IRF1 Is an upstream regulatory transcription factor of eosinophils in type 1 environments

Abstract

Background:

Eosinophils have specific immune phenotypes in type 2 and type 1 environments. The regulatory transcription factors (TFs) that control eosinophil activation in type 2 or type 1 immune phenotypes E2 (eosinophils treated with IL-4, GM-CSF, IL-33, and IL-5) or E1 (eosinophils treated with IFN-γ, TNF-α, and IL-5), respectively, are unknown.

Objective:

We sought to compare mouse and human eosinophil immune phenotypes following exposure to type 2 or type 1 polarizing cytokines and identify TFs that may regulate these responses.

Methods:

Peripheral blood eosinophils were isolated from wild-type mice and from healthy human donors. Cells were cultured with type 2 (IL-4, GM-CSF, and IL-33) or type 1 (IFN-γ and TNF-α) cytokines. Cells underwent characterization of morphology, gene or protein expression, and bulk RNA sequencing. Bone marrow–derived wild-type and interferon regulatory factor (IRF)-deficient mouse eosinophils were generated and analyzed.

Results:

Mouse and human eosinophils both demonstrated type 2 or type 1 cytokine/chemokine production as per E2 or E1 condition. Gene set enrichment revealed that similar pathways were upregulated in mouse and human E2 or E1 eosinophils, respectively. Upstream TF regulatory networks were identified as similar between species as per E2 or E1 condition. In particular, IRF1 expression increased significantly in E1 conditions for mouse and human eosinophils. IRF1-deficient mouse eosinophils had significant increases in type 2 cytokine and chemokine production concurrent with reduced Nos2, Stat1, IL-12b, and PDL1 when in E1 conditions.

Conclusions:

Mouse and human eosinophils have significant similarities in their transcriptomes for their responses to type 2 and type 1 cytokines. IRF1 is increased in mouse and human eosinophils in type 1 environments and regulates immune responses of mouse eosinophils stimulated with type 1 cytokines.

Graphical Abstract

Capsule summary:

The study demonstrates that interferon regulatory factor 1 is an upstream regulatory transcription factor that modulates eosinophil type 1 and type 2 immune responses.

Eosinophils are commonly thought to be rare granulocytes associated with parasitic infections and allergic diseases.¹ Although classically known as terminally differentiated cells that release cytotoxic granules in response to stimuli in disease,² eosinophils are increasingly identified as cells with significant immune functions in health and disease.^3,4 Studies in mice have demonstrated that eosinophils contribute to organ development and physiologic homeostasis. The diverse location of tissue eosinophils in human and mice in addition to the array of nonallergic diseases for which they are found (eg, cancer, infection, and organ transplant)^5-7 suggests that complex immune regulation of eosinophils likely exists but is incompletely defined.

Similar to other leukocytes, eosinophils respond to and promote specific type 2 and type 1 immune pathways. For example, activation states of eosinophils have been related to allergic type 2 immune responses. Both human and mouse eosinophils respond to cytokines found in type 2 environments (eg, IL-4, IL-5, and IL-33).^8-13 This leads to the release of type 2 cytokines (eg, IL-4 and IL-13) and chemokines (eg, CCL17) and increases cell surface expression of CD11b and ST2. Studies in mice have also demonstrated that type 2 eosinophils functionally interact with M2 macrophages, T_H2 cells, and group 2 innate lymphoid cells.^9,14-16 Conversely, eosinophils also have functions in type 1 immune environments, with some evidence in human eosinophils.^10,17,18 Eosinophils exposed to type 1 inflammation (IFN-γ and TNF-α) as found in viral¹⁹ and bacterial infection,^17,20 cancer,^21,22 or allograft responses^23,24 produce type 1 chemokines (eg, CXCL10), increase expression of nitric oxide synthase (NOS2), increase expression of cell surface molecules such as MHCI and programmed death ligand 1 (PDL1), and, as shown in mice, modulate T-cell activities. Altogether, these findings suggest that eosinophils activate into type 2 or type 1 immune subtypes, in part, depending on the activating cytokine milieu.

Regulatory transcription factors (TFs) are critical determinants of cell differentiation, phenotype, and fate. TF regulation of eosinophil hematopoiesis in the bone marrow is well characterized.²⁵ Cell fate and lineage TFs such as GATA-1, PU.1, the CCAAT-enhancer binding protein (C/EBP) family C/EBPα, FOG1, and X-box binding protein 1 generate mature eosinophils from other myeloid populations in the bone marrow. Yet, TFs involved in regulating eosinophil immune phenotype/activation state are less defined. TFs that polarize other immune cells, such as GATA-3 or signal transducer and activator of transcription 6 (STAT6) for type 2 responses and Tbx21 or STAT1 for type 1 responses,²⁶ are not well defined as critical regulators of eosinophils. Overall, the TFs that act as upstream regulators that determine the immune phenotype/activation states of eosinophils are unclear.

To gain insight into the TFs that regulate type 2 or type 1 cytokine–induced eosinophils (E2 and E1, respectively), we exposed mouse and human eosinophils to cytokines that represent type 2–associated (IL-4, GM-CSF, and IL-33) or type 1–associated (IFN-γ and TNF-α) in vivo immune environments. In brief, both mouse and human E2 eosinophils expressed low CD62L and increased levels of type 2 cytokines and chemokines. Similarly, both mouse and human E1 eosinophils expressed high PDL1, intercellular adhesion molecule 1 (ICAM-1), MHCI, and type 1 cytokines and chemokines and increased Nos2 expression. As a whole, mouse and human eosinophils transcriptome data revealed similar gene set enrichment pathways for E2 or E1 eosinophils. Upstream regulatory TFs included changes in the interferon regulatory factors (IRFs).^27,28 IRF1 was increased in mouse and human E1 eosinophils challenged by type 1 cytokines and not type 2 cytokines. This study reveals that IRF1 is a critical determinant of type 1 as compared with type 2 immune responses of mouse eosinophils when exposed to IFN-γ and TNF-α cytokines.

METHODS

Mice

Both male and female wild-type (WT), IL-5 transgenic NJ.1638, Irf4-deficient, and Irf1^−/− (The Jackson Laboratories, Bar Harbor, Maine) mice on a C57BL/6 background were maintained at Mayo Clinic, Arizona. Irf4-deficient mice were obtained by crossing eosinophil-Cre²⁹ to Irf4^fl/fl (The Jackson Laboratories) at Mayo Clinic, Arizona. Protocols and studies involving animals were performed in accordance with the National Institutes of Health and the Mayo Foundation Animal Care and Use Committee institutional guidelines.

Mouse eosinophil isolation and culture

Blood eosinophils (>97% purity and >99% viable) were isolated and purified from NJ.1638 mice as described previously.⁹ Briefly, tail vein blood eosinophils were separated from other leukocytes on Histopaque (Sigma-Aldrich, St Louis, Mo) 1.119 g/dL density gradient, depleted of remaining red blood cells by water lysis, and selected by negative selection using antibodies to CD90.2 and B220/CD45RB (Miltenyi Biotech, Bergish Glad-bach, Germany) as per manufacturer’s instructions. In some experiments, bone marrow–derived eosinophils from mice were generated as described previously³⁰ and used on day 14 for culture. Differentiation state was determined by cytospin and Hema 3 staining (Thermo Fisher Scientific, Cleveland, Ohio) to ensure no mitotic or immature eosinophils were present in the culture. Eosinophils were cultured at 2.0 × 10⁶/mL in RPMI-1640 medium supplemented with FBS, antibiotics, glutamine, sodium pyruvate, and minimal essential amino acids for 18 to 24 hours (unless noted elsewhere) as E0 eosinophils (IL-5 [10 ng/mL]), E2 eosinophils (IL-5 [10 ng/mL)/GM-CSF [10 ng/mL]/IL-4 [10 ng/mL]/IL-33 [40 ng/mL]), or E1 eosinophils (IL-5 [10 ng/mL]/IFN-γ [15 ng/mL]/TNF-α [15 ng/mL]). In some experiments, eosinophils were cultured with individual cytokines at the same aforementioned concentrations. All of these cytokines were from Peprotech (Cranbury, NJ) or R&D Systems (Minneapolis, Minn).

Human eosinophil isolation and culture

Eosinophils were purified from 30 mL of human blood by venous puncture from 3 otherwise healthy (nonatopic, nonallergic, and nonasthmatic) donors. Eosinophils were isolated from whole blood using the MACSxpress Whole Blood Eosinophil Isolation Kit (Miltenyi Biotech) following the manufacturer’s protocol. Viability was more than 99% as measured by trypan blue, and purity of eosinophils was more than 99% by Hema 3–stained cytospins. Cells were cultured at 2.0 × 10⁶/mL in RPMI-1640 medium supplemented with FBS, antibiotics, glutamine, sodium pyruvate, and minimal essential amino acids for 18 hours as E0 eosinophils (IL-5 [1 ng/mL]), E2 eosinophils (IL-5 [1 ng/mL]/GM-CSF [10 ng/mL]/IL-4 [10 ng/mL]/IL-33 [40 ng/mL]), or E1 eosinophils (IL-5 [1 ng/mL]/IFN-γ [15 ng/mL]/TNF-α [15 ng/mL]). All recombinant human cytokines were from R&D Systems. This study was approved by the Institutional Review Board at Mayo Clinic (no. 17-007025).

Cell imaging

Purified eosinophils from cultures were cytocentrifuged onto glass slides using Thermo Shandon Cytospin 4 (Thermo Fisher Scientific) and stained with Protocol Hema 3 (Thermo Fisher Scientific).

Flow cytometry

Surface staining of single-cell suspensions of cultured eosinophils was performed at 4°C for 30 minutes in fluorescence-activated cell sorting buffer (0.5% FBS + 2 mM EDTA in PBS). Anti-CD16/32 (clone 2.4G2) was used to block background staining, and fixable viability dye eFluor 455UV (eBiosciences, San Diego, Calif) was used to exclude dead cells. For intracellular staining of IRF1, the Foxp3/Transcription Factor Staining Buffer set (Thermo Fisher Scientific) was used as per manufacturer’s protocol. A full list of antibodies is provided in Table E1 (in the Online Repository available at www.jacionline.org) along with technical details of experimental procedures in this article’s Methods section in the Online Repository at www.jacionline.org. In some experiments, viability was determined with the BD Pharmingen FITC Annexin V Apoptosis Detection Kit I (BD Biosciences, San Jose, Calif). Flow cytometry was performed on a BD LSR II Fortessa Cytometer or BD Symphony A3 Cytometer (BD Biosciences) using the BD FACSDiva software (version 8.0; BD Biosciences) and analyzed using FlowJo 10.5 (TreeStar, Ashland, Ore).

Protein measurements

Cytokines and chemokines were measured in cell-free supernatant using the Multiplexing LASER Bead Assay (Eve Technologies, Inc, Calgary, Alberta, Canada).

RNA isolation

Eosinophils (>250,000 cells) were pelleted and then processed in Trizol (Thermo Fisher Scientific), quick-frozen on dry ice, and stored at −80°C until further processing as per the manufacturer’s protocol. During processing, 10 μg glycogen was added to aqueous phase, and SUPERase-In RNase inhibitor (1 U/μL) was added to the nuclease-free water at the end of purification for long-term storage at −80°C. RNA integrity and purity were analyzed by the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, Calif) and NanoDrop spectrophotometer (Thermo Fisher Scientific).

Real-time PCR

For both mouse and human gene expression analysis, cDNA was synthesized from 0.4 to 1 μg total RNA using SuperScript IV VILO with ezDnase kit (Invitrogen, Carlsbad, Calif) following the manufacturer’s protocol. Real-time PCR (RT-PCR) for mouse genes was performed using TaqMan probes and Taqman Universal Master Mix II, with UNG (Applied Biosystems, Waltham, Mass) on the 7900HT Fast Real-Time PCR System (Applied Biosystems) or the CFX384 Touch Real-Time PCR Detection System (Bio-Rad). For human genes, RT-PCR was performed using primers and Advanced Universal SYBR Green Supermix (Bio-Rad, Fort Worth, Tex) on the CFX384 Touch Real-Time PCR Detection System (Bio-Rad). Lists of primers and probes are provided in Tables E2 and E3 (in the Online Repository available at www.jacionline.org). Data were analyzed using the comparative Ct (ΔΔCt) method with β-actin as the endogenous control and normalized to E0 control eosinophils where noted.

RNA sequencing and processing

Mouse eosinophil total RNA (>100 ng/μL) was amplified using TruSeq RNA Library Prep Kit v2 (Illumina), and next-generation sequencing was performed using HiSeq 2000 PE (Illumina). Samples were sequenced at the Mayo Sequencing Core Facility. The expression differential analysis was performed by edgeR. Differentially expressed genes (DEGs) had either a false-discovery rate (FDR) less than 0.05 or a P value less than .05 and an absolute log₂ fold change greater than 1. Data are deposited in the Gene Expression Omnibus database (accession no. GSE243369).

Human eosinophil total RNA (>60 ng/μL) was processed by Genewiz LLC (South Plainfield, NJ) for next-generation sequencing. RNA was amplified using the NEBNext Ultra RNA Library Prep Kit (Illumina, San Diego, Calif) and next-generation sequencing was performed using Illumina HiSeq 2500. The expression analysis was performed using DESeq2, with a cutoff threshold of an FDR less than 0.05 and an absolute log₂ fold change greater than 1 to identify DEGs. Data are deposited in the Gene Expression Omnibus database (accession no. GSE243490).

RNA analysis

Upregulated DEGs unique to E1 or E2 eosinophils were processed with Enrichr³¹ for enrichment in functional pathways Kyoto Encyclopedia of Genes and Genomes (KEGG), Reactome, and DisGeNET. For TF enrichment analysis, we analyzed the up-regulated DEGs unique to E1 or E2 for analysis with the CHIP-X Enrichment Analysis 3 (ChEA3),³² which uses 6 libraries containing RNA-sequencing, TF CHIP-sequencing, and TF-gene cooccurrence from public-submitted gene lists. A threshold of the top 11 upstream regulatory TFs was chosen for each condition from the ChEA3 analysis. Eosinophil DEGs from rejecting lung transplant mice³³ were derived from flow-cytometric sorting of eosinophils from lungs (naive or rejecting) followed by bulk RNA sequencing (data set GSE223352).

Statistical analysis

Experiments were completed in duplicate at minimum with technical replicates. Data were analyzed using GraphPad Prism 10 for Windows (GraphPad Software, San Diego, Calif). Statistical analysis was performed using unpaired Student t tests or 1-way ANOVA with post hoc for comparisons depending on Gaussian distribution and SDs, using multiple comparisons test for comparison with control (Dunnet) or between all groups (Tukey). Some experiments were measured using the 2-way ANOVA with the Tukey post hoc multiple comparisons test. Error bars represent the mean ± SEM or mean ± SD. Differences between means were considered significant (****P < .0001; ***P <.001; **P <.01; *P < .05).

See this article’s Methods section in the Online Repository for remaining details.

RESULTS

Type 2 and type 1 cytokines are sufficient to induce type 2 and type 1 immune phenotypes of mouse eosinophils, respectively

We developed an in vitro cytokine model to induce eosinophil immune subtypes on the basis of the type 2 or type 1 cytokine microenvironments that eosinophils encounter in inflammatory environments. Activation of eosinophils with type 2–associated cytokines IL-4, GM-CSF, and IL-33 was previously shown in models of allergic asthma.^9,14,15 Eosinophil activation by type 1 cytokines IFN-γ and TNF-α was previously shown in cancer, infection, and lung transplant models.^19-24

Purified mouse blood eosinophils were cultured in vitro for 18 hours with IL-5 in addition to either type 2 (IL-4, GM-CSF, and IL-33) or type 1 (IFN-γ and TNF-α) cytokines to generate E2 and E1 eosinophils, respectively. As a control, E0 eosinophils were cultured with IL-5 only (Fig 1, A). We also chose these cocktail combinations because culture with single cytokines generally did not induce as significant expression of cell surface or cytokine/chemokines as the combinatorial cytokines of E1 or E2 conditions (see Fig E1 in this article’s Online Repository at www.jacionline.org). Mouse eosinophils have been shown to display morphological changes in vivo in type 2 environments.^34,35 In agreement, E2 eosinophils developed nuclear hypersegmentation and vacuole-like structures (Fig 1, A; see also Fig E2, A, in this article’s Online Repository at www.jacionline.org) and demonstrated secondary degranulation (Fig E2, B), whereas E1 and E0 eosinophils contained circular or figure 8 nuclei. Viability of all eosinophils was similar at 24 hours, yet at 48 hours E1 eosinophils were less than 65% viable (Fig E2, C and D). Metabolic activity is also an indication of cell activation state.³⁶ After 24 hours of culture, E1 eosinophils had lower ATP production, maximal respiration, and spare respiratory capacity, indicating lower bioenergetics than E0 or E2 eosinophils (see Fig E3 in this article’s Online Repository at www.jacionline.org). These results were similar to those reported for eosinophils in response to type 1 conditions of virus infection.¹⁹

FIG 1.

Type 2 and type 1 cytokines are sufficient to induce type 2 and type 1 immune phenotypes of mouse eosinophils, respectively. A, Scheme representing experiment. Isolated pure populations of peripheral blood mouse eosinophils were cultured in vitro with cytokines for 18 to 24 hours and then analyzed for various phenotype changes. All cultures had IL-5. E2 eosinophils (blue) were also cultured with IL-4, GM-CSF, and IL-33 and E1 eosinophils (orange) with IFN-γ and TNF-α. E0 (gray) indicates IL-5 only. Right panels are representative images of E0, E1, and E2 cells stained with Hema 3. B, Flow cytometry for cell surface molecules. Representative normalized to mode histograms are shown (top) as well as median fluorescent intensity (MFI) (bottom). Classic eosinophil markers are shown for CCR3, CD11b, Siglec-F, F4/80, and Gr1. C, Cells with higher MFI in E2 for ST2, CD80, CD11c, and CD69. D, Cells with higher MFI in E1 for MHCI, ICAM-1, and PDL1. E, Expression of other phenotype markers CD62L and CD101. F, Quantitative RT-PCR for IL-13 and Nos2 of eosinophils cultured for different time points in E2 or E1 conditions. Relative to β-actin and normalized to E0. G, Cell-free media from the cultures were measured for cytokines/chemokines by multiplex assay. For Fig 1, ***B-E, data are from 3 to 5 independent experiments using 1-way ANOVA with the Tukey post hoc test. Data are shown as mean ± SEM. For Fig 1, ***F and G, representative data are from 2 to 3 experiments with technical replicates using 1-way ANOVA with the Dunnet test to control E0. Data are shown as mean ± SD. ****P < .0001; ***P < .001; **P < .01; *P < .05.

In the literature, peripheral blood mouse eosinophils are identified as CCR3⁺, CD11b ⁺, Siglec-F⁺, F4/80^low-med, and Gr1^low-med.¹¹ The expression level can vary depending on environmental components (ie, cytokines) and tissue location.⁵ E0 eosinophils displayed the highest expression of CCR3 with decreased expression in E1 and E2 eosinophils (Fig 1, B). E2 eosinophils expressed more CD11b, Siglec-F, and F4/80 surface markers compared with E1 eosinophils, which expressed more Gr1. Next, we screened a panel of cell surface markers that have been previously reported on eosinophils for in vivo type 2 or type 1 disease microenvironments or regulated by tissue localization. E2 eosinophils expressed higher levels of ST2, CD80, and CD11c and trended higher for CD69 (Fig 1, C), similar to in vivo reports.^15,34,37 E1 eosinophils expressed increased MHCI, ICAM-1, and PDL1 (Fig 1, D), similar to in vivo reports.^19,22,24 Conversely, E2 eosinophils had reduced CD62L expression, and CD101 was unchanged (Fig 1, E). Overall, the trend is that cell surface expression of E2 and E1 eosinophils is similar to that found in type 2 and type 1 environments, respectively, in translational models of disease.

Next, we characterized the kinetic expression of genes that are established in the literature as increased in type 2 eosinophils or type 1 eosinophils; IL-13⁹ and Nos2,²³ respectively. Expression of IL-13 and Nos2 by E2 and E1 eosinophils, respectively, shows that these cells respond within 4 hours of exposure to cytokines and continue to increase for 24 hours postexposure (Fig 1, F). Protein analysis of the cell-free media revealed specific release of mediators, because E2 eosinophils released IL-13, IL-9, IL-6, CCL22, and CCL17, and CCL2 and E1 eosinophils released CCL5, CXCL9, and CXCL10 (Fig 1, G). Overall, in vitro cytokine stimulation was sufficient to induce expression of many chemokines and cytokines reported for eosinophils in type 1 and type 2 in vivo environments.

E2 and E1 human eosinophils can polarize into type 2 and type 1 immune subtypes

We wished to identify whether human eosinophils stimulated under E2 or E1 cytokine conditions displayed specific type 2 or type 1 immune characteristics, respectively. Blood eosinophils from 3 healthy donors were cultured under conditions similar to the methods for mouse eosinophils, yet with less IL-5 (1 ng/mL instead of 10 ng/mL) (Fig 2, A). All subtypes were more than 85% viable for the 18-hour culture (see Fig E4, A, in this article’s Online Repository at www.jacionline.org). Eosinophil nuclear morphology was a bi-lobed shape for all subtypes (Fig 2, B). Similar to mouse eosinophils, both E1 and E2 eosinophils had reduced CCR3 expression as compared with E0 eosinophils. Siglec-8, a paralog to Siglec-F, CD69, and CD11b were increased in both E1 and E2 eosinophils (Fig 2, C; see also Fig E4, B). Similar to mouse E1 eosinophils, human E1 eosinophils increased PDL1, ICAM-1, and MHCI expression in addition to species-specific CD66b, relative to E0 and E2 eosinophils. Similar to mouse E2 eosinophils, CD62L trended lower and CD101 was unchanged for human E2 eosinophils. Thus, some cell surface phenotypes were maintained between species, whereas others varied. Measuring protein production demonstrated that similar to mouse eosinophils, human E2 eosinophils released type 2 cytokines and chemokines (eg, IL-13, IL-9, IL-6, CCL22, CCL17, and CCL2) (Fig 2, D). Human E1 eosinophils released type 1 cytokines and chemokines (CCL5 and CXCL10) and increased mRNA for CXCL9 and Nos2 (Fig 2, E). Overall, human eosinophils can be polarized into specific immune phenotypes that are comparable, albeit not exact, with mouse eosinophil immune phenotypes with this in vitro stimulation protocol.

FIG 2.

Human eosinophils polarize into type 2 and type 1 immune subtypes on in vitro cytokine culture. A, Scheme representing experiment. Isolated pure populations of peripheral blood human eosinophils from healthy donors were cultured in vitro with cytokines for 18 hours and then analyzed for various phenotype changes. All cultures had IL-5. E2 eosinophils (blue) were also cultured with IL-4, GM-CSF, and IL-33 and E1 eosinophils (orange) with IFN-γ and TNF-α. E0 (gray) indicates IL-5 only. B, Representative images of E0, E1, and E2 cells after Hema 3 staining. C, Representative normalized to mode histograms are shown (top) as well as mean fluorescence intensity (MFI) (bottom) for cell surface markers. D, Cell-free media from the cultures were measured for cytokines/chemokines by multiplex assay. E, Quantitative RT-PCR for CXCL9 and NOS2 of E1 or E2 eosinophils relative to β-actin and normalized to E0. Data are shown as mean ± SEM. Data are from 3 independent donors using 1-way ANOVA with the Tukey multiple comparisons test (Fig 2, C and D) or the Student t test (Fig 2, E). ****P < .0001; ***P < .001; **P < .01; *P < .05.

E2 and E1 human eosinophils have transcriptomes similar to mouse E2 and E1 eosinophils

To better understand transcriptome changes between E0, E2, and E1 eosinophils, we performed bulk RNA next-generation sequencing on mouse and human eosinophils. Volcano plots and genes enriched in pathways for DEGs (FDR < 0.05 and absolute log₂ fold change > 1) are shown for E2 to E0 and for E1 to E0 for mouse (see Fig E5 in this article’s Online Repository at www.jacionline.org) and human (see Fig E6 in this article’s Online Repository at www.jacionline.org) (see also Supplementary Data 1 in this article’s Online Repository at www.jacionline.org). We compared the upregulated DEGs in E2 and E1 mouse and human eosinophil transcriptomes to better define potential similarities between species. In brief, both mouse and human E2 eosinophils expressed genes that have been previously described for eosinophils in type 2 environments, such as IL-6, IL-13, Ccl2, Ccl22, and Ccl17, in addition to genes associated with eosinophil migration and activation (Lgals1, Anxa2, CD52, S100a4, and S100a6), G protein–coupled receptors (Ffar1 and S1pr), remodeling and extracellular matrix (Plau and Timp1), protease inhibitors (Cst7), and survival (Myc and Bcl2). Genes associated with glucose and glycolysis (eg, HK3 [human] and Pgk1 and Pkfp [mice]) were increased as well as genes in fatty acid metabolism (eg, Faah).

E1 eosinophils, for human and mice, included upregulated genes that had been previously described in type 1 environments, such as IL-12, Ccl5, Cxcl9, Cxcl10, and CD274. Also increased were proteosome genes (eg, Psme and Psmb), antiviral genes (eg, Adar, Ifnar2, Ifit1, Ifih1, Dhx58, Mx1, and Oas2), E3 ubiquitin protein ligases and autophagy (Trim21 and Dtx3l), cell adhesion (eg, Vcam1), pathogen recognition responses (Casp1 and Nod1), apoptosis-related genes (Bak1 and Birc3), caspase genes (Casp1 and Casp7), complement genes (C4b and Cfb), and phosphatase genes (Dusp2). Genes associated with nicotinate metabolism (eg, Nampt, Parp9, Parp10, and CD38) were increased as well as genes important in itaconate metabolism pathways (ADOC1/Irg1).

Transcriptome pathways were enriched using unique DEGs for E2 and E1 (see Supplemental Data 2 in this article’s Online Repository at www.jacionline.org). Representative enriched pathways show some species and eosinophil subtype similarities (Fig 3). For additional pathways, see Supplemental Data 3 and 4 (in the Online Repository available at www.jacionline.org). In brief, human and mouse E2 eosinophils were increased in kinase signaling pathways, focal adhesion pathways, and IL-4/IL-13 and IL-10 pathways. Human and mouse E1 eosinophils were enriched in infection pathways, Toll-like receptor signaling, interferon signaling, and antigen presentation via MHCI. E2 eosinophils were associated with asthma, dermatitis, and cancers, whereas E1 eosinophils were associated with viral diseases, infections, and autoimmune diseases.

FIG 3.

Human and mouse E2 and E1 eosinophils have similar transcriptomes. Upregulated DEGs unique to E2 (blue; top) and E1 (orange; bottom) were enriched for significant pathways using Enricher. Representative top enriched pathways are shown and demonstrate similarities between human (left) and mouse (right) transcriptomes for KEGG, Reactome, and DisGeNET. Scale is log adjusted P value. KEGG, Kyoto Encyclopedia of Genes and Genomes; MAPK, mitogen-activated protein kinase.

Overall, mouse and human eosinophils expressed similar transcriptome profiles, resulting in overlap of many functional pathways for each unique E2 or E1 cytokine stimulation.

Upstream regulatory TFs and expression of IRFs in mouse and human eosinophils

Because similar pathways and genes were upregulated in human and mouse for each condition, this allowed for an opportunity to identify upstream regulatory TFs for E2 or E1 eosinophil phenotypes in both species. First, we measured the STAT gene family because it is associated with immune phenotypes in many cells.³⁸ Gene expression of the STAT gene family was higher in E1 eosinophils for both human (see Fig E7, A, in this article’s Online Repository at www.jacionline.org) and mouse (Fig E7, C) for Stat1, Stat2, Stat3, and Stat5a. RT-PCR confirmed Stat1 was increased in E1 eosinophils, whereas Stat6, which is associated with type 2 conditions, was not increased in E2 conditions for human (Fig E7, B) or mouse (Fig E7, D) eosinophils. TFs Tbx21 and Gata-3 also regulate type 1 and type 2 immune responses, respectively, particularly in lymphocytes. Both human (Fig E7, E) and mouse (Fig E7, F) E2 eosinophils had elevated Gata-3 in E2 conditions as compared with E0 conditions. Unexpectedly, Tbx21 was also increased in E2 conditions, but not significantly increased in E1 conditions.

Therefore, we completed gene enrichment analysis for upstream regulatory TFs. We used ChEA3 to identify the top 11 TFs for each group. ChEA3 uses 6 libraries to identify upstream TFs of transcriptomes.³² Both human and mouse E2 eosinophils had 7 matching TF genes: Etv3lL, Batf3, Nfkb2, Relb, NFkb1, Batf, and Csrnp1 (Fig 4, A and B). Both human and mouse E1 eosinophils had 10 matching TF genes: Irf1, Irf7, Relb, Batf, Batf3, Nfkb2, Tfec, Relb, Plscr1, and Sp100 (Fig 4, C and D). We then took advantage of our previously published transcriptome data of eosinophils isolated from the lungs of mice that had undergone lung transplant rejection,³³ which is rich in TNF-α and IFN-γ. These sorted lung eosinophils showed some similarities to the in vitro mouse and human E1 eosinophils matching genes Irf1, Irf7, Batf, Batf3, Nfkb2, and Tfec (Fig 4, E). These data demonstrate that there are several consistent upstream regulatory TFs per E1 or E2 phenotype for mouse and human eosinophils.

FIG 4.

Upstream regulatory TFs in E1 and E2 eosinophils. Upregulated DEGs were analyzed by ChEA3 for the top 11 upstream regulatory TFs in mouse and human E2 and E1 eosinophils. Network maps are shown for the TFs. A and B, Human (Fig 4, A) and mouse (Fig 4, B) E2 eosinophils matched for 7 TFs (blue). C and D, Human (Fig 4, C) and mouse (Fig 4, D) E1 eosinophils matched for 10 TFs (orange). E, When compared with isolated eosinophils from a rejecting lung transplant, which is rich in type 1 cytokines, 6 TFs matched. Genes are shown in tables to the right. F-H, To better characterize the pattern of the IRF gene family, normalized counts from RNA sequencing for human (Fig 4, F) and mouse (Fig 4, G) E0, E1, and E2 eosinophils were compared as well as for isolated eosinophils (Fig 4, H) from lung transplant as compared with naive resting lung eosinophils (ie, no transplant). Data are shown as mean ± SD using 1-way ANOVA with the Dunnet test to control E0 (Fig 4, F and G) or mean ± SEM and 1-way ANOVA with the Dunnet test to naive (Fig 4, H). ****P < .0001; ***P < .001; **P < .01; *P < .05.

We focused on the IRF gene family because it has not been well described in eosinophils. This 9-gene family (IRF1-9) was originally identified as regulators to type 1 and type 2 interferons that mount antiviral and anti-infection responses.²⁷ More recently, the IRF family has been found to have roles in cell activation, differentiation, metabolic programming, and immune responses as demonstrated in other leukocytes, particularly myeloid cells.^27,28,39 Normalized gene counts for human (Fig 4, F), mouse (Fig 4, G), and eosinophils from rejecting lung transplant (Fig 4, H) show that some IRF genes were increased in E1 conditions when compared with E0 conditions. Side-by-side comparison with the ingenuity pathway analysis of the IRF genes for all data sets showed relative increases in many IRF genes in E1 conditions, whereas IRF4 showed a relative increase in E2 eosinophils (see Fig E8, A, in this article’s Online Repository at www.jacionline.org). In a separate experiment, RT-PCR for Irf1-9 with mouse eosinophils confirmed increases in Irf1, Irf2, Irf5, Irf7, Irf8, and Irf9 for E1 eosinophils relative to E0, whereas Irf4 was the only significantly upregulated gene for E2 eosinophils (Fig E8, B).

An increase in Irf4 was unique to mouse E2 eosinophils, and RT-PCR of human E2 eosinophils did not reveal significant increases in IRF4 (see Fig E9, A-C, in this article’s Online Repository at www.jacionline.org). Because IRF4 was identified as an upstream TF in mouse E2 eosinophils by ChEA3, we expanded the analysis for human E2 to include the top 20 upstream regulatory TFs, which then identified IRF4 in human E2 eosinophils (Fig E9, D). The literature, also, suggests a role for IRF4 in human eosinophils in allergy.^40,41 To look at the role of IRF4 closer, we generated IRF4 knockout (Irf4^−/−) bone marrow–derived eosinophils, which expanded similarly to WT eosinophils (see Fig E9, E, in this article’s Online Repository at www.jacionline.org). It has been reported previously that IRF4 deficiency does not lead to eosinophil depletion in mice,⁴² whereas IRF8 is necessary for eosinophil development. IRF4 can regulate IRF1,⁴³ yet Irf1 expression was unchanged (Fig E9, F). Moreover, no major changes in cytokine or chemokine secretion were identified for WT or Irf4^−/− eosinophils, except a reduction in CCL2, CCL5, and M-CSF in IRF4-deficient eosinophils in E2 conditions (Fig E9, G). Thus, these data suggest that IRF4 likely has a limited role in cytokine/chemokine production of E1 and E2 mouse eosinophils in this in vitro system.

IRF1 is a significantly upregulated regulatory TF in E1 eosinophils and regulated by type 1 cytokines in mouse eosinophils

Our data demonstrated that Irf1 was highly expressed in E1 conditions and was identified as a top upstream regulatory TF in vitro and in vivo (see Figs 4 and 5, A and B). As reported by others, IRF1 was increased in mouse eosinophils in vivo in infection^17,20 and in vitro on culture with IFN-γ + Escherichia coli.⁴⁴ As noted in previous reports for IRF1, IFN-γ was sufficient to induce expression of Irf1 and was synergistically increased by TNF-α 1 IFN-γ (see Fig E10, A, in this article’s Online Repository at www.jacionline.org). Protein expression was confirmed by intracellular flow cytometry for IRF1 in mouse E1 eosinophils (Fig 5, C). To test the functional capacity of IRF1 to bind promoters, we used CHIP-PCR to identify direct binding targets of IRF1 as reported previously for macrophage.⁴⁵ IRF1 antibody pulled down the promoters for Nos2 and Batf2, indicating binding to these promoters in E1 eosinophils (Fig E10, B). To determine the plasticity of Irf1 levels to cytokine stimuli, we cultured eosinophils in E0, E1, or E2 conditions or switched E1 and E2 conditions at 24 hours. When mouse cultures were switched in condition from E1 to E2 cytokines or from E2 to E1 cytokines, it was found that Irf1 and the downstream targets Batf2 and Nos2 changed with condition (ie, only increased when type 1 cytokines were present) (Fig 5, D and E). Moreover, type 2 cytokine IL-13 and cell surface markers CD11b, CD62L, PDL1, and MHCI were found to modulate depending on cytokine environment (Fig 5, E and F; see also Fig E11 in this article’s Online Repository at www.jacionline.org). Overall, features of E1 and E2 immune phenotypes of mouse eosinophils were plastic to cytokine exposure, including expression of Irf1.

FIG 5.

IRF1 is highly upregulated and modulated by type 1 cytokines in eosinophils. A and B, Human eosinophils cultured as in Fig 2 (Fig 5, A) and mouse eosinophils as in Fig 1 (Fig 5, B) underwent quantitative RT-PCR for IRF1, showing significant upregulation in E1 eosinophils. C, Representative histogram for intracellular staining for IRF1 in mouse eosinophils showing elevated IRF1 protein in E1 eosinophils. D, Mouse eosinophils were cultured to test changing cytokine conditions on gene expression and surface protein expression. E1 eosinophils (orange) were cultured for 48 hours in E1 conditions (media refreshed at 24 hours); E2 eosinophils (blue) were cultured for 48 hours in E2 conditions (media refreshed at 24 hours); E1-2 eosinophils (green) were cultured for 24 hours in E1 conditions followed by 24 hours in E2 conditions; and E2-1 eosinophils (purple) were cultured for 24 hours in E2 conditions followed by 24 hours in E1 conditions. Quantitative RT-PCR for TFs Irf1 and Batf2. E, Nos2 and IL-13 were normalized to β-actin and E0 eosinophils. F, Representative normalized to mode histograms are shown (top) as well as median fluorescent intensity (MFI) (bottom) for cell surface markers by flow cytometry. For Fig 5, A, B, D-F, representative data were from 2 to 3 experiments with technical replicates using 1-way ANOVA with the Tukey test. Data are shown as mean ± SD. ****P < .0001; ***P < .001; **P < .01; *P < .05.

IRF1 is required to suppress type 2 immune responses when mouse eosinophils are exposed in vitro to type 1 cytokines

A role for IRF1 in inducing type 1 immune polarization and suppressing type 2 polarization has been identified in other cell types. For example, IRF1 is a critical regulator of the polarization of M1 macrophages^39,45 and T_H1 cells.^46,47 To better understand whether IRF1 has a role in eosinophil immune phenotype, we compared WT and IRF1 knockout bone marrow–derived eosinophils in our in vitro system. IRF1 deficiency did not alter the growth kinetics or differentiation of bone marrow–derived eosinophils (Fig 6, A and B). Similar number and purity of eosinophils were obtained between WT and IRF1 knockout (Irf1^−/−) eosinophils. Viability was similar between WT and Irf1^−/− eosinophils when cultured in E0, E1, and E2 conditions (see Fig E12, A, in this article’s Online Repository at www.jacionline.org). Because IRF family proteins may interact with each other to cooperate in transcription regulation, we measured the effect of IRF1 deficiency on expression of other IRF gene family members. In E0 conditions, Irfs1-9 expression levels were generally no different between WT and Irf1^−/− eosinophils, with some attenuation of Irf7 (Fig 6, C), indicating that at a basal state the absence of IRF1 was not having a major effect on other Irf levels. When cells were placed in E2 conditions, overall expression between WT and Irf1^−/− eosinophils was similar, yet reductions in expression were noted for Irf4, Irf5, and Irf7 (Fig 6, D). E1 conditions led to significant changes in expression of several IRF genes between WT and Irf1^−/− eosinophils (Fig 6, E). Irf1^−/− eosinophils expressed less Irf2 and Irf7 and increased Irf4 and Irf8. Strikingly, IRF1 target gene Batf2 was increased in IRF1-deficient eosinophils, whereas Nos2 was decreased (Fig 6, F). Stat1, which closely interacts with IRF1, was also reduced in Irf1^−/− eosinophils in E1 conditions (Fig 6, G). Analysis of other gene targets of IRF1 gave mixed results, with IL-12 being reduced and Cxcl10 increased in E1 Irf1^−/− eosinophils (Fig 6, H). Moreover, loss of IRF1 led to increased IL-4 gene expression (Fig 6, I) and protein released (Fig E12, B) in E1 cytokine conditions as compared with WT eosinophils, presumably because of loss of IRF1 acting as a TF repressor of IL-4.⁴⁸

FIG 6.

In the absence of IRF1, mouse eosinophils have type 2 immune characteristics when treated with IFN-γ and TNF-α. A, Bone marrow from WT or IRF1 knockout mice was cultured with Flt3 or stem cell factor (SCF) for 5 days and then with IL-5 to induce expansion and differentiation of eosinophils. The total number and percentage of eosinophils are shown. B, Hema 3–stained cytospins of eosinophils from day 14 show similar morphology. C, Day 14 cells were cultured for 18 hours as in Fig 1 for E0, E1, and E2 culture conditions. Quantitative RT-PCR was completed for the Irf1-9 gene families for E0 cells normalized to β-actin only. D and E, E2-treated cells (normalized to β-actin and WT E0) (Fig 6, D) and E1-treated cells (normalized to β-actin and WT E0) (Fig 6, E). F-I, Quantitative RT-PCR relative to β-actin and WT E0 was completed for Batf2 and Nos2 (Fig 6, F), Stat1 (Fig 6, G), IL-12b and Cxcl10 (Fig 6, H), and IL-4 (Fig 6, I). Cell-free media from the cultures were measured for cytokines by multiplex assay. For Fig 6, ***C-I, representative data were from 2 to 3 experiments with technical replicates using 1-way ANOVA with the Dunnet test to control. Data are shown as mean ± SD. For Fig 6, J, representative data were from 2 experiments with technical replicates using multiple t test between WT and IRF1-deficient eosinophils. Data are shown as mean ± SD. ****P < .0001;***P < .001; **P < .01; *P < .05.

To better define immune changes in Irf1^−/− eosinophils as compared with WT eosinophils, cells were measured for protein release and cell surface expression. Few measured changes in phenotype occurred in E0 conditions for Irf1^−/− eosinophils as compared with WT eosinophils. Some type 2 responses were increased in E2 Irf1^−/− eosinophils, such as increased IL-4 and IL-13, whereas CCL2 was reduced (Fig 6, J; see also Fig E12, B). Conversely, E1 cultured Irf1^−/− eosinophils gave mixed immune responses, with several type 1 and type 2 cytokines/chemokines being increased. In particular, E1 cultured Irf1^−/− eosinophils had significant increases in production of type 2 cytokines and chemokines, despite the stimulation of TNF-α and IFN-γ. E1 conditions induced significant release of IL-4, IL-13, IL-6, CCL17, CCL22 (Fig 6, J), and other cytokine/chemokines (eg, CCL3, CCL4, IL-16, and IL-10) (Fig E12, C) by Irf1^−/− eosinophils as compared with WT eosinophils.

Cell surface expression of E1 conditioned Irf1^−/− eosinophils was increased for CD11b and CD11c, which are normally expressed on WT E2 eosinophils (Fig 7, A; see also Fig E12, D). Moreover, E1 Irf1^−/− eosinophils had reduced expression of CD62L, similar to WT E2 eosinophils. Yet, Irf1^−/− eosinophils failed to increase PDL1 in E1 conditions, whereas ICAM-1 was higher and MCHI was unchanged. To better define the role of IFN-γ or TNF-α in these responses, we treated WTor Irf1^−/− eosinophils with individual cytokines IL-5, IFN-γ, TNF-α, and E1 conditions. Only the combination of IFN-γ and TNF-α as in the E1 condition led to significant increases in type 2 cytokines and chemokines (eg, IL-4, IL-6, and CCL22) and type 1 chemokines CXCL10 and CCL5 (see Fig E13, A, in this article’s Online Repository at www.jacionline.org). Moreover, E1 conditions, rather than individual cytokines, were required to increase CD11c and CD11b and reduce CD62L in Irf1^−/− eosinophils (Fig E13, B). PDL1 expression, conversely, was dependent on IFN-γ signaling alone (Fig 7, B) and failed to increase in Irf1^−/− eosinophils. Overall, these studies demonstrate that deletion of IRF1 in eosinophils stimulated with the combination of IFN-γ and TNF-α lead to dysregulated production of type 2 cytokines, chemokines, and cell surface expression. Moreover, some TF gene targets of IRF1 that are generally upregulated in type 1 conditions (eg, Nos2, IL-12, and PDL1) were not increased in response to type 1 cytokines in the absence of IRF1, whereas others were enhanced (eg, CXCL10 and CCL5).

FIG 7.

IRF1-deficient mouse eosinophils have altered cell surface expression with PDL1 regulated by IFN-γ. A, Bone marrow–derived WT or IRF1-deficient eosinophils were cultured as in Fig 1. Representative normalized to mode histograms are shown (left panels) as well as median fluorescent intensity (MFI) (right panels) for cell surface markers by flow cytometry. B, Bone marrow–derived WT or IRF1-deficient eosinophils were cultured with individual cytokines of IL-5, IFN-γ, TNF-α, or combinations of these cytokines for 24 hours and then measured for PDL1 expression. Representative normalized to mode histograms are shown (left panels) as well as MFI (right panels) for cell surface markers by flow cytometry. For Fig 7, A and B, representative data were from 2 to 3 experiments with technical replicates using multiple t test with WT as control to IRF1-deficient conditions. Data are shown as mean ± SD. ****P < .0001; ***P < .001; **P < .01; *P < .05.

DISCUSSION

Eosinophils have generally been considered a homogeneous population that is recruited into tissues with ongoing type 2 inflammation.² Recent paradigm shifting studies, however, are finding immune modulating roles for eosinophils in complex tissue microenvironments.^3,4 Indeed, although tissue extracellular matrices, microbiomes, and epigenetic factors likely imprint a phenotype into eosinophils,⁵ we propose that similar to other immune cells, cytokines can also implement specific immune responses by eosinophils. This concept of a type 2 or type 1 immune phenotype eosinophil has been proposed recently by others as well,⁴⁴ although the regulatory upstream TFs directing these responses are not well described. Here, we completed reductionist studies to complete a comparative analysis of mouse and human eosinophils exposed in vitro to type 2 or type 1 cytokines that are found in disease environments such as allergic asthma^9,13-16 or infection, cancer, and transplant rejection,^6,19-21,23 respectively. Analysis of whole transcriptome data identified that the IRF TF gene family was uniquely modulated by type 2 or type 1 cytokine treatment of eosinophils. In particular, IRF1 was identified as an increased upstream regulatory TF in eosinophils in type 1 environments in human and mouse eosinophils in vitro and in a translational model of disease. Deletion of IRF1 in mouse eosinophils resulted in an increase in type 2 immune functions despite stimulation with IFN-γ and TNF-α, suggesting that IRF1 is critical to regulating the balance between type 2 and type 1 immune responses of eosinophils.

Previous studies have used individual cytokines to stimulate eosinophils into activation states.^17,44 We found that individual cytokine stimulation was less effective at stimulating significant immune response (ie, cytokine/chemokine production and cell surface expression) in mouse blood eosinophils as compared with the cocktail of E2 or E1 cytokines that would be present in the in vivo environment. For example, mouse and human E2 eosinophils expressed IL-6, IL-4, and IL-13 and chemokines CCL17 and CCL22.^9,10,12,49 E2 eosinophils also expressed CCL3 (MIP1α), CCL4 (MIP1β), and CXCL2 (mouse), which likely were induced by IL-33 as per our studies and reports by others.¹⁷ Mouse and human type 1–challenged eosinophils expressed CXCL10, CXCL9, and CCL5 as reported previously.^{10,18,21,23,49}

Expression of cell surface molecules was more complex, partly because of species differences. A comprehensive analysis is not completed here, yet some notable changes occurred that were unique to each species or comparable between species. For example, although CD11b was increased in mouse E2 eosinophils, their levels in human E2 eosinophils did not reach significance, despite their reported increase in people with allergic asthma.⁵⁰ Most predominant was that both mouse and human E1 eosinophils demonstrated increased PDL1, MHCI, and ICAM-1, similar to reports for eosinophils in in vivo type 1 environments.^{19,20,44,51,52} Previous studies have identified inflammatory eosinophils (iEOSs) in type 2 allergic environments that have low CD62L and high CD101 on the cell surface.³⁵ We did not detect this pattern in E2 eosinophils, suggesting that in vitro culture alone may not regulate CD101, yet is sufficient for CD62L. Moreover, several recent studies have reported mixed expression of CD62L or CD101 in vivo,^5,17,53,54 indicating that iEOS may be an incomplete nomenclature for all immune subtypes of eosinophils. Similarly, cellular morphology has been a means of defining mouse eosinophil activation states, with mouse hypersegmented eosinophils being considered for activated or iEOSs.^34,35 Yet, we propose that morphology alone is insufficient to define eosinophil immune subtypes on the basis of our findings of nuclear morphological differences of E1 and E2 mouse eosinophils and lack of morphological changes in nuclei for human eosinophils.⁵⁵

Comparisons of transcriptomes of human and mouse eosinophils revealed many similar pathways for E2 or E1 subtypes, suggesting similar regulatory TFs controlled immune responses. None of the identified top upstream regulatory TFs included Gata-3 or Tbx21, which are important for lymphocytes type 2 and type 1 immune polarization, respectively.²⁶ Yet, Gata-3 was increased in both mouse and human E2 eosinophils and has been reported in eosinophils in allergic diseases.⁵⁶ Counterintuitively, Tbx21 was increased in E2 eosinophils, but not in E1 eosinophils. The reasons are unclear, particularly because E2 conditions did not induce IFN-γ, an activator of Tbx21.The STAT family also regulates many immune responses in leukocytes. Although E1 eosinophils had an increase in STAT1 as anticipated on the basis of reports by others,^10,44 STAT6 was not increased in E2 eosinophils in this culture system. Because STAT6 has been found to play a role in mouse and human eosinophils,^10,12,57 the low expression in E2 eosinophils may be due to timing, use of cytokine cocktails, or a need to measure phosphorylated proteins. The nuclear factor-κB family was identified as upstream TFs and is well known to contribute to activation responses of eosinophils,^10,12 yet this family was increased in both E2 and E1 subtypes, likely responding to IL-33¹² and TNF-α,⁵⁸ respectively. Analysis of transcriptomes for upstream regulatory TFs revealed members of the IRF gene family as modulated uniquely in E1 and E2 conditions. Little had been described previously for these TFs in eosinophil subtypes.

The IRF gene family (IRF1-9) is classically characterized as transcriptional regulators of interferon and pathogen recognition receptor pathways, yet increasingly has been found to play a central role in cell development, differentiation, metabolic function, and activation of specific immune responses.^27,28 A role for IRF4 in humans is suggested on the basis of expression patterns in allergic asthma^40,41 and regulation of IL-4.⁵⁹ Yet, despite regulation of IRF4 in eosinophils, our findings did not reveal major immune phenotype changes in IRF4-deficient eosinophils to either E2 or E1 conditions. The specific stimuli, microenvironment, or alternative TFs may contribute to the functions or compensate for IRF4 in our in vitro system.

In contrast to E2 eosinophils, E1 eosinophils had significant upregulation of other IRFs. IRF1, in particular, was an abundant IRF in mouse and human E1 eosinophils and was consistently identified as a major upstream regulatory TF in transcriptome enrichment analysis. IRF1 was increased by IFN-γ treatment, with a synergistic response to IFN-γ 1 TNF-α, as reported for other cells.⁴⁶ It was also upregulated in eosinophils from lung transplant,³³ mice infected with Helicobacter pylori,²⁰ and eosinophils treated with IFN-γ + E coli.⁴⁴ Certain subsets of intestinal eosinophils called A-EOS in mice and in human biopsies of colitis also showed expression of IRF1 and STAT1.¹⁷ In our cytokine system, mouse Irf1 expression was plastic and dependent on type 1 cytokines, indicating it may act as an on-off switch for downstream type 1 immune pathways. Here, Irf1 levels correlated with reductions in Stat1 and Batf2 levels. STAT1, Batf2, and IRF1 can act as cofactors to regulate inflammatory genes such as Nos2, CCL5, IL-12b, Cxcl10, and many others as well as levels of the TFs themselves.^46,60 Intriguingly, IRF1-deficient eosinophils had reduced Stat1, yet significantly increased Batf2 and Irf8, which also can form complexes with IRF1 to regulate transcription that is highly dependent on cell type and stimuli.^42,61 These TF complex interactions may explain some of the unexpected findings as compared with reports for other cells. For example, IRF1 has been reported as necessary for CXCL10 and CCL5 expression in other myeloid cells in response to IFN-γ,^46,60 yet it is increased in IRF1-deficient E1 eosinophils. Compensatory or altered pathways may contribute to these effects. For example, IFN-γ and TNF-α act synergistically to promote IRF1 functions in part through TNF-α–mediated upregulation of nuclear factor-κB.⁵⁸ These other activated TFs may change the nature of the TF complexes binding to IRF1 target genes. Additional studies that include measuring DNA binding sites of IRF1, cofactor interactions, and chromosome remodeling will aid in advancing our understanding of eosinophil-specific immune regulation by IRF1.

A most striking response of IRF1 knockout eosinophils was the significant upregulation of type 2 cytokines and chemokines (eg, IL-4, IL-13, and CCL17) in response to type 1 stimuli (ie, IFN-α and TNF-α) of E1 conditions. Moreover, CD11b and CD11c were increased and CD62L decreased, similar to a WT E2 eosinophil response. This dichotomy of type 2 responses upon exposure to type 1 stimuli suggests that IRF1 acts in part as a suppressor of type 2 immune responses in eosinophils. Such a role has been proposed for IRF1-deficient CD4 T-cell IL-4 production⁴⁷ and M1/M2 macrophage polarization⁴⁵ as well as other functions.⁴⁶

IRF1 performs these functions in part through metabolic reprogramming. For example, studies in macrophage have shown that IRF1 acts as a TF for the gene IRG1 that is necessary to catalyze cis-aconitate to itaconate, which is a metabolite that is important for promoting type 1 immune-related functions.^36,62 Irg1/ADOC was found to be increased in E1 eosinophil transcriptomes. In addition, Nos2, an enzyme to make nitric oxide, can limit mitochondrial respiration. These studies suggest that IRF1 in E1 eosinophils may modulate metabolic programming, although further studies are needed.

Despite these complex processes, we identified that surface expression of PDL1 was uniquely regulated by IFN-γ and IRF1 in mouse eosinophils. A study supports this finding and IRF1 was found to directly bind the promoter and regulate PDL1 induction by IFN-γ in cancer cells.⁶³ The role of PDL1 on eosinophils is an area of expanding research. PDL1 on eosinophils in a gut infection model of H pylori and an allograft lung transplant model demonstrated that it was critical to suppressing CD8 T-cell activities.^24,64 Gastrointestinal A-EOSs also express PDL1 and influence T cells in the gastrointestinal tract.¹⁷ In humans, PDL1 was found increased on a subpopulation of eosinophils in graft-versus-host disease,⁶⁵ eosinophilic esophagitis,⁶⁶ and on patients infected with severe acute respiratory syndrome coronavirus 2.⁵² The role of PDL1 on eosinophils in cancer, particularly with checkpoint blockade, remains to be defined.⁶ The studies here may indicate that an IFN-γ–IRF1 nexus regulates PDL1 expression in eosinophils.

Single nucleotide polymorphisms in the promotor site of IRF1 that result in reduced IRF1 expression are associated with increased IgE levels and atopy in patients with asthma⁶⁷ and increased risk and severity of asthma.⁶⁸ Elevated IRF1 is increased in airways of some subsets of people with severe asthma.^69,70 It has been noted in genome-wide association studies that IRF1 is associated with male-specific asthma susceptibility for asthma.⁷¹ We did not find sex-related differences in the studies completed here with specific cytokine stimuli induction. IRF1 also forms cooperative networks with other TFs that are extensive. This is exemplified in patients in whom abnormal IRF1 expression or function in other cell types is associated with increased risk of cancer, infection, autoimmune disease, transplant rejection, and ischemic or endotoxemia-induced injury.^28,72,73 Eosinophils are found to have immune modulating roles in many of these diseases. A greater understanding of the role of IRF1 in modulating eosinophil functions will require future endeavors that include in vivo models, human studies, and cell-specific deletion of IRF1 and related mediators. Nevertheless, the data here demonstrate that IRF1 is increased in mouse and human eosinophils exposed to type 1 environments and acts as an upstream regulatory TF of mouse eosinophil immune phenotype in response to type 1 cytokine environments.

Key messages.

• Human and mouse eosinophils have many similar type 1 and type 2 immune responses at the protein and transcriptome levels in response to in vitro type 1 and type 2 cytokine environments, respectively.

• The IRF TF family is modulated by cytokine environment in mouse and human eosinophils, with IRF1 increased as an upstream regulatory TF in mouse and human eosinophils in type 1 environments.

• IRF1 deficiency in mouse eosinophils results in an abnormal balance of type 1 and type 2 immune responses when challenged with IFN-γ and TNF-γ.

Abbreviations used

ChEA3

CHIP-X Enrichment Analysis 3

E0

Eosinophils treated with IL-5 only

E1

Eosinophils treated with IFN-γ, TNF-α, and IL-5

E2

Eosinophils treated with IL-4, GM-CSF, IL-33, and IL-5

FDR

False-discovery rate

ICAM-1

Intercellular adhesion molecule 1

iEOS

Inflammatory eosinophil

IRF

Interferon regulatory factor

PDL1

Programmed death ligand 1

RT-PCR

Real-time PCR

STAT

Signal transducer and activator of transcription

TF

: Transcription factor

WT

Wild type