INFECTION AND IMMUNITY, May 2010, p. 1841–1849 0019-9567/10/$12.00 doi:10.1128/IAI.01258-09 Copyright © 2010, American Society for Microbiology. All Rights Reserved.
Vol. 78, No. 5
Unique Host Iron Utilization Mechanisms of Helicobacter pylori Revealed with Iron-Deficient Chemically Defined Media䌤 Olga Senkovich, Shantelle Ceaser, David J. McGee, and Traci L. Testerman* Louisiana State University Health Sciences Center—Shreveport, Department of Microbiology & Immunology, 1501 Kings Highway, Shreveport, Louisiana 71130 Received 9 November 2009/Returned for modification 10 December 2009/Accepted 13 February 2010
Helicobacter pylori chronically infects the gastric mucosa, where it can be found free in mucus, attached to cells, and intracellularly. H. pylori requires iron for growth, but the sources of iron used in vivo are unclear. In previous studies, the inability to culture H. pylori without serum made it difficult to determine which host iron sources might be used by H. pylori. Using iron-deficient, chemically defined medium, we determined that H. pylori can bind and extract iron from hemoglobin, transferrin, and lactoferrin. H. pylori can use both bovine and human versions of both lactoferrin and transferrin, contrary to previous reports. Unlike other pathogens, H. pylori preferentially binds the iron-free forms of transferrin and lactoferrin, which limits its ability to extract iron from normal serum, which is not iron saturated. This novel strategy may have evolved to permit limited growth in host tissue during persistent colonization while excessive injury or iron depletion is prevented. H. pylori is intrinsically resistant to numerous antibiotics and can acquire resistance to other antibiotics, which makes treatment failure common. Cost and unacceptable drug side effects are also barriers to eradication of H. pylori infections, which makes development of new treatment methods a high priority. Based on the reported inability of H. pylori to utilize bovine Lf, this protein is being tested as a treatment adjuvant (11, 34, 41). Specific drugs are also being developed to inhibit heme utilization and other mechanisms of iron uptake by pathogens (7, 32). A clear understanding of a pathogen’s iron acquisition strategies is helpful in designing drugs with the purpose of starving the organism of iron. Iron may be periodically abundant in the gastric lumen as dietary iron is solubilized by gastric acid. H. pylori may therefore have a short window of opportunity to acquire iron before it is bound by mucus lactoferrin or taken up by epithelial cells. As H. pylori invades gastric crypts, free iron is expected to be scarce, and host iron-binding molecules may play an important role in maintaining colonization in this niche. In vitro studies of H. pylori iron utilization have been hampered by the lack of a suitable serum-free growth medium. Nonetheless, several investigators have made progress toward identifying host-derived iron sources utilized by H. pylori. One study reported that H. pylori can use heme as an iron source in a blood agar-based assay system (38). Another study claimed that H. pylori can use human Lf, but not bovine Lf, as an iron source (18). There are conflicting reports concerning the ability of H. pylori to obtain iron from Tf (18, 35). H. pylori has a tonB gene, but it lacks recognizable homologs of the Lf-, Tf-, and hemoglobin-binding proteins of other organisms. In our studies, we have found that H. pylori preferentially binds apo-Tf and apo-Lf rather than the iron-containing proteins. This mode of iron utilization is expected to reduce the potential for sepsis or overgrowth of the organism in tissue.
Like nearly all pathogenic bacteria, Helicobacter pylori must acquire iron to grow. Many pathogens produce siderophores, which are potent chelators that are capable of extracting iron from a variety of host proteins. Despite numerous efforts, no H. pylori-associated siderophores have been found (12). Without autologous siderophores or the ability to use siderophores produced by other species, pathogenic bacteria require specific receptors to utilize host iron-binding molecules as iron sources. The potential iron sources in the host include heme, hemoglobin (Hb), transferrin (Tf), lactoferrin (Lf), ferritin, hemosiderin, and iron-containing enzymes. Lactoferrin is present in gastric mucus, and the level of this molecule has been found to be increased in individuals with H. pylori-mediated gastritis (28). Bacteria normally extract iron from Tf or Lf before importing the iron and releasing the apoprotein (1). In the case of heme-containing proteins, the entire heme molecule is removed from the protein and imported into the cytoplasm, where it can be degraded to release iron or used as a prosthetic group (15). H. pylori iron acquisition is also of interest because infection with H. pylori is associated with iron-deficiency anemia (27). H. pylori could cause iron deficiency indirectly by suppressing gastric acid secretion, thereby reducing solubilization and uptake of dietary iron. There is also evidence obtained with a mouse infection model suggesting that H. pylori successfully competes for dietary iron when the diet is iron poor (20). Furthermore, a recent study showed that clinical isolates of H. pylori from patients with iron-deficiency anemia have higher rates of inorganic iron uptake than strains from patients with non-irondeficiency anemia (39). At present, it is not known to what extent H. pylori relies on iron in the host’s diet and how much iron is obtained from host molecules originating from host blood or tissue.
* Corresponding author. Mailing address: Louisiana State University Health Sciences Center—Shreveport, Department of Microbiology & Immunology, 1501 Kings Highway, Shreveport, LA 71130. Phone: (318) 675-8143. Fax: (318) 675-5764. E-mail:
[email protected]. 䌤 Published ahead of print on 22 February 2010.
MATERIALS AND METHODS Culture of organisms. H. pylori strains (Table 1) were routinely cultured on Campylobacter blood agar containing 10% defibrinated sheep blood and in
1841
1842
SENKOVICH ET AL.
INFECT. IMMUN.
TABLE 1. H. pylori strains used in this study Strain
26695m UMAB41 HpDJM17 43504 SS1 J166 J178 B134A J54 a
Source
Motile 26695 Clinical isolate ATCC R. R. R. R.
Peek, Peek, Peek, Peek,
duodenal ulcer isolatea gastritisa gastric ulcera duodenal ulcera
Reference
33 26 25 21 22
Kindly provided by R. Peek from his personal collection.
Ham’s F-12 medium (Mediatech, Inc., Manassas, VA) supplemented with 1% fetal bovine serum. Prior to experiments, organisms were inoculated into Ham’s F-12 medium without serum and allowed to reach mid- to late-log phase in order to minimize the effects of serum carryover. Preparation of an iron-deficient medium (TT18). Glassware was soaked for 4 to 7 days in 2% nitric acid to remove iron and then rinsed with ultrapure water. The formula for TT18, the chemically defined medium that we used, is based on the formula for Ham’s F-12 medium. The final (1⫻) concentrations of the components of TT18 are as follows. The final concentrations of amino acids were 8.9 mg/liter alanine, 211 mg/liter arginine-HCl, 15 mg/liter asparagine 䡠 H2O, 13 mg/liter aspartic acid, 35 mg/liter cysteine-HCl, 14.7 mg/liter glutamic acid, 146 mg/liter glutamine, 7.5 mg/liter glycine, 21 mg/liter histidine-HCl monohydrate, 4 mg/liter isoleucine, 13 mg/liter leucine, 36.5 mg/liter lysine-HCl, 4.5 mg/liter methionine, 5 mg/liter phenylalanine, 34.5 mg/liter proline, 10.5 mg/liter serine, 12 mg/liter threonine, 2 mg/liter tryptophan, 7.8 mg/liter tyrosine 2Na dihydrate, and 11.7 mg/liter valine. The metal-salt solution used contained 44 g/ml CaCl2 䡠 2H2O, 2.5 ng/ml CuSO4 䡠 5H2O, 22.4 g/ml KCl, 57.2 g/ml MgCl2 (anhydrous), 7.6 mg/ml NaCl, 0.86 ng/ml ZnSO4 䡠 7H2O, and 1.2 ng/ml phenol red. The medium also contained a sodium phosphate-sodium bicarbonate solution consisting of 1.176 mg/ml NaHCO3 and 0.142 mg/ml Na2HPO4, as well as 4.77 g/ml hypoxanthine, 0.11 g/ml sodium pyruvate, and 0.3 g/ml thiamineHCl. The medium was prepared by mixing appropriate ratios of stock solutions (10⫻ L-amino acids, 5⫻ metal-salt solution, 10⫻ sodium phosphate-sodium bicarbonate solution, 1,000⫻ hypoxanthine, 1,000⫻ sodium pyruvate, and 1,000⫻ thiamine-HCl) with ultrapure water. All solutions were filter sterilized, and all solutions except the metal-salt solution were treated with Chelex resin (Chelex 100 resin; catalog no. 143-2832; Bio-Rad Laboratories, Hercules, CA) to further reduce iron contamination. Since Chelex is not highly specific for iron, the metal-salt solution could not be treated without adversely impacting the zinc and magnesium concentrations. Deferoxamine mesylate (Sigma-Aldrich, Inc., St. Louis, MO) was added to the medium in some experiments, as indicated below. Preparation of iron-saturated serum and serum components. Dialysis chambers containing a protein sample were placed in 0.1 M sodium citrate-bicarbonate buffer (pH 8.2) containing 70 mM ferric chloride as previously described (24). Samples were subsequently dialyzed against phosphate-buffered saline (PBS) containing Chelex resin to remove any unbound iron. Total iron content and total iron-binding capacity assays were carried out by the LSUHSC Hospital Clinical Chemistry Laboratory before and after the iron saturation procedure. The unsaturated iron-binding capacity was assayed using a Ferrozine-based colorimetric assay (Equal Diagnostics, Exton, PA), and total iron was assayed using a FERENE-based method (Abbott Diagnostics, Abbott Park, IL). Proteins were stored at 4°C in the presence of Chelex resin, which absorbed any residual iron. Since Chelex has a lower binding affinity for iron than Lf, Tf, or heme, we assumed that protein preparations that had been determined to be iron saturated remained iron saturated after storage in the presence of Chelex. Measurement of growth by ATP assay or plate assay. Twenty-four well plates containing 1-ml aliquots of TT18 were supplemented with the iron-binding molecules indicated below and inoculated with 10 l of a logarithmic-phase H. pylori culture grown in serum-free F-12 medium (starting inoculum, approximately 1 ⫻ 104 CFU/ml). Experiments were performed in duplicate, triplicate, or quadruplicate, as indicated below. The culture plates were incubated at 37°C in an atmosphere containing 5% O2 and 10% CO2 for 16 to 42 h. Prior to the assay, the culture plates were evaluated by phase-contrast microscopy using an inverted microscope to assess the bacterial morphology and density of positive controls. If the positive-control cultures were not sufficiently dense, the plate was incubated for a few more hours. Assay plates were not analyzed if positive-control cultures
did not grow or if a significant percentage of the bacteria in the positive-control cultures were coccoid, which indicated that the viability was reduced. The ATP contents of adherent and planktonic populations were measured as described previously (36), except that 25-l samples instead of 100-l samples were mixed with 25 l of CellTiter-Glo reagent (Promega) instead of 100 l of CellTiter-Glo reagent. Although this assay gives reproducible results, levels of growth were compared only within an experiment, not between experiments. All experiments were carried out at least three times in order to ensure that the results were reproducible. All iron-binding molecules were tested to ensure that they did not interfere with the ATP assay. Numbers of CFU were assessed by preparing serial 10-fold dilutions in PBS. Aliquots (50 l) of each dilution were spotted in duplicate onto predried CBA plates and incubated at 37°C in an atmosphere containing 10% CO2 and 5% O2 for 4 to 5 days. Type 2 two-tailed Student’s t tests were performed with growth data using Microsoft Excel software. Flow cytometry analysis. Labeling of proteins with fluorescein isothiocyanate (FITC) was performed using an FITC labeling kit (Calbiochem catalog no. 343210; EMD Biosciences, La Jolla, CA). Unbound FITC was removed by passing proteins over a 5-ml desalting column (GE Healthcare), which was eluted with PBS containing 0.1% sodium azide. Removal of free FITC was confirmed by loss of a low-molecular-weight fluorescent band following polyacrylamide gel electrophoresis of labeled protein. One-milliliter aliquots of logarithmic-phase H. pylori cultures grown in Ham’s F-12 medium containing 1% fetal bovine serum (FBS) were pelleted with a microcentrifuge at 10,000 ⫻ g and resuspended in filtered phosphate-buffered saline (PBS) containing 500 g/ml bovine serum albumin. FITC-labeled protein was added to tubes, and samples were incubated at 37°C in the dark for 1 h. Samples were centrifuged and resuspended in a solution containing 2% (vol/vol) ultrapure, methanol-free formalin (catalog no. 04018; Polysciences, Warrington, PA) diluted in PBS. For competition experiments, labeled and unlabeled proteins were added at the same time. Flow cytometric determination of protein binding was performed with a BD LSRII (BD Biosciences, San Jose, CA) that was made available by the Research Core Facility at the Louisiana State University Health Sciences Center—Shreveport (Shreveport, LA). The LSRII has a Coherent Sapphire laser for 488-nm excitation, a JDS Uniphase HeNe laser for 633-nm excitation, and a Coherent VioFlame for 405-nm excitation. Data analysis was performed using FACS Diva software (BD Biosciences) and FlowJo software (Tree Star, Inc., Ashland, OR). A minimum of 20,000 events (bacteria) were collected for each sample. Gating of P2 was set to exclude at least 97.5% of the unstained population, and the same gate setting was used for all comparisons in a single experiment. Histograms were prepared by plotting the percentage of maximal fluorescence versus fluorescence.
RESULTS H. pylori can utilize hemoglobin as an iron source. Previously, studies of acquisition of iron by H. pylori from hostderived sources were hindered by the lack of a serum- or blood-free growth medium. We developed an iron-deficient chemically defined medium, TT18, which contains less than 0.1 M iron (based on atomic absorption spectroscopy). This medium is similar to the previously reported TT8 medium (33), but it contains NaCl at the same concentration as Ham’s F-12 medium and lacks iron. This medium supports minimal growth of H. pylori unless another source of iron is provided, which permits evaluation of iron utilization in the absence of serum or chelators. The absence of chelators is particularly important for studying iron acquisition from Tf and Lf because many commonly used iron chelators, such as deferoxamine, have affinities that are several orders of magnitude higher than that of transferrin (6), which makes them unsuitable for experiments with a relatively slow-growing organism such as H. pylori. We first examined the ability of H. pylori to obtain iron from hemoglobin. Since dehydrated hemoglobin powder may contain a small amount of degraded protein, free heme, and possibly even free iron, we dialyzed a hemoglobin solution using a membrane with a 10-kDa cutoff against PBS supplemented with Chelex resin. H. pylori growth was monitored by assaying
VOL. 78, 2010
FIG. 1. H. pylori utilizes hemoglobin as an iron source. Hemoglobin was added to TT18 at the concentrations indicated. Inocula containing the same amount of H. pylori strain 26695m were added to triplicate wells. Growth was measured by analysis of ATP after overnight incubation. rlu, relative light units. The error bars indicate standard deviations. *, P ⱕ 0.01 compared with the control.
the ATP content in 25-l aliquots of a culture. In this assay, 20,000 relative light units (RLU) corresponds to a culture density of 5 ⫻ 106 CFU/ml, while 100,000 RLU corresponds to 3 ⫻ 107 CFU/ml. The lower limit of detection is around 1 ⫻ 104 CFU/ml. As shown in Fig. 1, H. pylori exhibited dosedependent growth in the presence of hemoglobin, and toxicity was observed at concentrations greater than 3 M. Since hemoglobin is a tetramer, 1 M hemoglobin corresponds to 4 M heme iron. To further prove that the growth was due to the heme iron in hemoglobin and not due to free inorganic iron, we repeated the experiment using deferoxamine (DFO). DFO is a strong chelator which can extract iron from proteins such as Tf but cannot extract iron from heme (10). Addition of DFO to TT18 further suppressed growth of H. pylori. Addition of hemoglobin restored the ability to grow in the presence of DFO (data not shown). Free heme is toxic to H. pylori. All hemoglobin-utilizing bacteria studied to date extract the heme molecule from hemoglobin and transport it to the cytoplasm, where it can be used directly or degraded to release the iron. Therefore, organisms that utilize hemoglobin are expected to utilize free heme as well. It should be noted that when heme is free in solution, the iron is oxidized to Fe(III) and the molecule is properly referred to as hemin. For simplicity, here we use the term heme to refer to either the oxidized form or the reduced form. We attempted to culture H. pylori in the presence of heme concentrations ranging from 1 M to 50 M. In each case, the bacteria became coccoid, often appeared to be degraded, and lost ATP, indicating that there was toxicity (data not shown). Inorganic iron is not toxic to H. pylori at this range of concentrations (33). As positive controls, we used Vibrio cholerae and uropathogenic Escherichia coli strain CFT073. These organisms exhibited dose-dependent increases in growth with heme concentrations ranging from 1 M to 50 M, and no toxicity was noted (data not shown). Thus far, every H. pylori strain tested (J166, J178, B134A, J54, 43504, UMAB4, and G27) has shown sensitivity to heme at concentrations above 1 M. Consistent with the measured heme toxicity, predigestion of hemoglobin with trypsin made the hemoglobin solution toxic at concentrations that ordinarily stimulate
IRON UTILIZATION MECHANISMS OF H. PYLORI
1843
FIG. 2. Growth on heme iron in the presence of deferoxamine. Deferoxamine mesylate was added to TT18 at a final concentration of 1 M. Hemin chloride was added at a starting concentration of 62.5 nM, and serial 2-fold dilutions were prepared to obtain the concentrations shown. Inocula containing the same amount of strain 26695m were added to triplicate wells. Growth was assessed by measuring the ATP content after 30 h. rlu, relative light units. The error bars indicate standard deviations. *, P ⬍ 0.05 compared with the culture containing no heme.
growth (data not shown). Refrigerated hemoglobin solutions also become toxic after storage for more than a few weeks. Exogenous heme should be no more toxic than inorganic iron unless it is internalized. Therefore, we reasoned that toxicity could result from unregulated heme uptake and that heme might stimulate growth if it was provided at an appropriate concentration. We prepared serial 2-fold dilutions of heme in TT18 and found that the toxicity was reduced at concentrations below ⬃60 nM. We detected growth close to or slightly greater than that of the TT18 control at heme concentrations ranging from 15 to 30 nM in several experiments. These experiments clearly showed that there was a reduction in toxicity, but there was great variability between replicates. In order to reduce the variability and growth due to free iron, we added 1 M DFO to all wells and used a larger inoculum. As shown in Fig. 2, there was significant stimulation of growth by heme at concentrations above 15.6 nM. The RLU seen in the presence of DFO without added heme represent ATP from the inoculum along with limited growth using existing iron stores. The concentration of heme that results in toxicity appears to vary somewhat based on the inoculum size, and cultures started with a larger inoculum tolerated slightly higher heme concentrations. H. pylori can acquire iron from human and bovine lactoferrin and transferrin. Iron-saturated bovine and human Tf and Lf were dialyzed against Chelex-containing PBS and were stored in the presence of Chelex resin to remove all traces of unbound iron. These proteins were then added to TT18 at a concentration of 9 g/ml (⬃113 nM). Either human or bovine Tf or Lf consistently stimulated growth of H. pylori strain 26695m in iron-deficient TT18 (Fig. 3 and data not shown). In order to be certain that increased ATP levels represented real growth, we plated serial dilutions of cultures from one assay. The relative levels of growth were similar regardless of the assay method; 100,000 RLU corresponded to about 1 ⫻ 107 CFU/ml, and the lower limit of detection for the ATP assay corresponded to a culture density near 1 ⫻ 104 CFU/ml. Aside from speed, two major advantages of the ATP assay are that it permits enumeration of both planktonic and adherent bacteria
1844
SENKOVICH ET AL.
FIG. 3. Acquisition of iron from transferrin and lactoferrin. Ironsaturated human Tf (hTf), human Lf (hLf), bovine Tf (bTf), and bovine Lf (bLf) were added to TT18 at a concentration of 6 g/ml. As a positive control, 10 M FeCl3 was added. Inocula containing the same amount of H. pylori strain 26695m were added to quadruplicate wells. Growth was measured by analysis of ATP after overnight incubation. rlu, relative light units. The error bars indicate standard deviations. *, P ⬍ 0.01 compared with the TT18 control.
and that it is not affected by bacterial aggregation (33, 36). There are conflicting reports about the ability of H. pylori to use Tf, and use of Lf is reportedly species specific (18, 35). In contrast, we found that H. pylori grew with either Tf or Lf and that growth was similar when either human or bovine proteins were used. Strains J99 and B134A were also found to use bovine Tf and Lf (data not shown), leading us to believe that the lack of species specificity is common in H. pylori isolates. Interestingly, addition of apo-Tf or apo-Lf suppressed growth to levels below that in the control medium. Although these apoproteins are expected to bind free iron in the medium, one would expect that an organism capable of extracting iron from Tf or Lf would still be able to grow. When bovine apo-Lf and holo-Lf were mixed at different ratios, we found that growth was substantially inhibited when the proteins were mixed so that the total iron saturation was less than 75% (Fig. 4). A similar pattern was observed with bovine Tf (data not shown). These data suggest that H. pylori binds the apoproteins more strongly than the holoproteins or that the apoproteins have an inhibitory effect via a different mechanism. Lactoferrin is reported to have antibacterial properties unrelated to iron chelation (14), but we observed no evidence of toxicity as determined by culture, ATP, or microscopy. The growth of H. pylori was inhibited, but the bacteria were not killed. H. pylori cannot acquire iron from normal serum. Normal fetal bovine serum (FBS) contains Tf that is about 30% iron saturated and therefore has the capacity to bind free iron in bacteriological growth media. In fact, serum has been used as a chelator by some investigators (37). FBS consistently inhibits H. pylori growth in TT18 when it is added at a concentration of 1% (data not shown). A greater serum concentration (30%) was required to inhibit growth in Ham’s F-12 medium, consistent with the ⬃30-fold-higher iron concentration in F-12 medium (data not shown). When fetal bovine serum was iron saturated and dialyzed against Chelex-containing buffer, it stimulated growth of H. pylori in TT18 (data not shown). If H. pylori does not differentiate between holo- and apo-Tf, then one would expect that growth in serum that is partially iron
INFECT. IMMUN.
FIG. 4. Effect of lactoferrin iron saturation on H. pylori growth. Ferri-lactoferrin and apo-lactoferrin were added to TT18 at ratios that resulted in the saturation levels indicated. The total lactoferrin concentration was 9 g/ml. Inocula containing the same amount of strain 26695m were added to duplicate wells. Growth was assessed by measuring the ATP content. rlu, relative light units.
saturated would be slowed but not stopped. We therefore performed a time course experiment to follow growth of H. pylori in TT18 supplemented with 10% normal fetal bovine serum. The culture grew slightly during the first 24 h, reaching a peak ATP value of 7,700 RLU (⬃1.7 ⫻ 106 CFU/ml), it remained static for the next 24 h, and then the concentration of bacteria declined (data not shown). H. pylori cultured with 10% FBS in iron-replete medium normally reaches a density of 0.7 ⫻ 108 to 1 ⫻ 108 CFU/ml. The minimal growth observed during the first 24 h could have been due to utilization of existing intracellular iron stores. Organisms in wells containing TT18 with serum retained the bacillary morphology during the first 48 h, even though they did not grow after 24 h. The maintenance of bacillary morphology along with the ATP data suggested that H. pylori is inhibited, but not killed, by the apo-Tf and apo-Lf in normal serum; however, these results do not explain why H. pylori is unable to utilize the holo-Tf present in serum. H. pylori specifically binds transferrin, lactoferrin, and hemoglobin. Acquisition of iron from host proteins requires the presence of specific bacterial receptors. It should therefore be possible to detect binding of labeled host proteins to the bacterial surface. H. pylori was incubated with Tf, Lf, or hemoglobin that had been labeled with fluorescein isothiocyanate (FITC). As a positive control, we used rabbit IgG specific for H. pylori, followed by FITC-labeled goat anti-rabbit IgG as a secondary antibody. As a negative control, we used FITClabeled bovine serum albumin (BSA). We found that FITCBSA binds minimally to H. pylori, and a very great increase in fluorescence was detected following incubation with H. pylorispecific antibodies and an FITC-labeled secondary antibody (data not shown). BSA was also included with the FITC-labeled Tf, Lf, or Hb to reduce any nonspecific binding interactions that may have occurred. A significant increase in fluorescence was seen following incubation of H. pylori with human hemoglobin or Tf or Lf from either human or bovine sources (Fig. 5). Binding of FITC-labeled proteins (human LF, bovine Lf, human Tf, bovine Tf, human Hb) was compared for strains 26695m and SS1. The percentages of the bacterial populations with bound protein were comparable for the two strains over a range of ligand concentrations for all proteins except hemoglobin; strain SS1 bound hemoglobin significantly better than 26695m bound it (data not shown).
VOL. 78, 2010
FIG. 5. Binding of FITC-labeled proteins to H. pylori strain SS1. Three micrograms of each protein was incubated with 1 ml of H. pylori suspended in PBS for 1 h at 37°C. Samples were centrifuged and resuspended in PBS to remove unbound protein. The transferrin and lactoferrin were iron saturated. hTf, human Tf; hLf, human Lf; bTf, bovine Tf; bLf, bovine Lf; hHb, human Hb.
IRON UTILIZATION MECHANISMS OF H. PYLORI
1845
To test the specificity of the binding interactions, we added a 50-fold excess of unlabeled protein. In each case, fluorescence decreased substantially in the presence of unlabeled protein (data not shown). Next, we sought to determine whether the same receptor recognized both Tf and Lf. When FITC-labeled human Lf (40 g) was mixed with a 50-fold excess of unlabeled human Tf, only a small decrease in fluorescence was observed (Fig. 6A). On the other hand, human Lf was able to competitively reduce binding of FITC-labeled human Tf (Fig. 6B). These data suggest that the Tf receptor also binds Lf. Since Tf is not able to significantly reduce binding of Lf, it is possible that there are two Lf receptors and only one of them binds Tf or that the affinity for Lf is much higher than the affinity for Tf. When apo-Lf binding was measured over a broad concentration range, we observed binding greater than apo-Tf binding, and there was evidence of bimodal binding, in contrast with the pattern observed with apo-Tf (Fig. 7). Although not conclusive, these data are consistent with the pattern expected when two Lf receptors are present and only one of them is also able to bind Tf. This hypothesis is supported by the appearance of flow cytometry histograms created using the same samples. The bell curve of Tf binding is narrow over the
FIG. 6. Characteristics of transferrin and lactoferrin binding. H. pylori strain 26695m was incubated with FITC-labeled proteins at various concentrations (C and D) or in the presence or absence of excess unlabeled protein (A and B). The fluorescence of bacteria was measured by flow cytometry. (A) A 50-fold excess of transferrin does not prevent binding of lactoferrin. (B) A 50-fold excess of lactoferrin blocks binding by transferrin. (C) Effect of apo-Tf concentration on fluorescence distribution. (D) Effect of apo-Lf concentration on fluorescence distribution. hTf, human Tf; hLf, human Lf; bLf, bovine Lf.
1846
SENKOVICH ET AL.
FIG. 7. Percentages of bacteria displaying levels of fluorescence greater than the background level following incubation with different amounts of protein. Bacteria were incubated with FITC-labeled apotransferrin or apo-lactoferrin at concentrations ranging from 1 to 150 g/ml. Gating was set to exclude 99.5% of the unstained population. hTf, human Tf; bLf, bovine Lf.
INFECT. IMMUN.
Thus, there is no evidence that apoproteins are permanently bound to the receptors. Synthesis of receptors involved in iron uptake is typically iron regulated, and other workers have shown that H. pylori produces iron-repressible outer membrane proteins, although the identities of the proteins were not determined (38). In order to explore iron regulation of Tf, Lf, and Hb receptors, we split a culture of strain 26695 into subcultures that were grown in iron-replete and iron-deficient media. After 18 h of incubation, the abilities of the bacteria to bind FITC-labeled human hemoglobin or human or bovine Tf or LF were evaluated by flow cytometry. In each case, the percentage of bacteria displaying fluorescence greater than the gating threshold was higher when bacteria were grown in iron-deficient medium (Fig. 9). These findings suggest that the receptors involved in acquisition of iron from host proteins are induced when the iron supply is limited. DISCUSSION
entire concentration range (Fig. 6C), whereas the apo-Lf curve is narrow at the lower concentrations but broadens at higher concentrations, suggesting that there are two receptors with different Lf affinities (Fig. 6D). Of greatest interest, we examined the specificity of H. pylori receptors for apoproteins and holoproteins. Incubation of strain 26695m with 40 g/ml iron-saturated bovine Lf resulted in a substantial increase in fluorescence. Addition of a 50-fold excess of unlabeled bovine apo-Lf shifted fluorescence almost to the baseline level (Fig. 8A). In contrast, a 50-fold excess of unlabeled bovine holo-Lf did not reduce the fluorescence obtained when H. pylori was incubated with FITC-labeled bovine apo-Tf (Fig. 8B). Similar results were obtained using human apo- and holo-Tf (Fig. 8C and D). For comparison, we examined binding of bovine Lf to Neisseria gonorrhoeae. The binding pattern for this organism was the exact opposite of the pattern for H. pylori; apo-Lf was not able to compete with FITClabeled holo-Lf, but unlabeled holo-Lf dramatically decreased the fluorescence of organisms incubated with FITC-labeled apo-Lf (Fig. 8E and F). Binding curves generated by comparing different concentrations of apo- and holo-Lf and different concentrations of apo- and holo-Tf also showed that H. pylori had a clear preference for the apoproteins (data not shown). Thus, we concluded that H. pylori has a greater affinity for apo-Lf and apo-Tf, which is a pattern that differs substantially from the patterns for other pathogens. Given the ability of apoproteins to inhibit binding of the iron-containing versions, we wondered whether apoproteins are readily released from the receptors or remain bound, thus blocking subsequent binding of a holoprotein. In order to answer this question, we incubated bacteria with 2 mg of unlabeled holo- or apo-Tf or -Lf in order to saturate receptors. After 1 h, the bacteria were pelleted and resuspended in fresh buffer containing 40 g of FITC-labeled holo-Tf or holo-Lf. Bacteria were incubated for 1 h and formalin fixed prior to examination by flow cytometry. The fluorescence was similar regardless of whether bacteria had been preincubated with no protein, a holoprotein, or an apoprotein (data not shown).
H. pylori has hemolytic activity and has been found to adhere to red blood cells in capillaries in the lamina propria (3, 23). Although H. pylori has previously been reported to use hemoglobin for growth (12, 38), the methods used previously did not permit the workers to prove that growth was not due to free heme present as a contaminant or due to other heme-containing complexes, such as heme-hemopexin or hemoglobin-haptoglobin. By growing H. pylori in serum-free medium supplemented with different concentrations of dialyzed hemoglobin, we showed that growth is not a result of free heme and that haptoglobin is not required. We have not determined yet whether heme can be extracted from hemopexin or from hemoglobin bound to haptoglobin. A recent report showed that FrpB2 in strain J99, which is homologous to HP0915 in strain 26695, allows E. coli BL21/pLysS to grow on hemoglobin (16). We plan to pursue this finding in future studies. Since all hemoglobin-utilizing bacteria studied to date are also able to grow using free heme, the extreme toxicity of heme for H. pylori came as a surprise. Excess heme was not toxic to V. cholerae or uropathogenic E. coli when these organisms were grown in the same medium. These data led us to suspect that uptake of heme from hemoglobin in H. pylori is tightly regulated but that free heme can enter via a transporter that is not iron regulated, leading to iron toxicity. We believe that the toxicity observed at higher hemoglobin concentrations is the result of hemoglobin degradation and concomitant release of heme. Indeed, the hemoglobin concentration required to cause toxicity is different for different batches of dissolved hemoglobin and increases as solutions age. Attempts to isolate heme-resistant mutants from a transposon library have been unsuccessful so far, suggesting that the heme uptake mechanism used by H. pylori could have another essential role or that multiple transport pathways are involved. Although sensitivity to free heme seems to be a disadvantage for a hemeutilizing organism, it is unlikely that H. pylori encounters toxic free heme concentrations in vivo. Heme that is released by hemoglobin can be rapidly bound by plasma albumin or hemopexin, and the level of the heme released by epithelial cells is not likely to be toxic. A number of pathogens can use Tf or Lf as an iron source. Both of these proteins are found in blood, although the concen-
VOL. 78, 2010
IRON UTILIZATION MECHANISMS OF H. PYLORI
1847
FIG. 8. Competition between ferri-Tf or ferri-Lf and apo-Tf or apo-Lf. H. pylori strain 26695m (A to D) or N. gonorrhoeae (E and F) was incubated with FITC-labeled protein in the presence or absence of excess unlabeled protein. The fluorescence of bacteria was measured by flow cytometry. (A) FITC–ferri-Lf with or without a 50-fold excess of unlabeled apo-Lf. (B) FITC–apo-Lf with or without a 50-fold excess of unlabeled ferri-Lf. (C) FITC–ferri-Tf with or without a 50-fold excess of unlabeled apo-Tf. (D) FITC–apo-Tf with or without a 50-fold excess of unlabeled ferri-Tf. (E) N. gonorrhoeae incubated with FITC–ferri-Lf with or without a 50-fold excess of unlabeled apo-Lf. (F) N. gonorrhoeae incubated with FITC–apo-Lf with or without a 50-fold excess of unlabeled ferri-Lf. hTf, human Tf; bLf, bovine Lf.
trations of Lf are higher in mucus and other bodily secretions (30). Lactoferrin is secreted into gastric mucus, and this protein has been suggested to be important for H. pylori growth in vivo (18). Organisms lacking siderophores must have specific Tf or Lf receptors in order to extract iron from these proteins. For example, several pathogenic Neisseria species have specific receptors
for Tf and Lf (8), and Haemophilus influenzae can use Tf but not Lf (17). V. cholerae has been shown to bind Lf and, to a lesser extent, Tf (2). To date, there has not been a report of a receptor capable of binding both Lf and Tf, in spite of the significant amino acid homology (60% identity and 72% similarity) of these two proteins. Therefore, it was surprising that Lf could compete with
1848
SENKOVICH ET AL.
FIG. 9. Effect of iron restriction on binding of H. pylori to host ironcontaining proteins. Binding of FITC-labeled Lf, Tf, and Hb (40 g/ml) to wild-type strain 26695m was measured by flow cytometry following growth in iron-replete medium (F-12 with 1% FBS) or iron-restricted medium (TT18 with 1 M DFO) for 18 h. The data are the percentages of the fluorescent population (P2) in the total population (50,000 events). Fe⫹, iron-replete medium; Fe⫺, iron-restricted medium; hTf, human Tf; hLf, human Lf; bTf, bovine Tf; bLf, bovine Lf; hHb, human Hb.
Tf for binding to H. pylori. Our data are consistent with the presence of more than one H. pylori Lf receptor. We showed that the utilization of Tf and Lf by H. pylori is not species specific, contrary to a previous report (18). The previously reported inability of H. pylori to use bovine Lf led to a number of clinical trials in which Lf was used as a treatment adjuvant with the hope of starving H. pylori for iron. Our finding that H. pylori is able to utilize bovine holo-Lf may explain some of the disappointing results of clinical trials employing Lf as a treatment adjuvant (40, 41). A study comparing treatment of H. pylori-infected mice with bovine Lf and treatment of H. pylori-infected mice with human Lf concluded that neither protein reduced the bacterial load, and both groups of mice had higher levels of inflammation (based on myeloperoxidase levels) than controls (19). Our results suggest that bovine Lf inhibits H. pylori growth only if it is ingested in the absence of dietary iron that could be bound by the Lf. With sufficient iron intake, bovine Lf could augment rather than inhibit H. pylori growth, and ingested dairy products could also serve as an iron source. Although the Lf in dairy products is probably not completely iron saturated, it could absorb iron released from other foods as iron becomes solubilized by gastric acid. Transferrin from meat could also be a utilizable iron source under the right conditions. Several other pathogens have been shown to bind Tf from the host species but not Tf from nonhost species, suggesting that iron acquisition mechanisms can contribute to host tropism (4, 29). In contrast, we have found no evidence that iron acquisition contributes to host tropism in H. pylori. It has been reported that Lf has bactericidal activity for some microorganisms in addition to its ability to sequester iron (9, 13), but we have found no evidence that apo-Lf kills or damages H. pylori. The bacteria maintain the bacillary shape and do not die following exposure to apo-Lf; their growth is merely suppressed by iron limitation. H. pylori also appears to be unique in its preference for binding apo-Tf and apo-Lf. Several other pathogens, such as Neisseria meningitidis and Moraxella catarrhalis, specifically
INFECT. IMMUN.
FIG. 10. Free iron in the gastric mucosa. Lf, lactoferrin; Tf, transferrin; Hb, hemoglobin. Blue hexagons indicate iron-free Lf; red hexagons indicate iron-saturated Lf; curved blue bars indicate H. pylori cells.
bind holo-Tf (5, 31). We are not aware of any pathogen that preferentially binds apoproteins. This pattern of host iron utilization could explain why examination of predicted H. pylori protein sequences has not revealed homology with known Tfor Lf-binding proteins from other organisms. Since other pathogens that have been studied either require specific ferriproteins or show no preference, it is reasonable to speculate that receptors of these organisms recognize different amino acid residues of Tf and Lf. Much has been made of the ability of H. pylori to cause severe diseases, such as ulcers and gastric cancers, but it is perhaps more remarkable that in the vast majority of cases H. pylori is able to coexist with its host for decades while causing only asymptomatic inflammation. Many otherwise innocuous organisms, such as Staphylococcus epidermidis and Serratia marcescens, can cause severe invasive disease in immunocompromised individuals, yet H. pylori sepsis has never been reported. It is therefore possible that H. pylori has evolved mechanisms to avoid excessive tissue destruction in most cases. Decreased pathogenic potential is a distinct advantage for an organism that colonizes its host for life. Our data suggest that H. pylori limits its pathogenicity by binding apo-Tf and apo-Lf more strongly than the iron-containing forms, thus preventing growth in the bloodstream. This characteristic may also moderate competition with the host for iron. Although the preponderance of evidence suggests that there is an association between H. pylori and iron-deficiency anemia (27), more aggressive bacterial iron acquisition strategies might make anemia a more dominant symptom of H. pylori infection. The model shown in Fig. 10 suggests that H. pylori could easily acquire iron in the upper regions of the mucous layer but that growth in gastric crypts would be exceedingly difficult due to the prevalence of apo-LF. Since H. pylori is frequently attached to the epithelial surface and is present in crypts, it is likely able to acquire iron in this milieu. We have preliminary data suggesting that host epithelial cells do indeed improve the ability of H. pylori to grow in iron-limited media containing apo-Tf. This will be the topic of future investigations. In summary, our studies demonstrate that H. pylori uses iron acquisition strategies analogous to skimming cream from the top rather than stealing the whole milk can. Preferential binding of apoproteins ensures that the host does not suffer excessive iron
VOL. 78, 2010
IRON UTILIZATION MECHANISMS OF H. PYLORI
deprivation and prevents disseminated infection. Although the in vivo consequences of heme toxicity are unclear, heme toxicity could serve as a disincentive for acquisition of destructive extracellular proteases. H. pylori has coevolved with humans for millennia, and its iron acquisition strategies are further evidence suggesting that there has been an evolutionary path toward commensalism rather than toward pathogenesis; however, it is still unclear whether certain H. pylori strains are truly benign. ACKNOWLEDGMENTS This work was supported by Public Health Service grant K22 AI063307-01 from the National Institutes of Health and by a Bridge award from LSU Health Sciences Center. We thank Dalton Gossett for performing atomic absorption spectroscopy and Robert Perry for advice concerning iron removal. We are grateful to Shannon Mumphrey, Rob Chervenak, and Deborah Chervenak for their assistance with flow cytometry.
20.
21.
22.
23.
24.
25.
REFERENCES 1. Andrews, S. C., A. K. Robinson, and F. Rodriguez-Quinones. 2003. Bacterial iron homeostasis. FEMS Microbiol. Rev. 27:215–237. 2. Ascencio, F., A. Ljungh, and T. Wadstrom. 1992. Lactoferrin binding properties of Vibrio cholerae. Microbios 70:103–117. 3. Aspholm, M., F. O. Olfat, J. Norden, B. Sonden, C. Lundberg, R. Sjostrom, S. Altraja, S. Odenbreit, R. Haas, T. Wadstrom, L. Engstrand, C. SeminoMora, H. Liu, A. Dubois, S. Teneberg, A. Arnqvist, and T. Boren. 2006. SabA is the H. pylori hemagglutinin and is polymorphic in binding to sialylated glycans. PLoS Pathog. 2:e110. 4. Bahrami, F., A. Elkins, and D. F. Niven. 2003. Iron acquisition by Actinobacillus suis: identification and characterization of transferrin receptor proteins and encoding genes. Vet. Microbiol. 94:79–92. 5. Boulton, I. C., A. R. Gorringe, N. Allison, A. Robinson, B. Gorinsky, C. L. Joannou, and R. W. Evans. 1998. Transferrin-binding protein B isolated from Neisseria meningitidis discriminates between apo and diferric human transferrin. Biochem. J. 334:269–273. 6. Cabantchik, Z. I., O. Kakhlon, S. Epsztejn, G. Zanninelli, and W. Breuer. 2002. Intracellular and extracellular labile iron pools. Adv. Exp. Med. Biol. 509:55–75. 7. Carpenter, B. M., T. K. McDaniel, J. M. Whitmire, H. Gancz, S. Guidotti, S. Censini, and D. S. Merrell. 2007. Expanding the Helicobacter pylori genetic toolbox: modification of an endogenous plasmid for use as a transcriptional reporter and complementation vector. Appl. Environ. Microbiol. 73:7506–7514. 8. Cornelissen, C. N., and P. F. Sparling. 1994. Iron piracy: acquisition of transferrin-bound iron by bacterial pathogens. Mol. Microbiol. 14:843–850. 9. Dalmastri, C., P. Valenti, P. Visca, P. Vittorioso, and N. Orsi. 1988. Enhanced antimicrobial activity of lactoferrin by binding to the bacterial surface. Microbiologica 11:225–230. 10. Davis, B. A., and J. B. Porter. 2002. Results of long term iron chelation treatment with deferoxamine. Adv. Exp. Med. Biol. 509:91–125. 11. de Bortoli, N., G. Leonardi, E. Ciancia, A. Merlo, M. Bellini, F. Costa, M. G. Mumolo, A. Ricchiuti, F. Cristiani, S. Santi, M. Rossi, and S. Marchi. 2007. Helicobacter pylori eradication: a randomized prospective study of triple therapy versus triple therapy plus lactoferrin and probiotics. Am. J. Gastroenterol. 102:951–956. 12. Dhaenens, L., F. Szczebara, S. Van Nieuwenhuyse, and M. O. Husson. 1999. Comparison of iron uptake in different Helicobacter species. Res. Microbiol. 150:475–481. 13. Dionysius, D. A., and J. M. Milne. 1997. Antibacterial peptides of bovine lactoferrin: purification and characterization. J. Dairy Sci. 80:667–674. 14. Ellison, R. T., III. 1994. The effects of lactoferrin on gram-negative bacteria. Adv. Exp. Med. Biol. 357:71–90. 15. Genco, C. A., and D. W. Dixon. 2001. Emerging strategies in microbial haem capture. Mol. Microbiol. 39:1–11. 16. Gonzalez-Lopez, M. A., and J. J. Olivares-Trejo. 2009. The gene frpB2 of Helicobacter pylori encodes an hemoglobin-binding protein involved in iron acquisition. Biometals 22:889–894. 17. Herrington, D. A., and P. F. Sparling. 1985. Haemophilus influenzae can use human transferrin as a sole source for required iron. Infect. Immun. 48:248–251. 18. Husson, M. O., D. Legrand, G. Spik, and H. Leclerc. 1993. Iron acquisition by Helicobacter pylori: importance of human lactoferrin. Infect. Immun. 61:2694–2697. 19. Huynh, H. Q., M. A. Campbell, R. T. Couper, C. D. Tran, A. Lawrence, and R. N. Butler. 2009. Lactoferrin and desferrioxamine are ineffective in the
Editor: S. M. Payne
26.
27. 28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
1849
treatment of Helicobacter pylori infection and may enhance H. pylori growth and gastric inflammation in mice. Lett. Appl. Microbiol. 48:517–522. Keenan, J. I., R. A. Peterson, R. Fraser, C. M. Frampton, T. A. Walmsley, R. A. Allardyce, and J. A. Roake. 2004. The effect of Helicobacter pylori infection and dietary iron deficiency on host iron homeostasis: a study in mice. Helicobacter 9:643–650. Lee, A., J. O’Rourke, M. C. De Ungria, B. Robertson, G. Daskalopoulos, and M. F. Dixon. 1997. A standardized mouse model of Helicobacter pylori infection: introducing the Sydney strain. Gastroenterology 112:1386–1397. Loh, J. T., M. H. Forsyth, and T. L. Cover. 2004. Growth phase regulation of flaA expression in Helicobacter pylori is luxS dependent. Infect. Immun. 72:5506–5510. Martino, M. C., R. A. Stabler, Z. W. Zhang, M. J. Farthing, B. W. Wren, and N. Dorrell. 2001. Helicobacter pylori pore-forming cytolysin orthologue TlyA possesses in vitro hemolytic activity and has a role in colonization of the gastric mucosa. Infect. Immun. 69:1697–1703. Mazurier, J., J. M. Lhoste, J. Montreuil, and G. Spik. 1983. Comparative study of the iron-binding properties of human transferrins. II. Electron paramagnetic resonance of mixed metal complexes of human lactotransferrin. Biochim. Biophys. Acta 745:44–49. McGee, D. J., C. Coker, T. L. Testerman, J. M. Harro, S. V. Gibson, and H. L. Mobley. 2002. The Helicobacter pylori flbA flagellar biosynthesis and regulatory gene is required for motility and virulence and modulates urease of H. pylori and Proteus mirabilis. J. Med. Microbiol. 51:958–970. Mobley, H. L., M. J. Cortesia, L. E. Rosenthal, and B. D. Jones. 1988. Characterization of urease from Campylobacter pylori. J. Clin. Microbiol. 26:831–836. Muhsen, K., and D. Cohen. 2008. Helicobacter pylori infection and iron stores: a systematic review and meta-analysis. Helicobacter 13:323–340. Nakao, K., I. Imoto, N. Ikemura, T. Shibata, S. Takaji, Y. Taguchi, M. Misaki, K. Yamauchi, and N. Yamazaki. 1997. Relation of lactoferrin levels in gastric mucosa with Helicobacter pylori infection and with the degree of gastric inflammation. Am. J. Gastroenterol. 92:1005–1011. Ogunnariwo, J. A., and A. B. Schryvers. 1990. Iron acquisition in Pasteurella haemolytica: expression and identification of a bovine-specific transferrin receptor. Infect. Immun. 58:2091–2097. Otto, B. R., A. M. Verweij-van Vught, and D. M. MacLaren. 1992. Transferrins and heme-compounds as iron sources for pathogenic bacteria. Crit. Rev. Microbiol. 18:217–233. Retzer, M. D., R. Yu, Y. Zhang, G. C. Gonzalez, and A. B. Schryvers. 1998. Discrimination between apo and iron-loaded forms of transferring by transferrin binding protein B and its N-terminal subfragment. Microb. Pathog. 25:175–180. Stojiljkovic, I., V. Kumar, and N. Srinivasan. 1999. Non-iron metalloporphyrins: potent antibacterial compounds that exploit haem/Hb uptake systems of pathogenic bacteria. Mol. Microbiol. 31:429–442. Testerman, T. L., P. B. Conn, H. L. Mobley, and D. J. McGee. 2006. Nutritional requirements and antibiotic resistance patterns of Helicobacter species in chemically defined media. J. Clin. Microbiol. 44:1650–1658. Tursi, A., W. Elisei, G. Brandimarte, G. M. Giorgetti, M. E. Modeo, and F. Aiello. 2007. Effect of lactoferrin supplementation on the effectiveness and tolerability of a 7-day quadruple therapy after failure of a first attempt to cure Helicobacter pylori infection. Med. Sci. Monit. 13:CR187-CR190. Velayudhan, J., N. J. Hughes, A. A. McColm, J. Bagshaw, C. L. Clayton, S. C. Andrews, and D. J. Kelly. 2000. Iron acquisition and virulence in Helicobacter pylori: a major role for FeoB, a high-affinity ferrous iron transporter. Mol. Microbiol. 37:274–286. Williams, J. C., K. A. McInnis, and T. L. Testerman. 2008. Adherence of Helicobacter pylori to abiotic surfaces is influenced by serum. Appl. Environ. Microbiol. 74:1255–1258. Worst, D. J., M. M. Gerrits, C. M. Vandenbroucke-Grauls, and J. G. Kusters. 1998. Helicobacter pylori ribBA-mediated riboflavin production is involved in iron acquisition. J. Bacteriol. 180:1473–1479. Worst, D. J., B. R. Otto, and J. de Graaff. 1995. Iron-repressible outer membrane proteins of Helicobacter pylori involved in heme uptake. Infect. Immun. 63:4161–4165. Yokota, S. I., M. Konno, E. Mino, K. Sato, M. Takahashi, and N. Fujii. 2008. Enhanced Fe ion-uptake activity in Helicobacter pylori strains isolated from patients with iron-deficiency anemia. Clin. Infect. Dis. 46:e31–33. Zullo, A., V. De Francesco, G. Scaccianoce, C. Hassan, A. Panarese, D. Piglionica, C. Panella, S. Morini, and E. Ierardi. 2005. Quadruple therapy with lactoferrin for Helicobacter pylori eradication: a randomised, multicentre study. Dig. Liver Dis. 37:496–500. Zullo, A., V. De Francesco, G. Scaccianoce, G. Manes, C. Efrati, C. Hassan, G. Maconi, D. Piglionica, C. Cannaviello, C. Panella, S. Morini, and E. Ierardi. 2007. Helicobacter pylori eradication with either quadruple regimen with lactoferrin or levofloxacin-based triple therapy: a multicentre study. Dig. Liver Dis. 39:806–810.
