Inflammatory mechanisms are critical in the arterial response to injury. Both IL-1 and the naturally occurring inhibitor of IL-1, IL-1R antagonist (IL-1ra), are expressed in the arterial wall, and in particular in the endothelium. Previous studies suggest that endothelial cells only make the intracellular type I isoform of IL-1ra (icIL-1ra1), an isoform known to lack a secretory signal peptide. It is unclear how icIL-1ra is released from the endothelial cell to act as an antagonist on cell surface IL-1 type I receptors. IL-1β, which also lacks a secretory signal peptide, may be released by ATP stimulation of the P2X7R. Therefore, we examined whether icIL-1ra1 release occurs in an analogous manner, using both the mouse macrophage cell line RAW264.7 and HUVECs. P2X7R activation caused icIL-1ra1 release from LPS-primed RAW264.7 macrophages and from HUVECs. This release was inhibited in the absence of extracellular calcium, and attenuated by preincubation with oxidized ATP, KN62, and apyrase. Endogenous ATP release, which also facilitated release of icIL-1ra1, was detected during LPS treatment of both RAW264.7 macrophages and HUVECs. Annexin V assays showed that ATP stimulation resulted in a rapid phosphatidylserine (PS) exposure on the cell surface of RAW264.7 macrophages, and that PS-exposed microvesicles contained icIL-1ra1. However, PS flip and microvesicle shedding was not apparent in ATP-treated HUVECs. These data support a general role for the P2X7R in the release of leaderless cytokines into the extracellular medium, and indicate how icIL-1ra1 may act upon its extracellular target, the IL-1R.
The IL-1 family of cytokines are well-established agents in inflammation. The agonists IL-1α and -β signal proinflammatory events via the type I IL-1R (IL-1RI).3 The IL-1 agonist signaling is regulated by two inhibitory mechanisms; a nonsignaling or decoy receptor (type II IL-1R), as well as a naturally occurring inhibitory form of IL-1, the IL-1R antagonist (IL-1ra). The IL-1ra protein competitively inhibits binding of IL-1α and IL-1β to IL-1RI. However, IL-1ra does not activate the receptor, as its binding prevents recruitment of the accessory protein to the receptor complex (1). There is evidence that the balance between IL-1 and IL-1ra determines the overall inflammatory response (1).
The arterial response to injury hypothesis describes an inflammatory response in the arterial wall either to direct injury, which has the clinical equivalent in the response to angioplasty, or to chemical agents such as oxidized low density lipoprotein or homocysteine, which has clinical equivalence in atherosclerosis. Central to this hypothesis is endothelial cell (EC) activation, with up-regulation of adhesion molecule expression facilitating leukocyte recruitment (2).
IL-1 has been implicated in the arterial response to injury and in atherogenesis per se. Atherosclerotic coronary arteries express IL-1 (3) and IL-1ra (4) with significant levels within the endothelium. IL-1 is a smooth muscle cell mitogen (5), which is dependent upon autocrine platelet-derived growth factor production. IL-1 administration to the arterial wall induces neointima formation, whereas IL-1ra administration to apoE−/− mice reduces atherosclerosis (6), and the IL-1β−/− mouse, when crossed with the apoE−/− mouse, has reduced levels of atherosclerosis (7).
Experimental studies of cultured ECs have indicated that upon stimulation ECs synthesize both IL-1β and IL-1ra (4). IL-1ra is known to have a number of different isoforms; the IL-1ra gene generates a number of splice variants that fall into two patterns, the secreted isoform (sIL-1ra) and the intracellular isoform (icIL-1ra), the latter having a number of subtypes, all devoid of a signal peptide. The 18-kDa type 1 intracellular IL-1ra (icIL-1ra1) protein is generated by alternative splicing from an upstream exon of sIL-1ra, being expressed in keratinocytes, ECs, and monocytes (8). Expression of the 25-kDa type 2 icIL-1ra protein has not been detected in human cells (9). Whereas the 15-kDa isoform (type 3 icIL-1ra), which may be generated by alternative splicing as well as by alternative translation initiation, is found in monocytes, macrophages, neutrophils, and hepatocytes (10, 11).
ECs do not make sIL-1ra that contains a signal peptide; these cells express icIL-1ra1, devoid of a signal peptide, leaving the question of how this molecule is released to gain access to its likely target, IL-1RI, unresolved.
Many studies have demonstrated that the P2X7R is responsible for ATP-mediated IL-1β release from activated monocytes, macrophages, and microglia (12, 13, 14, 15, 16, 17, 18, 19, 20, 21); monocytes from P2X7R knockout mice show markedly impaired LPS-mediated IL-1 release (22) and exhibit attenuated inflammatory responses (23). Recently it was shown that the leaderless cytokine IL-1β is released from LPS-stimulated monocytes via a P2X7R-dependent microvesicle-shedding mechanism (24). Activation of the P2X7R by ATP leads to a rapid exposure of phosphatidylserine (PS) upon the cell surface, followed by microvesicle shedding. These small vesicles were found to contain the processed and biologically active mature form of IL-1β (24). These results suggest that microvesicle shedding may represent a general pathway for the release of cytoplasmic proteins that lack leader sequences. Therefore, we sought to determine whether icIL-1ra1 is also released into the extracellular medium via activation of the P2X7R. This would provide an insight into the cellular mechanisms used to balance IL-1 and IL-1ra levels. The release mechanism was analyzed using both transfected icIL-1ra1 in RAW264.7 cells, as a model macrophage system. In addition, we examined P2X7R dependence of icIL-1ra1 release from cultured HUVECs. The results indicate that in both cases release of the intracellular form of the IL-1R antagonist can be regulated by P2X7R activation, however, there are significant differences between the two cell types.
Materials and Methods
Cells and solutions
Murine RAW264.7 macrophages were purchased from American Type Culture Collection (Manassas, VA) and cultured in DMEM supplemented with 2 mM glutamine, 10% heat-inactivated FCS (BioWhittaker, Walkersville, MD), 100 U/ml penicillin, and 100 μg/ml streptomycin.
6 cells per well (∼90% confluence). Transfection efficiency was estimated using enhanced GFP (EGFP) transfection by counting the percentage of green cells in three separate fields, under ×20 objective by fluorescence microscopy. The icIL-1raI was expressed with a myc epitope tag, using a construct made with icIL-1ra1 fused to an EQKLISEEDL tag at the C terminus, in pcDNA3.1−. Cells were treated with LPS and P2X7R stimuli at 24 h posttransfection.
LPS (recombinant source from Escherichia coli, serotype 0111:B4; Sigma-Aldrich) incubations were performed at 37°C in normal growth medium at a concentration of 1 μg/ml for 3 h. Cells were then washed with PBS, and 3′-O-(4-benzoyl)benzyl-ATP (Bz-ATP; Sigma-Aldrich) was added in extracellular solution, consisting of 147 mM NMDG, 2 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 10 mM HEPES, 12 mM glucose, for 20 min unless otherwise stated, at 37°C. The P2X7R agonist Bz-ATP was used at 100 μM on RAW264.7 cells or 300 μM on HUVECs. Supernatants were collected, cells were washed with PBS, and then were extracted by scraping in PBS plus 1% Triton X-100 and protease inhibitors (Cocktail III; Calbiochem, San Diego, CA). The ATPase, apyrase (Sigma-Aldrich), was incubated at 20 U/ml. The biologically inactive ATP analog, periodate-oxidized ATP (Ox-ATP; Sigma-Aldrich) was at 500 μM. The isoquinoline KN62 (calmodulin kinase II and P2X7 inhibitor; Calbiochem) was incubated at 10 μM. These inhibitors were present during both the 3-h LPS incubation and the Bz-ATP stimulation. Staurosporin (Calbiochem) treatment was at 1 μM for 18 h at 37°C.
Measurement of IL-1ra release by immunoblotting
Supernatants from cell incubations were concentrated by centrifugation using Nano-sep 10-kDa cut-off filters (Flowgen, Leicestershire, U.K.), or by annexin biotin pull down (see below). These and cell extracts were analyzed by immunoblotting after electrophoresis on a 12% SDS polyacrylamide gel and transferred onto polyvinylidene difluoride membrane. The myc epitope was detected using a c-Myc Ab (9E10 clone; Santa Cruz Biotechnology, Santa Cruz, CA) at 1/1000 dilution in PBS with 0.1% Tween 20, 2% nonfat milk. Detection of IL-1ra used goat anti-human IL-1ra (AF-280-NA, R&D) at 1/1000 dilution. Band intensities from immunoblots were quantified using NIH Image analysis, and, where data were pooled, the maximal band integrated intensity was taken as the 100% value.
Measurement of ATP levels
Levels of ATP in the supernatants were determined using a luciferase-based ATP determination kit (A-22066; Molecular Probes, Eugene, OR) according to the manufacturer’s instructions.
Annexin V binding
PS exposure on the cell surface was measured by microscopy using annexin VFITC, or by annexin biotin pull-down experiments with streptavidin agarose, as described previously (24). Beads were allowed to settle by gravity to isolate PS-exposed microvesicles.
Lactate dehydrogenase (LDH) assay
LDH activity in supernatants or 1% Triton X-100-lysed cells was determined by measuring the change in absorbance over time at 340 nm using an LDH kit (Sigma-Aldrich) according to the manufacturer’s instructions.
P2X7R-dependent icIL-1ra1 release from the RAW264.7 macrophage cell line
To determine whether icIL-1ra1 is released via a P2X7R-dependent mechanism, we used RAW264.7 macrophages as a transfectable monocyte-derived cell line and model. The only Abs that are available to detect endogenous IL-1ra are unable to distinguish between the secreted and intracellular isoforms of this protein. Therefore, to distinguish the leaderless isoform of IL-1ra released into the extracellular medium following cellular stimulation, we tagged icIL-1ra1 with a myc epitope. Human icIL-1ra1-myc was transfected and expressed in RAW264.7 cells as shown by immunoblotting of cell extracts harvested 24 h posttransfection (Fig. 1⇓A). These cells were cultured under various conditions, and release of icIL-1ra1-myc into the extracellular medium was measured by immunoblotting of the supernatants after gel electrophoresis. Cells that had been pretreated for 3 h with LPS only or with LPS followed by Bz-ATP did release icIL-1ra1-myc into the supernatant; this release being inhibited in the presence of the ATP-degrading enzyme apyrase, inhibited by the absence of extracellular calcium, and partially blocked by the addition of Ox-ATP (Fig. 1⇓, B and C). This activation and inhibition profile indicates that the release of icIL-1ra1-myc is dependent upon activation of the P2X7R. The ability of LPS pretreatment alone to mediate release of icIL-1ra1-myc is likely due to LPS causing endogenous ATP release, as has been previously demonstrated in RAW264.7 cells (26). We also found that supernatant samples from cells treated with LPS contained detectable levels of ATP (using the luciferase assay; data not shown) compared with supernatants from unstimulated or apyrase-treated cells, where no ATP was present.
P2X7R-dependent microvesicle shedding by RAW264.7 cells
Release of the leaderless cytokine IL-1β from macrophages and monocytes has been shown to be via a microvesicle-shedding mechanism (24). The microvesicle shedding is preceded by a rapid exposure of PS on the outer surface of the plasma membrane, and also on the surface of the microvesicles containing the IL-1β protein. This may be a general mechanism of release of leaderless proteins, including icIL-1ra1, from the cell. Therefore, we tested, using annexin V conjugated to FITC, whether P2X7R activation in RAW264.7 macrophages also results in PS flip in the plasma membrane. Fig. 2⇓A shows RAW264.7 macrophages pretreated with LPS, then stimulated with Bz-ATP for various times, in the presence of annexin VFITC. Within 10–15 min, annexin VFITC was clearly seen bound to the cell surface indicating that PS exposure had occurred. Vesicle formation was observed at the cell surface from 10 to 15 min after ATP application (Fig. 2⇓A), and live microscopy showed vesicles detaching from the cell surface. Cells that were not stimulated with Bz-ATP did not show any significant annexin VFITC binding when incubated for up to 45 min (data not shown). To determine whether PS-exposed microvesicles containing icIL-1ra1 were released from these macrophages, we isolated microvesicles by pull-down assays using annexin V-coated beads (24). Supernatants from cells stimulated with LPS, or LPS and Bz-ATP, were incubated with annexin biotin linked to streptavidin beads. Fig. 2⇓B shows that the bound fraction in the pull-down assays contained icIL-1ra1-myc, if the protein had been released into the extracellular medium due to activation of the P2X7R, but that after LPS stimulation alone the amount was lower and not associated with PS-positive vesicles.
P2X7R-dependent icIL-1ra1 release from ECs
Only icIL-1ra1 has been detected in ECs from human coronary artery and human umbilical vein after treatment with LPS and PMA (4). Therefore, we determined whether icIL-1ra1 is released from ECs by a mechanism similar to that in the RAW264.7 macrophage model. Fig. 3⇓, A and B, shows results from HUVECs following various treatments, when supernatants were immunoblotted with IL-1ra Ab. LPS treatment, followed by Bz-ATP stimulation, resulted in a maximal release of IL-1ra into the extracellular medium. Significant release was also observed with LPS treatment alone and even in the absence of exogenous agonists (Fig. 2⇑B), this being inhibited in the absence of extracellular calcium, and attenuated in the presence of apyrase, Ox-ATP, or KN62. This profile of inhibition indicated that the release of IL-1ra was likely dependent upon activation of the P2X7R. The release of IL-1ra from LPS-treated and unstimulated cells was likely the result of the presence of endogenous ATP released from the HUVECs, as these released fractions were found to contain ATP when measured by luciferase assay (Fig. 3⇓C), and no IL-1ra was present in supernatant removed at time zero (Fig. 3⇓, A and B, left bar).
The results in Fig. 3⇑ show release of endogenous IL-1ra from HUVECs. However, use of the IL-1ra Ab meant that we were unable to distinguish what isoform of IL-1ra had been released into the extracellular medium. Therefore, HUVECs were transfected with an icIL-1ra1-myc construct to determine whether this specific leaderless isoform could be released from ECs. Detection of icIL-1ra1-myc was performed by immunoblotting of transfected HUVEC cells extracts (Fig. 4⇓A) and supernatants (Fig. 4⇓, B and C) with anti-myc Ab. Release of icIL-1ra1-myc occurred when cells were pretreated with LPS, then stimulated with Bz-ATP. This was inhibited in the absence of extracellular calcium, or by addition of apyrase or Ox-ATP. LPS treatment alone did lead to release of icIL-1ra1-myc from HUVECs, the levels varying significantly according to cell batch (data not shown).
P2X7R stimulation does not cause PS translocation in ECs
We determined whether P2X7R activation of HUVECs resulted in PS exposure, and whether PS-microvesicles contained icIL-1ra1-myc from transfected HUVECs. Fig. 5⇓A shows a typical example of HUVECs pretreated with LPS, then stimulated with Bz-ATP. No annexin VFITC binding was detected at 15 and 30 min after stimulation, or up to 60 min (data not shown). In contrast, cells treated with staurosporin for 18 h, then washed and incubated with annexin VFITC, did show binding (Fig. 5⇓B). Annexin biotin pull-down assays showed that icIL-1ra1-myc-containing supernatants from LPS and LPS + Bz-ATP-treated cells did not bind to annexin V-coated beads, being detected in the unbound fraction (Fig. 5⇓C).
P2X7R activation does not cause cell necrosis/lysis
To assess whether release of icIL-1ra1 from both RAW264.7 macrophages and HUVECs could have occurred as a result of cell lysis for all treatments, cells were observed for lysis by microscopy when each sample was taken, but no apparent necrotic or apoptotic appearance was noted. Supernatants were also measured for the presence of LDH, which is released upon cell lysis. Fig. 6⇓, A and B, shows LDH levels from RAW264.7 and HUVEC-released supernatants, respectively. None of the fractions contained significant levels of LDH, compared with control samples of cells stimulated with staurosporin, or cells lysed in 1% Triton X-100.
Our data demonstrate that icIL-1ra1 is released or secreted from cells by an extracellular ATP-dependent mechanism, in both the RAW264.7 macrophage cell model and in HUVECs. Transfection of myc-tagged icIL-1ra1 into both the mouse macrophages and HUVECs enabled us to distinguish between this isoform and sIL-1ra. The action of extracellular ATP in mediating this release is likely to be attributable to activation of the P2X7R in both cell types. The P2X7R is expressed and functional in RAW264.7 macrophages (27) and also in ECs (28, 29). Release of icIL-1ra1 from RAW264.7 macrophages and HUVECs was inhibited by the removal of extracellular calcium, and was significantly attenuated by incubation with Ox-ATP, KN62, and apyrase. This inhibition profile fits well with an action at the P2X7R. Therefore, we may conclude that ATP-stimulated release of icIL-1ra1 is most likely mediated by activation of the P2X7R.
Stimulation of the P2X7R on RAW264.7 macrophages resulted in PS translocation (PS flip), as demonstrated by annexin VFITC binding to the cell surface. By microscopy we observed the formation of microvesicles, which were able to bud off from the cell surface. The PS flip occurred within a relatively rapid time frame, within a few minutes of activation by ATP. As previously demonstrated by MacKenzie et al. (24), PS flip after brief (≥10 min) P2X7R activation in the human monocytic cell line THP-1 is reversible and not associated with cell death. In this study, we did not detect any LDH release from either ATP-stimulated RAW264.7 macrophages or from HUVECs. PS exposure is associated with microvesicle shedding from THP-1 cells (24), the vesicles containing biologically active 17-kDa processed IL-1β. By analogy, we were able to isolate PS-exposed vesicles from Bz-ATP-stimulated RAW264.7 macrophages using annexin-coated beads. The annexin bound fraction contained icIL-1ra1, indicating that, as with IL-1β, PS-exposed microvesicle formation and shedding provides a release mechanism for at least one icIL-1ra. Some icIL-1ra1-myc was detected in the “annexin unbound” fraction for macrophages treated with LPS alone. Therefore, it is possible that LPS may stimulate release of icIL-1ra by an independent mechanism, perhaps akin to that observed in HUVECs. However, the levels of icIL-1ra1 released due to LPS treatment alone were much lower than for the ATP-mediated PS-exposed release.
By microscopy we observed the predominant formation of microvesicles, with fewer large blebbing events, in ATP-activated RAW264.7 cells using high micromolar concentrations of Bz-ATP and in a low sodium extracellular solution. Release of icIL-1ra1 occurred in low sodium extracellular solution for both ATP-activated RAW264.7 murine macrophages and HUVECs. Several studies indicate that ATP-mediated cytokine-containing microvesicle shedding and cell surface “tethered” blebbing occur via separable mechanisms. Microvesicle formation occurs optimally at submillimolar concentrations of Bz-ATP or ATP in both HEK293 cells and BAC1 murine macrophages, associated with IL-1β release in the macrophages (30). Whereas P2X7R-mediated, large, tethered bleb formation requires millimolar ATP concentrations, and is dependent on the serine/threonine kinase ROCK I (31), RhoA activation, and Rho-effector kinase activity (30). In addition, removal of extracellular sodium results in a reduction in P2X7R-mediated cell blebbing (32), but does not inhibit microvesicle shedding (24). Hence, the conditions used in this study favored the microvesicle formation that was observed for RAW264.7 cell stimulation, resulting in release of icIL-1ra1.
Removal of extracellular calcium prevented the release of icIL-1ra1 in both RAW264.7 cells and in HUVECs. Release of IL-1β from THP-1 monocytes is also prevented in the absence of extracellular calcium (24). A recent detailed analysis of the calcium dependence of IL-1 release from murine peritoneal macrophages showed that ATP-mediated release IL-1β is independent of calcium influx, but requires release of intracellular calcium, and K+ efflux (33). In this study, removal of extracellular calcium will also have had the effect of depleting intracellular calcium stores, hence it is possible that intracellular stores are required for the release of icIL-1ra1, an area for future investigation.
In HUVECs, we were unable to detect PS flip following activation of the P2X7R, despite the fact that icIL-1ra1 is released from these ECs in a P2X7-dependent manner. It is possible that any PS flip in this cell type was at a lower level compared with the ATP-induced exposure in the monocytes and macrophages, and is below the threshold of detection in our experimental system. Alternatively, the P2X7R-mediated release of icIL-1ra1 from ECs may occur via a PS flip-independent mechanism, providing an important area for future investigation. Functional P2X7R channels have been measured in bovine aortic endothelium as assessed by whole cell electrophysiological recording, and by immunocytochemistry (28). However, there was an absence of YO-PRO-1 uptake in these cells, indicating that the functional regulation of the P2X7R may differ in this cell type, compared with monocyte-derived cells. This difference in channel function may be attributable to differences in protein-binding interactions due to tissue specific expression. This may in turn relate to cell-type-dependent differences in the mechanism of release of leaderless proteins following activation of the P2X7R. Differences in the expression of other purinergic receptor subtypes between ECs and macrophages may also lead to altered responses according to cell type. Alternatively, differences between the responses of the murine macrophages and human ECs may be due to the difference between species.
No release of the cytoplasmic protein LDH was observed from either RAW264.7 macrophages or HUVECs under conditions where icIL-1ra1 was released, suggesting that the secretory mechanism is selective. Therefore, the release of icIL-1ra1 does not occur via a general, and nonspecific, export route for any intracellular protein. Nor is it a consequence of breakdown of the plasma membrane barrier, concomitant, for example, with cell death.
Our investigation has shown that icIL-1ra1 can be released into the extracellular environment. Keratinocytes, like the endothelium, predominantly express the intracellular isoform of the receptor antagonist (8). Previously it was shown that icIL-1ra is released from keratinocytes into the supernatant via a leaderless brefeldin A-independent mechanism, which is not associated with cytotoxicity by LDH assay (34). The levels of release are altered according to the maturity of the keratinocytes (34) and are up-regulated in skin cells from atopic dermatitis patients compared with normal controls (35). icIL-1ra1 is released into the extracellular medium from airway epithelial cells, and is not associated with LDH-assayed cytotoxicity (36). In each of these studies it is possible that activation of the P2X7R, exerted via endogenous ATP, may be required for icIL-1ra1 release; P2X7Rs are expressed in keratinocytes (37) and epithelial cells (38).
In the endothelium the only detectable isoform of IL-1ra that is inducibly expressed is this intracellular leaderless protein. Increased levels of extracellular ATP at the EC surface are likely to result in the release of icIL-1ra1. Our results show that LPS-pretreated, and even unstimulated, HUVECs were able to release endogenous ATP sufficient to activate the release of icIL-1ra1, ATP was detectable in the culture supernatants, and secretion of icIL-1ra1 was blocked by antagonizing the P2X7R or by degradation of ATP by apyrase. This suggests that an autocrine mechanism controls icIL-1ra1 release, and endogenous ATP concentrations are sufficient to activate the P2X7R. Other work has shown that during LPS-mediated acute inflammation or ATP stimulation itself, an increase in ATP release was measured from ECs (39) (40). The endothelium may be exposed to high levels of extracellular ATP under inflammatory conditions, due to release from degranulating platelets, via sympathetic nerve stimulation, from damaged cells in atherosclerosis, and under hypertension, restenosis, or ischemia (41). Hence, activation of the P2X7R resulting in secretion of icIL-1ra1, plays an important role in a number of pathological and immunological processes at the endothelium. The balance between release of proinflammatory IL-1 and the IL-1ra protein is likely to be critical in determining the arterial wall response and resultant pathology (42). It will be important to understand whether there are subtle differences in the mechanisms of release of each of the leaderless proteins, proinflammatory IL-1 and anti-inflammatory icIL-1ra1, from the endothelium, or whether their secretion is under the same control by P2X7R activation.
In summary, we have shown that icIL-1ra1 may be secreted into the extracellular medium, via a P2X7R-dependent mechanism, from both macrophages and the endothelium. The results suggest that activation of the P2X7R by extracellular ATP may regulate a more general pathway for the release of leaderless cytoplasmic proteins.
We are grateful to Gary Shaw for isolation of HUVECs, to Ian Palmer for assistance with microscopy, and to Richard Varcoe for discussion.
↵1 This work was funded by the British Heart Foundation.
↵2 Address correspondence and reprint requests to Dr. Heather L. Wilson, Section of Functional Genomics, Division of Genomic Medicine, Royal Hallamshire Hospital, Glossop Road, Sheffield S10 2JF, U.K. E-mail address:
↵3 Abbreviations used in this paper: IL-1RI, type I IL-1R; Bz-ATP, 3′-O-(4-benzoyl)benzyl-ATP; EC, endothelial cell; EGFP, enhanced GFP; IL-1ra, IL-1R antagonist; icIL-1ra, intracellular isoform of IL-1ra; icIL-1ra1, type 1 icIL-1ra; LDH, lactate dehydrogenase; Ox-ATP, periodate-oxidized ATP; PS, phosphatidylserine; sIL-1ra, secreted isoform of IL-1ra.
- Received November 25, 2003.
- Accepted April 28, 2004.
- Copyright © 2004 by The American Association of Immunologists