Resident intestinal eosinophils constitutively express antigen presentation markers and include two phenotypically distinct subsets of eosinophils

Summary

Intestinal eosinophils are implicated in homeostatic and disease‐associated processes, yet the phenotype of intestinal tissue‐dwelling eosinophils is poorly defined and their roles in intestinal health or disease remain enigmatic. Here we probed the phenotype and localization of eosinophils constitutively homed to the small intestine of naive mice at baseline, and of antigen‐sensitized mice following intestinal challenge. Eosinophils homed to the intestinal lamina propria of naive mice were phenotypically distinguished from autologous blood eosinophils, and constitutively expressed antigen‐presenting cell markers, suggesting that intestinal eosinophils, unlike blood eosinophils, may be primed for antigen presentation. We further identified a previously unrecognized resident population of CD11c^hi eosinophils that are recovered with intraepithelial leucocytes, and that are phenotypically distinct from both lamina propria and blood eosinophils. To better visualize intestinal eosinophils in situ, we generated eosinophil reporter mice wherein green fluorescent protein expression is targeted to both granule‐delimiting and plasma membranes. Analyses of deconvolved fluorescent z‐section image stacks of intestinal tissue sections from eosinophil reporter mice revealed eosinophils within intestinal villi exhibited dendritic morphologies with cellular extensions that often contacted the basement membrane. Using an in vivo model of antigen acquisition in antigen‐sensitized mice, we demonstrate that both lamina propria‐associated and intraepithelium‐associated eosinophils encounter, and are competent to acquire, lumen‐derived antigen. Taken together these data provide new foundational insights into the organization and functional potential of intestinal tissue‐dwelling eosinophils, including the recognition of different subsets of resident intestinal eosinophils, and constitutive expression of antigen‐presenting cell markers.

Abbreviations

GFP

green fluorescent protein

GM‐CSF

granulocyte–macrophage colony‐stimulating factor

IE

intraepithelial

IL‐2

interleukin‐2

LP

lamina propria

OVA

ovalbumin

WT

wild‐type

Introduction

Eosinophils are innate immune granulocytes initially thought to function exclusively as end‐stage effector cells, recruited within the context of parasitic helminth infections or allergic diseases such as asthma. However, collective data now reveal that eosinophils naturally home to and reside within several tissues at baseline, including (but not limited to) thymus, mammary gland, uterine lining, adipose tissues and all regions of the gastrointestinal tract with the exception of the oesophagus (reviewed in refs 1, 2). Moreover, functions of tissue‐dwelling eosinophils are far more multifaceted than previously appreciated. It is now known that eosinophils contribute to tissue homeostasis, metabolism and immune regulation in both disease and the steady state. For example, adipose tissue eosinophils participate in beige fat thermogenesis and glucose homeostasis through regulation of alternatively activated macrophages,3, 4, 5 and bone marrow eosinophils are required for adjuvant‐induced B‐cell priming6 and maintenance of memory plasma cells.7 These findings suggest eosinophils that reside within distinct tissue niches exhibit different phenotypes (particularly when compared with the more commonly studied blood, or bone‐marrow‐derived eosinophils), reflective of their specialized functions.

An evolving recognition of the expanded functions of tissue‐dwelling eosinophils is particularly notable within the intestinal tract. Driven by the chemokine eotaxin‐1, eosinophils homing to gastrointestinal tissues precede colonization of the normal gastrointestinal flora, and occurs irrespective of an intact adaptive immune system.8, 9 Comparisons between wild‐type (WT) and eosinophil‐deficient mice suggest that at baseline, intestinal eosinophils promote Peyer's patch development, intestinal mucus secretions and generation of mucosal IgA, revealing homeostatic functions for eosinophils in normal gastrointestinal development and mucosal immunity.10, 11 In human diseases, eosinophilic inflammation is associated with multiple gastrointestinal disorders, ranging from food allergies to inflammatory bowel diseases, and also characterizes a heterogeneous group of diseases collectively called eosinophilic gastrointestinal diseases (i.e. eosinophilic oesophagitis, eosinophilic gastroenteritis and eosinophilic colitis) (reviewed in refs 12, 13).

The majority of studies to date have probed the phenotype and function of eosinophils in blood, derived from bone marrow ex vivo, or recovered from interleukin‐5 (IL‐5) over‐expressing transgenic mice. Fundamental studies of tissue‐resident eosinophils are largely lacking; therefore, despite their clear associations with inflammatory gastrointestinal diseases, the phenotype of intestinal tissue‐dwelling eosinophils is poorly defined and the roles of eosinophils in gastrointestinal health or disease remain enigmatic. This study directly assessed the phenotype and localization of eosinophils that have constitutively homed to the small intestine of naive WT mice. Here we demonstrate that lamina propria (LP) eosinophils are phenotypically distinct from peripheral blood eosinophils, with LP eosinophils constitutively expressing surface markers consistent with antigen presentation functions (i.e. MHC II and CD80), suggesting that LP eosinophils, unlike blood eosinophils, may be primed for antigen presentation. Moreover, we identify a resident population of eosinophils, recovered with intestinal intraepithelial (IE) leucocytes, that has previously not been recognized, and is phenotypically distinguishable from both blood and LP eosinophils. To more sensitively visualize resident intestinal eosinophils and their cellular morphologies at baseline in situ, we generated eosinophil reporter mice by targeting green fluorescent protein (GFP) specifically to eosinophil granule‐delimiting membranes and plasma membranes on a background of mice endogenously expressing membrane‐targeted tdTomato in all other cell lineages. Fluorescence microscopy revealed eosinophils exhibiting predominantly dendritic morphologies with extensive reaches that, within villi, often included contact points with the basement membrane. We further show, using an in vivo antigen sensitization model, that both LP‐associated and IE‐associated eosinophils of antigen‐sensitized mice acquired intestinal lumen‐derived antigen in vivo. These findings provide new phenotypic insights into tissue‐resident intestinal eosinophils, including the recognition that LP eosinophils are primed for antigen presentation functions at baseline, and identify a novel subset of intestinal eosinophils constitutively and intimately associated with the intestinal IE niche.

Material and methods

Animals

All studies received previous approval by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee. Female BALB/c mice, 6–8 weeks old, were obtained from Charles River Laboratory (Wilmington, MA). C57BL/6 EoCre^+/− mice14 were a kind gift from Dr Jamie Lee, Mayo Clinic, Arizona. C57BL/6 mTmG^fl/fl mice were obtained from the Jackson Laboratories (Bar Harbor, ME), and EoCre^+/− mTmG^fl/fl mice were generated and maintained in‐house. All mice were maintained under conventional animal housing conditions with food and water provided ad libitum.

Microscopy

For histological analyses of paraffin sections, formalin‐fixed small intestinal segments underwent standard tissue processing. Five micrometres of paraffin‐embedded tissue sections were stained with Fast Green (0·2% in 70% ethanol, Sigma F‐7252) and Neutral Red (0·5% in water, Fluka 72210) to identify eosinophils. For immunofluorescence analyses of frozen sections, small intestines were fixed overnight in 4% paraformaldehyde at 4° and cryoprotected in 30% sucrose before embedding in OCT. Frozen sections were mounted in medium supplemented with Hoechst dye to visualize nuclei. Image stacks were acquired using an Olympus BX62 microscope with a 60× UPlanApo objective, numerical aperture 1.42. SlideBook 6 (Intelligent Imaging Innovations (3i), Denver, CO) was used for data acquisition and deconvolution. Three‐dimensional rendering was performed using Volocity (PerkinElmer, Hopkinton, MA) and preparation of images for publication was performed using ivision software (BioVision Technologies, Exton, PA) or volocity.

Isolation of intestinal leucocytes

Intestinal leucocytes were isolated as previously described.15 Briefly, mouse intestines were harvested, depleted of Peyer's patches, flushed with CMF buffer (Hank's balanced salt solution (HBSS^−/−), 5% fetal bovine serum, 1 mm HEPES, 2·5 mm NaHCO₃), opened longitudinally, gently wiped to remove mucus, and cut into 5‐mm pieces. The pieces were incubated in dithiothreitol (DTT) buffer (HBSS^−/−, 10% fetal bovine serum, 1 mm HEPES, 2·5 mm NaHCO₃, 1 mm DTT) twice for 20 min each with shaking at 37° to release IE cells. Tissue pieces remaining after IE cell isolation were further processed to recover LP cells. Tissue pieces were washed with CMF buffer and incubated in an EDTA‐containing buffer (HBSS^−/−, 1 × penicillin/streptomycin, 1·3 mm EDTA) twice for 30 min each at 37° while shaking. Supernatants were discarded, and tissue pieces were washed with CMF buffer and transferred to collagenase‐containing buffer (RPMI, 10% fetal bovine serum, 1 mm CaCl₂, 1 mm MgCl₂, 1 × penicillin/streptomycin, 100 U/ml collagenase). Pieces were then passed through a 70‐µm pore‐size filter and resuspended in 8 ml of 44% Percoll (Sigma‐Aldrich, St Louis, MO) and layered over 5 ml of 67% Percoll. Gradients were spun for 20 min at 1000 × g and eosinophils were recovered from the interphase and washed.

Analysis of eosinophils from whole blood

For comparisons of intestinal eosinophils with autologous blood eosinophils, whole blood was collected by cardiac puncture immediately upon euthanasia. Red blood cell lysis was accomplished using BD PharmLyse lysing buffer (BD Biosciences, San Jose, CA) before staining for flow cytometry.

Flow cytometry

The following stains were used for flow cytometry: LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Invitrogen, Grand Island, NY), CD45‐phycoerythrin‐Cy7 (30‐F11), Siglec F‐phycoerythrin (E50‐2440) from BD Biosciences (San Jose, CA), and CD11c‐FITC (N418), MHCII‐FITC (M5/114.15.2), CD80‐allophycocyanin (16‐10A1) and CD11b‐allophycocyanin (M1/70) from BioLegend (San Diego, CA). Data were acquired using a BD LSR II (BD Biosciences, Franklin Lakes, NJ) flow cytometer, or a Gallios flow cytometer with kaluza software (Beckman Coulter, Indianapolis, IN). Cell sorting was performed using a FACSAria II (BD Biosciences). Data were analysed using flowjo analysis software.

Antigen sensitization and intestinal ligated loops surgical model

BALB/c WT mice (6–8 weeks old) were sensitized with 100 µl intraperitoneal injections containing 40 µg ovalbumin (OVA, Grade VI; Sigma‐Aldrich) mixed with 0·4 mg aluminium hydroxide and 0·4 mg magnesium hydroxide (imject; Thermo Scientific, Waltham, MA) in sterile phosphate‐buffered saline on days 0, 7 and 14. On day 21, to deliver antigen directly into the intestinal lumen in vivo in live mice, intestinal ligated loops surgery was performed as described.16 Briefly, mice were maintained under isoflurane anaesthesia while terminal ileum adjacent to the caecum was accessed through a small abdominal incision, and two intestinal loops were created by placing three sterile sutures, spaced approximately 5 cm apart. After injection of 20 µg of Alexa 647‐labelled or unlabelled OVA, in a total volume of 200 µl, into each loop, the peritoneum and skin were closed in one layer with sterile suture. Mice were maintained under anaesthesia for an additional 45 min to allow sufficient time for antigen uptake by intestinal leucocytes before the intestinal loops were excised and intestinal cells were isolated.

Statistical analyses

Unless otherwise indicated, direct comparisons between two groups used two‐tailed Student's t‐tests, and P‐values < 0·05 were considered statistically significant. Statistical analyses were performed with Microsoft excel or graphpad prism (San Diego, CA). Combined data included in the text and error bars within graphs show averages ± standard deviation (SD) from the mean.

Results

Resident intestinal LP eosinophils are phenotypically distinct from blood eosinophils

Using two complementary staining techniques for visualization of eosinophils (i.e. haematoxylin and eosin, or fast green and neutral red), intact eosinophils were readily identified within the LP of all regions of the small intestine (duodenum, jejunum and ileum) of eosinophil over‐expressing IL‐5 transgenic mice (not shown) and WT mice (Fig. 1a). Eosinophils were observed disseminated throughout the LP, both proximate to crypts and dispersed throughout intestinal villi. To investigate the phenotype of intestinal eosinophils, we modified an established protocol for recovery of leucocytes from LP,17 as described in the Materials and methods. By gating on live, CD45⁺ SSC^hi SiglecF^hi cells (see Supplementary material, Fig. S1), an eosinophil population was identified within the small intestinal LP cells from WT mice (Fig. 1b), representing 15(± 8)% of total live, CD45⁺ leucocytes in BALB/c WT mice (n = 12). This eosinophil gating strategy was validated using eosinophil over‐expressing IL‐5 transgenic mice (not shown), by spiking samples with purified splenic eosinophils (not shown), by morphological confirmation of sorted cells,15 and by demonstrating the absence of CD45⁺ SSC^hi SiglecF^hi cells in eosinophil‐deficient ΔdblGATA1^−/− mice (see ref. 15 and Fig. 1c).

Figure 1.

Resident intestinal eosinophils are phenotypically distinct from autologous blood eosinophils. (a) Paraffin sections of small intestinal tissue from BALB/c wild‐type mice stained with fast green and neutral red. Eosinophils (arrows) are identified within lamina propria associated with crypts and villi. (b,c) SSC^hi SiglecF^hi eosinophils are identified by flow cytometry of lamina propria (LP) cells recovered from digested small intestinal tissues from wild‐type (b) or eosinophil‐deficient (c) mice. Plots are gated on live, CD45⁺ leucocytes (see Supplementary material, Fig. S1). (d–f) Surface expression of Siglec F (d), CD11b (e) or CD11c (f) on eosinophils from autologous blood (empty histogram) or intestinal lamina propria (shaded histogram). Dashed line, isotype control staining.

Confirming our previous findings15 and those of others18, LP eosinophils stained positively for CCR3 (not shown), CD11b (Fig. 1e) and CD11c (Fig. 1f). CD11c expression distinguishes LP eosinophils from blood eosinophils, which do not express CD11c (Fig. 1f). Furthermore, expressions of Siglec F and CD11b were increased on LP eosinophils in comparison to autologous blood eosinophils (Fig. 1d,e). These findings complement and extend previous studies in mice and in healthy humans, which found intestinal eosinophils to be phenotypically distinct from blood eosinophils,18 and to exhibit an activated phenotype based on their cytokine expression and degranulation status.19

Intestinal LP eosinophils constitutively express antigen‐presenting cell markers

Cytokine‐primed and antigen‐primed eosinophils are competent to engage in professional antigen‐presenting cell functions, as demonstrated in vitro and in vivo by their capacity to take up and process antigen, traffic to draining lymph nodes, and activate naive, antigen‐specific T lymphocytes in an MHC II‐dependent manner (reviewed in refs 20, 21, 22). Eosinophils from the blood exhibit little to no detectable antigen presentation molecules (i.e. MHC II and co‐stimulatory molecules) on their cell surfaces; however, ex vivo exposure to granulocyte–macrophage colony‐stimulating factor (GM‐CSF) elicits surface expression of MHC II on eosinophils isolated from either mice23 or humans.24 GM‐CSF is constitutively expressed within intestinal tissues,25 therefore we queried whether eosinophils resident within the small intestine might be primed for antigen presentation functions by assessing surface expressions of MHC II and the co‐stimulatory molecule CD80 on autologous eosinophils from steady‐state blood and intestinal LP. As anticipated, blood eosinophils exhibited a resting phenotype of low to no detectable surface MHC II (Fig. 2a top panel). In contrast, surface‐expressed MHC II was detected on autologous LP eosinophils from naive WT mice (Fig. 2a middle panel, and b). Of note, expression of MHC II on intestinal eosinophils was modest in comparison to other antigen‐presenting cells (compare with SiglecF⁻ MHCII⁺ cells in Fig. 2a bottom panel). In parallel with MHC II expression, LP eosinophils also exhibited cell surface expression of the co‐stimulatory molecule CD80 (Fig. 2c).

Figure 2.

Resident intestinal lamina propria (LP) eosinophils constitutively express an antigen‐presenting cell phenotype. (a) MHC II expression on eosinophils gated from autologous whole blood (top), LP eosinophils (middle) and MHC II⁺ non‐eosinophils from the LP (bottom). (b) MHC II expression on LP eosinophils compared with isotype control staining. (c) CD80 expression on LP eosinophils.

A population of resident eosinophils is recovered with IE lymphocytes

Separated from the LP by the basement membrane, a single layer of epithelial cells forms the barrier between the external environment (intestinal lumen) and the intestinal LP. Interspersed within the epithelium are distinct subsets of IE leucocytes, the most predominant of these cells being CD8⁺ T cells. Although eosinophils are recognized to be constitutively present within gastrointestinal LP, their infiltration into the epithelial monolayer is thought to be restricted to inflammatory conditions, based on cross‐sectional histological analyses.26 To determine whether a more quantifiable and sensitive approach might reveal basal levels of eosinophils within the IE compartment, IE cells were recovered and analysed by flow cytometry. Using the same gating strategy described above, eosinophils were recovered with IE cells, where they represented 16% (± 7%) of live IE leucocytes from naive, WT mice on the BALB/c background (Fig. 3a, n = 12). Efficient separation of LP and IE compartments was confirmed by light microscopy of tissue pieces following removal of the epithelium and IE cells (see Supplementary material, Fig. S2a), and by comparing recoveries of CD8⁺ and immunoglobulin⁺ cells as a measure of CD8⁺ IE lymphocytes and LP B cells, respectively (see Supplementary material, Fig. S2b).

Figure 3.

A subset of eosinophils that is phenotypically distinct from both blood and lamina propria (LP) eosinophils is recovered with intestinal intraepithelial (IE) leucocytes. (a,b) Flow cytometry of IE cells isolated from BALB/c wild‐type (a) or C57BL/6 EoCre^+/− mTmG^fl/fl (b) mice. Plots shown are gated on total live, CD45⁺ cells. (b, i) Morphology and HEMA 3 staining of GFP⁺ cells sorted from IE preparations confirm the identity of IE eosinophils. (c) IE leucocytes were isolated from human resected tissue recovered from a patient following ileocaecectomy. Total IE cells recovered were spun onto a cytospin and stained using the HEMA 3 staining kit. (d–h) Surface expression of Siglec F (d), CD11b (e), CD11c (f), MHC II (g) and CD80 (h) on blood, LP and IE eosinophils recovered from BALB/c wild‐type mice is expressed as delta mean fluorescence intensity (ΔMFI). *P < 0·05. Data are pooled from n = 12 (LP, IE) mice. Pooled data expressing ΔMFI of autologous blood from three of the mice is provided for reference (open bars). (b, i; c) Scale bars, 10 µm.

To further confirm the identity of the live, CD45⁺ SSC^hi SiglecF^hi cells recovered with the IE preparations, we generated eosinophil reporter mice by crossing eosinophil‐targeted Cre recombinase‐expressing mice (EoCre) with Cre recombinase‐dependent membrane‐targeted GFP reporter mice (mTmG^fl/fl), both on the C57BL/6 background. Like BALB/c mice, live, CD45⁺ SSC^hi SiglecF^hi cells were identified in both the LP (not shown) and IE (Fig. 3b, left panel) preparations from EoCre^+/−mTmG^fl/fl mice. Although compared with BALB/c mice the frequencies of eosinophils within both LP and IE compartments were lower in the cohort of C57BL/6 reporter mice investigated here (3 ± 1% and 2 ± 1% of live, CD45⁺ leucocytes from LP and IE, respectively, n = 5), these data demonstrate that the recovery of eosinophils with IE leucocytes is not unique to BALB/c mice, but extends to C57BL/6 mice as well. GFP expression (Fig. 3b, right panel) and cell sorting (Fig. 3b, i) confirmed the identity of this IE‐associated cell population to be eosinophils.

Finally, pathologically normal margins of resected terminal ileum tissues recovered from patients undergoing ileocaecectomy for benign adenoma were subjected to the IE isolation procedure and assessed by light microscopy for the presence of eosinophils. Similar to the mouse, human IE leucocyte preparations included eosinophils (Fig. 3c, representative of n = 2).

Eosinophils recovered with IE leucocytes are phenotypically distinct from LP eosinophils

Phenotypically, IE eosinophils exhibited higher levels of surface‐expressed Siglec F, CD11b and CD11c when compared with eosinophils recovered from the LP, suggesting that IE eosinophils may exist in a higher activation state than LP eosinophils (Fig. 3d–f). Eosinophils isolated with IE cells also exhibited higher levels of surface‐expressed CD80 (Fig. 3h); however, MHC II expression by IE eosinophils was lower than that observed on LP eosinophils (Fig. 3g). It is unlikely that enzymatic tissue digestion is responsible for the differences in the surface phenotypes between LP (isolated post collagenase digestion) and IE (collagenase unexposed) eosinophils, as collagenase treatment has been shown to have no or minor effects on surface molecule expression, and in contrast to the decreased expression of CD11c observed on collagenase‐exposed LP eosinophils compared with IE eosinophils shown here (Fig. 3f), Autengruber et al.27 reported that collagenase exposure causes a modest increase in CD11c expression on immune cells. Moreover, in parallel experiments where isolated IE cells were subjected to the collagenase digestion step in parallel with LP tissues, CD11c expression remained elevated on IE eosinophils relative to CD11c expression on LP eosinophils (see Supplementary material, Fig. S3). Taken together, these data suggest that IE eosinophils represent a phenotypically distinct eosinophil subpopulation within the intestine.

Villous eosinophils extend dendritic processes that contact the basement membrane

EoCre^+/− mTmG^fl/fl mice provide a sensitive tool to visualize the morphology and localization of LP intestinal eosinophils in situ. Endogenous membrane‐localized tdTomato expression in non‐eosinophils enables visualization of cell borders, including that of epithelial cells, and targeting GFP expression to both granule‐limiting membranes and plasma membranes optimizes the visualization of eosinophil cellular extensions. Frozen sections of small intestinal tissues isolated from EoCre^+/− mTmG^fl/fl mice were analysed by fluorescence microscopy and revealed intestinal eosinophils (green) that generally exhibited elongated and/or dendritic morphologies. Eosinophils were readily observed within LP proximate to intestinal crypts (Fig. 4a) as well as throughout the full length of the villi (Fig. 4b,c). Eosinophil cell bodies were not readily observed within the epithelial monolayer (one GFP⁺ cell was identified in 40 high‐power fields examined). However, histopathological approaches probably lack the sensitivity to detect baseline levels of IE eosinophils, due to the low frequency of eosinophils among the already sparse CD45⁺ IE leucocyte population, particularly in mice on the C57BL/6 background. Of note, a sizeable proportion of eosinophils observed within villi localized at or just below the basement membrane. To fully appreciate the morphology and localizations of eosinophils and their dendritic processes, which extended through multiple focal planes, image stacks were captured using a 0·4‐µm Z‐step and deconvolved to provide three‐dimensional details. Deconvolved image stacks revealed villus eosinophils exhibiting dendritic extensions that often contacted or breached the basement membrane (Fig. 5).

Figure 4.

Visualization of eosinophils in situ in EoCre^+/− mTmG^fl/fl eosinophil reporter mice. Frozen sections of small intestinal tissues recovered from naive EoCre^+/− mTmG^fl/fl mice were mounted with Hoechst‐containing media and analysed by fluorescence microscopy. Membrane‐targeted tdTomato expression (left column) demarcates cell membranes of non‐eosinophils. GFP⁺ eosinophils (middle column, and arrows in merged images) are visualized within the lamina propria (LP) proximate to crypts (Cr, a) and throughout villi (b,c), often in apparent direct contact with the basement membrane (c, ci). Original magnification, 600×. Scale bars 20 µm (a‐c) and 10 µm (ci).

Figure 5.

Villous eosinophils extend dendritic processes that contact the basement membrane. (a) A deconvolved image stack (0·4‐µm z‐step) reveals eosinophil dendritic processes that traverse multiple focal planes. Even numbered slices of the image stack between 8 and 30 are shown. Note that slices 14 through 20 confirm continuity between the cell body visible in slice 10 and the dendritic processes in contact with the basement membrane (demarcated by arrows) in slices 20–24. Original magnification, 600×. Scale bars, 20 µm (top image) and 10 µm (slices). (b) Three‐dimensional rendering of panel (a). Scale bars, 10 µm.

Lumen‐derived antigen is detected in both LP‐ and IE‐associated eosinophils from antigen‐sensitized mice

We have previously demonstrated that eosinophils residing within the intestine of antigen‐sensitized mice acquire lumen‐derived antigen in vivo by an immunoglobulin‐dependent mechanism.15 With the identification here of two phenotypically and spatially distinct subpopulations of intestinal eosinophils (i.e. LP‐ and IE‐associated), we queried whether one or both subsets might acquire luminal antigens. BALB/c WT mice were systemically sensitized to OVA through intraperitoneal sensitizations with OVA mixed with alum adjuvant before undergoing intestinal loop surgeries wherein fluorescently tagged OVA (or unlabelled OVA as a negative control) were injected into the intestinal lumen of anaesthetized mice as described in ref. 15 and the Materials and methods. After 45 min, intestinal segments were excised, LP and IE leucocytes were recovered, and both cell populations were assessed for uptake of the fluorescently labelled OVA. As shown in Fig. 6(a and b), OVA⁺ eosinophils were isolated with both the LP and IE leucocyte preparations, indicating eosinophils in both of these compartments had acquired fluorescently labelled antigen from the intestinal lumen. Analysis of CD11c expression demonstrated the expected phenotypic distinction between LP and IE eosinophils (i.e. higher CD11c expression on the IE‐isolated eosinophils, Fig. 6c), a distinction maintained in the OVA⁺ eosinophils (Fig. 6d). Taken together, these data indicate that both LP and IE eosinophil subsets are exposed to and competent to acquire lumen‐derived antigens in vivo.

Figure 6.

Both lamina propria (LP) and intraepithelially (IE) associated eosinophils access and acquire lumen‐derived antigen in vivo. Fluorescently tagged ovalbumin (OVA) (or unlabelled OVA) was injected into intestinal loops of anaesthetized mice previously sensitized with OVA in alum adjuvant. After 45 min intestinal loops were removed and rinsed, and LP and IE cells isolated and analysed by flow cytometry for acquisition of the fluorescent signal. Percentages of OVA⁺ eosinophils (gated on live, CD45⁺ SSC^hi SiglecF⁺ cells) isolated with LP (a) or IE (b) cells are shown. In (c,d) surface expression of CD11c is assessed on total (c) and OVA⁺ (d) eosinophil populations.

Discussion

The prevailing paradigm of eosinophils as strictly end‐stage effector cells recruited to sites of allergic provocation has been replaced with a new and growing appreciation of tissue‐resident eosinophils as purveyors of homeostasis and immune regulation. The increasing recognition of unique phenotypes and functions for those eosinophils that localize within different tissue niches emphasizes the importance of directly studying the phenotype and function of steady‐state tissue eosinophils. In health, eosinophils naturally home to the LP of all regions of the gastrointestinal tract except the oesophagus.13 Eosinophil‐deficient mice exhibit alterations in Peyer's patch development, tolerance induction, secretory IgA production and microbiome composition, suggesting that intestinal tissue eosinophils uniquely engage in homeostatic functions (reviewed in ref. 13). Previous studies have also demonstrated that intestinal eosinophils are phenotypically distinguishable from circulating eosinophils,18 a finding supported by our data shown here (Fig. 1). In this study we extend these data by demonstrating that eosinophils residing within the intestinal LP constitutively express antigen‐presenting cell molecules, including MHC II and the co‐stimulatory molecule CD80 (Fig. 2). A previous study reported that intestinal eosinophils do not express MHC II molecules.18 This discrepancy may reflect a difference in interpretation due to the moderate level of MHC II expression on intestinal eosinophils when compared with MHC II expression by other intestinal leucocytes (see Fig. 2a). However, when compared with either isotype control staining (Fig. 2b) or anti‐MHC II antibody reactivity on autologous blood eosinophils (Fig. 2a), intestinal LP eosinophils display a clear shift in MHC II expression, comparable to that achieved in vitro following stimulation with GM‐CSF, and sufficient to promote antigen‐specific T‐cell activation.23

Studies in mice23, 28, 29 and humans30, 31, 32, 33 demonstrate that antigen and cytokine‐primed eosinophils are capable of functioning as professional antigen‐presenting cells in vitro and in vivo. Eosinophils recruited to the oesophagus in patients with eosinophilic oesophagitis express the antigen presentation markers HLA‐DR and CD80.34 Collectively, these data and ours shown here suggest that unlike peripheral blood eosinophils, gastrointestinal tissue eosinophils are primed for direct antigen‐specific cellular interactions, possibly driven by their constitutive exposure to GM‐CSF within intestinal tissues.25 Our previous work that demonstrated intestinal eosinophils from allergic mice acquire luminal antigen15 and new data shown here that extend antigen acquisition to both LP and IE subsets of eosinophils in vivo (Fig. 6) support this hypothesis. Further studies are needed to determine whether intestinal eosinophils that have acquired luminal antigen engage in direct antigen presentation functions in vivo, and to ascertain the biological significance of such antigen‐specific interactions. For example, might antigen‐loaded eosinophils migrate to draining lymph nodes to contribute to naive T‐cell activation, and/or participate in local immune modulation through direct antigen‐dependent interactions with activated T cells recruited into the local tissue microenvironment?

Here we also identify a novel subset of resident CD11c^hi intestinal eosinophils that are recovered with IE leucocytes and that are phenotypically distinguishable from both peripheral blood and LP eosinophils (Fig. 3). These data demonstrate the existence of distinct subsets of small intestinal tissue eosinophils and suggest some degree of spatial compartmentalization. Whether or not those eosinophils isolated with IE leucocyte preparations physically reside within the IE compartment (i.e. within the epithelial monolayer) remains to be determined. Although in general intestinal eosinophils are thought to localize to the LP in health and intercalate into the IE niche only in association with disease, eosinophils and/or their cellular extensions may be underappreciated in standard haematoxylin & eosin preparations of tissue cross‐sections. In support of this interpretation, a study that applied light and electron microscopy to delineate the IE leucocyte composition of the intestinal mucosa from normal subjects reported eosinophils to be among those IE cells observed.35 Fluorescence light microscopy shown here using EoCre^+/− mTmG^fl/fl mice demonstrate a proportion of villous eosinophils in close proximity to the basement membrane, with dendritic processes that contact the basement membrane (Figs 4 and 5); therefore it is also plausible that these IE‐proximate eosinophils are those that are recovered with the IE leucocytes. Regardless of their precise localizations (i.e. above or immediately below the basement membrane), these data demonstrate the existence of phenotypically distinct subsets of tissue‐dwelling eosinophils at baseline, providing foundational insights into the organization and functional potential of resident intestinal eosinophils.

These data also have technical implications that may impact the design and interpretation of studies using common methods of dendritic cell identification and deletion. For example, CD11c‐diphtheria toxin receptor (DTR) transgenic mice are commonly used to specifically probe dendritic cell functions. Although intestinal eosinophils are not globally absent in DTR transgenic mice,36 it will now need to be determined whether or not the CD11c^hi‐expressing IE eosinophil compartment remains fully functional and intact in CD11c⁺ DTR‐depleted mice.

Intestinal tissue eosinophils are poorly defined and their specific functions that promote health or disease remain enigmatic. Increasing prevalences of eosinophilic gastrointestinal diseases paired with an expanding appreciation for novel and distinct functions of those eosinophils residing within tissue niches at baseline and in disease highlight the need for understanding the basic immunobiology of intestinal tissue eosinophils. To this end, this work is the first to demonstrate that resident LP intestinal eosinophils constitutively express molecular machinery associated with antigen presentation, and to identify a novel resident population of CD11c^hi eosinophils associated with the IE microenvironment of the steady‐state intestine.

Author contributions

JJX, EDH, KMS, CLO, YH, DA, SM and EWC performed experiments. EDH, CLO and EWC generated the EoCre^+/− mTmG^fl/fl mice. LAS designed the study and wrote the manuscript.

Disclosures

The authors have no conflicts to disclose.

Supporting information

Figure S1. Gating strategy for intestinal eosinophils.

Figure S2. Efficiency of intraepithelial and lamina propria isolations.

Figure S3. Effects of collagenase digestion on CD11c expression.