Eosinophils are potent effector cells associated with allergic inflammation and parasite infections. However, limited information exists about their turnover, migration, and survival in vivo. To address these important questions, we determined murine eosinophil turnover under steady state and inflammatory conditions by flow cytometric analysis of BrdU incorporation and analyzed their migration pattern and survival in different tissues after adoptive transfer into recipient mice. In naive mice ∼50% of bone marrow eosinophils were labeled with BrdU during a 15-h pulse, whereas only 10% of splenic eosinophils were labeled within this time frame. Unexpectedly, the rate of eosinophil production did not change during acute infection with the helminth parasite Nippostrongylus brasiliensis despite massive eosinophilia in several tissues. Eosinophils present in lung and peritoneum remained largely BrdU negative, indicating that eosinophilia in end organs was mainly caused by increased survival of already existing eosinophils rather than increased production of new eosinophils in the bone marrow. Adoptive transfer experiments revealed that eosinophils preferentially migrated to the peritoneum in a macrophage-independent and pertussis toxin-sensitive manner, where they survived for several days. Peritoneal eosinophils expressed high levels of the inhibitory receptor Siglec-F, released less eosinophil peroxidase compared with eosinophils from the spleen, and could recirculate to other organs. These results demonstrate that the peritoneum serves as reservoir for eosinophils.
Helminth infections and allergic airway inflammation lead to increased numbers of eosinophils in the bone marrow, blood, and peripheral tissues like lung or intestines. IL-3, IL-5, and GM-CSF promote the generation of eosinophils in the bone marrow and prolong their survival in peripheral tissues (1). Mature eosinophils have a short half-life in vitro, which can be increased by adding IL-5 to the medium or coculture with endothelial cells (2). Ex vivo analysis of nasal polyps from bronchial asthma patients demonstrated that IL-5 delayed apoptosis of eosinophils in these tissues (3). IL-5 transgenic mice have large numbers of eosinophils in many tissues (4, 5). However, it is still unclear to what extent eosinophilia is regulated by increased eosinophil survival in peripheral tissues or increased de novo generation of eosinophils in the bone marrow. Early studies in humans determined eosinophilopoiesis by pulse-chase experiments using [3H]thymidine as tracer (6), but a detailed analysis of eosinophil turnover in different tissues and under inflammatory conditions has not been reported. This is mainly due to a lack of specific surface markers for these cells and their low abundance in different tissues. Administration of the thymidine analog BrdU can be used to mark proliferating cells in vivo. Immunocytochemical staining of BrdU-labeled cells has been used to detect newly generated cells in lung and bone marrow of mice exposed to allergic airway inflammation (7). The number of BrdU-labeled granulocytes increased in lung and bone marrow after repeated allergen challenges, but in that case it was impossible to differentiate between neutrophils, eosinophils, and basophils.
Eosinophil migration and recruitment to inflamed tissues is mediated by chemotactic factors including chemokines, lipid mediators, and complement components (reviewed in Ref. 8). Adoptive transfer of radioactively labeled eosinophils into helminth-infected rats demonstrated their recruitment to the lung and intestine according to the known migration pattern of the parasites (9). Recently, transgenic mice expressing luciferase in eosinophils were used to demonstrate eosinophil migration in mice during infection with Schistosoma mansoni (10). The strength of this technique is the possibility to perform noninvasive real-time measurements of the kinetics of eosinophil migration in the same animal over a long time period. At present the resolution of this technique is not high enough to visualize small cell numbers, and it is impossible to distinguish whether increased eosinophil numbers in a given tissue are due to enhanced recruitment or reduced apoptosis. A better understanding of the kinetics of eosinophil turnover and migration to inflamed tissues is crucial to develop new strategies for therapeutic interventions aiming at selectively reducing eosinophil numbers in affected tissues.
The activity of eosinophils can be regulated by surface receptors including Fc receptors, members of the leukocyte Ig-like receptor family, the CD2 family, or the CD33-related sialic acid binding Ig-like lectin (Siglec) family (11). Siglec-F, a recently identified inhibitory receptor on mouse eosinophils and its human functional homologue Siglec-8 have been shown to induce apoptosis when crosslinked in vitro by mAbs (12, 13). It remains to be determined whether Siglec-F can act as inhibitory receptor and regulate degranulation of eosinophils. In this study, we used Siglec-F as marker in combination with BrdU labeling to determine eosinophil turnover and migration in vivo by flow cytometry under steady-state conditions and during infection with the helminth parasite Nippostrongylus brasiliensis. In addition, we performed adoptive transfer experiments to determine the dissemination pattern and survival of eosinophils in different tissues.
Materials and Methods
IL-4 reporter mice (4get mice) have been described and were used on a BALB/c background (14). In brief, these mice carry an IRES-eGFP construct inserted after the stop codon of the IL-4 gene. IL-5 transgenic (tg)3 mice (strain NJ.1638) were obtained from James Lee (Mayo Clinic, Scottsdale, AZ) (5). GATA-1 mutant mice (ΔdblGATA), which lack eosinophils due to a mutation in the GATA-1 binding site of the gata-1 promoter, were obtained from Craig Gerard (Harvard Medical School, Boston, MA) (15). IL-4/IL-13-deficient mice were obtained from Andrew McKenzie (MRC, London, U.K.) (16). Rag2-deficient mice were originally obtained from Taconic Farms. BALB/c mice were originally obtained from The Jackson Laboratory. Rag2-deficient, IL-5tg mice and ΔdblGATA mice were crossed with 4get mice to generate 4get/rag, 4get/IL-5tg and 4get ΔdblGATA mice, respectively. All mice were on BALB/c background, housed according to institutional guidelines and used between 6 and 12 wk of age.
N. brasiliensis infection
Third-stage larvae (L3) of N. brasiliensis were recovered from the cultured feces of infected rats, washed extensively in sterile 0.9% saline (37°C) and injected (500 organisms) into mice subcutaneously at the base of the tail. Mice were provided antibiotica-containing water (2 g/L neomycin sulfate, 100 mg/L polymyxin B sulfate; Sigma-Aldrich) for the first 5 days after infection. Mice were analyzed for BrdU incorporation of eosinophils or used as recipients for eosinophil transfers on day 9 after infection.
For short term labeling (15 h), 1 mg of BrdU was injected i.p. in 100 μl of PBS. For long term labeling (36 or 84 h), BrdU was injected at 0, 12, and 24 h. Single-cell suspensions from total lung tissue, bone marrow, spleen, peritoneal lavage, or blood were stained for Siglec-F and washed twice with PBS. Cells were resuspended in 0.15 M NaCl and then fixed by adding drop wise −20°C ethanol (final concentration 70%) and incubated for 30 min on ice. After additional washing with PBS, the cells were permeabilized overnight at 4°C in PBS, 1% paraformaldehyde, and 0.01% Tween 20 (Sigma-Aldrich). Cells were washed in PBS and resuspended in 0.15 M NaCl, 4.2 mM MgCl2 (pH 5.0). Then genomic DNA was fragmented by adding 50 Kunitz units DNase I (Sigma-Aldrich) and incubating for 10 min at 20°C followed by 30 min on ice. Finally cells were stained with FITC labeled anti-BrdU Ab (BD Pharmingen) for 30 min at 20°C.
Flow cytometry and cell sorting
Single-cell suspensions were washed once in FACS buffer (PBS/2% FCS/1 mg/ml sodium azide), incubated with anti-CD16/CD32 blocking Ab (2.4G2) for 5 min at 25°C and stained with diluted Ab mixtures. The following mAbs were used: PE-labeled anti-Siglec-F (E50-2440; BD Pharmingen), APC-labeled anti-F4/80 (BM8; eBioscience), PE-labeled anti-Thy1.1 (HIS 51; eBioscience) and PE-Alexa700-labeled anti-CD4 (RM4–5; Caltag). Samples were acquired on a FACSCalibur instrument (BD Immunocytometry Systems) and analyzed by FlowJo software (Treestar). Siglec-Fhigh IL-4/eGFPhigh and Siglec-Flow IL-4/eGFPlow eosinophils were sorted from the spleen and peritoneum of 4get/IL-5tg mice based on the expression level of IL-4/eGFP because it correlates with Siglec-F expression, and we did not want to activate the Siglec-F receptor by staining with anti-Siglec-F Abs. Cells were sorted using a high speed cell sorter (FACSAria; BD Immunocytometry Systems) with ≥99% purity.
A total of 106 total spleen cells from 4get/IL-5tg mice containing ∼50% eosinophils were placed in the upper chamber of a Transwell assay plate (5 μm pore size, 6.5 mm diameter; Corning Costar) in 100 μl of RPMI 1640 with 2% FCS. Chemotactic factors were added at the indicated concentration to the bottom only or to the bottom and top of the Transwell system. After 3 h, migrated cells were collected from the bottom, counted, and analyzed by flow cytometry.
Adoptive eosinophil transfers
4get/IL-5tg mice were used as a source for eosinophils. Single-cell suspensions of the spleen, which contained ∼50% eosinophils, or peritoneal lavage, which contained ∼80% eosinophils, were washed in PBS and 5–10 × 107 total cells were transferred into naive or N. brasiliensis-infected recipient mice by injection into either the lateral tail vein or the peritoneum, as indicated. For some transfers, eosinophils were pretreated in vitro with 100 ng/ml pertussis toxin (Sigma-Aldrich), 10 μM leukotriene (LT)B4 (Cayman Chemical) or 1000 ng/ml CCL11 (ImmunoTools) for 2 h at 37°C in RPMI 1640 with 2% FCS. For transfer in 4get/rag2-deficient mice, eosinophils were labeled with CFSE using standard protocols to be able to distinguish endogenous from transferred eosinophils. Cells were washed extensively with PBS before transfer.
Cl2MDP (clodronate) was a gift of Roche Diagnostics. Clodronate liposomes were generated as described (17), and 200 μl of clodronate liposomes were injected i.p. 2 days before eosinophil transfer.
Eosinophil peroxidase assay
Single-cell suspensions from spleen and peritoneum or sorted eosinophil populations of 4get/IL-5tg mice were washed with PBS, 2% FCS and incubated overnight in 150 μl of PBS, 2% FCS with (total release) or without (spontaneous release) 0.1% (v/v) Triton X-100 (Fluka, Buchs). The next day, supernatants were analyzed by a colometric assay. In brief, 50 μl of 2× reaction buffer (100 mM Tris-HCl, (pH 8.0), 2 mM H2O2 (Sigma-Aldrich), 2 mM O-phenylenediamine (Sigma-Aldrich)) were added to 50 μl of supernatant, and reactions were incubated for 20 min (with Triton X-100) or up to 60 min at 25°C. Plates were read on a microplate reader (Molecular Devices), and eosinophil peroxidase activity was calculated as OD490 nm − OD650 nm.
Immune fluorescence staining
Cryosections of mesenteric lymph nodes of 4get mice or eosinophil-deficient 4get/ΔdblGata mice that had been infected with N. brasiliensis 10 days before were stained with purified anti-Siglec-F followed by Cy3-labeled donkey-anti-rat IgG (Jackson ImmunoResearch Laboratories). After blocking with rat serum, sections were further stained with Alexa 647-labeled anti-B220 (Caltag) and biotinylated anti-CD4 (Caltag) followed by streptavidin-HRP and Tyramide-FITC using the TSA fluorescein system (PerkinElmer Life Sciences). Pictures were acquired with a 10× UPlanSApo objective on a Olympus microscope BX41 equipped with a F-View II camera and cell F̂ software (Olympus). Original magnification was ×80.
Analysis of eosinophil turnover by BrdU incorporation and flow cytometry
Eosinophilia is often caused by allergic inflammation or parasite infection. An impressive example is the infection of mice with the gastrointestinal helminth parasite N. brasiliensis, which induces a thousand-fold increase in eosinophil numbers in the lung during the acute phase of infection (18). In this model, pronounced eosinophilia can also be observed in other organs like blood, spleen, and peritoneum and is largely dependent on IL-5 (Fig. 1 A) (19). However, it is currently unclear whether eosinophilia reflects increased eosinophil production in the bone marrow or prolonged survival of mature eosinophils in peripheral organs because IL-5 can be involved in both processes. Therefore, we determined the turnover of eosinophils in different tissues by flow cytometric analysis of BrdU incorporation. Because granulocytes leave the bone marrow as mature cells and do not undergo further proliferation in the periphery, BrdU labeling can be used to trace the de novo generation and turnover of eosinophils in vivo.
Mice were either left untreated or had been infected 9 days before analysis with the helminth parasite N. brasiliensis to induce eosinophilia. BrdU was administered i.p. either 15, 36, or 84 h before analysis by staining for BrdU and Siglec-F, a recently identified inhibitory receptor, which is predominantly expressed on mouse eosinophils (20). BrdU labeling resulted in clearly distinguishable populations of BrdU+ and BrdU− eosinophils (Fig. 1,B). It was critical to rule out endogenous BrdU synthesis in eosinophils because one major product of eosinophil peroxidase (EPO) is hypobromous acid (21), which has been shown to react with deoxycytidine to form 5-bromodeoxycytidine (BrdC). BrdC can be metabolized to BrdU, which is then integrated into the genomic DNA of proliferating cells (22). However, as shown in Fig. 1,B, mice that had not received BrdU remained BrdU negative, demonstrating that BrdU labeling can indeed be used to determine eosinophil turnover in vivo by flow cytometry. In uninfected mice only eosinophils in bone marrow and spleen could be analyzed due to extremely low frequencies of these cells in other tissues. At 15 h after BrdU labeling, ∼50% of eosinophils in the bone marrow were BrdU+ indicating a relatively high turnover of these cells under steady-state conditions (Fig. 1,C, top left). In the spleen only, ∼10% of eosinophils were labeled with BrdU during this time frame. When mice were analyzed at 36 h after the start of BrdU administration, essentially all eosinophils in the bone marrow and the majority of eosinophils in blood and spleen appeared to be BrdU+ (Fig. 1 C, middle left).
Interestingly, mice that had been infected with N. brasiliensis showed the same frequency of BrdU+ cells in bone marrow and spleen during the 15 h labeling time frame compared with noninfected mice whereas eosinophils in blood, lung, or peritoneum were not labeled during this short period (Fig. 1,C, top right). This demonstrates that >15 h are required for newly generated eosinophils to replace existing eosinophils in these organs. At 36 h after BrdU administration, the vast majority of eosinophils in bone marrow, spleen, and blood had incorporated BrdU, whereas still relatively few eosinophils in the lung and peritoneum of N. brasiliensis-infected mice were BrdU+ (Fig. 1 C, middle right). Only at 84 h after the start of BrdU injection, 50% of eosinophils in the lung and 75% of eosinophils in the peritoneum appeared BrdU+ which clearly shows that at least during infection the complete turnover of the eosinophil population in lung and peritoneum takes several days.
Next, we analyzed whether BrdU incorporation differs between eosinophil subsets in the lung. We and others (12, 23) have shown that Siglec-F is up-regulated whereas CD62L is down-regulated on activated eosinophils in peripheral tissues. The majority of eosinophils in the lung of N. brasiliensis-infected mice were Siglec-Fhigh, which indicates their activated phenotype (Fig. 1,D). When BrdU incorporation was analyzed in Siglec-Fhigh and Siglec-Flow eosinophils at 36 h after BrdU administration, it became clear that Siglec-Fhigh eosinophils were mainly BrdU− and therefore could not be directly derived from newly generated eosinophils. In contrast, the Siglec-Flow population which constituted 18% of total eosinophils in the lung was efficiently labeled with BrdU (>50% incorporation) indicating that only these cells represent newly generated and recruited eosinophils (Fig. 1,D). Therefore, eosinophils that had up-regulated Siglec-F showed a slower turnover compared with Siglec-Flow eosinophils. Because total eosinophil numbers in the lung increase dramatically during the acute phase of infection (18), but only few eosinophils show BrdU incorporation (Fig. 1, C and D), one must conclude that the Siglec-Fhigh population is mainly composed of eosinophils with increased survival that already existed before BrdU administration and were recruited from blood and spleen without further cell division. At the same time eosinophils in blood and spleen are replaced by newly generated eosinophils from the bone marrow, which explains the high frequency of BrdU+ eosinophils in these organs. Taken together, these results demonstrate that parasite-induced eosinophilia is mainly regulated by increased survival and recruitment of existing eosinophils rather than increased de novo production of eosinophils in the bone marrow.
Adoptive transfer of eosinophils reveals their migration pattern and survival in vivo
To analyze eosinophil migration and survival in vivo in more detail, 5 × 107 eosinophils were adoptively transferred into naive or N. brasiliensis-infected recipient mice by tail vein injection. We used the spleen of IL-5 tg mice, which had been crossed to IL-4/eGFP reporter mice (4get mice) as a source for eosinophils to be able to distinguish transferred eosinophils from endogenous eosinophils because eosinophils are constitutively eGFP+ in 4get mice (23, 24). Different organs of recipient mice were analyzed at 4, 24, and 72 h after transfer by flow cytometry. The frequency of transferred eosinophils in blood and spleen rapidly declined. Eosinophils also rapidly disappeared from the lung of naive mice, but remained present in the lung of N. brasiliensis-infected mice (Fig. 2), which could be explained by better survival due to the presence of IL-5-producing Th2 cells. Surprisingly, eosinophils in the peritoneum remained at a relatively constant level in both naive and N. brasiliensis-infected mice. The efficient migration to the peritoneum and prolonged survival at that site was unexpected. In fact, at two weeks after transfer the peritoneum was the only site where a population of donor-derived eosinophils could be found (data not shown). This phenomenon was unlikely to be caused by local accumulation of donor-derived IL-5 producing T cells in the peritoneum cells because only few donor-derived T cells migrated to the peritoneum (Fig. 2 C). The adoptive transfer experiments demonstrate that under the right conditions some eosinophils can survive for several days in vivo, which nicely complements the results from the BrdU-labeling experiments.
Eosinophil recruitment to the peritoneum can be blocked by pertussis toxin
Recruitment of eosinophils to peripheral tissues including lung, skin, or intestine is thought to be regulated by gradients of inflammatory chemokines like CCL11 (eotaxin-1) and CCL24 (eotaxin-2), lipid mediators including LTB4, platelet-activating factor, and prostaglandin D2 or complement components C3a and C5a, which all signal via Gi protein-coupled receptors (GPCR). To determine whether eosinophil recruitment to the peritoneum is mediated by an active process that requires signaling via GPCR, eosinophils from the spleen of 4get/IL-5tg mice were incubated for 2 h with pertussis toxin, washed, and adoptively transferred into naive recipient mice. Eosinophil recruitment to the lung, bone marrow, spleen, and peritoneum was analyzed 16 h later. Surprisingly, pertussis toxin-treated eosinophils efficiently migrated to lung, spleen, and even bone marrow but not to the peritoneum (Fig. 3, A and B). Therefore, a chemotactic factor which signals via GPCR seemed to be essential for eosinophil recruitment to the peritoneal cavity. Ligand induced in vitro desensitization of two major chemotactic receptors for murine eosinophils, CCR3 and the LTB4 receptor BLT1, caused efficient inhibition of chemotaxis in Transwell assays; however, in vivo migration to the peritoneum was not affected (Fig. 4). This suggests that recruitment is caused by GPCRs other than BLT1 or CCR3. Further studies are required to determine which GPCRs on eosinophils regulate migration to the peritoneum.
We have recently shown that during infection with N. brasiliensis, recruitment of eosinophils to the lung and peritoneum is regulated by alternatively activated macrophages (23). To analyze whether eosinophil recruitment to the peritoneum under homeostatic conditions is regulated by resting macrophages, eosinophils from 4get/IL-5tg mice were transferred into naive BALB/c mice that had been depleted of peritoneal macrophages by clodronate-containing liposomes. One day later recruitment to the peritoneum was determined by flow cytometry. Clodronate administration completely eliminated peritoneal macrophages but the frequency of endogenous and transferred eosinophils was only partially reduced (Fig. 3,C). This demonstrates that clodronate liposomes are not toxic for eosinophils and that homeostatic eosinophil recruitment to the peritoneum occurs in the absence of peritoneal macrophages. Normal recruitment was also observed after transfer into IL-4/IL-13- or Rag-deficient mice (Fig. 3 D). Therefore, homeostatic eosinophil recruitment to the peritoneum is not regulated by cells of the adaptive immune system or by resident peritoneal macrophages, which is in contrast to the macrophage-dependent accumulation of eosinophils in the peritoneum of N. brasiliensis-infected mice (23).
Next, eosinophils isolated from the peritoneum of 4get/IL-5tg mice were directly transferred to the peritoneum of naive BALB/c mice to analyze whether eosinophils can leave the peritoneum, enter the circulation and disseminate to other organs. Indeed, small numbers of eosinophils were detected in blood, spleen, and lymph nodes at 16 h after transfer (Fig. 5,A). The mediastinal lymph nodes, which drain the peritoneum (25), contained ∼10 times more eosinophils than the inguinal lymph nodes. Pretreatment of eosinophils with pertussis toxin did not block their exit from the peritoneum, because equal frequencies of eosinophils were found in the blood. Therefore, entry but not exit from the peritoneum is an active process mediated by GPCRs. GPCRs also seemed to be involved in recruitment of eosinophils to lymph nodes because pertussis toxin-treated eosinophils preferentially accumulated in the spleen and were less efficiently recruited to lymph nodes (Fig. 5,A). When the localization of eosinophils in lymph nodes of N. brasiliensis-infected mice was analyzed by immune fluorescence staining of tissue sections, they were found to accumulate in the subcapsular sinus and T cell area suggesting that under physiologic conditions eosinophils enter lymph nodes via the afferent lymph and migrate to the T cell zone where they might modulate immune responses (Fig. 5 B).
Modulation of Siglec-F expression on eosinophils in spleen and peritoneum
We have previously described that eosinophils in peritoneum, lymph nodes, and thymus of naive mice show an increased expression level of the inhibitory receptor Siglec-F compared with eosinophils in spleen, blood, or bone marrow (23). In contrast to wild-type mice, eosinophils in the spleen of 4get/IL-5tg mice show a biphasic expression of Siglec-F and IL-4/eGFP suggesting that the high level of IL-5 in these mice increased the expression of Siglec-F and IL-4/eGFP on about half of the eosinophils (Fig. 6,A). Siglec-F has been demonstrated to mediate proapoptotic signals (12). Therefore, one might expect that Siglec-Fhigh eosinophils are preapoptotic cells with a shorter lifespan compared with Siglec-Flow eosinophils. However, in vitro culture of total splenocytes from 4get/IL-5tg mice revealed that the Siglec-Fhigh eosinophil population survived better than the Siglec-Flow population under neutral culture conditions (Fig. 6,A). This finding fits with the increased survival of Siglec-Fhigh eosinophils in the peritoneum (Fig. 2). When eosinophils were cultured in the presence of IL-5, the expression of Siglec-F remained at a higher level compared with eosinophils cultured under neutral conditions indicating that Siglec-F expression is indeed induced or stabilized by IL-5 (Fig. 6,B). When eosinophils from the spleen of 4get/IL-5tg mice were transferred into naive recipient mice and recovered 24 h later, donor-derived eosinophils in the spleen appeared Siglec-Flow IL-4/eGFPlow whereas they had increased expression levels of both markers in the peritoneum suggesting preferential recruitment of Siglec-Fhigh IL-4/eGFPhigh eosinophils to the peritoneum or local up-regulation of both markers at that site (Fig. 6,C). Eosinophils in the peritoneum of 4get/IL-5tg mice consisted of a homogeneous population of Siglec-Fhigh cells similar to eosinophils found in wild-type mice (23) (Fig. 6,D). It is currently unclear whether eosinophils undergo a linear differentiation from Siglec-Flow to Siglec-Fhigh cells. To analyze whether Siglec-Fhigh eosinophils can revert to a Siglec-Flow phenotype, eosinophils were isolated from the peritoneum of 4get/IL-5tg mice and transferred into naive recipient mice by tail vein injection. Twenty four hours later, spleen and peritoneum of recipient mice were analyzed by flow cytometry. The expression level of Siglec-F and IL-4/eGFP was reduced on donor-derived eosinophils in the spleen but remained high on donor-derived eosinophils that had migrated to the peritoneum (Fig. 6 D). This shows that up-regulation of Siglec-F is reversible indicating that the activation state of eosinophils can be regulated by the tissue environment.
To further analyze whether eosinophils in peritoneum and spleen differ with respect to their effector cell potential, EPO release was determined from splenic or peritoneal eosinophils of 4get/IL-5tg mice. Splenic eosinophils contained and spontaneously released more than twice the amount of EPO compared with eosinophils from the peritoneum, which indicates that eosinophils in the peritoneum are partially degranulated (Fig. 7). Although eosinophils with low EPO content and increased survival might have lost immediate effector potential they could still serve an immune regulatory function by interaction with other cells of the innate and adaptive immune system.
This article is the first description of eosinophil turnover under steady-state and inflammatory conditions in vivo using BrdU labeling and flow cytometry. Our results demonstrate that 50% of eosinophils in the bone marrow were labeled with BrdU within 15 h. Because eosinophils do not accumulate in the bone marrow under steady-state conditions, 50% of eosinophils must have left the bone marrow during that time frame. In the periphery, the first BrdU-labeled eosinophils were detected in the spleen, which indicates efficient recruitment to that organ. However, only ∼10% of splenic eosinophils were labeled during the 15 h BrdU pulse. This can be explained by dilution of newly generated eosinophils within the pool of existing eosinophils in the periphery. At 36 h the peripheral eosinophil pool was almost completely replaced by newly generated eosinophils. Therefore, eosinophils are produced at a rapid rate in the bone marrow but complete replacement of the eosinophil pool in blood and spleen takes about 2 days. One might have expected a higher rate of eosinophil generation in the bone marrow during N. brasiliensis infection, because peripheral eosinophil numbers increase dramatically after infection. However, this was not the case. We conclude that parasite-induced eosinophilia results mainly from increased survival and recruitment of existing eosinophils rather than de novo generation of eosinophils. Indeed it has been shown that the number of eosinophil precursors in the bone marrow increased only 3-fold during acute infection with Trichinella spiralis, which is insufficient to explain massive tissue eosinophilia (26). Increasing eosinophil numbers by preventing apoptosis is obviously much faster than generating and recruiting new eosinophils from the bone marrow and therefore serves as a perfect mechanism to rapidly respond to infection. Neutrophils have a similar or even shorter lifespan compared with eosinophils, and their total numbers can increase rapidly by several mechanisms that reduce spontaneous apoptosis (27).
Increased eosinophil survival could also be observed in vitro under hypoxic conditions (28). Low partial oxygen pressure is found in the peritoneal cavity and might therefore contribute to the increased survival of eosinophils at that site (29). Crosslinking of Siglec-F has recently been shown to induce apoptosis in eosinophils (12). Here, we found that increased survival of eosinophils correlated with high expression levels of Siglec-F in vivo. It remains to be determined whether eosinophils with a low expression level of Siglec-F are more susceptible to induction of apoptosis by receptor crosslinking.
Apoptosis in eosinophils can occur spontaneously by withdrawal of survival factors like IL-5, IL-3, or GM-CSF or can be induced by signals from death-inducing receptors on the cell surface. Eosinophil survival is partially controlled by members of the bcl-2 family. The antiapoptotic protein Bcl-xL was up-regulated by IL-5 and GM-CSF and partially blocked apoptosis in human eosinophils cultured in vitro (30). IL-5 might also interfere with the activation of Bax, a proapoptotic bcl-2 family member, thereby preventing the release of cytochrome c from the mitochondrion (31). The proapoptotic BH3-only Bcl-2 homologue Bid is highly expressed in human eosinophils and gets activated by withdrawal of IL-5 or at least partially by Fas-mediated signals (32). It was further demonstrated that inhibitor of apoptosis proteins cIAP-2 and survivin were up-regulated in eosinophils by IL-5, IL-3, and GM-CSF and inhibited caspase-3 activation via the mitochondrial pathway but not via the Fas pathway (33). Although IL-5 has been shown to prolong eosinophil survival ex vivo, clinical trials aimed at reducing eosinophil counts in allergic asthma patients by anti-IL-5 therapy were rather unsuccessful. Therefore, other signals might control eosinophil survival in tissues. Unraveling the mechanisms that specifically induce apoptosis in eosinophils will have great therapeutic potential. A number of orphan nuclear receptors including members of the NR4A family could be promising drug targets (34).
Migration of eosinophils to end organs like lung and intestine is mainly regulated by chemokines and lipid mediators, demonstrated by using gene-deficient mice or blocking Abs (reviewed in Ref. 8). Injection of recombinant eotaxin into the peritoneum induced efficient eosinophil recruitment that could be partially blocked by anti-P- and anti-E-selectin Abs and was completely blocked by administration of the glucocorticoid dexamethasone (35). We have previously shown that s.c. infection with the helminth parasite N. brasiliensis, which does not migrate to the peritoneum, causes an accumulation of eosinophils in the peritoneum that depends on the presence of alternatively activated macrophages (23). Here we show that eosinophil migration to the peritoneum under homeostatic conditions is an active process that requires expression of a GPCR on eosinophils that can be blocked by pertussis toxin.
The peritoneum might serve as reservoir for eosinophils from where they can recirculate to other tissues, especially the lung-associated mediastinal lymph nodes. It has been shown that eosinophils can act as APCs for CD4+ T cells (36, 37, 38). A commonly used protocol to induce allergic airway inflammation in the mouse requires repeated injections of a model Ag with alum as adjuvant into the peritoneum before mice are challenged by Ag administration to the lung 2 to 3 weeks later. Therefore, it seems plausible that eosinophils that enter mediastinal lymph nodes from the peritoneum are involved in T cell priming, although dendritic cells might well be sufficient for T cell activation in these lymph nodes. Intratracheal instillation of eosinophils results in migration of these cells to paratracheal lymph nodes where they appear in the T cell area at 24 h after transfer (38). Using a more physiologic model, we could show here that eosinophils were localized in the subcapsular sinus and T cell zone of mesenteric lymph nodes after N. brasiliensis infection. It remains to be determined to what extent eosinophils can interact with T cells in vivo.
Taken together, we demonstrate that eosinophilia is mainly due to increased eosinophil survival in peripheral tissues rather than increased de novo generation in the bone marrow. Eosinophils were efficiently recruited to the peritoneum under homeostatic and inflammatory conditions in a pertussis toxin-sensitive and macrophage-independent manner. In the peritoneum, eosinophils up-regulated Siglec-F and survived for several days. Finally, they could exit the peritoneum and recirculate to other organs including lymph nodes where they might interact with T cells. Future studies should focus on the identification of pro- and antiapoptotic proteins that regulate eosinophil survival in end organs to develop drugs that can specifically induce or prevent apoptosis in these cells depending on whether the aim is protection against helminth parasites or amelioration of eosinophil-associated diseases.
We thank T. Brocker for support, S. King and R. Obst for comments on the manuscript, C. Cheminay for help with cell sorting, and A. Bol and W. Mertl for animal husbandry.
The authors have no financial conflict of interest.
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.
The work was supported by an Emmy Noether Grant (VO944/2-2) from the Deutsche Forschungsgemeinschaft (DFG).
Abbreviations used in this paper: tg, transgenic; GPCR, G protein-coupled receptor; LT, leukotriene; EPO, eosinophil peroxidase.