Oral antigen exposure under costimulation blockade induces Treg cells to establish immune tolerance

Oral tolerance is antigen-specific systemic immunological unresponsiveness induced by intestinal exposure to antigens derived from ingested food or commensal microbes (Chase, 1946; Cerovic et al., 2024). It is effective in preventing autoimmune diseases (e.g., arthritis, diabetes, and multiple sclerosis) and allergies (e.g., food allergy) in animal models (Rezende and Weiner, 2022). In humans, early exposure to peanut protein is reportedly capable of preventing peanut allergy in children (Du Toit et al., 2015; Du Toit et al., 2024). However, it has not been significantly successful in humans to induce oral tolerance for treating ongoing immunological diseases such as chronic allergy and autoimmune disease (Faria and Weiner, 2005). It is therefore a long outstanding issue to determine how harmful immune responses that have been already triggered and chronically progressing can be halted and cured by inducing antigen-specific oral tolerance (Mowat, 2018).

Naturally occurring CD4⁺ regulatory T (nTreg) cells expressing the transcription factor Foxp3, including thymus-derived Treg (tTreg) cells and peripherally derived Treg (pTreg) cells, are indispensable for the induction and maintenance of immunological tolerance, including oral tolerance, and instrumental in controlling immunological diseases (Sakaguchi et al., 2020; Mucida et al., 2005; Josefowicz et al., 2012; Tanoue et al., 2016; Hong et al., 2022). It remains obscure, however, how antigen-specific pTreg cells are generated and how they establish and maintain oral tolerance (Cerovic et al., 2024). It has been suggested that following antigen transportation through the intestinal epithelial cells, antigen-presenting cells (APCs), conventional dendritic cells (cDCs) in particular, play a pivotal role in the generation of antigen cognate pTreg cells in mesenteric lymph nodes (mLNs) and lymphoid tissues in the intestine (Mazzini et al., 2014; Hadis et al., 2011; Esterházy et al., 2016). Intestinal cDCs can efficiently convert TGF-β into its active form, and vitamin A into retinoic acid (RA); TGF-β and RA synergistically induce intestinal pTreg cells upon antigen stimulation (Mucida et al., 2007; Sun et al., 2007; Coombes et al., 2007). These pTreg-inducing cDCs appear to be immature DCs with low expression of the costimulatory molecules CD80 and CD86 (Morelli and Thomson, 2007; Benson et al., 2007; Lutz et al., 2021). However, the roles of CD28-mediated costimulation for pTreg generation and oral tolerance have not been well characterized. Addressing the issue, we have recently shown that in vitro antigen stimulation of CD4⁺ conventional T (Tconv) (i.e., CD4⁺CD25⁻Foxp3⁻) cells in the presence of TGF-β and IL-2 but without CD28 signal is able to efficiently induce Foxp3 Treg cells equipped with Treg-specific epigenetic changes, hence Treg cell-lineage stability, and that such functionally stable in vitro–induced Treg (iTreg) cells can be generated not only from naïve (i.e., CD4⁺CD25⁻Foxp3⁻CD62L^hiCD44^lo) but also from effector/memory (i.e., CD4⁺CD25⁻Foxp3⁻CD62L^loCD44^hi) Tconv cells (Mikami et al., 2020; Mikami et al., 2025). The question then arises whether such in vitro generation of iTreg cells functionally similar to nTreg cells shares a common developmental basis with in vivo pTreg generation, especially in the context of oral tolerance.

Here, we have attempted to determine whether a functionally and phenotypically distinct type of pTreg cells is generated upon antigen exposure via the oral route, how they survive upon antigen stimulation and maintain their function to sustain oral tolerance, and how they can be exploited to establish robust oral tolerance not only in antigen-nonsensitized naïve mice but also in those that have been already antigen-primed at other sites of the body.

To analyze the development of pTreg cells by distinguishing them from tTreg cells, we used OVA-specific DO11.10 (DO) TCR transgenic mice that were made Rag2-deficient and expressing Foxp3-eGFP reporter gene. DO Rag2KO Foxp3-eGFP mice thus produced are hereafter abbreviated as DORe mice. Foxp3⁺ (i.e., GFP⁺) cells were absent in the thymus, very few in the spleen and lymph nodes including mLNs in untreated DORe mice (see also below). To determine the effects of OVA feeding on anti-OVA T-cell responses in DORe mice, we kept feeding them for 4 wk with the diet containing egg white powder (EWP) as 10% of the protein content or with control normal chow (NC), and assessed ear swelling after four times of tape stripping (TS) and papain/OVA treatment on the right ear and control TS/papain treatment on the left ear (Fig. 1 A). Compared with NC-fed mice, which showed in the right ears severe ear swelling and histologically evident inflammation in the epidermis and dermis with abundant cellular infiltration, EWP-fed mice exhibited much lower degrees of ear swelling and histologically evident inflammation (Fig. 1, B and C). They had much smaller numbers of DO CD4⁺ T cells (Fig. S1 A) and much lower ratios of IFN-γ–, IL-4–, or IL-17A–producing DO CD4⁺ T cells in the draining lymph nodes (dLNs) (Fig. S1 B). They developed pTreg cells in mLNs, dLNs, and the right ear in much higher ratios compared with NC-fed mice (Fig. 1 D). Treg cells were found to be present in the spleen and also in the control left ear as well at high ratios, although in small numbers (Fig. S1 C and D), with no significant ear swelling or inflammation in EWP-fed mice (Fig. S1. D and E). In addition, diphtheria toxin (DTx) treatment of EWP-fed DO11.10 Rag2KO Foxp3-DTR-GFP (DORF) mice expressing the diphtheria toxin receptor (DTR) specifically in pTreg cells (Fig. 1 E) depleted them in the OVA-challenged ear and dLNs. It evoked severe ear swelling and inflammation (Fig. 1, F and G) with significant increases of inflammatory cytokine-producing T cells, especially in the ear and dLNs (Fig. S1, F and G).

To determine whether induction of oral tolerance would require continuous antigen exposure, we ceased EWP feeding for 4 or 8 wk before the OVA challenge to the ear (Fig. 1 H). Despite prior EWP feeding for 4 wk, the feeding-suspended DORe mice showed ear inflammation as severe as control NC-fed mice (Fig. 1 I), reduced ratios of pTreg cells, and increased cytokine-producing T cells in the ear tissue and dLNs when compared to DORe mice that had been kept EWP-fed until the ear testing (Fig. 1 J; and Fig. S1, H and I).

Next, to examine whether antigen-specific oral tolerance could be induced in precedingly antigen-primed mice, we first sensitized DORe mice by TS/papain/OVA treatment and then fed the mice with EWP or NC (Fig. 1 K). Both groups developed similar degrees of ear swelling and inflammation (Fig. 1 L). The frequencies of pTreg cells and Tconv cells producing inflammatory cytokines were comparable in mLNs, dLNs, and the ear tissues in both groups (Fig. 1 M; and Fig. S1, J and K). Moreover, the EWP-fed group showed significant body weight loss (Fig. 1 N), indicating elicitation of systemic inflammation by oral antigen feeding in antigen-sensitized mice.

Collectively, continuous feeding of the diet containing protein antigen establishes antigen-specific oral tolerance by generating pTreg cells from antigen-specific Tconv cells. The generated pTreg cells migrate through the whole body to various tissues. However, prior antigen sensitization at other sites or discontinuity of antigen feeding hampers oral tolerance induction and may even evoke an aberrant immune response upon antigen feeding.

To examine how the quality and quantity of antigens contained in the diet would affect the development and immunological characteristics of pTreg cells in oral tolerance, we fed DORe mice with the chow containing various concentrations of EWP, antigen-free AIN-93G chow containing only casein as protein (Reeves et al., 1993), or NC, starting from 2 wk of age before weaning. EWP feeding increased the frequency of pTreg cells in mLNs and the small intestine lamina propria (SI-LP) in an antigen dose-dependent manner, even if antigen-containing chow was taken ad libitum (Fig. 2, A and B). High doses of EWP increased CD25 expression by the pTreg cells induced in SI-LP (Fig. 2, A and C), indicating their enhanced activation status (Miyara et al., 2009). The EWP-induced pTreg cells were also maintained as CD44^hiCD62L^lo effector/memory type T cells in SI-LP (Fig. S2, A and B). These pTregs exhibited Treg-specific epigenetic modifications, including DNA hypomethylation at Treg signature gene loci, such as Foxp3-conserved noncoding sequence 2 (CNS2) region and other regions in Ikzf2, Ikzf4, and Ctla4 loci (Ohkura et al., 2012) (Fig. 2 D). By chromatin immunoprecipitation sequencing (ChIP-seq) of the genomic regions associated with histone H3 Lys27 acetylation (H3K27ac) as an indicator of active enhancer status, differential peak analysis revealed that Treg-specific enhancer regions were activated in pTreg cells from EWP-fed DORe mice, whereas Tconv-specific enhancer regions were not, as in nTreg cells in WT mice (Kitagawa et al., 2017; Kawakami et al., 2021; Dikiy et al., 2021) (Fig. 2 E). The pTreg cells thus generated had activated Treg-specific super-enhancers at the Foxp3 gene locus and other Treg signature gene loci including Ikzf2, Ikzf4, Ctla4, and Il2ra as observed in WT-nTreg cells (Kitagawa et al., 2017) (Fig. 2, F and G). Assay for transposase-accessible chromatin using sequencing (ATAC-seq) further confirmed chromatin remodeling also in these loci (Fig. 2, F and G).

In the experiments in Fig. 2 A, we noted that NC-fed mice also possessed detectable numbers of pTreg cells in mLNs and SI-LP in particular, at higher frequencies than antigen-free– or 0% EWP chow-fed mice (Fig. 2, A and B). To analyze the finding, we compared pTreg development in DORe and WT mice after NC feeding, and found that the former developed Foxp3⁺ cells at higher ratios than the latter in SI-LP, in the large intestine lamina propria (LI-LP), and among small intestine intraepithelial (SI-IE) cells, but not in the spleen and LNs (Fig. S2, C and D). The Foxp3⁺ cells in NC-fed DORe mice exhibited DNA hypomethylation at Treg-specific demethylation regions in Treg signature genes (Fig. S2 F), and expressed Helios and Neuropilin-1 (Fig. S2 E), at comparable levels as WT-nTreg cells. The results suggested that an antigen(s) cross-reactive with OVA in NC might activate DO Tconv cells to differentiate into functionally stable pTreg cells, although the number of pTreg cells thus generated was not sufficient to induce OVA-specific oral tolerance (Fig. 1, B and C). After depletion of these pTreg cells in NC-fed DORF mice by DTx administration (Fig. S2, G and H), EWP feeding efficiently induced pTreg cells, suggesting that the pTreg cells were derived from non-Treg cells upon EWP antigen stimulation, rather than from selective expansion or survival of the preexisting pTreg cells (Fig. S2 I).

Taken together, these results indicate that the diet containing cognate antigen drives Tconv cells to differentiate into activated pTreg cells in the intestine, and enables them to acquire Treg-specific epigenetic changes, hence Treg-lineage stability similar to that of nTreg cells. In addition, the quantity of pTreg cells is dependent on the amount of the antigen contained in the diet; thus, more than a certain amount of the antigen is required for generating pTreg cells quantitatively sufficient to establish oral tolerance.

We next attempted to characterize the dynamics of the differentiation/activation of pTreg and Tconv cells in EWP-fed DORe mice using single-cell RNA sequencing (scRNA-seq). Analysis of preprocessed 6,449 cells from four hash-tagged and pooled populations of DO⁺ CD4⁺ T cells (i.e., mLN or SI-LP cells in DORe mice NC- or 10% EWP-fed for 4 wk) identified 13 clusters by the expression of T-lineage signature genes (Fig. 3 A and Fig. S3 A). EWP feeding resulted in an increase of Cluster 8 (circulating 2), which was Myb^hi (indicative of augmented T cell survival and exhaustion) (Yuan et al., 2010; Dias et al., 2017; Tsui et al., 2022), reductions of Clusters 4 (Teff-mix) and 11 (Th1), and conversion of Treg population from Cluster 5 to Cluster 3 in SI-LP (Fig. 3 B). Compared with Cluster 5, Cluster 3 pTreg cells having increased after EWP feeding highly expressed activation-related genes (Tnfrsf9, Tnfrsf4, Tigit, and Tnfrsf18), Th2-type transcription factor (Gata3), and tissue repairing–related molecules (Areg and Penk) (Fig. 3 C). Supporting these results, gene ontology analysis between Cluster 3 and 5 SI-LP pTreg cells revealed that Cluster 3 pTreg cells expressed a highly activated and proliferative T cell phenotype compared with Cluster 5 pTreg cells (Fig. 3 D). To analyze further the kinetics of antigen-specific T cells in SI-LP, we reclustered Clusters 3, 4, 5, and 11 based on commonalities and newly identified 8 clusters (Fig. 3, E–G and Fig. S3 B). Together with pseudotime analysis (Fig. 3 G), the analysis suggested that among Foxp3⁺ cells (Clusters 0, 1, 4, 6, and 7), Clusters 0 and 7 became activated and differentiated upon EWP stimulation into Clusters 1 and 4, which expressed Klrg1, Areg, and Penk. Among Tconv cells (Clusters 2, 3, and 5), effector T-cell populations with Th1 and Th17 gene signatures appeared to diminish upon EWP feeding (Fig. 3, E and F; and Fig. S3 B). Mcl1, an anti-apoptotic gene, was relatively high in Sell⁺ (i.e., CD62L^hi) Foxp3⁺ cells (Cluster 7); Bcl2 was high in Th1 cells (Cluster 5) (Fig. S3 C).

We assessed protein expression to validate these findings by comparing EWP- and NC-fed SI-LP pTreg cells (Fig. 3, H–K). The former showed significantly higher expression of 4-1BB (Tnfrsf9), OX40 (Tnfrsf4), and TIGIT (Tigit), in good correlations with their mRNA expression levels (Fig. 3, H and I). EWP-fed pTreg cells also contained increased proportions of T-bet⁺, GATA3⁺, or RORγt⁺ pTreg cells compared with NC-fed pTreg cells (Fig. 3 J). Furthermore, continuous EWP feeding induced Ki-67⁺ and KLRG1⁺ pTreg cells (i.e., highly proliferative and differentiated pTreg cells), which were significantly reduced in ratio after cessation of the feeding for 4 wk (Fig. 3 K).

We also conducted bulk mRNA-seq to assess the characteristics of pTreg cells in lymphoid and nonlymphoid tissues in EWP-fed mice. pTreg cells in mLNs and skin dLNs possessed more common features compared with those in SI-LP or the ear (Fig. 3 L), suggesting that pTreg cells generated in the intestine underwent tissue-specific adaptation and acquired a tissue-specific Treg phenotype and function distinct from those of LN pTreg cells, as observed in scRNA-seq (Fig. S3 D) (Miragaia et al., 2019). For example, SI-LP and ear-tissue pTreg cells (corresponding to Clusters 0, 1, 4, 6, and 7 in Fig. 3 E) were distinct from LN pTreg cells in the expression of apoptosis-related molecules (Fig. S3 E), for example, their lower expression of Bcl2, hence being more prone to die by apoptosis after their antigen-specific activation (Fig. 3 M). Further, among chemokine receptors controlling tissue migration, SI-LP and ear-tissue pTreg cells distinctly expressed CCR9 and CCR5, respectively, in both the mRNA and protein levels (Fig. 3, N and O; and Fig. S3 F).

Collectively, these results demonstrate that EWP feeding generates and sustains pTreg cells that can differentiate into tissue-specific effector-type Treg cells that suppress the differentiation/activation of effector Tconv cells.

We next searched for specific markers that could differentiate EWP-induced intestinal pTreg cells from other Treg cells, particularly from tTreg cells, by preparing mixed bone marrow chimeric (BMC) mice in which DORe BM-derived pTreg cells and WT (Thy1.1-eFox) mouse BM-derived nTreg cells developed in the same environment (Fig. 4 A). While NC or 10% EWP feeding for 4 wk from immediately after BM transfer equally generated WT-nTreg cells, only EWP feeding generated DORe BM-derived pTreg cells in mLNs, SI-LP, and SI-IE, and to lesser extents in peripheral LN and spleen, but not in the thymus (Fig. 4, B–D). Bulk mRNA-seq of DORe-pTreg cells and WT-nTreg cells in BMC mice having been EWP-fed for 8 wk revealed the genes that were differentially expressed between the two Treg populations (Fig. 4 E). Among such genes normalized for transcript per million in the expression level, DORe-pTreg cells showed much higher expression of Cd101 and Gpr83 and lower expression of Icos when compared to WT-nTreg cells from NC- or EWP-fed BMC mice (Fig. 4 F). Regarding previously reported tTreg markers (Thornton et al., 2010; Weiss et al., 2012), Ikzf2 (Helios) was low, whereas Nrp1 (Neuropilin-1) was high, in DORe-pTreg cells (Fig. 4 G). Flow cytometry analysis further confirmed that DORe-pTreg cells strongly expressed the CD101 protein in BMC mice (Fig. 4 H). Similarly, in NC-fed WT (eFox) and DORe mice (Fig. S2), ∼80% of DORe-pTreg cells and ∼25% of WT-nTreg cells expressed CD101, while Foxp3⁻ Tconv cells scarcely expressed the molecule (Fig. 4 I).

To examine CD101 expression on Foxp3⁺ cells in a more physiological condition, DORe-derived naïve T cells were transferred into WT mice, which were then fed with EWP (Fig. 5 A). Detectable numbers of DORe-pTreg cells were found to be induced in mLNs (Fig. 5, B and C). They highly expressed CD101 and Nrp-1 (Fig. 5 D).

In the thymus of normal mice, CD101 was expressed by a fraction of DN1 (CD4/CD8 double-negative CD44⁺CD25⁻) thymocytes, losing its expression along differentiation and again becoming expressed at the CD25⁻Foxp3⁺CD4SP Treg precursor stage, with upregulation at mature CD25⁺Foxp3⁺ Treg cells (Fig. 5, E and F). In the periphery, CD101 expression was limited to Foxp3⁺ T cells, with higher expression by CD44⁺CD62L⁻ effector Treg cells compared with naïve ones (Fig. 5, G and H). In addition, CD101 expression by nTreg cells was well correlated with the expression of activation-associated molecules such as Nrp-1, GITR, CD69, CD103, and CD44, especially in mLNs (Fig. 5 I).

Taken together, these results indicate that intestinal pTreg cells generated by antigen feeding are distinct from tTreg cells in the expression levels of certain genes, especially CD101.

Based on the above results, we investigated how functionally stable iTreg cells similar to CD101⁺ intestinal pTreg cells induced by antigen feeding could be generated in vitro from Tconv cells. Previous studies have shown that iTreg cells can be produced in vitro by anti-TCR stimulation with plate-bound anti-CD3 mAb in the presence of TGF-β, IL-2, and agonistic anti-CD28 mAb (Zheng et al., 2002; Chen et al., 2003), and that in the presence of IL-2, iTreg generation was more efficient in the absence of agonistic anti-CD28 mAb (Mikami et al., 2020; Benson et al., 2007). Consistent with this, we found that iTreg cells generated without CD28 signal indeed expressed CD101 at higher levels than those produced with CD28 signal (Fig. 6, A and B). Moreover, CD28-deficient naïve Tconv cells more efficiently differentiated into iTreg cells with higher expression of CD101 compared with CD28-intact naïve Tconv cells; agonistic anti-CD28 mAb significantly hampered the generation of CD101⁺ iTreg cells from the latter, but not from the former (Fig. 6, C and D). Furthermore, TGF-β dose–dependent iTreg induction revealed that TGF-β significantly enhanced CD101 expression in a dose-dependent manner only in iTreg cells generated without CD28 signal (Fig. S4 A).

Next, to simulate in vitro the condition of intestinal pTreg generation in EWP-fed DORe mice, we examined the effects of CD80/CD86 blockade with CTLA-4-immunoglobulin (CTLA4-Ig) fusion protein on iTreg generation from naïve DORe-Tconv cells stimulated with OVA in the presence of splenic DCs (Fig. S4 B). CTLA4-Ig did not inhibit proliferative differentiation of Foxp3⁺ cells even at high doses (Fig. S4 C), increased CD101 expression levels in an IL-2 dependent fashion (Fig. S4 D), and conferred Foxp3-CNS2 hypomethylation on the generated iTreg cells (Fig. S4 E). With cDC1 and cDC2 cells prepared from SI-LP (Fig. S4 F), CTLA4-Ig in combination with all-trans retinoic acid (ATRA) significantly increased the generation of Foxp3⁺ cells and enhanced their expression of CD101 on both types of DCs (Fig. 6, E and F). CTLA4-Ig together with ATRA increased iTreg generation and CD101 expression even in the presence of various inflammatory cytokines (Fig. 6 G). Furthermore, with naïve or effector/memory Tconv cells from dLNs of OVA-immunized DORe mice (Fig. 6 H), CTLA4-Ig effectively augmented iTreg generation and their CD101 expression with naïve Tconv cells and, to lesser degrees, with effector/memory Tconv cells (Fig. 6, I and J).

Taken together, blockade of CD80/CD86 did not hamper iTreg generation with TGF-β, IL-2, and antigen stimulation, rather enhancing the generation of iTreg cells, CD101⁺ ones in particular. Thus, CD28 signaling is dispensable for iTreg generation, especially in the presence of IL-2. CD101 expression could be attributed to the reduction of CD28 costimulatory signal in iTreg generation.

The above result that CTLA4-Ig enabled iTreg cell induction from antigen-primed Tconv cells prompted us to determine whether in vivo CTLA4-Ig treatment would be able to establish oral tolerance even in precedingly antigen-primed mice (Fig. 7 A). When DORe mice were OVA-sensitized at the right ear, injected intraperitoneally with CTLA4-Ig at various doses or saline, and then fed with EWP chow for 4 days, the right ear swelling gradually subsided during the feeding whether the mice were treated with CTLA4-Ig or not (Fig. S5 A). However, mice treated with saline or low-dose (2 µg) CTLA4-Ig showed significant weight loss during EWP feeding (Fig. S5 B), contrasting with no weight loss in the mice CTLA4-Ig treated at high (10 or 50 µg) doses. The weight loss recovered after cessation of the feeding (Fig. 7 B). When these mice were OVA-challenged to the left ear after 4-day EWP feeding, those treated with high-dose (10 or 50 µg) CTLA4-Ig did not show ear swelling or succumb to histologically evident inflammation, contrasting with other groups suffering from severe ear swelling and dermatitis (Fig. 7 C and Fig. S5 C). Notably, contrasting with the CTLA4-Ig/EWP-treated mice exhibiting no ear swelling, the CTLA4-Ig/NC-treated mice developed severe ear swelling and dermatitis. The prevention of ear swelling was also observed with the higher antigen dose (30% EWP-containing chow), while the lower antigen dose (1% EWP) displayed a reduced preventive effect (Fig. S5, D–F). These results taken together indicated that both CTLA4-Ig treatment and antigen feeding were required for inducing the unresponsiveness.

There were little differences among these mouse groups in the ratio of Foxp3 cells on days 15 and 33, with a tendency of Treg increase in the CTLA4-Ig–treated mice (Fig. 7, D and E). The ratios of cytokine-secreting cells among DO Tconv cells were not significantly different among the mouse groups at day 15, except significant reduction of the ratio of IL-17⁺ cells in dLNs of the CTLA4-Ig–treated groups compared with the saline-treated or NC-fed mice (Fig. 7 F). IFN-γ⁺ cells and IL-17⁺ cells in the former were significantly lower compared with the latter at day 33 (Fig. 7 G). The results suggested that the high ratios of Foxp3⁺ cells in the low-dose CTLA4-Ig– or saline-treated group could be attributed to their secondary expansion due to inflammation.

The pTreg cells having developed in EWP-fed mice, whether treated with CTLA4-Ig or saline, showed Treg-specific DNA demethylation at the Foxp3 CNS2 region, suggesting their functional stability as Treg cells (Fig. S5 G). In contrast to no significant differences in total Foxp3⁺ cells in the ear or dLNs (Fig. 7, D and E), CD101 expression by Foxp3⁺ cells was significantly upregulated and the ratio of CD101⁺Foxp3⁺ cells increased at day 15 in the right ear and its dLNs of the CTLA4-Ig–treated mice depending on the dose of CTLA4-Ig, similarly in the left ear and its dLNs at day 33 (Fig. 7, H and I). Depletion of Foxp3⁺ cells by DTx administration on days 15 and 18 in CTLA4-Ig–treated and then EWP-fed DORF mice resulted in an exacerbated ear inflammation, a decrease in pTreg cells in the inflamed ear tissue, and an increase in IL-4⁺ T cells in dLNs (Fig. 7 J; and Fig. S5, H and I). CD101⁺ Foxp3⁺ cells were also significantly decreased in the inflamed ear tissue, dLNs, and mLNs (Fig. 7 K).

These results collectively demonstrate that CD80/CD86 blockade by CTLA4-Ig and subsequent antigen feeding is able to establish oral tolerance in antigen-presensitized mice through inducing CD101⁺ pTreg cells.

It has been controversial whether pTreg cells are as functionally stable as tTreg cells and how they can be functionally and phenotypically distinguished from the latter. We have shown here that intestinal pTreg cells induced by antigen feeding are similar to tTreg cells. They not only express Foxp3 but also possess Treg-type epigenetic changes such as DNA hypomethylation at the Treg-specific demethylation regions and activation of the Treg-specific super-enhancers at Treg signature gene loci including Foxp3, Il2ra, and Ctla4. Additionally, these cells are Helios^hi and Neuropilin-1^hi as are tTreg cells (Thornton et al., 2010; Weiss et al., 2012), while pTreg cells newly generated in BMC mice by EWP feeding were Helios^low. It is likely that chronic stimulation may gradually upregulate Helios and Neuropilin-1 in pTreg cells (Szurek et al., 2015; Thornton and Shevach, 2019). In addition, the generated pTreg cells are in a tissue-adapted state as seen with tissue-specific and tissue-resident tTreg cells (Muñoz-Rojas and Mathis, 2021; Dikiy and Rudensky, 2023), as illustrated by the expression of Tnfrsf9, Tnfrsf4, Gata3, and Areg, as well as skin-specific chemokine receptors such as Ccr5 and Ccr8 (Schiering et al., 2014; Miragaia et al., 2019). However, despite these immunological properties indicative of functional competence and stability of pTreg cells induced by antigen feeding, cessation of antigen administration resulted in decline of oral tolerance with diminution of proliferative and Bcl-2⁺ (i.e., apoptosis-resistant) pTreg cells. This could be, at least in part, attributed to the antigen specificity of the induced pTreg cells, as well as the strength and duration of antigen stimulation delivered to them in the periphery. tTreg cells are more self-reactive than Tconv cells, and hence tonically stimulated in the periphery for better survival (Hsieh et al., 2004; Hsieh et al., 2012). In contrast, pTreg cells once generated by antigen feeding may require persistent or frequent stimulation by the administered cognate non-self-antigen for their survival to maintain tolerance, despite their possession of Treg-specific epigenetic alterations and acquisition of tTreg-like immunological properties.

The generation of intestinal pTreg cells by antigen feeding depends on the quantity of cognate antigen contained in the diet. Notably, even DORe mice fed with NC or Ag-free diet developed Foxp3⁺ T cells particularly in SI-LP and LI-LP, and that they are phenotypically and epigenetically indistinguishable from pTreg cells generated by EWP-high diet. In NC-fed mice, a certain diet protein(s) or a microbial substance(s) cross-reactive with OVA could likely stimulate DO Tconv cells to differentiate into pTreg cells. The NC-induced pTreg cells were, however, smaller in number and less activated (e.g., lower in CD25 expression) than the high-EWP–induced pTreg cells and unable to suppress ear swelling upon antigen exposure. These findings indicate that induction of oral tolerance depends on the number of antigen-specific pTreg cells and/or their activation status, and that both events depend on the quantity of the ingested antigen. This dependency of oral tolerance on the quantity of administered antigen may also be the case in establishing mucosal tolerance via other routes, for example, by sublingual antigen administration for allergy treatment (Windom et al., 2024).

Intestinal pTreg cells induced by antigen feeding specifically express CD101, which is also highly specific for iTreg cells generated in vitro by the absence or blockade of CD28 costimulatory signal. Intestinal pTreg cells found in NC-fed mice, as discussed above, similarly expressed CD101, whereas nTreg cells in mLNs in WT mice hardly expressed the molecule even after high-EWP feeding. Furthermore, in vivo treatment with CTLA4-Ig for oral tolerance induction generated CD101⁺ pTreg cells in the antigen-challenged tissue, and the degree of expression depended on the dose of CTLA4-Ig. Recent studies have shown that CD101 is highly expressed by tissue-resident Treg cells compared with circulating ones (Kaminski et al., 2023); its expression is correlated with the level of FOXP3 expression in human Treg cells (Ferraro et al., 2014), and with in vivo Treg suppressive capacity on GvHD and IBD in mouse models (Fernandez et al., 2007; Schey et al., 2016). Furthermore, a genetic variant of CD101 contributes to genetic susceptibility to type 1 diabetes in an animal model through affecting Treg and other immune cells (Rainbow et al., 2011; Mattner et al., 2019). CD101 is reportedly expressed by exhausted CD8⁺ T cells as well (Hudson et al., 2019). These findings and ours, when taken together, suggest that CD101 plays a role in modulating TCR and CD28 signaling (Soares et al., 1998), especially the latter, for pTreg development in the intestine and iTreg generation from Tconv cells in vitro. CD101, also known as glycoprotein V7 containing EWI motif, is a member of the immunoglobulin superfamily; the molecular basis of the function of CD101 and its ligand remains to be determined. Nevertheless, CD101 could be a useful marker for pTreg cells, especially those induced by CD80/CD86^low tolerogenic DCs, which transduce TCR signal with reduced CD28 signal.

CD80/CD86 blockade by CTLA4-Ig facilitates in vivo CD101⁺ pTreg induction and in vitro CD101⁺ iTreg generation. The blockade and antigen stimulation allow both Treg populations to proliferate and enable them to acquire the Treg-type epigenome, hence their functional and cell-lineage stability (Mikami et al., 2020). IL-2 is essential for this Treg generation because it can compensate for the absence or reduction of CD28 signaling, hampering anergy induction and enabling more efficient generation of iTreg cells than without the blockade. It is plausible that in vivo provision of IL-2 could further facilitate pTreg cell generation by antigen feeding after CD80/CD86 blockade. In addition, APCs, especially DCs, play key roles in pTreg cell induction by antigen feeding (Esterházy et al., 2016; Brown and Rudensky, 2023). Both cDC1 and cDC2 were able to generate iTreg cells in vitro under CD80/CD86 blockade. In addition to the role of CD80/CD86 for Treg induction by APCs, it has been shown that PD-L1⁺ DCs are potent in inducing antigen-specific Treg cells and that PD-L1 blockade inhibits the induction (Wang et al., 2008; Francisco et al., 2009). We and others have shown that Treg cells are able to inhibit the T-cell stimulatory activity of APCs by reducing their CD80/CD86 expression via trogocytosis and transendocytosis, which are dependent on Treg-expressed CTLA-4 (Wing et al., 2008; Onishi et al., 2008; Qureshi et al., 2011; Tekguc et al., 2021). This CD80/CD86 reduction on APCs results in dissociation of PD-L1 from cis-bound CD80, increasing free PD-L1 available for the inhibition of PD-1–expressing effector T cells (Tekguc et al., 2021). CTLA4-Ig can have a similar effect by disrupting cis-CD80-PD-L1 interaction depending on the ratio of CD80 to PD-L1 on APCs (Tekguc et al., 2021; Oxley et al., 2024; Robinson et al., 2024). We have shown in this report that reduction or abrogation of CD28 signal alone (e.g., by utilizing CD28-deficient naïve Tconv cells), without modulating PD-L1/PD-1 interaction, suffices to generate iTreg cells effectively. It still needs to be determined, however, whether enhanced signaling from free PD-L1 on APCs to PD-1⁺ Tconv cells facilitates the conversion of the latter to pTreg cells.

Prior antigen sensitization and the following antigen feeding evoked systemic inflammation. CTLA4-Ig treatment before antigen feeding not only prevented the systemic inflammation but also suppressed local inflammation at the site of antigen exposure. These findings carry some clinical implications. First, it is conceivable that certain food allergies might be a consequence of prior antigen sensitization at other sites of the body, as proposed as the dual-allergen-exposure hypothesis for food allergy (Lack, 2008). Second, CTLA4-Ig has been used for treatment of allergy on the assumption that it may suppress effector T-cell activation and IgE production (Van Wijk et al., 2005). We have shown here that CTLA4-Ig can additionally facilitate the generation of antigen-specific functionally stable pTreg cells from antigen-primed Tconv cells when combined with subsequent antigen feeding. Third, this approach to establishing oral tolerance by CD80/CD86 blockade and antigen feeding can be useful for mucosal tolerance induced by other ways, for example, by sublingual or subcutaneous antigen treatment to treat chronic allergy. It could also be extended to the induction of immune tolerance toward microbial antigens derived from intestinal commensal bacteria to treat inflammatory bowel disease.

In conclusion, antigen-specific functionally stable CD101⁺ pTreg cells are generated in the intestine by antigen feeding and engaged in the establishment of long-term systemic antigen-specific immune tolerance. CD101⁺ iTreg cells with similar phenotype and function can also be produced in vitro from antigen-specific Tconv cells by depriving CD28 costimulation during their induction, suggesting a common developmental basis for these pTreg and iTreg cells. Furthermore, continuous antigen feeding combined with CD28 costimulation blockade is able to generate CD101⁺ pTreg cells and establish immune tolerance even in previously antigen-sensitized animals.

BALB/c mice were purchased from SLC or CLEA. DO11.10 TCR transgenic mice, Rag2-deficient mice, Foxp3-eGFP (eFox) reporter mice, Thy1.1-eFox mice, Foxp3-DTR-GFP mice, and CD28-deficient mice were previously described (Mikami et al., 2020; Kim et al., 2007; Shahinian et al., 1993). All mice were maintained under specific pathogen–free conditions. All procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Committee on Animal Research of The University of Osaka.

AIN-93G diet and EWP-added food were purchased from CLEA. The AIN-93G diet was used as antigen-free food (Table S1). CE-2 (CLEA) as NC was used as standard food in the animal facility. For DORe mice, food modification began at 2 wk of age under milk feeding or at 4 wk of age after weaning, and continued until 8 wk of age. For BMC mice, the food modification was started at the time of BM cell transfer, and continued for 4–8 wk.

The procedure was described in previous studies (Shimura et al., 2016). Mice were anesthetized with isoflurane. Both ears were treated with TS five or six times using surgical tape (21N; Nichiban). Then, a total of 12.5 μl of 10 mg/ml of papain with 10 mg/ml OVA was applied to each side of the surface of mouse ear. This treatment was repeated twice per week for 2 wk. During the experimental period, ear thickness was measured daily using a digital caliper. For depleting Foxp3⁺ cells in the experiment, DTx (400 ng in 200 μl PBS) was injected intraperitoneally twice per week into the Foxp3-DTR-GFP background mice.

Tissues were fixed with 10% formalin, embedded in paraffin blocks, and sectioned and stained by hematoxylin and eosin. For ear swelling test, histological and pathological scoring was performed by following criteria and summed up: disorder of epidermis: 1–3; disorder of dermis: 1–3; disorder of hypodermis: 1–3; cell invasion: 1–3; and range of inflammation: 1–3.

Dissected ears were minced into small pieces with scissors and digested in the isolation buffer (HBSS(−) with 2% fetal bovine serum [FBS], 2 mg/ml collagenase D [Roche], and 20 µg/ml DNase I [Roche]) with 1,300 rpm shaking at 37°C for 2 h. The suspension was then passed through a 70-μm mesh, and the debris was mashed by using a syringe plunger. The flow-through was centrifuged at 500 × g at 4°C for 5 min. The pellet was washed and resuspended with FACS buffer. Isolated cells were used for flow cytometry directly or first stimulated with PMA and ionomycin in the presence of GolgiStop (BD) for 4 h at 37°C for subsequent intracellular cytokine staining.

Intestines were harvested from mice, the lumen was opened longitudinally with scissors, the luminal contents were removed, and the tissues were washed with FACS buffer. After cutting the tissues into 2 cm, they were incubated with IE isolation buffer (HBSS[−] with 2% FBS, 1 mM EDTA [Nacalai]) at 37°C for 30 min. The tissues were then incubated with LP isolation buffer (HBSS[−] with 2% FBS, 1 mg/ml collagenase D [Roche], and 10 µg/ml DNase I [Roche]) at 37°C for 30 min. The suspension was then passed through a 70-μm mesh, and the debris was mashed using a syringe plunger. The flow-through was centrifuged at 500 × g at 4°C for 5 min. The pellet was resuspended with 40% Percoll solution, then layered with 80% Percoll solution, and centrifuged at 2,000 rpm at room temperature for 20 min without acceleration and brake. The interface was collected and washed with FACS buffer.

BM cells were harvested from the femur and tibia of mice, using a 23G needle and syringe. After lysing red blood cells, CD3e⁺ T cells were removed using the MACS negative selection system (Miltenyi Biotec) or BD IMag Cell Separation System (BD) following the manufacturer’s protocol. T cell–depleted BM cells from Thy1.1-eFox and DORe mice were mixed at a ratio of 1:2 in PBS buffer and injected intravenously into CD45.1⁺Rag2-deficient mice as recipients that had been irradiated at 4.0 Gy. After 4 wk or more later from BM reconstitution, chimeric mice were sacrificed and analyzed.

For the preparation of DORe-pTreg cells, CD4⁺DO11.10⁺CD25⁺Foxp3-eGFP⁺ cells were sorted from the peripheral LNs (pLNs) and peripheral tissues harvested from DORe mice. Peripheral Tconv cells were defined as these gating below: naïve Tconv cells (CD4⁺CD25⁻Foxp3-eGFP⁻CD62L^hiCD44^lo) and effector Tconv cells (CD4⁺CD25⁻Foxp3-eGFP⁻CD62L^loCD44^hi). As APCs, cDC1 (CD45⁺CD11c⁺IA-IE⁺CD103⁺CD11b⁻) and cDC2 (CD45⁺CD11c⁺IA-IE⁺CD103⁺CD11b⁺) fractions were sorted from SI-LP in CD45.1 WT mice. BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD) was used for intracellular staining for cytokines used following stimulation with PMA and ionomycin and treatment with Golgi-Stop (BD Biosciences), and the Foxp3 staining kit (eBiosciences) was used for transcriptional factors or intracellular molecules. Flow cytometry analysis and cell sorting were performed using FACSCanto II, FACSCelesta, FACSAria II, and FACSAria Fusion (BD Biosciences). Antibodies used in this study are listed in Table S2.

Bisulfite sequencing analysis was performed as previously described (Arai et al., 2022). Briefly, cells were sorted by FACSAria II, and gDNA was extracted by phenol extraction followed by ethanol precipitation. Bisulfite base conversion was carried out using Methyl Easy Xceed Rapid DNA Bisulfite Modification Kit (Human Genetic Signatures) or following our original methods previously described. PCR primer sequences for Treg-specific demethylated regions are available (Ohkura et al., 2012; Arai et al., 2022).

ChIP-seq experiments were performed as previously described (Kitagawa et al., 2017; Kawakami et al., 2021). Briefly, sorted cells were fixed with 1% formaldehyde (Thermo Fisher Scientific) for 10 min for anti-histone ChIP at room temperature. After nuclear extraction, chromatin lysate was fragmented using Picoruptor (Diagenode), 30-s sonication and 30-s cooling for 7 cycles at 4°C, before immunoprecipitation. The immunoprecipitated chromatin lysate was reverse-cross-linked at 65°C for 24 h, followed by purification and library preparation using NGS Library Preparation Kit for IonS5 (Thermo Fisher Scientific) or NEBNext Ultra II DNA Library Prep Kit for Illumina (Illumina) according to the manufacturer’s instructions. Raw data were generated using the IonS5 sequencing system (Thermo Fisher Scientific) or NextSeq 500 (Illumina).

ATAC-seq was performed as previously described (Corces et al., 2017). Briefly, sorted cells (up to 100,000 cells) were lysed with 100 µl of lysis buffer (0.01% digitonin, 0.1% NP-40, 0.1% Tween-20 in resuspension buffer; 10 mM Tris-HCl, pH 7.5, 100 mM NaCl, 3 mM MgCl₂) for 3 min on ice. After removal of lysis buffer by centrifugation, Tn5 tagmentation was performed using Illumina Tagment DNA TDE1 Enzyme and Buffer Kits (Illumina) at 37°C for 30 min, with shaking at 1,000 rpm. After purification using DNA Clean & Concentrator-5 (Zymo Research), tagmented DNA was amplified using NEBNext High-Fidelity PCR Master Mix (New England BioLabs) with the following prepared DNA libraries that were purified using DNA Clean & Concentrator-5 or AMPure XP (Beckman Coulter). Sequencing was performed using NextSeq 500 or NovaSeq (Illumina).

For ChIP-seq analysis, raw sequence reads were inspected and trimmed using Trim-galore (v0.6.6) (Babraham Bioinformatics). Trimmed reads were mapped to mm10 genome using Bowtie2 (v2.3.5) (Langmead and Salzberg, 2012) with the following options; --local --very-sensitive-local. Mapped reads were sorted using Samtools (v1.6) (Li et al., 2009), then converted into bigwig files for visualization by bamCoverage (v3.1.3), included deepTools package (Ramírez et al., 2016), with the following options: -of bigwig --binSize 5 -p max --normalizeUsing CPM --smoothLength 15 –ignoreDuplicates. Peak calling was performed using MACS2 (v2.1.1.20160309) (Feng et al., 2012) with the following options: macs2 callpeak -t ${filename} -name ${outputfilename} -f BAM -g mm --SPMR --nolambda --nomodel --bdg --call-summits. For defining Treg- or Tconv-specific enhancer regions, narrowPeak files were concatenated, sorted. The sorted narrowPeak file was merged using bedtools (v2.30.0) (Quinlan and Hall, 2010) and then converted into a SAF format file. The SAF peak file was annotated using HOMER (v4.11) (Heinz et al., 2010). Count files were generated using featureCounts (Liao et al., 2014) with the following options: -a annotated.saf -F SAF -p -o counts.txt ${BAM}. Differential expression analysis was performed using DESeq2. To assess the tag density of H3K27ac signals in Treg and naïve Tconv cells, matrix files were created using computeMatrix (v3.1.3), included deepTools package, with the following options: --beforeRegionStartLength 2000 --regionBodyLength 5000 --afterRegionStartLength 2000 --skipZeros, then visualized the plotProfile command.

For ATAC-seq analysis, sequence reads were mapped to mm10 genome using Bowtie2 (v2.3.5) by the following options: --very-sensitive. Mapped reads were sorted by Samtools (v1.9) and then converted into bigwig files for visualization by bamCoverage (v3.1.3) with the following options: -of bigwig --binSize 5 -p max --normalizeUsing CPM --smoothLength 15 –ignoreDuplicates. Peak calling was performed using MACS2 (v2.1.1.20160309) with the following options: macs2 callpeak -t ${filename} -name ${outputfilename} -f BAM -g mm --SPMR --nolambda --nomodel --shift -75 --extsize 150 --keep-dup all --bdg --call-summits.

Integrated Genomics Viewer was used to visualize peak or region data. Workflow of the analysis is deposited at: https://doi.org/10.5281/zenodo.15015798, https://doi.org/10.5281/zenodo.15024562, https://doi.org/10.5281/zenodo.15015786.

scRNA-seq was performed on a Chromium instrument (10X Genomics) following the user guide manual for 30 v3 or 50 v2 chemistry. Briefly, FACS-sorted DORe-derived T cells isolated from mLNs and SI-LP were stained with TotalSeq anti-mouse hashtag antibody (BioLegend), and washed three times with FACS buffer. Cells were resuspended with FACS buffer to a final concentration of ∼1,000 cells/μl with a viability above 90%. Capturing cells in droplets, reverse transcription, and cell barcoding were performed on the Chromium Controller (10x Genomics), followed by PCR amplification and library construction using Chromium Next GEM Single Cell 5′ Kit v2. Final libraries were sequenced on NovaSeq 6000 (Illumina), 20,000 reads/cell for mRNA and 5,000 reads/cell for antibody-derived tags.

Sequenced reads were quantified using Cell Ranger (v5.0.0) with the prebuilt reference refdata-gex-mm10-2020-A downloaded from 10x Genomics’ website. Quantified expressions were preprocessed and visualized using Seurat (v4.1.3 or v5.0.1) (Hao et al., 2021). Differentially expressed genes were calculated using the “FindMarkers” function based on the nonparametric Wilcoxon rank sum test in Seurat. Gene ontology analysis was performed using clusterProfiler (v4.10.0) (Wu et al., 2021). For the trajectory analysis, selected cells were reclustered and ordered using Monocle3 (v1.3.4) (Cao et al., 2019).

Workflow of the analysis is deposited at: https://doi.org/10.5281/zenodo.15015537.

Bulk RNA-seq was performed as below. Briefly, FACS-sorted cells were lysed by RLT RNA lysis buffer (Qiagen) containing 2-mercaptoethanol to a final concentration of ∼1,000 cells/10 μl followed by reverse transcription using the SMART-seq v4 Ultra Low Input RNA Kit for Sequencing (Clontech). Library preparation was performed using Kapa Library Preparation Kit for Ion Torrent or for Illumina (KAPA) by following the manufacturer’s protocol. Sequencing of cDNA libraries was performed on IonS5 (Thermo Fisher Scientific), NextSeq 500, or NovaSeq 6000 (Illumina).

Sequenced reads were inspected and trimmed using Trim-galore (v0.6.6) and aligned to the mouse genome GRCm38/mm10 using STAR (v2.7.7a) (Dobin et al., 2013). Reads counts at the gene level were calculated using RSEM (v1.3.1) (Li and Dewey, 2011). Normalization for library size and differential expression analysis were performed using DESeq2 (v1.32.0) (Love et al., 2014).

Workflow of the analysis is deposited at: https://doi.org/10.5281/zenodo.15015518, https://doi.org/10.5281/zenodo.15015530.

Cell culture was performed in RPMI 1640 supplemented with 10% (vol/vol) FBS, 60 g/ml penicillin G, 100 g/ml streptomycin, and 0.1 mM 2-mercaptoethanol. For in vitro Treg induction, sorted naïve Tconv cells were stimulated with plate-bound anti-CD3e mAb (clone: 145-2C11) and soluble anti-CD28 mAb (clone: 37.51) in the presence of 100 U/ml hIL-2 (Shionogi) and 5 ng/ml rhTGF-β₁ (R&D Systems). For the coculture experiment, DORe-derived naïve or effector/memory Tconv cells were stimulated by sorted APCs, 5 µM OVA_323-339 peptide (MBL Co., LTD., MedChemExpress), 100 U/ml rhIL-2, and 2.5 ng/ml rhTGF-β₁, and with or without 10 nM ATRA supplementation (Wako) or 10 µg/ml CTLA4-Ig (abatacept; Orencia, BMS) for 72 h.

Values were expressed as the mean ± SEM. Statistical significance was assessed by Student’s t test or Welch’s t test (two groups) following an F test. One-way ANOVA and Tukey’s HSD or Tukey–Kramer’s test were conducted for multiple comparisons. A P value <0.05 was considered statistically significant, and every P value was shown in the figures.

Fig. S1 shows pTreg generation and maintenance by antigen feeding. Fig. S2 shows the development of pTreg cells in NC-fed DORe mice. Fig. S3 shows scRNA-seq analysis of T cells in DORe mice and bulk mRNA-seq analysis of pTreg cells from different tissues after EWP feeding. Fig. S4 shows dose effects of CTLA-4-Ig, TGF-β, and IL-2 on iTreg generation and the gating strategy for isolating intestinal DCs. Fig. S5 shows the effects of CD80/CD86 blockade and subsequent antigen feeding on antigen-presensitized mice. Table S1 indicates nutritional facts of modified foods. Table S2 lists the antibodies used in the study.

The raw data (fastq files) from ATAC-seq, ChIP-seq, RNA-seq, and scRNA-seq experiments have been deposited in the DDBJ BioProject database, with links to the BioProject accession number PRJDB18773. All data needed to evaluate the conclusions are present in the paper or the supplementary materials.

We thank all Sakaguchi lab members, particularly: R. Ishii and M. Matsuura for their technical assistance, and C. Tay and S.K. Singh for critical reading of the manuscript. We would like to thank K. Kawade for advising us about the production of modified foods. The genomic sequencing and bioinformatics analyses were conducted using the facility at the Genome Information Research Center of the Research Institute for Microbial Diseases, The University of Osaka.

This study was supported by the Japan Society for the Promotion of Science (JSPS) grant-in-aid for JSPS Fellows (20J20443 to Masaya Arai), JSPS Grant-in-Aid for Early-Career Scientists (20K16279 to Ryoji Kawakami), JSPS Grant-in-Aid for Scientific Research (23H02733 to Ryoji Kawakami), INFRONT Office of Director’s Research Grants Program (to Ryoji Kawakami), Grant from Nipponham Foundation for the Future of Food (to Ryoji Kawakami), JSPS Grant-in-Aid for Specially Promoted Research (16H06295 to Shimon Sakaguchi), JSPS Grant-in-Aid for Scientific Research (24H00610 to Shimon Sakaguchi), and Leading Advanced Projects for Medical Innovation (LEAP; 18gm0010005 to Shimon Sakaguchi). Open access funding was provided by Kyoto University.

Author contributions: Masaya Arai: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, validation, visualization, and writing—original draft, review, and editing. Ryoji Kawakami: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, and writing—original draft, review, and editing. Yamami Nakamura: investigation. Yoko Naito: investigation. Daisuke Motooka: methodology, resources, software, and writing—review and editing. Atsushi Sugimoto: investigation. Tomiko Kimoto: resources. Naganari Ohkura: methodology and resources. Norihisa Mikami: conceptualization and methodology. Shimon Sakaguchi: conceptualization, funding acquisition, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing.