Human peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation (Ficoll-Paque PLUS, Cytiva) from buffy coats of healthy human donors aged between 18 and 66. Buffy coats were provided anonymously by the Etablissement Français du Sang, Toulouse, France. CD34⁺ precursor cells were isolated from the PBMCs by positive selection with anti-CD34 magnetic beads (EasySep™ Human CD34 Positive Selection Kit, STEMCELL Technologies). CD34⁺ cells were grown in serum-free StemSpan™ medium (STEMCELL Technologies) supplemented with recombinant human IL-6 (50 ng/mL; Peprotech), recombinant human IL-3 (10 ng/mL; Peprotech), and 3% supernatant of CHO transfectants secreting murine SCF (a gift from Dr. P. Dubreuil, Marseille, France; 3% corresponds to 50 ng/mL SCF) for 1 week at 37°C, 5% CO₂. Cells were next grown in IMDM GlutaMAX-I supplemented with sodium pyruvate, 2-mercaptoethanol, 0.5% BSA, insulin-transferrin selenium, penicillin/streptomycin (100 U/mL/100 µg/mL) (all from Invitrogen), IL-6 (50 ng/mL), and 3% supernatant of CHO transfectants secreting murine SCF for 8 weeks and tested phenotypically (CD117⁺, FcϵRI⁺, ST2⁺) and functionally (β-hexosaminidase release in response to FcεRI crosslinking) before being used for experiments. Only primary cell lines exhibiting more than 95% CD117⁺/FcϵRI⁺/ST2⁺ cells were used for experiments.

A total of 1 × 10⁵ memory CD4⁺ T cells freshly purified from PBMCs by immunomagnetic negative selection (EasySep™ Human memory CD4⁺ T cell enrichment kit, STEMCELL Technologies) were cocultured at a 1:1 ratio with human primary MCs (primed or not with 10 ng/mL of human recombinant IL-33 for 4 h and next washed) for 6 days in the presence of magnetic beads coated with anti-CD3 and anti-CD28 (anti-CD3/CD28 beads) mAbs (Dynabeads, Life Technologies) at 1 bead:10 T cells ratio in coculture medium [RPMI 1640 supplemented with 10% serum replacement medium (knockout medium, Life Technologies), GlutaMAX-I, sodium pyruvate, 2-mercaptoethanol, and 1% supernatant of CHO transfectants secreting murine SCF].

To induce an APC phenotype, primary human MCs were treated with recombinant human INF-γ (50 ng/mL, Peprotech) for 48 h at 37°C. MCs were next washed and primed with recombinant human IL-33 (10 ng/mL, Peprotech) for 4 h at 37°C. Subsequently, MCs were washed again and either pulsed or not with a cocktail of superantigens (50 ng/mL TSST-1, SEA, SEB, SEE, Sec-1; Toxin Technology, Sarasota, Florida) for 2 h. After two additional washes, 1 × 10⁵ MCs were cocultured at a 1:1 ratio with memory CD4⁺ T cells for 6 days in RPMI 1640 medium supplemented with 10% serum replacement (knockout medium, Life Technologies), GlutaMAX-I, sodium pyruvate, 2-mercaptoethanol, and 1% supernatant from CHO transfectants secreting murine SCF. When indicated, cocultures were treated with anti-OX40L mAb (10 µg/mL, R4930 Oxelumab, Novus Biologicals) or Ig control (10 µg/mL, Normal Rabbit IgG Control, R&D Systems). At day 6, cells were processed for intracellular cytokine staining.

To evaluate CD4⁺ T-cell cytokine production (IFN-γ/IL-17/IL-4/IL-13/IL-9) post-coculture, CD4⁺ T cells were sorted by FACS (FACSAria or FACSMelody cell sorter, BD Biosciences) (CD117^-, SSC^low) at day 6 and restimulated with PMA (50 ng/mL) and ionomycin (1 µg/mL) for 5 h. Supernatants were harvested and stored at −80°C. Cytokine concentrations were measured using a bead-based multiplex assay (LEGENDplex™, BioLegend) following the manufacturer’s recommendations. Flow cytometric acquisitions were performed with a MACSQuant^® Analyzer 10 instrument (Miltenyi Biotec) and analyzed with FlowJo V10 software (TreeStar).

For MC staining, cells were plated on poly-L-lysine (Sigma)-coated slides and fixed in 4% PFA for 15 min at RT and next blocked/permeabilized with PBS, 1% BSA, and 0.1% saponin (blocking buffer) for 30 min at RT. They were incubated with mouse anti-tryptase (IgG1 kappa, clone AA1, Dako) and anti-OX40L (polyclonal rabbit IgG, Invitrogen) at RT for 1 h in blocking buffer. MCs were washed and incubated with matched donkey AlexaFluor-conjugated secondary antibodies (Invitrogen) at RT for 1 h in blocking buffer and next counterstained with DAPI (1 µg/mL, Invitrogen).

For immunological synapse (IS) staining, SAg-pulsed or unpulsed MCs and CD3/CD28 pre-activated (24 h) CD4⁺ memory T cells were mixed at a 1:1 ratio and centrifuged at 70g for 30 s in a U-bottom 96-well plate and incubated for 10 min at 37°C in coculture medium to allow conjugate formation. They were next gently pipetted and plated on poly-L-lysine-coated slides for 7 min before fixation in methanol for 15 min at −20°C. Cells were blocked/permeabilized in blocking buffer for 30 min at RT and subsequently stained with mouse anti-OX40 mAb (IgG1 kappa, clone ACT35, BD Pharmingen), rabbit anti-OX40L (polyclonal rabbit, PA5-116057, Invitrogen), and anti-CD4 (polyclonal goat, AF-379-NA, R&D Systems). They were next incubated with matched AlexaFluor-conjugated secondary antibodies (Invitrogen) and counterstained with DAPI. Images were acquired using a Zeiss LSM 780 confocal microscope and analyzed using Zen (Carl Zeiss Microscopy) and ImageJ software.

To infer the helped MC-activated CD4⁺ memory T-cell communication network, we based our analysis on our previous study where MCs were stimulated by IL-33 for 4 h and processed for RNAseq analysis (GEO accession number GSE235240) using the publicly available database of ligand–receptor pairs collated by Noel and co-workers (28) and a computation method previously described (28, 29). A thresholding method was used to infer “active” L–R pairs, and communication score computation (expression product method) was used to rank the active L–R pairs. Briefly, average expression levels (FPKM) of the genes corresponding to the ligands in the database were selected from the list of DEGs between IL-33-primed and resting MCs (FDR <0.01 and log₂ fold change >1). Average expression levels (mean FPKM value) of the genes corresponding to the receptors in the database were selected from RNAseq data of memory CD4+ T cells activated with anti-CD3/CD28-coated beads for 24 h (GEO accession number GSE73214) (30). For thresholding, only ligands and receptors with FPKM > 1 were kept in the analysis. Intercellular communication score was calculated as previously described (28): ligand and receptor gene expression were next scaled by the maximum value of gene expression among ligands and receptors, respectively, and multiplied by 10 to obtain values ranging from 0 to 10. The maximum value of gene expression was calculated as the mean of the 10% highest values. Scaled values above 10 (outliers) were coerced to 10. Because the database employed here takes into account multichain receptors or ligands, we first calculated the expression level for the multichain molecules as the geometric average of the values of the ligand/receptor chain expression. Next, the ligand–receptor pair score was determined by the product of the expression level of the ligand by the expression level of the receptor (if L_i is the average expression level of ligand i by helped MC and R_i is the average expression level of the corresponding receptor by activated CD4⁺ memory T cell, the score S_i of this interaction is S_i = L_i.R_i). Inferred active ligand–receptor pairs (S_i > 0) are provided in Supplementary Table S1 . Custom R scripts were used for analyses. For circular visualization of the links between MC^IL-33 ligands and activated memory CD4⁺ T-cell corresponding receptors, the circlize R package was used (31) to represent the 30 top-ranked L–R pairs according to their computed communication score.

Statistical analyses for all experiments were conducted in GraphPad Prism 9 (GraphPad Software) or with rstatix and PMCMRplus R packages. Data were first analyzed for normality with Shapiro–Wilk test and variance homogeneity with Levene’s test. To determine statistical significance in assays comparing more than two groups, data showing homogeneity of variance and normality were analyzed with repeated-measures one-way ANOVA followed up by paired pairwise t-test with Bonferroni correction, otherwise with Friedman test followed up by pairwise Wilcoxon signed-rank tests with Bonferroni correction. All p-values are two-sided (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant).

To analyze the effects of IL-33 priming on MCs and their subsequent influence on Th cell responses, we devised a coculture system in which freshly isolated peripheral blood CD4⁺ memory T cells were stimulated with anti-CD3/CD28 beads to mimic antigenic recall in the presence of allogeneic human MCs (pure primary culture derived from peripheral blood precursors) either primed or unprimed with IL-33 ( Figure 1A ). Coculture with MCs, regardless of IL-33 priming, increased the proliferation ( Figure 1B ) and early (CD69 and CD25) or late (CD137) activation marker expression on the CD4⁺ T-cell surface ( Supplementary Figure S1 ), suggesting that MCs provide activating signals to CD4⁺ T cells independently of IL-33. We next analyzed typical Th cell cytokine (IFN-γ, IL-17, IL-4, IL-13, and IL-9) production by CD4⁺ T cells after 6 days of coculture. We observed that unprimed MCs increased the percentages of CD4⁺ T cells producing these cytokines, indicating that MC presence facilitates T-cell cytokine production ( Figure 1C ). MCs primed with IL-33 (MC^IL-33) significantly increased the percentages of IL-9⁺ (and, to a lesser extent, IL-13⁺) CD4⁺ T cells and decreased IL-17⁺ T-cell percentage without affecting the proportions of IFN-γ⁺ or IL-4⁺ T cells ( Figure 1C ). Moreover, the number of IL-9⁺ cells increased in the presence of MC^IL-33( Figure 1D ). Together, these results demonstrate that IL-33 enhances the capacity of MCs to drive Th cells toward IL-13 and IL-9 production.

Because IL-33 endows MCs with the capability to foster IL-9 production by Th cells, we sought to elucidate the underlying mechanism of this phenomenon. In a previous study, we analyzed the transcriptome of primary human MCs stimulated with IL-33 using RNA sequencing (GSE235240) (32), revealing that IL-33 alters the expression of more than 2,600 genes. Here, we investigated whether the genes upregulated in MCs following IL-33 stimulation might play a role in the interaction between MCs and activated memory CD4⁺ T cells.

To this end, we analyzed putative MC-activated memory CD4⁺ T-cell communication based on MC ligand expression upregulated by IL-33 in our RNAseq dataset and on receptor expression in memory CD4⁺ T cells activated for 24 h with anti-CD3/28 beads retrieved in the publicly available dataset (30). We computed a communication score for each L–R pair in a publicly available database by adapting the computational method described in the ICELLNET framework (28) dedicated to infer cell–cell communication. This score is based on both the expression level thresholding method (to determine active L–R pairs) and the R–L expression product method (to rank the active L–R pairs) (29) with MCs as sender cells and activated Th cells as receiver cells. This analysis inferred 53 active L–R pairs ( Supplementary Table S1 ), suggesting that IL-33 upregulated in MCs the expression of several molecules involved in T-cell costimulation such as CD80, CD58, ICAM1, and TNF superfamily (TNFSF) molecules CD70 (TNFSF7), OX40L (TNFSF4), and CD137L (TNFSF9) ( Figures 2A, B ). Among these molecules, we focused on OX40L because it was shown to be associated with Th2 promotion (10, 33, 34).

We confirmed by flow cytometry that IL-33 induced OX40L expression on the MC surface ( Figures 2C, D ). Because other type 2 response alarmins might prime MCs and it was reported that in human CD11c⁺ DCs express OX40L in response to TSLP but not to IL-33 (33), we analyzed the impact of MC priming by TSLP and IL-25 and observed that these alarmins did not induce OX40L expression on the MC surface ( Supplementary Figure S2 ). To test the effect of OX40/OX40L costimulation on Th cell polarization, we used anti-OX40L blocking mAb oxelumab. The addition of oxelumab significantly impaired MC’s ability to increase the percentage of IL-9⁺ Th cells ( Figures 2E, F ). A trend to reduce IL-13⁺ Th cell proportion in the presence of oxelumab was also noted. This result indicates that OX40 stimulation is required to promote IL-9 (and to a lesser extent IL-13) production by Th cells in our coculture system.

Because we and others previously showed that IFN-γ induced an APC phenotype in MCs (MHC class II and costimulatory molecule expression) (27, 35), we tested whether IL-33 also acts on MCs’ capability to polarize Th cells upon cognate interactions. We induced an APC phenotype in MCs by stimulating them with IFN-γ for 48 h and analyzed the effect of IL-33 priming on their phenotype (CD54, HLA-class II molecules and OX40L expression). IL-33 synergized with IFN-γ to induce MHC class II molecules and CD54 but IFN-γ did not affect OX40L expression on the MC surface ( Supplementary Figure S3 ).

IFN-γ-primed MCs were pulsed with a cocktail of bacterial superantigens to activate a substantial fraction of the polyclonal memory CD4⁺ T-cell population (approximately 70%) (36) and cocultured with CD4⁺ memory T cells for 6 days ( Figure 3A ). We observed that SAg-pulsed MC efficiently stimulated polyclonal T-cell proliferation and that IL-33 priming did not change MCs’ capacity to stimulate CD4⁺ T-cell proliferation ( Figure 3B ), indicating that IL-33 priming did not change the potential of MCs to activate Teff cells. To check whether MC^IL-33–Th cell cognate interactions did occur upon coculture and whether OX40L was involved at the IS, we analyzed IS formation by confocal laser scanning microscopy 2 h after coculture. No conjugates were observed with IFN-γ-primed MC^IL-33 unpulsed with SAg. Approximately 10% of IFN-γ-primed MC^IL-33 pulsed with SAg were found conjugated with memory CD4⁺ T cells after 2 h of coculture, indicating that MC^IL-33–CD4⁺ T-cell cognate interactions did occur. Conjugates were analyzed for OX40L and OX40 enrichment at the cell–cell contact area. We observed that approximately half of the conjugates showed OX40L enrichment at the IS, suggesting the involvement of Th cell OX40 costimulation upon cognate interaction with MCs ( Figures 3C–E ). We next analyzed cytokine production by Th cells at day 6 of coculture ( Figure 3F ). IL-33-primed MCs showed an enhanced capacity to increase the proportion of IL-9⁺ and IL-13⁺ Th cells and to reduce the percentage of IL-17⁺ Th cells ( Figure 3F ). The addition of blocking anti-OX40L mAb indicated that IL-9⁺ Th cell differentiation required OX40–OX40L interaction ( Figures 3G, H ). Taken together, these results show that MC^IL-33 pretreated with IFN-γ are capable of stimulating memory CD4⁺ T cells in a cognate manner and to foster IL-9⁺ Th cells via OX40L expression.