Human Polymorphonuclear Leukocytes Inhibit Aspergillus fumigatus Conidial Growth by Lactoferrin-Mediated Iron Depletion¹

Aspergillus fumigatus is a ubiquitous thermotolerant mold that thrives on decomposing organic materials. Conidia (spores) of this organism are resistant to chemical and physical stress and tend to remain airborne due to their small size (2–4 μm diameter) and hydrophobicity. It is estimated that the average human inhales at least 200 viable A. fumigatus conidia each day, and those people working around compost or in barns may be exposed to many multiples of this number (1). Normal humans are highly resistant to this fungus; however, immunosuppressed individuals, for example, patients undergoing organ transplant (2, 3) or chemotherapy for cancer (4), are highly susceptible to infection by A. fumigatus as well as other species in this genus, such as Aspergillus flavus, Aspergillus niger, and Aspergillus terreus. Iron overload is also a recognized risk factor for invasive aspergillosis in transplant patients (5, 6) and patients with hematologic malignancy (7). Although prophylactic antifungal drugs have reduced the incidence of aspergillosis in the immunosuppressed and immunodeficient (8, 9), the overall case fatality rate remains above 50% (9), making a better understanding of the basic biological processes governing host defenses against Aspergillus an urgent clinical need.

Aspergillosis is also a significant risk for individuals with chronic granulomatous disease (CGD),⁴ a primary immunodeficiency caused by mutations of or deletions in components of the NADPH oxidase of leukocytes (10). As a consequence of these defects, CGD patient leukocytes have a severely diminished (or absent) capacity to generate reactive oxygen species that normally help generate microbicides, such as hydrogen peroxide and hypochlorous acid (bleach). For CGD patients today, A. fumigatus is the most frequent microbial cause of death (11). Polymorphonuclear leukocytes (PMN) from CGD patients are unable to kill the hyphal form of Aspergillus in vitro unless a source of H₂O₂ is provided either by a cell-free H₂O₂ generating system (e.g., glucose/glucose oxidase) or by mixing as few as 1 normal PMN per 15 CGD PMN (12). Natural or innate immunity to Aspergillus has long been considered to depend on anticonidial activity of macrophages and antihyphal activity of PMN (13). The few studies that have examined the ability of PMN to inhibit Aspergillus conidia have reported conflicting results. Lehrer and Jan (14) found that whereas human PMN ingested A. fumigatus conidia, their germination was not altered within 3 h. Using purified myeloperoxidase, H₂O₂, and KI, Lehrer and Jan also demonstrated direct effects of these oxidants on germination of swollen conidia, but found resting conidia to be resistant to such cell-free oxidative attack. Later, Levitz and Diamond (15) found that swollen conidia are indeed killed by human PMN, but through in vitro experiments implicated oxidative products as the mechanism. Recent studies in mice have demonstrated the importance of PMN recruitment to the lung in response to A. fumigatus conidia (16). Interestingly, whereas somewhat greater conidial germination was noted in CGD mice (knockouts in the gp91 subcomponent of the NADPH oxidase), between 60 and 70% of conidia in airway fluids remained ungerminated even after 12 h (16). Effective orchestration of the host response is complicated by the rapid developmental transformation of Aspergillus upon inhalation and clearly requires several distinct pattern recognition pathways (17, 18, 19, 20). Likewise, host defenses must contend with two fundamentally distinct stages of this pathogen, leading to a complex and as yet incompletely understood host response.

Although A. fumigatus is the most common microbial cause of death in CGD, most CGD patients are free of aspergillosis for years despite constant exposure to its conidia. Therefore, oxygen-independent mechanisms must act to control the growth of the hundreds of A. fumigatus conidia inhaled every day. In this study, we show that PMN from both normal donors and CGD patients significantly inhibit A. fumigatus conidial germination in an O₂-independent manner and identify lactoferrin (LF) as an important contributor to this activity. In the lungs, the most common portal of entry for Aspergillus, PMN LF sequestration of iron would thus play an essential early role in arresting conidial germination, thereby allowing their removal by mucociliary action, a process thought to play an important role in pulmonary innate defenses (21).

A. fumigatus strain B-5233 is a pulmonary clinical isolate from a leukemic patient. It was maintained on Aspergillus minimal medium. Conidia were harvested from 1-wk-old cultures with PBS containing 0.1% Tween 20, washed, and stored in water at 4°C for <2 mo.

HBSS without phenol red, calcium, or magnesium, referred to as HBSS⁻ and RPMI 1640, was from Invitrogen Life Technologies. 2,3-bis-(2-methoxy-4-nitro-5-sulphenyl)-(2H)-tetrazolium-5-carboxanilide (XTT), coenzyme Q, DNase I, horse spleen ferritin, human iron-saturated LF, iron-poor LF, and rabbit anti-LF Western blotting Ab were purchased from Sigma-Aldrich. Yeast nitrogen base medium was from Difco. Carboxymethylfluoroscene diacetate (CMFDA; Molecular Probes/Invitrogen Life Technologies) was dissolved in DMSO and stored in aliquots at −80°C until use. PMA (Sigma-Aldrich or Upstate Biotechnology) was dissolved in DMSO at 1 mg/ml and stored at −20°C. LF was loaded with iron by overnight dialysis in a 3.5-kDa membrane vs 1 mM ferrous ammonium sulfate in water at 4°C, followed by dialysis for 2 days in 1000 vol/day of 150 mM NaCl buffered with 10 mM HEPES (pH 7.2).

All human blood samples used in these studies were collected after informed consent from normal subjects (National Institutes of Health Protocol 99-CC-0168) and patients with CGD (National Institutes of Health Protocol 93-I-0119). Patients with clinically apparent infectious disease were excluded. Blood was anticoagulated using acid citrate dextrose (BD Biosciences Vacutainer ACD Solution A) and centrifuged at 470 × g for 10 min to separate the plasma from the bottom phase containing hemocytes. All procedures were performed in sterile nonpyrogenic 15- or 50-ml polypropylene tubes. The hemocytes were diluted with HBSS⁻ to the lesser of 25 or 50 ml and mixed with an equal volume of 3% dextran 500 (Amersham) dissolved in HBSS⁻. While the erythrocytes were settling in the dextran, the plasma from the first spin was spun again to enhance leukocyte recovery. The resulting plasma was then centrifuged at 2060 × g for 10 min to remove platelets. After the erythrocytes formed a distinct phase (between 20 and 30 min later), the supernatant was removed from the aggregated erythrocytes and centrifuged for 5–10 min at 300 × g. To the resulting pellet, 10 ml of 33 mM NaCl was added, followed 30 s later by 10 ml of 267 mM NaCl to restore isotonicity. The tube was filled with HBSS⁻, spun as before, and the lysis was repeated. The pellet was then resuspended in 2 ml of HBSS⁻ and overlayed on a discontinuous gradient of Percoll (previously rendered isotonic with 10× PBS) diluted with autologous plasma to 51% (bottom layer) and 42% (top layer). The gradient was centrifuged for 10 min at 270–300 × g, and the PMN-rich pellet was resuspended and washed once in HBSS⁻. Purity of PMN prepared by this method was typically 95–99%.

The XTT assay of fungal viability described by Meshulam et al. (22) was modified as follows. To minimize loss of fungi, a 96-well 0.45-μm filter-bottom plate (Millipore Multiscreen-HV) was used. Each well received 100 μl of Aspergillus conidia at 1 × 10⁶/ml RPMI 1640 supplemented with 25 mM HEPES (pH 7.2). For some experiments, the conidial suspension was added to clear-bottom plates to observe fungal morphology. Within 4 h of incubation at 37°C in a 5% CO₂ incubator, ≥98% of the condia had swollen. After a total of 7–8 h, the swollen conidia germinated into hyphae. At indicated times, autologous plasma alone or plasma with indicated PMN numbers was added. After incubations of cells with fungus, the plates were filtered and bovine DNase I was added at 10 U/ml in 10 mM Tris (pH 7.5), 2.5 mM MgCl₂, and 0.5 mM CaCl₂ to enhance removal of PMN. PMN were lysed by addition of 1% Igepal CA-630 (Sigma-Aldrich) for 15 min at room temperature, and each well was washed with a total of 1 ml of water to remove debris. Yeast nitrogen base medium was added, and the plates were incubated for 12–16 h at 37°C. Plates were filtered, and 200 μl of PBS containing 0.5 mg of XTT/ml and 40 μg/ml coenzyme Q (2,3-dimethoxy-5-methyl-1,4-benzoquinone) was placed in each well. After 2–6 h at 37°C (depending on whether the experiment was measuring conidial or hyphal growth), 100 μl of supernatant was transferred to a clear microtiter plate and the absorbance at 450 nm was read in an Anthos Zenyth 3100 plate reader.

PMN were suspended in RPMI 1640 supplemented with 25 mM HEPES (pH 7.2) at a concentration of 2 × 10⁷/ml. Samples were incubated alone or with 10 ng of PMA/ml for 30 min at 37°C. After stimulation, suspensions were centrifuged at 2000 × g, and supernatants were stored at 4°C. Dialysis of neutrophil supernatants was performed at 4°C by placing 1 ml of supernatant into 3.5-kDa cutoff Pierce Snakeskin Dialysis tubing and stirring constantly for 16 h in 4 L of 1 mM (NH₄)₂Fe(SO₄)₂. In the first experiment (two donors), the dialysis bags were then placed in 1 L of 150 mM NaCl buffered with 10 mM HEPES (pH 7.2) for 7 h, followed by 500 ml of RPMI 1640 for 16 h, and a fresh bath of 500 ml of RPMI 1640 for another 7 h. In the second experiment, also with two different donors, the supernatants were loaded with iron, as described above, and then dialyzed sequentially against two changes of 4 L of 150 mM NaCl with 10 mM HEPES (pH 7.2) for 8 and 15 h each, followed with two changes of 500 ml of RPMI 1640 for 10 and 20 h each.

Frozen PMN pellets (1–5 × 10⁶ cells) were suspended in 1 ml of ice-cold HBSS with 0.2% Triton X-100 and sonicated with 25 × 1-s pulses using a microtip sonicator (Branson).

LF levels were determined with the OxisResearch LactofEIA kit, according to the manufacturer’s instructions. Elastase content of PMN lysates was determined by ELISA (Cell Sciences).

A fluorescence microplate assay of fungal growth using CMFDA has been published (23) and was modified as follows. Into a 96-well white polypropylene plate (Whatman UNIPLATE), 10⁵ conidia were added per well in 50 μl of RPMI 1640 buffered with 25 mM HEPES (pH 7.2–7.4). Proteins were diluted in HBSS⁻ and added at 1/10 vol with appropriate buffer controls. The final assay volume of 100 μl was made up with RPMI 1640 or PMN supernatants, depending on the experiment. After 16–18 h of incubation at 37°C/5% CO₂, 100 μl of 5 μM CMFDA in a buffer containing 110 mM glucose, 10 mM NaCl, and 10 mM HEPES (pH 7.2–7.4) was added to each well, and the plate was incubated as before for an additional hour. Fluorescence was measured using an Anthos Zenyth 3100 plate reader with a 485-nm excitation filter and a 535-nm emission filter for 1 ms/well.

Resting A. fumigatus conidia undergo a rapid developmental and metabolic transformation when placed in culture medium such as RPMI 1640. Conidia incubated for 4 h at 37°C were swollen, but had not formed germ tubes. After 8 h of incubation, almost all conidia germinated and fungi were growing as unbranched hyphae. To measure metabolically active fungi, we used the colorimetric XTT assay and compared dye reduction in cultures grown in medium and plasma alone vs cultures exposed to different E:T ratios of PMN to fungus. Similar to previous studies, we found that after a 2-h incubation of hyphae with normal PMN, but not CGD PMN, the growth of Aspergillus was significantly decreased, indicating NADPH oxidase-dependent damage (Fig. 1,A). Surprisingly, both normal and CGD PMN inhibited conidial germination in a dose-dependent manner (Fig. 1 B). There was no significant difference between normal and CGD PMN. Time-lapse video microscopy demonstrated that the majority of conidia were engulfed by neutrophils within 15 min (data not shown). Whereas conidia cultured in the absence of PMN grew robustly, most conidia treated continuously with either normal or CGD PMN remained ungerminated for up to 17 h (data not shown).

PMN contain several distinct populations of membrane-bound storage compartments (granules) that are generated during differentiation in the bone marrow with primary (azurophilic) granules being made earlier than secondary (specific) granules (24). Substantial release of secondary granules with little release of primary granules can be induced by concentrations of PMA between 10 and 100 ng/ml (25). We found that PMA at 10 ng/ml induced a release of an anticonidial factor from PMN from both normal and CGD PMN (Fig. 2, A and B). There was no significant difference in the potency of PMA-induced postsecretory supernatants from CGD or normal PMN at high doses. In the absence of PMA treatment, however, some, but not all of the CGD donor PMN displayed significant unstimulated secretion suggestive of prior activation. Release of anticonidial activity from PMN was not detectable at 1 ng PMA/ml and appeared to be maximal at 10 ng PMA/ml (data not shown). Growth of conidia treated with PMA alone at doses up to 1000 ng PMA/ml was normal (data not shown). Whereas in the absence of PMN supernatants A. fumigatus conidia grown for 16 h form extensive branched hyphae (Fig. 2, C–E), the addition of the indicated dose of PMN supernatant resulted in significant inhibition of fungal growth. Most conidia germinated and displayed an altered morphology (Fig. 2,F) compared with the growth of conidia in RPMI 1640 alone (Fig. 2 E).

Supernatants obtained from incubation of normal and CGD PMN with PMA were similarly active against A. fumigatus conidia. To identify what components might be responsible for this activity, supernatants were resolved by PAGE and gels were silver stained. The spectrum of factors detected was similar between normal and CGD PMN supernatants in the absence or presence of PMA (Fig. 3,A). The most abundant protein released under these conditions had a molecular mass of ∼80 kDa, similar to that of LF, an abundant secondary granule protein of PMN (26). As shown in Fig. 3 B, Ab to human LF detected an ∼80-kDa band that also increased significantly in supernatants after PMA stimulation.

LF levels in PMA-treated PMN supernatants from normal controls and CGD patients were determined by ELISA. Whereas resting PMN secreted ∼2 μg LF/ml, PMA stimulation caused release of 24.4 ± 8.5 μg LF/ml and 21.7 ± 1.8 μg LF/ml, respectively, from normal and CGD PMN. There was no significant difference in the amount of LF secreted from PMN between the 3 CGD and the 10 normal donors studied (p > 0.05, t test). However, given the slightly lower levels of LF detected by ELISA in postsecretory supernatants from CGD PMN and the variable secretion of activity in the absence of PMA (Fig. 2 B), we measured LF content of 10 normal and 30 CGD PMN samples. CGD PMN contained ∼20% less LF than normal PMN (2822 ± 145 ng per million CGD PMN vs 3523 ± 364 ng of LF per million normal PMN (mean ± SEM)), and the difference was statistically significant by t test, p = 0.037. Interestingly, the primary granule protein elastase did not differ between normal and CGD PMN (8328 ± 606 ng/million normal PMN vs 8412 ± 395 ng/million CGD PMN, p = 0.914 by t test).

Although there are exceptions, the ability of LF to prevent microbial growth depends on high-affinity binding and sequestration of iron, an essential growth factor for all organisms. To test whether iron sequestration played a role in PMN supernatant activity against A. fumigatus conidia, we supplemented conidial cultures with horse spleen ferritin, a soluble protein-iron complex. Ferritin alone had no effect on the growth of conidia, as indicated by the CMFDA assay, but the addition of ferritin to cultures grown in the presence of PMN supernatant completely reversed their growth-inhibitory effect on conidia (Fig. 3,C). Dialysis of PMA-induced PMN supernatants against iron, as described in Materials and Methods, resulted in a significant decrease in their anticonidial activity (Fig. 3 D). Immunoassay of dialyzed supernatants for LF content demonstrated that there was no loss of LF during dialysis (data not shown).

To test whether LF at concentrations occurring in PMN supernatants could account for their activity against A. fumigatus conidia, purified human milk LF (which is encoded by the same gene as leukocyte LF) was tested alone and in the presence of ferritin. As shown in Fig. 4,A, a modest inhibitory activity against conidia was evident at concentrations of purified LF as low as 5 μg/ml with an IC₅₀ of ∼10 μg/ml. In contrast, in separate experiments (data not shown) hyphae exposed to LF doses up to 80 μg/ml were inhibited by <50%. This activity was significantly inhibited by the inclusion of 1 μg of ferritin/ml (at each dose except control and 2.5 μg LF/ml, a t test comparing activity in the presence or absence of ferritin resulted in p < 0.05). To further demonstrate that LF acts by depleting iron, samples of LF were dialyzed against 1 mM ferrous ammonium sulfate, as described in Materials and Methods. As shown in Fig. 4,B, iron loading of LF completely abolished its ability to inhibit A. fumigatus conidial growth. Commercially available iron-saturated LF was also inactive against conidia, but activity could be recovered by depleting iron by dialysis against 0.1 M citrate (data not shown). The growth-inhibitory effects of LF on Aspergillus were also shown microscopically (compare control cultures in Fig. 4,C with fungus cultured in the presence of LF in Fig. 4,D). As shown in the inset of Fig. 4,D, the altered morphology of Aspergillus grown in the presence of LF was similar to that seen in the presence of PMN supernatants (Fig. 2 F).

Although CGD patients have increased susceptibility to Aspergillus attributable to defective oxidative killing of fungal hyphae, most CGD patients tolerate daily exposure to Aspergillus conidia without infection, suggesting that primary defenses such as mucociliary removal of conidia from lungs and nonoxidative antifungal host defenses are usually sufficient to avoid infection. We examined nonoxidative antifungal activities by comparing the ability of normal and CGD PMN to inhibit the growth of A. fumigatus. We found that inhibition of Aspergillus conidial germination by CGD and normal PMN was indistinguishable, and was therefore independent of attack by reactive oxygen derivatives. We demonstrate that LF released by PMN can inhibit fungal growth, and does so by depriving conidia of iron. We hypothesize that fungistasis by the PMN and perhaps by mucosal LF may increase the likelihood that conidia are physically removed from airways by mucociliary flow before they germinate into hyphae.

LF has previously been shown to inhibit the growth of Candida albicans (27). LF has not been reported to act on A. fumigatus, although LF is released by PMN in response to conidia and binding of purified LF to conidia has been mentioned (28). LF is produced in large amounts by lacrimal and mammary glands and is present in their secretions at concentrations up to several mg/ml (29). LF is also produced by PMN precursors in the bone marrow (30) and is stored in the secondary granules of mature PMN (26). Plasma LF concentrations correlate with PMN numbers, suggesting that PMN is the main source of LF in the circulation (31, 32). In human nasal secretions, LF occurs at concentrations between 80 and 200 μg/ml (33). Further down the respiratory tract, the cellular source of airway LF remains unclear. However, LF levels in bronchial lavage, no doubt representing a dilution of the concentration in the actual bronchial fluid, are ∼12 μg/ml and were elevated further during bronchitis (34). The doses of LF required to inhibit A. fumigatus conidial growth in our assays (∼10 μg/ml) were within reported physiologic concentrations. Pathophysiologic increases in LF have been reported in sputum from cystic fibrosis patients as disease worsens (35), and concentrations of LF occurring in cystic fibrosis sputum (900 μg/ml) have been shown to significantly decrease the amount of antibiotic required to achieve inhibition of Burkholderia and Pseudomonas in vitro (36). LF appears to be well tolerated in human clinical trials and has been shown recently to enhance clearance of Helicobacter pylori infections in conjunction with antibiotics (37). Although aerosolized antifungals such as amphotericin B have been beneficial in decreasing the incidence of aspergillosis in immunosuppressed individuals, the overall case fatality rate has remained high. It is an intriguing possibility that aerosolized LF or other metal-chelating substances may be a useful alternative prophylaxis against Aspergillus (and possibly other pulmonary pathogens) in high-risk patients.

The absence of LF has not yet been reported to predispose to aspergillosis; however, this may be expected due to other intact antifungal defenses, in particular NADPH oxidase products. LF knockout mice have been made; however, host defense functions were not directly tested nor were spontaneous infections by Aspergillus reported (38). Human neutrophil-specific granule deficiency, caused by a mutation in the C/EBPε trancription factor (39), also results in decreased or absent neutrophil LF (among many other specific granule proteins) (40, 41), and is associated with an increased risk of pyogenic bacterial infections, but not fungal infection. Several single nucleotide polymorphisms in human LF have been reported and are associated with an increased risk for periodontitis possibly by altering the oral microflora (42, 43); however, it is unknown whether these polymorphisms are risk factors for aspergillosis. Despite the overall similarity between normal and CGD PMN and their postsecretory supernatants in arresting the growth of A. fumigatus conidia, we found that CGD PMN actually contained less LF than normal PMN. It is likely that the chronic inflammation and prolonged survival of PMN (44) seen in CGD patients result in release of PMN granules so that less LF remained cell associated. This phenomenon may also contribute to the susceptibility of CGD patients to aspergillosis. It would be interesting to see whether Aspergillus conidia more readily infect mice deficient both in LF and NADPH oxidase components, although, as mice deficient in gp91^phox (45) or p47^phox (46, 47) are already highly susceptible to this organism, these studies may be difficult to perform. For most patients with aspergillosis, the most common cause is prolonged pathogenic or iatrogenic neutropenia of <100–500 PMN/μl (∼10-fold less than the normal range) (4). Decreased resistance to Aspergillus seen in these individuals could be due both to reduced oxidative killing of hyphae as well as the absence of nonoxidative PMN-derived host defenses such as LF.

Aspergillus, like most organisms, is dependent upon iron for various metabolic processes. Microbial competition for scarce bioavailable iron has favored evolution of siderophores, small secreted molecules that bind extracellular iron with high affinity and return it to the microbe for uptake. A. fumigatus synthesizes at least five siderophores (48). Deletion of sidA, a gene involved in siderophore synthesis, results in attenuated virulence (49, 50), and iron acquisition appears to be required for growth in human serum (48). Consistent with the ability of hyphae to acquire iron by these mechanisms, additional studies showed that hyphae were significantly less sensitive to the inhibitory actions of LF. Distant from the site of infection, where the host deploys LF to deprive the microbe of iron and the microbe deploys siderophores to acquire it, the host also causes a systemic reduction in available iron as part of the general acute-phase inflammatory response (51). A small peptide, hepcidin (also called liver expressed antimicrobial peptide-1), was initially characterized as an antibacterial and antifungal peptide produced by the liver (52), but later shown to be a peptide hormone acting as a central regulator of mammalian iron homeostasis (53). Hepcidin itself is an acute-phase reactant (54), and substantially decreases plasma iron levels within 1–2 h of injection (as reviewed (53, 55)). Conversely, iron overload has been associated with increased risk for several fungal infections. In a group of patients that died after hemopoietic stem cell transplant, increased hepatic iron was associated with increased risk of invasive aspergillosis (6). Iron overload was also associated with fungal infection in liver transplant (5), and in hematologic malignancies (7). Interestingly, patients being treated for iron overload with desferroxamine, a siderophore made by Streptomyces, are at increased risk for mucormycosis (56). This may be explained by the observation that Rhizopus microsporus, one of the causes of mucormycosis, is able to acquire iron from the iron-chelated form of the bacterial siderophore, ferroxamine (57). Similarly, Saccharomyces cerevisiae is able to take up and use iron from ferroxamine (58). We also found that desferroxamine antagonized LF activity against A. fumigatus condia, suggesting that A. fumigatus may also be able to acquire iron by uptake of ferroxamine (K. A. Zarember and J. I. Gallim, unpublished observations).

The ability of LF to inhibit A. fumigatus conidial growth suggests an essential role for iron in fungal development. We are currently studying the transcriptional responses of Aspergillus hyphae and conidia to attack by normal and CGD PMN and have found differential regulation of many genes putatively involved in metal uptake (J. H. Sugui, K. A. Earember, H. S. Kim, W. C. Nierman, J. I. Gallim, and K. J. Kwon-Chung, unpublished observations). A better molecular understanding of how conidia recognize the low iron levels in the host and the biochemical methods used by Aspergillus to acquire iron from the host and other sources may well reveal excellent targets for the development of novel antifungal drugs.

We thank Steven M. Holland and Harry L. Malech for helpful discussions and enabling access to their patients’ blood; Douglas Kuhns for performing the ELISAs on PMN pellets; and the staff of the National Institutes of Health Clinical Center Department of Transfusion Medicine, especially Amy Melpolder and Cynthia Matthews, for providing normal donors. We are grateful to Owen Schwartz, Meggan Czapiga, and Juraj Kabat for help with microscopy. Finally, we thank the CGD patients and the clinical staff that care for them for essential contributions to this study.

The authors have no financial conflict of interest.