Regulatory T cells (Tregs) are required to inhibit autoimmunity and limit excessive immune responses in colitis and graft-versus-host disease (GvHD) [1]. They inhibit T-cell responses and inflammation by producing tolerogenic cytokines such as IL-10 [2]. IL-10 antagonizes the maturation of dendritic cells and inhibits consequently the production of pro-inflammatory cytokines and the up-regulation of MHC class-II and costimulatory molecules that control T-cell activation [3]. The best-defined regulatory T cells express CD25 and FOXP3, and can develop in the thymus. FOXP3^– regulatory T cells that are generated exclusively in the periphery have also been described. FOXP3^–Tregs that produce IL-10 are often called type-1 regulatory T cells (Tr1, [1]). Recently, Eomesodermin (EOMES) was shown in several independent studies to program cytotoxicity and IL-10 production in Tr1-like cells in humans and mice [4-7]. Since EOMES also antagonizes CD4⁺ T-cell differentiation to alternative fates, it acts as a lineage-defining transcription factor [2, 5, 6]. Human EOMES⁺Tr1-like cells express high levels of granzyme-K (GzmK), CCR5, and IFN-γ, and have downregulated IL-7R, CCR6 and CD40L [4]. However, while GzmK and IFN-γ are directly induced by EOMES, IL-10 production and loss of IL-7R and CD40L are also regulated by extracellular cues, like TCR signal strength [8] and IL-27 [2, 4].

The role of EOMES in CD4⁺ T cells is multifaceted [9]. Thus, CD4⁺ cytotoxic T cells (CTL) are also EOMES⁺. They express GzmB and have lost CD27 and CD28 expression, suggesting that they represent terminally differentiated effector cells [10]. Moreover, EOMES is also expressed in “non-classical” human Th17-derived Th1 cells, which co-express CCR6 and CD161 and possess pro-inflammatory properties [11]. The relationships between pro-inflammatory EOMES⁺Th1 cells, CD4⁺CTL, and EOMES⁺Tr1-like cells are unclear [9]. We hypothesized that EOMES⁺Th1 cells contain pre-committed precursors of Tr1-like cells (“pre-Tr1”) that could switch their cytokine profiles to inhibit excessive inflammation upon chronic antigenic stimulation in a negative feed-back loop [2, 4]. EOMES⁺CD4⁺T cells play key roles in immune-mediated diseases [2]. Thus, it was reported that CD4⁺CTL are involved in progressive multiple sclerosis (MS) [12] and that EOMES in CD4⁺T cells plays a pathogenic role in experimental autoimmune encephalitis (EAE), a standard animal MS model [13]. Conversely, EOMES⁺Tr1-like cells play a protective role in GvHD [6]. Furthermore, they could inhibit protective CD8⁺ T-cell responses in human tumors [14], but may instead kill leukemia cells in mice [7]. These paradoxical findings may be explained by a heterogeneity of EOMES⁺CD4⁺ T cells. Here, we directly compared human EOMES⁺CD4⁺ T cell subsets to reconcile the different reported properties of CD4⁺EOMES⁺ T cells [4, 9, 11].

CD4⁺ T cells from human peripheral blood of healthy blood donors were analyzed ex vivo by multi-dimensional flow cytometry, dimensionality reduction, and FlowSOM clustering (Supporting Information Table S1, Fig. 1A and B, Supporting Information Fig. S1A–C). 11 CD4⁺ T cell clusters were identified that separated into two superclusters, one containing naïve and central memory T cells (T_CM, Clusters 1–4), and the other one containing effector memory (T_EM) and cytotoxic subsets (Clusters 5–9, 11, Fig. 1A). Cluster 10 corresponded to FOXP3⁺Tregs and clustered distantly with T_CM. Cluster 11 in the effector supercluster corresponded to CD4⁺GzmB⁺CTL, while Cluster 9 displayed the canonical phenotype of Tr1-like cells (i.e., EOMES⁺GzmK⁺CD27⁺CCR5⁺PD1⁺IL-7R^loCCR6⁻). Notably, the latter co-clustered with GzmK⁺IL-7R⁺CCR6^+/−pre-Tr1-like cells (Cluster 8, Fig. 1A). DiffusionMap (Fig. 1B), which predicts possible cell differentiation, suggested a progressive differentiation from naïve cells to T_CM, T_EM, and CD4⁺CTL as expected. Notably, Tr1 cells were positioned after T_EM clusters in this hypothetical differentiation trajectory, before CTL and immediately after pre-Tr1-cells, consistent with the hypothesized effector–precursor relationship of EOMES⁺GzmK⁺pre-Tr1- and Tr1-like subsets [2, 4]. We next analyzed CD4⁺ T cells by conventional gating. We subdivided them into four main subsets: FOXP3⁺Tregs, GzmB⁺ or GzmK⁺ cytotoxic cells [10], and Th cells that lacked FOXP3 and ranzymes (Supporting Information Fig. S2A, Supporting Information Table S2). CD4⁺GzmK⁺ T cells and GzmB⁺CD4⁺CTL expressed similar high levels of EOMES as CD8⁺ T cells, while Th cells and Tregs lacked EOMES expression (Fig. 1C). GzmK⁺ T cells were largely GzmB⁻ and were further subdivided according to CCR6 [11] and IL-7R expression [4] (Supporting Information Fig. S2A). Three subsets (Fig. 1D, Supporting Information Fig. S2A) were identified: non-classical Th1 cells (CCR6⁺), Tr1-like cells (CCR6⁻IL-7R^−/lo), and cells with an intermediate phenotype (“pre-Tr1” cells, CCR6⁻IL-7R⁺, Supporting Information Table S2) [2]. All three subsets expressed EOMES and GzmK, but Tr1 cells expressed the highest levels (Fig. 1D, Supporting Information Fig. S2A and B). CD161, a second marker of non-classical Th1 cells, was highest on CCR6⁺GzmK⁺ cells but low on EOMES⁺Tr1-like cells. CCR5 was expressed at high levels on cytotoxic T cells, including GzmB⁺ and all GzmK⁺ subsets, at low levels on Th cells and at intermediate levels on Tregs. CCR5 mediates migration to non-lymphoid tissues and is expressed on T_EM. Consistently, GzmK⁺ T cells and CTL had predominantly a T_EM phenotype, but CD4⁺CTL contained also some CD45RA⁺T_EMRA (Fig. 1C, Supporting Information Fig. S2C). Moreover, CD27 was selectively downregulated on CTL [10]. Among GzmK⁺ subsets, there was a gradient from CCR6⁺cells, which had partially downregulated CD27, to Tr1-like, which expressed high levels. Conversely, IL-7R was highly expressed on Th cells, was reduced on CD4⁺CTL, and low on FOXP3⁺Tregs. Total GzmK⁺ T cells expressed high levels of IL-7R, but among GzmK⁺ subsets there was an inverse gradient of IL-7R expression from IL-7R^loTr1 cells to IL-7R^hiCCR6⁺ cells, while pre-Tr1 cells expressed intermediate levels. PD1, which is associated with chronic TCR stimulation, was expressed rather selectively on cytotoxic CD4⁺ T-cell subsets, with Tr1-like cells expressing the highest levels [15]. Notably, CCR6⁺GzmK⁺ cells were, in contrast, largely PD1-negative, confirming that PD1 is characteristic for Tr1-like cells [15]. Finally, the activation marker CD69 was expressed at low levels on CTL, EOMES⁺Tr1-like cells, and Tregs, but was largely absent on Th cells. Consistently, ex vivo expression of Ki67, which reflects recent proliferation in vivo, was expressed in some Tregs and EOMES⁺Tr1-like cells, but was hardly detectable in Th cell, pre-Tr1 cell, and CCR6⁺GzmK⁺T cell.

We next analyzed the cytokine profiles of GzmK⁺CD4⁺T cell subsets following brief stimulation with PMA and ionomycin. Production of IFN-γ, a direct transcriptional target of EOMES [4, 9, 11], was as expected highest in cytotoxic T cells, including all GzmK subsets (Fig. 2A and B). IL-17 was produced at low levels by Th cells and by FOXP3⁺Tregs. IL-17 is inhibited by EOMES [4, 9, 11], but was nevertheless produced at low levels by CCR6⁺GzmK⁺T cells and CTL. Conversely, it was hardly detectable in pre-Tr1 and Tr1-like cells. GM-CSF, which plays a key encephalitogenic role in the EAE mouse model, was highly expressed in cytotoxic CD4⁺ T-cell subsets (Fig. 2A). However, while GM-CSF was highly expressed in CTL, pre-Tr1, and CCR6⁺GzmK⁺T cells, it was reduced in Tr1-like cells. Finally, IL-10 was produced at low levels by Th cells, at intermediate levels by FOXP3⁺Tregs and total GzmK⁺T cells, but was hardly detectable in GzmB⁺CTL (Fig. 2A). As expected, Tr1-like cells produced the highest levels of IL-10. Pre-Tr1 cells also produced some IL-10, whereas CCR6⁺GzmK⁺T cells did not. We also analyzed the upregulation of CD40L, which has a key role in T-cell-dependent B-cell responses [16]. CD40L was highly expressed on activated Th cells, but was lower on FOXP3⁺Tregs and CD4⁺CTL. Among GzmK⁺ subsets, it was efficiently upregulated by CCR6⁺ and pre-Tr1 cells, but was reduced on Tr1-like cells. Notably, IL-10 producing Th and pre-Tr1 cells co-expressed CD40L, while IL-10 producing FOXP3⁺Tregs and EOMES⁺Tr1-like subsets were largely CD40L negative (Supporting Information Fig. S2D). Overall, while Tr1-like cells produced mainly IFN-γ and relevant amounts of IL-10, CTL, CCR6⁺GzmK⁺, and also pre-Tr1 cells displayed pro-inflammatory effector cytokine profiles (Fig. 2B, Supporting Information Fig. S2D).

Regulatory T cells inhibit CD4⁺ T cell responses. Suppression of CD8⁺ T cells is only rarely assessed, but is relevant for anti-tumor and anti-viral immune responses [14]. We showed previously that EOMES⁺Tr1-like cells, but not CD4⁺CTL, possess regulatory functions [4], but if CCR6⁺GzmK⁺ T cells or pre-Tr1 possess regulatory capabilities is unknown. FACS-purified naïve CD4⁺ or CD8⁺ T cells were stimulated with anti-CD3/28-coated beads, and cell division measured by CellTrace dilution (Fig. 2C). FOXP3⁺Tregs inhibited the proliferation of naïve CD4⁺ and CD8⁺ T cells to approximately 50% and 60%, respectively. Conversely, CD4⁺CD25⁻IL-7R^loCCR5⁻ control T cells (Fig. 2C) [15] and unlabeled CD4⁺ naive T cells (data not shown) had only a weak effect on responder T-cell proliferation. Tr1 cells, pre-Tr1, and CCR6⁺CD4⁺T cells were sorted as CCR5⁺CD27⁺ [4]. Not all, but the majority of CCR5⁺CD27⁺ cells expressed EOMES and/or GzmK (Supporting Information Fig. S2E). Unexpectedly, all three subsets suppressed T-cell proliferation, and CCR6⁺ and pre-Tr1 cells were even more suppressive then CD25⁺IL-7R^lo-sorted Tregs, which were as expected mostly FOXP3⁺ (Supporting Information Fig. S2F).

In conclusion, although Th17-derived EOMES⁺Th1 cells and Tr1-like cells possess largely opposing pro- and anti-inflammatory cytokine profiles [4, 9, 11], they share nevertheless important features like expression of GzmK, CCR5, and IFN-γ. EOMES⁺CCR6⁻Th1 cells had in general intermediate properties, consistent with the view that they contain precursors of Tr1 cells [2]. Surprisingly, suppressive capabilities are correlated with GzmK expression, but not with IL-10 production. Notably, we showed previously that EOMES⁺Tr1-like cells can suppress T-dependent B-cell responses independently of IL-10 [16]. In the future, it should be assessed if EOMES⁺Th1 cells also suppress T-cell responses in vivo, as EOMES⁺Tr1-like cells do [6, 15, 16]. Moreover, it would be important to identify the molecular mechanisms of suppression. Despite these open questions, our findings challenge the notion that suppression and IL-10 production by Tr1 cells are necessarily linked, but suggest that they are acquired early and late upon EOMES⁺Tr1-cell differentiation.

Conflict of interest

The authors declare no commercial or financial conflict of interest.