TSLP shapes the pathogenic responses of memory CD4⁺ T cells in eosinophilic esophagitis

Abstract

The cytokine thymic stromal lymphopoietin (TSLP) mediates type 2 immune responses, and treatments that interfere with TSLP activity are in clinical use for asthma. Here, we investigated whether TSLP contributes to allergic inflammation by directly stimulating human CD4⁺ T cells and whether this process is operational in eosinophilic esophagitis (EoE), a disease linked to variants in TSLP. We showed that about 10% of esophageal-derived memory CD4⁺ T cells from individuals with EoE and less than 3% of cells from control individuals expressed the receptor for TSLP and directly responded to TSLP, as determined by measuring the phosphorylation of STAT5, a transcription factor activated downstream of TSLP stimulation. Accordingly, increased numbers of TSLP-responsive memory CD4⁺ T cells were present in the circulation of individuals with EoE. TSLP increased the proliferation of CD4⁺ T cells, enhanced type 2 cytokine production, induced the increased abundance of its own receptor, and modified the expression of 212 genes. The epigenetic response to TSLP was associated with an enrichment in BATF and IRF4 chromatin-binding sites, and these transcription factors were induced by TSLP, providing a feed-forward loop. The numbers of circulating and esophageal CD4⁺ T cells responsive to TSLP correlated with the numbers of esophageal eosinophils, supporting a potential functional role for TSLP in driving the pathogenesis of EoE and providing the basis for a blood-based diagnostic test based on the extent of TSLP-induced STAT5 phosphorylation in circulating CD4⁺ T cells. These findings highlight the potential therapeutic value of TSLP inhibitors for the treatment of EoE.

INTRODUCTION

Allergic diseases are characterized by an impaired epithelial barrier and augmented type 2 immune responses that involve multiple cell types in a highly orchestrated fashion (1, 2). After initial allergen-driven priming, naïve CD4⁺ T cells differentiate into T helper type 2 (T_H2) cells and serve a pivotal role in maintaining allergic responses by secreting type 2 cytokines (3-5). These T_H2 cells survive as memory cells and exhibit a rapid and robust response to low doses of antigen after reexposure. Pathogenic effector T_H2 (peT_H2) cells, a subset of T_H2 cells that produces large amounts of T_H2 cytokines upon T cell receptor (TCR) activation, have an essential role in allergic disorders (6-11) and are substantially enriched in the esophageal tissue of patients with the eosinophilic gastrointestinal disorder (EGID) eosinophilic esophagitis (EoE) (12, 13). Upon breach of the mucosal barrier, thymic stromal lymphopoietin (TSLP) is one of the early cytokines that is expressed by the activated epithelium (1, 14-18) and causally contributes to the pathogenesis of asthma and other allergic diseases, such as atopic dermatitis (AD), allergic rhinitis, food allergies, and EoE (19-22). TSLP amounts are increased in the skin cells of patients with AD, the esophageal epithelia of patients with EoE (23), and the airway epithelia of patients with asthma and strongly associate with disease severity (24-28). At the genetic level, the human TSLP locus 5q22 represents a notable risk allele for allergic diseases, as determined by multiple, independent genome-wide association studies (29-33). In addition, children carrying genetic variants for TSLP and IL4 have a synergistically increased risk for EoE compared with those with a risk allele in only one of these genes (34). As indicated from transcriptome data, EoE endotype 2 was characterized as an inflammatory subtype with high production of TSLP, representing the most inflammatory type 2–associated subgroup of EoE (35). Although several studies have focused on assessing TSLP abundance in the context of allergic diseases (23-27, 35-37), the direct responsiveness of human T cells to TSLP in the context of allergic disease pathogenesis remains underinvestigated. Understanding the TSLP-allergy axis has the potential to better define patient subsets that may have variable responses to agents that directly or indirectly block TSLP [for example, tezepelumab or Janus kinase (JAK) inhibitors, respectively] (38, 39).

TSLP is a pleiotropic cytokine that affects both innate and adaptive immune cells. The biological functions of TSLP require heterodimer formation between the TSLP receptor (TSLPR) and interleukin-7 receptor α (IL-7Rα) (40). TSLP polarizes dendritic cells to induce type 2 inflammation, maintains the homeostasis of T cells, and directly expands or activates group 2 innate lymphoid cells, mast cells, basophils, and other immune cells (17, 18, 21, 41-43). Although the role of TSLP in the differentiation and maintenance of T_H2 cells in mice is well documented (44-47), its role in human T cells remains less clear. In human and mouse T cells, TSLP signals through activation of the JAK1-JAK2–signal transducer and activator of transcription 5 (STAT5) pathway (48). Murine TSLPR is constitutively expressed in T helper (T_H) cells, with the highest abundance on T_H2 cells (49), whereas human CD4⁺ T cells from the blood of normal donors requires strong TCR-mediated activation to induce detectable TSLPR expression in vitro (50).

There are reports of TSLPR expression on a small population of human CD4⁺ T cells from patients with allergic disorders, such as EGID and AD (51, 52), as well as colorectal cancer (53), but not from normal donors, suggesting a disease-associated increase in the abundance of TSLPR on human T cells. Moreover, CRLF2 mRNA (which encodes TSLPR) is expressed in a unique T_H2A (likely peT_H2) cell population from the blood of atopic individuals; these T_H2A cells produce increased amounts of T_H2 cytokines compared with those produced by conventional T_H2 cells (9). Despite evidence that TSLPR is expressed on a small population of human CD4⁺ T cells, the functional importance of TSLPR in CD4⁺ T cells and the molecular mechanisms underlying TSLP-receptor interactions in human T cells are unclear. Because anti-TSLP is in clinical usage for severe asthma (38), gaining a more comprehensive understanding of the activity and mechanism of action of TSLP holds the promise of refining its clinical applications, including potential benefits for diseases other than asthma.

Here, we focused on key outstanding questions in the field: How does TSLP shape the response of human CD4⁺ T cells, and what molecular and cellular changes does it induce in these T_H cells? These are more than academic questions because TSLP having direct effects on pathogenic T cells would imply that anti-TSLP–based therapeutics may have consequences on both the development and suppression of established type 2 memory responses. Accordingly, to investigate this in the context of human disease, we examined autologous blood and tissue CD4⁺ T cells from patients with EoE, including individuals with active and inactive inflammation (remission), and from normal controls and compared their responses to TSLP stimulation ex vivo. We found that a subpopulation of human memory, but not naïve, CD4⁺ T cells, responded to TSLP and that patients with EoE had increased numbers of these human memory cells in the circulation and allergic tissue. In addition, the frequency of these cells correlated with the count of esophageal eosinophils. TSLP induced transcriptional and epigenetic changes at target loci in CD4⁺ T cells, regulating T cell activation, proliferation, differentiation, and cytokine production. TSLP markedly increased the amounts of BATF and IRF4 mRNAs together with an increase in the abundance of IRF4. These effects likely led to subsequent chromatin changes, as evidenced by enriched motifs for their binding sites in TSLP-stimulated memory CD4⁺ T cells. Together, our results show that TSLP promotes proliferation and robust T_H2 cytokine production in a subset of human memory CD4⁺ T cells, a process that likely contributes to EoE, and that the presence of an increased percentage of TSLP-responsive CD4⁺ T cells in the blood may serve as a diagnostic tool for EoE.

RESULTS

A subset of human CD4⁺ T cells expresses TSLPR in EoE biopsies

We aimed to determine the cells that expressed the TSLPR at sites of allergic inflammation. Accordingly, we probed single-cell RNA sequencing (scRNA-seq) data from biopsies of patients with EoE (12) and found that CRLF2 mRNA (which encodes TSLPR) was mainly detected in CD4⁺ T cell and mast cell subpopulations (Fig. 1A). Through bulk RNA-seq analysis of tissue CD3⁺CD45⁺ fluorescence-activated cell sorting (FACS)–sorted T cells, we found that CRLF2 mRNA was increased in abundance in active and some remission EoE esophageal tissue samples compared with that in control individuals (fig. S1A) and was associated with expression of genes encoding T_H2 cytokines (IL4, IL5, and IL13).

Fig. 1. TSLPR abundance is increased on human CD4⁺ T cells from EoE biopsies.

(A) scRNA-seq analysis was performed with published data (12). UMAP projection of cells obtained from the esophageal biopsies of six patients with EoE, colored by cell population (left), and with CRLF2 expression mapped onto the corresponding UMAP projection (right). (B and C) Cell surface TSLPR on CD4⁺CD3⁺ and CD8⁺CD3⁺ gated cells was analyzed by flow cytometry. (B) Representative plots of T cells from biopsies of control individuals and patients with active EoE. (C) Summary of flow cytometry staining for control individuals (n = 14; eosinophils/HPF = 0) and patients with active EoE (n = 28; eosinophils/HPF > 15). Data are presented as a box and whiskers plot. *P ≤ 0.05 and ****P ≤ 0.0001 by nonparametric paired or unpaired two-tailed t tests. (D and E) Cells from biopsies were left unstimulated [medium (Med.)] or were stimulated with TSLP (40 ng/ml) or IL-2 (100 U/ml) for 20 min and subjected to flow cytometry analysis of pSTAT5. (D) Live cells from biopsies were gated on the CD3⁺ population and then gated on CD4⁺ or CD8⁺ cells. Representative plots of pSTAT5 and CD45RO staining are shown. (E) Summary of pSTAT5 flow cytometry analysis of CD4⁺ and CD8⁺ cells from eight esophageal biopsies of active EoE. Data are presented in a box and whiskers plot. *P ≤ 0.05, **P ≤ 0.01, and ****P ≤ 0.0001 by nonparametric, paired one-way ANOVA; ns, not significant.

To substantiate the transcriptional findings, we compared the abundance of TSLPR protein on the surfaces of CD4⁺CD3⁺ and CD8⁺CD3⁺ T cells isolated from inflamed and noninflamed tissue samples. CD4⁺ T cells in the tissue primarily exhibited a memory phenotype characterized by CD45RO surface expression (Fig. 1B). Consistent with the mRNA data (Fig. 1A and fig. S1A) and a central pathogenic role of tissue residential memory T_H2 cells in EoE (12, 13), markedly increased surface TSLPR was detected on CD45RO⁺CD4⁺ cells from the esophageal tissue of patients with active EoE compared with that of control individuals, whereas TSLPR was barely detectable on CD8⁺ counterparts regardless of EoE disease status (Fig. 1, B and C). Short stimulation of the single-cell suspension from the whole biopsy with TSLP induced STAT5 phosphorylation in CD4⁺ but not CD8⁺ T cells (Fig. 1, D and E), demonstrating functionality of the TSLPR complex and responsiveness of a subset of tissue-resident CD4⁺ T cells to TSLP.

TSLPR⁺CD4⁺ T cells in human allergic tissue have T_H2 cell properties

To investigate whether TSLPR expression in tissue-resident T cells was associated with type 2 responses in a disease context, we performed intracellular flow cytometry staining to detect major T cell cytokines and key transcriptional factors in populations of TSLPR⁺ and TSLPR⁻ CD4⁺CD3⁺ cells from single-cell suspensions of esophageal biopsies from patients with EoE (Fig. 2A). CD8⁺ T cells were used as an internal negative control for TSLPR, GATA3, FOXP3, and cytokine staining (fig. S1, B and C). Paired analysis of TSLPR⁺ and TSLPR⁻ CD4⁺CD3⁺ cells obtained from the same donors demonstrated that TSLPR⁺ CD4⁺CD3⁺ cells exhibited a notable enrichment in the production of IL-4, IL-13, or both. Similarly, the key T_H2 regulator GATA3 exhibited a substantial increase in abundance within the TSLPR⁺ compared with that in the TSLPR⁻ population (Fig. 2B). In contrast, IL-10, interferon-γ (IFN-γ), and FOXP3 were mainly produced by TSLPR⁻ CD4⁺CD3⁺ cells (Fig. 2B). At the single-cell mRNA level, there was coexpression of CRLF2 with IL5, IL13, and GATA3, but not IFNG (Fig. 2C). Moreover, transcripts that define the phenotype of T_H2A cells, including PTGDR2 (encoding CRTH2), KLRB1 (CD161), and ITGA4 (CD49d), but excluding CD27 (9), exhibited coexpression with CRLF2 (Fig. 2D). Together, these results indicate that TSLPR is predominantly expressed in cells characterizing a T_H2A phenotype and that these cells represent a source of type 2 cytokines.

Fig. 2. TSLPR⁺CD4⁺ T cells from inflamed tissue exhibit a T_H2 cell phenotype.

(A and B) Cells from human esophageal biopsies obtained from patients with active EoE were activated for 5.5 hours with PDBu and ionomycin; brefeldin A and monensin were added for the last 4 hours. The cells were then subjected to intracellular staining for the indicated cytokines and transcription factors. To characterize TSLPR⁺ and TSLPR⁻ populations of CD4⁺ T cells, the lymphocyte population was first gated on CD3⁺CD4⁺ cells and then on subpopulations of TSLPR-positive and TSLPR-negative cells. (A) Data are from one experiment, representative of seven. (B) Percentages of TSLPR⁻ (negative) and TSLPR⁺ (positive) cells expressing the individual factors from seven independent experiments. Data were analyzed by paired, two-tailed t test. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. (C and D) Feature plot analyses of (C) CRLF2, IL5, IL13, GATA3, and IFNG and (D) PTGDR2, KLRB1, ITGA4, and CD27 of the six biopsies obtained from patients with active EoE (>15 eosinophils/HPF) (12).

TSLP potentiates cytokine production by circulating CD4⁺ T_H2 cells

To determine whether CD4⁺ T cells from the blood of patients with EoE had an increased response to TSLP and whether TSLP could potentiate the T_H2 responses induced by IL-4, as previously shown in murine systems (44), we purified CD4⁺ T cells from the blood of EoE patients (active and in remission) and control individuals (Fig. 3A, purity >98%) and activated them with anti-CD3 and anti-CD28 antibodies in the presence of TSLP, IL-4, or both. TSLP alone and in combination with IL-4 augmented the expression of IL4, IL5, and IL13 mRNA in cells from patients with EoE, but not in those from control individuals (Fig. 3B). Note that the effect of TSLP in cells from EoE patients was independent of the disease status because cells from patients with active and inactive (remission) esophagitis both exhibited increased T_H2 cytokine expression compared with control. The secretion of IL-5 protein by total CD4⁺ T cells was also increased in cells derived from EoE patients but not in those from control individuals (Fig. 3C). Furthermore, the effect of TSLP was more pronounced on the induction of IL5 and IL13 and less pronounced on the expression of IL4, suggesting that TSLP has a late-phase role in T_H2 cell differentiation (6). These results demonstrated that TSLP acted alone and in concert with IL-4 to induce pathogenic T_H2 responses directly in human CD4⁺ T cells purified from the blood, a phenomenon that was more substantial in cells from patients with EoE (active and remission) than in cells from control individuals.

Fig. 3. TSLP and IL-4 act in concert to induce the production of T_H2 cytokines by human CD4⁺ T cells.

(A) Representative plots of cells before and after human CD4⁺ T cell purification. (B and C) Purified human CD4⁺CD3⁺ cells from control individuals (Control), patients with EoE in remission (Remission), or patients with active EoE (Active) were treated with anti-CD3 (5 μg/ml), anti-CD28 (2 μg/ml), and anti–IFN-γ (5 μg/ml) alone as a medium control (Med.) or together with IL-4 (40 ng/ml) and TSLP (40 ng/ml) alone or in combination. (B) The relative abundances of the indicated mRNAs were measured on day 3 (Control: n = 6; Remission: n = 18; Active: n = 16). Data were normalized to the expression of the housekeeping gene EIF3K. (C) The amounts of IL-5 secreted by the indicated cells were assessed on day 3 (Normal: n = 7; Remission: n = 9; Active: n = 10). Data in (B) and (C) are presented in box and whiskers plots and were analyzed by nonparametric one-way ANOVA for each condition. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.

TSLP induces STAT5 phosphorylation in circulating memory CD4⁺ T cells

We aimed to better understand TSLP-induced responses in circulating CD4⁺ T cells. Given that TSLPR is only weakly detectable on circulating CD4⁺ T cells (fig. S1A) (50, 53, 54), we examined signaling downstream of TSLPR with phosphorylated STAT5 (pSTAT5) as a readout of STAT5 activation (48). Human peripheral blood mononuclear cells (PBMCs) were acutely stimulated with TSLP (40 ng/ml) and analyzed for pSTAT5 by flow cytometry. As controls, PBMCs cultured in medium alone did not demonstrate STAT5 activation in CD3⁺ T cells, whereas stimulation with IL-2 induced robust STAT5 phosphorylation in these cells (fig. S2A). With this surrogate approach, we found that CD3⁺ cells were major responders to TSLP in human blood (fig. S2A). By gating CD3⁺ cells for CD4⁺ and CD8⁺ cells, we detected a population of CD4⁺CD3⁺ cells that responded to TSLP; in contrast, human blood CD8⁺ T cells showed a reduced response to TSLP despite having a normal response to IL-2 (fig. S2, A and B).

Because the proportions of memory and naïve T cells varied between donors, we compared the response of naïve (CD45RO⁻) versus memory (CD45RO⁺) CD4⁺ T cells to TSLP from the blood of control individuals or patients with EoE, aiming to detect differences driven by disease status. To enrich the CD4⁺ T cell population and avoid the potentially confounding effect of TSLP on other blood cell types, CD4⁺ T cells were purified from PBMCs and then activated for 20 min in the presence of TSLP or the positive control IL-2. A noticeable percentage of memory (CD45RO⁺) CD4⁺ T cells in the blood, compared with naïve (CD45RO⁻) CD4⁺ T cells (<1%), responded to TSLP by activation of STAT5 (Fig. 4A). The percentage of cells exhibiting TSLP-induced pSTAT5 was increased in CD45RO⁺CD4⁺ T cells from patients with EoE compared with that in cells from control individuals (Fig. 4A). To substantiate these findings, cells from an independent, relatively large cohort of patients were similarly stimulated; a substantial difference in the percentage of TSLP-responsive memory T_H cells was observed in cells isolated from patients with active EoE (n = 52) compared with patients with EoE in remission (n = 45) and control individuals (n = 18) (Fig. 4B). Cells from patients in remission retained a greater amount of TSLP-induced pSTAT5 cells than did cells from control individuals. Note that the percentage of CD4⁺ T cells with TSLP-induced pSTAT5 correlated with tissue eosinophil numbers (Fig. 4C) and with the percentage of TSLPR⁺CD4⁺ memory cells in autologous esophageal tissue (Fig. 4D). The detection of TSLP responsiveness in human blood CD4⁺ T cells exhibited high sensitivity and specificity in distinguishing between control samples and active EoE or remission EoE samples (fig. S2C). As an internal control for STAT5 activation, IL-2–induced STAT5 activation in both memory and naïve cells did not show any disease context (Fig. 4B and fig. S2D).

Fig. 4. A subset of human memory CD4⁺ T cells from the blood responds to TSLP by activating the STAT5 signaling pathway.

(A and B) CD4⁺ T cells were enriched from human PBMCs and stimulated with the indicated cytokines for 20 min. Cells were gated on CD3⁺CD4⁺ cells and assessed for CD45RO and pSTAT5 content by flow cytometry. (A) Data are from representative experiments of control individuals (top) and patients with active EoE (bottom). (B) Data represent a summary of the percentage of pSTAT5⁺ memory T cells gated on CD3⁺ CD4⁺CD45RO⁺ and obtained after stimulation with TSLP (left) or IL-2 (right) for 20 min. Samples were collected from control individuals (Control: n = 18) and patients with EoE in remission (Remission: n = 45) or with active EoE (Active: n = 52). Data are means ± SEM, with each circle representing an individual sample. **P ≤ 0.01 and ****P ≤ 0.0001 by one-way ANOVA. (C) Correlation between the percentage of pSTAT5⁺ cells among purified CD45RO⁺CD4⁺ T cells stimulated with TSLP for 20 min and autologous tissue eosinophilia [eosinophils per high-power field (Eos/HPF), n = 112, with Spearman R shown]; each circle represents an individual sample. (D) Correlation of the percentage of TSLP-induced pSTAT5⁺ cells among blood memory CD4⁺ T cells and the percentage of TSLPR⁺CD4⁺ T cells in autologous esophageal tissue. Each circle represents an individual donor (n = 32). Data are shown with Spearman R (R_S) and its P value. (E) CRLF2 mRNA abundance in purified naïve and memory CD4⁺ T cells from the blood of control individuals (Control: n = 6) and patients with EoE (active or in remission: n = 8). Data are means ± SEM, with each circle or square representing an individual sample. *P ≤ 0.05, ***P ≤ 0.001, and ****P ≤ 0.0001 by two-way ANOVA. (F) Correlation between the percentage of TSLP-induced pSTAT5⁺ cells among blood memory CD4⁺ T cells and the abundance of CRLF2 mRNA in purified blood memory CD4⁺ T cells from all donors, autologously. Each circle represents an individual donor (n = 20). Data are shown with Spearman R (R_S) and its P value.

We subsequently examined whether the variation in STAT5 activation was associated with the abundance of TSLPR, which is encoded by CRLF2. By analyzing CRLF2 mRNA abundance in naïve versus memory CD4⁺ T cells from human blood, we observed that naïve T cells had low amounts of CRLF2 mRNA, whereas memory T cells had increased amounts of CRLF2 mRNA, and that CRLF2 mRNA abundance was associated with EoE disease status (Fig. 4E). Moreover, in memory CD4⁺ T cells from the blood, CRLF2 mRNA abundance significantly correlated with TSLP-mediated STAT5 activation (pSTAT5 abundance), substantiating the blood-based TSLP-mediated STAT5 activation findings (Fig. 4F). Together, these results demonstrated that a subset of memory CD4⁺ T cells in the blood directly respond to TSLP and that patients with EoE have an increased number of circulating CD4⁺ T cells that respond to TSLP, supporting TSLP-induced STAT5 activation as a potential diagnostic and disease activity biomarker.

TSLP modifies gene transcription in activated human memory CD4⁺ T cells

To study the mechanism by which TSLP activates human memory CD4⁺ T cells, we performed RNA-seq analysis of purified circulating CD45RO⁺ CD4⁺ T cells (purity >98%) (fig. S3A) that were activated by anti-CD3 and anti-CD28 antibodies in the presence or absence of TSLP. TSLP increased the expression of 204 genes and decreased the expression of eight genes (table S1). Functional enrichment analyses showed the strongest enrichment in pathways associated with cell cycle and cell division (Fig. 5A). Accordingly, carboxyfluorescein succinimidyl ester (CFSE) staining demonstrated increased T cell proliferation in response to TSLP with anti-CD3 and anti-CD28 activation, but not the effect of TSLP alone (Fig. 5B). There was also enrichment in biological processes mediated by the IL-7 signaling pathway (Fig. 5A), consistent with TSLPR forming a heterodimer with IL-7Rα and both TSLP and IL-7 activating the STAT5 pathway (40). TSLP induced the expression of genes involved in leukocyte differentiation, including transcription factors, cell surface receptors, and cytokines (Fig. 5C). These findings were largely validated in an additional cohort of patients with EoE (n = 10) (fig. S3B), and TSLP-induced production of CD25 and IRF4 was confirmed at the protein level (fig. S3C, n = 10).

Fig. 5. RNA-seq and ATAC-seq analyses of TSLP-stimulated human CD4⁺ T cells.

(A to H) Memory CD3⁺CD4⁺CD45RO⁺ T cells were sorted from the PBMCs of patients with EoE whose cells were highly responsive to TSLP. Cells were then activated with anti-CD3 and anti-CD28 antibodies in the presence of anti–IFN-γ antibody with or without TSLP for 3 days (A and C to H) or 7 days (B). (A) Biological processes for the RNA-seq analysis of five pairs of independent samples (TSLP-treated versus untreated) with a twofold change threshold (presence versus absence of TSLP), P adjustment ≤ 0.1, and gene expression levels ≥ 1 RPKM in at least one sample were identified with ToppGene Suite with Bonferroni statistical P value. (B) Purified memory CD4⁺ T cells were labeled with CFSE and cultured as indicated. Left: Data are from a representative experiment. Right: Summary of cell divisions for the indicated conditions in individual samples (n = 6). Data are means ± SEM. *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001 by two-way ANOVA. Asterisks represent P values between anti-CD3/28 and anti-CD3/28 + TSLP samples at different cycles. (C) Box and whiskers plot of the log₂(fold change) of T_H2-related genes identified by RNA-seq analysis of five pairs of human blood memory CD4⁺ T cell samples that were untreated or were treated with TSLP. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001; P adjustment values were calculated by DESeq2. (D) The diagram of RNA-seq genes that were epigenetically changed upon TSLP stimulation in accordance with the ATAC-seq analysis of six pairs of independent samples (TSLP-treated versus untreated) with P ≤ 0.05 for the identified peaks: 28 genes gained an accessible chromatin state (UP), one gene changed from an accessible to a closed chromatin state (DOWN), and 183 genes did not show significant changes in chromatin accessibility. (E) Volcano plot showing changes in accessibility at ATAC-seq peaks in TSLP-treated versus untreated samples. Peaks near differentially expressed genes are indicated [50 kb in both directions from the transcription start site (TSS)]. (F) Representative plots of ATAC-seq analysis. Black arrows show significant peaks observed in the DiffBind analysis of six pairs of samples: *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. (G) Biological processes involving genes nearest to the significantly different ATAC peaks were identified using ToppGene Suite with Bonferroni statistical P value. (H) Homer motif enrichment analysis of DNA sequences in TSLP-treated samples in peaks with P ≤ 0.05. The heatmap shows the percentage of target (fold change ≥0.4) and background (fold change <0.4) sequences with motifs. The line shows log₁₀ of the P value for motif enrichment. Asterisks represent transcription factors (TFs) significantly increased in the EoE transcriptome according to the EGID Express database: https://egidexpress.research.cchmc.org/data/.

The epigenetic landscape of memory CD4⁺ T cells is shaped by TSLP

Given that the TSLP-regulated transcriptome was enriched in genes encoding factors involved in chromatin organization (Fig. 5A), we hypothesized that the T_H2-polarized responsiveness of human memory CD4⁺ T cells to TSLP resulted from chromatin remodeling. Thus, we examined the chromatin landscape by assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq) analysis (55). Overall, 798 gene loci were modified by TSLP, with 673 sites opened and 125 closed. Within these loci, 29 genes were identified by RNA-seq analysis (Fig. 5, D and E, and table S2). Twenty-eight of these genes gained an accessible chromatin state and were enriched in genes involved in type 2 responses, including BATF, IRF4, NFIL3, IL13, CSF2, IL3, IL2RA, TNFRSF8, and IL13/RAD50 regulatory regions within the T_H2 cytokine locus (Fig. 5, E and F). Functional enrichment analysis of genes identified by ATAC-seq revealed enrichment for T cell activation, proliferation, and cytokine production processes (Fig. 5G), which are analogous to the cell cycle and cell division processes that were observed in the RNA-seq analysis (Fig. 5A) and corroborated the TSLP-mediated increase in the abundance of cytokine-encoding mRNAs (Fig. 5C).

Motif enrichment analysis of ATAC-seq peaks identified the likely involvement of BATF, IRF4, and STAT6, key transcriptional factors in T_H2 cell differentiation (Fig. 5H) (3, 56, 57), as well as ATF3, FRA1/2, JUN-AP1, BACH2, and HIF-1β, which have roles in T cell activation and chromatin remodeling (58). The production of both IRF4 and BATF was induced by TSLP (Fig. 5C and fig. S3C), and their genes had an open chromatin structure (Fig. 5E), providing evidence for a feed-forward loop. Moreover, the genes encoding several of these transcription factors, including ATF3, BATF, IRF4, HIF-1β, and STAT6, were significantly increased in expression in the EoE transcriptome (fig. S4) (59).

TSLP stimulates type 2 heterogeneous responses in memory CD4⁺ T cells

RNA-seq and ATAC-seq data indicated a direct enhancing action of TSLP on the type 2 cytokine locus (Fig. 5, C and F). To substantiate the role of TSLP in type 2 cytokine production in the context of allergic disease, we analyzed memory CD4⁺ T cells from a biologically independent cohort consisting of control individuals and patients with EoE (active and in remission) and activated them with anti-CD3 and anti-CD28 antibodies in the presence or absence of TSLP. T cells from patients with EoE had varying responsiveness to TSLP, as was determined by measuring TSLP-induced STAT5 phosphorylation (Fig. 4B). Therefore, the EoE samples were subcategorized by TSLP-induced pSTAT5⁺ cells as TSLP high-responsive EoE (≥4% pSTAT5⁺CD45RO⁺ cells) and TSLP low-responsive EoE (<4% pSTAT5⁺CD45RO⁺ cells). At the mRNA level, there were marked increases in IL4, IL5, and IL13 mRNA abundance induced by TSLP in T cells from TSLP high-responsive EoE individuals (Fig. 6A), whereas cells from control individuals and TSLP low-responsive EoE patients showed only a modest increase in the expression of genes encoding T_H2 cytokines in the presence of TSLP (Fig. 6A). TSLP responsiveness, corresponding with T_H2 cytokine production, was substantiated by flow cytometry analysis (Fig. 6, B and C). During T cell activation by anti-CD3 and anti-CD28 antibodies in the presence of TSLP, more memory CD4⁺ T cells from TSLP high-responsive EoE individuals were positive for IL-4, IL-5, and IL-13 compared with memory CD4⁺ T cells from control individuals and TSLP low-responsive EoE patients. In contrast, IFN-γ production by these cells was unaffected by EoE status (Fig. 6D). Furthermore, TSLP not only induced T_H2 cytokine production but also promoted the proliferation of T_H2 cytokine–producing cells (fig. S5A). As an internal control, naïve CD4⁺ T cells from either control or EoE donors were incapable of producing T_H2 cytokines in response to TSLP (fig. S5B). The production of all T_H2 cytokines in response to TSLP on day 7 strongly correlated with STAT5 activation (pSTAT5 abundance) induced by acute TSLP exposure (20 min) at the initiation of cell activation (day 0) (fig. S5C), suggesting a likely causal role of TSLP signaling in EoE pathogenesis.

Fig. 6. A subset of human memory CD4⁺ T cells exhibits increased production of T_H2 cytokines upon exposure to TSLP in the context of allergic inflammation.

(A) Purified memory CD4⁺ T cells from the control participants (Control: n = 8), EoE high-responsive to TSLP (EoE high-resp.: n = 14), and EoE low-responsive to TSLP (EoE low-resp.: n = 6) were left nonactivated or were activated with anti-CD3 and anti-CD28 antibodies in the presence of anti–IFN-γ antibody with or without TSLP. The relative abundances of the indicated mRNAs were measured on day 3. Data are presented in a box and whiskers plot. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001 by nonparametric one-way ANOVA. (B to D) Memory CD4⁺ T cells activated in the absence (medium) or presence of TSLP were collected after 7 days, stimulated with PDBu and ionomycin in the presence of brefeldin A and monensin for 4 hours, and analyzed by flow cytometry. Representative plots of cells are from control individuals (Control, B) and patients with EoE whose cells have a high response to TSLP (high-responsive, C). (D) Data represent a summary of the percentages of the indicated cytokine-producing cells from control individuals (Control: n = 6) and patients with EoE whose cells are highly responsive to TSLP (high: n = 21) or less responsive to TSLP (low: n = 5). Data are means ± SEM. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001 by para-metric one-way ANOVA. (E and F) Culture medium was collected from TSLP high-responsive memory CD4⁺ T cells isolated from patients with active EoE (n = 4) and activated for 3 days with anti-CD3 and anti-CD28 antibodies in the presence (TSLP) or absence (medium) of TSLP. The amounts of the indicated secreted proteins were determined by cytokine multiplex. (E) Heatmap represents a log₁₀ conversion of concentrations (pg/ml) of the indicated proteins secreted upon 3 days of activation. (F) Individual plots for each of the indicated secreted proteins. *P ≤ 0.05 and **P ≤ 0.01 by paired, parametric, two-tailed t test.

We next expanded the analysis of cytokine production by TSLP high-responsive cells beyond classical type 1 and type 2 cytokines (Fig. 6, E and F, and table S3). TSLP markedly increased the secretion of granulocyte-macrophage colony-stimulating factor (GM-CSF) and IL-3, which were also increased at the mRNA level (Fig. 5C), and acquired chromatin changes in the IL3/CSF2 locus (Fig. 5F). Furthermore, significant increases in the amounts of secreted IL-9 and IL-10 were also detected in cultures of CD4⁺ T memory cells activated in the presence of TSLP (Fig. 6F), whereas IFN-γ, IL-17A, and IL-6 were barely detected in the presence or absence of TSLP (Fig. 6E).

Given that anti-CD3 and IL-4 signals strongly enhance TSLPR expression on human CD4⁺ T cells (50, 51) and that TSLP induced an increase in the abundance of IL-4 in human T cell cultures (Fig. 6, E and F), we aimed to determine whether TSLP regulated the expression of its own receptor. TSLP exposure during activation with anti-CD3 and anti-CD28 antibodies increased the abundance of CRLF2 mRNA (Fig. 7A) and percentage of TSLPR⁺ cells (Fig. 7B), suggesting the presence of a positive feedback mechanism to enhance and maintain this cell population. Moreover, IRF4 abundance was notably increased in CD4⁺T cells after treatment with anti-CD3 and anti-CD28 antibodies in the presence of TSLP, which is in agreement with mRNA-seq data (Fig. 5C). Increased cell surface TSLPR and intracellular IRF4 abundances were not observed in samples with a low response to TSLP (fig. S5D). Together, these results demonstrate the functional and pathogenic activities of TSLP signaling in human memory CD4⁺ T cells, resulting in the increased abundance of its own receptor, as well as increases in a broader panel of type 2 cytokines and transcription factors as major downstream targets of TSLP in human memory CD4⁺ T cells.

Fig. 7. TSLP increases the abundance of TSLPR.

(A) Memory CD3⁺CD4⁺ CD45RO⁺ T cells obtained from patients with EoE whose cells were highly responsive to TSLP (n = 7) were not activated (fresh isolated) or were activated for 3 days with anti-CD3 and anti-CD28 antibodies in the presence or absence of TSLP. The cells were subjected to quantitative PCR analysis of CRLF2 mRNA abundance. The abundance of CRLF2 mRNA was normalized to that of the housekeeping gene EIF3K. *P ≤ 0.05 and ***P ≤ 0.001 by paired, nonparametric, one-way ANOVA. (B) Memory CD4⁺ T cells were left untreated or were activated with anti-CD3 and anti-CD28 antibodies in the presence or absence of TSLP for 7 days before being analyzed by flow cytometry to detect TSLPR and IRF4. Data are from one representative experiment of cells from a patient with EoE with cells that were highly responsive to TSLP. Right: Summary of the percentage of memory CD4⁺ T cells co-expressing TSLPR and IRF4 (n = 8). *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001 by paired, nonparametric, one-way ANOVA.

DISCUSSION

Here, we analyzed the response of human CD4⁺ T cells to TSLP in the context of EoE and uncovered several key findings, including (i) the existence of a subset of human memory TSLPR⁺CD4⁺ cells in the blood and esophagi of patients with EoE; (ii) that stimulation of these cells with TSLP induced T_H2 cytokine production and cellular proliferation; (iii) that TSLP had additive effects with IL-4 in terms of inducing T_H2 cytokine production; (iv) that TSLP induced the expression of its own receptor, providing evidence for a feed-forward loop; (v) that TSLP induced open chromatin at sites across the genome, particularly enriched at sites associated with genes involved in cellular division, differentiation, and cytokine production; and (vi) that TSLP increased the expression of several transcriptional factors, including BATF and IRF4, whose DNA binding motifs were significantly enriched in genes regulated by TSLP, supporting an epigenetic and transcriptional mechanism for the action of TSLP in human CD4⁺ T cells.

TSLPR protein detection on human CD4⁺ T cells has been reported in patients with atopic conditions, colorectal cancer (9, 51-53), and after T cell antigen receptor (TCR) activation of purified CD4⁺ T cells (50, 51). Furthermore, TSLPR expression has been detected in T_H2A cells (9), which have a crucial role in allergic diseases, including AD, food allergy, asthma, and EoE (60). Although we could detect abundant TSLPR mRNA and high amounts of TSLPR protein in the subset of tissue-resident CD4⁺ T cells from patients with EoE (Fig. 1 and fig. S1A) and these cells expressed markers of T_H2A cells (Fig. 2D), we were unable to identify this protein on human CD4⁺ T cells from the blood regardless of EoE status, despite cells responding to TSLP by phosphorylating STAT5. The surrogate measure of TSLP-induced STAT5 activation correlated with CRLF2 mRNA abundance in circulating memory CD4⁺ T cells (Fig. 4F), substantiating the accuracy of the blood-based, TSLP-induced pSTAT5 assay. The finding that TSLP increased the abundance of its own receptor in vitro (Fig. 7) suggests that this positive feedback mechanism might be operational in the tissues of patients with EoE, which overexpress TSLP (59, 61), and that it may at least partially explain the detection of substantial amounts of TSLPR protein on tissue-resident CD4⁺ T cells, but not in the blood. Furthermore, there was a significant correlation between the number of TSLP-responsive memory CD4⁺ T cells in the blood and the number of TSLPR⁺ CD4⁺ T cells from autologous tissue samples (Fig. 4D). This latter finding substantiates the connection between tissue-resident and circulating cells (62-64) and offers a noninvasive approach to monitor EoE disease activity in allergic tissue. Given that TSLP induces IL-4 secretion and increases IL4R expression in human CD4⁺ T cells and that IL-4 contributes to TSLPR expression in human CD4⁺ T cells (51), it is possible that TSLP enhances the expression of its own receptor directly or indirectly through IL-4. In mice, TSLP and IL-4 synergistically generate peT_H2 cells (44, 47). Similarly, we found that the combination of TSLP and IL-4 further increased T_H2 cytokine production by human CD4⁺ T cells from patients with active EoE and with EoE in remission (Fig. 3). In support of this model, there is epistasis between genetic variants of TSLP and IL4, together being associated with an increased risk for EoE compared with variants in either gene alone (34).

To gain insight into the functional role of TSLP in human memory CD4⁺ T cells, we analyzed the transcriptional and epigenetic changes induced by this cytokine. The combination of TSLP stimulation with TCR activation enhanced the expression of genes associated with cell division and cell cycle processes, which was validated by the increased proliferation rate of TSLP-responsive cells (Fig. 5, A and B). Consistent with the heterodimeric receptor for TSLP including both TSLPR and IL-7Rα and given that both TSLP and IL-7 activate the STAT5 pathway (40), we observed enrichment in biological processes mediated by IL-7 signaling (Fig. 5A). Transcriptomic analyses showed that TSLP substantially potentiated leukocyte differentiation processes, which include increased production of specific cytokines (IL-4, IL-5, IL-9, IL-13, and GM-CSF) and increased expression of T_H2 cell–associated markers (for example, IL-2Rα and IL-4R) and specific transcription factors (BATF, GFI, IRF4, and NFIL3). On the basis of our ATAC-seq analysis, we uncovered substantial changes in chromatin state induced by TSLP that were enriched in loci encoding genes involved in T cell activation, proliferation, and cytokine production and included T_H2-associated gene loci (for example, IL13/RAD50, IL3/CSF2, IL2RA, and NFIL3) (Fig. 5, D to G). TSLP-induced open chromatin regions were strongly enriched in motifs for IRF4, HIF-1β, STAT6, and AP-1 family members, including ATF3, BATF, FRA1/2, and Jun (Fig. 5H), which represent major transcription factors for T cell activation, proliferation, and differentiation (58, 65, 66).

On the basis of previously and currently reported findings, it is tempting to speculate a role for TSLP in expanding peT_H2 cells in allergic disease based on the findings that (i) TSLP-mediated IL-13 production in CD4⁺ T cells further increases TSLP secretion from epithelial cells (67); (ii) TSLP-dependent secretion of GM-CSF induces the maturation of antigen-presenting cells, such as dendritic cells and macrophages, and, as a result, enhances their capacity to activate CD4⁺ T cells (68); (iii) TSLP-mediated expression of IL-2Rα and IL-4Rα potentiates the autocrine response to secreted IL-2 and IL-4, which further amplifies T_H2 cell responses and expands peT_H2 cells specifically from a selected subpopulation (3, 4); and (iv) subsequently, by enhancing the proliferation of T_H2 cytokine–producing cells (fig. S4A), TSLP may promote the expansion of peT_H2 cells in the context of allergic disease. Together, these findings suggest that the TSLP axis seems to actively control a key checkpoint in allergic responses by providing feed-forward signals for the terminal differentiation and expansion of T_H2 cells within a subpopulation of memory T_H cells (Fig. 8).

Fig. 8. Summary schema of the mechanisms of TSLP-induced generation and expansion of human pathogenic T_H2 cells.

Gray circles represent TSLPR⁻ memory CD4⁺ T cells, and the blue circle represents TSLPR⁺ memory CD4⁺ T cells. Exposure to TSLP during antigenic activation of human memory TSLPR⁺CD4⁺ T cells induces epigenetic and transcriptional changes, as a consequence of which, feed-forward mechanisms are activated. The increased secretion of IL-13 leads to further stimulation of epithelial cells to produce TSLP, representing a positive feedback loop for pathogenic effector T_H2 (peT_H2) cell differentiation. Increased GM-CSF production induces the maturation of dendritic cells and macrophages, which enhance the activation, and thus the proliferation, of T cells. TSLP-mediated expression of IL-2Rα, IL-4Rα, and IL-4 amplifies T_H2 cell differentiation and proliferation. The peT_H2 cells also provide a priming and survival cytokine milieu to nourish the eosinophils and mast cells in the allergic tissue, promoting type 2 pathogenesis. Pathways discovered in this study are represented by solid arrows, whereas published pathways are shown by dashed arrows.

TSLP is produced in inflamed tissue and is genetically and functionally linked to various types of EoE, including lymphocytic (69) and classical (2, 19, 31, 35) forms of the disease, as well as other allergic diseases (23-25, 28, 36, 70). Although our study focused on classic EoE as a prototype of allergic disease, it is possible that our findings extend to other allergic disorders that involve a high frequency of TSLP-responsive cells. In addition, it is important to note that patients with EoE may also have a history of other allergic disorders. This aspect expands the possibility of detecting TSLP/TSLPR-dependent atopy, although it does limit our conclusions specifically to EoE.

Despite the robust disease context in blood and tissue pSTAT5/TSLPR assessment, a portion of memory CD4⁺ cells were less responsive to TSLP in some patients with active EoE and EoE in remission. The cause of this variation is unknown and might be associated with many factors, including differential response to treatment, natural history of diseases and vaccine exposure, genetic and epigenetic variabilities, or a combination of these factors. Studies showed that treatment with anti-TSLP (tezepelumab) has a substantial effect on uncontrolled asthma and moderate and severe AD (38, 71). However, with response rates ranging from 61 to 71% in asthma, a substantial group of patients remains unresponsive to the treatment (38, 72). In light of these results and the high clinical cost of anti-TSLP biologics, our study expands the rationale for a noninvasive screening test before treatment, aiming to offer anti-TSLP treatment only to patients with memory CD4⁺ T cells that exhibit high responsiveness to TSLP.

Most of the previous work concerning the pathogenic role of TSLP in T cells has predominantly focused on murine models (44-47). Here, using EoE as a human allergic disease prototype, we have identified a subpopulation of memory CD4⁺ T cells in allergic tissue and blood that has a high responsiveness to TSLP, resulting in a T_H2 pathogenic phenotype. These findings do not challenge the conventional concept that TSLP polarizes dendritic cells for type 2 responses; rather, we propose a parallel, direct mechanism involving TSLP-mediated human T_H2 cell activation and proliferation. Because TSLP is a pleotropic cytokine that can affect different cell populations (17, 18, 21, 43, 73), the direct action of TSLP on T cells deserves further attention in the context of human diseases, because it can expand the value of anti-TSLP therapy by offering personalized and precision medicine opportunities.

MATERIALS AND METHODS

Study design

The aims of this study were to elucidate the role of TSLP in the generation of peT_H2 in the context of human allergic disease and to understand the molecular mechanism of its action. We used esophageal tissue and blood of patients with EoE versus control individuals and determined the responsiveness of their T cells to TSLP. TSLPR expression was examined on human CD4⁺ T cells by RNA-seq (single-cell and bulk populations) and flow cytometry. TSLP-induced STAT5 activation in CD4⁺ T cells was determined by detecting pSTAT5 by flow cytometry, and this was used as a direct method to validate the responsiveness of cells to TSLP. To measure the functional properties of TSLP-responsive CD4⁺ T cells, cells were activated with TSLP in the presence or absence of TCR signaling, which was followed by measurement of cytokine secretion and flow cytometry analysis. To gain insight into the molecular mechanism of TSLP action on human CD4⁺ T cells, RNA-seq and ATAC-seq were performed on cells stimulated in the presence of TSLP. We correlated the responsiveness of memory CD4⁺ T cells to TSLP with tissue eosinophil numbers and the abundance of TSLPR in the tissue and blood cells. There was no intervention, blinding, or randomization conducted as part of study recruitment.

Study approval

Samples were obtained after informed consent, under the auspices of the Institutional Review Board (IRB) of Cincinnati Children’s Hospital Medical Center (CCHMC) (2016-0123). All human participant recruitment complied with relevant ethical regulations according to our protocol approved by CCHMC’s IRB. Written informed consent was received from participants before inclusion in the study.

Human samples

Esophageal biopsies and blood samples were acquired at the Allergy and Gastroenterology outpatient clinics (CCHMC) systemically from patients who were having an endoscopy for EoE or related symptoms. Active EoE was defined by a diagnosis of EoE and a histologic finding of ≥15 eosinophils per microscopic high-power field (HPF) with clinical symptoms. EoE remission (inactive EoE control) was defined as any patient (with EoE history) whose tissue eosinophil count was <15 eosinophils/HPF. Complete EoE remission was defined by tissue with 0 eosinophils/HPF. Control individuals were defined as patients without a history of EoE or esophageal pathology. Clinical and demographic parameters of participants are summarized in table S4.

Cell isolation from blood and biopsies

Human PBMCs were prepared by Ficoll gradient centrifugation. CD4 MicroBeads (Miltenyi Biotec) were used to enrich the total CD4⁺ T cell population. The EasySep Human Naïve CD4⁺ T Cell Isolation Kit and EasySep Human Memory CD4⁺ T Cell Enrichment Kit (STEMCELL Technologies, Vancouver, Canada) were used to isolate naïve and memory cells, respectively. Memory CD4⁺ T cells for RNA-seq and cytokine multiplex panel analyses were also isolated by a Sony/SH800S cell sorter (Sony Biotechnology) after enrichment of CD4⁺ T cells by CD4 MicroBeads. Purity of isolated CD3⁺CD4⁺CD45RO⁺ or CD3⁺CD4⁺CD45RA⁺ cells was >97% as assessed by flow cytometry. A single biopsy from the distal esophagus was collected into RPMI medium supplemented with 10% fetal bovine serum (FBS), kept on ice, and transported to the research laboratory for processing as previously described (13). Briefly, the biopsy was transferred into EDTA buffer for 15 min at 37°C, washed once with phosphate-buffered saline (PBS), minced, and then subjected to collagenase A digestion for 30 min. The resulting suspension was diluted with ice-cold PBS, passed through a 19-gauge needle, filtered through two layers of gauze, and washed with ice-cold PBS.

Human CD4⁺ T cell activation

Human T cells were activated as previously described (44) in RPMI 1640 medium (Thermo Fisher Scientific) containing 10% FBS, l-glutamine, β-mercaptoethanol, and antibiotics. Purified human CD4⁺ T cells were activated with soluble anti-CD3 (5 μg/ml), anti-CD28 (2 μg/ml; Bio X Cell), and anti–IFN-γ (5 μg/ml; BioLegend) in the presence or absence of TSLP (40 ng/ml; R&D Systems) and/or IL-4 (40 ng/ml; PeproTech). Cell culture medium, mRNA, and DNAwere collected after 3 days, and flow cytometry analysis of surface markers, cytokines, and intracellular proteins was performed after 7 days of cell activation.

Proliferation assay

Purified naïve or memory CD4⁺ T cells were labeled for 8 min at room temperature with 2.5 μM CFSE (Molecular Probes) and then activated with soluble anti-CD3, anti-CD28, and anti–IFN-γ antibodies with or without TSLP for 7 days. From the distribution of the proportion of cells in each CFSE peak, the expected cell yield was calculated according to the formula

$$n=n_0∕\sum _{i=0}^k\frac{fi}{2i}$$

where $n$ is the expected cell yield, $n_0$ is the initial cell number, $f_i$ is the fraction of the cells at a given $i$ cycle, and $k$ is the highest cell cycle.

Flow cytometry analysis

Freshly obtained or cultured cells were stained with fluorescent antibodies against human CD4, CD3, CD8a, CD11c, CD45RA, CD45RO, and TSLPR (see table S5). For pSTAT5 detection, freshly isolated cells were stimulated with TSLP (40 ng/ml; R&D Systems) or IL-2 (100 U/ml; Roche) for 20 min and fixed with BD Cytofix Fixation Buffer for 10 min, which was followed by permeabilization in 90% methanol. Cells were stained with fluorescent antibodies for pSTAT5, CD4, CD3, CD8, and CD45RO (see table S5). To detect intracellular proteins, cells were first stained with fluorescent antibodies against human CD4, CD25, and TSLPR surface markers for 20 min and then fixed and permeabilized with a FoxP3 fixation kit (eBioscience). Cells were stained with fluorescent antibodies against IRF4 (table S5). For intracellular cytokine detection, cells were stimulated with phorbol 12,13-dibutyrate (PDBu) (500 ng/ml; Sigma-Aldrich) and 1 μM ionomycin (Sigma-Aldrich) for 1.5 hours, and then monensin (BioLegend) and brefeldin A (eBioscience) were added for an additional 4 hours. Cells from biopsies were stained for 15 min with fluorescent antibodies against the human surface markers CD3, CD4, CD8, and TSLPR (table S5) and then were fixed and permeabilized with the FoxP3 fixation kit (eBioscience). Intracellular proteins were stained with fluorescent antibodies against human IL-4, IL-10, IFN-γ, IL-13, GATA3, and FOXP3 (table S5). Cultured memory CD4⁺ cells from blood were fixed and permeabilized with a Cytofix/Cytoperm kit (BD Biosciences) and stained with fluorescent antibodies against human IL-4, IL-5, IL-13, and IFN-γ (table S5). Freshly isolated cells were stained in the presence of anti-Fc blocking antibody (BD Biosciences). Data acquired with a BD LSRFortessa flow cytometer (BD Biosciences) were analyzed by FlowJo software (Tree Star Inc.).

Multiplex cytokine assay and ELISA

Culture medium from 3 days of memory CD4⁺ T cell culture was analyzed with the Human Cytokine/Chemokine 65-Plex Panel (Eve Technologies Corp.). The DuoSet human IL-5 enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems) was used to measure the concentration of IL-5 in the culture medium on day 3.

mRNA-seq and quantitative PCR

Human total or memory CD4⁺ T cells were activated with anti-CD3, anti-CD28, and anti–IFN-γ antibodies in the presence or absence of TSLP, IL-4, or both for 3 days. mRNA was isolated with the Zymo mini-elute kit (Zymo Research). For real-time polymerase chain reaction (PCR) analysis, mRNA was reverse-transcribed with the ProtoScript cDNA synthesis kit (New England BioLabs). Primers specific for human IL4, IL5, IL13, IRF4, BATF, GFI1, IL2RA, IL4R, CRLF2, and EIF3K were obtained from Integrated DNATechnologies (see table S6 for sequences). The ABI QuantStudio 7 Flex Real-Time PCR System (Thermo Fisher Scientific) was used to quantitate PCR products. The abundances of the transcripts of interest were determined relative to that of EIF3K cDNA. For bulk RNA-seq analysis of CD3-sorted blood and tissue lymphocytes, samples were processed as previously described (13). Single-cell suspensions were made from freshly obtained biopsies and autologous circulating blood. Samples were stained for CD3 (HIT3a)–AF647, CD45 (2D1)–APC-H7, and Zombie Violet viability dyes and subjected to FACS sorting for CD45⁺CD3⁺ live singlet events. The Nextera XT DNA Sample Preparation Kit (Illumina) was used to generate the cDNA libraries, which were subsequently processed by the CCHMC DNA Core for RNA-seq at 30 million read depth, 75–base pair (bp) × 2 pair-ended next-generation sequencing (NGS) by Illumina’s HiSeq 2500 machine. Raw reads were aligned to hg19 annotation (GRCh37/hg19), and the subsequent quantification and analysis were performed with Strand NGS software (Agilent Inc.). For RNA-seq analysis of blood memory CD4⁺ T cells activated by anti-CD3 and anti-CD28 antibodies in the presence or absence of TSLP, mRNA was purified with the Zymo mini-elute kit. The initial cDNA amplification step for all samples was performed at the CCHMC RNA Expression Core with the NuGEN Ovation RNA-seq System v2 Assay. The Nextera XT DNA Sample Preparation Kit (Illumina) was used to generate cDNA libraries, which were subsequently processed by the CCHMC DNA Core for RNA-seq at 30 million read depth, 75 bp × 2 pair-ended NGS by Illumina’s HiSeq 2500 machine. RNA-seq data are available from the Sequence Read Archive (SRA) database (accession number: PRJNA916236). RNA-seq preprocessing analysis was performed in the Scientific Data Analysis Platform (SciDAP) (https://scidap.com) (Datirium LLC) with the “RNA-seq paired-end” pipeline. The containerized Common Workflow Language pipeline is available at https://scidap.com/public/workflows/beXePHzgPzDWFB6W5. Raw sequencing data were aligned to the GRCh38/hg38 reference genome with STAR aligner, enabling a maximum of five mismatches per read. Gene expression was calculated from uniquely mapped reads based on University of California Santa Cruz (UCSC) genome browser annotations for the GRCh38/hg38 genome. Differential gene expression analysis was conducted in the SciDAP (https://scidap.com) data analysis platform with the “DESeq2 Multi-Factor Analysis” pipeline. The pipeline was written according to the Common Workflow Language (CWL v1.0) specification and was processed with CWL-Airflow runner (74, 75). Samples from five donors that were left untreated or were treated with TSLP were split into two groups by the treatment condition (with or without TSLP) and compared with DESeq2. To model the difference caused by both treatment and donor factors, the likelihood ratio test with full “~treatment + donor” and reduced “~donor” formulas was used. The resulting differentially expressed genes were filtered on the basis of the absolute log₂(fold change) value being ≥ 1. In addition, those genes for which the maximum reads per kilobase per million mapped reads (RPKM) values among all compared samples were <1 were excluded. As a significance threshold, the adjusted P value of <0.1 was used. Biological process analysis was performed with the ToppGene Suite website (76).

Assay for transposase-accessible chromatin with high-throughput sequencing

Human memory CD4⁺ T cells from the blood were activated with anti-CD3, anti-CD28, and anti–IFN-γ antibodies in the presence or absence of TSLP for 3 days, and 5 ×10⁴ cells were collected and processed according to the OMNI-ATAC protocol as described previously (77). Libraries were sequenced as PE 2 ×150 on HiSeq 4000 at Novogene. ATAC-seq data are available from the SRA database (accession number: PRJNA916236).

ATAC-seq analysis

ATAC-seq samples (six TSLP-treated and six untreated paired samples of TSLP high-responsive memory CD4⁺ T cells from patients with EoE) were analyzed with the ATAC-seq paired-end pipeline in SciDAP (https://scidap.com). The containerized Common Workflow Language pipeline is available at https://scidap.com/public/workflows/qzE4WBqpG5vSYBiGS. Briefly, reads were trimmed with Trim Galore (https://bioinformatics.babraham.ac.uk/projects/trim_galore/) and aligned to the hg38 genome with BowTie (78), enabling a maximum of three mismatches per read. Uniquely aligned reads were cleaned from PCR duplicates by Sequence Alignment/Map tools (79) and used for peak calling in MACS2 (80) with a false discovery rate (FDR) of 0.05. Only peaks present in all samples within each group (TSLP-treated or untreated) were used for differential binding analyses, comparing TSLP-treated and untreated samples with DiffBind: http://bioconductor.org/packages/release/bioc/vignettes/DiffBind/inst/doc/DiffBind.pdf. DiffBind was performed in “paired” fashion with the Empirical Analysis of Digital Gene Expression Data in R (EdgeR) package (81) used for P value calculations. Loci with P < 0.05 in the combined analysis of six pairs of samples and log₂(fold change) ≥0.4 were considered differentially accessible and assigned to the nearest genes if the distance between the gene transcription start site and peak was <50 kb. The gene list associated with differentially bound sites was intersected with the gene list obtained from differential expression analyses of RNA-seq data. To analyze known motif enrichment analyses, all differentially bound sites were divided into two groups on the basis of log₂(fold change) value. Those sites with log₂(fold change) ≥ 0.4 and P ≤ 0.05 formed a group of TSLP-treated binding sites, and those with log₂(fold change) < 0.4 formed an untreated group. All MACS2 peaks of the same condition (TSLP-treated or untreated) were first concatenated and then intersected with the corresponding group of differentially bound sites and merged to remove overlaps. The resulting intervals from TSLP-treated samples formed target regions, whereas those of the untreated samples formed background regions. Both target and background regions were used by HOMER Motif Analyses (http://homer.ucsd.edu/homer/motif/) within SciDAP to identify enriched motifs.

Reanalysis of published scRNA-seq datasets

Whole-transcriptome sequencing data from six esophageal biopsies obtained from patients with EoE with >15 eosinophils per HPF in the esophageal biopsy by Morgan et al. (12) (accession number: GSE175930) were used. For the analysis of all sequenced esophageal samples, cells were filtered on the basis of unique feature counts of >200 or <4000, with less than 20% mitochondrial counts. In addition, genes that were expressed in fewer than three cells were excluded. In summary, 8473 cells and 21,994 genes passed these filter criteria. The relative expression of a given gene was calculated by log normalization and centering. Principal components analysis was performed with the list of the top 2000 variable genes. Harmony integration was then performed, and the data were subjected to Uniform Manifold Approximation and Projection (UMAP) and shared nearest neighbor (SNN) modularity optimization–based clustering. Top marker genes with high specificity were used to classify cell markers into 10 known cell types: epithelium, fibroblast, endothelium, CD4⁺ T cells, CD8⁺ T cells, B cells, myeloid cells, neutrophils, mast cells, and eosinophils.

Statistical analysis

The results are presented as means ± SEM or displayed with box and whiskers plots. Paired samples were analyzed with paired, two-tailed t tests, whereas unpaired samples were analyzed with unpaired, two-tailed t tests. Multiple-group comparisons were conducted using one- or two-way analysis of variance (ANOVA), with FDR correction applied. Nonparametric, two-tailed t tests or nonparametric one-way ANOVA was used for data that did not pass normality tests. Spearman correlation was used to evaluate the strength and direction of the monotonic relationship between two variables. All statistical analyses were performed with Prism 9 software (GraphPad Software Inc.). To calculate adjusted P values for mRNA-seq and ATAC-seq analyses, DESeq2 and DiffBund were used, respectively, as described earlier. P < 0.05 was considered statistically significant.