Cytotoxic programming of CD4⁺ T cells is regulated by opposing actions of the related transcription factors Eos and Aiolos

Abstract

In contrast to the “helper” activities of most CD4⁺ T effector subsets, CD4⁺ cytotoxic T lymphocytes (CD4-CTLs) perform functions normally associated with CD8⁺ T and natural killer cells. Specifically, CD4-CTLs secrete cytotoxic molecules and directly target and kill compromised cells in an MHC class II-restricted fashion. The functions of these cells have been described in diverse immunological contexts, including their ability to provide protection during antiviral and antitumor responses, as well as being implicated in autoimmunity. Despite their significance to human health, the complete mechanisms that govern their programming remains unclear. Here, we identify the Ikaros zinc finger transcription factor Eos (Ikzf4) as a positive regulator of CD4-CTL differentiation during murine immune responses against influenza virus infection. We find that the frequency of Eos⁺ cells is elevated in lung CD4-CTL populations and that the cytotoxic gene program is compromised in Eos-deficient CD4⁺ T cells. Consequently, we observe a reduced frequency and number of lung-residing, influenza virus-responsive CD4-CTLs in the absence of Eos. Mechanistically, we determine that this is due, at least in part, to reduced expression of IL-2 and IL-15 cytokine receptor subunits on the surface of Eos-deficient CD4⁺ T cells, both of which support the CD4-CTL program. Finally, we find that Aiolos, a related Ikaros family member and known CD4-CTL antagonist, represses Eos expression by antagonizing STAT5-dependent activation of the Ikzf4 promoter. Collectively, our findings reveal a mechanism wherein Eos and Aiolos act in opposition to regulate cytotoxic programming of CD4⁺ T cells.

Introduction

CD4⁺ T helper cells function as critical coordinators of the adaptive immune response through their ability to differentiate into specific effector subsets that secrete cytokines and aid in the effective recruitment and activation of other immune cell populations (1). Beyond serving in these classical “helper” roles, an additional subset of CD4⁺ T cells exists that exhibits cytotoxic features typically attributed to CD8⁺ cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells (2–5). These CD4-CTLs are characterized by the ability to secrete granzymes and perforins and to directly target and kill compromised cells in a major histocompatibility complex (MHC) class II-dependent fashion (6, 7). CD4-CTLs have been found to contribute to antiviral and antitumor responses in both humans and mice, with their presence largely corresponding to favorable outcomes (2–5). Conversely, CD4-CTLs have been implicated in the pathogenesis of autoimmune diseases, including multiple sclerosis, ulcerative colitis, and systemic lupus erythematosus (2, 8–11). While the importance of CD4-CTLs in a broad range of immunological contexts has been demonstrated, the mechanisms that govern their differentiation and function remain incompletely understood.

The differentiation of individual CD4⁺ T cell effector subsets is dictated by complex, and sometimes overlapping, transcription factor regulatory networks. Transcriptionally, it is known that CD4-CTLs are positively regulated by the T helper 1 (T_H1) lineage-defining transcription factor T-bet, encoded by Tbx21 (2, 7). The related T-box transcription factor Eomesodermin (Eomes) and Runt-related transcription factor 3 (Runx3), each of which harbors prominent roles in CD8⁺ cytotoxic T cells, have also been identified as drivers of CD4-CTL responses (3, 12–14). The transcription factor class I MHC-restricted T cells-associated molecule (CRTAM) has also been implicated in the positive regulation of CD4-CTLs, acting as an upstream inducer of Eomes and other cytotoxic-related genes (15). Finally, the transcription factor B lymphocyte-induced maturation protein 1 (Blimp-1) is also important for the acquisition of cytotoxic features while also repressing alternative cell fates, including T follicular helper (T_FH) cell programming (16–18). In contrast, several T_FH-associated transcription factors, including T cell factor 1 (Tcf-1, encoded by Tcf7) and B cell lymphoma 6 (Bcl-6), have been identified as repressors of CD4-CTL differentiation (17).

The expression and activity of the above factors is driven by signals received from environmental cytokines. Notably, Interleukin-2/Signal Transducer and Activator of Transcription 5 (IL-2/STAT5) signaling is a known promoter of CD4-CTL differentiation, at least partially through the induction of Blimp-1 and Eomes expression (16, 19, 20). IL-15 signaling, like IL-2, is also propagated via STAT5 and has been implicated in the differentiation of CD4-CTLs (21). While STAT5 is necessary for the direct induction of several CD4-CTL effector molecules and transcription factors, it is also capable of competing with and antagonizing the activity of STAT3 (22). This antagonism may allow for the safeguarding of the CD4-CTL program, as STAT3 is critical for promoting transcriptional repressors that oppose the cytotoxic program, such as Bcl-6 (22). Despite these insights, the comprehensive identity of the regulatory network that directs CD4-CTL formation is unclear.

Members of the Ikaros zinc finger family of transcription factors (IkZF) play numerous roles in the differentiation and function of mature T helper cell populations (reviewed extensively in (23–25)). IkZF family members exhibit a high degree of homology, and each contains N-terminal zinc finger DNA-binding and C-terminal zinc finger protein-protein interaction domains (23–25). IkZF family members were initially defined as key regulators of immune cell development, mediating their activity via direct interaction with target gene loci and the recruitment of chromatin remodeling complexes (23–25). Additionally, IkZF family members have emerged as regulators of cytokine signaling pathways (23, 26). We have previously reported that Aiolos (Ikzf3) both cooperates with STAT3 to induce Bcl-6 expression and represses IL-2/STAT5 signaling to promote T_FH cell programming (27, 28). Aiolos has also been implicated in the promotion of T_H17 populations, at least in part, by repressing IL-2 expression (29). A second Ikaros family member, Eos (Ikzf4), has been identified as a positive regulator of T_H2 and T_REG cell formation, while also functioning to repress T_H17 differentiation (23, 27, 28, 30–32). Thus, the collective literature suggests that there are subset-specific roles for Aiolos and Eos in regulating CD4⁺ T cell differentiation decisions.

Here, we identify Aiolos and Eos as oppositional regulators of the CD4-CTL differentiation program. We find that Aiolos-deficient CD4⁺ T cells have elevated Eos expression and that this positively correlates with an increase in cytotoxic programming. We further find that murine antigen-specific CD4-CTLs responding to influenza virus infection have elevated Eos expression relative to other effector CD4⁺ T cell subsets. Consistent with this observation, Eos deficiency results in a decrease in cytotoxic programming and, consequently, a reduction in influenza virus-specific CD4-CTLs. Mechanistically, we find that Eos deficiency results in reduced expression of CD25 and CD122, cytokine receptor subunits that propagate signals received from IL-2 and IL-15. Finally, we find that Eos itself is induced by both IL-2 and IL-15 signaling and that Aiolos represses Eos expression by antagonizing STAT5 association with the Ikzf4 promoter. Collectively, this work identifies a counter-regulatory axis composed of the Ikaros factors Aiolos and Eos that plays a role in modulating the cytotoxic potential of CD4⁺ T cells.

Materials and Methods

This research complies with all ethical regulations defined by the Institutional Animal Care and Use Committee (IACUC) and Institutional Biosafety Committee (IBC) of The Ohio State University.

Mouse strains and cell lines

Wildtype (WT) CD45.1 and CD45.2 C57BL/6 mice were originally obtained from the Jackson Laboratory. Aiolos-deficient (Ikzf3^−/−) mice were originally obtained from Riken BRC and were backcrossed onto the C57BL/6J background for over 10 generations. Eos-deficient (Ikzf4^−/−) mice were a generous gift of Dr. Ethan Shevach (NIH) and Dr. Bruce Morgan (Harvard Medical School). OT-II mice, which have the transgene located on the Y-chromosome, were originally generated by the Carbone laboratory (33). For these studies, OT-II mice were a generous gift of Dr. Haitao Wen (The Ohio State University). For adoptive transfer studies, Ikzf3^−/− and Ikzf4^−/− mice were crossed onto the OT-II background mice to generate OT-II-Ikzf3^−/− and OT-II-Ikzf4^−/− mice. For all adoptive transfer studies, only male mice were utilized due to the Y-linked nature of the OT-II transgene. For all in vitro experiments, male and female mice were used to avoid unintentional sex bias. For all experiments and replicates (both in vivo and in vitro) individual mice were age- and sex-matched. All the above experiments performed utilizing mice were done with the approval of the Institutional Animal Care and Use Committee (IACUC) at The Ohio State University in Columbus, OH.

The murine EL4 thymoma T cell line (TIB-39) was obtained from the American Type Culture Collection (ATCC) and was maintained as previously described in complete RPMI (RPMI-1640, 10% FBS [26140079, Life Technologies], 1% Pen/Strep [Life Technologies])(34).

CD4⁺ T cell isolation and culture

Naïve CD4⁺ T cells were isolated from the spleens and lymph nodes of 5-8 week old mice using the BioLegend MojoSort naïve CD4⁺ T cell isolation kit per the manufacturer’s protocol. For the in vitro polarization of T_H1 cell populations, naïve cells were plated at a density of 300,000 cells/well in complete IMDM (IMDM [Life Technologies], 10% FBS [26140079, Life Technologies], 1% Penicillin Streptomycin [Life Technologies], and 50 μM 2-mercapto-ethanol [Sigma-Aldrich]). Plates were coated with plate-bound anti-CD3 (clone 145-2C11; 5 μg/mL) and anti-CD28 (clone 37.51; 2 μg/mL) overnight and washed twice with PBS just prior to addition of cells in cIMDM. Immediately upon plating, cells were cultured in the presence of IL-4-neutralizing antibody (clone 11B11, BioLegend, 5 μg/mL) and where appropriate, IL-2-neutralizing antibody (JES6-1A12, BioLegend, 5 μg/mL) for 3-5 hours. Following this, the T_H1-polarizing cytokine mIL-12 was added (R&D, 5 ng/mL). For some experiments, concurrent with IL-12 addition, rhIL-2 was added (Peprotech, 50-250U/mL). For experiments where IL-2 was neutralized, an IL-15/IL-15Rα complex was added at a concentration of 125 ng/mL. The IL-15/IL-15Rα complex was generated as previously described(35). Briefly, soluble IL-15 (447-ML, R&D Systems) was combined with a recombinant mouse IL-15Rα Fc chimera (551-MR, R&D Systems) at a 1:2 ratio, respectively. This combination was then incubated at 37°C for 45 minutes just prior to addition to cell suspensions. Cells were cultured under the above conditions for 48-72 hours prior to harvesting for downstream applications and analyses.

RNA isolation and qRT-PCR

RNA was isolated from cells generated as described above utilizing the Machery-Nagel Nucleospin RNA kit per the manufacturer’s guidelines. cDNA was then generated using the Superscript IV First Strand Synthesis System (Thermo Fisher Scientific). qRT-PCR reactions were performed utilizing the SYBR Select Mastermix for CFX (Thermo Fisher Scientific) with 4-20 ng cDNA per reaction. The following primers were used to assess Ikzf4 expression: Forward: 5’-GACGCACTCACTGGCCACCTCC-3’, Reverse: 5’-GGCACCTCTCCTTGTGCTCCTCC-3’. All qRT-PCR reactions were performed on the CFX Connect (BioRad). Data outputs were normalized to Rps18 and are presented as either relative to Rps18 or as relative to their relative control, as indicated.

RNA sequencing analysis

Naïve WT, Ikzf3^−/− and Ikzf4^−/− T cells were cultured under T_H1-polarizing conditions with or without IL-2 as described above. 0.5-1.0x10⁶ cells were harvested using the Machery-Nagel Nucleospin kit as above, and purified total RNA was provided at 1 μg to Azenta Life Sciences for polyA selection, library preparation, and sequencing. For Aiolos-deficient RNA-seq, data are representative of 3 biological replicates per genotype from 3 independent experiments. For Eos-deficient RNA-seq, data are representative of 4 biological replicates per genotype from 4 independent experiments and were compiled from 2 sequencing runs. Briefly, sequence reads were trimmed by Azenta to remove possible adapter sequences and nucleotides with poor quality (% bases < 30) using Trimmomatic v.0.36. The trimmed reads were mapped to the Mus musculus GRCm38 reference genome available on ENSEMBL using the STAR aligner v.2.5.2b. BAM files were generated as a result of this step. Unique gene hit counts were calculated by using featureCounts from the Subread package v.1.5.2. The hit counts were summarized and reported using the gene_id feature in the annotation file. Only unique reads that fell within exon regions were counted. If a strand-specific library preparation was performed, the reads were strand-specifically counted. After extraction of gene hit counts, the gene hit counts table was used for downstream differential expression analysis. For Aiolos-deficiency RNA-seq, DESeq2, a comparison of gene expression between the customer-defined groups of samples was performed by Azenta, and the DESeq2 Wald test was used to generate p-values and log2 fold changes. For Eos-deficiency RNA-seq, following Trimmomatic trimming, alignment to Mus musculus GRCm38 was performed by P.L.C. and D.M.J. using Kallisto 0.43.1, and gene level outputs were generated as a result. DESeq2 analysis and batch normalization were performed by P.L.C. and D.M.J. utilizing the Bioconductor DESeq2 package in RStudio via the standard pipeline (36). Genes with a p-value of ≤ 0.05 were considered differentially expressed. For gene set analysis, genes were pre-ranked by multiplying the sign of the fold-change by −log₁₀(p-value) and were analyzed using the Broad Institute Gene Set Enrichment Analysis (GSEA) software, version 4.3.2, for comparison against “gene ontology,” “hallmark,” “immunological signatures,” and “curated” gene sets. Enrichment gene sets with a false discovery rate (FDR) q-value of <0.25 were considered significant. Heatmaps were generated and clustering (by Euclidian distance) was performed using normalized log2 counts from DESeq2 analysis and the Morpheus software (https://software.broadinstitute.org/morpheus). ClustVis (https://bio.tools/clustvis) was used to generate PCA plots.

Immunoblot analysis

Immunoblot analysis was performed as previously described (28, 34). Briefly, cell pellets were harvested, lysed in 1X loading dye, (50 mM Tris [pH 6.8], 100 mM DTT, 2% SDS, 0.1% bromophenol blue, and 10% glycerol) and boiled for 15 minutes. Cell lysates were then separated via SDS-PAGE on 10% Bis-Tris Bolt gels (Thermo Fisher Scientific) and then transferred onto a 0.45 μm nitrocellulose membrane. Following transfer, nitrocellulose membranes were blocked with 2% nonfat dry milk in 1X TBST (10 mM Tris [pH 8], 150 mM NaCl, 0.05% Tween-20). Detection of the indicated proteins was performed using the following antibodies and concentrations: α-Eos (1:500, W15169A, BioLegend), α-Aiolos (1:20,000, 39293, Active Motif), α-STAT5b (1:5,000, sc-1656, Santa Cruz), α-V5 (1:20,000 R960-25, Invitrogen) α-β-actin-HRP (1:15,000, A00730, GenScript), goat α-mouse (1:5,000-1:10,000, 115-035-174, Jackson Immunoresearch), and mouse α-rabbit (1:15,000, sc-2357, Santa Cruz).

Influenza virus infection and tissue preparation

Naïve CD45.2⁺ OT-II CD4⁺ T cells were purified from Ikzf3/4^+/+ OT-II-WT, OT-II-Ikzf3^−/−, or OT-II-Ikzf4^−/− mice using negative selection as described above. Cells were washed and resuspended in sterile 1X PBS for retro-orbital transfer (5x10⁵ cells/animal) into WT CD45.1 recipient mice. 24 hours after retro-orbital transfer, mice were infected intranasally with approximately 40 PFU of OT-II-specific OVA_323-339-expressing influenza virus strain A/Puerto Rico/8/34, referred to as “PR8-OVA”. Influenza virus was propagated for 48 hours in 10-day embryonated chicken eggs and tittered in MDCK cells. 8 days after influenza virus infection, draining lymph node (DLN), spleen, and/or lungs were harvested, and single-cell suspensions were generated as previously described (27). Briefly, for DLN and spleen, single-cell suspensions were generated in tissue processing media (IMDM + 4% FBS) by passing the tissue through a 100 μm nylon mesh strainer. For lungs, single-cell suspensions were generated by incubating whole lung tissue with HBSS (Gibco) + 1.3 mM EDTA for 30 minutes at 37°C. Following this, lungs were dissociated in Collagenase IV-supplemented media (RPMI + 4% FBS) using the Miltenyi Biotec gentleMACS Dissociator for 30 minutes according to the manufacturer’s protocol. Dissociated tissues were then filtered through a 40 μm mesh strainer, layered via Percoll density gradient and centrifugation, and mononuclear layers were harvested. For all tissues, erythrocytes were lysed and cells were washed once in FACS buffer (PBS + 4% FBS) prior to staining for flow cytometry. For ex vivo peptide stimulation, whole tissue homogenates generated as above were cultured in complete IMDM supplemented with OVA_323-339 peptide (AnaSpec) at 5 μg/mL for 48 hours, with protein transport inhibitors added at 45 hours. Cells were then harvested, washed once with FACS buffer, and stained for flow cytometry.

Flow cytometry

Single-cell suspensions were generated as described above from the DLN, lung, and spleen. Prior to extracellular staining, cells were incubated with an appropriate volume of Fc Block (clone 93; BioLegend) for a minimum of 5 minutes at 4°C. Extracellular staining antibodies were then added on top of Fc block and incubated for 30 minutes at 4°C protected from light. The following antibodies and dilutions were utilized for extracellular staining: CD4 (1:300; clone GK1.5; R&D, catalog # for anti-CD4:A700 FAB554N100), CD45.2 (1:300; clone 104; BioLegend; catalog # for anti-CD45.2:AF488: 109816), CD45.1 (1:300; clone A20; BD Biosciences, catalog # for anti-CD45.1:BUV395: 565212), CD44 (1:300; clone IM7, BioLegend, catalog # for anti-CD44:BV785: 103041), CD62L (1:300; clone MEL-14, BioLegend, catalog # for anti-CD62L:BV650: 104453), CD122 (1:100; clone 5H4, BD Biosciences, catalog # for anti-CD122:BUV661: 741-537), CD25 (1:300; clone PC61, BioLegend, catalog # for anti-APC:CD25: 102012), NKG2A/C/E (1:100; clone 20D5; BioLegend, catalog # for anti-NKG2A/C/E:BUV737: B7418), and Cxcr5 (1:100; clone SPRCL5; Thermo Fisher Scientific; catalog # for anti-Cxcr5:PE: 12-7185-8). Samples were concurrently stained with a set of exclusionary (dump gate) antibodies which include: CD45R/B220 (1:300; clone RA3-6B2; BioLegend, catalog # for anti-D45R/B220:BV510: 103247), CD11b (1:300; clone M1/70; BioLegend, catalog # for antiCD11b:BV510: 101245), F4/80 (1:300, clone BM8, BioLegend, catalog # for anti-F4/80:BV510: 123135), and CD11c (1:300, clone N418, BioLegend, catalog # for anti-CD11c:BV510: 117337). Cells were also stained with viability dye (1:750, Tonbo Biosciences, catalog # for Red780: 13-0865-T100). Cells were then washed once with FACS buffer prior to intracellular staining. For intracellular staining, cells were fixed and permeabilized using the eBioscience Foxp3 transcription factor staining kit (Thermo Fisher Scientific) for 30 minutes at room temperature, or overnight at 4°C. Following fixation, samples were stained with the following antibodies in 1X eBiosciences permeabilization buffer for 30 minutes at room temperature: Foxp3 (1:300; clone FJK-16s; Thermo Fisher Scientific; catalog # for anti-Foxp3:PerCP-Cy5.5: 45-5773-82); Aiolos (1:100; clone S48-791; BD Biosciences; catalog # for anti-Aiolos:AF647: 565265); Eos (1:50; clone ESB7C2; Invitrogen; catalog # for anti-Eos:PE: 12-5758-82); Eomes (1:100; clone Dan11mag; Invitrogen; catalog # for anti-Eomes:PE-Cy7: 25-4875-80); IFN-γ (1:400; XMG1.2; BioLegend; catalog # for anti-IFNg:BV650: 505831); Granzyme B (1:300; clone GB11; Thermo Fisher Scientific; catalog # for anti-Granzyme B:PE-Cy5.5: GRB18); and Perforin (1:100; S16009A; BioLegend; catalog # for anti-Perforin:APC: 154304). Cells were washed with 1X eBiosciences permeabilization buffer and resuspended in FACS buffer for analysis. Samples were run on a BD FACS Symphony and analyzed using FlowJo software (version 10.8.1).

ATAC-seq analysis

Naïve WT and Aiolos-deficient CD4⁺ T cells were cultured as above. Assay for Transposase-Accessible Chromatin with sequencing (ATAC-seq) analysis was performed as described (37). Briefly, 5×10⁴ cells with greater than 95% viability were processed with DNA Library Preparation Kit (Illumina Nextera) per the manufacturer’s guidelines to generate paired-end 42 bp sequencing reads. Resultant sequences were trimmed and aligned to mm10 using Bowtie2 (version 2.4.4). All subsequent analyses were performed utilizing the indicated tools in Galaxy (usegalaxy.org). Samples with low read quality (% bases < 30), duplicates, mitochondrial reads, and improperly paired reads were filtered out. Statistically significant peaks were identified using MACS2 callpeak. DiffBind was used to identify regions of significant differential accessibility between WT and Aiolos-deficient samples for each cell type.

Promoter-reporter assay

An Ikzf4 promoter-reporter construct (pGL3-Ikzf4) was generated by cloning the regulatory positions of the Ikzf4 gene (2 kb upstream of the transcriptional start site) into the Promega pGL3-Basic vector (Promega). Expression vectors for constitutively active STAT5b (STAT5b^CA), Aiolos, and Aiolos DNA-binding mutant (Aiolos^DBM) were generated as previously described (28). Briefly, STAT5b and Aiolos vectors were generated by cloning the coding sequence into the pcDNA3.1/V5-His-TOPO^® vector (K4800, Life Technologies). Appropriate sequences were confirmed by sequencing with T7 and BGH primers. Coding sequences were then transferred into pEF1/V5-His vector (V920, Life Technologies). Generation of STAT5b^CA and Aiolos^DBM utilized the above methods combined with the Agilent QuikChange kit (200519) as previously described (28). For generation of STAT5b^CA, point mutations were introduced into STAT5b (H298R and S715F), resulting in constitutive phosphorylation of the Y699 residue. For generation of Aiolos^DBM, point mutations were introduced into the N-terminal zinc finger domain to disrupt the first and second zinc fingers. Expression of the described proteins was confirmed via immunoblot as described above, using both anti-V5 and protein-specific antibodies. EL4 cells were then nucleofected with pGL3-Ikzf4 in conjunction with the indicated overexpression vectors (STAT5b^CA, Aiolos, or Aiolos^DBM) and SV40 Renilla as a control for transfection efficiency using the Lonza 4D Nucleofection system (program CM-120, SF buffer). Cells were incubated at 37°C for 18-24 hours. Following this, cells were harvested, and luciferase expression analyses were performed using the Promega Dual Luciferase Reporting System per the manufacturer’s guidelines.

Primary T cell transduction

For Aiolos overexpression in primary murine CD4⁺ T cell populations, the Aiolos (Ikzf3) coding sequence was cloned into the pMSCV-IRES-GFP II vector backbone (pMIG II, Addgene 52107) vector backbone, resulting in “pMIG-Aiolos”. Vector sequence was validated via sequencing and overexpression was validated via both qRT-PCR and immunoblot using both anti-V5 and anti-Aiolos antibodies.

For generation of retroviral supernatant, the Platinum-E (Plat-E) Retroviral Packaging Cell Line (RV-101, Cell Biolabs, Inc) was maintained per the manufacturer’s protocol. Plat-E cells were seeded in 10 cm culture dishes overnight. The following day, empty pMIG vector (pMIG(-)) or pMIG-Aiolos in combination with pCL-Eco (12371, Addgene) and polyethylenimine (PEI) were added into the Plat-E cell culture for 4-6 hours. Media was changed and viral supernatant was collected once 48 hours later, immediately frozen, and stored at −80°C until use. Viral supernatant was then added to primary murine T cells on anti-CD3/anti-CD28 stimulation with T_H1-polarizing cytokines, as described above. Cells were then transduced via spin-infection in the presence of 8 μg/mL polybrene (Sigma-Aldrich) as previously described (27). Spin-infection was performed on two consecutive days (days 1 and 2 post-plating). Cells were harvested on day 3.

Chromatin immunoprecipitation (ChIP)

ChIP assays were performed as described previously (34). Briefly, naïve WT and Aiolos-deficient CD4⁺ T cells were cultured under T_H1-polarizing conditions as described above. Chromatin was isolated, fragmented, and immunoprecipitated with antibodies against STAT5 (R&D AF2168, 5 μg/IP) or IgG control (Abcam ab6709; 5 μg/IP). Enrichment of the indicated proteins was analyzed via qPCR (Ikzf4 promoter primers: Forward 5’-CACATACACCCTGGGCTGAG-3’, Reverse 5’-GGAAGCGTGGATTCCTGGAAGTG-3’). Samples were normalized to total DNA input controls and percent enrichment from isotype control antibodies.

Statistics and reproducibility

All statistical analyses were performed using the GraphPad Prism software (version 9.5.1). For single comparisons, unpaired or paired Student’s t-test were performed. For multiple comparisons, one-way ANOVA with Tukey’s multiple comparisons test was performed. Error bars indicate the standard error of the mean. P values <0.05 were considered statistically significant.

Software summary

Data were collected utilizing the following open-source or commercially available software programs: BD FACSDiva (version 8.0.2), BioRad Image Lab (version 6.0.1, build 34), BioRad CFX Manager (version 3.1). ATAC-seq data was generated by Illumina Novaseq SP. RNA-seq was performed by Azenta Life Sciences (formerly Genewiz). DESeq2 analysis for Eos-deficient samples was performed using the DESeq2 package in RStudio (version 2023.03.0+386) (36). Analyses and/or manuscript preparation were conducted using the Microsoft Office Suite, version 16.43 (including Microsoft Word, Microsoft Excel, and Microsoft PowerPoint), BD FlowJo (version 10.8.1), and open-source software, including tools available on Galaxy (http://usegalaxy.org): Bowtie2 (2.4.4), MACS2, Integrative Genomics Viewer (version 2.9.4), Morpheus (Broad Institute), and ClustVis (https://bio.tools/clustvis). All statistical analyses were performed using the GraphPad Prism software (version 9.3.1). Data preparation for this manuscript did not require the use of custom code or software.

Data availability

RNA-seq and ATAC-seq data have been made publicly available on GEO for WT vs Ikzf3^−/− RNA-seq (GSE203065, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE203065) and ATAC-seq (GSE203064, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE203064) together under GSE203066, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE203066, as well as WT vs Ikzf4^−/− RNA-seq (GSE147428). Publicly available data sets under the GEO accession number GSM1865310 were also used in this study. All other data sets and materials from this study will be made available upon reasonable request, which should be sent to the corresponding author.

Results

Eos expression is elevated in the absence of Aiolos

We previously demonstrated that the IkZF transcription factor Aiolos (Ikzf3) restrains acquisition of cytotoxic features by CD4⁺ T cells (27). Intriguingly, analysis of previously published RNA-seq data from wildtype (WT) and Aiolos-deficient (Ikzf3^−/−) T_H1-polarized cells revealed that the expression of a second IkZF transcription factor, Eos (Ikzf4), was elevated in Ikzf3^−/− cells and that its expression positively correlated with cytotoxic gene signatures (Fig. 1A, Supplemental Table I). Gene set enrichment analyses (GSEA) of “gene ontology”, “immunological signatures”, and “curated” gene sets also showed that cytotoxic programming was upregulated in the absence of Aiolos (Fig. 1B). Subsequent transcript and protein analyses confirmed that Eos expression was elevated in Aiolos-deficient CD4⁺ T cells cultured under T_H1-polarizing conditions, relative to WT (Fig. 1C,D).

Figure 1. Eos expression is elevated in the absence of Aiolos.

(A-D) Naïve WT and Aiolos-deficient (Ikzf3^−/−) CD4⁺ T cells were cultured under T_H1-polarizing conditions for 3 days. (A-B) RNA sequencing was performed to assess differentially expressed genes (DEGs). (A) Representative heatmap of DEGs positively and negatively associated with the CD4-CTL program in WT versus Aiolos-deficient cells. Differential expression is presented as row (gene) Z-score and are compiled from 3 independent experiments. (B) Genes were pre-ranked using (sign of fold change x −log₁₀(p-value)) and analyzed using the Broad Institute’s Gene Set Enrichment Analysis (GSEA) software for comparison against “gene ontology”, “immunological signatures”, and “curated” gene sets. Enrichment plots for the indicated gene sets are shown. (C) qRT-PCR analysis of the indicated genes, normalized to Rps18 and presented as relative to the WT sample. Data are compiled from 4 independent experiments (n = 4 ± s.e.m.; **P<0.01; two-tailed, unpaired Student’s t-test). (D) Immunoblot analysis of the indicated proteins. β-actin serves as a loading control. Representative image from 3 independent experiments is shown. (E,F) Naïve CD4⁺ T cells were harvested from the spleens and lymph nodes of OT-II-WT and OT-II-Ikzf3^−/− CD45.2⁺ mice. 500,000 cells/animal were adoptively transferred via retro-orbital injection into CD45.1 recipients. 24 hours after adoptive transfer, recipient mice were infected with 40 PFU OVA_323–339-expressing A/Puerto Rico/8/34 (“PR8-OVA”). 8 days after infection, spleens were harvested and viable CD4⁺CD45.2⁺CD44⁺CD62L⁻Cxcr5⁻Foxp3⁻ effector cell populations were analyzed via flow cytometry. Counts data represent the number of cells collected per 800,000 total events for all samples. Data are representative of 2 independent experiments (n = 9-10 ± s.e.m.; *P<0.05; two-tailed, unpaired Student’s t-test).

To assess any reciprocal nature of Eos and Aiolos expression in vivo, we adoptively transferred naïve CD4⁺CD45.2⁺ OT-II WT or Aiolos-deficient (OT-II-WT or OT-II-Ikzf3^−/−, respectively) cells into WT CD45.1 mice. Importantly, OT-II cells express a transgenic T cell receptor that is specific to the chicken ovalbumin 323-339 peptide. We then infected recipient animals with OVA_323-339-expressing PR8 influenza virus (PR8-OVA) 24 hours post-transfer (Supplemental Fig. 1A). Flow cytometry analysis of antigen-specific, CD45.2⁺ splenic donor cell populations at 8 days post-infection (dpi) revealed an increase in the frequency of Eos⁺ cells in Aiolos-deficient effector cells relative to WT counterparts (Fig. 1E and Supplemental Fig. 1A,B). We observed no difference in the number of activated, CD45.2⁺ donor cells in the spleen at day 8 dpi, suggesting that the increase in Eos expression was not due to alterations in cell survival or proliferation in the absence of Aiolos (Fig. 1F). Taken together, these data demonstrate that Eos expression inversely correlates with that of Aiolos and suggest that Eos may promote cytotoxic programming of CD4⁺ T cells.

Eos expression is enhanced in lung CD4-CTL populations

We next assessed the relative levels of Eos protein expressed across different CD4⁺ T cell subsets and tissues during immune responses to influenza virus infection. We again adoptively transferred naïve CD4⁺CD45.2⁺ OT-II-WT cells into CD45.1 recipient mice, which were then infected with PR8-OVA and sacrificed at 8 dpi for analysis (Supplemental Fig. 1A). To compare Eos expression in numerous T helper cell populations, we first analyzed recipient animal (CD4⁺CD45.1⁺) T cell populations in the lung-draining lymph node (DLN) (Supplemental Fig. 2A). Consistent with previously defined roles for Eos in regulating T_REG populations, we found that Eos expression was significantly elevated in T_REG cells (CD4⁺CD45.1⁺CD44⁺CD62L⁻Cxcr5⁻Foxp3⁺) when compared to both naïve (CD4⁺CD45.1⁺CD44⁻CD62L⁺) and non-T_FH effector (CD4⁺CD45.1⁺CD44⁺CD62L⁻Cxcr5⁻Foxp3⁻) populations (Supplemental Fig. 2A) (31, 38–40). We noted that non-T_FH effector CD4⁺ T cell populations also displayed a significant increase in Eos expression when compared to naïve cell populations, suggesting that Eos may have a functional role in effector CD4⁺ T cell subsets during a type 1 immune response (Supplemental Fig. 2A).

We next examined Eos expression in CD45.2⁺ donor cell populations, including non-T_FH/non-T_REG effector cells (CD4⁺CD45.2⁺CD44⁺CD62L⁻Cxcr5⁻Foxp3⁻) in the spleen and CD4-CTL cells (CD4⁺CD45.2⁺CD44⁺CD62L⁻Cxcr5⁻Foxp3⁻NKG2A/C/E⁺), which express NKG2A/C/E and are found predominantly in the lungs of infected animals (Fig. 2A) (41). Consistent with a role for Eos in regulating cytotoxic programming, we observed significantly increased frequencies of Eos⁺ cells in lung CD4-CTLs compared to either NKG2A/C/E⁻ effector cells from the lungs or non-T_FH/non-T_REG cells from the spleen (Fig. 2B). Collectively, these data demonstrate that the frequency of Eos⁺ cells is increased in antigen-specific effector CD4⁺ T cell subsets relative to naïve CD4⁺ T cells, with the highest levels found in lung CD4-CTL populations.

Figure 2. Eos is differentially expressed in effector CD4⁺ T cell populations.

(A-D) Naïve CD4⁺ T cells were harvested from the spleens and lymph nodes of OT-II-WT or OT-II-Ikzf4^−/− CD45.2 mice. 500,000 cells/animal were adoptively transferred into CD45.1 recipients. Recipient mice were infected with 40 PFU OVA_323–339-expressing A/Puerto Rico/8/34 (“PR8-OVA”) 24 hours post-transfer. 8 days post-infection, lungs and spleens were harvested. WT CD4⁺CD45.2⁺CD44⁺CD62L⁻Cxcr5⁻Foxp3⁻ antigen-specific donor effector cell populations were analyzed via flow cytometry. (A) Percent of NKG2A/C/E⁺ cells in spleens and lungs. Data are representative of 3 independent experiments (n = 5 ± s.e.m; ***P<0.001; two-tailed, unpaired Student’s t-test). (B) Percent of Eos⁺ cells in spleens and lungs. Data are representative of 3 independent experiments (n = 5 ± s.e.m.; *P<0.05, **P<0.01; one-way ANOVA with Tukey’s multiple comparisons). (C,D) Analysis of whole lung homogenates via flow cytometry for the indicated proteins. Counts data represent the number of cells collected per 400,000 total events for all samples. Data are from 4 independent experiments (n = 12 ± s.e.m.; *P<0.05, ***P<0.001; two-tailed, unpaired Student’s t-test).

Eos deficiency compromises CD4-CTL formation in the lungs of influenza-infected mice

Given the increased frequency of Eos in CD4-CTLs, we next sought to examine the cell-intrinsic impact of Eos deficiency on CD4-CTL formation. To do so, we adoptively transferred naïve WT or Eos-deficient (OT-II-WT or OT-II-Ikzf4^−/−, respectively) CD4⁺CD45.2⁺ T cells into CD45.1 recipient mice (Supplemental Fig. 1A). As before, recipients were infected with PR8-OVA influenza virus and CD45.2⁺ donor cells in the lungs were examined at 8 dpi. Indeed, consistent with a role for Eos in positively regulating CD4-CTL differentiation, we found that Eos-deficient donor cells from the lungs exhibited significantly reduced frequencies and numbers of NKG2A/C/E⁺ CD4-CTL cells relative to their WT counterparts (Fig. 2C,D).

Eos deficiency results in reduced CD4-CTL effector function

We next examined functional features associated with CD4-CTLs. For these experiments, we again employed the above approach by transferring OT-II-WT and OT-II-Ikzf4^−/− CD4⁺CD45.2⁺ donor T cells into CD45.1 recipients and this time stimulated lung-derived cell homogenates at 8 dpi ex vivo with OVA_323-339 peptide. Consistent with the observed decrease in CD4-CTLs, CD45.2⁺ Eos-deficient donor cells produced significantly lower levels of the cytolytic molecules perforin and granzyme B (GzmB) (Fig. 3A,B). CD4-CTLs are also known to produce elevated levels of the pro-inflammatory cytokine IFN-γ. We similarly found that Eos-deficient cells had significantly lower IFN-γ expression compared to WT controls (Fig. 3C).

Figure 3. Cytotoxic features and Eomes expression are reduced in the absence of Eos.

(A-D) Naïve CD4⁺ T cells were harvested from the spleens and lymph nodes of OT-II-WT or OT-II-Ikzf4^−/− CD45.2 mice. 500,000 cells/animal were adoptively transferred into CD45.1 recipients. Recipient mice were infected with 40 PFU OVA_323–339-expressing A/Puerto Rico/8/34 (“PR8-OVA”) 24 hours post-transfer. 8 days post-infection, lungs were harvested. (A-C) Lung homogenates were stimulated ex vivo for 48 hours with OVA_323–339 peptide. Homogenates were treated with protein transport inhibitors for the last 4 hours of ex vivo stimulation. (A-D) Antigen-specific CD4⁺CD45.2⁺CD44⁺ or CD4⁺CD45.2⁺CD44⁺CD62L⁻ cells were then analyzed via flow cytometry for the expression of the indicated proteins. Data are representative of 4 independent experiments (n = 10-11 ± s.e.m.; *P<0.05; two-tailed, unpaired Student’s t-test).

We also assessed expression of the T-box transcription factor Eomes, a known positive regulator of cytotoxic programming in CD4⁺ T cells. Indeed, Eomes expression was reduced in Eos-deficient cells in the lungs, and this observation was consistent with the decreased frequency of CD4-CTL cells and reduced cytotoxic capabilities (Fig. 3D). T_H1 cells have been implicated as precursors for the generation of lung CD4-CTLs in certain T_H1-biased immunological contexts (3, 42). Therefore, we also assessed whether T-bet^HI T_H1 cell populations were impacted by Eos deficiency, potentially explaining the reduction in CD4-CTLs. However, we did not detect a significant difference in either the frequency or number of T-bet^HI T_H1 between OT-II-WT and OT-II-Ikzf4^−/− donor cells in the lungs or DLN (Supplemental Fig. 2B,C). Collectively, these data demonstrate that Eos preferentially supports Eomes expression and cytotoxic programming in a CD4⁺ T cell-intrinsic manner.

Eos deficiency results in broad dysregulation of the cytotoxic gene program

We next wanted to gain insight into how Eos deficiency may impact cytotoxic transcriptional programming of CD4⁺ T cells. As CD4-CTLs and T_H1 cells share environmental and transcriptional regulatory features, we cultured naïve WT and Eos-deficient (Ikzf4^−/−) CD4⁺ T cells under T_H1-polarizing conditions and performed RNA-seq analyses (5). Both hierarchical clustering by Euclidian distance and principal component analysis (PCA) of DESeq2-normalized reads found that WT and Ikzf4^−/− samples formed distinct clusters by genotype (Fig. 4A,B). Further, of the top 200 differentially expressed genes (DEGs), we found that the Ikzf4^−/− samples exhibited increased expression of 51 genes and decreased expression of 149 genes when compared to WT (Fig. 4A, Supplemental Table I). We next examined hallmark CD4-CTL genes and found that the absence of Eos resulted in a decrease in transcriptional regulators (Hopx and Prdm1), effector molecules (Gzmb, Prf1 and Ifng), and cell surface receptors (Klrk1, Klrc1, Nkg7, Il2ra, and Il2rb) associated with cytotoxic programming (Fig. 4C). The decrease in CD4-CTL hallmarks in the absence of Eos was also consistent with an increase in the expression of key T_FH transcription factors including Tcf7, which has been implicated in the repression of CD4-CTLs, and Zfp831, an upstream regulator of both Tcf7 and Bcl6 (Fig. 4C) (17, 43). Notably, GSEA of “hallmark” gene sets revealed that IL-2/STAT5 signaling was downregulated in Eos-deficient cells, while Myc and E2F targets, both of which are associated with the opposing T_FH cell program, were upregulated (Fig. 4D–F) (44). Collectively, these data further support a role for Eos in promoting CD4-CTL gene programming through regulation of cytokine signaling pathways, effector molecules, and transcription factors.

Figure 4. Eos deficiency results in broad dysregulation of the cytotoxic gene program.

(A-F) WT and Eos-deficient cells were cultured under T_H1-polarizing conditions for 3 days. RNA sequencing analyses were performed to assess differentially expressed genes. Counts were normalized via DESeq2 analysis. Data are representative of 4 independent experiments and 2 independent sequencing runs. (A) Heatmap of top 200 DEGs between WT and Eos-deficient cells. Changes in gene expression are presented as row (gene) Z-score from normalized counts. Genes are clustered by Euclidian distance. (B) PCA analysis of normalized counts. (C) Representative heatmap of DEGs positively and negatively associated with the CD4-CTL program in WT versus Eos-deficient cells. Differential expression is presented as row (gene) Z-score. (D-F) Genes were pre-ranked using (sign of fold change x −log₁₀(p-value)) and analyzed using the Broad Institute’s Gene Set Enrichment Analysis (GSEA) software for comparison against “hallmark” gene sets. Enrichment plots for the indicated gene sets are shown.

Eos deficiency results in reduced expression of CD25 and CD122

As discussed above, both IL-2 and IL-15 signaling have been implicated in the positive regulation of CD4-CTL programming, largely through activation of STAT5 (3, 16, 21). Our RNA-seq analyses found that Eos-deficient cells exhibited a reduction in Il2ra, which is required for high-affinity IL-2 signaling, and Il2rb, which is required for signaling in response to both IL-2 and IL-15 (Fig. 4C) (45). Furthermore, GSEA revealed that the “hallmark” IL-2/STAT5 signaling gene set was significantly downregulated in Eos-deficient cells (Fig. 4D). To assess whether the surface expression of either receptor subunit was altered in the absence of Eos in vivo, we examined donor CD45.2⁺ WT and Eos-deficient effector cells generated in response to influenza infection utilizing the previously described adoptive transfer strategy before protein analysis with flow cytometry (Supplemental Fig. 1A). Indeed, Eos deficiency resulted in a trending decrease in the surface expression of the alpha subunit of the IL-2 receptor (IL-2Rα or CD25) and a significant reduction in expression of the beta subunit (IL-2Rβ or CD122) (Fig. 5A,B). Thus, these findings support the possibility that Eos may promote CD4-CTL formation by propagating the IL-2/STAT5 and IL-15/STAT5 signaling pathways.

Figure 5. Eos deficiency results in reduced expression of CD25 and CD122.

(A,B) Naïve CD4⁺ T cells were harvested from the spleens and lymph nodes of OT-II-WT or OT-II-Ikzf4^−/− CD45.2 mice. 500,000 cells/animal were adoptively transferred into CD45.1 recipients. Recipient mice were infected with 40 PFU OVA_323–339-expressing A/Puerto Rico/8/34 (“PR8-OVA”) 24 hours post-transfer and lungs were harvested 8 days later for analysis. Viable, antigen-specific CD4⁺CD45.2⁺CD44⁺CD62L⁻ effector cells were analyzed via flow cytometry for expression of CD122 and CD25, as indicated. Data are representative of 4 independent experiments (n = 11-12 ± s.e.m.; **P<0.01; two-tailed, unpaired Student’s t-test).

Eos expression is induced by both IL-2 and IL-15 signaling

Our data thus far suggest that Eos promotes the formation of CD4-CTLs and that Eos expression is regulated either directly or indirectly by the CD4-CTL suppressor and IkZF family member, Aiolos (27). To further investigate the mechanism(s) by which Eos expression is regulated, we examined previously published Assay for Transposase-Accessible Chromatin with sequencing (ATAC-seq) data of WT and Ikzf3^−/− cells cultured under T_H1 conditions (27). Using these data, we sought to identify regulatory regions which display altered chromatin accessibility at the Ikzf4 locus and, thus, may be involved in transcriptional regulation of Eos. Consistent with elevated Eos expression in the absence of Aiolos, we found that the promoter region of the Ikzf4 gene exhibited an increase in accessibility in Aiolos-deficient cells (Fig. 6A). Further, using publicly available Chromatin Immunoprecipitation sequencing (ChIP-seq) data for STAT5 DNA binding (GEO Accession #GSM1865310, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM1865310), we found that this region of increased accessibility at the Ikzf4 promoter corresponds with a known region of STAT5 binding in CD8⁺ T cells (Fig. 6A).

Figure 6. Eos expression is induced by both IL-2 and IL-15 signaling.

(A) Naïve WT and Aiolos-deficient CD4⁺ T cells were cultured under T_H1-polarizing conditions for 3 days. ATAC-seq was performed to assess changes in chromatin accessibility. Data are representative of 3 independent experiments. WT and Aiolos-deficient samples are displayed as CPM-normalized Integrative Genomics Viewer (IGV) tracks and are representative of 3 independent experiments. Data are overlaid with previously published CD8⁺ T cell STAT5 ChIP-seq data set (GSM1865310). The Ikzf4 promoter region with increased accessibility and STAT5 binding is indicated by a gray box. (B) EL4 T cells were transfected with an Ikzf4 promoter-reporter construct in conjunction with STAT5b^CA or empty vector control. Cells were concurrently transfected with SV40 Renilla to serve as a control for transduction efficiency. Luciferase promoter-reporter values were normalized to SV40 Renilla control and presented relative to empty vector. Data are representative of 2 independent experiments (n = 3 ± s.e.m.; *P<0.05; two-tailed, unpaired Student’s t-test). Immunoblot of indicated proteins was performed to validate vector overexpression. β-actin serves as a loading control. Data are representative of 3 independent experiments. (C-F) Naïve CD4⁺ T cells were isolated from the spleens and lymph nodes of WT mice and cultured in vitro under T_H1-polarizing conditions for 2 days. In addition to T_H1-polarizing cytokines, cells were given either 1) nothing additional, 2) IL-2, or 3) IL-2-neutralizing antibodies with IL-15/IL-15Rα complex (IL-15_TRANS). (C,E) Ikzf4 transcript was analyzed via qRT-PCR. Data are representative of 5 independent experiments (n = 8 ± s.e.m.; **P<0.01; two-tailed, paired Student’s t-test). (D,F) Cells were analyzed via flow cytometry for Eos protein expression. Representative flow plots are presented. The same isotype control is used for comparison in both primary flow cytometry plots. Data are representative of 2 independent experiments (n = 5 ± s.e.m.; **P<0.01; two-tailed, paired Student’s t-test).

Previously published work from our group established that overexpression of a constitutively active form of STAT5b (STAT5b^CA) resulted in a significant increase in Ikzf4 promoter activity, which we replicated here (Fig. 6B) (30). As STAT5 is activated downstream of environmental IL-2 and IL-15 signals, both of which have been implicated in promoting CD4-CTLs, we next sought to determine whether signals from IL-2 and IL-15 could induce Eos expression in a type 1 immune response (16, 19). We cultured WT naïve CD4⁺ T cells in vitro under T_H1 conditions, as previously described, with the addition of either 1) IL-2 or 2) IL-15 complexed with IL-15Rα (IL-15_TRANS) (27, 35). For experiments involving IL-15_TRANS, we also included neutralizing antibodies against IL-2 to ensure that IL-15 signaling was the primary activator of STAT5. Indeed, we found that addition of IL-2, or IL-15_TRANS during IL-2 neutralization, resulted in a significant increase in Eos expression at both the gene and protein levels relative to the untreated control (Fig. 6C–F). Indeed, the increase in Eos expression coincided with a decrease in Aiolos expression at both the gene and protein levels, further highlighting potential antagonism and disparate regulation between these two IkZF transcription factors (Supplemental Fig. 3A–D). Collectively, these data highlight a potential feed-forward mechanism wherein Eos promotes CD4-CTL programming, at least in part, via increased IL-2/IL-15/STAT5 signaling, which reciprocally drives Eos expression.

Aiolos antagonizes STAT5-dependent induction of Ikzf4 (Eos) promoter activity

We next wanted to investigate the mechanism(s) by which Aiolos represses Eos expression. Initially, we utilized a retroviral transduction system to overexpress Aiolos in primary T_H1 cells and found that forced Aiolos expression was sufficient to repress Eos at both the transcript and protein levels (Fig. 7A). In previously published work, we found that Aiolos was capable of repressing IL-2/STAT5 signaling, suggesting that Aiolos may antagonize STAT5-dependent induction of Eos expression (27). Indeed, ChIP analysis of the Ikzf4 promoter in WT and Aiolos-deficient T_H1 cells revealed that STAT5 association with the Ikzf4 promoter was enhanced in the absence of Aiolos (Fig. 7B).

Figure 7. Aiolos represses Eos expression by antagonizing STAT5-dependent induction of Ikzf4 promoter activity.

(A) Naïve CD4⁺ T cells were isolated from the spleens and lymph nodes of Aiolos-deficient mice and cultured under T_H1-polarizing conditions for 3 days. Concurrently, cells were retrovirally transduced with either empty vector control or a vector with the gene encoding Aiolos. Ikzf4 gene transcript was assessed via qRT-PCR, normalized to Rps18 control, and presented relative to empty vector control. Data are representative of 3 independent experiments (n = 3 ± s.e.m.; *P<0.05; two-tailed, paired Student’s t-test). Immunoblot of Eos and Aiolos protein analysis is shown. β-actin serves as a loading control. Data are representative of 3 independent experiments. (B) Naïve WT or Aiolos-deficient cells were cultured under T_H1-polarizing conditions for 3 days. ChIP analysis was performed using anti-STAT5 and IgG control antibodies for the Ikzf4 promoter region. Data are normalized to total input sample and IgG values and are displayed relative to the WT sample. Data are compiled from 3 independent experiments (n = 3 ± s.e.m; **P<0.01; two-tailed, unpaired Student’s t-test). (C) Schematic depicting the conserved IkZF domains of Aiolos versus the Aiolos DNA-binding domain mutant (Aiolos^DBM). (D-F) EL4 T cells were transfected with Ikzf4 promoter-reporter construct in conjunction with STAT5b^CA, Aiolos, Aiolos^DBM and/or empty vector control. Cells were concurrently transfected with SV40 Renilla to serve as a control for transduction efficiency. Luciferase promoter-reporter values were normalized to Renilla control and presented relative to control sample. Data are representative of 3-5 independent experiments (n = 3-5 ± s.e.m.; *P<0.05, **P<0.01, ***P<0.001; one-way ANOVA with Tukey’s multiple comparisons). Immunoblot of indicated proteins was performed to validate vector overexpression. β-actin serves as a loading control. Data are representative of 3-5 independent experiments.

In our prior study, we also noted that Aiolos shares a core DNA-binding motif with STAT5 (27). Thus, we investigated the possibility that Aiolos directly represses Eos expression by competing with and antagonizing STAT5 binding at the Ikzf4 locus (27). To do so, we generated overexpression vectors for Aiolos and an Aiolos DNA-binding mutant (Aiolos^DBM) (Fig. 7C). Indeed, we found that Aiolos was able to repress Ikzf4 promoter activity and that this depended on its DNA-binding capacity, as Ikzf4 repression was lost upon Aiolos^DBM overexpression (Fig. 7D). To determine if Aiolos-mediated repression of Ikzf4 promoter activity was due to competition with STAT5, we next overexpressed Aiolos in combination with STAT5b^CA. We found that the addition of Aiolos largely blocked the ability of STAT5b to induce Ikzf4 promoter activity (Fig. 7E). In contrast, we found that overexpression of Aiolos^DBM enhanced STAT5-mediated induction of Ikzf4 promoter activity, demonstrating again that DNA-binding is necessary for Aiolos repression of STAT5b activity (Fig. 7F). As the Aiolos^DBM still contains an intact protein-protein interaction domain and IkZF family members are known to homodimerize at target gene loci, it is possible that this further induction may be attributed to sequestering of endogenous (repressive) Aiolos by Aiolos^DBM. Collectively, these data demonstrate that Aiolos represses Eos expression by disrupting STAT5-dependent activation of the Ikzf4 promoter.

Discussion

The mechanisms that govern the differentiation and functional programs of CD4-CTLs have remained enigmatic despite their established protective and pathogenic roles during distinct immune responses. Here, we identify an oppositional transcription factor network that functions to regulate CD4-CTL differentiation and effector molecule production during influenza virus infection. Our recent work defined the Ikaros family member Aiolos as a repressor of CD4-CTL programming (27). Here, we find that an additional Ikaros factor, Eos, functions in opposition to Aiolos to promote CD4-CTL formation. Mechanistically, Eos supports cytotoxic programming, at least in part, by inducing CD25 and CD122 expression, thus sensitizing differentiating effector CD4⁺ T cells to the known CD4-CTL-driving cytokines IL-2 and IL-15. Further, we find that Aiolos directly represses Eos expression by antagonizing STAT5 association with the Ikzf4 promoter. Collectively, this work identifies a novel, counterbalancing Aiolos/Eos regulatory axis that modulates the formation of effector CD4⁺ T cells with cytotoxic properties.

To our knowledge, this is the first study to examine the role of Eos in the generation of effector CD4⁺ T cell responses to infection, as much prior work has been conducted in the context of allergy and autoimmunity. Eos was first identified as a contributor to T_REG differentiation and stability, largely due to its inclusion in the Foxp3 transcriptional repressive complex (31, 38–40). Indeed, Eos-deficiency leads to immune dysregulation and the spontaneous onset of autoimmunity (32, 38). Further, a prior study from our laboratory demonstrated that Eos supports T_H2 differentiation during allergic responses to house dust mite (HDM) challenge (30). In this work, Eos was found to propagate STAT5 activity by interacting with STAT5 and enhancing its tyrosine phosphorylation via an unknown mechanism (30). Importantly, this suggests that Eos-dependent augmentation of STAT5 activity may represent a conserved regulatory mechanism that functions across STAT5-responsive immune cells and represents a key area for future investigation. Further, our finding that either IL-2 or IL-15 can induce Eos expression highlights the possibility of a feed-forward Eos-mediated mechanism that may broadly propagate STAT5-dependent cytokine signaling pathways.

Our study also defines Aiolos as a direct antagonist of Eos expression, implicating Aiolos-dependent repression of Eos as a regulatory component influencing the differentiation of CD4-CTLs (27). It is also notable that Aiolos antagonizes STAT5 association with the Ikzf4 promoter, which suggests a broader mechanism wherein Aiolos may antagonize STAT5 binding throughout the genome of differentiating CD4⁺ T cells. Thus, our collective work highlights multifactorial roles for Aiolos in restraining STAT5 signaling pathways: 1) repression of CD25 and CD122 expression, 2) antagonism of STAT5 association with target genes, and 3) direct repression of Eos. Indeed, prior work has indicated that Aiolos positively regulates both T follicular helper (T_FH) and T helper 17 (T_H17) cell programs, both of which are repressed by IL-2/STAT5 signaling (27, 29). Prior work has also demonstrated that Eos expression is markedly reduced in T_FH and T_H17 populations (28, 32, 46). These studies, combined with our current work, suggest that counteracting Aiolos/Eos or Aiolos/STAT5 regulatory axes may function as significant contributors to T helper cell programming across effector subsets.

In addition to larger implications for T helper cell programs, the regulatory mechanisms established in this work also have the potential to extend to additional immune cell populations. Indeed, Aiolos has been implicated in the regulation of numerous lymphoid populations, including NK cells, innate lymphoid cells (ILC), and B cells (23). Work showing that Aiolos is necessary for the appropriate maturation and activation of B cells has been expanded to show that Aiolos also supports mature B cell function via antibody generation (47–49). While the roles for Aiolos in NK and ILC populations are less defined, early data is emerging to indicate that Aiolos also regulates these cell populations. Indeed, Aiolos is expressed at relatively high levels in both NK cells and ILC1s (37, 50, 51). In the context of ILCs, Aiolos has been shown to cooperate with T-bet to repress regulatory elements active in ILC3s, thus potentially playing a role in ILC tissue-dependent plasticity (52). Further work has highlighted that Aiolos plays a key part in the regulation of NK cell programs, as its absence resulted in hyper-reactive NK cells with increased proliferation in response to IL-15 (53). The role of Eos in additional immune cell populations outside of T cells is limited, but Eos is expressed at low levels in both NK and ILCs, suggesting that it may function similarly in these cell programs (54). Due to a number of shared regulatory requirements between CD4-CTLs and CD8⁺ cytotoxic T cells, as well as the relatively high levels of expression of Aiolos and Eos in CD8⁺ T cells, it is possible that this Aiolos/Eos regulatory axis may broadly function to regulate cytotoxic populations (32, 54). However, to date, the roles of Aiolos versus Eos in CD8⁺ T cell programs have yet to be thoroughly investigated.

Finally, it is clinically important to note that Aiolos represents a current therapeutic target. Specifically, the immunomodulatory therapeutic lenalidomide is known to degrade Aiolos, along with the IkZF family member Ikaros (55). Lenalidomide is currently in use for the targeting of pathogenic B cell populations in the contexts of multiple myeloma and systemic lupus erythematosus (56–60). Intriguingly, our data suggest that the therapeutic targeting of Aiolos may represent an avenue to promote the cytotoxic activity of CD4⁺ T cells. In line with this, treatment of CAR T cells with lenalidomide has been shown to promote their antitumor activity (57). Beyond T cell populations, treatment with lenalidomide has been shown to increase the cytotoxic activity of NK cells, possibly through similar mechanisms (61). Hence, the identification of the Aiolos/Eos regulatory axis described here not only provides insight into the mechanisms that govern CD4-CTL programming but may also represent a conserved regulatory feature and promising therapeutic avenue that may be exploited across numerous immune cell populations and clinical contexts.

Key Points.

1. Eos positively regulates cytotoxic programming in CD4⁺ T cells

2. Eos promotes IL-2 and IL-15 cytokine receptor expression

3. Aiolos antagonizes STAT5-dependent induction of Eos expression