of 8 or 16 µg ml−1 to chelate iron. M9 minimal media were prepared as previously described (Sambrook and Russell, 2001) and supplemented with 100 µg ml−1 nicotinic acid. Glutamate at 0.012% (w/v) was added to M9 where indicated. Antibiotics were used at the following concentrations (per millilitre): 125 µg of carbenicillin, 25 µg of kanamycin, 15 µg of chloramphenicol, 12.5 µg tetracycline and 200 µg of streptomycin.

Table 3.  Bacterial strains and plasmids.
Strain or plasmidCharacteristicsReference or source
E. coli strains
 DH5αendA1 hsdR17 supE44 thi-1 recA1 gyrA relA1 Δ(lacZYA-argF)U169 deoR [Φ80 dLacΔ(lacZ)M15]Sambrook and Russell (2001)
 MFT5fur::Tn5M. McIntosh
 SM10 λpirpirR6KTaylor et al. (1989)
 DH5αλpirpirR6KSambrook and Russell (2001)
 MM294/pRK2013Mobilizing helper strain for R6K plasmidsR. Meyer
 JI130sodA::MudPR13 (Camr)Carlioz and Touati (1986)
 JI131sodB::kanCarlioz and Touati (1986)
S. flexneri strains
 SA100S. flexneri wild-type serotype 2aPayne et al. (1983)
 SA1301SA100 fur::Tn5This work
 SA101SA100 derivative with deletion in virulence plasmidDaskaleros and Payne (1987)
 SA211SA101 fur::Tn5Schmitt and Payne (1988)
 SM100SA100 StrrS. Seliger
 SM1301SM100 fur::Tn5This work
 SM100; ryhB::camSM100 ryhB::camThis work
 SM1302SM100; ryhB::cam fur::Tn5This work
 SM1303SM100 sodA::MudPR13 (Camr)This work
 SM1304SM100 sodB::kanThis work
 SM1305SM1304 sodA::MudPR13 (Camr)This work
 SM1306SM100 ydeP::camThis work
Plasmids
 pACYC184Cloning vectorChang and Cohen (1978)
 pFS2pACYC184 carrying furSchmitt (1989)
 pCVD442N2Allelic exchange vector containing a NotI siteE. Wyckoff
 pERM104pCVD442N2 carrying ryhB::camThis work
 pCVD-ydeP::campCVD442N2 carrying ydeP::camThis work
 pQE-2IPTG-inducible overexpression vectorQiagen
 pERM111pQE-2 carrying ryhBE. Murphy
 pGEM-T easyCloning vectorQiagen
 pAGO-sodB1pGEM-T easy carrying sodBThis work
 pWKS30Cloning vectorWang and Kushner (1991)
 pWKS-ydePpWKS30 carrying ydePThis work

Construction of plasmids

To clone ydeP, PCR was carried out using the following primers: ydePclone2.for (5′-GAGGAGGATAAGTAGATGAA GAAAA-3′) and ydePclone2.rev (5′-GAGCTCGAAGAGGC ATTAATTTGATGG-3′), and the resulting product was cloned into pGEM-T easy (Promega, Madison, WI). The ydeP fragment obtained by digestion of the plasmid with PstI and SacI was ligated into pWKS30 digested with the same enzymes. To clone sodB, PCR was carried out using sodBclone.for (5′-TTTGCTACCCTATCCATACG-3′) and sodBclone.rev (5′-CGATTTATCTTGCCGAATGC-3′). The PCR product was cloned into pGEM-T easy, yielding pAGO-sodB1.

Construction of mutants

Splice overlap PCR as previously described (Sambrook and Russell, 2001) was used to create a ryhB::cam mutation in SM100. Sequence upstream and downstream of the ryhB gene was amplified, replacing the gene with a SmaI recognition site, using the following primers: ryhB-9 (5′-GCTCTAGAGTGAGAGCGGTATTCTG-3′) with ryhB-2 (5′-AAAGATCCCCGGGGATGTTGAAAGGGACAT-3′), and ryhB-3 (5′-CAACATCCCCGGGGATCTTTCGTTCTCAAT-3′) with ryhB-10 (5′-GCGAGCTCCAACGCATCATAAACACG-3′). The amplified DNA fragment was cloned into pGEM-T easy. The resulting plasmid was digested with SmaI and a chloramphenicol resistance cassette (cam) was introduced. ryhB::cam was excised by digestion with NotI and ligated into pCVD442N2 digested with the same enzyme. This plasmid, pERM104, was electroporated into E. coli SM10 λpir, then introduced into SM100 by bi-parental conjugation. Primary integrants were selected for by growth in the presence of chloramphenicol and carbenicillin. The final ryhB::chl mutation was selected for by growth in the presence of sucrose and chloramphenicol. The ryhB::cam mutation was confirmed by PCR using ryhB-7 (5′-GGACGGTGAGTTGTAC-3′) and ryhB-8 (5′-TGCCGCTTTCAATCG-3′).

The ydeP::cam mutant in SM100 was constructed by allelic exchange. A chloramphenicol cassette was excised from pMTLchl using the restriction endonuclease SmaI and was ligated into pGEM-ydeP digested with the same enzyme. ydeP::cam was then excised from pGEM-ydeP::cam using NotI and ligated into pCVD442N2 also digested with NotI. The resulting plasmid, named pCVD-ydeP::cam, was transformed into DH5α-λpir. pCVD-ydeP::cam was then moved into SM100 by tri-parental conjugation, using MM294/pRK2013 as a helper strain. Primary integrants were selected for by growth in the presence of carbenicillin and chloramphenicol. The final ydeP::cam mutation was selected for by growth in the presence of sucrose and chloramphenicol. The ydeP::cam mutation was confirmed by PCR using ydePclone2.for and ydePclone2.rev (described above).

The fur, sodA, and sodB mutations were introduced into S. flexneri by P1 transduction as previously described (Sambrook and Russell, 2001). Lysates were prepared following growth of PI phage on the E. coli fur mutant MFT5, sodA mutant JI130 and sodB mutant JI131. The MFT5/P1 lysate was used to transduce fur::Tn5 into SA100, SM100, and SM100 ryhB::cam. Insertional inactivation of the fur gene in each strain was confirmed by PCR using a primer internal to the Tn5 (5′-GCGATAACTCAAAGAGGTGGTGTC-3′) and another primer in fur (5′-GGGACTTGTGGTTTTCATT TAGGC-3′). The JI131 and JI130/P1 lysates were used to transduce the sodB::MudPR13 and sodA::kan mutations into SM100 and SM1304. Insertion of MudPR13 into sodA was confirmed by PCR using primers that flank the sodA gene: sodAclone.for (5′-TCGTCGCTTCACATCTCC-3′) and sodAclone.rev (5′-GCCTATACGCCTCATTGC-3′). Insertion of the kanamycin cassette into sodB was confirmed by PCR using primers that flank the sodB gene: sodBclone.for and sodBclone.rev (described above).

Construction of microarrays

Microarrays containing open reading frames from both E. coli and Shigella were designed by two methods. First, a PCR-based microarray was constructed by amplifying every gene in the E. coli genome (Blattner et al., 1997), as well as a number of Shigella-specific genes. Primers for the 4290 genes in the E. coli K-12 genome were purchased from Sigma-Genosys (The Woodlands, TX). Primers specific for Shigella-specific genes located on the sequenced 210 kb virulence plasmid (Venkatesan et al., 2001) and chromosome of S. flexneri 2a (Jin et al., 2002; Wei et al., 2003) were custom designed and synthesized by Sigma-Genosys. PCR amplification was carried out according to the instructions provided by Sigma-Genosys. Approximately 93% of all PCR amplifications were successful, and the corresponding genes are represented on these arrays. These arrays were used for expression analysis of the Fur and iron regulons during logarithmic growth.

Subsequent arrays were constructed using 70-mer oligonucleotides (Qiagen, Santa Clarita, CA) that were specific for each gene of the E. coli K-12 and enterohemorrhagic E. coli sequenced genomes. Oligonucleotides corresponding to Shigella-specific genes, as well as putative small RNA sequences previously identified in E. coli (Hershberg et al., 2003), were custom designed and synthesized by Qiagen. These arrays were used for all subsequent array analysis. Glass microscope slides were coated with poly-L-lysine as described at http://chipmunk.icmb.utexas.edu/ilcrc/protocols/index.shtml, and PCR products or oligonucleotides were printed on poly-L-lysine coated slides using a robotic Arrayer and Array Maker 2.4 as described by MGuide (cmgm.stanford.edu/pbrown/mguide/). Microarrays were then post-processed as described by MGuide.

Microarray analysis of iron and Fur regulation

Strains were grown as described in the text and RNA was isolated from cultures using RNEasy Midiprep Spin columns (Qiagen). Where indicated, L broth was supplemented with either 8 or 16 µg EDDA per millilitre to create iron-restricted conditions. When comparing gene expression in wild-type S. flexneri and fur, L broth was supplemented with 40 µM FeSO4 to fully repress Fur-regulated genes. When using PCR-amplified arrays, reverse transcription of RNA was carried out as follows: 2.4 µg of pd(N)6 randomized hexamer primers (Amersham Biosciences, Piscataway, NJ) were annealed with 50 µg total RNA for 10 min at 65°C. Primed RNA was incubated at 42°C with 0.5 mM dATP, 0.5 mM dCTP, 0.5 mM dGTP, 0.2 mM dTTP, 0.3 mM amino-allyl modified dUTP (Sigma-Aldrich, St. Louis, MO), 0.01 M dithiothreitol (DTT), 120 units RNasIn (Promega, Madison, WI), 800 units SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA) and 1X SuperScript RT Buffer. After 1 h of incubation, 400 additional units of SuperScript II were added and the reaction was continued for an additional hour at 42°C. RNA was hydrolysed by the addition of NaOH to a final concentration of 50 mM and was incubated at 65°C for 10 min; the reactions were then neutralized by the addition of HCl to a final concentration of 50 mM. When using oligonucleotide arrays, 5 µg of pd(N)6 was annealed to 15 µg of RNA during cDNA preparation.

Ribosomal cDNA generated during reverse transcription was removed when using PCR-amplified arrays by hybridization with biotinylated DNA encoding 16S and 23S ribosomal RNA as follows. 16S ribosomal DNA was amplified using the following primers: 16S.for (5′-CCACACTGGAACTGAGA CAC-3′) and 16S.rev (5′-GGACTACGACGCACTTTATG-3′). 23S ribosomal DNA was amplified using the following primers: 23S.for (5′-TGGTGTTACTGCGAAGGG-3′) and 23S.rev (5′-TCATCTCGGGGCAAGTTTCG-3′). DNA encoding 16S and 23S ribosomal RNA was then biotinylated using the Psoralen Biotin Kit (Ambion, Austin, TX) according to the manufacturer's instructions. A total of 225 ng of biotinylated DNA encoding 16S and 23S ribosomal RNA was added to cDNA samples, denatured at 95°C for 5 min, and hybridized to cDNA samples for 30 min at 60°C. Hybridized DNA was removed by incubation with magnetic streptavidin-coated beads (Promega) for 10 min at room temperature and ribosomal DNA/cDNA complexes were pulled down on a magnetic stand and removed. Samples were washed and concentrated to 9 µl on a Microcon-30 column (Millipore, Billerica, MA).

Amino allyl-dUTP incorporated in cDNA samples was coupled to reactive Cy3 or Cy5 dyes (Amersham Biosciences) as follows: 1 µl of 10 M sodium bicarbonate (pH 10) was added to the concentrated cDNA samples, and the buffered cDNA was combined with either Cy3 or Cy5 reactive dye. Reference samples derived from the wild-type strain grown without EDDA or the wild-type strain grown with 40 µM FeSO4 were routinely labelled with Cy3, and test samples derived from the wild-type strain grown in the presence of EDDA or the fur mutant grown with 40 µM FeSO4 were labelled with Cy5. The coupling reactions were incubated for 1 h in the dark, and then uncoupled dye was removed by purification of labelled samples on QiaQuick columns (Qiagen). Cy3- and Cy5-labelled samples were combined with 3X SSC and 0.25% SDS. cDNA probes were incubated at 65°C with the arrays for either 6 h (PCR-derived) or 3 h (oligonucleotide), then washed as described at http://chipmunk.icmb.utexas.edu/ilcrc/protocols/index.shtml.

Microarrays were scanned using the Genepix Array Scanner 4000 A (Axon Instruments, Union City, CA). Preliminary analysis of microarrays was performed using the Genepix 5.0 software, and normalization of microarray data was carried out by the Longhorn Array Database (powered by the Stanford Microarray Database). Normalized data were filtered so that spots with a regression correlation lower than 0.6 in any experiment were excluded from further analysis. Hierarchical clustering in Fig. 1 was carried out using Cluster 3.0 software and the expression map was generated using TreeView (programs available at bonsai.ims.u-tokyo.ac.jp/∼mdehoon/software/cluster/index.html).

Real time PCR

RNA was isolated from strains grown as indicated using RNeasy Midi Columns (Qiagen) and DNase-treated using DNase I, Amplification Grade from Invitrogen according to the manufacturer's instructions. cDNA was generated from approximately 5 µg of each RNA sample using the High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA). Real time PCR reactions in a total volume of 50 µl were set up as follows: 1X TaqMan Universal PCR Master Mix (available from Applied Biosystems), 900 nM of forward primer, 900 nM of reverse primer, 250 nM TaqMan MGB Probe and one-twentieth (< 10 µg) of the cDNA reaction. ryhB cDNA was detected using the primers RyhB.for (CGCG GAGAACCTGAAAGC) and RyhB.rev (CGGCTGGCTAAG TAATACTGGAA) and the TaqMan MGB Probe RyhB (6FAM-CGACATTGCTCACATTG-MGB-NFQ). sodB cDNA was detected using the primers SodB.for (GCGCATCATCCCCT TGAG) and SodB.rev (CTACTACAACGTGCGCATTGG) and the TaqMan MGB Probe SodB (6FAM-AACGCCATATCGCC -MGB-NFQ). ydeP cDNA was detected using the primers YdeP.for (GCGCATCATCCCCTTGAG) and YdeP.rev (CTAC TACAACGTGCGCATTGG) and the TaqMan MGB Probe YdeP (6FAM-AACGCCATATCGCC-MGB-NFQ). evgA cDNA was detected using the primers EvgA.for (CGGGAAACAT TGTGCTGATG) and EvgA.rev (TTGTTCATGCCTTCTTTTT TACTCA) and the TaqMan MGB Probe EvgA (6FAM-TGGCGCTAATGGTTT-MGB-NFQ). rrsA cDNA was detected using the primers RrsA.for (CACGATTACTAGCGATTCC GACTT) and RrsA.rev (CGTCGTAGTCCGGATTGGA) and the TaqMan MGB Probe RrsA (6FAM-ATGGAGTCGAGTTG CAG-MGB-NFQ). Primers and probes were designed using ABI Prism Primer Express software and synthesized by Applied Biosystems. Real time PCR was carried out in an Applied Biosystems 7300 Real Time PCR System, and analysis was performed using the 7300 Real Time PCR System software. A CT value, defined as the number of the first cycle in which the signal intensity exceeds a designated threshold value, was determined for each reaction. Standard curves for each primer/probe set were generated using cDNA generated from 10-fold dilutions of SM100 RNA, and the relative amount of each cDNA in each sample was extrapolated from the standard curve. The relative amounts of ryhB, sodB, evgA and ydeP cDNA were normalized by dividing the values by the relative amounts of rrsA cDNA in each sample.

Batch culture growth

Batch cultures were grown in a BIOFLO 110 Fermenter/Bioreactor (New Brunswick Scientific, Edison, NJ). Overnight cultures grown in L broth were diluted 1:100 into L broth in the bioreactor vessel. Cultures were grown in batch at 37°C. pH was monitored by the bioreactor and was recorded every 5 min. Growth readings (OD650) were taken every 30 min until cultures reached early stationary phase (OD650 ≈ 1.0).

Acid-resistance assays

Acid-resistance assays were carried out as described by Lin et al. (1995) with modifications. Strains were grown for 20 h at 30°C shaking in MOPS-buffered L broth at pH 7 (+ carbenicillin, IPTG or EDDA as indicated). Cultures were then diluted 1:50 into M9 minimal medium with the pH adjusted to 2.5. Acid shock cultures were incubated for 2 h at 37°C with aeration, and serial dilutions were plated at the start and the end of the incubation time. Per cent survival was determined by dividing the number of colony forming units (cfu) at 2 h by the cfu at 0 h.

Acknowledgements

We gratefully thank Elizabeth Wyckoff and Alexandra Mey for helpful scientific discussion and critical reading of the manuscript, and the members of the Iyer Laboratory for their assistance in microarray analysis. This work was supported by Grant AI16935 from the National Institutes of Health.

Ancillary

Advertisement