ICOS-Deficient Regulatory T Cells Can Prevent Spontaneous Autoimmunity but Are Impaired in Controlling Acute Inflammation

Inducible costimulator is a member of the CD28 family of T cell costimulatory receptors (1). Engagement of ICOS with its unique ligand, ICOS ligand (ICOSL), enhances TCR-mediated expansion of T cells and their differentiation into effector Th cells (2, 3). Despite similarities in structure between CD28 and ICOS, ICOS costimulation has indispensable roles in T cell immunity, in part due to the expression patterns of ICOS/ICOSL and the signal transduction pathways it can evoke (1, 4–7). The most remarkable role of ICOS is its ability to control germinal center reactions by promoting the differentiation of T follicular helper (Tfh) cells (8, 9). Thus, both in human patients and in mice, germline ICOS deficiency leads to reduced Tfh cell generation, germinal center reaction, and reduced Ab titers (8, 9).

Foxp3⁺ regulatory T (Treg) cells develop from CD4⁺ T cells in the thymus or in the periphery and suppress autoreactive T cells and inflammatory T cells (10). Thus, Treg cells play key roles in preventing autoimmunity and downregulating inflammation. Genetic modifications that damage the generation and/or suppressive function of Treg cells leads to spontaneous autoimmune symptoms, including multiorgan inflammation in humans and animal models. CD28 has been shown to be critical for the development of Treg cells from the thymus (2, 11), Treg homeostasis in the periphery (12, 13), and conversion of CD4 conventional T (Tcon) cells into Foxp3⁺ Treg cells (14–16). However, the roles of ICOS in Treg cell development, homeostasis, and function remain less clear.

Although ICOS was originally described as a costimulatory receptor acutely upregulated in effector T cells after activation (17), a portion of Treg cells constitutively expresses ICOS (1). Because ICOS is preferentially expressed in Ag-experienced Treg cells such as a CD44⁺CD62L⁻ subset, many studies have been using ICOS as one of the activation markers that are upregulated in activated Treg cells (18–20). Furthermore, several studies also suggested positive roles of ICOS in Treg cell proliferation, survival, migration, and suppressive functions (18, 21–24). However, these studies used either purified Treg cells stimulated in vitro (23, 24), Abs or recombinant proteins injected into mice to manipulate ICOS–ICOSL interactions (18), or mice lacking ICOS or ICOSL expression in all of the cells in the body (21, 22). Because each approach has its own limitations, the results need to be complemented with more precise experimental systems.

To assess the role of ICOS costimulation in Treg cell homeostasis and function without interference of ICOS deficiency in other cells, we generated a mouse model in which ICOS expression is specifically abrogated in Foxp3⁺ Treg cells. Through flow cytometric and single-cell transcriptome analyses, we demonstrate that the absence ICOS in Foxp3⁺ Treg cells does not heavily affect the activation status, suppressive capacity, and maintenance of Treg cell subpopulations. Consistently, mice born with ICOS-deficient Treg cells did not show any spontaneous autoimmune symptoms upon aging. In contrast, mice lacking ICOS-expressing Treg cells had more severe skin inflammation in an experimental model of atopic dermatitis. Single-cell transcriptome analysis of Treg cells in the draining lymph node (LN) suggests an impaired differentiation of a Treg cell subset equipped with highly effective anti-inflammatory functions.

Foxp3^YFP-Cre (JAX 016959) and C57BL/6 (JAX 00664) mice were purchased from The Jackson Laboratory and bred with other lines of mice to make composite mouse lines. Foxp3 reporter knock-in (B6.Foxp3GFP^ki) mice were provided by Alexander Rudensky and bred into the congenic background (CD45.1) for >10 generations. ICOS conditional knockout mice were generated in C57BL/6 background as described (25). ICOS germline knockout mice were backcrossed on the C57BL/6 background (8). All mouse lines had been backcrossed on the C57BL/6 background for >10 generations before breeding with other lines. Throughout the study, we used male Foxp3^YFP-Cre Icos^+/+ (ICOS wild-type [WT]) and Foxp3^YFP-Cre Icos^F/F (ICOS FC) mice confirming Treg cell–specific loss of ICOS expression by flow cytometry. All mice were housed and bred under specific pathogen-free conditions, and all of the experiments were performed in compliance with animal use protocols approved by the Institut de Recherches Cliniques de Montréal Animal Care Committee or McGill University.

For flow cytometry, single-cell suspensions of LNs and spleen were prepared by mechanical disruption via a 70-µm mesh filter (BD Biosciences) in PBS or staining buffer (PBS containing 1% BSA; Wisent). To prevent nonspecific blinding, anti-CD16/32 (Bio X Cell) was used before surface staining. For intracellular staining, cells were fixed and permeabilized with Fix/Perm buffer (Thermo Fisher Scientific) according to the manufacturer’s instruction. Abs for flow cytometry are anti-TCRβ (H57-597, BioLegend), anti-CD4 (RM4-5, Invitrogen), anti-CD8 (53-6.7, Invitrogen), anti-CD25 (PC61.5, Invitrogen), anti–CTLA-4 (UC10-4F10-11, BD Biosciences), anti–PD-1(J43, BD Biosciences), anti-GITR (YGITR 765, BioLegend), anti-FOXP3 (FJK-16s, Invitrogen), anti–T-bet (eBio4B10, Invitrogen), anti-ICOS (C398.4A, BioLegend), anti-CD103 (2E7, BioLegend), anti-CD44 (IM7, BD Biosciences), anti-CD62L (MEL-14, BioLegend), anti-Ki67 (16A8, BioLegend), anti-Bcl2 (3F11, BD Pharmingen), anti–IFN-γ (XMG1.2, Invitrogen), anti–TNF-α (MP6-XT22, Invitrogen), anti–IL-13 (eBio13A, Invitrogen), and anti–IL-17A (eBio17B7, eBioscience). Viability dye was from Thermo Fisher Scientific. Data were acquired using LSRFortessa (BD Biosciences) and analyzed using FlowJo v10 (BD Biosciences).

Mice were anesthetized by isoflurane (CDMV) and sensitized by topical application of 1% oxazolone (4-ethoxymethylene-2-phenyl-2-oxazolin-5-1, Sigma-Aldrich) in acetone/olive oil (4:1 v/v) on the shaved abdomen (50 µl). Five days later, the mice were challenged by topical application of 1% oxazolone on the left ear (20 µl). Mice were sacrificed at day 1 or 3 postchallenge for analysis. In some experiments, the severity of contact hypersensitivity was monitored for >16 d by measuring ear thickness. For flow cytometric analysis of T cells infiltrated in the skin, mouse ears were separated into dorsal and ventral halves, then minced and incubated in 1.5 ml of digestion solution (RPMI 1640 supplemented with 10% FBS, 200 µg/ml Liberase [Roche], and 40 µg/ml DNase I [Sigma-Aldrich]). The samples were incubated at 37°C for 45 min with agitation (270 rpm). The digested samples were strained through a 40-µm filter (Falcon) and washed with PBS for ex vivo FACS staining. In some experiments, the digested samples were incubated for 2 d after adding soluble anti-CD3 (1 µg/ml, 145-2C11, Bio X Cell), anti-CD28 (2 µg/ml, 37.51, Invitrogen), and mouse (m)IL-2 (100 U/ml, PeproTech) to expand skin-infiltrating T cells.

In vitro Treg cell suppression assays were performed by stimulating CFSE-labeled responder cells in the presence of Treg cells. CD4⁺ responder T cells were FACS sorted from ICOS WT mice based on YFP signal (YFP⁻), and YFP⁺ Treg cells were prepared either from ICOS WT mice or ICOS FC mice. After CFSE labeling, responder cells (5 × 10⁴) were seeded on 96-well plates mixed with titrated numbers of ICOS-deficient or WT Treg cells. T cell stimulation was achieved by adding anti-CD3 (1 µg/ml) in the presence of irradiated (900 cGy) splenic cells (2 × 10⁵) or isolated dendritic cells (5 × 10⁴). Dendritic cells were prepared using an EasySep mouse CD11c positive selection kit (STEMCELL Technologies). Proliferation of responder cells was assessed by flow cytometric analysis of CFSE dilution after 3 d of culture.

For Treg cell assays under various Th polarizing conditions, FACS-sorted CD4⁺Foxp3⁻ responder T cells from CD45.1⁺ B6.Foxp3^GFPki mice were labeled with CellTrace Violet (Thermo Fisher Scientific) and plated in 96-well plates (5 × 10⁴ per well) together with mitomycin C–treated feeder cells (2 × 10⁵ per well, CD4-depleted splenocytes using MACS beads, Miltenyi Biotec). For the Th0 condition, responder T cells were stimulated with soluble anti-CD3 (1 μg/ml) in the presence or absence of FACS-sorted CD4⁺Foxp3⁺ Treg cells (2.5 × 10⁴) from either Foxp3^YFP-Cre Icos^+/+ (CD45.2) or Foxp3^YFP-Cre Icos^F/F (CD45.2) mice for 72 h. To induce Th cell polarization, the following cytokines were added to the culture: mIL-12 (10 ng/ml, PeproTech) and mIL-18 (10 ng/ml; R&D Systems) for Th1; mIL-4 (10 ng/ml, BioLegend) and mIL-33 (10 ng/ml, R&D Systems) for Th2; mTGF-β (1 ng/ml, R&D Systems), mIL-6 (10 ng/ml, R&D Systems), and mIL-1β (50 ng/ml, R&D Systems) for Th17.

To measure cytokine-expressing cells, cells (3 × 10⁶) were seeded in 24-well plates in 1 ml of culture media (RPMI 1640 medium, 10% FBS, 2 mM l-glutamine, 0.05 mM 2-ME, 10 mM HEPES) and stimulated with PMA (50 ng/ml) and ionomycin (1 µg/ml) for 4 h at 37°C incubator in the presence of GolgiPlug (1 μl/ml, BD Biosciences). After stimulation, the cells were collected and analyzed by flow cytometry according to the standard intracellular staining protocol.

Organs were fixed in 10% neutral buffered formalin for 24 h at 4°C. The samples were washed in PBS, embedded in paraffin, and cut in 5-µm sections. Slides were stained with H&E and examined under a microscope to assess immune cell infiltration.

Single-cell suspensions were prepared from either spleens (for steady state) or cervical draining LNs (for oxazolone challenge on day 1) from ICOS WT and ICOS FC mice. Cells were stained with viability dye, anti-CD4, and anti-TCRβ. Live Treg cells (CD4⁺TCRβ⁺YFP⁺) were sorted using a BD FACSAria (BD Biosciences) with >95% purity. A total of 12,000 cells from ICOS WT and ICOS FC mice were prepared for the library. Libraries were generated at the Institut de Recherches Cliniques de Montréal Molecular Biology and Functional Genomics Core Facility using the following components from 10× Genomics: Chromium Next GEM Chip G Single Cell kit, Chromium Next GEM Single Cell 3′ GEM, Library & Gel Bead kit v3.1, and Chromium i7 Multiplex Kit. Sequencing was performed by Genome Québec using a NovaSeq 6000 (Illumina) with a flow cell SP PE28x91.

The alignment of the reads was performed using Cell Ranger 4.0.0 (from 10x Genomics) with the GRCm38.p6 (mm10) reference genome procured from Ensembl. The resulting expression matrix was loaded into R version 3.6.1 (from the R Foundation for Statistical Computing) to conduct analysis.

The expression matrices were stored into an R Seurat object available in the package Seurat v3.0 (26). For each condition (steady state and oxazolone), we filtered ICOS WT and ICOS FC samples by eliminating any cell that presented >10% mitochondrial RNA contamination as well as any cell with <200 unique genes expressed, and we merged the ICOS FC and ICOS WT samples in each condition. The expression matrix was then log normalized and scaled. The most differentially expressed genes within the samples were identified. As several cell cycle–related genes were identified, we used the CellCycleScoring function available in Seurat to score cell cycle phases. Afterwards, we scaled the data by regressing cell cycle score variables (phases S and G₂/M) to minimize the importance of cell cycle genes. We then proceeded with a dimensional reduction using a principal component analysis approach based on the 2000 most variable features. We selected the first 30 most important eigenvectors produced by the principal component analysis and constructed a shared nearest neighbor graph, and we used Modularity Optimizer version 1.3.0 (27) to identify seven clusters for the steady-state condition and eight clusters for the oxazolone condition. The cells were projected on a two-dimensional space using a uniform manifold approximation and projection (UMAP) method (L. McInnes, J. Healy, and J. Melville, manuscript posted on arXiv, DOI: 10.48550/arXiv.1802.03426). We then analyzed gene expression patterns across samples and conditions using Seurat functions FeaturePlot, VlnPlot, and DoHeatmap. The single-cell RNA sequencing datasets are available at Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/), accession no. GSE185254).

All data are presented as mean ± SEM. Because the datasets were normally distributed, we used the Student t test to judge statistical significance (Prism 9.0; GraphPad Software). A p value <0.05 was considered statistically significant.

To assess the role of ICOS in Treg cell homeostasis and function, we created mice that specifically lack ICOS expression in Foxp3⁺ Treg cells by generating ICOS FC mice. We confirmed that ICOS expression is selectively abrogated in CD4⁺Foxp3⁺ T cells (Fig. 1A). In the control ICOS WT mice, a small portion of non–Treg cell CD4⁺ Tcon cells expressed a low level of ICOS, whereas Treg cells expressed elevated levels of ICOS in steady state. In ICOS FC mice, the pattern of ICOS expression in Tcon cells remained unaltered, yet Foxp3⁺ Treg cells completely lost ICOS expression. This Treg-specific abrogation of the Icos gene in ICOS FC mice was confirmed under inflammatory conditions in which CD4⁺ and CD8⁺ effector T cells upregulated ICOS expression whereas Treg cells were unable to do so (Supplemental Fig. 1).

We observed a reduction in the frequency of Treg cells in the spleen and peripheral LNs in ICOS FC mice (Fig. 1B), but the numbers of Treg cells were not heavily affected in these lymphoid organs (Fig. 1C). Also, individual Treg cells expressed similar levels of Foxp3, Ki67, and Bcl2 (Fig. 1B, 1D, 1E). Importantly, there was no indication that Tcon cells were highly activated in ICOS FC mice at 8–10 wk of age (Fig. 1F). We reasoned that a deficit in Treg cell frequency combined with some potential functional defects of Treg cells in ICOS FC mice may lead to a breach of tolerance upon aging. However, we have not detected any visible sickness in ICOS FC mice until 1 y of age. Also, there was no immune cell infiltration observed in typical autoimmune targets such as liver, lung, pancreas, brain, eyes, and stomach at the age of 8 mo (Fig. 1G). These results indicate that ICOS-deficient Treg cells are capable of suppressing spontaneous activation of autoreactive T cells in ICOS FC mice.

ICOS was shown to be highly expressed in the CD44^hiCD62L^lo effector Treg (eTreg) cell population and possibly involved in maintaining the eTreg cell population (18). Compared with the CD44^loCD62L^hi central Treg (cTreg) cell population, eTreg cells express higher levels of CD103 and spend more time in the peripheral tissues. Also, eTreg cells express lower levels of CD25 compared with cTreg cells, and manipulation of ICOS–ICOSL interaction by putative ICOSL blocking Ab reduced the pool size of eTreg cells. However, it remains unclear whether altered ICOS signaling changed the expression levels of markers of eTreg cells or induced death of some of the eTreg cell population. To address the role of ICOS in the eTreg/cTreg balance in vivo, we examined the ratio of eTreg versus cTreg cells in ICOS FC mice. Compared to ICOS WT mice, there was no alteration in the eTreg-to-cTreg ratio (Fig. 2A). These results indicate that Treg-specific loss of ICOS expression does not drastically change the composition of the eTreg population.

Next, we examined the expression levels of Treg cell markers that have been shown to be important for Treg cell fitness and suppressive functions. As shown in (Fig. 2B, ICOS-deficient Treg cells expressed slightly elevated levels of CD25 and CTLA-4 that are known to be important for Treg cell suppressive functions (28, 29). In addition, PD-1 expression was lower in both splenic and LN Treg cells in ICOS FC mice, and GITR was slightly elevated in splenic Treg cells. In contrast, ICOS-deficient Treg cells expressed normal levels of CD103 and Helios. Despite these minor differences in Treg cell characteristics, ICOS-deficient Treg cells were equally good at suppressing polyclonal expansion of CD4⁺ Tcon cells in vitro (Fig. 2C). Furthermore, their suppressive capacities remained intact under Th1, Th2, and Th17 polarizing conditions (Supplemental Fig. 2).

Collectively, these data indicate that under steady state, ICOS-deficient Treg cells can be activated to form eTreg cell populations and are capable of suppressing conventional T cell expansion and polarization.

It has been shown that constitutive expression of CD25 is one of the key features of healthy Foxp3⁺ Treg cells (19, 30). A high level of CD25 appears to enhance their sensitivity to a low concentration of IL-2 facilitating p-STAT5–mediated Foxp3 expression (31). However, CD25⁻Foxp3⁺ Treg cell populations do exist both in lymphoid and nonlymphoid tissues (32). Because we saw an increase of CD25⁺ Treg cells in ICOS FC mice (Fig. 2B), we wondered whether there is a negative correlation between CD25 and ICOS expression levels in Treg cells. Indeed, CD25⁻ Treg cells expressed a 2-fold higher level of ICOS compared with their CD25⁺ Treg cell counterparts (Fig. 3A). We then asked whether elimination of ICOS in Treg cells has any impact on the CD25⁻ Treg cell population. Importantly, the ratios between CD25⁻ Treg cells and CD25⁺ Treg cells were reduced 2- to 4-fold in the spleen and LN of ICOS FC mice compared with ICOS WT control mice (Fig. 3B). These results raise two possibilities. First, ICOS costimulation suppresses CD25 expression itself directly or indirectly. If this is the case, ICOS-deficient Treg cells would quickly upregulate CD25. Alternatively, there might be a CD25⁻ Treg cell subset whose maintenance is highly dependent on ICOS costimulation. In this scenario, a group of cells sharing gene signatures may disappear in ICOS FC mice.

Flow cytometric analysis of ICOS-deficient Treg cells identified only a few minor changes in the Treg cell characteristics. To confirm these findings and further explore potential homeostatic changes established at the level of Treg cell subpopulations, we analyzed single-cell transcriptomes using splenic Treg cells sorted from an ICOS WT and ICOS FC mouse.

All of the cells (12,000 cells with 6,000 cells per genotype) were split into seven clusters based on the similarity of gene expression patterns (Fig. 4A). Importantly, none of the clusters was unique to ICOS WT or ICOS FC mice (Fig. 4A). We further compared the relative expression levels of key signature genes that are differentially expressed in each cluster using a heatmap. Although each cluster had two to five signature genes that distinguished them from other clusters, the expression levels of each markers had only marginal differences between genotypes (Fig. 4B). These results suggest that ICOS does not provide any indispensable function to form and maintain a particular subset of Treg cells.

The Icos gene is highly expressed in clusters 1 and 2 (Fig. 4C). These clusters express low levels of Sell (CD62L) and high levels of Cd44, indicating that ICOS expression is associated with the expression of activation markers for Treg cells. However, the absence of ICOS did not drastically change Sell and Cd44 expression levels, consistent with our FACS data (Fig. 2A). Also consistent with FACS data, clusters 1 and 2 had higher Ctla4 in ICOS FC mice. Lastly, clusters 1 and 2 had lower Il2ra expression compared with ICOS^lo cluster 0. Despite these correlations, ICOS deficiency did not abrogate any clusters seen in ICOS WT control, nor did it generate new clusters.

Overall, these single-cell transcriptome analyses confirmed only subtle changes in Treg cell characteristics in the steady-state splenic Treg cell population and support the view that ICOS-deficient Treg cells preserve intact capacities to prevent spontaneous autoimmunity.

Most in-depth studies regarding the role of ICOS in Treg cells have been performed in BDC2.5 TCR transgenic mice in the NOD background (23, 33, 34). In the BDC2.5-NOD model, autoreactive pathogenic T cells accumulate in a prediabetic insulitis lesion, but ICOS-expressing Treg cells can delay onset of overt diabetes (23, 33). One of the roles of ICOS in the early phase of T cell priming in the draining LNs is to promote maturation of a Th1-like Treg cell population. This unique population expressed a high level of T-bet and CXCR3 in response to IFN-γ produced by inflammatory Th1 cells (34). To test whether Treg cell–specific ICOS deficiency hinders generation of this Th1-Treg population, we decide to use an oxazolone-mediated contact hypersensitivity model, a mouse model of atopic dermatitis driven by IFN-γ (35). In this context, an early study showed that a putative ICOS agonist (B7RP1-Fc, soluble rICOSL fused with human Fc fragment) increased skin inflammation in this model, although the authors have not considered the impact of B7RP1-Fc on Treg cells (36).

Our data showed that ICOS FC mice had more severe skin inflammation as shown by the increase in ear thickness (Fig. 5A) and immune cell infiltration in histological images (Fig. 5B). This elevated skin inflammation was highly correlated with a reduced number of Treg cells in the skin and an increased number of inflammatory CD4⁺ cells (and to some extent CD8⁺ T cells) that express IFN-γ or TNF-α at day 1 postchallenge (Fig. 5C, 5D). These skin phenotypes are well reflected in the reduced number of Treg cells and elevated inflammatory T cells in the draining LN (Fig. 5E, 5F). We tried to detect T-bet⁺CXCR3⁺ Th1-Treg cells under this condition in the draining LN. However, the number of cells in this population was too small to allow any meaningful comparisons between genotypes. We presume that Th1-Treg cells continuously differentiate and migrate to the inflamed skin. To support this idea, we expanded skin-infiltrated T cells in vitro for 2 d and assessed the number of T-bet⁺ Treg cells. Indeed, inflamed ear skin (at day 1 postchallenge) in ICOS FC mice had a substantially reduced number of T-bet⁺ Treg cells compared with the ICOS WT counterpart with a concomitant increase of IFN-γ⁺CD4⁺ Tcon cells (Supplemental Fig. 3).

In germline ICOS knockout mice, we did not find significant differences in ear swelling compared with control mice (Supplemental Fig. 4A). This is presumably due to reduced activation of both inflammatory CD4⁺ T cells and Treg cells (Supplemental Fig. 4B–E).

Taken together, ICOS FC mice had reduced ability to control Th1-mediated skin inflammation with reduced Treg cell numbers in the draining LN and inflamed tissue.

To gain more insights into dynamic changes in the gene expression profile of Treg cells responding to oxazolone-induced inflammation, we performed single-cell transcriptome analysis. The collection of ∼12,000 cells was split into eight clusters and there were no unique clusters appearing or disappearing in ICOS FC Treg cells (Fig. 6A). The most drastic changes in the gene expression profile happened in cells belonging to cluster 0 as shown in the heatmap (Fig. 6B). ICOS FC Treg cells had reduced levels of Cd44, Gzmb, and S100a4, but increased levels of Sell, suggesting reduced activation/maturation levels (Fig. 6C, 6D). Notably, we found a small number of Treg cells belonging to cluster 0 that coexpress genes representing Th1-Treg: Tbx21 (T-bet) and Cxcr3 (Fig. 6E, red arrows). These cells also expressed IL-10 and CTLA-4, known effector molecules with suppressive functions (Fig. 6E). Importantly, the ICOS FC Treg cell population had fewer cells in this specific UMAP area, suggesting that there could be a defect leading to delayed or impaired maturation of Th1-Treg–like cells in the absence of ICOS costimulation. This may explain why there were reduced numbers of T-bet⁺ Treg cells in the inflamed ear skin (Supplemental Fig. 3).

Thus, our data suggest that ICOS promotes maturation of a T-bet⁺CXCR3⁺ Th1-Treg subset that has the capacity to migrate into inflamed tissues and suppress inflammatory T cells.

In this study, we developed an ICOS FC mouse model to most precisely address in vivo function of ICOS costimulation with minimal complications that may arise from ICOS deficiency in other immune cells. In contrast with some of the data in the literature, we could not detect any major differences in terms of number, subset composition, and suppressive function of Treg cells in the C57BL/6 background at homeostasis. Consequently, we could not detect any spontaneous autoimmune manifestations in ICOS FC mice. However, ICOS FC mice responded more vigorously to experimentally induced skin inflammation. We provide evidence that this is probably due to an impaired maturation of Treg cells in the draining LN hampering their migration to the inflammatory sites.

Minimal alterations in Treg behavior seen in ICOS FC mice in homeostasis are in contrast with some of the data in literature. First, it has been shown that ICOS^hi Treg cells are more proliferative and more resistant to cell death in vitro (18). Perhaps this is due to supraphysiological TCR stimulation used in the experiments. Second, injection of recombinant ICOSL–human Fc fusion protein (B7RP1-Fc) in C57BL/6 mice undergoing oxazolone-mediated skin inflammation worsened the disease (36). The authors proposed that this is due to putative agonistic actions of B7RP1-Fc upon ligation of ICOS on effector T cells. However, our data suggest another possibility: B7RP1-Fc may have interfered with ICOS–ICOSL interaction in an antagonistic way, and the main target of B7RP-Fc could have been Treg cells. Lastly, eTreg cells expressing a high level of ICOS were diminished in mice that have been treated with multiple rounds of ICOSL blocking Ab (18). The authors provided evidence that eTreg cells are more prone to apoptosis and highly dependent on ongoing ICOS signaling for survival. However, our data show that ICOS-deficient Treg cells maintain a normal eTreg cell compartment. Thus, it appears that chronic absence of ICOS in an eTreg cell population can be compensated by other mechanisms.

Although CD28 and ICOS belong to the same family of T cell costimulatory receptors, they have several differences including those in their signal transduction mechanisms (1). In terms of Treg cell homeostasis, CD28 appears to play a more dominant role than ICOS. CD28 deficiency leads to a drastic numerical deficit in peripheral Treg cell populations (13). In contrast, we detected only a marginal defect in ICOS FC mice. Considering that ICOS has a limited role in the maturation of Treg cells during inflammation, combined treatment of CD28 and ICOS may confer synergistic effects.

Importantly, note that ICOS has dual stimulatory roles for effector (or pathological) T cells and Treg cells. For example, in our contact hypersensitivity model, Treg cell–specific ICOS deficiency worsened the disease, whereas germline ICOS deficiency did not alter the disease outcome. Similar observations were made in autoimmune diabetes models in a NOD background (37, 38). Furthermore, ICOS is essential for the generation of Tfh cells (39) and follicular Treg cells (40), with effector T cell and Treg cell subsets controlling germinal center reactions. In our recent study, we found that ICOS FC mice have impaired follicular Treg cell differentiation from Treg cells and are more prone to produce autoantibodies after immunization or infection (V. Panneton, B.C. Mindt, Y. Bouklouch, J. Chang, M. Witalis, J. Li, A. Stancescu, A. Bouchard, J.E. Bradley, T.D. Randall, et al., manuscript posed on bioRxiv, DOI: 10.1101/2021.09.14.460384). This dual impact of ICOS signaling in the effector and Treg cell populations and the immunological contexts should be considered for clinical application of agonists or antagonists of ICOS–ICOSL pathways.

Because Foxp3 expression starts early during thymic development, the Icos gene is lost in Treg cells during its early stage of Treg cell development. However, in a separate work, we found that ICOS is not highly expressed in de novo thymic Treg cells in WT mice (41). The absence of spontaneous autoimmunity in ICOS FC mice also supports the view that ICOS does not play a critical role in the thymic selection of Treg cells. In contrast, lack of ICOS signaling partially impaired negative selection of CD8⁺ thymocytes (41).

In summary, we confirmed that the main role of ICOS-expressing Treg cells is to fine-tune the maturation of peripheral Treg cells during inflammation. This implies that, if applied at the right timing with preferential delivery to Treg cells, ICOS agonists could be beneficial to control Th1-driven inflammation.

We thank Dominic Filion, Manon Laprise, Julie Lord, Éric Massicotte, and Simone Terouz for technical assistance, and Viviane Beaulieu for animal husbandry.

The authors have no financial conflicts of interest.