Tumor-infiltrating regulatory T cells (Tregs) have been extensively studied as therapeutic targets. However, not all infiltrating T cells exert their functions equally, presumably because of their heterogeneity and substantial turnover in tissues. In this study, we hypothesized that intertissue migration underlies the functional heterogeneity of Tregs. To test this, we applied in vivo photolabeling to examine single-cell diversity of immunosuppressive molecules in mouse Tregs migrating to, remaining in, and emigrating from MC38 tumors. Neuropilin-1 (Nrp1) expression was inversely correlated with that of six other molecules associated with Treg function. Unsupervised clustering analyses revealed that clusters containing Tregs that were retained in tumors expressed high levels of the six functional molecules but not of Nrp1. However, these clusters represented only half of the Tregs migrating to the tumor, suggesting evolving heterogeneity of tumor-infiltrating Tregs. Thus, we propose progressive pathways of Treg activation and migration between tumors and draining lymph nodes.

Visual Abstract

Regulatory T cells (Tregs) play essential roles in tissue homeostasis, autoimmunity control, chronic infection, and cancer (1). In lymphoid tissues, Tregs display well characterized immunoinhibitory mechanisms, whereas in nonlymphoid tissues, Tregs mediate tissue-specific immunoinhibition (2). For example, Treg-specific deletion of the immunosuppressive cytokine IL-10 or surface molecule CTL-associated protein 4 (CTLA-4) causes peripheral or systemic immune disorders (1). In several cancers, Tregs infiltrating tumors exhibit more suppressive signatures than Tregs in lymphoid tissues (3) and use more than 10 molecules to suppress immunity (2). However, single-cell functional diversity (4) complicates the investigation of Treg function in both lymphoid and nonlymphoid tissues.

Treg tissue tropism is well characterized. CCR4 mediates Treg recruitment to the tumor sites, lung tissue, and skin (5). In tumors and inflamed tissues, Tregs strongly express other migration molecules, such as CCR5, CCR8, and integrins β1 and αE (3, 6). Although previous reports focused on tissue-infiltrating Tregs, it is likely that only a proportion of them are retained in tissues to mediate immune regulation. Tissue adaptation through the expression of T helper–inducing POZ/Krüeppel–like factor (ThPOK) is required for Tregs to reside in their preferred microenvironments in the intestine (7). Furthermore, our established photoconvertible protein-expressing mice revealed rapid Treg replacement in the inflamed tissue and tumors via emigration to draining lymph nodes (dLNs) (8, 9).

In this study, we investigated Treg function and residence in nonlymphoid pathological tissues by focusing on single-cell Treg functional heterogeneity and intertissue migration between the dLN and tumor. We used mice expressing a photoconvertible protein, KikGR (10), to distinguish the Tregs migrating (Mig) to, remaining (Rem) in tumors, and emigrating (Emi) to dLNs. We identified partially overlapping functional heterogeneity between Treg subpopulations with specific migratory profiles and Treg clusters that were enriched in Rem Tregs and defined by distinct expression profiles of immunoinhibitory molecules.

KikGR knock-in (10, 11), IL10Venus (12), and FucciG1-#639 and FucciS/G2/M-#474 mice (11, 13) were crossed with Foxp3hCD2/hCD52 mice (14) and maintained in specific pathogen–free facilities at Osaka Ohtani University. These mice enabled us to measure KikGR-Green, KikGR-Red, IL-10, and mitosis indicators Kusabira orange and Azami green expression without using Ab staining. Foxp3 expression was measured using the surface staining of human CD2. All experiments were approved by the Animal Research Committee of Osaka Ohtani University.

Twelve hours before the analysis, intradermally injected MC38 colon carcinoma was exteriorized in the anesthetized mice and subjected to 2-min exposure to violet light (436 nm, 200 mW/cm2), which did not affect inflammatory gene expression (8). The tumors and dLNs were digested with collagenase IV (Worthington Biochemical), hyaluronidase (Sigma-Aldrich), and DNase I (Calbiochem). Cells were stained with Abs described in Supplemental Table I. For the intracellular staining of CTLA-4 and granzyme B (GzmB), Fixation Buffer (BioLegend) and 10X Permeabilization Wash Buffer (BioLegend) were used. Flow cytometry (FCM) data were obtained using the SP6800 Spectral Cell Analyzer (Sony) and compensated using a spectral unmixing algorithm (15).

All live Tregs were identified as KikGR-Green+ CD45+ CD4+ hCD2+ cells by FlowJo software (Tree Star). After conventional FCM gating analysis, single-cell expression data of Tregs were analyzed using R 3.5.1 software (R Foundation for Statistical Computing). As partial normalization of the contribution of each animal (n = 7) (16), we extracted and pooled equal cell numbers from each replicate. Data included 2289, 2611, and 786 Mig, Rem, and Emi Tregs, respectively, in addition to 4900 KikGRRed Tregs in dLNs. Values of molecule expression were logicle transformed. Manhattan distance and group average methods were used for clustering analysis. We used t-distributed stochastic linear embedding (tSNE) analysis (perplexity, 500; θ, 0.2; max_iter, 1,000) (16, 17), isometric feature mapping (Isomap; k = 8) (16), and PhenoGraph algorithm (18) (k = 60 in Supplemental Fig. 1; k = 120 in (Fig. 3). We performed Spearman rank-order correlation for the correlation analysis between the molecule expression, as the values did not follow a Gaussian distribution.

KikGR-Red or KikGR-Red+ Tregs harvested from tumor tissues between days 9 and 11 in Foxp3hCD2/hCD52 KikGR mice by cell sorter SH800 (Sony) were transferred into the ear skin of the B6 mice stained using 0.75% 2,4-dinitro-1-fluoro-benzene (Nacalai Tesque) 5 d before. Immediately afterward, 0.45% 2,4-dinitro-1-fluoro-benzene was painted on the ear skin. Before cell transfer and at 8 and 16 h afterward, ear thickness was measured with a thickness gauge (Teclock).

CD4+ effector and CD4 feeder cells were isolated from human CD2 spleen cells using human CD2 and CD4 MicroBeads (Miltenyi Biotec). After the CellTrace Violet (Thermo Fisher Scientific) staining of the CD4+ cells, 625 or 1250 spleen Treg cells, KikGR-Red, or KikGR-Red+ tumor Tregs were cocultured with 5000 CD4+ and 20-Gy–irradiated 50,000 CD4 cells for 66 h in the presence of 500 ng/ml anti-CD3 Ab (BioLegend). The proliferation was determined by CellTrace Violet dilution with FCM analysis.

To block CCR4, CCR5, or G protein pathways, 50 µg of C021 (Cayman Chemical), 8 µg of D-Ala-peptide T-amide (DAPTA; Tocris Bioscience), or 1.2 µg of pertussis toxin (Wako Chemicals), respectively, were dissolved in PBS and administrated intratumorally overnight before analysis.

One-way ANOVA with Tukey post hoc test and Wilcoxon matched-pairs test were performed using GraphPad Prism v7.05 (GraphPad Software). Data in scatter plots represent means ± SEM. The p values < 0.05 were considered statistically significant.

We identified a set of seven functional phenotypic Treg markers (2): Nrp1, CD25, CD39, IL-10, lymphocyte activation gene-3 (LAG-3), CTLA-4, and GzmB. They are expressed heterogeneously in individual Tregs in the inflamed skin (19) and can be simultaneously and reliably measured by FCM. We compared their expression profiles in Tregs derived from adenocarcinoma MC38 tumors (Fig. 1A, 1B). Nrp1 expression was negatively correlated with that of the other six molecules, similar to our previous findings in a skin inflammation model (19), suggesting a distinct functional phenotype of Nrp1+ tumor Tregs. Meanwhile, the other six markers were somewhat positively correlated with each other, indicating that they may be useful markers for multifunctional Tregs.

FIGURE 1.

Expression profiles of immunosuppressive molecules in tumor Tregs. (A) Expression of seven functional molecules in tumor Tregs. FCM data are representative of eight independent experiments. (B) Cross-correlation analysis of expression profiles of the seven molecules using pooled data (n = 8 in three independent experiments). Colors indicate positive (red) or negative (blue) correlations. (C) Definition of expression levels. Nonexpressing cells were defined by fluorescence minus-one control (FMO). Positive cells were divided into five equal levels from the highest expression level (red circle). (D and E) tSNE maps of tumor Tregs derived from mice receiving PBS, C021, DAPTA, or pertussis toxin (n = 10, 4, 4, and 2, respectively, in three independent experiments). Colors indicate expression levels of the five molecules (black, blue, cyan, green, yellow, and red for each expression level) (D) or density of individual cells (yellow to red) (E). Contour plots indicate distributions of individual cells.

FIGURE 1.

Expression profiles of immunosuppressive molecules in tumor Tregs. (A) Expression of seven functional molecules in tumor Tregs. FCM data are representative of eight independent experiments. (B) Cross-correlation analysis of expression profiles of the seven molecules using pooled data (n = 8 in three independent experiments). Colors indicate positive (red) or negative (blue) correlations. (C) Definition of expression levels. Nonexpressing cells were defined by fluorescence minus-one control (FMO). Positive cells were divided into five equal levels from the highest expression level (red circle). (D and E) tSNE maps of tumor Tregs derived from mice receiving PBS, C021, DAPTA, or pertussis toxin (n = 10, 4, 4, and 2, respectively, in three independent experiments). Colors indicate expression levels of the five molecules (black, blue, cyan, green, yellow, and red for each expression level) (D) or density of individual cells (yellow to red) (E). Contour plots indicate distributions of individual cells.

Close modal

To investigate the effect of migration on functional heterogeneity in tumor Tregs, we measured Nrp1, CD25, CD39, IL-10, and LAG-3 expression at the single-cell level in tumor Tregs from mice treated with chemokine inhibitors. CTLA-4 and GzmB, which require intracellular staining, were omitted to ensure sufficiently high cell numbers. Positive and negative cells were distinguished by fluorescence minus-one control FCM gating, and positive cells were divided into five groups based on expression levels (Fig. 1C). Individual Tregs were clustered by tSNE analysis based on functional marker expression profiles (Fig. 1D). Two clear clusters were observed: IL-10+ and IL-10.

Individual cells from mice treated with inhibitors of CCR4 (C021), CCR5 (DAPTA), or G protein (pertussis toxin) were distributed differently than PBS-treated samples (Fig. 1E). Both C021 and DAPTA treatments significantly altered the proportions of some clusters when defined by PhenoGraph analysis (Supplemental Fig. 1A, 1B). We expected phenotypic alteration by chemokine inhibition to parallel correlations between chemokine receptors and functional molecules (Supplemental Fig. 1C). However, no clear trend linking these phenotypes could be observed, suggesting that the complexity of the mechanisms that determine the Treg phenotypic diversity cannot be explained by the influence of individual chemokines. Pertussis toxin treatment increased the proportion of cells in the IL-10 cluster (Fig. 1E), although a low sample number (n = 2) precluded statistical analysis. These trends in cell distribution were observed in almost all tSNE plots that were computed with perplexities between 30 and 1000 (data not shown).

To test the hypothesis that the migration behavior of Tregs contributes to their functional heterogeneity in tumors, we investigated the expression profiles of Treg subpopulations with distinct intertissue migration behavior using KikGR-expressing mice. The exposure to violet light induces photoconversion of the KikGR protein from green (KikGR-Green) to red (KikGR-Red) (10). The targeted tumor photoconversion converted all Tregs in the tumor into KikGR-Red without affecting the fluorescence of the Tregs in the dLNs (Fig. 2A). We evaluated Rem (KikGR-Red+), Mig (KikGR-Red), and Emi Tregs (KikGR-Red+ cells in the dLN). KikGR-Red Tregs in the dLNs are a mix of Tregs retained in the dLNs and Tregs derived from other tissues or the tumor before photoconversion.

FIGURE 2.

Distinct profiles of Mig and Rem Tregs in tumors. (A) FCM gating strategies for Mig, Rem, and Emi Tregs from photoconverted tumor tissues. Values in plots indicate percentages of parent populations. (B) Proportions of KikGR-Red+ cells in Tregs from tumor tissues 12 h after photoconversion at days 7 and 9–11 (n = 4 and 18, respectively, in seven independent experiments). (C) Proportions of cells expressing seven molecules in Mig and Rem Tregs (n = 9 in four independent experiments). (D) Ear swelling at 8 and 16 h following induction of contact hypersensitivity and intradermal transfer of PBS (black, n = 12), Mig Tregs (green, n = 8) and Rem Tregs (red, n = 10 in two independent experiments). (E) Proportions of divided effector T cell cocultured without Tregs (white, n = 4) or with 625 or 1250 spleen Tregs (black, n = 4), Mig Tregs (green, n = 6 or 5) or Rem Tregs (red, n = 8 in three independent experiments). Pound sign (#) indicates statistical differences with all the other subsets (p < 0.05). Asterisks indicate statistical difference (p < 0.05).

FIGURE 2.

Distinct profiles of Mig and Rem Tregs in tumors. (A) FCM gating strategies for Mig, Rem, and Emi Tregs from photoconverted tumor tissues. Values in plots indicate percentages of parent populations. (B) Proportions of KikGR-Red+ cells in Tregs from tumor tissues 12 h after photoconversion at days 7 and 9–11 (n = 4 and 18, respectively, in seven independent experiments). (C) Proportions of cells expressing seven molecules in Mig and Rem Tregs (n = 9 in four independent experiments). (D) Ear swelling at 8 and 16 h following induction of contact hypersensitivity and intradermal transfer of PBS (black, n = 12), Mig Tregs (green, n = 8) and Rem Tregs (red, n = 10 in two independent experiments). (E) Proportions of divided effector T cell cocultured without Tregs (white, n = 4) or with 625 or 1250 spleen Tregs (black, n = 4), Mig Tregs (green, n = 6 or 5) or Rem Tregs (red, n = 8 in three independent experiments). Pound sign (#) indicates statistical differences with all the other subsets (p < 0.05). Asterisks indicate statistical difference (p < 0.05).

Close modal

We investigated changes in the proportions of Rem Tregs and tumor growth (Fig. 2B). We found that the highest tumoral Treg turnover occurred at the early stage of tumor growth, with only 5.3% ± 0.9% of photoconverted Rem Tregs in tumors 12 h after photoconversion, indicating substantial Treg influx. We found a greater proportion of Rem Tregs at days 9–11 (29.8% ± 4.2% of tumor Tregs were KikGR-Red+; (Fig. 2B). Therefore, tumors at days 9–11 were used for subsequent experiments.

Because we previously observed that Mig or Rem Tregs in the inflamed skin possess distinct expression profiles of immunoinhibitory molecules (19), we analyzed the functional profiles of Mig and Rem Tregs in tumors. The Nrp1 expression was higher in Mig Tregs, whereas those of CD25, CD39, IL-10, LAG-3, and GzmB were higher in Rem Tregs (Fig. 2C). The transfer of Rem Tregs but not of Mig Tregs to the inflamed skin decreased swelling compared with control mice (Fig. 2D) in a contact hypersensitivity model, in which Treg transfer and CD39 deletion reduced the inflammatory response (4, 8, 20). In contrast, Rem Tregs but not Mig Tregs showed less inhibitory activity than spleen Tregs in the in vitro proliferation assay (Fig. 2E) in which the Nrp1+ Treg inhibitory activity was higher than that of the Nrp1 Tregs (21, 22). These data suggest that tumor-infiltrating Tregs are not phenotypically and functionally homogeneous.

The CCR4 expression in Rem Tregs was the highest in the four Treg subsets (Supplemental Fig. 2a). Therefore, we expected that the C021 treatment would lead to a change in functional heterogeneity of Tregs (Fig. 1E) due to the inhibition of tumor recruitment and retention. However, C021 did not change Rem Treg proportions and, unexpectedly, reduced proportions of the Emi Tregs (Supplemental Fig. 2b). Furthermore, we did not observe any changes in the Rem Treg proportions by the CD29 and CCR5 blockade (data not shown). These data suggested that there may be some redundancies in the chemokine pathways that control Treg migration, and it would be hard to manipulate bulk Treg migration by blocking individual chemokines. This led us to analyze the relationship between functional heterogeneity and migration status.

Given the progressive phenotypic adaptation of Mig Tregs to nonlymphoid tissues (7, 23), single-cell data analysis of Tregs in tumors and dLNs could clarify the phenotypic progression and migration behavior. To this end, we performed an unsupervised Isomap analysis, a more appropriate method to map potential relationships between cell subsets in dimensionality-reduced plots than tSNE (17), and subdivided the Tregs into clusters using a PhenoGraph algorithm (18). Rem, Mig, and Emi Tregs, as well as KikGR-Red Tregs in the dLNs (2289, 2611, 786, and 4900 cells, respectively) were distributed sequentially along the first dimension of Isomap (Isomap 1 in (Fig. 3A), wherein Nrp1 expression decreased and the six other molecules concurrently increased (Fig. 3B, 3C), consistent with the correlation analysis (Fig. 1B). Unlike Rem Tregs, which showed high expression of CD39, IL-10, LAG-3, and GzmB, dLN Tregs exhibited a significantly lower expression of these markers, suggesting very distinct functional phenotypes between tumor-tropic and dLN Tregs. We found a cluster chiefly composed of cells with very low/negative expression of all seven molecules (0, 15, 17, and 512 cells in each of the four populations, respectively) in the Isomap 1–low Isomap 2–high region (Fig. 3B–D), suggesting naive or anergic Tregs.

FIGURE 3.

Comparative heterogeneity of functional phenotypes of Mig, Rem, and Emi Tregs in tumors. (A and C) Isomaps of tumor and dLN Treg data (n = 7 in three independent experiments) colored by migration behavior (A) and expression levels of seven molecules (C). Contour plots indicate distributions of individual cells. (B) Expression levels of seven molecules with the first and second Isomap dimensions. (D) Individual cells with very low or negative expression of all seven molecules (blue) in the Isomap map. (E) Isomaps of Tregs with four types of migration behavior colored by cluster were defined by PhenoGraph. The location and size of numbers indicate centroids of clusters and proportions of clusters in each Treg subpopulation, respectively. (F) Heatmap indicating expression levels of the seven molecules in PhenoGraph-defined clusters (black to yellow) and proportions of the clusters in Tregs with different migration behavior (red-yellow-white). The same expression data were used in (A)–(D) and (F) (left panel). (G) Proportions of PhenoGraph-defined clusters in Tregs with migration behavior. Number in bar plot indicates cluster with the largest proportions compared with the other three Treg subpopulations. An asterisk indicates statistical significance (p < 0.05). The same data of cell frequency were used in (E), (F) (right panel), (G), and Supplemental Fig. 3. (H) Proportions of cells in S/G2/M phase (mAzami Green+ mKusabira Orange2low/−) in Tregs derived from dLNs and tumors. FCM data are representative of three independent experiments. (I) Hypothetical pathways of migration, activation, and residence in Tregs.

FIGURE 3.

Comparative heterogeneity of functional phenotypes of Mig, Rem, and Emi Tregs in tumors. (A and C) Isomaps of tumor and dLN Treg data (n = 7 in three independent experiments) colored by migration behavior (A) and expression levels of seven molecules (C). Contour plots indicate distributions of individual cells. (B) Expression levels of seven molecules with the first and second Isomap dimensions. (D) Individual cells with very low or negative expression of all seven molecules (blue) in the Isomap map. (E) Isomaps of Tregs with four types of migration behavior colored by cluster were defined by PhenoGraph. The location and size of numbers indicate centroids of clusters and proportions of clusters in each Treg subpopulation, respectively. (F) Heatmap indicating expression levels of the seven molecules in PhenoGraph-defined clusters (black to yellow) and proportions of the clusters in Tregs with different migration behavior (red-yellow-white). The same expression data were used in (A)–(D) and (F) (left panel). (G) Proportions of PhenoGraph-defined clusters in Tregs with migration behavior. Number in bar plot indicates cluster with the largest proportions compared with the other three Treg subpopulations. An asterisk indicates statistical significance (p < 0.05). The same data of cell frequency were used in (E), (F) (right panel), (G), and Supplemental Fig. 3. (H) Proportions of cells in S/G2/M phase (mAzami Green+ mKusabira Orange2low/−) in Tregs derived from dLNs and tumors. FCM data are representative of three independent experiments. (I) Hypothetical pathways of migration, activation, and residence in Tregs.

Close modal

We identified 12 PhenoGraph clusters, numbered in order of proportions of Rem Tregs in each cluster (Fig. 3E–G). Approximately 80% of Rem Tregs and 50% of Mig Tregs were distributed in Isomap 1–high areas, where clusters 1, 2, 3, 4, and 8 are located (Fig. 3E, 3G). Phylogenetic tree analysis placed these in one of two main roots (Fig. 3F) that differ in Nrp1 expression. For Emi Tregs, 14.3% ± 1.9% were in cluster 1 (significantly more than the number of KikGR-Red Tregs in dLNs; Supplemental Fig. 3) and not in clusters 2, 3, or 4, indicating that cells in cluster 1 have the potential to emigrate from tumors. These data suggest that Nrp1 Tregs expressing high levels of multiple immunoinhibitory molecules, such as those of clusters 2, 3, and 4, comprise the main populations of tumor-remaining Tregs.

In addition to these five clusters, Mig Tregs also included clusters 6 and 7 (13.3% ± 2.9% and 13.6% ± 1.2%, respectively; (Fig. 3G, Supplemental Fig. 3) that were also present in ∼10% of KikGR-Red dLN Tregs. Cells in clusters 6 and 7 may be unadapted or unqualified to remain in tumor tissues. Clusters 6 and 7 displayed higher expression of Nrp1 and lower CD25, CD39, IL-10, LAG-3, and GzmB than clusters 1, 2, 3, and 4 (Fig. 3F). Given that cluster 1 was present in >10% of Emi Tregs and expressed Nrp1, although at very low levels, activation pathways for the reciprocal expression of Nrp1 and the other five molecules may be important determinants for the retention of Tregs in tumors, and Nrp1 could work before tumor residence, such as for the recruitment to tumors (24). These data also suggest that Treg polyfunctionality is characteristic for tumor-resident Tregs, although it is unknown whether this relationship is correlative or causative.

To ascertain a likely source of tumor Treg renewal, we assessed the in situ proliferation of tumor-associated Tregs using a mouse line expressing Fucci, a probe for cell cycle (13). Compared with dLN Tregs, a smaller proportion of tumor Tregs could be observed in the S/G2/M phases (Fig. 3H), suggesting that tumor Treg renewal might rely primarily on migration from other tissues, such as dLNs, which contain thousands of proliferating Tregs.

In this study, we observed that tumors typically contained on the order of several thousand total Tregs, whereas several hundred thousand Tregs could be detected in dLNs. Therefore, tens of thousands of dLN Tregs, such as the cells in clusters 6, 7, 9, 11, and 12, exhibit the potential to be activated by the tumor microenvironment. This could increase their expression of functional molecules, except for Nrp1, along the Isomap 1 axis after recruitment to tumor tissues (Fig. 3I, upper arrow). Approximately 50% of Mig Tregs showed a dLN-like signature that supports this model. However, in future studies, Treg conversion from conventional T cells should be examined in the peripheral tissues (25). A recent study using intravital imaging revealed that T cell residence in the epithelium was driven by ThPOK (7). Alternatively, key markers of each hypothetical Treg subpopulation with distinct stages during skin adaptation were estimated from single-cell sequencing data (23). These two studies also support the presented model (Fig. 3I, upper arrow).

In addition, our observations that ∼5% of KikGR-Red− Tregs in dLNs were in clusters 1, 2, 3, and 4 (Fig. 3G) and that Treg proliferation in dLNs was more intensive than in tumors (Fig. 3H) encouraged us to propose another model regarding new recruitment of Tregs that have already acquired tumor-specific signatures in the dLN (Fig. 3I, lower arrow). A 5% value of dLN Tregs represents thousands of cells. If most of these highly activated Tregs can migrate to tumor sites, they would account for most of the immunoinhibitory cells in the tumor mass, harboring thousands of Tregs.

Although multiple reports compared lymphoid and nonlymphoid tissue-derived Tregs to uncover tissue-specific function and recruitment (3, 5, 6), careful analysis is required for a complete understanding of the tissue-specific phenomenon of Tregs, especially during the rapid Treg replacement in the target tissues. In comparison with Mig Tregs, we found distinct expression profiles and higher anti-inflammatory capacity in Rem Tregs. The single-cell analysis revealed that half of the Mig Tregs showed profiles similar to either tumor-remaining or dLN-specific signatures (Fig. 3E). Based on the proportions of KikGR-Red+ cells (Fig. 2B), Treg migration to the tumor during the period of 12 h accounts for 20–80% of total tumor Tregs, suggesting that tumor Tregs have heterogeneous functional phenotypes.

Given that systemic Treg depletion results in strong side effects (25), the control of functional heterogeneity along with the alteration of tissue residence of Tregs is an effective strategy to establish efficient immunotherapy with fewer side effects. Our data demonstrating the modulation of Treg heterogeneity by chemokine inhibitors provide precious basis for future studies aimed at manipulating tumor-associated Tregs. Our approach using photoconvertible protein-expressing mice will help elucidate tissue-specific functions and tissue residence mechanisms. Furthermore, it would provide crucial information to develop new therapeutic strategies to modulate Treg immunosuppression.

Materials provided by Dr. Kiyoshi Takeda, Osaka University, and Dr. Shohei Hori, IMS, RIKEN, Japan, were of great assistance in this study.

This work was supported by a Japan Society for the Promotion of Science Grant-in-Aid for Young Scientists (B) (15K19687), National Breast Cancer Foundation Grant IIRS-19-027 (to T.C.), a PanKind Foundation Grant (to T.C.), and a grant from the Kato Memorial Bioscience Foundation.

Author contributions: R.I. and M.T. designed experiments. R.I., T.M., and M.T. performed experiments, statistical analysis, and data visualization. R.I., T.M., I.Y., M.U., Y.K., and M.T. analyzed data. R.I., T.C., and M.T. wrote the manuscript.

The online version of this article contains supplemental material.

Abbreviations used in this article

CTLA-4

CTL-associated protein 4

DAPTA

D-Ala-peptide T-amide

dLN

draining lymph node

Emi

emigrating

FCM

flow cytometry

GzmB

granzyme B

Isomap

isometric feature mapping

LAG-3

lymphocyte activation gene-3

Mig

migrating

Rem

remaining

Treg

regulatory T cell

tSNE

t-distributed stochastic linear embedding

1.
Campbell
D. J.
,
M. A.
Koch
.
2011
.
Phenotypical and functional specialization of FOXP3+ regulatory T cells.
Nat. Rev. Immunol.
11
:
119
130
.
2.
Burocchi
A.
,
M. P.
Colombo
,
S.
Piconese
.
2013
.
Convergences and divergences of thymus- and peripherally derived regulatory T cells in cancer.
Front. Immunol.
4
:
247
.
3.
Plitas
G.
,
C.
Konopacki
,
K.
Wu
,
P. D.
Bos
,
M.
Morrow
,
E. V.
Putintseva
,
D. M.
Chudakov
,
A. Y.
Rudensky
.
2016
.
Regulatory T cells exhibit distinct features in human breast cancer.
Immunity
45
:
1122
1134
.
4.
Ikebuchi
R.
,
S.
Teraguchi
,
A.
Vandenbon
,
T.
Honda
,
F. H. W.
Shand
,
Y.
Nakanishi
,
T.
Watanabe
,
M.
Tomura
.
2016
.
A rare subset of skin-tropic regulatory T cells expressing Il10/Gzmb inhibits the cutaneous immune response.
Sci. Rep.
6
:
35002
.
5.
Sather
B. D.
,
P.
Treuting
,
N.
Perdue
,
M.
Miazgowicz
,
J. D.
Fontenot
,
A. Y.
Rudensky
,
D. J.
Campbell
.
2007
.
Altering the distribution of Foxp3(+) regulatory T cells results in tissue-specific inflammatory disease.
J. Exp. Med.
204
:
1335
1347
.
6.
Tan
M. C. B.
,
P. S.
Goedegebuure
,
B. A.
Belt
,
B.
Flaherty
,
N.
Sankpal
,
W. E.
Gillanders
,
T. J.
Eberlein
,
C. S.
Hsieh
,
D. C.
Linehan
.
2009
.
Disruption of CCR5-dependent homing of regulatory T cells inhibits tumor growth in a murine model of pancreatic cancer.
J. Immunol.
182
:
1746
1755
.
7.
Sujino
T.
,
M.
London
,
D. P.
Hoytema van Konijnenburg
,
T.
Rendon
,
T.
Buch
,
H. M.
Silva
,
J. J.
Lafaille
,
B. S.
Reis
,
D.
Mucida
.
2016
.
Tissue adaptation of regulatory and intraepithelial CD4+ T cells controls gut inflammation.
Science
352
:
1581
1586
.
8.
Tomura
M.
,
T.
Honda
,
H.
Tanizaki
,
A.
Otsuka
,
G.
Egawa
,
Y.
Tokura
,
H.
Waldmann
,
S.
Hori
,
J. G.
Cyster
,
T.
Watanabe
, et al
2010
.
Activated regulatory T cells are the major T cell type emigrating from the skin during a cutaneous immune response in mice.
J. Clin. Invest.
120
:
883
893
.
9.
Torcellan
T.
,
H. R.
Hampton
,
J.
Bailey
,
M.
Tomura
,
R.
Brink
,
T.
Chtanova
.
2017
.
In vivo photolabeling of tumor-infiltrating cells reveals highly regulated egress of T-cell subsets from tumors.
Proc. Natl. Acad. Sci. USA
114
:
5677
5682
.
10.
Tomura
M.
,
A.
Hata
,
S.
Matsuoka
,
F. H. W.
Shand
,
Y.
Nakanishi
,
R.
Ikebuchi
,
S.
Ueha
,
H.
Tsutsui
,
K.
Inaba
,
K.
Matsushima
, et al
2014
.
Tracking and quantification of dendritic cell migration and antigen trafficking between the skin and lymph nodes.
Sci. Rep.
4
:
6030
.
11.
Tomura
M.
,
R.
Ikebuchi
,
T.
Moriya
,
Y.
Kusumoto
.
2021
.
Tracking the fate and migration of cells in live animals with cell-cycle indicators and photoconvertible proteins.
J. Neurosci. Methods
355
:
109127
.
12.
Atarashi
K.
,
T.
Tanoue
,
T.
Shima
,
A.
Imaoka
,
T.
Kuwahara
,
Y.
Momose
,
G.
Cheng
,
S.
Yamasaki
,
T.
Saito
,
Y.
Ohba
,
T.
Taniguchi
,
K.
Takeda
,
S.
Hori
,
I. I.
Ivanov
,
Y.
Umesaki
,
K.
Itoh
,
K.
Honda
.
2011
.
Induction of colonic regulatory T cells.
Science
331
:
337
341
.
13.
Tomura
M.
,
A.
Sakaue-Sawano
,
Y.
Mori
,
M.
Takase-Utsugi
,
A.
Hata
,
K.
Ohtawa
,
O.
Kanagawa
,
A.
Miyawaki
.
2013
.
Contrasting quiescent G0 phase with mitotic cell cycling in the mouse immune system.
PLoS One
8
:
e73801
.
14.
Miyao
T.
,
S.
Floess
,
R.
Setoguchi
,
H.
Luche
,
H. J.
Fehling
,
H.
Waldmann
,
J.
Huehn
,
S.
Hori
.
2012
.
Plasticity of Foxp3(+) T cells reflects promiscuous Foxp3 expression in conventional T cells but not reprogramming of regulatory T cells.
Immunity
36
:
262
275
.
15.
Futamura
K.
,
M.
Sekino
,
A.
Hata
,
R.
Ikebuchi
,
Y.
Nakanishi
,
G.
Egawa
,
K.
Kabashima
,
T.
Watanabe
,
M.
Furuki
,
M.
Tomura
.
2015
.
Novel full-spectral flow cytometry with multiple spectrally-adjacent fluorescent proteins and fluorochromes and visualization of in vivo cellular movement.
Cytometry A
87
:
830
842
.
16.
Becher
B.
,
A.
Schlitzer
,
J.
Chen
,
F.
Mair
,
H. R.
Sumatoh
,
K. W. W.
Teng
,
D.
Low
,
C.
Ruedl
,
P.
Riccardi-Castagnoli
,
M.
Poidinger
, et al
2014
.
High-dimensional analysis of the murine myeloid cell system.
Nat. Immunol.
15
:
1181
1189
.
17.
Amir
A. D.
,
K. L.
Davis
,
M. D.
Tadmor
,
E. F.
Simonds
,
J. H.
Levine
,
S. C.
Bendall
,
D. K.
Shenfeld
,
S.
Krishnaswamy
,
G. P.
Nolan
,
D.
Pe’er
.
2013
.
viSNE enables visualization of high dimensional single-cell data and reveals phenotypic heterogeneity of leukemia.
Nat. Biotechnol.
31
:
545
552
.
18.
Levine
J. H.
,
E. F.
Simonds
,
S. C.
Bendall
,
K. L.
Davis
,
A. D.
Amir
,
M. D.
Tadmor
,
O.
Litvin
,
H. G.
Fienberg
,
A.
Jager
,
E. R.
Zunder
, et al
2015
.
Data-driven phenotypic dissection of AML reveals progenitor-like cells that correlate with prognosis.
Cell
162
:
184
197
.
19.
Ikebuchi
R.
,
M.
Fujimoto
,
Y.
Nakanishi
,
H.
Okuyama
,
T.
Moriya
,
Y.
Kusumoto
,
M.
Tomura
.
2019
.
Functional phenotypic diversity of regulatory T cells remaining in inflamed skin.
Front. Immunol.
10
:
1098
.
20.
Ring
S.
,
S. J.
Oliver
,
B. N.
Cronstein
,
A. H.
Enk
,
K.
Mahnke
.
2009
.
CD4+CD25+ regulatory T cells suppress contact hypersensitivity reactions through a CD39, adenosine-dependent mechanism.
J. Allergy Clin. Immunol.
123
:
1287
96.e2
.
21.
Campos-Mora
M.
,
P.
Contreras-Kallens
,
F.
Gálvez-Jirón
,
M.
Rojas
,
C.
Rojas
,
A.
Refisch
,
O.
Cerda
,
K.
Pino-Lagos
.
2019
.
CD4+Foxp3+T regulatory cells promote transplantation tolerance by modulating effector CD4+ T cells in a neuropilin-1-dependent manner.
Front. Immunol.
10
:
882
.
22.
Battaglia
A.
,
A.
Buzzonetti
,
G.
Monego
,
L.
Peri
,
G.
Ferrandina
,
F.
Fanfani
,
G.
Scambia
,
A.
Fattorossi
.
2008
.
Neuropilin-1 expression identifies a subset of regulatory T cells in human lymph nodes that is modulated by preoperative chemoradiation therapy in cervical cancer.
Immunology
123
:
129
138
.
23.
Miragaia
R. J.
,
T.
Gomes
,
A.
Chomka
,
L.
Jardine
,
A.
Riedel
,
A. N.
Hegazy
,
N.
Whibley
,
A.
Tucci
,
X.
Chen
,
I.
Lindeman
, et al
2019
.
Single-cell transcriptomics of regulatory T cells reveals trajectories of tissue adaptation.
Immunity
50
:
493
504.e7
.
24.
Hansen
W.
2013
.
Neuropilin 1 guides regulatory T cells into VEGF-producing melanoma.
OncoImmunology
2
:
e23039
.
25.
Pere
H.
,
C.
Tanchot
,
J.
Bayry
,
M.
Terme
,
J.
Taieb
,
C.
Badoual
,
O.
Adotevi
,
N.
Merillon
,
E.
Marcheteau
,
V. R.
Quillien
, et al
2012
.
Comprehensive analysis of current approaches to inhibit regulatory T cells in cancer.
OncoImmunology
1
:
326
333
.

The authors have no financial conflicts of interest.

Supplementary data