Identification of defensin-1, defensin-2, and CAP37/azurocidin as T-cell chemoattractant proteins released from interleukin-8-stimulated neutrophils.

Reports that interleukin-8 (IL-8) induces the infiltration of neutrophils followed by T-cells into injection sites led us to postulate that by stimulation of neutrophil degranulation IL-8 may cause the release of factors with chemoattractant activity for T-lymphocytes. Extracts of human neutrophil granules were chromatographed to isolate and purify T-lymphocyte chemoattractant factors. Two major peaks of T-cell chemotactic activity were purified by C18 reversed phase high pressure liquid chromatography (HPLC). The first peak was resolved further by C4 reversed phase HPLC and yielded an active fraction shown by NH-terminal amino acid sequence analysis to contain defensins HNP-1, HNP-2, and HNP-3. Purified defensins HNP-1 and HNP-2 (kindly provided by Dr. R. I. Lehrer, UCLA) were also potent chemoattractants for human T-cells, while HNP-3 was inactive. The second peak of T-cell chemoattractant activity was also further purified to homogeneity by C4 reversed phase HPLC and identified by NH-terminal sequence analysis as CAP37/azurocidin, a protein with sequence homology to serine proteases. 0.1-100 ng of defensins and 1.0-100 ng/ml CAP37 were able to stimulate in vitro T-cell chemotaxis. Neutrophil activating factors, i.e. IL-8, phorbol 12-myristate 13-acetate/ionomycin, and formylmethionylleucylphenylalanine each induced the release of CAP37 and defensins from neutrophil granules. Subcutaneous administration of defensins or CAP37/azurocidin into BALB/c mice resulted in a moderate neutrophil and mononuclear cell infiltrate by 4 h, which was greater by 24 h at the site of injection. Additionally, subcutaneous injection of defensins into chimeric huPBL-SCID mice resulted in significant infiltration by human CD3+ cells within 4 h. These results identify the antimicrobial proteins, CAP37/azurocidin and defensins HNP-1 and HNP-2, as potent neutrophil-derived chemoattractants for T-cells. These proteins represent primordial antimicrobial peptides which may have evolved into acute inflammatory cell-derived signals that mobilize immunocompetent T-cells and other inflammatory cells.

IL-8 has also been reported to be an in vitro and in vivo chemoattractant for T-cells (16). Furthermore, we have demonstrated that subcutaneous injection of human IL-8 into human T-cell-engrafted SCID mice causes an initial infiltration of murine neutrophils by 4 h into the site of injection, which is followed by a marked infiltration of human T-cells by 72 h (17). This delayed infiltration of human T-cells suggested that IL-8 may recruit T-cells indirectly, perhaps by stimulating the release of T-cell chemotactic factors by activated neutrophils. This hypothesis was tested by incubating purified peripheral blood neutrophils with a number of degranulating agents, including IL-8, and resulted in the release of T-cell chemotactic activity into the supernatants within 2 h. This led us to detect, purify, and identify several T-cell chemotactic factors from extracts of neutrophil granules.

MATERIALS AND METHODS
Reagents-Rabbit anti-human CAP37 antisera was prepared by immunization with purified CAP37/azurocidin in Freund's adjuvant (both the CAP37 and antisera were kindly provided by Dr. A. Pereira of the University of Oklahoma). Purified defensins and rabbit anti-defensin antisera were kind gifts from Dr. R. I. Lehrer, UCLA.
T-cell Preparations-Human peripheral blood enriched in mononuclear cells or lymphocytes was obtained from normal donors by leukapheresis (National Institutes of Health Clinical Center, Department of Transfusion Medicine, Bethesda, MD). The blood was centrifuged through Ficoll-Hypaque at 800 ϫ g for 30 min. The peripheral blood mononuclear cells at the interface were washed twice with PBS and centrifuged through an isosmotic Percoll gradient as described to remove monocytes. Human T-cell enrichment columns (R&D Systems, Minneapolis, MN) were then used according to the manufacturer's instructions to rapidly purify the T-cells using high affinity negative selection. This isolation procedure typically yields greater than 90% * This project has been funded in part with funds from the Department of Health and Human Services under Contract NO1-CO-74102 with Program Resources, Inc. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed.
CD3ϩ T-cells. The cells were resuspended in chemotaxis medium (RPMI 1640 containing 1% bovine serum albumin and 25 mM HEPES). Chemotaxis Assay-T-lymphocyte migration was assessed using a 48-well microchemotaxis chamber (NeuroProbe, Cabin John, MD) as reported (18). 25 l of the sample to be tested (diluted in chemotaxis medium) was placed in the lower compartment, and 50 l of cell suspension (at 5 ϫ 10 6 cells ml Ϫ1 ) was placed in the upper compartment. The two compartments were separated by a polycarbonate filter (5-m pore size) coated with 10 g/ml collagen type IV overnight at 4°C. The apparatus was incubated at 37°C for 3 h in humidified air with 5% CO 2 . At the end of the incubation period the filter was removed, fixed, and stained with LeukoStat (Fisher). The number of cells per high power field that migrated through the filter were counted by light microscopy. The results are expressed as the mean Ϯ S.D. value of the migration in triplicate samples and are representative of at least three experiments. The statistical significance of the number of cells migrating in response to stimuli versus control medium was calculated using the Student's t test.
Neutrophil Granule Preparations-Human neutrophils were isolated from granulocyte packs obtained from the Department of Transfusion Medicine. Briefly, erythrocytes were removed by sedimentation with 1.5% dextran. Mononuclear cells were centrifuged on a Ficoll-Hypaque cushion, and the residual erythrocytes were removed by hypotonic lysis. The neutrophils were resuspended in PBS, counted, and assessed for viability. Following a wash with PBS, the neutrophils were resuspended in disruption buffer (0.25 M sucrose and 10 mM HEPES, pH 7.4, containing 4 mM EGTA). Neutrophils were lysed by nitrogen cavitation, and granule fractions were separated on Percoll gradients (19). The cells were equilibrated at 450 p.s.i. for 20 min in a nitrogen bomb and lysed by dropwise release. Nuclei and cell debris were removed by centrifugation at 600 ϫ g for 10 min. The supernatant was layered over 48% Percoll and centrifuged at 29,000 rpm for 26 min. The Percoll gradient was fractionated, and the fractions were assessed for granule enzymes and T-cell chemotactic activity. Granule fractions were pooled and recentrifuged at 35,000 rpm for 3 h. The granules were recovered as a white flocculent material just above the Percoll pellet. At this point aliquots of the granule-containing fraction were made and stored at Ϫ70°C until further use.
Purification of T-cell Chemoattractant Activity-Neutrophil granules (0.5 ml) were lysed by freeze-thawing (three times) in 3 ml of 0.1% trifluoroacetic acid containing 2 M NaCl. The suspension was centrifuged for 15 min at 800 ϫ g. 1 ml of clarified supernatant was applied to a C18 -300A reversed phase HPLC cartridge column (8 ϫ 100 mm, Delta-Pak, Millipore), equilibrated with 0.1% trifluoroacetic acid (Buffer A). Absorbed proteins were eluted with a linear gradient of 0 -90% Buffer B (0.05% trifluoroacetic acid in acetonitrile) over 60 min at a flow rate of 1 ml min Ϫ1 (Fig. 1). Absorbance at 206 nm was monitored as a measure of protein concentration. Collected fractions were lyophilized and assayed for chemotactic activity.
Identification of Chemoattractant Proteins-Fractions containing chemoattractant activity were pooled, lyophilized, redissolved in Buffer A, and applied to a C4 reversed phase HPLC column (4.6 ϫ 250 mm, Bio-Rad, RP-304). Proteins were eluted with a linear gradient of 0 -90% Buffer B over 60 min at a flow rate of 0.5 ml min Ϫ1 (Figs. 2 and 3). Fractions with activity were analyzed by SDS-PAGE (20) and used for automatic sequence analysis on an Applied Biosystems sequencer model 477A. The presence of cysteine residues was inferred from the absence of detectable amino acid.
Release of Granule Contents-Neutrophils (2 ϫ 10 6 ml Ϫ1 ) suspended in AIM-V medium (Life Technologies, Inc.) were incubated at 37°C in 6-well plates that had been coated with fibronectin or vascular-cell adhesion molecule-1. Exocytosis was initiated by addition of the appropriate stimulus and stopped by centrifugation and removal of the supernatant. Lactoferrin and ␤-glucuronidase were assayed (20,21) and detected to demonstrate that IL-8 was effective in degranulating the neutrophils. Protein in the cell-free medium was concentrated using StrataClean resin (Stratagene, La Jolla, CA), separated by SDS-PAGE, and Western blotted using a rabbit anti-defensin antisera. In order to quantitate these results the blots were scanned and analyzed using the NIH Image 1.58f program.
In Vivo Studies-BALB/c and CB-17 scid/scid (SCID) mice were obtained from the Animal Production Area (NCI, Frederick Cancer Research and Development Center). Mice were used at 8 -12 weeks of age and kept in pathogen-free conditions as described (22). SCID mice were treated with anti-asialoganglioside-M1 and injected with 1 ϫ 10 8 huPBL intraperitoneally as described (22). Immediately thereafter 0.1 ml of PBS containing our purified preparations of 1.0 g of defensins, CAP37/azurocidin, or control PBS were injected daily one time subcu-taneously into the same injection site. Chromogenic Limulus amebocyte lysate assays for endotoxin contamination in our purified peptides revealed 0.05 endotoxin unit/20 g of defensins and 0.114 endotoxin unit/20 g of CAP37/azurocidin. The injection site was examined histologically either at 4 h after the first injection or 24 h later 4 h after a second injection. Experiments were performed on three to four mice/ group in duplicate. Histological and immunohistochemical examinations were performed as described previously (22).

Identification of Defensins as T-cell Chemoattractants-Hu-
man neutrophil granule proteins were extracted by repeated freeze-thawing in a solution of high salt and acid pH. These granule extracts yielded a T-cell chemoattractant activity that was concentrated and partially purified by C18 reversed phase HPLC chromatography. This resulted in two major peaks of chemotactic activity (Fig. 1). The proteins with chemotactic activities were further resolved by chromatography on a C4 reversed phase HPLC column. The first peak of activity was purified to apparent homogeneity ( Fig. 2), appearing as a single band that migrated with an apparent molecular mass of approximately 3 kDa on Tris/Tricine SDS-PAGE (23). NH 2 -terminal sequence analysis of the preparation revealed the presence of three related sequences (Table I) corresponding to defensin HNP-1 (50% of the material), HNP-2 (30%), and HNP-3 (20%). The fact that 1) we detected only defensin sequences by sequence analysis, 2) no impurities were detected by silver staining, and 3) amino acid analysis of our preparation corresponds well with amino acid composition of defensins (data not shown) suggests that our preparation consisted of at least 95% pure defensins.
The T-cell chemotactic activity of homogeneous preparations of neutrophil-derived HNP-1, HNP-2 and HNP-3 (kindly provided by Dr. R. I. Lehrer, UCLA) were compared with our preparation. The dose response to individual HNP-1 and HNP-2 was practically identical to that of our preparation containing a mixture of HNP-1, HNP-2, and HNP-3 (Fig. 3). Checkerboard analysis of our HNP-1 and HNP-2 showed that their effect was chemotactic rather than chemokinetic (data not shown). HNP-3, which differs from HNP-1 only by substitution of aspartate for the amino-terminal alanine, was not able to stimulate T-cell chemotaxis. It has been proposed that HNP-3 is a precursor of HNP-2 (24) and proteolytic removal of the amino-terminal aspartate residue converts the inactive HNP-3 into chemotactically active HNP-2 (24). While serum proteins have been reported to inhibit the antimicrobial activity of defensins (24), both HNP-1 and HNP-2 retained in vitro chemo- tactic activity for human T-cells in the presence of 10% human serum (data not shown).
Identification of CAP37/Azurocidin as T-cell Chemoattractants-The second major peak of T-cell chemotactic activity ( Fig. 1) was also further purified using the same C4 reversed phase HPLC column. In this case T-cell chemotactic activity co-eluted with a major protein peak corresponding to a homogeneous 30-kDa band on Tris/glycine SDS-PAGE (Fig. 4). Amino-terminal sequence analysis identified this protein as CAP37 (cationic antimicrobial protein of molecular mass 37 kDa), also known as azurocidin (Table I). By amino acid analysis, our preparation of CAP37/azurocidin corresponds well with the known amino acid composition of CAP37/azurocidin. In addition, upon sequence analysis of our preparation we detected no other sequence except CAP37/azurocidin. These facts indicate that our preparation must be at least 95% pure. CAP37/azurocidin is a previously identified azurophilic granule protein with sequence homology to the serine protease family and has potent oxygen-independent bactericidal activity (25,26). Sequencing of the polypeptide eluting just before CAP37/azurocidin revealed it to be a lysozyme that does not induce T-cell chemotaxis. CAP37 was chemotactic from 1.0 to 100 ng/ml, which is comparable in molar concentration with the range of defensin activity (Fig. 3). In addition, preparations of purified as well as recombinant human CAP37/azurocidin (kindly provided to us by Dr. A. Pereira) were equally potent chemoattractants for human T-cells (data not shown).
IL-8-stimulated Release of Defensins from Neutrophils-Experiments were performed to ascertain that defensins and CAP37/azurocidin would be released into the extracellular environment by appropriate stimuli. Neutrophils exposed to IL-8 could be shown to degranulate with the release of lactoferrin and ␤-glucuronidase (27). The release of defensins following degranulating stimuli was examined by concentration of the culture supernatants and quantitative detection of the defensins released using a Western blot assay (Fig. 5) (28). We established by scanning of the Western blots that IL-8 released about 10% of the total content of defensins in the granules of 2 ϫ 10 6 neutrophils, while FMLP released 17% and phorbol 12-myristate 13-acetate/ionomycin 23% of the total content of defensin in the granules. The addition of cytochalasin increased the degranulating effect of IL-8 only minimally to 13% of the total and did not effect FMLP release. Tumor necrosis factor-␣ has also been shown to have a synergistic effect on IL-8-stimulated degranulation (27). Similarly, as determined by Western blotting, IL-8 also induced the release of CAP37/ azurocidin, which is also contained in the azurophilic granules (data not shown).
In Vivo Inflammatory Effects of Human Defensins and CAP37/Azurocidin in Mice-BALB/c mice were injected subcutaneously with 1 g of the defensin or CAP37/azurocidin preparations. After 4 or 24 h the injection site was excised, and the extent and types of cells infiltrating the site were examined histologically. Within 4 h of injection the defensin and CAP37/ azurocidin each resulted in infiltration by PMNs and mononuclear cells in the dermis and subcutaneous tissues (Table II). This is in contrast to a single 1-g injection of recombinant human IL-8, which produced a marked infiltration of PMNs by 4 h with little mononuclear cell infiltration (17). By 24 h, 4 h after a second injection at the same site, an even greater infiltration of PMNs and mononuclear cells was elicited by defensins as well as CAP37/azurocidin. Doses of 10 and 100 ng of defensins and CAP37 also induced neutrophil and mononuclear cell infiltrates at 24 h. Thus, human defensins and CAP37/azurocidin are capable of inducing considerable local murine neutrophil and mononuclear cell infiltration. Immunohistochemical studies of sites of defensin injection in chimeric huPBL-SCID mice were performed to establish whether there were T-cells in the infiltrate (22). A single injection of 1 g of defensins (HNP-1 and HNP-2) within 4 h resulted in the infiltration by low to modest numbers of human CD3ϩ T-lymphocytes in four of six mice examined (Table III). In contrast, the sites of PBS injection in six control mice did not contain any human CD3ϩ T-cells. These in vivo results support the in vitro evidence that defensins are T-cell chemoattractants. DISCUSSION The results presented in this report demonstrate that neutrophils are an important source of factors chemotactic for T-lymphocytes. The polypeptides derived from neutrophil granules were identified as T-cell chemotactic factors: defensins NP-1 and -2 and CAP37/azurocidin. Defensins are a family of small (29 -30 amino acids, approximately 3.5 kDa) cationic antimicrobial proteins (24) whose structure is stabilized by three conserved disulfide bridges, including one disulfide that cyclizes the protein by linking the carboxyl-terminal cysteine to the amino-terminal ultimate (HNP-2) or penultimate (HNP-1, HNP-3) cysteine (29). Indeed the defensins HNP-1, HNP-2 and HNP-3 differ by only a single NH 2 -terminal amino acid. Ap-FIG. 2. Purification of defensins by C4 reversed phase chromatography. The first peak of activity from the C18 column was lyophilized, loaded onto a C4 reversed phase HPLC column and eluted with a gradient of acetonitrile (---). Fractions were tested for activity using the microchamber assay. The active fraction, indicated by an open rectangle beneath the absorbance trace, showing one peptide band on gradient (10 -20%) Tris/Tricine gel with Coomassie staining (see inset) was sequenced and identified as HNP-1, HNP-2, and HNP-3.

TABLE I Identification of the T-cell chemotactic proteins by amino-terminal sequences analysis
The complete sequences of HNP-1, -2, and -3 are shown with the amino-terminal sequences obtained from the chemotactic peptides underlined. The presence of cysteine residues was inferred from the absence of detectable amino acid in that cycle. The N-terminal sequence of CAP37/ azurocidin is shown with the obtained sequence underlined.  25-30% of the human azurophilic granule protein consists of defensins. It has been estimated that up to 3-5 pg of defensins can be released by each neutrophil. They are believed to be released into the phagocytic vesicle of the neutrophil and into the medium (30) where they contribute to the respiratory burst-independent cytotoxic killing of microbes. The cytotoxic effect of defensins occurs at micromolar concentrations and is thought to be due to their ability to form numerous transmembrane channels that permeabilize lipid bilayers of microorganisms. Even though their amino acid sequences are virtually identical, HNP-3 does not kill Candida albicans (31) whereas HNP-1 and HNP-2 have been reported to have this activity. HNP-3 also lacked T-cell chemotactic activity, but cleavage of the amino-terminal aspartate residue can convert it into active HNP-2.
CAP37/azurocidin, like defensins, was previously identified as a neutrophil granule protein also with antimicrobial activity at micromolar concentrations (32). Structurally, CAP37 belongs to the serine protease superfamily and has 45% sequence identity with human neutrophil elastase; however, due to replacements of crucial amino acids at the active site, CAP37/ azurocidin is inactive as a protease (26). It is unlikely that there are any other T-cell chemotactic peptides present that we have not detected in the neutrophil granule extracts because the defensins and CAP37 are major protein constituents of neutrophil granules and are chemotactic at low (nanomolar) concentrations.
Both the defensins HNP-1 and HNP-2 and CAP37/azurocidin have been reported to chemoattract monocytes (35,36). How- The presence of infiltrating mast cells was noted. ever, we have repeatedly been unable to detect any monocyte or neutrophil chemotaxis with doses of 1-1000 ng/ml defensins. The report that defensins chemoattract human monocytes is based on assays of the leading front of mononuclear cells migrating into a filter (35). It is difficult to identify cells lodging in a filter, and this may have led to a misidentification of the migrating mononuclear cell type. Our assay method more readily permits the identification of cells adhering to the underside of polycarbonate filters as T-cells. In the case of our preparations of natural CAP37/azurocidin and the purified and recombinant CAP37 (provided us by Dr. A. Pereira), we also could not detect significant monocyte chemotaxis (chemotaxis index Յ 1.5). Perhaps this is based on technical discrepancies in our assays, or possibly tissue macrophages express more receptors for CAP37 than do human peripheral monocytes. However, we have identified another protein present in the chromatography fractions of the neutrophil granule extracts as having monocyte chemoattractant activity but not T-cell chemotactic activity. This protein migrates very closely to CAP37/ azurocidin and may therefore have been a contaminant of CAP37 preparations that attract monocytes. 2 Our data indicate that defensins and CAP37/azurocidin are potent in vitro T-cell chemoattractants.
Because defensins account for 25-30% of the neutrophil granule protein content, they may serve as a major source of T-cell chemoattractant whose release is under the control of neutrophil activators such as IL-8. Although serum proteins effectively inactivate the antimicrobial activity of defensins, their in vitro chemotactic activity is not impaired by the presence of 10% fetal calf serum (data not shown). This predicts that defensins can deliver a T-lymphocyte activating signal at a distance from the site of acute inflammation where they originate. This prediction was borne out by our experiments showing that BALB/c mice injected subcutaneously with 1.0-g injections of defensins or CAP37/azurocidin developed mononuclear as well as neutrophilic infiltrates by 4 h. By 24 h the number of accumulated neutrophils and mononuclear cells was markedly increased even in response to lower 10-or 100-ng doses. The infiltration by neutrophils and mononuclear cells was not predicted by the in vitro assays. It is possible that defensins and CAP37/azurocidin initiate a cascade of proin-flammatory signals in vivo resulting in the recruitment of neutrophils and monocytes. The inflammatory response may be a result of the tissue-damaging effects of these microbicidal agents, although necrosis became evident only at 24 h in response to the 1000-ng dose. In the case of the defensins, immunohistochemical studies revealed that in four of the six chimeric huPBL-SCID mice, CD3ϩ human T-cells appeared at the injection site already by 4 h. However, there are also unstained infiltrating mononuclear cells which are murine macrophages. These macrophages may express more receptors for human defensins and CAP37/azurocidin. The possibility that these agents are immunomodulators and may have adjuvant activity is currently being tested.
Our quantitative assays of the Western blots revealed that IL-8, FMLP, and phorbol 12-myristate 13-acetate/ionomycin resulted in comparable release of defensins by neutrophils in vitro. On the presumption that IL-8 induces similar degranulating effects in vivo, we therefore estimate that 2 ϫ 10 6 neutrophils could release 600 ng of defensins in response to IL-8. This would produce a concentration gradient sufficient to attract T-lymphocytes. Preincubation of neutrophils with tumor necrosis factor-␣ has also been shown to augment the IL-8stimulated degranulation of both azurophilic and specific granules of neutrophils (27), suggesting that secondary inflammatory stimuli may have a priming effect on IL-8-stimulated degranulation and can thus enhance leukocyte recruitment to sites of inflammation.
The physiological importance of these neutrophil granule T-lymphocyte chemoattractants remains to be established. Tantalizing clues are provided by an experiment of nature namely beige mutant mice with the Chédiak-Steinbrinck-Higashi Syndrome. Beige mice have also been reported to be deficient in defensins (37,38). An even less common disorder, specific granule deficiency, is also characterized by a deficiency in defensins and increased susceptibility to microbial infections (39). The defective host defenses in these mice may be due in part to deficient secretion of defensins and CAP37/azurocidin. Our results indicate that the release of these proteins from neutrophils can be triggered by proinflammatory mediators, such as IL-8, in addition to the release caused by phagocytosis of microorganisms (30,35). The possibility that these potent T-lymphocyte chemoattractants can provide a means by which acute inflammatory reactions develop into immunologically mediated chronic inflammatory responses dominated by mononuclear cells is supported by data showing that depletion of neutrophils inhibits subsequent chronic inflammatory reactions in mice (17).  Table II. Immunohistology was performed on skin sections 4 h after huPBL injection using antibodies to human CD3ϩ. Injection sites were evaluated microscopically without knowledge of the experimental treatment. Immunostaining was graded from minimal (1ϩ) to extensive (4ϩ) based on the intensity and distribution of positive cells.