Th17 cells have been investigated in mice primarily for their contributions to autoimmune diseases. However, the pathways of differentiation of Th17 and related Th cells (type 17 cells) and the structure of the type 17 memory population in humans are not well understood; such understanding is critical for manipulating these cells in vivo. By exploiting differences in levels of surface CCR6, we found that human type 17 memory cells, including individual T cell clonotypes, form an elongated continuum of type 17 character along which cells can be driven by increasing RORγt. This continuum includes cells preserved within the memory pool with potentials that reflect the early preferential activation of multiple over single lineages. The phenotypes and epigenomes of CCR6+ cells are stable across cell divisions under noninflammatory conditions. Nonetheless, activation in polarizing and nonpolarizing conditions can yield additional functionalities, revealing, respectively, both environmentally induced and imprinted mechanisms that contribute differentially across the type 17 continuum to yield the unusual plasticity ascribed to type 17 cells.

Immunological memory mediated by the adaptive immune system can provide long-standing natural and vaccine-mediated protection against infectious agents. Notwithstanding controversies regarding the meaning and effectiveness of the T cell component of immunological memory (1), pathogen-specific CD8+ and CD4+ T cells can be detected in humans many decades after exposure (2, 3), and for CD4+ memory T cells, which are the focus of the current studies, there are data from mice demonstrating their protective capabilities (4, 5). T cell memory depends on epigenetic memory, including DNA methylation, with resting memory T cells maintaining an altered epigenome that is poised to allow cells to respond to activation in a biased but flexible fashion based on the pattern of initial differentiation/polarization (6).

The ontogeny of memory T cells has long been a subject of ongoing investigation and debate (7). It is clear that descendants of individual CD8+ T cells can assume multiple phenotypes of effector and memory cells (8–10). Further, based on studies of human CD4+ memory T cells specific for pathogens or tetanus toxoid, individual T cells can and often do give rise to cells of different lineages as defined by Th1, Th2, and Th17 signature cytokines (11). Moreover, the preponderance of evidence from mouse and human studies favors effector T cells as the direct precursors of memory cells (8, 12–18). Critical experiments using single-cell transfers in mice have shown that the ability to form memory cells is a general and not exceptional property of individual naive CD4+ T cells, and that, of equal importance, the composition of the memory population mirrors the composition of the effector population (18). Moreover, studies of human memory CD8+ T cells have suggested that phenotypes and epigenomes are stable under homeostatic conditions (19). The correspondence between effector and memory populations provides a fundamental benefit often lacking in models explaining the origins of T cell memory, namely, that the information contained in the acute and evolving effector response is fully preserved in the memory population. A further consequence of these findings is that the memory T cell population preserves a record of the acute response and can be used to reconstruct pathways of differentiation taken by activated T cells. The ability to use studies of the memory population in this way is particularly important in understanding the process of memory T cell differentiation in humans, where it is both the case that the time frames and complexity of environmental exposures far exceed those of laboratory mice (20, 21) and that detailed observations of immune responses in real time are difficult.

Based on the conventional definitions of CD4+ T cell effector/memory subsets, Th17 cells were the third major subset to be identified as a “lineage,” with IL-17A as their signature cytokine and RORγt as their signature transcription factor (22). In initial studies in mice, Th17 cells were found to contribute to autoimmune pathology (23), and subsequent work in humans demonstrated the importance of the Th17 pathway in host defense against Candida, Staphylococcus aureus, and Mycobacterium (24–27), as well as in inflammatory disease (28). Further, mouse studies revealed that Th17 cells can assume dichotomous phenotypes as both protective and pathogenic (29), and that although some studies in mice have questioned the durability of Th17 memory (30), other studies in mice (31) and humans (32) have reported that Th17 cells can establish persistent memory.

Prominent features described for mouse Th17 cells include their abilities to coexpress genes of alternative lineages, lose expression of IL-17A/F, and acquire the ability to express a different signature cytokine, primarily IFN-γ (33–37), manifesting properties broadly described as plasticity (37). Nonetheless, whether plasticity in cytokine profiles indicates that a fully differentiated memory cell is able to transdifferentiate remains controversial (38), and some studies of Th17 cells in mice have shown that their phenotype can be stable during both homeostasis and inflammation (39, 40). Investigations of the phenotype and epigenome of human Th17 cells suggest that they may be more stable than their murine counterparts (41), and that IFN-γ+IL-17A cells that arise under Th17 conditions are best viewed as a subset of Th17 cells rather than as Th1 cells that have arisen by transdifferentiation (42, 43).

The chemokine receptor CCR6 and its ligand, CCL20, have broad roles in humoral (44) and cellular (45) barrier defense. We and others reported that virtually all human Th17 cells express CCR6 (46–48), which is inducible on T cells by RORγt (49). Cells expressing other cytokines whose expression is associated with, but not precisely matching the expression of IL-17A, such as IL-22 and CCL20, are also found exclusively as subsets of CCR6+ cells (46, 50, and this article), and CCR6 is found on a subset of human peripheral regulatory T cells (51) that can be stimulated to produce IL-17A (52). In addition, CCR6 is used to identify the human Th17-related, IFN-γ+IL-17A cells that have been referred to as nonclassical Th1 (43) or Th1* (46) cells and that have been implicated in defense against mycobacteria (26) and autoimmune disease (53). Together, these data suggest that human Th17 cells that are defined by their production of IL-17A/F represent a limited subset within a broader family of “type 17,” CCR6-expressing cells. Our studies started from the premise that investigating the interrelationships of cells within this larger type 17 family would reveal how human Th17 and Th17-related effector/memory cells are formed and organized, thereby providing additional insights into their roles in host defense and inflammation.

We discovered previously in analyzing relationships between the expression of signature chemokine receptors and a limited number of other lineage-associated genes in bulk populations of human Th cells that levels of surface CCR6 correlated positively with levels of expression of IL17A and RORC, and that the CCR6+ cells showed a wide gradient in their expression of these genes as compared with gradients for expression of type 1 genes in CXCR3+ cells or type 2 genes in CCR4+ cells (54). In this study, by analyzing human CD4+ T cells based on their levels of surface CCR6, both as populations and as single cells, we have found that the CCR6+ cells form an extended, progressive continuum of type 17 character, with cells of individual T cell clonotypes found throughout the continuum, albeit with changing frequencies. Although cells could be driven along the continuum by overexpression of RORC/RORγt, analysis of genome-wide patterns of CpG methylation in dividing cells showed that epigenotypes/phenotypes are stable under noninflammatory conditions. Single-cell analysis suggested pan-potentiation of effector pathways early after activation of CD4+ T cells in vivo that persist within the memory population, and most type 17 cells can coexpress genes of alternative lineages, with coexpression diminishing along the type 17 continuum. Restimulation of CCR6-defined subgroups under polarizing and nonpolarizing conditions yielded additional functionalities by both environmentally induced and previously imprinted mechanisms, which contributed differentially across the type 17 continuum to type 17 cell plasticity.

Elutriated lymphocytes and whole blood were obtained from healthy donors by the Department of Transfusion Medicine, National Institutes of Health, under an Institutional Review Board–approved protocol.

CD4+ T cells were isolated from elutriated lymphocytes by negative selection using RosetteSep CD4+ T cell enrichment mixture (StemCell Technologies) followed by Ficoll/Hypaque (Amersham Biosciences) density centrifugation according to the protocol from StemCell Technologies. If necessary, RBCs were removed using Pharm Lyse (BD Biosciences) according to the manufacturer’s protocol. For staining, cells were suspended in 250 μl of HBSS (Mediatech) containing 4% FBS (GeminiBio) and incubated with the following Abs: anti-CD4, anti-CD45RO, and anti-CCR6 for 30 min at room temperature. In some experiments, in addition to these Abs, cells were incubated with anti-CD25, anti-CD62L, anti-CXCR5, anti-CXCR3, and anti-CCR4 Abs. Abs were purchased from eBioscience and all were against human Ags. Cells were washed and resuspended in HBSS plus 4% FBS. CD4+ T cells were sorted into CD45RO (naive) and CD45RO+ subsets, and the CD45RO+ subset was further separated into CCR6neg and CCR6pos, and within the CCR6pos cells, into CCR6low, CCR6int, and CCR6high subsets. For in vitro polarization experiments, cells were sorted into naive (CD4+CD45ROCD25CXCR5CCR4CXCR3), CCR6neg (CD4+CD45RO+CD25CXCR5CXCR3CCR6 or CD4+CD45RO+CD25CXCR5 CCR4CCR6), CCR6low (CD4+CD45RO+CD25CXCR5CXCR3CCR6low or CD4+CD45RO+CD25CXCR5CCR4CCR6low), and CCR6high (CD4+CD45RO+CD25CXCR5CXCR3CCR6high or CD4+CD45RO+CD25CXCR5CCR4CCR6high). All cell sorting was done using a FACSAria Cell Sorter (BD Biosciences). By postsort analysis, the purity of populations was 95–99%. Cells were used for experiments immediately after sorting.

T cells were activated with 20 ng/ml PMA and 1 mM ionomycin in RPMI 1640 (Life Technologies-Thermo Fisher Scientific) supplemented with 10% FBS, 1% l-glutamine, and 1% penicillin/streptomycin (complete medium) for 3 h at 37°C under 5% CO2, total cellular RNA was purified using PureLink RNA kit (Life Technology) or RNeasy Mini Kit (Qiagen), and the concentration and RNA Integrity Number was determined with an RNA 6000 nano kit on 2100 Bioanalyzer instrument (Agilent Technologies). Real-time PCR was performed on samples in duplicate on an ABI 7900HT Real-Time PCR System (Applied Biosystems) or Eco Real-Time PCR System (Illumina) using SuperScript One Step RT-PCR kit (Life Technology). For most experiments, primer and probe sets (FAM/MGB-labeled) were purchased from Applied Biosystems. In some cases, DELTAgene Assay primers were purchased from Fluidigm, and amplicon DNA was quantified using SsoFast EvaGreen Supermix with Low ROX (Bio-Rad Laboratories). Gene expression data were normalized based on the values for GAPDH mRNA. For some assays, for cells from each donor, relative levels of expression, based on values for 2−ΔCT, are shown after normalization to the single highest value, which was set to 100. For other assays, values of 2−ΔCT or ΔCT are shown without additional normalization or shown as fold changes relative to the values for naive cells, which were set at 1.

T cells purified by cell sorting were stimulated with Leukocyte Activation Cocktail with GolgiPlus (BD Pharmingen) for 6 h at 37°C under 5% CO2 before being stained with the following Abs alone or in combinations: anti–IL-17A (eBioscience), anti–IL-17F, anti–IL-22, anti–IL-6, anti-CCL20, and anti–IFN-γ (R&D Systems) by using Cytofix/CytoPerm Plus kit (BD Pharmingen) or anti-RORγt, anti-FOXP3, anti–T-bet, or anti-GATA3 using Foxp3 staining buffer set (eBioscience) according to the manufacturer’s instructions. All flow cytometry was done using FACSCalibur, LSR II System, or Fortessa flow cytometers (BD Biosciences), and the data were subsequently analyzed using FlowJo software (TreeStar).

T cells purified by cell sorting were activated with 20 ng/ml PMA and 1 mM ionomycin in complete medium at 37°C under 5% CO2 for 12 h, and supernatants were collected and stored in −80°C freezer until analyzed. GM-CSF, TNF-α, IFN-γ, IL-4, IL-5, IL-10, IL-13, IL-17A, CCL3, CCL4, CCL5, and CCL20 were measured using Search Light Proteome Array (Pierce Biotechnology).

FACS-sorted naive and memory CCR6neg, CCR6low, CCR6int, and CCR6high subsets of CD4+ T cells were used for isolating RNA without and with prior activation with 20 ng/ml PMA and 1 mM ionomycin for 3 h at 37°C under 5% CO2 using RNeasy Mini Kit (Qiagen), and gene expression levels were determined using Illumina Human WG6 Expression BeadChip Version 3 arrays (Illumina) according to the manufacturer’s protocol. In brief, 500 ng of total RNA was used to make cDNA and label cRNA using the Ambion Illumina TotalPrep RNA Amplification kit (Invitrogen). The resulting cRNA samples were hybridized to an Illumina Human WG6 Expression BeadChip array for 14–20 h at 58°C in an Illumina Hybridization Oven (Illumina). Arrays were washed and scanned following the protocols in the Illumina Whole-Genome Gene Expression Direct Hybridization Assay Guide (Illumina). Signal data were extracted from the image files with the Gene Expression module of the GenomeStudio software (Illumina). Signal intensities were converted to log2 scale. Detection p values were calculated as described in the GenomeStudio Gene Expression Module User Guide. Data for array probes with insufficient signal were removed from the dataset (detection was defined as at least two arrays with detection p < 0.1). ANOVA was then performed on the normalized log2 expression estimates to test for mRNA expression differences across cell subgroups, with contrasts for the CCR6low, CCR6int, and CCR6high subsets versus the CCR6neg subgroup. A p value of 0.05 was used for the significance cutoff, after adjusting for the familywise error rate using the Benjamini-Hochberg method to account for multiple testing. One-way hierarchical clustering (on probes, i.e., gene-wise) was performed on the standardized LS means for all genes whose differential expression reached significance, separately for stimulated and nonstimulated cells. Genes whose differential expression reached significance and whose differences were the most extreme (>4 SDs [|Z| > 4] of log2 differences within any contrast) were then analyzed to determine patterns across the CCR6-expressing subsets. Log2 differences from the contrasts were standardized across each probe, and the resulting z scores were clustered by K-means, where k = 12. Genes within each cluster for each contrast were summarized as mean z scores and shown as a heatmap to illustrate relative expression patterns across the CCR6-expressing subsets. Statistical analyses were performed using JMP/Genomics software version 7.0 (SAS Institute).

Genomic DNAs from FACS-sorted naive and memory CCR6neg, CCR6low, CCR6int, and CCR6high subgroups were purified using the DNeasy kit (Qiagen). In some experiments, after purification by cell sorting, a portion of cells was used as starting cells (day 0) and the remaining cells were labeled with CFSE, activated with soluble anti-CD3 (100 ng/ml; eBioscience), and cultured with the homeostatic cytokines IL-7 and IL-15 (10 ng/ml each; R&D Systems) for 7 d. Cells from the different subgroups were again sorted, based on CFSE dilution, into nonproliferated and proliferated cells, and genomic DNAs were purified using the DNeasy kit. Bisulfite conversion of genomic DNA (500 ng) was performed using the Zymo Research EZ DNA Methylation kit (Zymo Research) according to the manufacturer’s recommendations for the Illumina Infinium assay. The bisulfite conversion incubation was processed using alternative thermocycler settings as outlined in the “Alternative Incubation Conditions When Using the Illumina Infinium® Methylation” section of the Zymo instruction manual appendix (chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://files.zymoresearch.com/protocols/_d5001_d5002_ez_dna_methylationga_o_kit.pdf). The bisulfite-converted DNA samples were then processed for hybridization and staining using either the Illumina Infinium Human Methylation 450 BeadChip Kit or the Illumina Infinium MethylationEPIC BeadChip Kit (Illumina) according to the manufacturer’s recommendations. DNA methylation ratios were calculated on a scale of 0–1, in which 0 represents 100% unmethylated and 1 represents 100% methylated. The methylation dataset was visualized using Tidyverse (an R open-source package).

Purified CD4+ T cells were FACS sorted into memory CCR6neg, CCR6low, CCR6int, and CCR6high subgroups, loaded with CFSE, and cultured with autologous dodecyldimethylamine oxide (DDAO)-loaded, irradiated PBMCs in the presence of an extract of CMV-infected cells (EastCoast Bio, North Berwick, ME). On day 7, the proliferated, CFSElow T cells were sorted, counted for cell viability, and immediately used for library preparation using Chromium Single Cell V (D) J Reagent Kit (10× Genomics) according to the manufacturer’s instructions.

The single-cell TCR libraries were sequenced on an Illumina NextSeq500 to a minimum sequencing depth of 5000 reads/cell using paired-end sequencing with single indexing as follows: read length of 150 bp read 1; 8 bp i7 index, 150 bp read 2. The raw data were downloaded using Python Run Downloader (Illumina) and analyzed using Cellranger (Illumina). TCR repertoires were processed using custom scripts to produce Circos-compatible files. In each Circos plot (Circos, version 0.69), the relative size of a clonal expansion is represented by arc length, whereas overlapping clonotypes are represented by ribbons connecting arcs. In some cases, to increase interpretability of the Circos plots, ribbons were drawn only between clonotypes that represented ≥4% of either of the connected TCR repertoires. Morisita-Horn dissimilarity values between TCR repertoires were calculated using the R package vegan (v. 2.5-7) and plotted in R using reciprocal values to represent similarity.

The single-cell 5′ gene expression libraries were sequenced on Illumina NextSeq500 to a minimum sequencing depth of 50,000 reads/cell using pair-end sequencing as follows: read length of 150 bp read 1; 8 bp i7 index, 150 bp read 2. The raw data were downloaded using Python Run Downloader (Illumina) and analyzed using Cell Ranger (Illumina).

CD4+ T cells were sorted into naive and memory CCR6neg, CCR6low, CCR6int, and CCR6high subgroups; total cellular RNAs were purified from using PureLink RNA kit (Life Technology); and the concentrations and RNA Integrity Numbers were determined with an RNA 6000 nano kit on a 2100 Bioanalyzer (Agilent Technologies). Libraries for TCR sequencing were prepared as described previously (55). TCRB annotation was performed by first determining which paired reads were of the expected library insert size by merging the paired-end reads with the bioinformatic tool, PEAR (56). Paired-end reads that could be merged were discarded because they represented library inserts that were too short and assumed to be off-target products. Unmerged read pairs were then annotated with MiXCR (57). Heatmaps and Circos plots were then generated in the same manner as earlier.

Individual cells from naive and memory CCR6neg, CCR6low, CCR6int, and CCR6high subgroups were sorted by FACS into single wells of a 96-well plate, where each well contained 5 μl of CellDirect 2× reaction mixture solution (Invitrogen) supplemented with 0.05 U of SUPERase-In RNase inhibitor (Ambion). After incubating at 4°C for 15 min, each plate containing sorted cells was frozen and stored at −30°C until analysis. For analysis, plates were thawed on ice, and each well was supplemented with 0.2 μl SuperScript III RT Platinum Taq Mix (Invitrogen), 2.8 μl DNA suspension buffer (Invitrogen), and 1.0 μl of a mixture of 48 pooled DELTAgene Assay primer pairs containing each primer at a final concentration of 500 nM (Fluidigm). mRNAs from lysates of single cells were reverse transcribed into cDNA at 50°C for 15 min, followed by inactivation of the reverse transcriptase at 95°C for 2 min. cDNAs were preamplified for 20 cycles by denaturing at 95°C for 15 s followed by annealing and amplification at 60°C for 4 min. To digest unincorporated primers, we added to each well 3.6 μl of Exonuclease I reaction mixture containing 0.36 μl of 10× Exonuclease reaction buffer, 0.72 μl of 20 U/μl Exonuclease I (New England BioLabs), and 2.52 μl of nuclease-free water, followed by an incubation at 37°C for 30 min, and finally treatment at 80°C for 15 min. The preamplified cDNA was diluted 5-fold with TE buffer (10 mM Tris [pH 8.0], 1 mM EDTA) before gene expression assays using the 96.96 (M96) Dynamic Array Microfluidic Chip on a BioMark HD System (Fluidigm) according to the manufacturer’s protocol. In brief, 2.25 μl of preamplified cDNA was mixed with 2.5 μl of 2× SsoFast EvaGreen Supermix with low ROX (Bio-Rad) and 0.25 μl of 20× DNA binding dye sample loading reagent (Fluidigm) and transferred into one of the array microfluidic chip sample inlets. Individual DELTAgene assays were prepared by combining 0.25 μl of primer pairs (100 μM), 2.25 μl of DNA suspension buffer, and 2.5 μl of 2× Assay Loading Reagent (Fluidigm) and transferring into one of the array microfluidic chip assay inlets. Samples and assay mix solutions were loaded on the microfluidic chip using an IFC Controller MX (for 48.48 Dynamic array IFC) or IFC Controller HX (for M96 Dynamic array IFC, Fluidigm) and transferred to a BioMark HD System and run following the manufacturer’s protocol. For each subgroup, cells were sorted from three individual donors. All cycle threshold (Ct) values obtained from the BioMark HD System were normalized to the endogenous control genes, GAPDH and PGK1, and cells with very low (Ct > 30) or nondetected endogenous control genes were removed from the analysis. For further analysis, normalized Ct values for individual genes were calculated by subtracting from the Ct value the mean Ct values of that gene on all the samples from a given subgroup on an individual chip and dividing by three times the SD as described previously (58). Data acquired from the BioMark HD system were processed using Singular Analysis Toolset Software v3.5.2 (Fluidigm). We used open-source R packages (Tidyverse) to process the datasets and visualization. In some experiments, preamplified samples were used for TaqMan quantitative PCR, which was performed on Eco Real-Time PCR System (Illumina) using gene-specific primers purchased from Fluidigm (DELTAgene Assay) and detected using SsoFast EvaGreen Supermix with Low ROX (Bio-Rad Laboratories). Results from these assays were normalized based on the values for GAPDH mRNA.

All the lentivirus particles expressing sequences from RORC, TBX21, and GATA3 or control virus particles contained the pReceiver-LV201 vector and were obtained from GeneCopoeia. A CMV promoter was used to drive the RORC, TBX21, and GATA3 sequences, and an SV40 promoter was used to drive enhanced GFP sequences. Purified CD4+ T cells were sorted by FACS into memory cells that were CCR6neg, CCR6low, and CCR6high and transduced with lentivirus particles at an MOI of 5 using retronectin (TaKaRa) according to the manufacturer’s instructions. In brief, 1 × 106 sorted cells/well of a 24-well plate were activated with anti-CD3/anti-CD28–coated beads, one bead per cell, for 24 h and transferred to a well precoated with retronectin (TaKaRa). Lentivirus transduction was performed by spinoculation at 1200 × g for 2 h at 32°C in the presence of virus particles and 8 μg/ml polybrene (Millipore). The next day, virus particles were removed, and cells were cultured in complete RPMI 1640 supplemented with 10 ng/ml IL-2 for an additional 3 d. Cells were stimulated with PMA and ionomycin for 3 h, and GFP+ cells were sorted by FACS. For single-cell gene expression experiments, cells were sorted by FACS into each well of a 96-well plate containing 5 μl of CellDirect 2× reaction mixture solution supplemented with 0.05 U of SUPERase-In RNase inhibitor and used for analysis using the M96 Dynamic Array Microfluidic Chip on a BioMark HD System according to the manufacturer’s protocol or analyzed using quantitative RT-PCR. For some experiments, cells were loaded with DDAO before lentivirus transduction to enable cells to be sorted based on numbers of cell divisions.

Cells were sorted by FACS as described earlier into naive and memory CCR6neg, CCR6low, and CCR6high memory subgroups and cultured at 1 × 106 cells/ml in 24-well plates in ImmunoCult XF T cell expansion medium (Stem Cell Technologies) supplemented with 10% FBS and IL-2 (200 U/ml). Stimulation was done using ImmunoCult human CD3/CD28/CD2 T Cell Activator (STEMCELL Technologies) in nonpolarizing conditions using anti–IL-4 (5 μg/ml) and anti–IFN-γ (10 μg/ml), in Th1 conditions using recombinant human IL-12 (5 ng/ml) and anti–IL-4 (4 μg/ml), or in Th2 conditions using recombinant human IL-4 (10 ng/ml) and anti–IFN-γ (10 μg/ml). After 3 d, proliferating cells were washed in PBS and expanded under the same conditions in the absence of CD3/CD28/CD2 T Cell Activator.

Data were analyzed using GraphPad Prism software version 9 (GraphPad, San Diego, CA). The p values were calculated using Wilcoxon matched pairs signed rank test. Linear regression in R was used to generate p values for the relationship between single-cell quantitative PCR Ct values for genes of interest. ANOVA was performed on the normalized log2 expression values to estimate differences between cell subgroups, and adjusted p values were generated using the Benjamini-Hochberg method to account for multiple testing: *p < 0.05, **p < 0.01, and ****p < 0.0001.

To investigate the relationship between the surface expression of CCR6 and genes associated with effector/memory phenotypes, we used FACS to separate CD4+ T cells from healthy adult blood donors into naive cells and memory-phenotype cells that were CCR6neg and CCR6pos, and the CCR6pos cells were further separated into three equal subgroups based on their levels of CCR6: CCR6low, CCR6int, or CCR6high (Fig. 1A). Intracellular staining of FACS-purified cells activated with PMA and ionomycin revealed that virtually all cells expressing type 17 cytokines/chemokines, such as IL-17A, IL-17F, IL-22, and CCL20, are found within the CCR6-expressing subset (Supplemental Fig. 1A, 1B). Building on our previous studies (54), using data from gene arrays in cells from eight donors, we analyzed changes in gene expression in cells that were either left nonstimulated or stimulated ex vivo with PMA and ionomycin. As shown in volcano plots in Fig. 1B, in comparing the CCR6+ subgroups with CCR6neg cells as a common comparator, we found that type 17 genes such as IL17A, IL17F, IL22, IL23R, CCL20, RORA, and RORC were progressively upregulated (shown in red), and type 1 genes such as IFNG, CXCR3, CCL3, CCL5, EOMES, GZMK, and GZMH and type 2 genes such as IL4 and IL5 were downregulated (shown in green), also generally progressively, in CCR6low to CCR6int to CCR6high cells. Pairwise comparisons of gene-specific probes among the subgroups revealed ∼600 genes that were differentially expressed in any of the comparisons for the nonstimulated and/or stimulated cells based on a false discovery rate of 0.05. It is important to note, as can be appreciated from the mRNAs of diminishing abundance across the subgroups, that the CCR6neg subgroup is particularly heterogeneous, containing, for example, highly differentiated Th1 and Th2 cells. All differentially expressed genes were grouped by k-means into 12 clusters with a variety of patterns across the CCR6low, CCR6int, and CCR6high subgroups. Genes in most clusters display monotonic changes in expression. A salient gene within each cluster is shown, such as the signature genes for the Th lineages, IL17F, IFNG, and IL5 (Fig. 1C). Heatmaps of gene expression from the naive and CCR6-defined (CCR6neg, CCR6low, CCR6int, and CCR6high) memory subgroups, both nonstimulated and stimulated, show large differences in patterns between naive and memory cells and between CCR6neg and CCR6pos cells and progressive changes from CCR6low to CCR6high cells (Supplemental Fig. 1C).

FIGURE 1.

Progressive upregulation of type 17 genes and downregulation of type 1 and type 2 genes in CCR6-expressing memory subgroups. (A) Elutriated lymphocytes from blood that had been enriched for CD4+ T cells were separated based on CD45RO staining into CD45RO (naive) and CD45RO+ (memory) cells, and the CD45RO+ cells were either separated into CCR6neg and CCR6pos subgroups or into CCR6neg, CCR6low, CCR6int, and CCR6high subgroups. (B) Volcano plots showing fold changes (log2 ratios) versus significance (−log10p values) for genes in nonstimulated (top) or stimulated (bottom) cells for CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup. The red dotted line marks the cutoff for significance of differential expression at a false discovery rate ≤0.05. On each volcano plot is indicated the number of genes upregulated (red) or downregulated (green), and some genes of interest are labeled. (C) Heatmaps illustrating the relative patterns of gene expression differences between CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup. The standardized differences for the gene-specific probes showing the greatest differences in these comparisons were clustered by K-means (Supplemental Fig. 1C); the color/value in each small box represents the mean across the genes in the cluster for each contrast. The cluster number, the number of unique gene-specific probes in that cluster, and a representative gene in each cluster are shown. Analyses shown are combined from four individual donors for the nonstimulated cells and eight individual donors for stimulated cells, including four common donors.

FIGURE 1.

Progressive upregulation of type 17 genes and downregulation of type 1 and type 2 genes in CCR6-expressing memory subgroups. (A) Elutriated lymphocytes from blood that had been enriched for CD4+ T cells were separated based on CD45RO staining into CD45RO (naive) and CD45RO+ (memory) cells, and the CD45RO+ cells were either separated into CCR6neg and CCR6pos subgroups or into CCR6neg, CCR6low, CCR6int, and CCR6high subgroups. (B) Volcano plots showing fold changes (log2 ratios) versus significance (−log10p values) for genes in nonstimulated (top) or stimulated (bottom) cells for CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup. The red dotted line marks the cutoff for significance of differential expression at a false discovery rate ≤0.05. On each volcano plot is indicated the number of genes upregulated (red) or downregulated (green), and some genes of interest are labeled. (C) Heatmaps illustrating the relative patterns of gene expression differences between CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup. The standardized differences for the gene-specific probes showing the greatest differences in these comparisons were clustered by K-means (Supplemental Fig. 1C); the color/value in each small box represents the mean across the genes in the cluster for each contrast. The cluster number, the number of unique gene-specific probes in that cluster, and a representative gene in each cluster are shown. Analyses shown are combined from four individual donors for the nonstimulated cells and eight individual donors for stimulated cells, including four common donors.

Close modal

We next performed a more targeted analysis of the gradients of lineage-related genes and proteins across the naive and CCR6-defined memory subgroups. Consistent with the gene array data, real-time RT-PCR of cells activated with PMA and ionomycin showed increases in expression of type 17 genes and decreases in signature type 1 (IFNG and TBX21) and type 2 genes (IL4, IL5, IL13, and GATA3) across the CCR6-defined memory subgroups (Fig. 2A). Similar patterns were found for lineage-associated proteins as detected by intracellular staining or ELISA (Fig. 2B, Supplemental Fig. 2A). We further investigated pairwise relationships among type 17 proteins across the naive and CCR6-defined memory subgroups by intracellular staining. For IL-17F, as depicted in Supplemental Figs. 1A and 2B, virtually all positive-staining cells were found as a subset of cells expressing IL-17A. For IL-17A/F versus CCL20, both CCL20+IL-17A/F and CCL20+IL-17A/F+ cells increased from the CCR6low to the CCR6high subgroup, although the fold increase in CCL20+IL-17A/F+ cells was greater across the subgroups, leading to an increase in the ratio of IL-17A/F+CCL20+ to IL-17A/FCCL20+ cells across the subgroups. For IL-17A versus IL-22, there was a significant increase in the frequency of IL-22+IL-17A+ cells from the CCR6low to the CCR6high subgroup (although the ratio of IL-17A+IL-22+ to IL-17A+IL-22 cells was constant across the subgroups), but very little change in the frequency of IL-22+IL-17A cells, suggesting that IL-22 may be upregulated through different pathways in cells that do versus cells that do not coexpress IL-17A. For CCL20 versus IL-22, based on the percentages of single-positive versus double-positive cells, their increases across the subgroups showed no evidence of being coregulated (Supplemental Fig. 2B). In contrast with the patterns for the type 17 cytokines, total percentages of IFN-γ+ cells declined across the subgroups, although there was an increase in IFN-γ+IL-17A+ cells. Taken together, these data support the value of CCR6 as a marker of activity of the broad type 17 program in which type 17 genes show varying degrees and pathways of coordinated versus independent regulation, demonstrate increasing type 17 and decreasing type 1 and type 2 character across the CCR6-defined memory subgroups, and suggest that the patterns of gene expression and resulting phenotypes for the type 17 cells progress along an extended continuum.

FIGURE 2.

Gradients of expression of lineage-associated genes with increasing CCR6. (A) Relative mRNA expression was measured using real-time RT-PCR. Ct values of all genes were normalized to GAPDH, and relative mRNA expression is shown for means (left panel) or for individual values (right panel) from four or more donors as percentages of the value in the highest expressing subgroup. (B) Protein levels in naive and CCR6-defined memory subgroups for IL-2, IL-4, IL-5, and IL-13 were measured by ELISA from cell supernatants after 12 h of activation with PMA and ionomycin. For other proteins, cells were stimulated with the leukocyte activation mixture for 6 h, fixed and permeabilized, stained for indicated genes, and analyzed by flow cytometry for percent of cells staining positive. In the left panel, means are shown of values from three or more donors, calculated as percentages of the value in the highest expressing subgroup. The right panel shows percentages of positive-staining cells from five or more donors. Symbols in the right panels indicate data from individual donors, and bars indicate means. p values were determined using the Wilcoxon matched pairs signed rank test. **p < 0.01, ****p < 0.001.

FIGURE 2.

Gradients of expression of lineage-associated genes with increasing CCR6. (A) Relative mRNA expression was measured using real-time RT-PCR. Ct values of all genes were normalized to GAPDH, and relative mRNA expression is shown for means (left panel) or for individual values (right panel) from four or more donors as percentages of the value in the highest expressing subgroup. (B) Protein levels in naive and CCR6-defined memory subgroups for IL-2, IL-4, IL-5, and IL-13 were measured by ELISA from cell supernatants after 12 h of activation with PMA and ionomycin. For other proteins, cells were stimulated with the leukocyte activation mixture for 6 h, fixed and permeabilized, stained for indicated genes, and analyzed by flow cytometry for percent of cells staining positive. In the left panel, means are shown of values from three or more donors, calculated as percentages of the value in the highest expressing subgroup. The right panel shows percentages of positive-staining cells from five or more donors. Symbols in the right panels indicate data from individual donors, and bars indicate means. p values were determined using the Wilcoxon matched pairs signed rank test. **p < 0.01, ****p < 0.001.

Close modal

For genes/proteins other than CCR6, the patterns that we described for type 17 genes across the CCR6-defined memory subgroups could represent mixtures of homogeneous subsets of, for example, negative and positive cells in varying proportions versus individual cells with ranges of phenotypes, or some combination thereof. We addressed this issue by using Fluidigm Dynamic Arrays for RT-PCR, which allowed quantitative analysis of expression of a panel of lineage-associated, activation-associated, and housekeeping genes in single cells from each of the naive and memory CCR6neg, CCR6low, CCR6int, and CCR6high memory subgroups. Hierarchical clustering of the normalized expression profiles shows coexpression in single cells of type 1 (IFNG, TBX21, TNF, and CXCR3; highlighted in blue), Treg-associated (IKZF2 and FOXP3; highlighted in green), type 2 (ETS1, GATA3, IL13, and IL4; highlighted in orange), and type 17 (RORA, KLRB1, CCR6, IL22, IL17A, IL23R, RORC, and CCL20; highlighted in red) genes (Fig. 3A). Using mean values of gene expression, calculations of correlation coefficients between pairs of the CCR6-defined memory subgroups revealed the ordered relationship of CCR6low to CCR6int to CCR6high (Fig. 3B), consistent with the analyses from the bulk subgroups.

FIGURE 3.

Single-cell data reveal a continuum of expression levels among lineage-associated genes. (A) Heatmap representation of gene expression profiles using hierarchical clustering showing coexpression of lineage-specific genes in individual cells across naive cells and CCR6-defined memory subgroups for 282 cells per subgroup. ΔCt values were calculated using the average of housekeeping genes GAPDH and PGK1, and after excluding genes with low and/or inconsistent expression, a total of 38 nonhousekeeping genes were included in the final analysis. Rows are genes, and columns are cells. Some clustered, lineage-associated genes are indicated. (B) Correlations among CCR6low, CCR6int, and CCR6high memory subgroups based on means of expression of the genes in (A). (C) Gene expression values normalized as described in Materials and Methods for BioMark HD data for the 1410 cells from naive and CCR6-defined memory subgroups. Each symbol represents a single cell, and each cell is colored based on its subgroup as indicated. The 1410 cells analyzed in (A)–(C) were from three donors.

FIGURE 3.

Single-cell data reveal a continuum of expression levels among lineage-associated genes. (A) Heatmap representation of gene expression profiles using hierarchical clustering showing coexpression of lineage-specific genes in individual cells across naive cells and CCR6-defined memory subgroups for 282 cells per subgroup. ΔCt values were calculated using the average of housekeeping genes GAPDH and PGK1, and after excluding genes with low and/or inconsistent expression, a total of 38 nonhousekeeping genes were included in the final analysis. Rows are genes, and columns are cells. Some clustered, lineage-associated genes are indicated. (B) Correlations among CCR6low, CCR6int, and CCR6high memory subgroups based on means of expression of the genes in (A). (C) Gene expression values normalized as described in Materials and Methods for BioMark HD data for the 1410 cells from naive and CCR6-defined memory subgroups. Each symbol represents a single cell, and each cell is colored based on its subgroup as indicated. The 1410 cells analyzed in (A)–(C) were from three donors.

Close modal

For the type 17 genes, IL17A, CCL20, IL22, and RORC, the numbers of cells with detectable expression and the levels of expression within single cells increased from the CCR6low to CCR6high subgroups (Fig. 3C). The data suggest that for the type 1 and type 2 genes, the decreases in expression along the CCR6 continuum in analysis of the bulk population were predominantly due to decreases in the numbers of expressing cells (IFNG and TBX21), the expression levels per cell (IL4 and IL13), or both (IL5). Together these data demonstrate that the gradient of type 17 character among the CCR6-defined memory subgroups at the level of populations reflects a continuum at the level of individual cells, with single cells showing many intermediate phenotypes characterized by a wide range in their expression of signature type 17 genes, as well as signature genes of alternative lineages.

We investigated the stability of phenotypes across the type 17 continuum by analyzing genome-wide CpG methylation. Methylation of cytosines at CpG dinucleotides is the epigenetic change best associated with stability in patterns of gene expression and reflects the reshaping of these patterns in memory T cells (59, 60). The global methylation profiles reveal a progressive shift from higher to lower CpG methylation from naive to CCR6high cells (Fig. 4A). The corresponding unsupervised principal component analysis shows the large distance between the naive and memory cells along principal component 1 and the separation along principal component 2 in the order from CCR6neg to CCR6low to CCR6int to CCR6high (Fig. 4B). Volcano plots of changes in overall methylation levels at individual sites for each of the CCR6-expressing subgroups versus the CCR6neg cells show the predominance of demethylated versus hypermethylated sites from the CCR6low to CCR6high cells and the prominent demethylation at multiple type 17 genes, such as CCR6, IL-17A, IL-17RE, IL-26, IL-22, CCL20, RORC, RORA, FURIN, SGK1, and RUNX3 (Fig. 4C). To visualize better the changes in methylation status at genes of interest across the CCR6 subgroups, we identified the gene-associated CpG sites showing the greatest variabilities in their β values across the subgroups and displayed these values in each subset (Fig. 4D). We found progressively lower overall β values at sites within multiple type 17 genes, and either little change or progressively higher β values at sites within type 1 and type 2 genes, such as CXCR3, IFNG, TBX21, IL4, and IL13, along the continuum from naive to CCR6high cells.

FIGURE 4.

Progressive changes in CpG methylation across CCR6-defined subgroups. (A) Distribution of β values across 450,000 CpG sites using the Illumina Infinium Methylation Assay with naive cells and the CCR6-defined memory subgroups. β values (ratios) of 0.0 and 1.0 correspond to 0 and 100% methylation, respectively, of the sequences at a given CpG site. Data were combined from six donors. (B) Correlation and principal variance component projections of naive cells and CCR6-defined memory subgroups based on genome-wide CpG methylation using the Illumina Infinium Methylation Assay. (C) Volcano plots showing differences in methylation β values at CpG sites in the CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup versus significance (−log10 p values). The red dotted line marks the cutoff for significance of differences between β values at a false discovery rate ≤0.05. Some CpG sites at genes of interest are labeled. (D) For each of the indicated genes, box plots of methylation β values are displayed for the single CpG site with the greatest variability among the cell subgroups. For (A)–(D), data from six donors were combined for the analysis.

FIGURE 4.

Progressive changes in CpG methylation across CCR6-defined subgroups. (A) Distribution of β values across 450,000 CpG sites using the Illumina Infinium Methylation Assay with naive cells and the CCR6-defined memory subgroups. β values (ratios) of 0.0 and 1.0 correspond to 0 and 100% methylation, respectively, of the sequences at a given CpG site. Data were combined from six donors. (B) Correlation and principal variance component projections of naive cells and CCR6-defined memory subgroups based on genome-wide CpG methylation using the Illumina Infinium Methylation Assay. (C) Volcano plots showing differences in methylation β values at CpG sites in the CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup versus significance (−log10 p values). The red dotted line marks the cutoff for significance of differences between β values at a false discovery rate ≤0.05. Some CpG sites at genes of interest are labeled. (D) For each of the indicated genes, box plots of methylation β values are displayed for the single CpG site with the greatest variability among the cell subgroups. For (A)–(D), data from six donors were combined for the analysis.

Close modal

We were interested to know whether the cells’ phenotypes and genotypes were stable under homeostatic conditions. We reasoned that evaluating cells that had divided in a noninflammatory environment, under conditions mimicking homeostatic proliferation and more challenging than nonproliferative homeostasis, would serve as a more than adequate test. To this end, we activated CFSE-loaded naive and memory CCR6neg, CCR6low, CCR6int, and CCR6high cells using low concentrations of anti-CD3 plus homeostatic cytokines for 7 d and purified nonproliferated and proliferated cells (Fig. 5A). To check whether there had been a change in the cells’ type 17 character, we compared expression of IL17A, RORC, and CCL20 from day 0 and day 7 proliferated cells after activation with PMA and ionomycin. Proliferation under these conditions did not change the expression of these genes (Fig. 5B). We analyzed methylation states in cells at day 0 and on day 7 in cells that had either failed to divide (d7NP) or divided three or more times (d7P). In general, principal component analysis positioned the memory cell samples according to their CCR6-defined subgroup in a pattern like that in Fig. 4B, regardless of whether the cells had proliferated (Fig. 5C). Comparing each of the CCR6-expressing subgroups versus the CCR6neg cells at day 0 (Fig. 5D, upper panel) showed a similar pattern as in Fig. 4C, whereas the methylation levels at individual sites for day 0 versus day 7 proliferated cells for each of the CCR6-expressing subgroups showed no significant changes across the genome (Fig. 5D, lower panel). Similarly, for genes of interest, methylation at the single CpG sites that displayed the greatest variability across the naive and CCR6-defined memory subgroups showed no significant changes resulting from culturing or proliferation over 7 d (Fig. 5E). Together, these data demonstrate that the progressive pattern in lineage-related gene expression across the CCR6-expressing memory subgroups correspond to epigenetic changes at these loci, and the genome-wide and gene-specific patterns are unchanged across cell divisions under noninflammatory conditions, suggesting that the cells’ phenotypes will remain stable during homeostasis in vivo.

FIGURE 5.

Stable epigenomes of CCR6-defined memory subgroups across cell divisions. (A) After purification by cell sorting, cells were loaded with CFSE, activated with soluble anti-CD3, and cultured in the presence of IL-7 and IL-15 for 7 d, after which cells were again FACS purified into nonproliferated (d7NP) cells and cells that had undergone three or more cell divisions (d7P) according to the gating shown in these density-colored dot plots. Data are from one of the two donors used for these experiments. (B) Relative mRNA expression of IL17A, RORC, and CCL20 in cell subgroups in the starting cells at day 0 (d0) and d7P cells. Cells were stimulated with PMA and ionomycin for 3 h, and mRNAs were measured using RT-PCR. Values for these genes were normalized to values for GAPDH and then to the results for naive cells, which were set at 1. Bars indicate means. Data are shown for cells from the two donors used for these experiments. (C) Correlation and principal variance component projections based on the CpG methylation data of cell subgroups for d0, d7NP, and d7P cells. (D) Volcano plots showing differences in methylation β values between CpG sites in the CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup at day 0 (top) or for each subgroup between d0 and d7P cells (bottom). The red dotted lines mark the cutoffs for CpG sites with significant differences at a false discovery rate ≤0.05. Some CpG sites at genes of interest are labeled. (E) Box plots of methylation β values for the single CpG site with the greatest variability among the cell subgroups at lineage-associated genes for d0, d7NP, and d7P cells. Data for (C)–(E) were combined from two donors.

FIGURE 5.

Stable epigenomes of CCR6-defined memory subgroups across cell divisions. (A) After purification by cell sorting, cells were loaded with CFSE, activated with soluble anti-CD3, and cultured in the presence of IL-7 and IL-15 for 7 d, after which cells were again FACS purified into nonproliferated (d7NP) cells and cells that had undergone three or more cell divisions (d7P) according to the gating shown in these density-colored dot plots. Data are from one of the two donors used for these experiments. (B) Relative mRNA expression of IL17A, RORC, and CCL20 in cell subgroups in the starting cells at day 0 (d0) and d7P cells. Cells were stimulated with PMA and ionomycin for 3 h, and mRNAs were measured using RT-PCR. Values for these genes were normalized to values for GAPDH and then to the results for naive cells, which were set at 1. Bars indicate means. Data are shown for cells from the two donors used for these experiments. (C) Correlation and principal variance component projections based on the CpG methylation data of cell subgroups for d0, d7NP, and d7P cells. (D) Volcano plots showing differences in methylation β values between CpG sites in the CCR6low, CCR6int, and CCR6high subgroups compared with the CCR6neg subgroup at day 0 (top) or for each subgroup between d0 and d7P cells (bottom). The red dotted lines mark the cutoffs for CpG sites with significant differences at a false discovery rate ≤0.05. Some CpG sites at genes of interest are labeled. (E) Box plots of methylation β values for the single CpG site with the greatest variability among the cell subgroups at lineage-associated genes for d0, d7NP, and d7P cells. Data for (C)–(E) were combined from two donors.

Close modal

To investigate the clonal relationships among the CCR6-expressing memory cells, we determined the sequences of TCRβ-chain mRNAs from naive cells and CCR6-defined memory subgroups. Using the Morisita-Horn method to compute similarity, we found that the naive cells show little similarity with the memory cells and the memory subgroups are ordered from CCR6neg to CCR6low to CCR6int to CCR6high (Fig. 6A, left panels). Circos plots of the 20 most abundant clonotypes in each subgroup reveal that the frequencies of shared clonotypes are most similar between adjacent subgroups and differ most between nonadjacent subgroups (Fig. 6A, right panels). For example, several clonotypes are found at high frequencies in both the CCR6int and CCR6high subgroups, whereas clonotypes shared between the CCR6low and CCR6high cells that are present at high frequencies in the former are generally at low frequencies in the latter and vice versa.

FIGURE 6.

TCR sequencing reveals shared clonotypes whose frequencies change across CCR6-defined memory subgroups. (A) Two million T cells from each subgroup from two donors were used for sequencing TCRβ-chain cDNAs in bulk at a depth of 6.3–9.3 reads/cell and V(D)J sequences were used for identifying clonotypes. Morisita-Horn similarity indexes were calculated for pairwise comparisons to yield values between 0 (lowest similarity) and 1 (left panels). Circos plots display all clonotypes with the 20 highest frequencies in any subgroup (right panels). Bars in the circle have widths that are proportional to clonotype frequencies and are arranged within the cell subgroups clockwise in order of decreasing frequencies. Ribbons show connections between clonotypes among the cell subgroups for clonotypes with frequencies of ≥0.04 within any cell subgroup. (B) Cells from each CCR6-defined memory T cell subgroup from a CMV-seropositive donor that proliferated in response to CMV Ags were used for single-cell sequencing of TCRα- and β-chains. Paired productive sequences were obtained from ∼4000–6000 cells from each subgroup. Morisita-Horn similarity indexes were calculated for pairwise comparisons as in (A) (left panel). A Circos plot was produced using all clonotypes (right panel). Bars in the circle have widths that are proportional to clonotype frequencies and are arranged within the cell subgroups clockwise in order of decreasing frequencies. Ribbons show connections between clonotypes among the cell subgroups for clonotypes with frequencies ≥0.03 within any cell subgroup.

FIGURE 6.

TCR sequencing reveals shared clonotypes whose frequencies change across CCR6-defined memory subgroups. (A) Two million T cells from each subgroup from two donors were used for sequencing TCRβ-chain cDNAs in bulk at a depth of 6.3–9.3 reads/cell and V(D)J sequences were used for identifying clonotypes. Morisita-Horn similarity indexes were calculated for pairwise comparisons to yield values between 0 (lowest similarity) and 1 (left panels). Circos plots display all clonotypes with the 20 highest frequencies in any subgroup (right panels). Bars in the circle have widths that are proportional to clonotype frequencies and are arranged within the cell subgroups clockwise in order of decreasing frequencies. Ribbons show connections between clonotypes among the cell subgroups for clonotypes with frequencies of ≥0.04 within any cell subgroup. (B) Cells from each CCR6-defined memory T cell subgroup from a CMV-seropositive donor that proliferated in response to CMV Ags were used for single-cell sequencing of TCRα- and β-chains. Paired productive sequences were obtained from ∼4000–6000 cells from each subgroup. Morisita-Horn similarity indexes were calculated for pairwise comparisons as in (A) (left panel). A Circos plot was produced using all clonotypes (right panel). Bars in the circle have widths that are proportional to clonotype frequencies and are arranged within the cell subgroups clockwise in order of decreasing frequencies. Ribbons show connections between clonotypes among the cell subgroups for clonotypes with frequencies ≥0.03 within any cell subgroup.

Close modal

In addition to analyzing clonotypes using TCRβ-chain sequences from the bulk populations, we determined clonotype identities with increased specificity by sequencing both TCRα- and TCRβ-chain mRNAs from single cells. Given the limitations in sample sizes with this method and the significant diversity of the repertoire of nonselected memory T cells, we analyzed cells from the CCR6-defined memory subgroups purified from a donor seropositive for CMV based on the cells’ proliferative responses to CMV Ags (Supplemental Fig. 3A). We processed a total of 27,645 cells and recovered paired, productive sequences for TCRα- and TCRβ-chains from ∼67–71% of the cells, depending on the subgroup (Supplemental Fig. 3B, 3C), and only these sequences were used for subsequent analysis. Computing similarities again ordered the subgroups from CCR6neg to CCR6low to CCR6int to CCR6high (Fig. 6B, left panel). Circos plots including all the clonotypes (Fig. 6B, right panel) or the 20 most abundant clonotypes (Supplemental Fig. 3D) show that frequencies of shared clonotypes are most similar between adjacent subgroups and differ most between nonadjacent subgroups. These analyses of both the nonselected and CMV-reactive clonotypes reinforce the validity of considering all the CCR6-expressing Th cells as one family, because they show, for example, that the CCR6low subgroup is closer to the CCR6int subgroup than to the CCR6neg cells and demonstrate the same relationships among subgroups as reflected in the gene and protein expression data. Together, the results suggest that at least some cells within the subgroups originate from common T cell clones whose progeny have differentiated, with dynamic changes in their frequencies, along the continuum of type 17 character. To lend further support to the directionality of this pathway, we performed single-cell droplet-based mRNA sequencing on cells from the CCR6-expressing subgroups from an additional donor. Trajectory analysis of these cells suggested progression from the CCR6low and CCR6int to the CCR6high cells in pseudotime (Supplemental Fig. 3E).

To understand the relationships among the Th lineages across the type 17 continuum we analyzed coexpression of lineage-associated genes in the single cells from the naive and CCR6-defined memory subgroups. We found that coexpression of type 17, type 1, and type 2 genes is common. From the 251 of the 1410 total naive and memory CD4+ T cells from three donors where, after activation, we could detect expression of IL17A, 44% coexpressed IFNG, 33% coexpressed a type 2 cytokine gene (IL4 and/or IL5 and/or IL13), and 54% coexpressed IFNG and/or a type 2 cytokine gene. In analyzing those cells that coexpressed pairs of lineage-associated genes, we compared normalized values of expression (Fig. 7A). Cells coexpressing type 17 genes generally showed positive correlations in their levels, with cells expressing the highest levels enriched in the CCR6high subgroup. Positive correlations were also found for genes within the type 1 and type 2 gene sets. However, positive correlations were not found for IL17A versus type 1 and type 2 genes, and these coexpressing cells were found less commonly in the CCR6high subgroup. For the cells coexpressing IL17A and IL4 and/or IL13, virtually no cells coexpressed IL17A and IL4 or IL13 at high levels.

FIGURE 7.

Coexpression of Th lineage-associated genes in single cells. (A) Normalized values for expression of the indicated genes from 1410 cells from naive and CCR6-defined memory subgroups from three donors were pooled, and the values from cells coexpressing pairs of genes were plotted for comparison. Each dot corresponds to a single cell, and dots are colored based on their subgroups. (B) The normalized gene expression data from 1062 cells from the CCR6neg, CCR6low, CCR6int, and CCR6high memory cells from three donors were pooled, and cells were divided into positive and negative for IL17A, IFNG, or IL4, and the positive cells were equally divided into low, intermediate, and high groups based on their expression values for the respective genes. Within each of the four IL17A-, IFNG-, or IL4-defined and color-coded groups of cells, the quadrilaterals show the frequencies of cells expressing the other indicated genes.

FIGURE 7.

Coexpression of Th lineage-associated genes in single cells. (A) Normalized values for expression of the indicated genes from 1410 cells from naive and CCR6-defined memory subgroups from three donors were pooled, and the values from cells coexpressing pairs of genes were plotted for comparison. Each dot corresponds to a single cell, and dots are colored based on their subgroups. (B) The normalized gene expression data from 1062 cells from the CCR6neg, CCR6low, CCR6int, and CCR6high memory cells from three donors were pooled, and cells were divided into positive and negative for IL17A, IFNG, or IL4, and the positive cells were equally divided into low, intermediate, and high groups based on their expression values for the respective genes. Within each of the four IL17A-, IFNG-, or IL4-defined and color-coded groups of cells, the quadrilaterals show the frequencies of cells expressing the other indicated genes.

Close modal

In an alternative analysis, we combined all the memory cells regardless of their CCR6 expression, and for each of the genes, IL17A, IFNG, and IL4 separated cells into four groups, including a group without any expression and three equal-size groups of low, intermediate, and high expressers. For each group, we determined the frequencies of cells with detectable expression of other signature genes. For nonexpressing cells or cells that were expressing various levels of IL17A, the frequencies of cells coexpressing IFNG diminished from IL17Alow to IL17Aint to IL17Ahigh cells (Fig. 7B), consistent with the trend found in our other data. Surprisingly, however, the IL17Alow cells had higher frequencies of cells expressing IFNG, as well as IL4, IL5, and IL13, than did the IL17Aneg cells. For IL4, IL5, and IL13, the IL17Aint cells also had higher frequencies of expressers as compared with the IL17Aneg cells. Patterns that were generally analogous were also found in analyses of cells separated based on their expression of IFNG or IL4. These data suggest that initial stimulation of CD4+ T cells in vivo leads to pan-activation of multiple effector pathways, and that these cells can persist in the resting memory population.

The gradients of phenotypes, clonal relationships, and patterns of lineage-gene coexpression that we identified among the CCR6-expressing CD4+ T cells suggested that cells could move down this continuum to acquire more type 17 character while diminishing their type 1 and type 2 features. To demonstrate this and understand the outcomes for individual cells, we overexpressed RORC in CCR6low CD4+ T cells by transduction using lentiviral vectors encoding RORγt and GFP or GFP alone and purified GFP-positive cells. The bulk cultures of transduced cells showed increases in mRNAs for RORC, IL17A, and CCL20, a modest decrease in the mRNA for IFNG, and more significant decreases for the mRNA for IL4 (Supplemental Fig. 4A).

For determining the effects of transduction of RORC on individual CCR6low CD4+ T cells, we did single-cell analysis using Fluidigm Dynamic Arrays for RT-PCR. Comparative data on the frequencies of cells expressing detectable expression of signature genes in RORC virus-transduced versus control virus-transduced cells revealed significant increases in RORC and IL17A, decreases in IL4, IL5, and IL13, and little change in IFNG (Fig. 8A, left). In analyzing only the cells expressing IL17A, the RORC-transduced cells showed an increase in frequencies of cells coexpressing IFNG and decreases in cells coexpressing IL4, IL5, and IL13 (Fig. 8A, right). For the RORC-transduced cells, violin plots show increases in the frequencies and levels of expression of the type 17 genes IL17A, IL22, CCL20, and IL23R and the opposite effects for IL4 and IL13 (Fig. 8B). We also investigated the “dose response” of cytokine genes to RORγt by comparing the levels of mRNAs of RORC with those of the cytokine genes in the RORC-transduced cells. The levels of expression of RORC showed positive correlations with expression of IL17A and IL22. By contrast, even though transduction with RORC decreased the frequencies of cells with detectable expression of IL4, there was no correlation between the levels of RORC and IL4 mRNAs in individual cells in which expression of IL4 remains detectable. Expression of RORC and IFNG also failed to show any correlation (Fig. 8C). To address whether forced expression of RORγt might be altering frequencies of cells with relevant patterns of gene expression by selective effects on cell proliferation during the 3-d cultures, we analyzed mRNA levels of cytokine genes in transduced and DDAO-loaded cells and found that patterns of cytokine gene expression did not differ as a function of cell divisions, suggesting that RORγt was having direct effects on the cells’ phenotypes (Supplemental Fig. 4B). Together these data suggest that within the milieu of the CCR6low cells, RORγt alone is capable of driving expression of signature type 17 genes, and that increasing expression of RORγt is a mechanism for driving these cells further down the type 17 continuum.

FIGURE 8.

Position in the type 17 continuum determines sensitivity to master regulators and mechanisms of plasticity. (A) Numbers on the lower and upper bar segments show frequencies of cells expressing the indicated genes in the CCR6low cells that had been transduced with control or RORC lentiviral vectors, respectively. Each bar is divided according to the relative frequencies of the control- and RORC-transduced cells expressing the indicated genes. Left panel shows data from all transduced cells, and right panel shows data only from cells with detectable expression of IL17A. (B) Violin plots with expression distributions for the indicated genes in the CCR6low cells that had been transduced with control or RORC lentiviral vectors. (C) Normalized Ct values are plotted for the CCR6low cells that had been transduced with control or RORC lentiviral vectors and showed coexpression of RORC with IL17A, IL22, IFNG, or IL4. Each dot represents a single cell, red line is the least squares regression line, and the number in red in each plot is the correlation coefficient, r. For (A)–(C), data are pooled from four donors. (D) Dot plots showing intracellular expression on days 0, 5 and 12 after culturing in nonpolarizing (NP), Th2, or Th1 conditions of IL-17A versus IL-4 and IL-17A versus IFN-γ in naive cells (CD4+CD45ROCD25CXCR5CCR4CXCR3), IL-17A versus IL-4 in CCR6neg (CD4+CD45RO+CD25CXCR5CCR4CCR6), CCR6low (CD4+CD45RO+CD25CXCR5CCR4CCR6low), and CCR6high (CD4+CD45RO+CD25CXCR5CCR4CCR6high) subgroups, and IL-17A versus IFN-γ in CCR6neg (CD4+CD45RO+CD25CXCR5CXCR3CCR6), CCR6low (CD4+CD45RO+CD25CXCR5CXCR3CCR6low), and CCR6high (CD4+CD45RO+CD25CXCR5CXCR3CCR6high) subgroups. Cells were stimulated with the leukocyte activation mixture for 6 h, fixed and permeabilized, and stained for indicated genes. For determining percentages of cells staining positive, quadrants were drawn based on staining with isotype-matched control Abs. The numbers in black indicate the percentage of cells in each quadrant, whereas purple and red numbers show the sums of percentages of positive cells in the upper two or rightmost two quadrants, respectively. Data are representative of results from three donors.

FIGURE 8.

Position in the type 17 continuum determines sensitivity to master regulators and mechanisms of plasticity. (A) Numbers on the lower and upper bar segments show frequencies of cells expressing the indicated genes in the CCR6low cells that had been transduced with control or RORC lentiviral vectors, respectively. Each bar is divided according to the relative frequencies of the control- and RORC-transduced cells expressing the indicated genes. Left panel shows data from all transduced cells, and right panel shows data only from cells with detectable expression of IL17A. (B) Violin plots with expression distributions for the indicated genes in the CCR6low cells that had been transduced with control or RORC lentiviral vectors. (C) Normalized Ct values are plotted for the CCR6low cells that had been transduced with control or RORC lentiviral vectors and showed coexpression of RORC with IL17A, IL22, IFNG, or IL4. Each dot represents a single cell, red line is the least squares regression line, and the number in red in each plot is the correlation coefficient, r. For (A)–(C), data are pooled from four donors. (D) Dot plots showing intracellular expression on days 0, 5 and 12 after culturing in nonpolarizing (NP), Th2, or Th1 conditions of IL-17A versus IL-4 and IL-17A versus IFN-γ in naive cells (CD4+CD45ROCD25CXCR5CCR4CXCR3), IL-17A versus IL-4 in CCR6neg (CD4+CD45RO+CD25CXCR5CCR4CCR6), CCR6low (CD4+CD45RO+CD25CXCR5CCR4CCR6low), and CCR6high (CD4+CD45RO+CD25CXCR5CCR4CCR6high) subgroups, and IL-17A versus IFN-γ in CCR6neg (CD4+CD45RO+CD25CXCR5CXCR3CCR6), CCR6low (CD4+CD45RO+CD25CXCR5CXCR3CCR6low), and CCR6high (CD4+CD45RO+CD25CXCR5CXCR3CCR6high) subgroups. Cells were stimulated with the leukocyte activation mixture for 6 h, fixed and permeabilized, and stained for indicated genes. For determining percentages of cells staining positive, quadrants were drawn based on staining with isotype-matched control Abs. The numbers in black indicate the percentage of cells in each quadrant, whereas purple and red numbers show the sums of percentages of positive cells in the upper two or rightmost two quadrants, respectively. Data are representative of results from three donors.

Close modal

Given the plasticity reported for Th17 cells in mice, particularly in their transformation into IL-17 IFN-γ+ cells (36), we investigated plasticity among the CCR6-defined subgroups by activating and culturing them under nonpolarizing, Th2, or Th1 conditions. Starting cells were either CCR4 or CXCR3 to eliminate cells with pre-existing activity of the type 2 and/or type 1 programs, including all cells able to make IL-4 or IFN-γ, respectively (61). In all experiments, IL-17A+ cells were found principally in the CCR6high subgroup. In the Th2 experiments, there was little difference in frequencies of IL-17A+ cells in the Th2 versus nonpolarizing conditions, and IL-17A+ cells generally remained IL-4. At 12 d, nonpolarizing cultures showed prominent increases in percentages of IL-4–producing cells in the CCR6neg and CCR6low subgroups as compared with the naive and CCR6high cells. The Th2-polarizing conditions resulted in further increases in IL-4+ cells, with the fewest IL-4+ cells found in the CCR6high subgroup. However, the fold increase in IL-4 production in Th2 versus nonpolarizing conditions was greater in the CCR6high (and naive) cells as compared with the CCR6neg and CCR6low subgroups (Fig. 8D, left panels).

In the Th1 experiments, a significant percentage of CCR6high cells, but fewer cells in the CCR6low or other subgroups, acquired the ability to produce IFN-γ within 5 d under nonpolarizing conditions. At the same time, both at 5 and 12 d the fold increases in IFN-γ production in Th1 versus nonpolarizing conditions were smaller in the CCR6high cells as compared with the CCR6neg and CCR6low subgroups. Th1 conditions also led to a loss in IL17+IFN-γ cells and, by day 12, a shift of most IFN-γ+ cells into the IFN-γ+IL17+ subset (Fig. 8D, right panels). When considered together, these experiments suggest that progression along the type 17 continuum results in a diminished ability to express the type 2, but not the type 1, program. Of greater interest, the comparisons between cells cultured under nonpolarizing versus polarizing conditions suggest that prior conditioning versus activating environments contribute differentially to the expression of the programs for the non-Th17 lineages depending on the program and a cell’s position along the type 17 continuum. The cells’ prior imprinting contributes more than subsequent environmental inputs for the CCR6low than CCR6high cells in acquiring an ability to produce IL-4, whereas prior imprinting contributes more than the environmental inputs for the CCR6high than CCR6low cells in acquiring an ability to produce IFN-γ.

In an alternative approach to studying plasticity, we transduced CCR6low and CCR6high memory cells (here without removing cells expressing CXCR3 or CCR4) using GFP-expressing vectors encoding RORC, TBX21, and GATA3 and a control vector expressing GFP alone. All experiments used GFP+ cells purified by FACS. We checked the activities of these transcription factors initially by transducing CCR6low cells and analyzing mRNAs in the bulk populations. Given the levels of the endogenous mRNAs, the fold increase in the RORC mRNA was much greater than for the mRNAs for TBX21 and GATA3 (Supplemental Fig. 4C). As in the experiments described earlier, transduction with RORC led to a large increase in the expression of IL17A. Transductions with TBX21 and GATA3 led to increased expression of IFNG and IL4, respectively. We next transduced CCR6low and CCR6high cells with these transcription factors and analyzed mRNAs from individual cells by RT-PCR (Supplemental Fig. 4D). For IL17A, transduction with RORC led to large increases in the frequency of expressing cells and levels in the individual expressing cells in the CCR6low subgroup, and a small increase in the CCR6high subgroup. For each of the mRNAs, IFNG, GATA3, and IL4, transduction with RORC led to a decrease in the CCR6low subgroup and little change in the CCR6high subgroup, with the possible exception of some decrease in IL4 expression also in the latter cells. Transduction with TBX21 led to modest increases in the frequency and levels (in the individual expressing cells) of IFNG expression in the CCR6low subgroup, but no changes in CCR6high cells, and no consistent changes in expression of other genes. Transduction with GATA3 led to modest increases in the frequency and levels of IL4 expression in the CCR6low subgroup, but no changes in CCR6high cells, and no consistent changes in expression of other genes. Taken together, these data reveal significant responsiveness for CCR6low cells, but a diminished ability to affect the expression of lineage-specific genes in the CCR6high cells by overexpression of signature transcription factors, suggesting, as we found in our ex vivo cross-polarization experiments, that mechanisms of plasticity differ depending on the cells’ positions along the type 17 continuum. For example, the CCR6low cells, which are more responsive to Th1-polarizing conditions, are also more responsive to overexpression of TBX21, whereas the CCR6high cells are those showing the greatest prior imprinted ability to upregulate IFN-γ while being insensitive to the overexpression of TBX21.

Our studies focused on understanding the composition and organization of the memory type 17 Th cell population in healthy humans. One central finding is that these cells, including the progeny of single clones, lie at different positions along an extended continuum of type 17 differentiation on which is superimposed the coactivation of alternative lineages. Although coexpression of alternative lineages diminishes along the continuum, such coexpression is more the rule than the exception and contributes to the population’s extensive heterogeneity. Based on a wealth of data from mouse studies establishing a direct relationship between effector and memory T cells (8, 12–18), we presume that the phenotypes of the cells within this continuum reflect pathways of effector cell activation and differentiation. Our data suggest that these in vivo pathways are initiated by activating changes at genes associated with multiple effector lineages, and that the less differentiated type 17 memory cells bearing these changes are particularly capable of expressing a variety of alternative effector phenotypes.

The character of the type 17 continuum was revealed by analyzing cells that differed in their expression of the trafficking/chemokine receptor, CCR6. Studies in mice and humans have suggested that there is a family of Th17-related cells that cannot be identified by relying on the expression of a single cytokine, particularly when limited, as in human studies, to examining cells at one point in time (35, 42). In addition, the increasing appreciation of memory T cell diversity has led to proposing that the phenotypes of these cells be viewed along multiple vectors, with trafficking patterns representing a major defining feature (62). Whether all CCR6-expressing Th cells belong within a single type 17 family has been questioned (38, 63). However, in toto, our data strongly suggest that CCR6 does, in fact, identify a subgroup of related cells, and the expression of IL17A, IL17F, IL22, CCL20, IL23R, and RORC almost exclusively within the CCR6+ subset of memory T cells (46–48, 50, and this article) suggests that CCR6 is an appropriate marker for human type 17 cells understood more broadly. We reasoned that analyzing this more inclusive collection of type 17 cells would reveal relationships that might be obscured by studies of the Th17 population narrowly construed, and thereby illuminate both the pathways for differentiation and the nature of interlineage relationships in human Th cells in vivo.

Expressions of type 1 genes/proteins, such as IFNG, CXCR3, CCL3, CCL5, TBX21, EOMES, GZMK, and GZMH, and type 2 genes/proteins, such as IL4, IL5, IL13, and GATA3, were downregulated along the CCR6-defined, type 17 continuum. Both T-bet (64, 65) and eomesodermin (66, 67) suppress type 17 genes and support the development of Th17 cells that acquire a pathogenic profile and that, in humans, display the CCR6+CXCR3+ phenotype of “Th1*” or “nonclassic” Th1 cells (46, 68). Eomesodermin also induces granzyme genes in human Th cells (66) and is critical for the differentiation of a subset of type 1 regulatory, cytotoxic cells that coexpress IL-10 and IFN-γ (67, 69). EOMES and granzyme genes are signature type 1 genes in human Th cells (70), and their overall lower expression in the CCR6-expressing versus CCR6neg subgroups reflects the diminishing type 1 character and increasing type 17 character across the CCR6-defined continuum.

Analysis of individual cells along the gradient of CCR6 expression showed that expression levels of type 17 genes vary, generally in a coincident fashion, from cell to cell, with most type 17 genes showing continua of expression across a wide range of values. High-dimensional studies have identified linear patterns of differentiation with continua of phenotypes as general features of human CD8+ (71) and CD4+ (72) memory cell populations and of mouse Th cells activated under polarizing conditions (73).

If, as we presume, the memory cells within the CCR6-expressing subgroups are derived from activated cells that had differentiated to various points along a type 17 continuum, then we might expect the subgroups to share T cell clones in a pattern reflecting the subgroups’ relationships. From both the sequences of β-chain mRNAs from the cells analyzed in bulk and the sequences of paired α/β-chains of CMV-reactive single T cells, we found many shared clonotypes among the memory cell subgroups with changes in frequencies of clonotypes from subgroup to subgroup. The clonotype compositions of the CCR6-negative cells were closest to the CCR6low cells, and the ordered relationship among the CCR6-expressing subgroups matched the pattern derived from the gene and protein expression data, CCR6low to CCR6int to CCR6high, consistent with the descendants of individual clones having progressed to different points during the process of activation-driven differentiation.

To investigate the stability of the type 17 continuum, we analyzed CpG methylation in the naive and CCR6-defined memory subgroups, both in cells immediately after isolation and in cells that had undergone proliferation after gentle activation under nonpolarizing conditions in vitro. Epigenetic changes have been studied for their roles in establishing and maintaining T cell memory (74); genome-wide changes in DNA methylation are characteristic of effector/memory differentiation in mouse (75, 76) and human (72) Th cells; demethylation at lineage-specific cytokine genes has been shown to be important for their expression in differentiating Th cells (76, 77); and the preservation of patterns of gene expression in memory T cells has been associated with the stability of gene-specific demethylation across multiple cell divisions in vitro and/or in vivo (17, 19, 60, 78). Of particular relevance for our studies, demethylation of a region at the CCR6 promoter was shown to be important for stable expression of CCR6 in human Th cells (59).

We found an overall loss of methylated CpGs along the CCR6-defined continuum, consistent with increasing memory cell character, and progressive decreases in methylated CpGs at multiple type 17 genes together with increasing methylation at type 1 and type 2 genes. The patterns of inducibility of type 17 genes and degrees of their site-specific CpG methylation as analyzed at the level of cell populations were also unaltered across the CCR6-expressing memory subgroups after proliferation in a noninflammatory environment in vitro. These results suggest that the phenotypes of the cells within the CCR6-expressing subgroups would remain stable under homeostatic conditions in vivo.

The stability of the type 17 continuum under noninflammatory conditions notwithstanding, the strong suggestion from the linear relationship among the phenotypes and the clonotype data were that cells could progress along the continuum under the proper conditions of activation. We found that, indeed, a forced increase of expression of RORγt in the CCR6low cells upregulated the expression of type 17 genes and downregulated expression of some type 1 and type 2 genes. These changes were evident both in analyzing the cells in bulk and at the level of individual cells, and the single-cell data showed that expression levels of RORC mRNA correlated with the levels for a subset of type 17 genes, such as IL17A and IL22, suggesting that varying expression of RORC is one mechanism for establishing the cell-to-cell gradients in gene expression and for moving cells along the continuum. This is not to claim that the level of RORγt is the sole determinant of the complex profiles of gene and protein expression displayed by cells at different positions along this continuum. For example, it is well established by us and others that RORγt supports the expression of CCR6 in human T cells (49, 54, 79). Nonetheless, we were not able to detect an increase in CCR6 expression in the CCR6low cells after transduction with the RORC lentiviral vector (data not shown), so other factors must be important for enhancing CCR6 expression in this context.

In our analysis of gene expression levels in single, freshly isolated cells, we found positive correlations for genes within the type 17, type 1, and type 2 lineages, and negative correlations for genes across lineages, especially when a gene from one lineage was expressed at high levels. Overall, cross-suppression was most notable between the type 17 and type 2 pathways, as reflected in the single-cell data in which almost no cells coexpressed high levels of both IL17A and IL4 or IL13. Nonetheless, coexpression of signature genes of different lineages was common, with most cells that had detectable expression of IL17A coexpressing one or more of the genes that are characteristic of alternative lineages. Although we cannot know whether each cell with measurable expression of a given gene could produce the encoded protein in amounts that are biologically relevant, at a minimum, the detection of these mRNAs indicates epigenetic changes at the gene loci that reflect activation of the lineage-specific pathways. For mouse Th cells, complex continua of mixed-lineage phenotypes have been described after activation in vitro with combinations of “opposing” cytokines (80), and cells with stable mixed lineage phenotypes have been identified from inflammatory reactions in vivo (73). Human Th cells coexpressing IL-17A and IFN-γ (46–48) have been well described, and coexpression of IL-17A and type 2 cytokines has also been reported (11).

Of particular interest, our analysis of single cells showing different levels of expression of lineage-associated cytokine genes revealed that the frequencies of cells expressing IFNG or IL4 were higher among those cells expressing low levels of IL17A than among cells in which expression of IL17A was undetectable. Analogous patterns were found for cells with undetectable versus low-level expression of IFNG or IL4 in their expression of the cytokine genes of alternative lineages. These data suggest that initial activation of Th cells in vivo leads preferentially to pan-activation of effector pathways and that “early” cells with this history persist within the memory population. Previous experiments analyzing the initial events in the differentiation of activated Th cells have shown, for example, that genes for both Th1 and Th2 transcription factors and cytokines are activated before polarized patterns are established (81). During an acute challenge, when time is of the essence, pathway pan-activation could jump-start effector functions until the response can be focused appropriately. In studies of pathogen-reactive human memory Th cells, the analysis of clonotypes has strongly supported the model of “one cell, many fates,” suggesting that the dominance of a particular Th lineage in response against a pathogen results principally from the differential expansion of a clonotype within a given lineage rather than initial commitment being restricted to a single lineage (11). Within a recall response, such a process would follow naturally from individual memory cells with the phenotypes of pan-activation that we have described. Such cells would maintain within the memory pool the capabilities resulting from the previous multipathway coactivation and form cellular substrates with the flexibility to expand along the optimal pathway(s) to meet the demands of host defense.

Coexpression of genes in the type 17 and alternative Th pathways raises the issue of memory/effector Th cell plasticity (82). Soon after the identification of Th17 cells, it was reported that they showed an unusual degree of plasticity, particularly in the apparent conversion of mouse IL-17A/F–producing Th cells to cells that were making IFN-γ and often no IL-17A/F, and that these “transdifferentiated” cells were mediators of organ injury in models of autoimmune disease (33, 34, 83).

We addressed plasticity in human type 17 cells by activating and culturing naive and/or CCR6-defined subgroups of memory cells under nonpolarizing and polarizing conditions and by transducing these cells with genes for the “master regulators” RORγt, T-bet, and GATA3. In CCR4 memory cells, none of which can produce detectable IL-4 initially, the CCR6high subgroup contained fewer cells producing IL-4 after culturing cells under Th2-polarizing conditions and was less able to increase numbers of IL4-expressing cells after transducing with GATA3 when compared with cells in the CCR6low subgroup. These findings are consistent with the significant progressive decrease in expression of type 2 genes in the freshly isolated cells along the CCR6-defined continuum. However, the contribution of Th2-polarizing conditions on IL-4 production was greater for the CCR6high (and naive) cells than for the CCR6neg and CCR6low cells, for which much of the ability to produce IL-4 was acquired in the nonpolarizing environment. In CXCR3 memory cells, none of which can produce detectable IFN-γ initially, a notable finding was the ready acquisition by the CCR6high cells of the ability to produce IFN-γ after activation under nonpolarizing conditions. For these cells, the added effect of a Th1-polarizing environment on IFN-γ production was significantly less marked than it was for the CCR6neg and CCR6low (and naive) cells.

The results of our polarization experiments suggest two distinguishable processes at work. The first process is “imprinted partial plasticity” as proposed by Crotty (38), whereby the potential to express aspects of alternative programs has been established during initial activation but is not manifest until reactivation in a permissive environment. The second process is plasticity induced by the environmental inputs during reactivation. Our data suggest that the outcomes of reactivating type 17 memory cells will reflect differing contributions of imprinted plasticity and retained flexibility depending on which pathways are activated and, importantly, the cells’ positions along the type 17 continuum.

Our work has limitations typical of many human studies. In vivo processes are inferred from the analyses of cells isolated at one time from a limited numbers of donors who vary in genetic makeups and histories. Not only are the donors heterogeneous, but the cell subsets that were studied are highly heterogeneous in their Ag specificities and other phenotypes, being characterized and purified based on a limited number of surface proteins. Consequently, we do not have direct data on the fates of individual cells and their progeny after activation and during differentiation in vivo. A further important limitation is that our analyses were restricted to cells isolated from the blood. Although recent analyses of T cells from human tissue donors have shown overlap between memory T cell clones found in blood and tissues, as compared with the T cells in blood, the clonal diversity of memory T cells in tissue is better maintained with age, and some abundant tissue-resident memory clones are poorly represented in blood (84). Consequently, samples of human memory T cells from blood (or any single tissue site) cannot include the full diversity of the body’s memory T cell population.

In summary, we have found that CCR6 identifies a broad but connected family of human type 17 memory Th cells that fall along an extended gradient of type 17 activity. This continuum is evident in cells analyzed in bulk as subgroups or, more importantly, as single cells, and is likely to reflect pathways of clonal differentiation during activation in vivo. The epigenomes and phenotypes of cells along the continuum are stable, including across cell divisions under noninflammatory conditions. Nonetheless, increasing expression of RORC in CCR6low cells demonstrated a mechanism whereby cells could “move to the right” to acquire additional type 17 character, and cells across the type 17 continuum showed differing mechanisms of plasticity in the inducible expression of signature genes of non–type 17 lineages. Whereas the CCR6high cells were more resistant than the CCR6low cells in activating IL4, the CCR6high cells were superior in activating IFNG under nonpolarizing conditions, representing, to our knowledge, a novel example of imprinted plasticity, which was greater in the more highly differentiated type 17 cells. Considered together, the features of the extended pathway of type 17 differentiation that we have described yield both a linear progression of type 17 activity and a range of other attributes that, importantly, differ as a function of the cells’ positions along the continuum, including mixed-lineage capabilities and varied mechanisms of plasticity to create a multifunctional and adaptable collection of type 17 memory cells. Finally, our sensitive single-cell RT-PCR assays of cytokine gene expression suggest that pan-induction of effector pathways, rather than choice of a single lineage, is a general feature of early Th cell activation in vivo, and that cells with the resulting multilineage character persist in the memory pool. We believe that, these studies provide fundamental information on the composition of the human type 17 memory population and allow for informed inferences and hypotheses as to pathways and mechanisms through which this highly heterogeneous population is formed and can adapt to novel demands of host defense or be misdirected to produce inflammatory damage.

The authors have no financial conflicts of interest.

We thank Jinfang Zhu, NIAID, NIH, and Jean K. Lim, Icahn School of Medicine at Mount Sinai, for critical reading of the manuscript and Chaim A. Schramm, NIAID, NIH, for help in depositing TCR sequencing data in the BioProject database. We are grateful to Calvin Eigsti and members of the Research Technologies Branch, NIAID, for help with cell sorting.

The bulk TCR sequence data presented in this article have been submitted to the National Center for Biotechnology Information (NCBI) BioProject (https://www.ncbi.nlm.nih.gov/bioproject/?term=PRJNA930733) under accession number PRJNA930733 and the gene array, DNA methylation, and single-cell sequencing data presented in this article have been submitted to the NCBI Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE227488) under accession number GSE227488.

This work was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health.

The online version of this article contains supplemental material.

Ct

cycle threshold

d7NP

day 7 in cells that failed to divide

d7P

day 7 in cells that divided three or more times

DDAO

dodecyldimethylamine oxide

1
Farber
,
D. L.
,
M. G.
Netea
,
A.
Radbruch
,
K.
Rajewsky
,
R. M.
Zinkernagel
.
2016
.
Immunological memory: lessons from the past and a look to the future
.
Nat. Rev. Immunol.
16
:
124
128
.
2
Hammarlund
,
E.
,
M. W.
Lewis
,
S. G.
Hansen
,
L. I.
Strelow
,
J. A.
Nelson
,
G. J.
Sexton
,
J. M.
Hanifin
,
M. K.
Slifka
.
2003
.
Duration of antiviral immunity after smallpox vaccination
.
Nat. Med.
9
:
1131
1137
.
3
Naniche
,
D.
,
M.
Garenne
,
C.
Rae
,
M.
Manchester
,
R.
Buchta
,
S. K.
Brodine
,
M. B.
Oldstone
.
2004
.
Decrease in measles virus-specific CD4 T cell memory in vaccinated subjects
.
J. Infect. Dis.
190
:
1387
1395
.
4
Khader
,
S. A.
,
G. K.
Bell
,
J. E.
Pearl
,
J. J.
Fountain
,
J.
Rangel-Moreno
,
G. E.
Cilley
,
F.
Shen
,
S. M.
Eaton
,
S. L.
Gaffen
,
S. L.
Swain
, et al
.
2007
.
IL-23 and IL-17 in the establishment of protective pulmonary CD4+ T cell responses after vaccination and during Mycobacterium tuberculosis challenge
.
Nat. Immunol.
8
:
369
377
.
5
Teijaro
,
J. R.
,
D.
Verhoeven
,
C. A.
Page
,
D.
Turner
,
D. L.
Farber
.
2010
.
Memory CD4 T cells direct protective responses to influenza virus in the lungs through helper-independent mechanisms
.
J. Virol.
84
:
9217
9226
.
6
Kanno
,
Y.
,
G.
Vahedi
,
K.
Hirahara
,
K.
Singleton
,
J. J.
O’Shea
.
2012
.
Transcriptional and epigenetic control of T helper cell specification: molecular mechanisms underlying commitment and plasticity
.
Annu. Rev. Immunol.
30
:
707
731
.
7
Ahmed
,
R.
,
M. J.
Bevan
,
S. L.
Reiner
,
D. T.
Fearon
.
2009
.
The precursors of memory: models and controversies
.
Nat. Rev. Immunol.
9
:
662
668
.
8
Stemberger
,
C.
,
K. M.
Huster
,
M.
Koffler
,
F.
Anderl
,
M.
Schiemann
,
H.
Wagner
,
D. H.
Busch
.
2007
.
A single naive CD8+ T cell precursor can develop into diverse effector and memory subsets
.
Immunity
27
:
985
997
.
9
Gerlach
,
C.
,
J. W.
van Heijst
,
E.
Swart
,
D.
Sie
,
N.
Armstrong
,
R. M.
Kerkhoven
,
D.
Zehn
,
M. J.
Bevan
,
K.
Schepers
,
T. N.
Schumacher
.
2010
.
One naive T cell, multiple fates in CD8+ T cell differentiation
.
J. Exp. Med.
207
:
1235
1246
.
10
Buchholz
,
V. R.
,
T. N.
Schumacher
,
D. H.
Busch
.
2016
.
T cell fate at the single-cell level
.
Annu. Rev. Immunol.
34
:
65
92
.
11
Becattini
,
S.
,
D.
Latorre
,
F.
Mele
,
M.
Foglierini
,
C.
De Gregorio
,
A.
Cassotta
,
B.
Fernandez
,
S.
Kelderman
,
T. N.
Schumacher
,
D.
Corti
, et al
.
2015
.
T cell immunity. Functional heterogeneity of human memory CD4+ T cell clones primed by pathogens or vaccines
.
Science
347
:
400
406
.
12
Löhning
,
M.
,
A. N.
Hegazy
,
D. D.
Pinschewer
,
D.
Busse
,
K. S.
Lang
,
T.
Höfer
,
A.
Radbruch
,
R. M.
Zinkernagel
,
H.
Hengartner
.
2008
.
Long-lived virus-reactive memory T cells generated from purified cytokine-secreting T helper type 1 and type 2 effectors
.
J. Exp. Med.
205
:
53
61
.
13
Harrington
,
L. E.
,
K. M.
Janowski
,
J. R.
Oliver
,
A. J.
Zajac
,
C. T.
Weaver
.
2008
.
Memory CD4 T cells emerge from effector T-cell progenitors
.
Nature
452
:
356
360
.
14
Hu
,
H.
,
G.
Huston
,
D.
Duso
,
N.
Lepak
,
E.
Roman
,
S. L.
Swain
.
2001
.
CD4(+) T cell effectors can become memory cells with high efficiency and without further division
.
Nat. Immunol.
2
:
705
710
.
15
Opferman
,
J. T.
,
B. T.
Ober
,
P. G.
Ashton-Rickardt
.
1999
.
Linear differentiation of cytotoxic effectors into memory T lymphocytes
.
Science
283
:
1745
1748
.
16
Youngblood
,
B.
,
J. S.
Hale
,
H. T.
Kissick
,
E.
Ahn
,
X.
Xu
,
A.
Wieland
,
K.
Araki
,
E. E.
West
,
H. E.
Ghoneim
,
Y.
Fan
, et al
.
2017
.
Effector CD8 T cells dedifferentiate into long-lived memory cells
.
Nature
552
:
404
409
.
17
Akondy
,
R. S.
,
M.
Fitch
,
S.
Edupuganti
,
S.
Yang
,
H. T.
Kissick
,
K. W.
Li
,
B. A.
Youngblood
,
H. A.
Abdelsamed
,
D. J.
McGuire
,
K. W.
Cohen
, et al
.
2017
.
Origin and differentiation of human memory CD8 T cells after vaccination
.
Nature
552
:
362
367
.
18
Tubo
,
N. J.
,
B. T.
Fife
,
A. J.
Pagan
,
D. I.
Kotov
,
M. F.
Goldberg
,
M. K.
Jenkins
.
2016
.
Most microbe-specific naïve CD4+ T cells produce memory cells during infection
.
Science
351
:
511
514
.
19
Abdelsamed
,
H. A.
,
A.
Moustaki
,
Y.
Fan
,
P.
Dogra
,
H. E.
Ghoneim
,
C. C.
Zebley
,
B. M.
Triplett
,
R. P.
Sekaly
,
B.
Youngblood
.
2017
.
Human memory CD8 T cell effector potential is epigenetically preserved during in vivo homeostasis
.
J. Exp. Med.
214
:
1593
1606
.
20
Kumar
,
B. V.
,
T. J.
Connors
,
D. L.
Farber
.
2018
.
Human T cell development, localization, and function throughout life
.
Immunity
48
:
202
213
.
21
Davis
,
M. M.
,
P.
Brodin
.
2018
.
Rebooting human immunology
.
Annu. Rev. Immunol.
36
:
843
864
.
22
Korn
,
T.
,
E.
Bettelli
,
M.
Oukka
,
V. K.
Kuchroo
.
2009
.
IL-17 and Th17 cells
.
Annu. Rev. Immunol.
27
:
485
517
.
23
Langrish
,
C. L.
,
Y.
Chen
,
W. M.
Blumenschein
,
J.
Mattson
,
B.
Basham
,
J. D.
Sedgwick
,
T.
McClanahan
,
R. A.
Kastelein
,
D. J.
Cua
.
2005
.
IL-23 drives a pathogenic T cell population that induces autoimmune inflammation
.
J. Exp. Med.
201
:
233
240
.
24
Milner
,
J. D.
,
J. M.
Brenchley
,
A.
Laurence
,
A. F.
Freeman
,
B. J.
Hill
,
K. M.
Elias
,
Y.
Kanno
,
C.
Spalding
,
H. Z.
Elloumi
,
M. L.
Paulson
, et al
.
2008
.
Impaired T(H)17 cell differentiation in subjects with autosomal dominant hyper-IgE syndrome
.
Nature
452
:
773
776
.
25
Puel
,
A.
,
S.
Cypowyj
,
J.
Bustamante
,
J. F.
Wright
,
L.
Liu
,
H. K.
Lim
,
M.
Migaud
,
L.
Israel
,
M.
Chrabieh
,
M.
Audry
, et al
.
2011
.
Chronic mucocutaneous candidiasis in humans with inborn errors of interleukin-17 immunity
.
Science
332
:
65
68
.
26
Okada
,
S.
,
J. G.
Markle
,
E. K.
Deenick
,
F.
Mele
,
D.
Averbuch
,
M.
Lagos
,
M.
Alzahrani
,
S.
Al-Muhsen
,
R.
Halwani
,
C. S.
Ma
, et al
.
2015
.
IMMUNODEFICIENCIES. Impairment of immunity to Candida and Mycobacterium in humans with bi-allelic RORC mutations
.
Science
349
:
606
613
.
27
Ogongo
,
P.
,
L. B.
Tezera
,
A.
Ardain
,
S.
Nhamoyebonde
,
D.
Ramsuran
,
A.
Singh
,
A.
Ng’oepe
,
F.
Karim
,
T.
Naidoo
,
K.
Khan
, et al
.
2021
.
Tissue-resident-like CD4+ T cells secreting IL-17 control Mycobacterium tuberculosis in the human lung
.
J. Clin. Invest.
131
:
e142014
.
28
Patel
,
D. D.
,
V. K.
Kuchroo
.
2015
.
Th17 cell pathway in human immunity: lessons from genetics and therapeutic interventions
.
Immunity
43
:
1040
1051
.
29
Stockinger
,
B.
,
S.
Omenetti
.
2017
.
The dichotomous nature of T helper 17 cells
.
Nat. Rev. Immunol.
17
:
535
544
.
30
Pepper
,
M.
,
J. L.
Linehan
,
A. J.
Pagán
,
T.
Zell
,
T.
Dileepan
,
P. P.
Cleary
,
M. K.
Jenkins
.
2010
.
Different routes of bacterial infection induce long-lived TH1 memory cells and short-lived TH17 cells
.
Nat. Immunol.
11
:
83
89
.
31
Muranski
,
P.
,
Z. A.
Borman
,
S. P.
Kerkar
,
C. A.
Klebanoff
,
Y.
Ji
,
L.
Sanchez-Perez
,
M.
Sukumar
,
R. N.
Reger
,
Z.
Yu
,
S. J.
Kern
, et al
.
2011
.
Th17 cells are long lived and retain a stem cell-like molecular signature
.
Immunity
35
:
972
985
.
32
Kryczek
,
I.
,
E.
Zhao
,
Y.
Liu
,
Y.
Wang
,
L.
Vatan
,
W.
Szeliga
,
J.
Moyer
,
A.
Klimczak
,
A.
Lange
,
W.
Zou
.
2011
.
Human TH17 cells are long-lived effector memory cells
.
Sci. Transl. Med.
3
:
104ra100
.
33
Bending
,
D.
,
H.
De la Peña
,
M.
Veldhoen
,
J. M.
Phillips
,
C.
Uyttenhove
,
B.
Stockinger
,
A.
Cooke
.
2009
.
Highly purified Th17 cells from BDC2.5NOD mice convert into Th1-like cells in NOD/SCID recipient mice
.
J. Clin. Invest.
119
:
565
572
.
34
Lee
,
Y. K.
,
H.
Turner
,
C. L.
Maynard
,
J. R.
Oliver
,
D.
Chen
,
C. O.
Elson
,
C. T.
Weaver
.
2009
.
Late developmental plasticity in the T helper 17 lineage
.
Immunity
30
:
92
107
.
35
Hirota
,
K.
,
J. H.
Duarte
,
M.
Veldhoen
,
E.
Hornsby
,
Y.
Li
,
D. J.
Cua
,
H.
Ahlfors
,
C.
Wilhelm
,
M.
Tolaini
,
U.
Menzel
, et al
.
2011
.
Fate mapping of IL-17-producing T cells in inflammatory responses
.
Nat. Immunol.
12
:
255
263
.
36
Bonelli
,
M.
,
H. Y.
Shih
,
K.
Hirahara
,
K.
Singelton
,
A.
Laurence
,
A.
Poholek
,
T.
Hand
,
Y.
Mikami
,
G.
Vahedi
,
Y.
Kanno
,
J. J.
O’Shea
.
2014
.
Helper T cell plasticity: impact of extrinsic and intrinsic signals on transcriptomes and epigenomes
.
Curr. Top. Microbiol. Immunol.
381
:
279
326
.
37
DuPage
,
M.
,
J. A.
Bluestone
.
2016
.
Harnessing the plasticity of CD4(+) T cells to treat immune-mediated disease
.
Nat. Rev. Immunol.
16
:
149
163
.
38
Crotty
,
S.
2018
.
Do memory CD4 T cells keep their cell-type programming: plasticity versus fate commitment? Complexities of interpretation due to the heterogeneity of memory CD4 T cells, including T follicular helper cells
.
Cold Spring Harb. Perspect. Biol.
10
:
a032102
.
39
Nurieva
,
R.
,
X. O.
Yang
,
Y.
Chung
,
C.
Dong
.
2009
.
Cutting edge: in vitro generated Th17 cells maintain their cytokine expression program in normal but not lymphopenic hosts
.
J. Immunol.
182
:
2565
2568
.
40
Krebs
,
C. F.
,
J. E.
Turner
,
H. J.
Paust
,
S.
Kapffer
,
T.
Koyro
,
S.
Krohn
,
F.
Ufer
,
M. A.
Friese
,
R. A.
Flavell
,
B.
Stockinger
, et al
.
2016
.
Plasticity of Th17 cells in autoimmune kidney diseases
.
J. Immunol.
197
:
449
457
.
41
Cohen
,
C. J.
,
S. Q.
Crome
,
K. G.
MacDonald
,
E. L.
Dai
,
D. L.
Mager
,
M. K.
Levings
.
2011
.
Human Th1 and Th17 cells exhibit epigenetic stability at signature cytokine and transcription factor loci
.
J. Immunol.
187
:
5615
5626
.
42
Boniface
,
K.
,
W. M.
Blumenschein
,
K.
Brovont-Porth
,
M. J.
McGeachy
,
B.
Basham
,
B.
Desai
,
R.
Pierce
,
T. K.
McClanahan
,
S.
Sadekova
,
R.
de Waal Malefyt
.
2010
.
Human Th17 cells comprise heterogeneous subsets including IFN-gamma-producing cells with distinct properties from the Th1 lineage
.
J. Immunol.
185
:
679
687
.
43
Mazzoni
,
A.
,
V.
Santarlasci
,
L.
Maggi
,
M.
Capone
,
M. C.
Rossi
,
V.
Querci
,
R.
De Palma
,
H. D.
Chang
,
A.
Thiel
,
R.
Cimaz
, et al
.
2015
.
Demethylation of the RORC2 and IL17A in human CD4+ T lymphocytes defines Th17 origin of nonclassic Th1 cells
.
J. Immunol.
194
:
3116
3126
.
44
Reboldi
,
A.
,
T. I.
Arnon
,
L. B.
Rodda
,
A.
Atakilit
,
D.
Sheppard
,
J. G.
Cyster
.
2016
.
IgA production requires B cell interaction with subepithelial dendritic cells in Peyer’s patches
.
Science
352
:
aaf4822
.
45
Ito
,
T.
,
W. F.
Carson
IV
,
K. A.
Cavassani
,
J. M.
Connett
,
S. L.
Kunkel
.
2011
.
CCR6 as a mediator of immunity in the lung and gut
.
Exp. Cell Res.
317
:
613
619
.
46
Acosta-Rodriguez
,
E. V.
,
L.
Rivino
,
J.
Geginat
,
D.
Jarrossay
,
M.
Gattorno
,
A.
Lanzavecchia
,
F.
Sallusto
,
G.
Napolitani
.
2007
.
Surface phenotype and antigenic specificity of human interleukin 17-producing T helper memory cells
.
Nat. Immunol.
8
:
639
646
.
47
Annunziato
,
F.
,
L.
Cosmi
,
V.
Santarlasci
,
L.
Maggi
,
F.
Liotta
,
B.
Mazzinghi
,
E.
Parente
,
L.
Filì
,
S.
Ferri
,
F.
Frosali
, et al
.
2007
.
Phenotypic and functional features of human Th17 cells
.
J. Exp. Med.
204
:
1849
1861
.
48
Singh
,
S. P.
,
H. H.
Zhang
,
J. F.
Foley
,
M. N.
Hedrick
,
J. M.
Farber
.
2008
.
Human T cells that are able to produce IL-17 express the chemokine receptor CCR6
.
J. Immunol.
180
:
214
221
.
49
Manel
,
N.
,
D.
Unutmaz
,
D. R.
Littman
.
2008
.
The differentiation of human T(H)-17 cells requires transforming growth factor-beta and induction of the nuclear receptor RORgammat
.
Nat. Immunol.
9
:
641
649
.
50
Rivino
,
L.
,
P.
Gruarin
,
B.
Häringer
,
S.
Steinfelder
,
L.
Lozza
,
B.
Steckel
,
A.
Weick
,
E.
Sugliano
,
D.
Jarrossay
,
A. A.
Kühl
, et al
.
2010
.
CCR6 is expressed on an IL-10-producing, autoreactive memory T cell population with context-dependent regulatory function
.
J. Exp. Med.
207
:
565
577
.
51
Kleinewietfeld
,
M.
,
F.
Puentes
,
G.
Borsellino
,
L.
Battistini
,
O.
Rötzschke
,
K.
Falk
.
2005
.
CCR6 expression defines regulatory effector/memory-like cells within the CD25(+)CD4+ T-cell subset
.
Blood
105
:
2877
2886
.
52
Beriou
,
G.
,
C. M.
Costantino
,
C. W.
Ashley
,
L.
Yang
,
V. K.
Kuchroo
,
C.
Baecher-Allan
,
D. A.
Hafler
.
2009
.
IL-17-producing human peripheral regulatory T cells retain suppressive function
.
Blood
113
:
4240
4249
.
53
Paulissen
,
S. M.
,
J. P.
van Hamburg
,
W.
Dankers
,
E.
Lubberts
.
2015
.
The role and modulation of CCR6+ Th17 cell populations in rheumatoid arthritis
.
Cytokine
74
:
43
53
.
54
Singh
,
S. P.
,
H. H.
Zhang
,
H.
Tsang
,
P. J.
Gardina
,
T. G.
Myers
,
V.
Nagarajan
,
C. H.
Lee
,
J. M.
Farber
.
2015
.
PLZF regulates CCR6 and is critical for the acquisition and maintenance of the Th17 phenotype in human cells
.
J. Immunol.
194
:
4350
4361
.
55
Boritz
,
E. A.
,
S.
Darko
,
L.
Swaszek
,
G.
Wolf
,
D.
Wells
,
X.
Wu
,
A. R.
Henry
,
F.
Laboune
,
J.
Hu
,
D.
Ambrozak
, et al
.
2016
.
Multiple origins of virus persistence during natural control of HIV infection
.
Cell
166
:
1004
1015
.
56
Zhang
,
J.
,
K.
Kobert
,
T.
Flouri
,
A.
Stamatakis
.
2014
.
PEAR: a fast and accurate Illumina Paired-End reAd mergeR
.
Bioinformatics
30
:
614
620
.
57
Bolotin
,
D. A.
,
S.
Poslavsky
,
I.
Mitrophanov
,
M.
Shugay
,
I. Z.
Mamedov
,
E. V.
Putintseva
,
D. M.
Chudakov
.
2015
.
MiXCR: software for comprehensive adaptive immunity profiling
.
Nat. Methods
12
:
380
381
.
58
Dalerba
,
P.
,
T.
Kalisky
,
D.
Sahoo
,
P. S.
Rajendran
,
M. E.
Rothenberg
,
A. A.
Leyrat
,
S.
Sim
,
J.
Okamoto
,
D. M.
Johnston
,
D.
Qian
, et al
.
2011
.
Single-cell dissection of transcriptional heterogeneity in human colon tumors
.
Nat. Biotechnol.
29
:
1120
1127
.
59
Steinfelder
,
S.
,
S.
Floess
,
D.
Engelbert
,
B.
Haeringer
,
U.
Baron
,
L.
Rivino
,
B.
Steckel
,
A.
Gruetzkau
,
S.
Olek
,
J.
Geginat
, et al
.
2011
.
Epigenetic modification of the human CCR6 gene is associated with stable CCR6 expression in T cells
.
Blood
117
:
2839
2846
.
60
Weng
,
N. P.
,
Y.
Araki
,
K.
Subedi
.
2012
.
The molecular basis of the memory T cell response: differential gene expression and its epigenetic regulation
.
Nat. Rev. Immunol.
12
:
306
315
.
61
Sallusto
,
F.
,
A.
Lanzavecchia
.
2009
.
Heterogeneity of CD4+ memory T cells: functional modules for tailored immunity
.
Eur. J. Immunol.
39
:
2076
2082
.
62
Jameson
,
S. C.
,
D.
Masopust
.
2018
.
Understanding subset diversity in T cell memory
.
Immunity
48
:
214
226
.
63
Rydyznski Moderbacher
,
C.
,
S. I.
Ramirez
,
J. M.
Dan
,
A.
Grifoni
,
K. M.
Hastie
,
D.
Weiskopf
,
S.
Belanger
,
R. K.
Abbott
,
C.
Kim
,
J.
Choi
, et al
.
2020
.
Antigen-specific adaptive immunity to SARS-CoV-2 in acute COVID-19 and associations with age and disease severity
.
Cell
183
:
996
1012.e19
.
64
Yang
,
Y.
,
J.
Xu
,
Y.
Niu
,
J. S.
Bromberg
,
Y.
Ding
.
2008
.
T-bet and eomesodermin play critical roles in directing T cell differentiation to Th1 versus Th17
.
J. Immunol.
181
:
8700
8710
.
65
Wang
,
Y.
,
J.
Godec
,
K.
Ben-Aissa
,
K.
Cui
,
K.
Zhao
,
A. B.
Pucsek
,
Y. K.
Lee
,
C. T.
Weaver
,
R.
Yagi
,
V.
Lazarevic
.
2014
.
The transcription factors T-bet and Runx are required for the ontogeny of pathogenic interferon-γ-producing T helper 17 cells
.
Immunity
40
:
355
366
.
66
Mazzoni
,
A.
,
L.
Maggi
,
F.
Siracusa
,
M.
Ramazzotti
,
M. C.
Rossi
,
V.
Santarlasci
,
G.
Montaini
,
M.
Capone
,
B.
Rossettini
,
R.
De Palma
, et al
.
2019
.
Eomes controls the development of Th17-derived (non-classic) Th1 cells during chronic inflammation
.
Eur. J. Immunol.
49
:
79
95
.
67
Gruarin
,
P.
,
S.
Maglie
,
M.
De Simone
,
B.
Häringer
,
C.
Vasco
,
V.
Ranzani
,
R.
Bosotti
,
J. S.
Noddings
,
P.
Larghi
,
F.
Facciotti
, et al
.
2019
.
Eomesodermin controls a unique differentiation program in human IL-10 and IFN-γ coproducing regulatory T cells
.
Eur. J. Immunol.
49
:
96
111
.
68
Sallusto
,
F.
,
A.
Cassotta
,
D.
Hoces
,
M.
Foglierini
,
A.
Lanzavecchia
.
2018
.
Do memory CD4 T cells keep their cell-type programming: plasticity versus fate commitment? T-cell heterogeneity, plasticity, and selection in humans
.
Cold Spring Harb. Perspect. Biol.
10
:
a029421
.
69
Zhang
,
P.
,
J. S.
Lee
,
K. H.
Gartlan
,
I. S.
Schuster
,
I.
Comerford
,
A.
Varelias
,
M. A.
Ullah
,
S.
Vuckovic
,
M.
Koyama
,
R. D.
Kuns
, et al
.
2017
.
Eomesodermin promotes the development of type 1 regulatory T (TR1) cells
.
Sci. Immunol.
2
:
eaah7152
.
70
Ono
,
C.
,
Z.
Yu
,
Y.
Kasahara
,
Y.
Kikuchi
,
N.
Ishii
,
H.
Tomita
.
2014
.
Fluorescently activated cell sorting followed by microarray profiling of helper T cell subtypes from human peripheral blood
.
PLoS One
9
:
e111405
.
71
Newell
,
E. W.
,
N.
Sigal
,
S. C.
Bendall
,
G. P.
Nolan
,
M. M.
Davis
.
2012
.
Cytometry by time-of-flight shows combinatorial cytokine expression and virus-specific cell niches within a continuum of CD8+ T cell phenotypes. [Published erratum appears in 2013 Immunity 38: 198–199.]
Immunity
36
:
142
152
.
72
Durek
,
P.
,
K.
Nordström
,
G.
Gasparoni
,
A.
Salhab
,
C.
Kressler
,
M.
de Almeida
,
K.
Bassler
,
T.
Ulas
,
F.
Schmidt
,
J.
Xiong
, et al
DEEP Consortium
.
2016
.
Epigenomic profiling of human CD4+ T cells supports a linear differentiation model and highlights molecular regulators of memory development
.
Immunity
45
:
1148
1161
.
73
Tortola
,
L.
,
A.
Jacobs
,
L.
Pohlmeier
,
F. J.
Obermair
,
F.
Ampenberger
,
B.
Bodenmiller
,
M.
Kopf
.
2020
.
High-dimensional T helper cell profiling reveals a broad diversity of stably committed effector states and uncovers interlineage relationships
.
Immunity
53
:
597
613.e6
.
74
Tough
,
D. F.
,
I.
Rioja
,
L. K.
Modis
,
R. K.
Prinjha
.
2020
.
Epigenetic regulation of T cell memory: recalling therapeutic implications
.
Trends Immunol.
41
:
29
45
.
75
Hashimoto
,
S.
,
K.
Ogoshi
,
A.
Sasaki
,
J.
Abe
,
W.
Qu
,
Y.
Nakatani
,
B.
Ahsan
,
K.
Oshima
,
F. H.
Shand
,
A.
Ametani
, et al
.
2013
.
Coordinated changes in DNA methylation in antigen-specific memory CD4 T cells
.
J. Immunol.
190
:
4076
4091
.
76
Ichiyama
,
K.
,
T.
Chen
,
X.
Wang
,
X.
Yan
,
B. S.
Kim
,
S.
Tanaka
,
D.
Ndiaye-Lobry
,
Y.
Deng
,
Y.
Zou
,
P.
Zheng
, et al
.
2015
.
The methylcytosine dioxygenase Tet2 promotes DNA demethylation and activation of cytokine gene expression in T cells. [Published erratum appears in 2015 Immunity 42: 1214.]
Immunity
42
:
613
626
.
77
Thomas
,
R. M.
,
H.
Sai
,
A. D.
Wells
.
2012
.
Conserved intergenic elements and DNA methylation cooperate to regulate transcription at the il17 locus
.
J. Biol. Chem.
287
:
25049
25059
.
78
Fitzpatrick
,
D. R.
,
K. M.
Shirley
,
A.
Kelso
.
1999
.
Cutting edge: stable epigenetic inheritance of regional IFN-gamma promoter demethylation in CD44highCD8+ T lymphocytes
.
J. Immunol.
162
:
5053
5057
.
79
Castro
,
G.
,
X.
Liu
,
K.
Ngo
,
A.
De Leon-Tabaldo
,
S.
Zhao
,
R.
Luna-Roman
,
J.
Yu
,
T.
Cao
,
R.
Kuhn
,
P.
Wilkinson
, et al
.
2017
.
RORγt and RORα signature genes in human Th17 cells
.
PLoS One
12
:
e0181868
.
80
Eizenberg-Magar
,
I.
,
J.
Rimer
,
I.
Zaretsky
,
D.
Lara-Astiaso
,
S.
Reich-Zeliger
,
N.
Friedman
.
2017
.
Diverse continuum of CD4+ T-cell states is determined by hierarchical additive integration of cytokine signals
.
Proc. Natl. Acad. Sci. USA
114
:
E6447
E6456
.
81
Grogan
,
J. L.
,
M.
Mohrs
,
B.
Harmon
,
D. A.
Lacy
,
J. W.
Sedat
,
R. M.
Locksley
.
2001
.
Early transcription and silencing of cytokine genes underlie polarization of T helper cell subsets
.
Immunity
14
:
205
215
.
82
Zhu
,
J.
2018
.
T helper cell differentiation, heterogeneity, and plasticity
.
Cold Spring Harb. Perspect. Biol.
10
:
a030338
.
83
Harbour
,
S. N.
,
C. L.
Maynard
,
C. L.
Zindl
,
T. R.
Schoeb
,
C. T.
Weaver
.
2015
.
Th17 cells give rise to Th1 cells that are required for the pathogenesis of colitis
.
Proc. Natl. Acad. Sci. USA
112
:
7061
7066
.
84
Miron
,
M.
,
W.
Meng
,
A. M.
Rosenfeld
,
S.
Dvorkin
,
M. M. L.
Poon
,
N.
Lam
,
B. V.
Kumar
,
Y.
Louzoun
,
E. T.
Luning Prak
,
D. L.
Farber
.
2021
.
Maintenance of the human memory T cell repertoire by subset and tissue site
.
Genome Med.
13
:
100
.

Supplementary data