ABSTRACT
Siderophores are low-molecular-weight iron chelators secreted by microbes to obtain iron under deprivation. We hypothesized that the catecholate siderophore enterobactin, produced by Enterobacteriaceae, serves as a proinflammatory signal for respiratory epithelial cells. Respiratory tract responses were explored, since at this site siderocalin, an enterobactin-binding mammalian gene product, is expressed inducibly at high levels and enterobactin-secreting respiratory flora is rare, suggesting selection against a dependence on enterobactin. Addition of aferric, but not iron-saturated, enterobactin elicits a dose-dependent increase in secretion of the proinflammatory chemokine interleukin-8 by human respiratory epithelial cells in culture. This response to purified enterobactin is potentiated by recombinant siderocalin at physiologically relevant concentrations. Conditioned media from genetically modified Escherichia coli strains expressing various levels of enterobactin induce an enterobactin-mediated proinflammatory response. Siderocalin has been shown to deliver enterobactin to other mammalian cell types, exogenously supplied siderocalin can be detected within epithelial cells, and siderocalin increases delivery of enterobactin to the intracellular compartment. Although many siderophores perturb labile cellular iron pools, only enterobactin elicits interleukin-8 secretion, suggesting that iron chelation is necessary but not sufficient. Thus, aferric enterobactin may be a proinflammatory signal for respiratory epithelial cells, permitting detection of microbial communities that have disturbed local iron homeostasis, and siderocalin expression by the host amplifies this signal. This may be a novel mechanism for the mucosa to respond to metabolic signals of expanding microbial communities.
Of the myriad of competitive interactions known to occur between host and colonizing or infecting microbes, the struggle for micronutritional iron is among the most prominent. Iron is required by virtually all forms of cellular life, microbes included, for a number of important biochemical processes, including oxidative respiration, DNA and RNA processing, and others. The host is thought to actively restrict the concentration of free iron, especially in tissues that interface with microbial populations, a concept known as nutritional immunity (40). In turn, microbes have evolved a variety of countermeasures with which to overcome host iron restrictions (28), and a number of specific molecular mechanisms have been elucidated (32).
One of the means by which microbes can acquire iron is by the secretion of extracellular iron chelators, called siderophores, which are then reacquired by highly specific uptake systems (28). More than 500 microbial siderophores have been identified and characterized thus far, and despite an impressive diversity of structure, all are small soluble molecules (<1 kDa) with a high affinity for iron (KD ≤ 10−30 M) (39). Siderophores can be categorized by structure, and the tricyclic catecholate siderophores are among the better-studied groups, exemplified by the bacterial siderophore enterobactin (29), originally identified in Escherichia coli (26) and Salmonella (27).
Recently, it has been proposed that iron acquisition by siderophores could serve to modulate the inflammatory responses of host animals (9, 16). We sought to test this question in an experimental model of the human respiratory tract by exposing the epithelial cell line A549 to a variety of microbial siderophores. Responses were determined by measuring secreted interleukin-8 (IL-8), an important proinflammatory chemokine for neutrophils in humans, as neutrophil recruitment is thought to be a critical element of host defense during a variety of bacterial respiratory infections. The respiratory tract was chosen for study for two reasons.
First, the mammalian gene product siderocalin (also known as lipocalin 2 [Lcn2], neutrophil gelatinase-associated lipocalin, 24p3, and others) is known to bind enterobactin with high affinity (1, 15) and is inducibly expressed in human (10) and mouse (20, 25) respiratory tracts and in A549 cells (10). In a previous report, we described upregulation of mucosal secretion of mouse siderocalin in an animal model of upper respiratory tract colonization by both gram-positive and gram-negative organisms (25). It has been suggested that siderocalin competes with at least some microbes for secreted siderophores, and therefore for iron, thereby restricting the growth of siderophore-dependent organisms (13, 15). This hypothesis has not been explored in the context of the mucosal surface, where the vast majority of host-microbial interactions and competition occur.
A second justification for studying the biology of enterobactin, siderocalin, and the respiratory tract is the observation that few enterobactin-producing organisms are known to colonize or infect the respiratory tract in healthy individuals (19, 23, 37). This is in contrast to the intestinal flora, where many enterobactin biosynthesis or utilization genes can be detected, particularly among members of the Enterobacteriaceae. In the respiratory tract, increasing concentrations of enterobactin may signal the expansion of an alien microbial community and, therefore, serve as an unrecognized inflammatory signal. We hypothesize that epithelial cells derived from tissues where enterobactin-producing organisms are commonly found are likely to be tolerant of the siderophore to prevent the induction of inflammation. Siderocalin production is abundant in the respiratory tract and sparse in the gut, an inverse distribution compared with the enterobactin-producing microflora, raising the possibility that this siderophore-binding protein contributes to the tropism of microbial populations.
In this report, we demonstrate that enterobactin serves as a proinflammatory signal to human epithelial cells in an iron-dependent manner and that siderocalin potentiates the response to the siderophore, possibly by intracellular delivery.
MATERIALS AND METHODS
Cell culture and cell stimulation.A549 (ATCC CCL-185), a human type II pneumocyte cell line, Detroit 562 (D562, ATCC CCL-138), a human nasopharyngeal carcinoma cell line, and 16HBE14o-, a transformed human bronchial epithelial cell line (gift of Dieter Gruenert) (11), were maintained in minimal essential medium (Invitrogen, Carlsbad, CA) as described elsewhere (17) in 24-well plates for all experiments other than fluorescence microscopy (see below). Cells were weaned from serum and antibiotics overnight, and the following day the medium was replaced with serum-free and antibiotic-free medium containing siderophore, iron salts, protein, E. coli supernatant, or a combination thereof (see below). Cells were stimulated overnight, unless otherwise indicated, and the following day supernatant was collected for a lactate dehydrogenase (LDH) release assay or stored at −20°C for subsequent analysis by IL-8 enzyme-linked immunosorbent assay (ELISA).
An IL-8 OptEIA ELISA (BD Pharmingen, Franklin Lakes, NJ) and the LDH release assay (Roche, Branchburg, NJ) were performed according to the manufacturers ' instructions. Recombinant IL-1β (1 ng/ml; PeproTech, Rocky Hill, NJ) was used as a positive control for IL-8 secretion (10), and 0.1% Triton X-100 was used as a positive control for LDH release, as per the manufacturers' instructions. All experiments included a vehicle control well, accounting for the solvents for all reagents included in the experiment. For experiments with multiple independent replicates, data were normalized to the mean of the vehicle control conditions to account for variation in baseline IL-8 secretion.
Western blotting.After collection of the supernatant, cell monolayers were washed five times with sterile phosphate-buffered saline (PBS) and lysed for protein with Laemmli buffer, followed by a 5-min incubation at 100°C. Total cell lysate equivalent to 2.5 × 104 cells per lane was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis on a 4 to 15% Tris-HCl gel (Bio-Rad, Hercules, CA) and transferred to Immobilon-P transfer membranes (Millipore, Billerica, MA). Human transferrin receptor 1 (hTrfR1) was detected using a mouse anti-human monoclonal antibody to hTrfR1 (Invitrogen) and detected with a sheep anti-mouse-horseradish peroxidase-conjugated secondary antibody (GE Healthcare, Piscataway, NJ). Actin was detected using a goat anti-human actin antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and detected with a rabbit anti-goat-horseradish peroxidase-conjugated secondary antibody (Santa Cruz Biotechnology). Siderocalin was detected using a polyclonal anti-mouse siderocalin antibody (41) and detected with an anti-rabbit-horseradish peroxidase-conjugated secondary antibody (GE Healthcare). Secondary binding was detected with the enhanced chemiluminescence plus Western blotting detection system (GE Healthcare).
Densitometric analysis of hTrR1 Western blots was performed using the gel analysis tools of ImageJ (Wayne Rasband; NIH [http://rsb.info.nih.gov/j/ ]). Intensity values of hTrfR1 were normalized to actin intensity values to correct for gel loading, and all experimental conditions were then normalized to a baseline untreated condition to calculate a fold-change value.
Fluorescence microscopy.Recombinant siderocalin was incubated with Alexa568 succinimidyl ester (Invitrogen) in 0.1 M HCO3− buffer (pH 8.0) for 30 min followed by quenching with 0.1 M Tris buffer, dialysis against PBS, washing, and concentration on a Microcon YM-10 filter (Millipore). For these studies, A549 cells were grown on collagen type I-coated eight-well chamber slides (BD Biosciences Discovery, Bedford, MA). Protein-fluorophore conjugate was fed to serum-weaned cells for 5 h, and the cells were washed five times in sterile PBS, fixed in 4% paraformaldehyde in PBS, and imaged directly. Epifluorescence imaging was performed on a Nikon E600 Eclipse microscope equipped with a high-resolution charge-coupled-device digital camera (CoolSnap CF; Roper Scientific, Tucson, AZ) with Nomarski optics.
Radioactive 55Fe tracing. 55FeCl2 (see below) was bound to aferric enterobactin at room temperature for 1 hour in the dark. Separately, ferric-holo-siderocalin was prepared from recombinant apo-siderocalin and ferric enterobactin (EMC Microcollections, Tubingen, Germany) mixed at equimolar ratios. Labeled enterobactin was divided into three aliquots and added to vehicle control solution, recombinant apo-siderocalin, or ferric-holo-siderocalin, respectively, and incubated for 1 hour at room temperature in the dark, followed by dilution in serum-free antibiotic-free medium. A549 cell monolayers were stimulated for 7 h, after which the medium containing labeled enterobactin and protein was removed. The monolayers were washed 11 times with sterile PBS and lysed (50 mM Tris pH 7.4, 300 mM NaCl, 5 mM EDTA, 1% Triton X-100) for scintillation counting.
Reagents.Recombinant mouse siderocalin was produced as a glutathione S-transferase fusion protein in E. coli BL21 grown in iron-rich medium to minimize endogenous enterobactin expression, captured on glutathione beads (GE Healthcare), and cleaved with human thrombin (Sigma-Aldrich, St. Louis, MO), as previously described (41). Digested protein was gel filtered on an Agilent 1100 series high-performance liquid chromatograph (Agilent Technologies, Palo Alto, CA) and a Superdex 200 10/300 GL column (GE Healthcare) in PBS at 0.75 ml/minute and detected based on absorbance at 280 nm. Purified protein appeared as a single band of the expected size upon electrophoresis, reacted with anti-mouse siderocalin antibody by Western analysis, and bound enterobactin by spectroscopic and radioisotope analyses (data not shown).
Other investigators have described mouse and human siderocalin as having similar properties (13, 15, 20). Although amino acid sequence similarity overall is 62%, the majority of amino acids considered critical for function are conserved or similar (15). In this regard, we showed specific uptake of mouse siderocalin and binding of enterobactin consistent with a functional equivalency. Moreover, it has been shown that both mouse and human siderocalin protein are endocytosed and induce nephrons in both mouse and human embryonic renal mesenchyme (reference 41 and unpublished observations).
Purified E. coli enterobactin was resuspended in methanol for storage. 2,3-Dihydroxybenzoic acid (DHBA), ferrichrome, and ferric (Fe3+) ammonium citrate salt were purchased from Sigma-Aldrich. Desferal, desferrioxamine mesylate, was purchased from Ciba Pharmaceuticals (Summit, NJ). 55FeCl2 was purchased from Perkin-Elmer (Boston, MA).
Bacterial strains and culture conditions.Single colonies of E. coli strains were inoculated from Luria-Bertani plates into Luria-Bertani liquid culture for overnight growth and then diluted 1:10 in M9 minimal medium (Invitrogen). Bacteria were grown in liquid culture to late log phase (optical density at 620 nm, 0.6 to 0.8; approximately 108 CFU/ml) and centrifuged, and the supernatants were collected and filtered with a 0.22-μm filter (Whatman, Florham Park, NJ) and diluted 1:100 into serum-free, antibiotic-free cell culture medium for cell stimulations. Strains utilized included wild-type parental E. coli K-12 E541 (MG1655) (6), entA mutant E792 (BN1071; donated by P. Klebba) (21), and fur mutant E793 (BN4020; obtained from the E. coli Genetic Stock Center, strain 7540) (3).
Statistical analysis.IL-8 ELISA data were analyzed with a one-way analysis of variance (ANOVA) with the Tukey-Kramer multiple comparisons test performed using GraphPad InStat version 3.00 for Windows 95 (GraphPad Software, San Diego, CA). Figure 5, below, was generated and statistics performed using GraphPad Prizm 4 (GraphPad).
RESULTS
Aferric enterobactin elicits IL-8 secretion from respiratory epithelial cells.The proinflammatory properties of the bacterial catecholate siderophore enterobactin were tested in a cell culture model of the respiratory tract. Purified aferric enterobactin was added to the tissue culture medium of A549 cell monolayers in the presence or absence of equimolar concentrations of ferric (Fe3+) ammonium citrate salt. A dose-dependent increase in IL-8 secretion was observed in A549 cells cultures treated with aferric enterobactin (Fig. 1A). A significant increase in IL-8 secretion, however, required exogenous enterobactin concentrations exceeding 50 μM, with no further increases at higher concentrations (Fig. 1A).
Aferric enterobactin elicits an increase in IL-8 secretion in A549 respiratory epithelial cells. (A) A549 cell monolayers were stimulated overnight with aferric enterobactin (closed circles) or aferric enterobactin and equimolar ferric ammonium citrate (open circles) at a range of concentrations as indicated, after which supernatants were collected and assayed for IL-8 by ELISA and normalized to a vehicle-alone control well. Values are expressed as means ± standard deviations of five independent determinations. *, P ≤ 0.01 using ANOVA when relative IL-8 values were compared for stimulation with aferric and ferric enterobactin. (B) A549 cell monolayers were stimulated overnight with aferric enterobactin (Ent), aferric DHBA, aferric ferrichrome (Ferr), or aferric deferrioxamine (DFO), each at 50 μM. After stimulation, culture medium was collected, assayed for IL-8, and normalized to a vehicle-alone control well. Values are expressed as means ± standard deviations of four independent determinations. *, P ≤ 0.01 using ANOVA when all groups were compared with one another.
No change in IL-8 secretion over baseline was observed in A549 cells treated with ferric enterobactin saturated by prior treatment with equimolar iron salt. Moreover, addition of iron salts at increasing concentrations failed to diminish the secreted IL-8 until the concentration of iron was equivalent to, or greater than, the concentration of enterobactin (data not shown). Treatment with iron salt alone did not alter IL-8 secretion (Fig. 2 and data not shown), indicating that ferric ammonium citrate salt is not a general inhibitor of epithelial inflammation at micromolar concentrations.
Siderocalin potentiates the proinflammatory response to aferric enterobactin. A549 cell monolayers were stimulated overnight with aferric enterobactin (Ent), ferric ammonium citrate (Fe), or siderocalin (Scn), each at 50 μM, after which supernatants were collected and assayed for IL-8 and normalized to a vehicle-alone control well. IL-1β served as a positive control. Values represent the means ± standard deviations of three independent determinations. *, P ≤ 0.001 using ANOVA when all groups were compared with one another.
No increase in IL-8 secretion was observed in A549 cell monolayers stimulated with aferric forms of other microbial siderophores, including deferrioxamine, ferrichrome, or DHBA, an iron-binding precursor of enterobactin (38) (Fig. 1B). All siderophores were evaluated for proinflammatory properties to concentrations as high as 250 to 500 μM, and still no significant increase in IL-8 secretion was detected with these siderophores (data not shown).
Only at siderophore concentrations exceeding 100 μM were elevated levels of cytotoxicity (≥15%) detected by an increase in LDH release. Therefore, all additional studies in this report utilized siderophore concentrations at or below 50 μM unless otherwise noted.
These data demonstrate that aferric enterobactin can elicit IL-8 secretion from cultured human respiratory epithelial cells and that stimulation is dependent upon the ability of enterobactin to chelate iron and the particular siderophore employed.
Siderocalin potentiates the proinflammatory response to aferric enterobactin.We tested whether siderocalin could modulate the proinflammatory response to enterobactin by A549 cells. Siderocalin potentiates the IL-8 secretory response to enterobactin when added to the enterobactin at equimolar concentrations prior to cell stimulation (Fig. 2, columns 3 and 6). As observed for enterobactin alone, addition of iron to the protein-siderophore complex abrogated the proinflammatory response to the level of protein alone (Fig. 2, columns 5 to 7). Siderocalin protein alone, in the absence of enterobactin, could elicit a smaller increase in IL-8 secretion (Fig. 2, column 5). It is not clear at this time if this response is due to the siderocalin protein itself or to copurified contaminants. The proinflammatory effect of siderocalin in the absence of exogenous enterobactin could not be attributed to enterobactin contamination during protein purification, as addition of exogenous iron salts did not diminish the response (Fig. 2A, columns 5 and 8). Regardless, to eliminate the possible effects of contaminants, we focused only on the iron-dependent increase in IL-8 secretion upon stimulation with enterobactin or the enterobactin-siderocalin complex.
The siderocalin-enterobactin complex induces IL-8 secretion in a dose-dependent manner, tested from 1 to 50 μM, with a fivefold increase in IL-8 secretion over baseline observed at 2 μM (data not shown). Further, potentiation was observed even when siderocalin protein was added at concentrations lower than the concentration of enterobactin, i.e., a molar ratio of <1.0 (data not shown).
Similar responses were seen in two other human cell lines derived from the respiratory tract, Detroit 562 and 16HBE (data not shown).
In all studies described using siderocalin protein, cytotoxicity was ≤15% (data not shown), as detected by an LDH release assay.
These data demonstrate that, in the presence of siderocalin, an increase in IL-8 secretion was seen at micromolar enterobactin concentrations, which is comparable to reported enterobactin concentrations in broth-grown gram-negative enteric bacteria (3, 31).
Enterobactin production determines the inflammatory response to E. coli supernatants.To determine if the proinflammatory response to purified enterobactin is relevant to microbial physiology, A549 cells were stimulated with media conditioned by isogenic E. coli strains that varied in production of enterobactin. Culture supernatants were used instead of live infections to permit restriction of iron during bacterial growth, maximize enterobactin production, and exclude proinflammatory effects of contact between bacteria and epithelia. The entA gene encodes a 2,3-dihydro-2,3-dihydroxybenzoate dehydrogenase enzyme required for synthesis of the 2,3-dihydroxybenzoic acids (38) that are cyclized via serine to generate enterobactin (14). Thus, entA mutants are unable to produce detectable enterobactin (21). The fur gene product is an important iron-dependent negative regulator, and fur mutants are locked in a physiologic state adapted for low concentrations of iron, regardless of their actual environment. fur mutants, therefore, secrete approximately 100-fold more enterobactin into broth culture supernatants compared with wild-type organisms (3). Supernatants obtained from the parental wild-type K-12 strain were also included as a physiologic control, producing intermediate concentrations of enterobactin that were sensitive to the iron concentration during culture.
Stimulation of A549 monolayers with conditioned medium increased IL-8 section over vehicle control in a manner proportional to the enterobactin secretion (Fig. 3A). That is, the IL-8 secretion by cells stimulated with entA supernatants was indistinguishable from unconditioned medium, wild-type K-12 supernatants elicited an increase, and fur supernatants elicited yet more. These data demonstrate that proinflammatory responses to enterobactin are not an artifact of the use of purified enterobactin and are present in response to naturally produced siderophore.
Enterobactin mediates the proinflammatory response to E. coli supernatants by respiratory epithelial cells. (A) Bacterial enterobactin secretion can be manipulated using genetically defined strains of E. coli. A549 cell monolayers were stimulated overnight with late-log-phase supernatants from E. coli entA, wild-type (wt), and fur strains, with relative enterobactin secretion levels indicated graphically below. After stimulation, culture medium was collected, assayed for IL-8, and normalized to a vehicle-alone control well. IL-1β served as a positive control. Values represent the means ± standard deviations of three independent determinations. *, P ≤ 0.05 using ANOVA when all groups were compared with one another. (B) Exogenous enterobactin can complement the proinflammatory defect in entA E. coli. A549 cell monolayers were stimulated overnight with entA E. coli supernatant without or with supplementation with exogenous aferric enterobactin at 12 to 100 μM, as indicated below, or with fur E. coli supernatant, after which culture medium was collected, assayed for IL-8, and normalized to a vehicle control well. Values represent the means ± standard deviations of three independent determinations. *, P ≤ 0.01 using ANOVA between responses to entA- and fur-conditioned media in the absence of exogenous enterobactin. (C) Siderocalin and iron modulate the proinflammatory response of respiratory epithelial cells to wild-type E. coli supernatants. A549 monolayers stimulated overnight with supernatants from wild-type K-12 E. coli with vehicle control (control), 40 μM exogenous ferric ammonium citrate (+Fe), 20 μM siderocalin (+Scn), or both 40 μM iron and 20 μM siderocalin (+S/F). Addition of siderocalin yields statistically significant increases in IL-8 secretion, and addition of iron significantly abrogates this potentiation. Values represent the means ± standard deviations of three independent determinations. **, P ≤ 0.01; *, P ≤ 0.05 (using ANOVA).
To ensure that differences in siderophore concentration were sufficient to explain differences in IL-8 secretion responses between E. coli mutant strains, exogenous aferric enterobactin was added to entA supernatants. Exogenous enterobactin (0 to 100 μM) added to entA supernatants increased its proinflammatory properties above baseline control levels and, at high concentrations of enterobactin, approximated the response to fur supernatants (Fig. 3B). These data indicate that enterobactin is sufficient to enhance the proinflammatory properties of entA supernatants following growth in minimal medium.
Further, in the case of wild-type K-12 E. coli, representative of a physiologically relevant, enterobactin-expressing strain, addition of siderocalin potentiated the IL-8 response to supernatant, and addition of iron salt partially abrogated the siderocalin-mediated potentiation (Fig. 3C). Both of these biochemical observations are consistent with an enterobactin-mediated inflammatory response. Similar observations were made using fur conditioned medium, such that siderocalin potentiated the response, iron attenuated the siderocalin-mediated potentiation, and addition of iron salts to fur-conditioned medium abrogated the IL-8 response in the absence of siderocalin (data not shown).
Airway epithelial cells take up siderocalin and enterobactin.It has been reported that siderocalin can bind to enterobactin (1, 15) and deliver it into the cytosol of nonrespiratory cells (12, 24, 41), and specific siderocalin receptors have been proposed (12, 18). The possibility that siderocalin might capture a microbial siderophore and transport it to host cells, resulting in changes in host signaling, has been suggested (4) but, to date, not tested directly. We therefore considered the hypothesis that siderocalin may inhibit the persistence of enterobactin-producing organisms in the respiratory tract by a second mechanism, triggering host inflammation.
To determine if siderocalin protein can interact directly with A549 cells, these cells were stimulated with siderocalin under conditions identical to those used for IL-8 assays. Retention of siderocalin protein by cells washed extensively was demonstrated by Western blotting (data not shown). In order to explore siderocalin binding and/or uptake further, we incubated fluorophore-conjugated siderocalin with A549 cells and examined them by fluorescence microscopy (Fig. 4A and B). High-power epifluorescence microscopy revealed a punctate distribution consistent with endocytosis and vesicular sorting (Fig. 4A), as previously described for other cell types (24, 41). Confocal microscopy detected fluorescent siderocalin several micrometers deep into cells (data not shown), suggesting the protein is internalized. Addition of a 10-fold excess of unlabeled siderocalin reduced the amount of fluorescent material found in association with the cells, suggesting a specific uptake mechanism (Fig. 4B, panels i and ii), perhaps by the recently identified putative siderocalin receptor (12).
Siderocalin can transport enterobactin into A549 respiratory epithelial cells. (A) Siderocalin may enter endocytic intracellular sorting pathways. Cells treated with Scn-fluorophore visualized at high power demonstrated staining patterns consistent with vesicular distribution. A representative field from one of two experiments is shown. Bar, 10 μm. (B) Siderocalin uptake occurs through specific recognition pathways. A549 cell monolayers were treated with recombinant siderocalin protein conjugated to Alexafluor 468 in the absence (i) or presence (ii) of a 10-fold excess of unlabeled protein. After incubation, the monolayers were washed extensively, fixed, and visualized. Note that labeled siderocalin protein is qualitatively found at lower intensities if competing protein is present, suggesting specific recognition of siderocalin by A549 cells, possibly via a specific mechanism. Bars, 100 μm. (C) Siderocalin delivers enterobactin into A549 cells. Enterobactin was loaded with 55Fe and added to A549 cell monolayers in the presence of vehicle control, apo-siderocalin (Scn), or holo-ferric siderocalin (Scn-Ent-Fe). After uptake, cell monolayers were extensively washed and lysed, and scintillation counting was performed on the lysates. Note that apo-siderocalin significantly increases the labeled enterobactin detected in A549 cell lysates. Values represent the means ± standard deviations of four independent experiments. *, P ≤ 0.05 using ANOVA when all groups were compared with one another.
We sought to determine whether the internalized siderocalin could transport siderophore into A549 cells. To this end, aferric enterobactin was loaded with 55Fe as a means of following the siderophore. Labeled siderophore was incubated in the presence of apo-siderocalin (siderocalin in the absence of ligand), holo-ferric siderocalin (siderocalin bound with ferric enterobactin), or with vehicle control solution and subsequently applied to A549 cell monolayers. Addition of apo-siderocalin significantly increased radioactivity detectable in monolayer lysates, suggesting that siderocalin protein with an empty binding pocket can deliver siderophore to epithelial cytosol. Further support for this model is offered by the inability of holo-ferric siderocalin, with an occupied binding pocket, to increase the labeled siderophore detected in epithelial lysates (Fig. 4C).
These data are consistent with reports of siderocalin-mediated enterobactin transport to the intracellular compartment of other cell types (12, 24, 41), and we hypothesize that siderocalin-mediated endocytosis of enterobactin is likewise important for potentiation of an IL-8 response.
Chelation of cellular labile iron pools is not sufficient to drive proinflammatory responses.As aferric, but not ferric, enterobactin can elicit IL-8 secretion from A549 cells, we hypothesized that perturbation of cellular iron pools triggered proinflammatory stimulation. hTrfR1 expression is upregulated when labile cellular iron pools are depleted, and expression is downregulated when iron pools are elevated (8), although these data do not permit distinction of direct chelation of intracellular iron or indirect effects due to chelation of extracellular iron. Immunoblotting of A549 cell lysates for hTrfR1 demonstrated significantly different expression in cell monolayers treated with siderophores, iron, siderocalin, or a combination of these reagents (Fig. 5).
Chelation of cellular labile iron pools is not sufficient for proinflammatory responses to enterobactin. A549 cell monolayers were stimulated overnight with 10 μM siderocalin (S), 40 μM aferric enterobactin (E), 40 μM ferric ammonium citrate (F), 40 μM deferrioxamine (D), or combinations thereof, after which cell monolayers were lysed and assayed for hTrfR1 expression by Western blotting, with actin serving as a loading control, and quantified by densitometry. The changes in transferrin receptor expression level over a baseline untreated control well are shown. Values represent the means ± standard deviations of three independent blots.
Enterobactin treatment increased expression of hTrfR1 by Western densitometry, while iron treatment decreased expression, as expected (Fig. 5). hTrfR1 expression was suppressed by treatment with iron-saturated enterobactin, indicating that hTrfR1 upregulation is dependent upon the chelation capacity of enterobactin. However, treatment of A549 cells with equivalent concentrations of deferrioxamine also resulted in an upregulation of hTrfR1 (P < 0.05), indicating that other siderophores were capable of perturbing labile cellular iron pools but not elicit IL-8 secretion (Fig. 1B). Thus, the ability of enterobactin to alter cellular iron pools is necessary, but not sufficient, to elicit IL-8 secretion, as desferrioxamine is capable of perturbing cellular iron without evoking a proinflammatory response.
We also hypothesized that siderocalin could potentiate the effect of enterobactin by increasing the ability of enterobactin to access the intracellular iron pools. However, hTrfR1 expression in A549 cells treated with enterobactin-siderocalin complex was not distinguishable from the level seen with enterobactin treatment (P > 0.05) (Fig. 5). This finding suggests that siderocalin does not potentiate the proinflammatory properties of enterobactin by increasing the efficiency of intracellular iron chelation by enterobactin. Although the differences are not statistically significant, it is not clear why siderocalin protein alone elicits an upregulation of hTrfR1 (Fig. 5, column 6). The upregulation of hTrfR1 in response to siderocalin alone may be due to copurified enterobactin, a hypothesis supported by the failure of siderocalin to upregulate hTrfR1 in the presence of iron salts (Fig. 5, column 10). However, given that addition of iron salts does not attenuate the IL-8 response to siderocalin protein, we instead hypothesize that siderocalin protein is capable of perturbing cellular iron homeostasis, as reported in HeLa cells (12).
Analysis of ferritin expression in A549 cells under these experimental conditions was compared but was found to be low to undetectable except upon treatment with iron, whether bound or not, where an increase in ferritin detection was observed (data not shown), indicating an increase in labile cellular iron.
DISCUSSION
This report demonstrates that aferric enterobactin can elicit IL-8 secretion, a measure of proinflammatory response, in human respiratory epithelial cells in culture. That enterobactin-mediated IL-8 secretion is effectively inhibited by addition of iron indicates first that the response is dependent upon the ability of enterobactin to chelate iron and, second, that the cells were responding specifically to the aferric enterobactin and not to copurified contaminants. This is, to our knowledge, the first report of iron-dependent proinflammatory responses to a microbial signal and supports a novel mechanism by which pathogens might be detected.
Four additional lines of evidence were utilized to demonstrate the role enterobactin might play in mediating proinflammatory responses to growing bacterial populations. First, a proinflammatory response was seen to medium conditioned by E. coli K-12, suggesting that wild-type bacteria may be sufficient to elicit an enterobactin-dependent proinflammatory response. Second, the proinflammatory response to wild-type supernatants could be potentiated by addition of exogenous siderocalin, and this potentiation was abrogated by addition of exogenous iron, providing biochemical support for a role for enterobactin in mediating proinflammatory responses. Third, genetic manipulation of enterobactin production can control the proinflammatory response of A549 cell monolayers to conditioned medium. Finally, supplementation of exogenous purified enterobactin complemented the defect in siderophore production in entA mutants, proving that enterobactin itself is sufficient to explain different responses to different genotypes. These data support a role for enterobactin in the proinflammatory response to replicating E. coli under iron-limiting growth conditions.
To the best of our knowledge, there are no data published on siderophore concentrations produced in vivo by microbes colonizing or infecting mammalian hosts. The E. coli fur mutant used in these studies can produce more than 100-fold more enterobactin than the parent strain when grown in vitro in iron-replete medium (3), conditions under which Salmonella enterica serovar Typhimurium fur mutants may produce more than 140 μM enterobactin (42). Iron is virtually undetectable in respiratory secretions, such as sputum (30, 36) or bronchial-alveolar lavage (35), and free iron is thought to be as low as 10−24 M in the serum (2). In contrast, the cell culture system employed for these studies contains supraphysiologic levels of iron (≤2 × 10−6 M [personal communication from the manufacturer]). We therefore hypothesize that in vitro conditions used for cell culture may underestimate the magnitude of proinflammatory responses seen in vivo, where severe iron restriction may significantly increase sensitivity to aferric enterobactin. Siderocalin concentration in the urine has been reported at 2 μM (24), suggesting the concentration at the renal epithelial surface where it is produced may be higher. While purified exogenous aferric enterobactin alone does not elicit a detectable increase in IL-8 except at concentrations above 20 μM, the siderocalin-enterobactin complex yielded a severalfold increase in IL-8 secretion at concentrations as low as 2 μM, which may be physiologically relevant for bacterial siderophore production and host mucosal siderocalin secretion and is well above the resident environmental iron concentration (3, 31). In this regard it is worth highlighting that enterobactin produced during in vitro growth was sufficient to elicit proinflammatory responses and that siderocalin potentiated the response to wild-type E. coli culture supernatant. We therefore conclude that siderocalin amplifies enterobactin-dependent inflammatory responses.
Siderocalin may have complementary functions to both inhibit siderophore-mediated iron uptake by pathogenic bacteria (13, 15) and to potentiate proinflammatory signaling by the epithelium, simultaneously mediating detection and growth inhibition of offending microbial populations. Independent reports of siderocalin-deficient mice demonstrate increased susceptibility to septic challenge with clinical isolates of enterobactin-dependent E. coli in an iron-dependent manner (5, 13). It was hypothesized that these defects were due to failure to iron restrict the bacteria in question, based on the in vitro bacteriostasis observation described above (15). However, we hypothesize that at least some of the deficiency in antimicrobial defense exhibited by siderocalin knockout mice may be due to attenuation of enterobactin-mediated proinflammatory signaling. Mice that lack siderocalin expression would be unable to capture microbial enterobactin, and proinflammatory responses to aferric enterobactin may be reduced, leading to diminished IL-8 secretion and a less effective inflammatory responses.
More broadly, we hypothesize that direct competition for microbial siderophores and indirect proinflammatory signaling by siderocalin are significant barriers to life in the respiratory tract for enterobactin-dependent organisms. As noted, enterobactin-dependent microbes are widespread in the intestinal tract but very rarely colonize or infect the respiratory tract of healthy individuals (19, 23, 37). Thus, it is possible that the respiratory tract is sensitive to enterobactin, while the epithelium of the intestinal tract would be predicted to be relatively tolerant, as enterobactin-producing Enterobacteriaceae produce high concentrations of enterobactin.
Another question raised by this work is how enterobactin-mediated signaling interacts with classical innate immune pathways, such as toll-like receptor 4 (TLR4)-mediated recognition of lipopolysaccharide (LPS). The choice of A549 cells as the experimental model may have been fortuitous, in that these cells are considered relatively unresponsive to LPS due to low expression of TLR4 (34), possibly biasing the response of these cells towards secondary ligands. It is clear, for instance, that TLR4 is critical for host defense during gram-negative pneumonia (7, 33), while no previous role for enterobactin has been described. We suggest that the proinflammatory response to enterobactin is relevant in vitro and note that we were able to observe responses to enterobactin and siderocalin in two other human respiratory epithelial cell lines. Further, we suggest this response is relevant in vivo but that appropriate experimental conditions to demonstrate this have not yet been tested. Given the clear and significant role played by TLR4 and LPS in defense of the respiratory tract, it is likely that enterobactin is of secondary importance, but this does not rule out an important role in experimental or clinical outcomes. However, to our knowledge, there are no reports comparing the inflammatory responses to enterobactin-sufficient and enterobactin-deficient infection models in vivo. Moreover, it is possible that low concentrations of enterobactin produced during bacterial growth signal synergistically with low concentrations of other proinflammatory microbial signals, such as LPS. While the studies described herein do not address this possibility directly, synergy may explain the robust responses of epithelial cells to diluted culture supernatants which contain enterobactin and other proinflammatory microbial by-products.
Although many questions about possibly mechanistic explanations for these observations remain, data presented herein suggest A549 cells take up siderocalin carrying enterobactin, likely by specific endocytosis, and sort the protein within the intracellular vesicular trafficking system, as shown for other cell types (41). While no in vivo demonstration of siderocalin uptake in the respiratory tract has yet been reported, it is possible that some portion of the siderocalin protein detected within upper respiratory epithelial cells reported previously was derived from lumenal secretions, rather than endogenous expression (25). Like enterobactin, several other noninflammatory siderophores deplete the epithelial labile iron pool, as assayed by changes in hTrfR1 expression, suggesting that the ability of enterobactin to chelate iron is necessary but not sufficient to elicit inflammation. It is possible that aferric enterobactin possesses biochemical characteristics aside from the capacity to chelate iron that initiate a proinflammatory response in respiratory epithelial cells. Alternatively, it may be that siderocalin potentiation is due to stabilization of bound enterobactin molecules, preventing degradation, and effectively increasing enterobactin concentration over time. Nor is it clear why A549 cells do not secrete IL-8 in response to deferrioxamine, in contrast to HT29 intestinal epithelial cells (9), although other groups have shown differential cytokine responses to treatment of human cells with different siderophores (9, 16, 22). Future studies will examine the cellular mechanisms of the response to enterobactin and enterobactin-siderocalin complex and may clarify the relative contributions of the bacteriostatic and proinflammatory properties of siderocalin.
In conclusion, we have described a novel innate immune proinflammatory signaling pathway based on recognition of an important aspect of microbial metabolism, iron acquisition by siderophore secretion. We propose that this pathway may play an important role in the respiratory tract, where siderocalin expression is abundant, and is induced by the presence of bacteria. Further, we propose that this pathway permits recognition of expanding populations of organisms that deplete iron, avoiding detrimental inflammatory responses to small microbial communities.
ACKNOWLEDGMENTS
We thank Andrew Dancis and Simon Knight for contributing reagents, Young Hwang for technical assistance with protein production, and Michael Buckstein and Harvey Rubin for assistance with gel filtration.
This work was supported by grants from the U.S. Public Health Service to J.N.W. (AI44231 and AI38446) and the Morphology Core of the Center for the Molecular Studies of Liver and Digestive Disease (center grant P30 DK50306).
FOOTNOTES
- Received 26 October 2006.
- Returned for modification 17 December 2006.
- Accepted 26 March 2007.
- Copyright © 2007 American Society for Microbiology
