IL-33 Precedes IL-5 in Regulating Eosinophil Commitment and Is Required for Eosinophil Homeostasis

Eosinophils are immune cells that circulate in the blood and also reside in several tissues, including the intestine, thymus, and adipose tissue (1). In addition to their roles in homeostatic processes, eosinophils contribute to the pathology of many type 2–mediated diseases, such as asthma, eosinophilic esophagitis, and atopic dermatitis (2). Many studies have established the important effector functions of eosinophils and their ability to modulate inflammation through the release of granule contents and cytokines. However, the development of eosinophils in the bone marrow is less understood. It has been established that granulocyte–macrophage progenitors (GMPs) give rise to eosinophil lineage–committed progenitors (EoPs), which then develop into fully granulated mature eosinophils (EoMs) (3). Although IL-3, GM-CSF, and IL-5 can drive this eosinophilopoiesis process in vitro (1), IL-5 appears to be the critical cytokine specific to eosinophil development (4–6) and mechanistically acts to drive expansion and survival of EoMs within the bone marrow (7). In contrast, the factors involved in driving the initial commitment of GMP into the eosinophil lineage are less clear.

IL-33 is the most recently discovered member of the IL-1 family of cytokines. In its initial description by Schmitz et al. (8), rIL-33 was shown to promote several type 2–associated responses, including type 2 cytokine expression (IL-4, IL-5, and IL-13) and IgE production. Furthermore, ST2, the IL-33R, is expressed on many cell types involved in type 2 effector responses, including Th2 cells (9), mast cells, basophils, eosinophils (10), and type 2 innate lymphoid cells (ILC2s) (11). Subsequently, IL-33 has been extensively studied in the setting of helminth infections and allergic diseases. Studies in asthma (12–14), food allergy (15), and hookworm models (16) have reported the presence of reduced eosinophilic inflammation in IL-33– or ST2-deficient mice, suggesting a positive interplay between IL-33 and eosinophils. Indeed, the initial description of IL-33 demonstrated that in vivo administration of rIL-33 was sufficient to increase peripheral blood eosinophil numbers (8). Similarly, in vitro IL-33 was proposed to support eosinophil differentiation from bone marrow (17). In sharp contrast, Dyer et al. (18) examined the effects of IL-33 on eosinophil development using in vitro differentiation approaches and concluded that IL-33 antagonized IL-5–dependent eosinophilopoiesis and supported monocyte development. Macrophage activation has also been implicated in driving IL-33–induced lung eosinophilia (19).

In this study, we sought to reconcile these conflicting results by examining the role of IL-33 in eosinophil development in vivo and in vitro. We demonstrate that IL-33– and ST2-knockout (KO) mice show homeostatic dysregulation of granulocyte responses in both the blood and the bone marrow compartments. Furthermore, our data show not only that IL-33 is a potent stimulus for expansion of the Siglec-F⁺ eosinophil pool, but also that the functional influence of IL-33 lies in expansion of an eosinophil precursor (EoPre) population, as well as in upregulation of the IL-5Rα on this population. As already established, IL-33 also strongly induces IL-5, which further fuels the development of EoPre cells into an EoM phenotype. Consequently, we propose that IL-33 and IL-5 are cooperative cytokines for eosinophilopoiesis and that IL-33 precedes the need for IL-5 support in the progression toward eosinophil maturity.

Wild-type (WT) C57BL/6J mice were purchased from Jackson Laboratories (Bar Harbor, ME). ST2 KO mice were previously generated by Andrew McKenzie and backcrossed to C57BL/6J for eight generations. IL-33 KO mice on the C57BL/6J background were provided by Dr. Dirk Smith (Amgen, Seattle, WA). IL-5–transgenic mice (strain NJ.1638, previously described [20]) were provided to Dr. Sergejs Berdnikovs by Dr. James Lee (Mayo Clinic, Phoenix, AZ) and crossed with ST2 KO mice. Depending on the experimental requirements, both male and female mice (aged 6–36 wk) were used. Animals were housed under specific pathogen-free conditions at Northwestern University. All experiments were approved by the Northwestern University Animal Care and Use Committee.

Mice were given IL-33 (eBioscience, San Diego, CA) daily by i.p. injection at 0.4 μg/d for a total of seven injections. For some experiments, mice were also given either anti–IL-5 (TRFK5; eBioscience) or isotype control (rat IgG1 κ; eBioscience) at a dose of 25 μg per mouse by i.p. injection on days −1, 2, and 5. One day after the last injection, cells and serum were collected from blood for eosin staining, flow cytometry, and cytokine protein analysis. In addition, bone marrow was collected for RNA, cytologic, and flow cytometric analysis.

Real-time RT-PCR was used to determine IL-5 expression, as previously described. Myeloid-to-erythroid ratios in bone marrow were assessed by histologic inspection of nuclear and staining profiles on cytologic preparations (Shandon Cytospin; Thermo Fisher Scientific, Waltham, MA) stained with Kwik-Diff stain (Thermo Scientific) using established approaches (21); the relative proportions of granulocytic and erythrocytic cells were calculated after lymphocyte elimination.

Blood was collected into EDTA-coated tubes, and absolute eosinophil numbers were determined after staining with Discombe’s fluid.

Bone marrow from the femur and tibia was recovered by brief centrifugation, and the pellet was resuspended in 10 ml of complete media (RPMI 1640 with 2 mM of l-glutamine, 10% FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, 1% nonessential amino acids, 1 mM of sodium pyruvate, 25 mM of HEPES, 0.05 mM of 2-ME). Cells were counted and seeded at 3 × 10⁶ cells/ml in 1 ml of complete media supplemented with one of the following conditions for 3 d: 100 ng/ml stem cell factor (SCF) and 100 ng/ml Flt3 ligand (Flt3L), 10 ng/ml IL-5, or 10 ng/ml IL-33. Alternatively, cells were seeded at 0.5 × 10⁶ cells/ml in 6 ml of complete media supplemented with 100 ng/ml SCF and 100 ng/ml Flt3L with or without 10 ng/ml IL-33; on days 3 and 7, nonadherent cells were collected, counted, and readjusted to 0.5 × 10⁶ cells/ml with fresh medium containing 10 ng/ml IL-5.

Bone marrow neutrophils were isolated as previously described (22). The purity of neutrophils (>95%) was confirmed using flow cytometry against Ly6C⁺Gr-1⁺Siglec-F⁻ cells.

Bone marrow cells or whole blood was lysed with RBC lysis buffer (eBioscience) following the manufacturer’s protocol. Bone marrow cells were counted, and 5 × 10⁶ cells were used for staining. Cells were washed with PBS and stained with 0.25 μl of LIVE/DEAD Fixable Aqua Dead Cell Stain (Thermo Fisher Scientific) in 500 μl of PBS for 20 min at room temperature in the dark. In some experiments, cells were washed with FACS buffer (1% FBS in PBS) and incubated with allophycocyanin-Cy7–labeled anti-CD16/32 for 30 min (Supplemental Table I, panel 5). After washing in FACS buffer, cells were blocked with anti-CD16/CD32 (BD Biosciences, San Jose, CA) for 10 min and then stained in 100 μl of Ab mixture in FACS buffer (as detailed in Supplemental Table I) for 30 min at 4°C in the dark. Cells were then washed in FACS buffer and fixed in 4% paraformaldehyde. Samples were run on an LSRII flow cytometer (BD Biosciences) or sorted on a FACSAria SORP system. Data were analyzed on FlowJo 10.7 (Tree Star, Ashland, OR). Compensation on samples collected by the LSRII was performed in FlowJo postcollection.

Appropriate statistical testing was performed using GraphPad Prism 6 software (GraphPad, La Jolla, CA).

Because recently it had been shown that ST2 was expressed on hematopoietic stem cells (23), we initially asked whether ST2 was necessary for competent hematopoiesis. Naive ST2 KO mice had lower spleen weight/body weight ratios and total cell numbers in bone marrow than WT mice (Fig. 1A, 1B), suggesting a defect in the hematopoietic compartment. Upon further cytologic examination of hematopoietic-derived cell populations in peripheral blood, ST2 KO mice had fewer peripheral blood eosinophils than WT mice as determined by staining with Discombe’s fluid (Fig. 1C). Because IL-33 had been proposed to promote a macrophage-like phenotype in vitro (18) and neutrophilic inflammation in vivo (24), the frequency of lymphocytes, neutrophils, eosinophils, and monocytes in the blood of ST2 KO and WT mice was also assessed using flow cytometry (Fig. 1D, 1E). In addition to eosinophils, ST2 KO mice had fewer neutrophils than WT mice (Fig. 1E). Notably, there were no differences in total monocytes, CD115⁺Ly6C⁺ monocyte progenitors, or CD115⁺Ly6C⁻ mature monocytes. Similar to ST2 KO mice, IL-33 KO mice also showed lower numbers and frequency of both eosinophils and neutrophils in peripheral blood than WT mice (Fig. 2B–D and data not shown). Thus, IL-33 and ST2 are necessary for neutrophil and eosinophil homeostasis in the periphery.

Next, we wanted to determine whether exogenous IL-33 was sufficient to induce hematopoiesis in vivo and whether such an effect was dependent on ST2. Following the approach used by Schmitz et al. (8), we injected WT, IL-33 KO, and ST2 KO mice with 0.4 μg of rIL-33 or PBS for 7 d and analyzed them 18 h after the last injection. As predicted, exogenous IL-33 increased splenic weight and peripheral blood eosinophils in WT and IL-33 KO mice, but not ST2 KO mice (Fig. 2A, 2B). This increase in eosinophils by cytology was also confirmed by flow cytometry, with eosinophils being defined as CD45⁺SSC^hiLy6G^neg−loCD11b⁺Siglec-F⁺ cells (Fig. 2C, 2D). Although the levels of eosinophils observed after IL-33 treatment were lower in IL-33 KO mice than in WT, this was not significantly different, and the relative increases were similar when the differences in basal numbers were considered. In terms of the effect of exogenous IL-33 on other cell populations in the blood, IL-33 did not increase the frequency of neutrophils in the blood (Supplemental Fig. 1) despite ST2 KO mice displaying impaired neutrophil numbers in the basal state (Fig. 1E); it did, however, increase the numbers of ILC2s in agreement with previous literature (11). Because we observed differences in total bone marrow cell numbers (Fig. 1B) and eosinophil numbers in the peripheral blood between ST2 KO and WT mice (Fig. 1C), we queried whether the reduced numbers in ST2 KO mice might be caused by an effect on the bone marrow itself (and perhaps eosinophilopoiesis) rather than just the known effect on the periphery. Indeed, in addition to the changes seen in the blood, the bones of the IL-33–treated WT mice were noticeably lighter in color than in PBS-treated controls, and this color change was ST2 dependent (Fig. 2E). Upon further examination, IL-33 treatment increased the myeloid-to-erythroid ratio in the bone marrow, indicating an expansion in the myeloid compartment within the bone marrow compartment and that IL-33 might exert direct function at this important eosinophil development site (Fig. 2F). Taken together, these data suggest that, within granulocyte populations, IL-33 can expand EoMs in the blood and alter myeloid cells in the bone marrow.

The increased myeloid-to-erythroid ratio led us to further examine eosinophil development specifically in the bone marrow using flow cytometry. Because we observed an increase in the pool of Siglec-F⁺ eosinophils in peripheral blood after IL-33 treatment, we used this Siglec-F marker to initially examine the cells in the bone marrow. Indeed, the total Siglec-F⁺ cell population in the bone marrow was similarly increased in WT and IL-33 KO, but not ST2 KO, mice after IL-33 treatment (Fig. 3A, 3B). Interestingly, strict gating based on the fluorescence minus one controls incorporated a range of Siglec-F–expressing cells and higher overall percentages than have been previously reported for EoMs (25). To further examine eosinophil-related populations within this Siglec-F⁺ pool, we defined a population of GMP-like cells (Lin⁻Sca1⁻Siglec-F⁺IL-5Rα⁻SSC^lockit^hiCD34⁺ST2⁺), EoPre cells (Lin^hiSca1⁻Siglec-F^loIL-5Rα⁺SSC^lockit⁻CD34⁻ST2⁻), and EoM cells (Lin^loSca1⁻Siglec-F^hiIL-5Rα⁺SSC^hickit⁻CD34⁻ST2^lo) (characterization and fluorescence minus one control staining are shown in Supplemental Fig. 2). Because these GMP-like and EoPre cells expressed slightly different markers than the classically defined GMP and EoP cells, we also used traditional staining regimens to examine the GMP (Lin⁻Sca1⁻ckit^hiCD34⁺CD16/CD32^hi), EoP (Lin⁻Sca1⁻IL-5Rα⁺ckit^loCD34⁺), and common myeloid progenitor (Lin⁻Sca1⁻ckit^hiCD34⁺CD16/CD32^hi), which precedes the GMP (Supplemental Fig. 3) (3). Importantly, EoM frequency in the bone marrow was significantly lower in PBS-treated ST2 KO and IL-33 KO mice than in WT mice (Fig. 3C, 3D), which is in agreement with our analysis of EoMs in the peripheral blood (Fig. 1C, 1E). Similar frequencies of EoPre, GMP-like, and GMP cells were observed at baseline for all genotypes (Fig. 3C, 3E, 3F, Supplemental Fig. 3). ST2 KO mice had significantly lower EoPs at baseline (Supplemental Fig. 3). After treatment with IL-33 for 7 d, WT and IL-33 KO mice had dramatically increased frequency of EoMs (Fig. 3C, 3D), lower GMP-like cells (Fig. 3F), and similar EoPre (Fig. 3C, 3E). Similar results were observed when CD11b was used instead of the full lineage mixture (Supplemental Fig. 4A, 4B). Because IL-33 was previously shown to induce IL-5 (8), we hypothesized that the expanded EoM pool might be because of the influence of IL-5. Indeed, IL-33 treatment significantly increased IL-5 mRNA levels in bone marrow (Fig. 3G) and IL-5 protein levels in serum (Fig. 3H). Taken together, these data define that IL-33 supports both the expansion of the EoM pool and the elevation of IL-5 in the bone marrow and periphery.

To address the contribution of this elevated IL-5 in IL-33–driven eosinophil expansion, we treated mice with i.p. injection of 0.4 μg of IL-33 for 7 d in the presence of a neutralizing anti–IL-5 Ab or its isotype control (Fig. 4A). In this particular experiment, we noticed that anti–IL-5 treatment dramatically altered IL-5R expression levels on cells from those treated mice, most likely because of feedback from IL-5R being internalized upon binding to IL-5 (26). We therefore felt it was inappropriate to define the EoPre and EoM populations in this study, because proper identification of these populations relied on IL-5R staining as a defining marker. Instead, we focused on the significant increase in Siglec-F⁺ cells, as shown in Fig. 3B, and the significant decrease in GMP-like cells, as shown in Fig. 3F. In mice that received PBS, anti–IL-5 treatment did not have any significant effect on the frequency of Siglec-F⁺ cells (Fig. 4B, 4C) or GMP-like cells (Fig. 4D) in the bone marrow; moreover, despite a trend toward being lower, no significant difference was observed in the number of eosinophils in the peripheral blood after anti–IL-5 treatment (Fig. 4E). As before, the addition of IL-33 significantly increased the Siglec-F⁺ population, and this was prevented by anti–IL-5 treatment (Fig. 4C), suggesting that this response was regulated by the elevated IL-5 levels upon IL-33 treatment. Anti–IL-5 treatment was also sufficient to completely block the increases in peripheral EoMs (Fig. 4E). In contrast, the significant decrease in GMP-like cells upon IL-33 treatment was unaffected by blockade of IL-5. Collectively, these data suggest that the IL-33–driven expansion of EoMs we observed is dependent on IL-5. In contrast, the decrease in the GMP-like population is IL-5 independent and demonstrates that IL-33 exerts its influence on less mature populations that are separable from the influences of elevated systemic IL-5.

We next sought to better define whether the requirement for IL-33 on basal eosinophil homeostasis lay upstream or downstream of IL-5. We used the NJ.1638 strain of mice, which possess transgenic overexpression of IL-5 and develop a profound, age-dependent hypereosinophilia in the blood (27), and examined the effects of ST2 deficiency on this response. Evaluating the homeostatic peripheral blood eosinophil numbers over time, we observed that the substantial elevations in eosinophils seen in the NJ.1638 mice were significantly diminished in the absence of IL-33 signaling in the NJ.1638/ST2KO mice (Fig. 4F). Although we observed a trend toward higher serum IL-5 at early time points in the NJ.1638/ST2KO mice versus NJ.1638, this was not statistically significant and there were no significant differences as we tracked these animals during aging (Fig. 4G). Interestingly, NJ.1638/ST2KO mice still had more eosinophils than WT and ST2 KO mice, implying that ST2 is not an essential checkpoint for eosinophilopoiesis and that ST2-independent mechanisms exist to support IL-5–responsive eosinophil development. Regardless, these studies do establish that ST2 clearly regulates IL-5–driven eosinophil homeostasis and that IL-33 signaling lies upstream of the effect of IL-5 on eosinophils.

With our in vivo data showing that IL-33 and IL-5 both play a role in increasing eosinophils, we next turned to in vitro culture systems to assess the mechanisms by which IL-33 and IL-5 cooperate to promote development of EoMs. Previously, Dyer et al. (18, 28) used a protocol for generating eosinophils in vitro in which bone marrow cells were cultured in SCF and Flt3L for the first 4 d to expand the progenitor cells, followed by a switch to IL-5 in the culture to promote eosinophil development; the effect of IL-33 was assessed during these later stages of culture. Because our data suggest that IL-33 is upstream of IL-5, we sought to examine the effects of IL-33 during the initial progenitor expansion phase. We cultured freshly isolated bone marrow cells with SCF and Flt3L, IL-5, or IL-33 for 3 d and assessed for changes in EoPre and EoM populations by flow cytometry (Fig. 5A, 5B). As shown in Fig. 5A and quantified in Fig. 5B, both populations were maintained after 3 d when cultured in SCF and Flt3L or IL-5. In contrast, IL-33 drastically increased the EoPre population in WT and IL-33 KO cultures, but not in ST2 KO cultures. Assessment of cytology after cell sorting of the EoPre and EoM populations showed clear multilobed nuclear morphology consistent with eosinophils but poor granule staining in the EoPre, whereas the EoM population possessed clear eosin-stained granules (Fig. 5C). Moreover, bone marrow cultured with IL-33, but not SCF and Flt3L or IL-5, led to a substantial increase in IL-5Rα mean fluorescence intensity on the EoPre (Fig. 5D), further indicating their commitment to an eosinophil lineage. When CD11b was specifically used instead of the full lineage mixture, a similar decline in EoMs and expansion of EoPre was observed, as well as an upregulation of IL-5Rα on the EoPre population (Supplemental Fig. 4C–F).

Although the difference in granularity we see by cytology is consistent with the differences in side scatter we observe in our flow cytometry analysis, it contrasts with the previously described EoPre, which have been reported to possess granule proteins (29). Moreover, IL-5Rα can also be expressed on neutrophils under certain conditions (30). Therefore, to further confirm that this population we considered to be EoPre cells was truly within the eosinophil lineage, we interrogated the gene expression profiles of these EoPre cells to assess whether they expressed eosinophil-associated genes. In this experiment, we used the EoM population as a positive control and bone marrow neutrophils as a negative control lying still within the granulocyte lineage. Whereas bone marrow neutrophils, EoPre, and EoMs had similar expression of granulocyte-associated genes Csf2ra (GM-CSFRα) and Spi1 (PU.1), EoPre and EoMs had significantly lower expression of the neutrophil-associated gene Csf3r (G-CSFR) than neutrophils (Fig. 5E). Instead, the EoM and EoPre populations expressed significantly higher levels of the eosinophil-associated genes Epx (eosinophil peroxidase), Prg2 (major basic protein), Cebpa (C/EBPα), Gata1 (GATA-1), and Gata2 (GATA-2) than neutrophils (Fig. 5E). Furthermore, we noted that the EoPre population expressed intermediate levels of these eosinophil-associated genes, suggesting that they possessed an immature eosinophil phenotype. Together, these findings suggest that IL-33 may function as a growth factor for the early commitment toward the EoPre population and/or drive their expansion.

To test this idea further, we cultured bone marrow cells in SCF and Flt3L with or without IL-33 for 3 d before reculturing them in IL-5 for another 7 d; cell populations were assessed by flow cytometry at various stages of the culture. Although no difference was seen in the GMP populations (data not shown), WT cultures grown with IL-33 exhibited high numbers of EoPre populations at early stages of culture, as well as a rapid increase in total EoMs that was sustained throughout the entire culture period (Fig. 5F). In contrast, the WT culture without early IL-33 treatment showed little change in EoPre cells, and increases in EoM appeared only after day 10 of culture. ST2 KO cells showed no expanded EoPre on IL-33 treatment and generated even fewer EoMs over the course of culture with IL-5. Taken together, these data suggest that IL-33 precedes the need for IL-5 and functions mainly to promote not only the expansion of EoPre cells but also IL-5Rα upregulation, which then sustains the EoM pool if IL-5 is provided.

IL-33 has emerged as an important cytokine in allergic diseases, largely because of its potential to activate cells that are hallmarks of allergy, including eosinophils, mast cells, and basophils (31). Outside of allergy, IL-33 has also been proposed to be involved in bacterial and viral infections, tumorigenesis, autoimmunity, fibrosis (32), and more recently, hematopoiesis (23, 33). In this article, we define a previously unappreciated mechanism for IL-33 in regulating eosinophil commitment.

Our data demonstrate that IL-33 directs the eosinophil compartment by expanding the EoPre frequency and upregulating IL-5Rα to license the responsiveness of these precursors to IL-5 within the bone marrow. Importantly, the defects in basal eosinophil populations we identified in the IL-33 KO and ST2 KO mice strongly implicate a homeostatic contribution of this cytokine that functions outside of a disease pathogenesis setting. Indeed, the previously defined function of IL-33 as an alarmin released upon tissue damage or injury (34) seems unlikely to explain such homeostatic regulation in healthy animals. Thus, the underlying alteration in eosinophil homeostasis in IL-33 KO and ST2 KO mice we describe in this article may impact the numerous interpretations others have made from studying the IL-33 KO and ST2 KO mice in disease models; consequently, reconsideration of some conclusions made from data using eosinophil numbers as a key response readout in previous studies may be warranted.

Although our data clearly show a role for IL-33 in the bone marrow, other studies examining ST2 expression on hematopoietic stem and progenitor cells in the bone marrow have shown conflicting findings. Initial reports claimed that ST2 was present on multiple subsets of Lin⁻ckit⁺ progenitor cells, including GMP cells (23). In contrast, recent work using a spontaneous mutant mouse model of myeloproliferative neoplastic tumorigenesis failed to detect ST2 expression on a variety of hematopoietic stem cell and progenitor lineages (35). However, this work did demonstrate a functional role for IL-33 in regulating myeloid cell fate in this model. From our own characterization of the Siglec-F–expressing compartment (Supplemental Fig. 2) in which we define the stages of eosinophil development based on surface markers as well as size and granularity, ST2 does appear to be expressed on GMP-like cells and EoMs, but its expression is lost at the EoPre stage. One possibility for this diminished ST2 expression on the EoPre population is that IL-33 binding leads to receptor internalization, as occurs when IL-5 binds to IL-5R (26). ST2 can also be shed from the surface of cells after activation (36). Alternatively, differences in gating strategy in terms of how to define GMP cells has influenced our conclusions; certainly, the expression of low Siglec-F levels on a progenitor cell, which we are choosing to term GMP-like because of the other marker profiles (Lin⁻Sca1⁻IL-5Rα⁻c-kit⁺CD34⁺), has not been previously described to the best of our knowledge. Interestingly, publicly available gene expression microarray data sets do support the potential for expression of Siglec-F on GMP cells, as well as on other progenitor subsets (http://biogps.org/gene/233186).

Recently, significant increases in EoMs, CD34⁺ hematopoietic progenitor cells, and committed eosinophil progenitors were reported within the lungs of asthmatic patients; this report further showed that IL-33 treatment primed the CD34⁺ population for migration toward stromal cell–derived factor 1α (CXCL12) (37). It has also been shown that CD34⁺ cells circulating in the blood express ST2, and activation of these cells with IL-33 leads to significant production of Th2-type cytokines, including IL-5 (38). Although our data define the responses to IL-33 that occur within the bone marrow compartment, the recruitment of progenitor populations after IL-33 exposure might support similar eosinophil developmental processes that occur within peripheral tissues; however, whether this might occur under homeostasis, as has been suggested for eosinophil maintenance (39), or only under disease conditions remains to be determined.

In agreement with others, we show that administration of IL-33 is capable of inducing significant elevation of IL-5, both systemically and within the bone marrow itself (Fig. 3G, 3H). However, the source of this IL-5 remains to be determined. Because the mice in this study were not primed toward a Th2 adaptive response, it seems reasonable to predict that innate cells, such as ILC2s, might be a significant source. ST2-expressing ILC2 populations that produce IL-5 were previously shown to regulate eosinophil homeostasis in the intestine and lung, but this production was spontaneous; moreover, these cells were shown to be largely absent from the bone marrow compartment (39). Interestingly, a Sca-1⁺ precursor population within the bone marrow that expresses ST2 and produces IL-5 in response to IL-33 has been reported (40), suggesting that both the expansion of the IL-5Rα⁺ EoPre population and the elevation of IL-5 itself could occur locally within the bone marrow and be driven by IL-33. In our in vitro cultures, IL-5 was detected in the media after 3 d of culture with SCF, Flt3L, and IL-33 (54.6 ± 6.7 pg/ml) but was undetectable (<31.25 pg/ml) with SCF and Flt3L alone. This might explain the rapid transition from precursors toward EoMs that we observed. Importantly, through the use of the IL-5–transgenic mouse, our studies clearly define a role for IL-33 signaling in eosinophil frequency that is independent of the IL-33–induced IL-5 response and establishes that the contribution of IL-33 lies upstream of IL-5–driven eosinophilopoiesis. Similarly, it has previously been shown that IL-33 cannot elicit eosinophilia in IL-5 KO mice, helping to define that IL-5 lies downstream of IL-33–driven responses (18). Furthermore, blocking IL-5 in vivo prevented the increased mature blood eosinophil expansion that occurs in response to IL-33 but failed to prevent the observed decreases in the GMP-like cells in the bone marrow. Currently, we postulate that the GMP-like cells, shown as expressing ST2 in Supplemental Fig. 2, are being driven toward the EoPre phenotype by IL-33, and that the high expression of IL-5R on these cells combined with the elevated IL-5 (from currently unknown cells) results in rapid increases in EoMs. This model is supported by our studies in which we neutralized IL-5 and affected only the IL-33–induced changes in EoMs, but this concept requires further studies to define fully.

Previous studies using in vitro culture approaches to eosinophilopoiesis have led to confusing findings. Whereas Stolarski et al. (17) claimed that culture with IL-33 alone for 5 or 8 d was sufficient to induce eosinophil differentiation, Dyer et al. (18) concluded that IL-33 did not support eosinophilopoiesis and, instead, antagonized the effects of IL-5 and promoted monocyte differentiation. Data from our initial in vitro studies were in agreement with Dyer et al. (18) in that IL-33 failed to sustain the cell cultures to day 5 or beyond, with cells dying through apoptosis (data not shown). We also observed that the addition of IL-33 to SCF and Flt3L during the first 3 d of culture led to significant expansion of total cell numbers at day 3 compared with cultures with SCF and Flt3L alone and that this was not seen with ST2KO cells. As shown in Fig. 5, these cells, although clearly not EoMs, possessed hallmark characteristics indicative of EoPre. Interestingly, they expressed mRNA for eosinophil granule proteins and eosinophil-associated transcription factors, but lacked eosin-positive granules. Because of this and the clearly defined nuclear morphology of an eosinophil, we postulate that these cells represent an alternative precursor state than those defined from long-term IL-5 culture approaches (29). Initially, when these cultures were switched into IL-5, many of these expanded cells (>50%) were dying by apoptosis at day 7, even when a 10-fold higher concentration of IL-5 was used (data not shown). We subsequently found that readjusting the cell density with each media change helped facilitate survival of these cells, and these cultures were used as the source of the data shown in Fig. 5F. Potentially, because the IL-33–treated cultures have significantly more IL-5Rα⁺ EoPre cells, consumption of the available IL-5 and outgrowth of other cell types might explain the elevated monocyte numbers seen by Dyer et al. (18). An interesting aspect of our data in Fig. 5F is that we observed a rapid expansion of both the EoPre and EoM populations within the first few days, but that the overall numbers of EoM plateaued, rather than continued to increase after removal of IL-33. Although further work is required to understand this fully, one possibility is that the pool of precursors became a limiting factor and that the balance between development of mature EoMs and the apoptotic death we observed in the expanded IL-33 cultures simply maintained the EoM populations during the IL-5 treatment. Nonetheless, the overall numbers of EoMs generated by IL-33 treatment during the early days of bone marrow cultures significantly outpaced those cultures without IL-33 and supports the concept that early expansion of precursors supports a more rapid establishment of eosinophil populations.

Our findings showing that NJ.1638/ST2KO mice have diminished eosinophil numbers (Fig. 4F) also provide in vivo evidence suggesting that IL-33 signaling positively regulates eosinophil differentiation rather than antagonizing it. Furthermore, we did not find basal differences in blood monocyte subsets between IL-33 KO and ST2 KO mice or WT mice (Fig. 1E). Instead, we did observe a diminished frequency of blood neutrophils that, unlike the eosinophil response, was not altered by exogenous IL-33 treatment (Fig. 1E, Supplemental Fig. 1). Sustaining eosinophil and neutrophil populations requires a competent GMP population, but we did not see any significant effect on the homeostatic frequency of GMP or GMP-like cells in the bone marrow of either ST2 KO or IL-33 KO (Fig. 3F, Supplemental Fig. 3B), although exogenous IL-33 treatment did diminish the GMP-like pool (Figs. 3F, 4D). Intriguingly, ST2KO bone marrow did seem to generate fewer EoM cells by day 10 of culture. Although we saw no obvious difference in the frequency of either EoPre (Fig. 5B) or basal IL-5R levels (Fig. 5D) between ST2KO and WT that did not receive IL-33 treatment, subtle differences in the basal numbers of GMP-like cells or their basal levels of IL-5R expression could explain this. Alternatively, WT bone marrow might contain a cell capable of generating endogenous IL-33 that supports eosinophil development during the course of culture. These ideas require further investigation to define properly. Previous reports suggested that IL-33 might directly influence neutrophils during inflammation (41), and mice overexpressing IL-33 under the CMV promoter exhibited elevated eosinophilia and neutrophilia (42). We have been unable to observe ST2 expression on neutrophils (data not shown), and our data show that the doses of IL-33 required to increase eosinophils are separable from any effect on neutrophils. Interestingly, subsequent studies related to IL-33–induced neutrophil responses have proposed an indirect regulation via activation of mast cells (24). In contrast with the neutrophils, IL-33 treatment did increase the ILC2 population in our study, indicating that the doses of IL-33 needed to influence these lymphoid cells were likely similar to the eosinophils.

Eosinophils are involved in many diseases, including asthma, eosinophilic esophagitis, atopic dermatitis, and hypereosinophilic syndromes. Since IL-5 is needed to support eosinophil development and survival, two Abs targeting IL-5 (mepolizumab and reslizumab) and one targeting the IL-5R (benralizumab) have been therapeutically tested. Particularly in the setting of severe eosinophilic asthma, these therapies have shown success in reducing symptoms and dependence on oral glucocorticoids (43). Targeting IL-33 and/or ST2 has already been extensively suggested for allergic diseases (44); importantly, our findings suggesting that IL-33 participates in maintaining the eosinophil pool upstream of IL-5 further predict the usefulness of targeting IL-33 and/or ST2 in disease settings similar to those targeted by IL-5–based therapies.

In conclusion, our data demonstrate a previously unappreciated role for IL-33 in supporting eosinophil development. Because basal homeostasis is affected in both ST2 KO and IL-33 KO mice, this role seems to represent a homeostatic function for IL-33. Furthermore, our findings suggest that an elevation in IL-33 levels can induce eosinophil development. Mechanistically, we show that IL-33 is vital for promoting IL-5Rα upregulation on EoPre to facilitate their responsiveness to IL-5. Moreover, although IL-33 signaling does not appear to be a necessary checkpoint in eosinophil development, our data demonstrate that it functions as an important regulator over the numbers of EoPre, as well as the output of eosinophils from the bone marrow.

We thank Dr. James Lee and Dr. Nancy Lee (Mayo Clinic) for providing the IL-5–transgenic mice and Dr. Dirk Smith (Amgen) for providing the IL-33 KO mice. We also thank Dr. Mendy Miller for assistance in editing this manuscript.

The authors have no financial conflicts of interest.