Published online before print September 18, 2008
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
* Department of Medical Sciences, Gastroenterology Research Group, University Hospital, Uppsala, Sweden; and
Karolinska Institute, Department of Medicine, Unit of Clinical Allergy Research, Karolinska University Hospital, Stockholm, Sweden
1 Correspondence: Department of Medical Sciences, Gastroenterology Research Group, University Hospital, S-751 85, Uppsala, Sweden. E-mail: maria.lampinen{at}medsci.uu.se
ABSTRACT
The aim of this investigation was to study the involvement of eosinophil and neutrophil granulocytes in different stages of Crohn’s disease (CD) and ulcerative colitis (UC). Biopsy samples were taken from the right flexure of the colon and from the rectum in patients with active (n=12) and inactive colonic CD (n=7), patients with active (n=33) and inactive UC (n=24), and from control subjects (n=11). Cell suspensions from biopsies and blood were analyzed by flow cytometry with regards to activation markers and viability. Immunohistochemistry was used to evaluate cell number and degranulation. Blood eosinophils were cultured with Th1 and Th2 cytokines, and the expression of activity markers was assessed by flow cytometry. Eosinophil number, viability, and activity were increased during active CD and UC compared with controls. The activity, assessed as CD44 expression, tended to diminish during inactive CD but was increased further in quiescent UC. Neutrophil number and activity were increased only during inflammation in both diseases. Culture of blood eosinophils with IL-5 and IL-13 caused increased CD44 expression, whereas IL-5 and IFN-
induced elevated CD69 expression. We observed different patterns of eosinophil activation in CD and UC, with the highest CD44 expression during quiescent UC. Our in vitro experiments with recombinant cytokines suggest that the diverse mechanisms of eosinophil activation in CD and UC are a result of different cytokine milieus (Th1 vs. Th2). In contrast, neutrophil activation reflects the disease activity in CD and UC, irrespective of Th cell skewing.
Key Words: eosinophil • survival
INTRODUCTION
Inflammatory bowel disease (IBD) is a group of chronic, relapsing disorders characterized by uncontrolled inflammation of the intestinal mucosa. Crohn’s disease (CD) and ulcerative colitis (UC) are the most common of the IBDs. They are considered two separate conditions with distinguishing clinical, endoscopical, and histological findings, but they have overlapping features and may even represent several different diseases with similar characteristics [1 ]. Under normal conditions, the intestinal mucosa is in a state of "controlled inflammation." Different subsets of T cells are present in the healthy intestine, as are eosinophils and APCs. However, there is a predominance of anti-inflammatory and regulatory cytokine responses that keep the mucosal homeostasis intact [2 , 3 ]. In IBD, the balance between mucosal responsiveness and tolerance toward antigens is disturbed, and there is an exaggerated immune response to the commensal flora.
The presence of neutrophil granulocytes is a hallmark of active IBD [4 ]. Neutrophils are recruited from the circulation to take part in the defense against infectious agents, but they may also cause tissue destruction in the host by secretion of toxic granule proteins, such as myeloperoxidase (MPO) and reactive oxygen species [5 ].
The role of eosinophil granulocytes in IBD is largely unknown. They are normally present in the intestinal mucosa, participating in the host defense [6 ], but the number of eosinophils is increased in patients with IBD [7 , 8 ], probably resulting from increased influx of eosinophils to the gut [9 ] and from dysfunctional eosinophil apoptosis. The eosinophils are proinflammatory cells with the capacity to kill invading parasites. They also take part in immunological events [10 ], and it has been suggested that eosinophils may act as APCs that stimulate T cell proliferation and activation [11 ]. Recent studies indicate that the eosinophils are involved in remodeling and tissue repair through fibroblast stimulation by the release of eosinophil cationic protein and TGF-β [12 , 13 ].
The finding of increased levels of eosinophil granule proteins in intestinal perfusion fluid from patients with UC [14 ] and CD [15 ] indicates an active involvement of eosinophils in the inflammatory process of both diseases, possibly by contributing to alterations in the colonic epithelial barrier function [16 ]. However, in our recently published study, we proposed an alternative role for the eosinophils: We found that the number of activated (CD44high) eosinophils was higher in inactive UC than in the active phase of the disease [17 ]. This new finding may indicate a dual role for the eosinophil, with involvement in tissue destruction and repair in the different stages of UC.
Given the different features of UC and CD, we found it of great interest to investigate eosinophil activity also in CD. Thus, the main objective of the present study was to compare the activity of eosinophils in active and inactive CD with our findings in UC.
We have previously demonstrated increased neutrophil activation in active UC, but in contrast to eosinophils, neutrophils were normalized completely in inactive UC. To compare the involvement of neutrophils in UC and CD, neutrophil number and CD66b expression were studied in all patients.
PATIENTS AND METHODS
Patients and control group
The patients and controls in the study were recruited from the Department of Gastroenterology at Uppsala University Hospital (Sweden) from September 2002 to May 2006 (Table 1
).
|
View this table: [in a new window] |
Table 1. Patient Characteristics
|
The patients with UC were considered to be in a phase of inactive disease if they had no clinical symptoms of disease activity, and the endoscopical picture was normal or at most showed a slight disturbance of the mucosal vessels. Patients with clinical symptoms of UC and endoscopical signs of inflammation with a score of 3–4 on the Binder scale [19 ] were considered to have active disease. Fourteen of the patients with active UC had total colitis and could be included for the study of the right flexure of the colon and rectum. Nineteen of the patients with active UC had distal colitis or proctitis and were only included for the study of the rectum. Among the patients with inactive UC, 12 of the patients had total colitis, and 13 had distal colitis or proctitis, and they were grouped accordingly.
The controls were recruited among patients examined for anemia (n=5) or healthy volunteers (n=6). Some of the patients with UC (10 with active and 10 with inactive UC) as well as the control subjects were also included in our previous study [17 ]. However, all patient groups were included in parallel, and there was no time gap between the assessment of UC and CD patients. Also, all of the samples were handled by the same technician and with the same equipment during the entire study. There was no correlation between eosinophil activation and any kind of medication in UC or CD patients (data not shown).
Biopsy samples were taken during colonoscopy (Olympus 160 AL endoscope with standard endoscopy forceps) after bowel preparation: 2 days of diet restriction and an oral purgative in the morning and afternoon of the day before the examination. The procedure does not induce inflammation [20 ] and is therefore appropriate for this setting.
The Ethical Committee of the Medical Faculty, Uppsala University, approved the project, and all patients gave their informed consent to participate in the study.
Collection and preparation of samples
During colonoscopy, four adjacent biopsy samples were taken from the right flexure of the colon and from the rectum. Two of the samples from each location were sent for routine histological analysis. The remaining two samples were transferred immediately into tubes filled with room-tempered physiological saline solution and were processed further within 1 h. Peripheral blood was also collected from all participating patients and control subjects.
Single-cell suspensions of biopsy cells were obtained using a loosely fit glass homogenizer, and the cells were then washed twice with a buffer assigned for FACS containing 0.05% NaN3, 0.1% BSA, and 0.4% trisodium citrate dihydrate in PBS. Heparinized peripheral blood from the same individuals was hemolyzed with a 0.83% ammonium chloride solution and washed twice in the FACS buffer to obtain a suspension of blood leucocytes. Both types of cell suspensions were incubated with fluorochrome-conjugated mAb for 30 min at room temperature in the dark. After a final wash, the cells were suspended in 500 µL FACS buffer and analyzed.
Antibodies for flow cytometry
Mouse anti-human mAb conjugated to FITC, PE, or PerCP were used for all antigens. Isotype-matched control labeling was also performed, using fluorochrome-conjugated mouse anti-human IgM
and IgG2b as controls for nonspecific staining. All antibodies used for flow cytometry were purchased from BD Biosciences/PharMingen (San Diego, CA, USA).
Flow cytometry assay
The flow cytometry assay was performed on a two-laser FACSCalibur cytometer (BD Immunocytometry Systems, San Jose, CA, USA). Fluorescence measurements were collected using a logarithmic amplifier; forward- and side-scatter was studied using a linear amplifier. Cells (10,000) were counted and analyzed in each sample. For data analyses, CellQuest Pro software from Becton Dickinson (San Jose, CA, USA) was used.
Eosinophil and neutrophil granulocytes from peripheral blood or biopsy samples were identified as described in detail previously [17 ]. They were gated by their forward- and side-scatter properties and identified further by specific surface markers. Eosinophils were identified as cells double-positive for CD9 and CDw125. Eosinophils with a high-surface expression of CD44 were classified as activated. CD15 was used as a marker for neutrophils, and the mean fluorescence intensity (MFI) of CD66b was used as a measure of neutrophil activation [21 ].
Apoptosis assay
We assessed the viability of intestinal eosinophils from a number of patients and controls. Biopsy samples were taken from inflamed areas of the colon from six patients with CD and from eight patients with UC, as well as from noninflamed areas from eight patients with CD, seven with UC, and from six control patients. Single-cell suspensions were prepared by enzymatic digestion of the biopsy tissue: Epithelial cells were detached by incubating the samples for 30 min at 37°C in HBSS/1 mM EDTA with frequent shaking. The tissue was then digested by incubation in RPMI-1640 medium with 10% FCS (Life Technologies, Paisley, Scotland) and collagenase type IV (1 mg/mL) and DNase (0.1 mg/mL; both from Roche Diagnostics GmbH, Penzberg, Germany) for 1 h at 37°C with gentle shaking. The cell suspension was then passed through a 40-µm nylon mesh, centrifuged, and washed twice with FACS buffer. Eosinophils were identified as described above, and death/survival was evaluated by staining with propidium iodide (PI; 5 mg/L; Orpegen Pharma, Heidelberg, Germany) and fluorescein di-acetate (FDA; 0.17 mg/L; Sigma Chemical Co., St. Louis, MO, USA), respectively, for 30 min at 4°C. PI enters dead or dying cells, and living cells are identified by the fluorescein formed by intracellular hydrolysis of FDA. Phosphatidylserine staining of apoptotic cells was performed with FITC-labeled Annexin V (BD PharMingen) in Annexin V binding buffer, according to the instructions of the manufacturer. The cells were washed with FACS buffer and analyzed by flow cytometry.
Immunohistochemistry
Immunohistochemical analyses were performed on biopsy samples from five patients from each patient group. A mAb to eosinophil peroxidase (EPO; Department of Medical Sciences, Clinical Chemistry, University of Uppsala) was used to identify eosinophil granulocytes. CD69 (Abcam, Cambridge,UK) was stained on consecutive sections and used as an additional activation marker for eosinophils. Neutrophil granulocytes were identified by a polyclonal antibody to MPO (obtained from Department of Medical Sciences, Clinical Chemistry, University of Uppsala), and the activation marker CD66b (Biolegend, San Diego, CA, USA) was stained on consecutive sections to ensure that CD66b is expressed on neutrophils.
Sections cut from wax-embedded blocks (prepared for routine histological analysis) were deparaffinized in xylene, rehydrated through decreasing concentrations of alcohol, and finally rinsed in distilled water. To expose antigenic sites and reduce background, the sections were heated in a citrate buffer using a pressure cooker. The slides were subsequently placed in an automated slide-processing system (AutostainerPlus, Dako Cytomation, Glostrup, Denmark), where sections were blocked in hydrogen peroxide/methanol, washed, and stained in several steps. The antigen-antibody complex was visualized with 3,3'-diaminobenzidine using a commercial Envision kit (Dako Cytomation), according to the instructions given in the manual. The samples were then counterstained with Mayer’s hematoxylin (Histolab Products AB, Gothenburg, Sweden), dehydrated, and mounted. The sections were examined with an Olympus BH2-MDO microscope (Olympus Optical Co. Ltd., Tokyo, Japan).
Culture of peripheral blood eosinophils with IL-5, IL-13, and IFN-
We assessed the impact of the Th2 cytokines IL-5 and IL-13 and the Th1 cytokine IFN- on eosinophil CD44 and CD69 expression by culturing peripheral blood eosinophils from healthy donors with these cytokines. Purified peripheral blood eosinophils were obtained by Percoll centrifugation and subsequent magnetic bead sorting (removal of neutrophils with anti-CD16) [22
]. Differential counts were obtained using a cytospin preparation (Cytospin, Shandon Southern Instruments, Sewickley, PA, USA) stained with May Grünewald and Giemsa and examined under light microscope. Eosinophil purity was 99 ± 1%, and the viability was 98 ± 1%, determined by staining with Trypan blue. Eosinophils were plated in flat-bottomed 96-well plates at a concentration of 1 x 105 cells/100 µL in RPMI-1640 medium supplemented with 10% FCS (both from Life Technologies), penicillin (100 IU/mL), and streptomycin (100 µg/mL; Biochrom KG, Berlin, Germany). Just after the onset of culture, the cells were treated with recombinant human (rh)IL-5, rhIL-13, or rhIFN- (R&D Systems Europe, Abingdon, UK) at three different concentrations between 10 and 1000 ng/mL or with culture medium. In some of the wells, the cells were treated with IL-5 in combination with IL-13 or IFN-. The plates were kept at 37°C with 5% CO2 for 24 h. Eosinophil activation was then evaluated by their expression of CD44 and CD69, as described in a previous section.
Statistical evaluation
Kruskal-Wallis ANOVA and the Mann-Whitney U test were used to evaluate statistical differences between groups of patients. For paired analyses, we used Friedman ANOVA and the Wilcoxon matched pairs test. Correlation coefficients were obtained by Spearman’s rank order correlation test. A P value of <0.05 was adopted as significant. All calculations were performed on a personal computer by means of the statistical software Statistica (Statsoft Inc., Tulsa, OK, USA).
RESULTS
Eosinophil granulocytes in the right flexure of the colon
The proportion of CD44high eosinophils was significantly larger in patients with active CD and UC compared with control subjects (Fig. 1
). No significant difference in the proportion of CD44-expressing cells was seen when comparing active CD and active UC. The proportion of CD44-expressing cells tended to be lower in patients with inactive CD, but there was no significant difference between patients with active and inactive CD or between patients with inactive CD and control subjects. In contrast, a significantly higher percentage of CD44high eosinophils was found in patients with inactive UC than in patients with inactive CD, and it was also significantly higher than in patients with active UC.
 View larger version (24K): [in a new window] |
Figure 1. Percentage numbers of activated (CD44high) eosinophils (eos). Significant differences between the different patient groups are indicated. The results are expressed as mean ± SEM. Kruskal-Wallis ANOVA and the Mann Whitney U test were used for statistical evaluation.
|
Neutrophil granulocytes in the right flexure of the colon and in the rectum
The neutrophil activity, assessed as MFI of CD66b, was significantly higher in patients with active CD than in patients with inactive CD and control subjects in both of the intestinal locations (Fig. 2
). However, in comparison with patients with active UC, the expression was substantially lower in active CD.
 View larger version (18K): [in a new window] |
Figure 2. MFI of CD66b on neutrophils. Significant differences between the different patient groups are indicated. The results are expressed as mean ± SEM. Kruskal-Wallis ANOVA and the Mann Whitney U test were used for statistical evaluation.
|
 View larger version (15K): [in a new window] |
Figure 3. Viability of intestinal eosinophils obtained from inflamed and noninflamed colonic tissue of patients with CD, UC, and noninflamed tissue from control patients. Results are expressed as percentage (±SEM) of FDA-positive cells and PI-positive cells. (A) Eosinophil survival measured as FDA metabolism (living cells are identified by the fluorescein formed by intracellular hydrolysis of FDA). Results are expressed as percentage of FDA-positive cells. (B) Dying or dead eosinophils (PI-positive cells), expressed as percentage of PI-positive cells.
|
Eosinophil and neutrophil counts in colonic biopsy samples
Immunohistochemical staining for EPO revealed larger numbers of eosinophils in the lamina propria of patients with active CD and UC compared with control patients. The numbers were lower in inactive CD than in active CD, whereas they were only decreased slightly in biopsy samples from patients with inactive UC compared with active UC (Table 2
; Fig. 4 A-E
). The number of neutrophils, identified as MPO-positive cells, was significantly higher in active CD and UC than in control patients and inactive CD and UC (Table 2)
.
|
View this table: [in a new window] |
Table 2. Eosinophil and Neutrophil Numbers in Rectal Biopsy Samples Evaluated as EPO- and MPO-Positive Cells, Respectively, by Immunohistochemistry
|
 View larger version (97K): [in a new window] |
Figure 4. Immunohistochemical identification of eosinophils with EPO in colonic biopsy specimens from patients with active CD (A), active UC (B), control patients (C), inactive CD (D), and inactive UC (E). The figure includes one representative patient from each group of five patients.
|
The proinflammatory activity of eosinophils can be estimated by their release of EPO into the tissue. We found extensive EPO release in biopsy samples from patients with active UC and from some of the patients with active CD. The release was less pronounced in patients with inactive UC and CD. No released EPO was found in control patients (Fig. 4 A-E) .
Neutrophil activity assessed by immunohistochemistry
We found extensive MPO release into the tissue of patients with active UC, indicating neutrophil activity. In samples from patients with CD, we also observed released MPO, however slightly less than in the UC patients. Staining with CD66b in consecutive sections revealed colocalization with MPO in active CD and active UC (not shown).
Effect of IL-5, IL-13, and IFN- on purified peripheral blood eosinophils
Incubation with IL-5 at the concentrations of 10 and 100 ng/mL for 24 h resulted in increased proportions of CD44high (Fig. 5D
) and CD69+ (Fig. 5G)
eosinophils. IL-13 had no effect on eosinophil expression of CD44 (Fig. 5E)
or CD69 (Fig. 5H)
. IFN- did not influence the expression of CD44 (Fig. 5F)
but caused a small increase in eosinophil CD69 expression (Fig. 5I)
.
 View larger version (45K): [in a new window] |
Figure 5. Identification gates of purified peripheral blood eosinophils (A) by forward- and side-scatter properties and (B) by the expression of CD9 and CDw125. (C) Isotype-matched controls for CD9 and CDw125. Eosinophils thus gated are presented in the following figures, where filled curves represent unstimulated and unfilled curves, cytokine-stimulated eosinophils. (D) Eosinophil CD44 expression after incubation with 10 ng/mL IL-5. (E) Eosinophil CD44 expression after incubation with 10 ng/mL IL-13. (F) Eosinophil CD44 expression after incubation with 10 ng/mL IFN-. (G) Eosinophil CD69 expression after incubation with 10 ng/mL IL-5. (H) Eosinophil CD69 expression after incubation with 10 ng/mL IL-13. (I) Eosinophil CD69 expression after incubation with 10 ng/mL IFN-. SSC-H and FSC-H, Side-scatter-height and forward-scatter-height, respectively; FL2-H, fluorescence 2-height.
|
in the presence of IL-5 and evaluated the expression of CD44 and CD69 after 24 h. Interestingly, IL-13 had an additive effect to that of IL-5 on the expression of CD44, whereas IFN- had an additive effect on the CD69 expression (Table 3
). |
View this table: [in a new window] |
Table 3. The Proportions of CD44high and CD69+ Eosinophils after 24 h Culture with IL-5 Together with IL-13 or IFN-
|
The role of the eosinophil granulocyte in IBD is not yet fully understood. It is a cell with many potential functions—in some inflammatory diseases, such as eosinophilic gastroenteritis and allergic disorders, eosinophils are believed to be the principal effector cells, whereas in other diseases (e.g., gastroesophageal reflux), they may have a regulatory function [6 , 10 ].
In the present study, we compared eosinophil survival and activity in two apparently related inflammatory diseases—CD and UC.
We found increased numbers of colonic eosinophils in the active phases of CD and UC compared with control patients, and the cells were viable to a higher extent than in controls. The numbers were significantly lower in quiescent CD but only slightly lower in quiescent UC, where no significant difference was found compared with active UC.
CD44 and CD69 are two established markers for eosinophil activation: The surface expression of CD44 increases upon eosinophil activation, and CD69 is only expressed on eosinophils with an activated phenotype [23 ]. EPO release is another sign of eosinophil activation and indicates proinflammatory activity. We found that CD44 was increased significantly in active CD and UC compared with control patients. Notably, during inactive UC, the expression was increased further as compared with active UC, whereas it tended to be lower during inactive CD than in the active phase of the disease. We also found CD69+ eosinophils in active CD and UC and in patients with inactive UC, but only a few of the eosinophils in inactive CD were CD69+. Thus, eosinophils show opposite patterns of activation during the inactive phases of CD and UC, suggesting different mechanisms may be operative, possibly as a result of different cytokine milieus in the two diseases. The release of EPO, however, was lower in quiescent UC than in active UC, indicating different functions of eosinophils in the two phases of the disease. The increased expression of CD44 on eosinophils may be a clue to their function in inactive UC, as CD44 is a receptor for hyaluronate, through which it stimulates eosinophils to produce TGF-β [24 ]. This cytokine is not only a stimulatory factor for fibroblasts but is also involved in the generation of regulatory T cells [25 ]. A beneficial effect of eosinophils in UC has been suggested in studies where patients with high eosinophil numbers in the affected mucosa had a better disease outcome than patients with eosinophil paucity [26 , 27 ].
CD is often referred to as a Th1-type condition, with predominant production of IFN-, TNF-
, and IL-12 [2
, 28
]. UC has not been classified using the Th1/Th2 paradigm, but the Th2 cytokine IL-5, a key factor for eosinophilia, seems to be an abundant protein in this disease: It has been found in increased concentrations in UC compared with CD and control patients by immunohistochemical stainings of human intestinal mucosa [7
]. Also, lamina propria T cells from patients with UC secreted high amounts of IL-5 after stimulation via the TCR; this is in contrast with T cells from patients with CD that secreted IFN- [29
]. IL-13 is another Th2 cytokine that is up-regulated in UC but not in CD [30
, 31
] and is present in the intestinal mucosa, even in quiescent UC [32
]. IL-5 and IL-13 stimulate eosinophil migration and prolong eosinophil survival [33
, 34
]. Thus, they may be important stimulators of eosinophils in UC, whereas other cytokines are operative in CD.
In our experiments, we found that rIL-5, but not IL-13, causes up-regulation of CD44 and CD69 on peripheral blood eosinophils. However, IL-13 had an additive effect to that of IL-5 on CD44 expression. IFN-, on the other hand, did not influence the expression of CD44 but stimulated eosinophil expression of CD69, especially in the presence of IL-5. This again indicates that different cytokines may stimulate eosinophils to diverse phenotypes.
The presence of eosinophils in CD remains to be elucidated. Eosinophils express IFN-Rs, and this cytokine has been shown to enhance eosinophil cytotoxicity and to induce their production of T cell-attracting chemokines [35
, 36
]. TNF- enhances the effects of IFN- on chemokine production. Thus, eosinophils may have immunoregulatory roles in active CD but perhaps of a different character than in UC. The increased expression of CD44 observed during disease relapse is not maintained during inactive CD, and a reason for this may be the low concentrations of IL-5 and other Th2 cytokines.
In comparison with eosinophils, the role of neutrophils in IBD seems less complicated. We found increased numbers of neutrophils in active CD and UC but normal neutrophil counts in quiescent disease. The activity of neutrophils in CD displayed similar patterns as in UC, with increased activity during active disease and a resting phenotype during the inactive phases. However, neutrophil activity was significantly lower in CD than in UC. This is consistent with a previous study from our group, where higher feces concentrations of the neutrophil protein human neutrophil lipocalin were detected in patients with UC than in patients with CD [37 ]. Interestingly, defective neutrophil recruitment has been proposed as a primary pathogenic abnormality in CD [38 ], and stimulation of the innate immune system of CD patients by GM-CSF treatment has demonstrated beneficial effects [39 ].
In summary, our results indicate different regulation of eosinophil activity in CD compared with UC. Whereas in active disease, CD and UC display increased numbers of eosinophils with increased viability and activation, we found important differences in quiescent disease: The eosinophils became activated further in UC and less active in CD.
Previous investigations have shown differential Th1/Th2 skewing of CD4+ T cells in the two diseases. Hence, the diverse mechanisms of eosinophil activation in CD and UC may be a result of different cytokine milieus, a notion that is supported by our experiments with rTh1 and rTh2 cytokines. Taken together, eosinophils may have different roles in CD and UC and may even have a dual role in UC.
ACKNOWLEDGEMENTS
This study was supported by grants from the Lisa and Johan Grönbergs Foundation, Ruth and Richard Juhlin’s Foundation, The Lars Erik Lundberg Foundation for Research and Education, the Swedish Research Council, and the Medical Faculty at Uppsala University (Uppsala, Sweden). We appreciate the skillful technical assistance of Ingrid Stolt.
Received August 2, 2007; revised August 14, 2008; accepted August 17, 2008.
REFERENCES
participate in the recruitment of eosinophils to intestinal mucosa in ulcerative colitis Dig. Dis. Sci. 46,2004-2009[CrossRef][Medline]): inhibition by immunoglobulin A complexes J. Exp. Med. 190,487-495 J. Leukoc. Biol. 76,685-691
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
