Visual Abstract

Classic T cell subsets are defined by a small set of cell surface markers, while single-cell RNA sequencing (scRNA-seq) clusters cells using genome-wide gene expression profiles. The relationship between scRNA-seq clustered populations (scCPops) and cell surface marker–defined classic T cell subsets remains unclear. In this article, we integrated six bead-enriched T cell subsets with 62,235 single-cell transcriptomes from human PBMCs and clustered them into nine scCPops. Bead-enriched CD4+/CD45RA+/CD25 naive T and CD8+/CD45RA+ naive T cells were mainly clustered into their scCPop counterparts, while cells from the other T cell subsets were assigned to multiple scCPops, including mucosal-associated invariant T cells and NKT cells. The multiple T cell subsets forming one scCPop exhibit similar expression patterns, but not vice versa, indicating scCPop is a more homogeneous cell population with similar cell states. Interestingly, we discovered and named IFN signaling–associated gene (ISAG) high T (ISAGhi T) cells, a T cell subpopulation that highly expressed ISAGs. We further enriched ISAGhi T cells from human PBMCs by FACS of BST2 for scRNA-seq analyses. The ISAGhi T cell cluster disappeared on t-distributed stochastic neighbor embedding plot after removing ISAGs, whereas the ISAGhi T cell cluster showed up by analysis of ISAGs alone, indicating ISAGs are the major contributor of the ISAGhi T cell cluster. BST2+ and BST2 T cells showing different efficiencies of T cell activation indicate that a high level of ISAGs may contribute to quick immune responses.

The T cells also known as T lymphocytes are an essential component of adaptive immunity for protecting us against various pathogens. In fact, T cells not only play a key role in elimination of invasive pathogens and cancer cells but also are important targets in the treatment of autoimmune diseases (15). The major T cells can be classified into CD4 T cells and CD8 T cells based on cell functions and cell surface markers. CD4 T cells, also known as Th cells, play a central role in adaptive immune response to pathogens, such as assisting B cells to produce Abs, recruiting granulocytes to infected sites, and producing cytokines and chemokines to orchestrate immune response (6). Many distinct CD4 T cell subsets have been identified since the identification of Th1 by Mosmann et al. (6, 7). CD4+CD25+ T cells, namely, regulatory T cells (Tregs), suppress self-reactive lymphocytes to maintain immunologic self-tolerance (8, 9). CD8 T cells are important for immune defense against intracellular pathogens (10, 11). In contrast, naive T cells egress from the thymus and are fairly quiescent, which could differentiate into distinct functional T cell subsets after they interact with cognate Ags, while memory T cells recognize previously encountered pathogens and respond to them quickly (3). These T cell subsets could be enriched or sorted out by FACS or magnetic beads because each of them is defined by a set of specific cell surface markers. Then population-average assays were routinely used to characterize the biochemical and molecular features of each T cell subset.

Single-cell RNA sequencing (scRNA-seq), profiling gene expression at single-cell resolution, has become an ideal approach for identifying cellular heterogeneity, searching new and/or rare cell types, pseudotime inference, and inferring regulators underlying lineage changes (5, 1215). Most studies focused on searching cellular heterogeneity and new cell types in well-defined T cell subsets because of the high statistical power of scRNA-seq to distinguish cellular heterogeneities (1621). Only a few studies compared classic T cell subsets with scRNA-seq clustered populations (scCPops) (22); thus, our knowledge about the cellular relationship between classic T cell subsets and scCPops is limited. It remains unknown whether we could find the scCPop counterpart for each classic T cell subset, whether classic T cell subsets could represent well their scCPop counterparts, and whether more T cell subpopulations could be identified by reclustering multiple classic T cell subsets.

In this study, we integrated scRNA-seq data of six bead-enriched T cell subsets for investigating the cellular relationships between classic T cell subsets and scCPops. We found bead-enriched CD4+/CD45RA+/CD25 naive T cells (CD4 naive), CD8+/CD45RA+ naive T (CD8 naive) cells, and CD4 memory cells were mainly clustered into their scCPop counterparts, while cells from the other T cell subsets were assigned to multiple scCPops. The multiple T cell subsets that form a single scCPop exhibited similar expression patterns, but not vice versa, indicating that scCPops are more homogeneous cell populations. We unexpectedly identified a new T cell subpopulation that highly expresses IFN signaling–associated genes (ISAGs); we named it ISAG high T (ISAGhi T). Because BST2 is located on the cell membrane and is highly expressed in ISAGhi T cells, it was selected to enrich ISAGhi T cells. BST2+ T cells showed quick T cell activation upon anti-CD3/CD28 costimulation, indicating a high level of ISAGs may contribute to quick immune responses.

This study was approved by Institutional Review Boards at Southern University of Science and Technology (SUSTech). All experiments were conducted following the protocols approved by Institutional Review Boards at SUSTech. Human peripheral blood samples were obtained from a healthy adult blood donor (donor B). PBMCs were isolated by density-gradient sedimentation using Lymphoprep (Axis-Shield, Oslo, Norway) following the manufacturer’s instructions. In brief, a peripheral blood sample was diluted with an equal volume of PBS containing 2% FBS, then transferred into a Lymphoprep tube. After centrifugation at 800 × g for 20 min at room temperature, the middle white phase was transferred to a fresh tube. The enriched mononuclear cells were washed with PBS containing 2% FBS and centrifuged at 500 × g for 5 min twice. Cell count and viability were assessed by hemocytometer counting.

PBMCs were resuspended in FACS buffer (PBS containing 2% FBS) and incubated with Human TruStain FcX (Fc Receptor Blocking Solution; catalog [Cat.] #422302; BioLegend) for 15 min on ice. After anti-human CD14-allophycocyanin (61D3, Cat. #17-0149-42; eBioscience) and anti-human CD3-FITC (HIT3a, Cat. #11-0039-42; eBioscience) were added into PBMCs, they were incubated on ice for 30 min. These cells were washed twice with FACS buffer and counterstained with 1 μg/ml DAPI. These cells were sorted on FACS Aria-III (Becton Dickinson) with CD14-CD3+ gating to obtain T cells. FlowJo (version 10) was used for FACS data analyses and figure plot. Besides, we simultaneously added anti-human CD317-PE (RS38E, Cat. #348405; BioLegend) during Ab incubation for sorting BST2+ T or BST2hi T cells.

Following the manufacturer’s instructions, FACS-sorted BST2+ and BST2 T cells were stimulated with Dynabeads Human T-Activator CD3/CD28 for T Cell Expansion and Activation (Cat. #11161D; Thermo Fisher). In brief, 8 × 104 BST2+ or BST2 T cells sorted by FACS were suspended with 100–200 µL medium, respectively. Prewashed Dynabeads were added at the bead/cell ratio of 1:1 and incubated in a humidified CO2 incubator at 37°C. After 12 h, the bead-stimulated T cells were harvested by removing Dynabeads and then stained with anti-CD25 (Cat. #4302545; eBioscience) for FACS analyses of the efficiency of T cell activation.

FACS-sorted BST2+ T cells were directly processed for scRNA-seq with 10X Genomics 3′ kit v3 (10X Genomics, Pleasanton, CA) following the manufacturer’s instructions. In short, ∼1.6 × 104 BST2+ T cells were loaded into a single inlet of a 10X Genomics Chromium controller to generate the Gel Beads-in-Emulsions. Two independent biological replicates were prepared, with each replicate recovering ∼8000 cells. Gel Beads-in-Emulsions were used to complete the reverse transcription, cDNA amplification, and library preparation. scRNA-seq libraries were prepared using Chromium Single Cell 3′ Library Construction Kit v3 (P/N 1000078; 10X Genomics). Sequencing libraries were loaded at 2.4 pM on Illumina NovaSeq 6000 with 2 × 150 paired-end kits using the following read length: 28 bp Read1, 8 bp I7 Index, 8 bp I5 Index, and 91 bp Read2. About 40,000 reads per cell were generated for each sample.

We downloaded scRNA-seq data of six bead-enriched T cell subsets and PBMCs from the Web site of 10X Genomics (https://support.10xgenomics.com/single-cell-gene-expression/datasets). The six bead-enriched T cell subsets are CD4+ Th cells (CD4 Th), CD4 naive, CD4+/CD25+ Tregs (CD4 Tregs), CD4+/CD45RO+ memory T cells (CD4 memory) cells, CD8+ CTLs, and CD8 naive cells. FACS analyses showed the high purity of these T cell subsets: CD4 Th (99%), CD4 naive (98%), CD4 memory (98%), CD4 Treg (95%), CD8 naive (99%), and CD8 CTL (98%). All six bead-enriched T cell subsets and PBMCs in this part are from one healthy blood donor (donor A), and the basic information is available in Zheng et al. (23). These T cell subsets were processed by 10X Genomics Chromium with the aim of capturing 10,000 cells for each sample.

Cell ranger was used for generating the single-cell expression matrix. We filtered out the cells of bead-enriched T subsets meeting any of the following criteria: (1) <200 unique genes, (2) >1500 unique genes, (3) >50,00 unique molecular identifier (UMI) counts, and (4) >5% mitochondrial counts. However, the criterion for FACS-sorted BST2+ T cells is (1) <200 unique genes, (2) >6000 unique genes, (3) >45,000 UMI counts, or (4) >20% mitochondrial counts. After quality control, we normalized the expression level of each gene to the total number of UMIs for each cell.

Seurat (v3.0) (24) was used for data integration, data normalization, dimension reduction, cell clustering, and other basic scRNA-seq data analyses following our previous studies (5, 15). We calculated the cell-to-cell variation of each gene and selected the top 2000 genes exhibiting the highest variation for further analysis. To avoid highly expressed genes dominating in later analyses, we scaled the mean and variance of each gene across cells as 0 and 1, respectively. t-Distributed stochastic neighbor embedding (t-SNE) (15, 25, 26) in Seurat was used for data visualization. Cell embeddings of merged six bead-enriched T subsets cells, CD4 T cell subsets, and CD8 T cell subsets in the t-SNE plot were calculated based on 20 principal components (PCs), 0.6 resolution, and 35 perplexity. For fine analyses of CD4 T cell subsets, 1.0 resolution was used to obtain detail clusters. Although cell embeddings of FACS-sorted BST2+ T cells were calculated based on 50 PCs, 2.0 resolution, and 30 perplexity, we displayed cluster-specific expression genes on the t-SNE plot, which provide nice visualization for distinguishing different cell clusters.

To examine whether particular Gene Ontology (GO) terms were enriched in certain gene sets, we carried out GO enrichment analysis using DAVID (27). GO categories with adjusted p < 0.05 were considered as significant.

The Sankey diagram is an intuitive visualization approach for depicting flows from one set of values to another. We plotted the Sankey diagram using Highcharts (https://www.highcharts.com/demo/sankey-diagram) to explore the relationships between bead-enriched T cell subsets and scCPops.

We implemented Slingshot (28) on scRNA-seq data to infer the trajectory of CD4 and CD8 T cells. The scCPops were used as input for Slingshot, and the top 20 PCs were used to infer the pseudotime. Cell lineages were visualized on the PC analysis three-dimensional diagram using plot3D in R.

The statistical tests and plots were conducted using R version 3.5.2 (December 12, 2018) (29).

To analyze the relationship between classic T cell subsets and scCPops, we designed a workflow for single-cell analysis of six T cell subsets (Fig. 1A). The scRNA-seq data of six bead-enriched T cell subsets from a healthy donor (donor A), namely, CD4 Th cells, CD4 naive cells, CD4 memory cells, CD4 Tregs, CD8 CTLs, and CD8 naive cells, were obtained from Zheng et al. (23) (Table I). Analyses of each T cell subset showed there was no obvious subpopulation in each T cell subset except CD8 CTLs (Supplemental Fig. 1A), which is essentially consistent with the original report that these T cell subsets are relatively pure (23). The 62,235 cells from the six bead-enriched T cell subsets were projected on the t-SNE plot and formed a single connected entity (Fig. 1B). The CD4 T cell subsets, namely, CD4 naive, CD4 memory, CD4 Treg, and CD4 Th cells, are mixed together on the t-SNE plot. There are five distinct cell clusters (T cells, NK cells, B cells, CD14 monocytes, and CD16 monocytes) in the scRNA-seq data of PBMCs from donor A (Supplemental Fig. 1B, 1C). The six bead-enriched T cell subsets are overlapped with the T cells from PBMCs on the t-SNE plot (Fig. 1C), indicating the cellular heterogeneity among T cells are quite small compared with that between T cells and other cell types. Highlighting the T cell subsets one by one showed that each T cell subset was concentrated on a specific location of the t-SNE plot (Fig. 1D).

FIGURE 1.

The six bead-enriched T cell subsets and their features. (A) The schematic of this study. (B) t-SNE projection of the six bead-enriched T cell subsets, colored by each T cell subset. (C) t-SNE projection of PBMCs and the six T cell subsets. The PBMCs and six T cell subsets are colored differently. (D) Distribution of T cell subset in a t-SNE plot of the six bead-enriched T cell subsets, highlighting one T cell subset in each panel. (E) t-SNE projection of the six T cell subsets, with each cell colored based on their normalized expression of CD3D, CD3E, CD4, CD8B, CCR7, and GZMA, respectively.

FIGURE 1.

The six bead-enriched T cell subsets and their features. (A) The schematic of this study. (B) t-SNE projection of the six bead-enriched T cell subsets, colored by each T cell subset. (C) t-SNE projection of PBMCs and the six T cell subsets. The PBMCs and six T cell subsets are colored differently. (D) Distribution of T cell subset in a t-SNE plot of the six bead-enriched T cell subsets, highlighting one T cell subset in each panel. (E) t-SNE projection of the six T cell subsets, with each cell colored based on their normalized expression of CD3D, CD3E, CD4, CD8B, CCR7, and GZMA, respectively.

Close modal
Table I.

The scRNA-seq data of six bead-enriched T cell subsets and FACS-sorted BST2hi T cells after quality control

T Cell Subsets/Sample NamesMedian Genes per CellMedian UMI Counts per CellNo. of GenesNo. of Cells
Donor A [Zheng et al., 2017 (23)]     
 CD4 Th cells 546 1328 13,572 10,776 
 CD4 Tregs 547 1222 13,486 9,806 
 CD4 memory cells 547 1503 13,730 9,982 
 CD4 naive cells 493 1193 13,033 10,132 
 CD8 CTLs 572 1626 13,704 9,946 
 CD8 naive cells 502 1442 13,247 11,593 
Donor B (this study)     
 BST2hi #1 1406 4935 18,585 6,352 
 BST2hi #2 1373 5604 18,799 7,325 
T Cell Subsets/Sample NamesMedian Genes per CellMedian UMI Counts per CellNo. of GenesNo. of Cells
Donor A [Zheng et al., 2017 (23)]     
 CD4 Th cells 546 1328 13,572 10,776 
 CD4 Tregs 547 1222 13,486 9,806 
 CD4 memory cells 547 1503 13,730 9,982 
 CD4 naive cells 493 1193 13,033 10,132 
 CD8 CTLs 572 1626 13,704 9,946 
 CD8 naive cells 502 1442 13,247 11,593 
Donor B (this study)     
 BST2hi #1 1406 4935 18,585 6,352 
 BST2hi #2 1373 5604 18,799 7,325 

We further calculated the cell–cell variation of each gene. The genes showing the lowest cell–cell variation are enriched in housekeeping genes and T lineage–associated genes (Supplemental Fig. 1D, 1E). T cell marker genes CD3D and CD3E, encoding the δ- and ε-chain of CD3 protein, respectively, are expressed in all T cell subsets (Fig. 1E, Supplemental Fig. 1D). In contrast, the genes showing the highest cell–cell variation are enriched in immune effector process, T cell activation, immune effector process, cytokine signaling, and cytotoxic-related genes, including GNLY, CCL5, NKG7, GZMH, LYZ, GZMB, and GZMK (Supplemental Fig. 1F, 1G). CD4 and CD8B (encoding β-chain of CD8 protein) are highly expressed in the upper part and lower part of the t-SNE plot, respectively (Fig. 1E), consistent with the observation that CD4 and CD8 T cell subsets are in the upper part and lower part of the t-SNE plot, respectively. Expression of CCR7 (naive T cell marker gene in scRNA-seq) and GZMA indicate the naive/naive-like T cells are on the left of the t-SNE plot and cells secreting cytotoxic protein are on the right of the t-SNE plot, respectively. Overall, expression of marker genes on the t-SNE plot indicates that clustering cells by scRNA-seq data could provide further biological insights.

The 62,235 T cells from the six bead-enriched T cell subsets were clustered into nine clusters/scCPops by Seurat (24, 30) (Fig. 2A). The cell type of each scCPop was inferred based on scCPop-specific highly expressed genes (Fig. 2B, 2C, Supplemental Fig. 2A, 2B). The scCPops were named after their classic T cell counterpart with “sc” prefix. Among the nine T cell scCPops, five exhibit strong signals matching well-known classic T cell subsets: CD4 naive T cell (scCD4 TN; CD4, CCR7, LEF1, SELL), CD8 naive T cell (scCD8 TN; CD8B, CCR7, LEF1, SELL), CD4 memory T cell (scCD4 TM; CD4, TNFRSF4, TIMP1, USP10), CD8 effector memory T cell with CD45RA or CD8 effector cells (scCD8 TEMRA/scCD8 TEFF; CD8B, FGFBP2, GZMH, GZMB), and mucosal-associated invariant T cell (scMAIT; KLRB1, NCR3, ZBTB16, SLC4A10, RORC). Two scCPops exhibit strong signals matching CD4 Tregs (FOXP3, IL2RA, CCR10), which were designated as CD4 Treg 1 (scTreg#1) and CD4 Treg 2 (scTreg#2), respectively. Compared with scTreg#1, the scTreg#2 population is significantly higher expressed in MIR4435-1HG, LINC00152, KLRB1, LIMS1, NOSIP, and CD59 (Supplemental Fig. 2C), indicating scTreg#2 is in active proliferation (31, 32). The high expression of TIMP1 and TNFRSF4 in scCD4 TM potentially explains the long life of memory T cells because both genes have antiapoptotic function (33, 34).

FIGURE 2.

The nine scCPops and their features. (A) t-SNE projection of the 62,235 T cells, where each cell was grouped into one of the nine clusters distinguished by colors. Numbers in the panel indicate cluster identifier. (B) Normalized expression level and expression percentage of the cell-type-specific genes in the nine scCPops. Color represents normalized expression level, and dot size represents expression percentage. (C) t-SNE projection of the 62,235 T cells, colored by the nine scCPops, with inferred cell types on the panel. (D) Dot plot of scISAGhi-specific expressed genes. Color represents normalized expression level, and dot size represents expression percentage. (E) GO enrichment analysis of scISAGhi-specific expressed genes. (F) The number of CD4 T cells and CD8 T cells almost equally contributed to scTM. (G) scTM-specific expressed genes in their CD4 T subset and CD8 T subset. (H) DEGs between the CD4 T subset and the CD8 T subset of scTM.

FIGURE 2.

The nine scCPops and their features. (A) t-SNE projection of the 62,235 T cells, where each cell was grouped into one of the nine clusters distinguished by colors. Numbers in the panel indicate cluster identifier. (B) Normalized expression level and expression percentage of the cell-type-specific genes in the nine scCPops. Color represents normalized expression level, and dot size represents expression percentage. (C) t-SNE projection of the 62,235 T cells, colored by the nine scCPops, with inferred cell types on the panel. (D) Dot plot of scISAGhi-specific expressed genes. Color represents normalized expression level, and dot size represents expression percentage. (E) GO enrichment analysis of scISAGhi-specific expressed genes. (F) The number of CD4 T cells and CD8 T cells almost equally contributed to scTM. (G) scTM-specific expressed genes in their CD4 T subset and CD8 T subset. (H) DEGs between the CD4 T subset and the CD8 T subset of scTM.

Close modal

We identified 36 significantly high expressed genes in scCPop #5 by comparing scCPop #5 with the other cells. These most significant genes include ISG15, IFI6, OAS1, MX1, MX2, IRF7, IFI44L, XAF1, STAT1, IFIT1, IFIT3, and BST2 (Fig. 2B, 2D), which does not match any predefined T cell subsets. GO analyses of the 36 genes showed the most significantly enriched GO terms were type I IFN signaling pathway (4.9 × 10−25), defense response to virus (1.8 × 10−18), negative regulation of viral genome replication (3.1 × 10−17), response to virus (3.5 × 10−14), response to IFN-β (1.1 × 10−8), and IFN-γ–mediated signaling pathway (9.0 × 10−8) (Fig. 2E). We named this scCPop as scISAGhi T because all those significantly enriched GO terms belong to IFN signaling. The discovery of scISAGhi T in bead-enriched T cell subsets further pinpointed the cellular heterogeneity in well-defined classic T cell subsets.

Different from the other scCPops either belonging to the CD4 T cell subset or the CD8 T cell subset, scCPop #7 contains 41.39% CD4 T cells and 58.61% CD8 T cells (Fig. 2F). This scCPop was named scTM because it highly expressed memory T cell–specific genes, such as CXCR3, GZMK, TIMP1, LYAR, KLF6, DUSP1, JUN, and FOS (Fig. 2B, 2G). The CD8 T cells in scTM highly expressed NKG7, CCL5, CTSW, and CD7 (Fig. 2H), indicating those T cells are scCD8 TEM, while CD4 T cells highly expressed IL32, LTB, GZMA, S100A11, CXCR3, and GIMAP7 (Fig. 2H), indicating those cells are CD4 effector memory T cells re-expressing CD45RA (scCD4 TEMRA) or CTLs (scCD4-CTLs). It is well known that CD4 and CD8 T cells diverged into two functional different lineages after double-positive T cells (4). The scCD4 TEMRA/scCD4-CTL and scCD8 TEM cells were clustered into one subpopulation (scTM) potentially because both cell subsets express cytotoxic associated genes. The observation that CD4 and CD8 T cells could differentiate into similar phenotypes could be a model of functional convergence development.

The relationship between classic T cell subsets and scCPops was demonstrated by Sankey diagram (Fig. 3A). Bead-enriched CD4 naive and CD8 naive cells mainly clustered into their counterparts in scCPops, namely, scCD4 TN and scCD8 TN, respectively, indicating the consistency between classic T cell enrichment and scRNA-seq clustering on those cell populations. These bead-enriched T cells solely clustering into their scCPop counterparts also indicated those T cell subsets are relatively homogeneous. In contrast, CD4 Tregs and CD4 memory cells were assigned to multiple scCPops, indicating high heterogeneity in the two T cell subsets. For instance, Tregs were mainly assigned to five different scCPops, namely, scCD4 TN, scCD4 TM, scTM, scTreg#1, and scTreg#2. As expected, CD4 Th cells and CD8 CTLs contributed to all CD4 scCPops and all CD8 scCPops, respectively. Although there are differences between scCPops and their classic counterparts, classic T subsets and their scCPop counterparts showed the highest similarity among all T cell populations (Supplemental Fig. 2E). Furthermore, the correlations between naive T subsets and their counterparts are higher than that of the other T cell subsets. Differentially expressed gene (DEG) analyses showed the number of DEGs between naive T cells and their counterparts was much less than that of other T cell subsets (Supplemental Fig. 2D). These results are consistent with the Sankey diagram that naive T cell subsets are much more similar to their counterparts than other T cell subsets.

FIGURE 3.

Relationships between bead-enriched T cell subsets and scCPops. (A) Cellular relationships between the six bead-enriched T cell subsets and the nine scCPops by Sankey diagram. Each node represents a cell population, and the width of line between two nodes represents number of shared cells. (B) Cell compositions of bead-enriched T cell subsets by bar chart. (C) Bar plot displays cell compositions of scCPops based on bead-enriched T cell subsets. (D) Gene expression heatmap of cell subpopulations generated by combinations of scCPops and bead-enriched T cell subsets.

FIGURE 3.

Relationships between bead-enriched T cell subsets and scCPops. (A) Cellular relationships between the six bead-enriched T cell subsets and the nine scCPops by Sankey diagram. Each node represents a cell population, and the width of line between two nodes represents number of shared cells. (B) Cell compositions of bead-enriched T cell subsets by bar chart. (C) Bar plot displays cell compositions of scCPops based on bead-enriched T cell subsets. (D) Gene expression heatmap of cell subpopulations generated by combinations of scCPops and bead-enriched T cell subsets.

Close modal

Cell composition analyses of bead-enriched T cell subsets based on scCPops directly demonstrated the cellular heterogeneity in classic T cell subsets. The CD4 naive subset consists of 91.21% scCD4 TN, 5.56% scCD8 TN, 1.74% scCD4 TM, and 0.63% scTreg#1 (Fig. 3B). The CD8 naive subset consists of 85.23% scCD8 TN and 12.79% scCD4 TN. The CD4 memory subset consists of 70.35% scCD4 TM, 10.79% scTreg#1, 7.47% scTM, and 7.35% scCD4 TN (Fig. 3B). The CD4 Treg subset consists of 41.72% scCD4 TM, 33.86% scTreg#1, 16.57% scCD4 TN, 3.26% scTreg#2, and 3.07% scTM (Fig. 3B). The CD4 Th cell subset consists of 50.75% scCD4 TN, 34.11% scCD4 TM, 6.98% scTReg#1, 4.32% scTM, and 2.14% scCD8 TN (Fig. 3B). CD8 CTLs consist of 50.96% scCD8 TN, 21.03% scTM, 10.82% scCD8 TEMRA, 10.49% scMAIT, 3.66% scCD4 TN, and 2.31% scCD4 TM, which should be the most heterogeneous among the six bead-enriched T cell subsets.

We further calculated the cell composition of scCPops based on classic T cell subsets. Both scMAIT and scCD8 TEMRA were almost solely derived from CD8 CTLs (Fig. 3A, 3C). However, the cell compositions of the other seven scCPops were quite complex, with each scCPop deriving from several bead-enriched T cell subsets (Fig. 3C). In particular, all six bead-enriched T cell subsets contributed to scISAGhi T, arranging from 2.54 to 26.92% (Fig. 3C). These results indicated scISAGhi T was a rare cell population accounting for 0.26% of the analyzed cells, with CD4 T cell subsets containing a high fraction of scISAGhi T. scISAGhi T accounts for ∼0.4% of CD4 naive, CD4 Th, and CD4 Tregs and 0.24% of CD4 memory cells, while it accounts for only 0.04% of CD8 CTLs and 0.09% of CD8 naive cells (Supplemental Fig. 2F).

The T cells were separated into 54 cell subpopulations by combinations of nine scCPops and six bead-enriched T cell subsets. Finally, 45 cell subpopulations were left after we removed the subpopulations with the number of cells <10. Cell subpopulations deriving from different bead-enriched T cell subsets while belonging to same scCPop always showed similar expression patterns (Fig. 3D). In contrast, cell subpopulations deriving from different scCPops while belonging to same bead-enriched T cell subset usually showed quite different expression patterns (Fig. 3D, Supplemental Fig. 2G). These results are consistent with our assumption that scCPops better represent the cell states than that of bead-enriched T cell subsets.

The CD4 T cell subsets (Supplemental Fig. 3A) were further analyzed to better understand the subpopulations and lineages of CD4 T cells. The 40,696 cells from the CD4 Th, CD4 naive, CD4 memory, and CD4 Treg populations formed a single connected entity on the t-SNE plot and different CD4 T cell subsets mixed together (Fig. 4A). We clustered these T cells into six scCPops (Fig. 4B, 4C), among which four scCPops have been identified as mentioned earlier, namely, scCD4 TN (CCR7, SELL, LEF1), scISAGhi T (OAS1, IFI6, MX1, BST2), scCD4 TEMRA (GZMA, CCL5, GZMK, CST7, LYAR), and scTreg (FOXP3, IL2RA, IL10RA, CD59). The original scCD4 TM cells were separated into central memory T cells (scCD4 TCM) (CCR7, SELL, TCF7) and effector memory T cells (scTEMs) (CCR7, TIMP1, LGALS1, USP10), respectively (Fig. 4B, 4C). The Sankey diagram showed that the cellular relationships between the four bead-enriched CD4 T cell subsets and the six scCPops were consistent with the earlier-mentioned analyses of all T cells. In brief, bead-enriched CD4 naive cells almost solely clustered into the counterpart scCD4 TN. In contrast, CD4 Th, CD4 memory, and CD4 Tregs contributed to all six scCPops (Fig. 4D).

FIGURE 4.

Reclassification of CD4 T cells and single-cell trajectory. (A) t-SNE projection of CD4 T cells, with each cell colored according to bead-enriched T cell subsets that are distinguished by colors. (B) Normalized expression level and expression percentage of the cell-type-specific genes in six CD4 T cell scCPops. (C) t-SNE projection of CD4 T cells, where each cell is grouped into one of the six scCPops, with inferred cell types on the panel. (D) Cellular relationships between the four CD4 T cell subsets and the six scCPops by Sankey diagram. Each node represents a cell population, and the width of line between two nodes represents number of shared cells. (E) The original 6 scCPops were further separated into 10 scCPops by setting a high resolution. (F) Single-cell trajectories of CD4 T cells inferred by Slingshot. The color of each scCPop is the same as that in (C).

FIGURE 4.

Reclassification of CD4 T cells and single-cell trajectory. (A) t-SNE projection of CD4 T cells, with each cell colored according to bead-enriched T cell subsets that are distinguished by colors. (B) Normalized expression level and expression percentage of the cell-type-specific genes in six CD4 T cell scCPops. (C) t-SNE projection of CD4 T cells, where each cell is grouped into one of the six scCPops, with inferred cell types on the panel. (D) Cellular relationships between the four CD4 T cell subsets and the six scCPops by Sankey diagram. Each node represents a cell population, and the width of line between two nodes represents number of shared cells. (E) The original 6 scCPops were further separated into 10 scCPops by setting a high resolution. (F) Single-cell trajectories of CD4 T cells inferred by Slingshot. The color of each scCPop is the same as that in (C).

Close modal

We further clustered the CD4 T cells into 10 scCPops for better understanding of its fine substructure (Fig. 4E). The original scTregs were separated into central memory Tregs (FOXP3, CCR7, LEF1, CTLA4low) and effector memory Tregs (FOXP3, IL2RA, LGALS1, ANXA2, CTLA4) (Supplemental Fig. 3B). The original scCD4 TN cells were separated into scCD4 TN#1 and scCD4 TN#2, with most CD4 naive-specific genes, such as CCR7, SELL, LEF1, and TCF7, expressing higher in scCD4 TN#1 (Supplemental Fig. 3B). The original scCD4 TEM cells were divided into scCD4 TEM#1 and scCD4 TEM#2 (Fig. 4E), with KLRB1 and PTGER2 expressing higher in scCD4 TEM#1, potentially indicating scCD4 TEM#1 is an intermediate state between scCD4 TEM#2 and scCD4 TEMRA. The Sankey diagram showed that CD4 Th, CD4 memory, and CD4 Tregs contributed to almost all 10 fine scCPops, indicating the high cell heterogeneity of bead-enriched T cells subsets (Supplemental Fig. 3C).

Application of Slingshot (28) on these CD4 T cells identified four pseudotime lineages: (1) scCD4 TN → scCD4 TCM → scCD4 TEM → scCD4 TEMRA; (2) scCD4 TN → scCD4 TCM → scCD4 TEM; (3) scCD4 TN → scCD4 TCM → scCD4 Treg; and (4) scCD4 TN → scISAGhi (Fig. 4F). The CD4 T cell differentiation was a continuous process in which naive cells first progressed into central memory–like cells and then developed into different functional CD4 T cells. The first inferred lineage was consistent with recent reports on CD4 T cell differentiation (20, 35, 36), which was the most important lineage because it included most of the CD4 T cells.

We integrated three T cell populations, namely, bead-enriched CD8 CTLs, bead-enriched CD8 naive cells, and scCD4 TEMRA/scCD4-CTLs in scTM (Supplemental Fig. 3D), for detecting fine substructure in these T cell subsets and providing insight into the relationship between CD4 CTLs and CD8 CTLs. The three T cell populations formed one major group, which mainly consisted of CD8 naive cells and several small groups in which scCD4 TEMRA overlapped with some CD8 CTLs on the t-SNE plot (Fig. 5A). We further clustered these T cells into six scCPops (Fig. 5B, 5C), namely, scCD8 TN (CCR7, SELL, LEF, CD69low), scCD8 TCM (CCR7, SELL, LEF1, CD69), scTM (CXCR3, TRADD, GZMA, GZMK, CCL5), scCD8 TEMRA (FGFBP2, GZMH, GZMB, GNLY), scMAIT (KLRB1, ZBTB16, NCR3, SLC4A10, RORC), and NK cells (scNKT; NCR3, KLRC2, ZNF683, XCL1) (Fig. 5B, 5C). The expression pattern and cell composition of scTM is the same as the aforementioned scTM in (Fig. 2. The expression of commonly used T cell markers on the t-SNE plot showed obvious scCPop-specific features (Fig. 5D). A volcano plot showed that both MAIT and NKT expressed many cell-type-specific genes (Supplemental Fig. 3E, 3F), further indicating that MAITs and NKTs are outliers in CD8 CTLs.

FIGURE 5.

Reclassification of potential CTLs based on scRNA-seq and single-cell trajectory. (A) t-SNE projection of CD8 naive cells, CD8 CTLs, and scCD4 TEMRA cells, with each cell colored according to cell population. (B) Normalized expression level and expression percentage of the cell-type-specific genes in the six scCPops. (C) t-SNE projection of CTLs, where each cell is grouped into one of the six scCPops, with inferred cell types indicated in the panel. (D) t-SNE projection of CTLs, with each cell colored based on its normalized expression of CD4, CD8B, CCR7, GZMA, GZMH, and KLRB1, respectively. (E) Relationships between the initial T cell subsets and the six scCPops by Sankey diagram. Each node represents a cell population, and the width of line between two nodes represents the number of shared cells. (F) Single-cell trajectory of CD8 CTL inferred by Slingshot. The color of each scCPops is the same as that in (C).

FIGURE 5.

Reclassification of potential CTLs based on scRNA-seq and single-cell trajectory. (A) t-SNE projection of CD8 naive cells, CD8 CTLs, and scCD4 TEMRA cells, with each cell colored according to cell population. (B) Normalized expression level and expression percentage of the cell-type-specific genes in the six scCPops. (C) t-SNE projection of CTLs, where each cell is grouped into one of the six scCPops, with inferred cell types indicated in the panel. (D) t-SNE projection of CTLs, with each cell colored based on its normalized expression of CD4, CD8B, CCR7, GZMA, GZMH, and KLRB1, respectively. (E) Relationships between the initial T cell subsets and the six scCPops by Sankey diagram. Each node represents a cell population, and the width of line between two nodes represents the number of shared cells. (F) Single-cell trajectory of CD8 CTL inferred by Slingshot. The color of each scCPops is the same as that in (C).

Close modal

Bead-enriched CD8 naive cells were mainly clustered into its counterpart scCD8 TN, with a small fraction clustered into scCD8 TCM (Fig. 5E). scCD4 TEMRA, together with a fraction of cells from CD8 CTL, formed scTM, consistent with our earlier results. The cells from CD8 CTLs contributed to each of the six scCPops (Fig. 5E), indicating multiple T cell subsets in commonly used CD8 CTLs. Especially, the existence of an unignorable amount of MAITs and NKTs in CD8 CTLs indicated that the composition of classic CD8 CTL is complex.

We removed scMAIT, scCD4 TEMRA, and scNKT cells for inferring CD8 T cell differentiation lineage because the three cell populations do not belong to well-defined CD8 T cells. By conducting Slingshot (28), we inferred the pseudotime trajectory of CD8 T cell differentiation: scCD8 TN → scCD8 TCM → scCD8 TEM → scCD8 TEMRA (Fig. 5F), which was similar to the CD4 main lineage. Based on inferred lineage and cell projection on the top PCs, the differentiation from CD8 naive T to scCD8 TEMRA cells was a continuous process (Fig. 5F).

Although BST2 showed weak specificity for only scISAGhi T cells (Fig. 2D), it is the only cell surface gene with a well-established Ab among ISAGs. We enriched scISAGhi T by FACS of BST2hi cells from PBMCs and made two independent scRNA-seq libraries, BST2hi T#1 and BST2hi T#2 (Fig. 6A, Table I, Supplemental Fig. 4A). A total of 13,677 FACS-sorted BST2hi T cells were clustered into ISAGhi T, CD4 TN, CD8 TN, CD4 TM, CD8 TM, proliferated CD4 T, MAIT, γδ T, NKT, effector Treg, and naive Tregs (Fig. 6B, Supplemental Fig. 4B). Although the enrichment of ISAGhi T cells has significantly increased, ISAGhi T accounts for only a moderate fraction of the BST2hi cell because BST2 is expressed in many T cell subsets. The expression level of BST2 in ISAGhi T is significantly higher than that of other cells based on the scRNA-seq data (Fig. 6C). We further separated the non-ISAGhi T cells into BST2+ and BST2 T cells based on whether BST2 was detected in scRNA-seq data (Fig. 6D). The expression profile of ISAGhi T cells is quite different from that of BST2+ and BST2 T cells, while the expression profiles of BST2+ and BST2 T cells are more similar (Fig. 6E, 6F). ISAGhi T cell has 89 genes with significantly higher expression, including IF144L, MX1, IF1616, XAF1, ISG15, OAS3, IFI44, IRF7, MX2, IFITM1, OASL and IFI35, than BST2+ T cells (Fig. 6G). These ISAGhi T cell–specific genes were significantly enriched in IFN signaling, defense response to virus, and response to IFN-γ and antiviral mechanism by IFN-stimulated genes (Supplemental Fig. 4C). In summary, high expression of ISAGs is the most significant feature of ISAGhi T cells.

FIGURE 6.

Enrichment of ISAGhi T cells by FACS and features of ISAGhi T cells. (A) Sorting of BST2hi T cells from PBMCs by FACS. (B) t-SNE projection of 13,644 sorted BST2hi T cells, with each cell colored according to cell subpopulation. (C) Box plot of BST2 expression of ISAGhi T cells and other cells in BST2hi T cells. Asterisks represent a significant difference (***p < 0.001). (D) t-SNE projection of BST2hi T cells, with each cell colored according to ISAGhi T, BST2+ T, and BST2 T cells. BST2+ and BST2 T cells are defined by expression of BST2 in scRNA-seq data. (E) Spearman correlation coefficient between ISAGhi T, BST2+ T, and BST2 T cells. Color represents correlation coefficient between two groups. (F) Heatmap of DEGs between ISAGhi T, BST2+ T, and BST2 T cells. Color represents average expression of genes in each group. (G) Volcano plot for DEGs between ISAGhi T and BST2+ T cells (p < 0.01). Red points represent ISAGhi T cell–specific high expressed genes, while blue points represent BST2+ T cell–specific high expressed genes. CD4 Tp, proliferated CD4 T; eTreg, effector Treg; nTreg, naive Treg.

FIGURE 6.

Enrichment of ISAGhi T cells by FACS and features of ISAGhi T cells. (A) Sorting of BST2hi T cells from PBMCs by FACS. (B) t-SNE projection of 13,644 sorted BST2hi T cells, with each cell colored according to cell subpopulation. (C) Box plot of BST2 expression of ISAGhi T cells and other cells in BST2hi T cells. Asterisks represent a significant difference (***p < 0.001). (D) t-SNE projection of BST2hi T cells, with each cell colored according to ISAGhi T, BST2+ T, and BST2 T cells. BST2+ and BST2 T cells are defined by expression of BST2 in scRNA-seq data. (E) Spearman correlation coefficient between ISAGhi T, BST2+ T, and BST2 T cells. Color represents correlation coefficient between two groups. (F) Heatmap of DEGs between ISAGhi T, BST2+ T, and BST2 T cells. Color represents average expression of genes in each group. (G) Volcano plot for DEGs between ISAGhi T and BST2+ T cells (p < 0.01). Red points represent ISAGhi T cell–specific high expressed genes, while blue points represent BST2+ T cell–specific high expressed genes. CD4 Tp, proliferated CD4 T; eTreg, effector Treg; nTreg, naive Treg.

Close modal

We reconducted dimension reduction and clustering analysis on BST2hi T cells using gene set–excluded ISAGs; the scCPops did not change, except the ISAGhi T cluster disappeared (Fig. 7A). On the t-SNE plot based on gene set without ISAGs, ISAGhi T cells dispersed into other scCPops, in particular concentrated on CD8 TN (39.41%), CD4 TN (28.39%), and CD4 TM (28.39%) cells, although only a few ISAGhi T cells dispersed into innate-like T cells, such as MAIT, γδ T, and NKT cells (Fig. 7B, Supplemental Fig. 4D). These results are expected because the expression pattern of ISAGhi T cells is very similar to CD4 TN, CD8 TN, and CD4 TM cells (Supplemental Fig. 4B). After normalizing the cell number, the percentage of ISAGhi T cells in CD8 TN, CD4 TN, and CD4 TM subsets is similar (Supplemental Fig. 4E). These results are consistent with our earlier-mentioned observations that bead-enriched CD4 T cell subsets have a much higher fraction of cells that were clustered into ISAGhi T cells (Supplemental Fig. 2F). Interestingly, the ISAGhi T cell cluster appeared when using only ISAGs to do the same analysis (Fig. 7C), further indicating that ISAGs are a major contributor for the formation of ISAGhi T cell cluster.

FIGURE 7.

ISAGs are the major contributor of ISAGhi T cell cluster, and a high level of ISAGs may contribute to quick T cell activation. (A) t-SNE projection of BST2hi T cells using gene sets that removed ISAGs, in which ISAGhi T cell cluster disappears. (B) t-SNE projection of BST2hi T cells using gene sets that removed ISAGs, with red points highlighting the original identified ISAGhi T cells. (C) t-SNE projection of BST2hi T cells using ISAGs, with red points highlighting original identified ISAGhi T cells. (D and E) FACS analysis of CD25 showed almost all BST2+ T cells become activated upon anti-CD3/CD28 beads for 12 h, while only a fraction of BST2 T cells become activated at the same time (two experiment repeats).

FIGURE 7.

ISAGs are the major contributor of ISAGhi T cell cluster, and a high level of ISAGs may contribute to quick T cell activation. (A) t-SNE projection of BST2hi T cells using gene sets that removed ISAGs, in which ISAGhi T cell cluster disappears. (B) t-SNE projection of BST2hi T cells using gene sets that removed ISAGs, with red points highlighting the original identified ISAGhi T cells. (C) t-SNE projection of BST2hi T cells using ISAGs, with red points highlighting original identified ISAGhi T cells. (D and E) FACS analysis of CD25 showed almost all BST2+ T cells become activated upon anti-CD3/CD28 beads for 12 h, while only a fraction of BST2 T cells become activated at the same time (two experiment repeats).

Close modal

Constitutive expression of ISAGs plays pivotal roles in host responses to malignant cells and pathogens, such as virus (3739). We used FACS to sort the T cells into BST2+ and BST2 T cells because BST2 level indicates activation level of ISAGs in some way. The sorted BST2+ and BST2 T cells were activated by Dynabeads human T activator CD3/CD28 for 12 h. FACS analysis of CD25 showed almost all BST2+ T cells become activated upon anti-CD3/CD28 costimulation, whereas only a fraction of BST2 T cells become activated (Fig. 7D, 7E, Supplemental Fig. 4F, 4G). These results may indicate that almost all BST2+ T cells have been well prepared for T cell activation, whereas only a fraction of BST2 T cells have been prepared for quick T cell activation.

It is assumed that each T cell subset has specific functions, and several T cell subsets may coordinate with each other to complete a specific function (40, 41). The recent advance of scRNA-seq provided an alternative approach to identify the T cell subpopulations and cell heterogeneity. In this study, we reclustered the six bead-enriched T cell subsets using scRNA-seq data. Among the six T cell subsets, naive T cells and memory T cells account for the majority of the T cells in this study (Figs. 2B, 2C, 3A), and the signals from memory T cells and naive T cells were strong. Therefore, we mainly focused on memory T cells and naive T cells, while ignoring other T cell subsets, such as Th1, Th2, and Th17.

We inferred the cell type of each scCPop and its counterpart in bead-enriched T cell subsets using scCPop-specific genes. The naive T subsets were mainly clustered into their scCPop counterparts, while other classic T subsets were dispersed into multiple scCPops. Correlation analyses and DEG analyses showed that the similarities between naive T cell subsets and their scCPop counterparts are much greater than that of other T cell subsets, indicating the naive T cell subsets are relatively pure compared with other bead-enriched T cell subsets. Because the expression of population-specific genes in scCPops are stronger and more homogeneous than that of their bead-enriched T cell counterpart, the statistical parameters of scCPops should be good representatives of features of the T cell population. Furthermore, we increased clustering resolution and identified very specific T cell subpopulations, such as scCD4 TCM, scCD4 TEM, scCD4 TEMRA, central memory Tregs, effector memory Tregs, scCD8 TCM, scCD8 TEM, and scCD8 TEMRA. These results showed that scRNA-seq clustering is essentially consistent with the classic FACS/bead-enriched approach. In addition, scRNA-seq clustering identified rare cell subpopulations, such as ISAGhi T, and inferred very specific subpopulations, indicating its advantage compared with the classic approach. Further analyses revealed that ISAGhi T showed specific features and may contribute to the quick T cell activation.

Although CD4 and CD8 T cells essentially are distinguishable based on cell surface markers, the boundaries between CD4 and CD8 T cells were blurred based on scRNA-seq data (Fig. 1B), indicating the overall transcriptional profiles between CD4 and CD8 T cells were not completely different. In particular, we identified the scTM cluster that contains both CD4 and CD8 T cells. Furthermore, we inferred the trajectories of CD4 and CD8 T cells, which showed that CD4 naive T cells and CD8 naive T cells differentiated into CD4 TEMRA and CD8 TEMRA, respectively. Reinvestigation of T cell subsets by scRNA-seq is of great significance for understanding T cell subsets and helps us to further understand the mechanism of the human adaptive immune response, which also has important implications for the ongoing immunotherapy.

We thank Dr. Keji Zhao for critical reading of the manuscript. We thank members of the Flow Cytometry facility in SUSTech for the cell sorting. We thank members of the Center for Computational Science and Engineering in SUSTech for computational support.

This work was supported by the National Key Research and Development Program of China Grant 2018YFC1004500, National Natural Science Foundation of China Grants 81872330 and 31771603, Shenzhen Science and Technology Program Grants KQTD20180411143432337, JCYJ20170412153453623, and JCYJ20180504170158430, and Shenzhen Innovation Committee of Science and Technology Grant ZDSYS20200811144002008.

W.J. conceived and designed the project. S.C. did the experiment. X.W. and X.S. analyzed the data, with contributions from H.L. W.J., N.H., H.Z., and X.C. supervised the study. W.J. and X.W. wrote the manuscript. All authors have read and approved the manuscript.

The raw sequence data reported in this article have been submitted to the Genome Sequence Archive in Beijing Institute of Genomics Data Center (https://ngdc.cncb.ac.cn/gsa-human/browse/HRA000313) under accession number HRA 000313.

The online version of this article contains supplemental material.

Abbreviations used in this article

Cat.

catalog

CD4 memory

CD4+/CD45RO+ memory T

CD4 naive

CD4+/CD45RA+/CD25 naive T

CD8 naive

CD8+/CD45RA+ naive T

CD4 Th

CD4+ Th cell

CD4 Treg

CD4+/CD25+ regulatory T cell

DEG

differentially expressed gene

GO

Gene Ontology

ISAG

IFN signaling–associated gene

ISAGhi T

IFN signaling–associated gene high T

MAIT

mucosal-associated invariant T cell

PC

principal component

scCD4 TEMRA

CD4 effector memory T cell re-expressing CD45RA scCPop

scCD8 TEMRA/scCD8 TEFF

CD8 effector memory T cell with CD45RA or CD8 effector cell

scCD4 TM

CD4 memory T cell scCPop

scCD4 TN

CD4 naive T cell scCPop

scCD8 TN

CD8 naive T cell scCPop

scCPop

scRNA-seq clustered population

scMAIT

mucosal-associated invariant T cell scCPop

scRNA-seq

single-cell RNA sequencing

scTM

scCPop #7

SUSTech

Southern University of Science and Technology

Treg

regulatory T cell

t-SNE

t-distributed stochastic neighbor embedding

UMI

unique molecular identifier.

1.
Zhang
L.
,
X.
Yu
,
L.
Zheng
,
Y.
Zhang
,
Y.
Li
,
Q.
Fang
,
R.
Gao
,
B.
Kang
,
Q.
Zhang
,
J. Y.
Huang
, et al
2018
.
Lineage tracking reveals dynamic relationships of T cells in colorectal cancer.
Nature
564
:
268
272
.
2.
Farh
K. K.
,
A.
Marson
,
J.
Zhu
,
M.
Kleinewietfeld
,
W. J.
Housley
,
S.
Beik
,
N.
Shoresh
,
H.
Whitton
,
R. J.
Ryan
,
A. A.
Shishkin
, et al
2015
.
Genetic and epigenetic fine mapping of causal autoimmune disease variants.
Nature
518
:
337
343
.
3.
Farber
D. L.
,
N. A.
Yudanin
,
N. P.
Restifo
.
2014
.
Human memory T cells: generation, compartmentalization and homeostasis.
Nat. Rev. Immunol.
14
:
24
35
.
4.
Hu
G.
,
K.
Cui
,
D.
Fang
,
S.
Hirose
,
X.
Wang
,
D.
Wangsa
,
W.
Jin
,
T.
Ried
,
P.
Liu
,
J.
Zhu
, et al
2018
.
Transformation of accessible chromatin and 3D nucleome underlies lineage commitment of early T cells.
Immunity
48
:
227
242.e8
.
5.
Qin
P.
,
Y.
Pang
,
W.
Hou
,
R.
Fu
,
Y.
Zhang
,
X.
Wang
,
G.
Meng
,
Q.
Liu
,
X.
Zhu
,
N.
Hong
, et al
2021
.
Integrated decoding hematopoiesis and leukemogenesis using single-cell sequencing and its medical implication.
Cell Discov.
7
:
2
.
6.
Zhu
J.
,
W. E.
Paul
.
2008
.
CD4 T cells: fates, functions, and faults.
Blood
112
:
1557
1569
.
7.
Mosmann
T. R.
,
H.
Cherwinski
,
M. W.
Bond
,
M. A.
Giedlin
,
R. L.
Coffman
.
1986
.
Two types of murine helper T cell clone. I. Definition according to profiles of lymphokine activities and secreted proteins.
J. Immunol.
136
:
2348
2357
.
8.
Sakaguchi
S.
,
N.
Sakaguchi
,
M.
Asano
,
M.
Itoh
,
M.
Toda
.
1995
.
Immunologic self-tolerance maintained by activated T cells expressing IL-2 receptor alpha-chains (CD25). Breakdown of a single mechanism of self-tolerance causes various autoimmune diseases.
J. Immunol.
155
:
1151
1164
.
9.
Fontenot
J. D.
,
M. A.
Gavin
,
A. Y.
Rudensky
.
2003
.
Foxp3 programs the development and function of CD4+CD25+ regulatory T cells.
Nat. Immunol.
4
:
330
336
.
10.
Suni
M. A.
,
S. A.
Ghanekar
,
D. W.
Houck
,
H. T.
Maecker
,
S. B.
Wormsley
,
L. J.
Picker
,
R. B.
Moss
,
V. C.
Maino
.
2001
.
CD4(+)CD8(dim) T lymphocytes exhibit enhanced cytokine expression, proliferation and cytotoxic activity in response to HCMV and HIV-1 antigens.
Eur. J. Immunol.
31
:
2512
2520
.
11.
Intlekofer
A. M.
,
N.
Takemoto
,
E. J.
Wherry
,
S. A.
Longworth
,
J. T.
Northrup
,
V. R.
Palanivel
,
A. C.
Mullen
,
C. R.
Gasink
,
S. M.
Kaech
,
J. D.
Miller
, et al
2005
.
Effector and memory CD8+ T cell fate coupled by T-bet and eomesodermin. [Published erratum appears in 2006 Nat. Immunol. 7: 113.]
Nat. Immunol.
6
:
1236
1244
.
12.
Wang
W.
,
G.
Ren
,
N.
Hong
,
W.
Jin
.
2019
.
Exploring the changing landscape of cell-to-cell variation after CTCF knockdown via single cell RNA-seq.
BMC Genomics
20
:
1015
.
13.
Macosko
E. Z.
,
A.
Basu
,
R.
Satija
,
J.
Nemesh
,
K.
Shekhar
,
M.
Goldman
,
I.
Tirosh
,
A. R.
Bialas
,
N.
Kamitaki
,
E. M.
Martersteck
, et al
2015
.
Highly parallel genome-wide expression profiling of individual cells using nanoliter droplets.
Cell
161
:
1202
1214
.
14.
Paul
F.
,
Y.
Arkin
,
A.
Giladi
,
D. A.
Jaitin
,
E.
Kenigsberg
,
H.
Keren-Shaul
,
D.
Winter
,
D.
Lara-Astiaso
,
M.
Gury
,
A.
Weiner
, et al
2015
.
Transcriptional heterogeneity and lineage commitment in myeloid progenitors. [Published erratum appears in 2016 Cell. 164: 325.]
Cell
163
:
1663
1677
.
15.
Zhou
B.
,
W.
Jin
.
2020
.
Visualization of single cell RNA-seq data using t-SNE in R.
Methods Mol. Biol.
2117
:
159
167
.
16.
Yu
Y.
,
J. C.
Tsang
,
C.
Wang
,
S.
Clare
,
J.
Wang
,
X.
Chen
,
C.
Brandt
,
L.
Kane
,
L. S.
Campos
,
L.
Lu
, et al
2016
.
Single-cell RNA-seq identifies a PD-1hi ILC progenitor and defines its development pathway.
Nature
539
:
102
106
.
17.
Proserpio
V.
,
A.
Piccolo
,
L.
Haim-Vilmovsky
,
G.
Kar
,
T.
Lönnberg
,
V.
Svensson
,
J.
Pramanik
,
K. N.
Natarajan
,
W.
Zhai
,
X.
Zhang
, et al
2016
.
Single-cell analysis of CD4+ T-cell differentiation reveals three major cell states and progressive acceleration of proliferation. [Published erratum appears in 2016 Genome Biol. 17: 133.]
Genome Biol.
17
:
103
.
18.
Waickman
A. T.
,
K.
Victor
,
T.
Li
,
K.
Hatch
,
W.
Rutvisuttinunt
,
C.
Medin
,
B.
Gabriel
,
R. G.
Jarman
,
H.
Friberg
,
J. R.
Currier
.
2019
.
Dissecting the heterogeneity of DENV vaccine-elicited cellular immunity using single-cell RNA sequencing and metabolic profiling.
Nat. Commun.
10
:
3666
.
19.
Zemmour
D.
,
R.
Zilionis
,
E.
Kiner
,
A. M.
Klein
,
D.
Mathis
,
C.
Benoist
.
2018
.
Single-cell gene expression reveals a landscape of regulatory T cell phenotypes shaped by the TCR. [Published erratum appears in 2018 Nat. Immunol. 19: 645.]
Nat. Immunol.
19
:
291
301
.
20.
Patil
V. S.
,
A.
Madrigal
,
B. J.
Schmiedel
,
J.
Clarke
,
P.
O’Rourke
,
A. D.
de Silva
,
E.
Harris
,
B.
Peters
,
G.
Seumois
,
D.
Weiskopf
, et al
2018
.
Precursors of human CD4+ cytotoxic T lymphocytes identified by single-cell transcriptome analysis.
Sci. Immunol.
3
:
eaan8664
.
21.
Chattopadhyay
P. K.
,
T. M.
Gierahn
,
M.
Roederer
,
J. C.
Love
.
2014
.
Single-cell technologies for monitoring immune systems.
Nat. Immunol.
15
:
128
135
.
22.
Hao
Y.
,
S.
Hao
,
E.
Andersen-Nissen
,
W. M.
Mauck
III
,
S.
Zheng
,
A.
Butler
,
M. J.
Lee
,
A. J.
Wilk
,
C.
Darby
,
M.
Zager
, et al
2021
.
Integrated analysis of multimodal single-cell data.
Cell
184
:
3573
3587.e29
.
23.
Zheng
G. X.
,
J. M.
Terry
,
P.
Belgrader
,
P.
Ryvkin
,
Z. W.
Bent
,
R.
Wilson
,
S. B.
Ziraldo
,
T. D.
Wheeler
,
G. P.
McDermott
,
J.
Zhu
, et al
2017
.
Massively parallel digital transcriptional profiling of single cells.
Nat. Commun.
8
:
14049
.
24.
Butler
A.
,
P.
Hoffman
,
P.
Smibert
,
E.
Papalexi
,
R.
Satija
.
2018
.
Integrating single-cell transcriptomic data across different conditions, technologies, and species.
Nat. Biotechnol.
36
:
411
420
.
25.
van der Maaten
L.
,
G.
Hinton
.
2008
.
Visualizing data using t-SNE.
J. Mach. Learn. Res.
9
:
2579
2605
.
26.
Mahfouz
A.
,
M.
van de Giessen
,
L.
van der Maaten
,
S.
Huisman
,
M.
Reinders
,
M. J.
Hawrylycz
,
B. P.
Lelieveldt
.
2015
.
Visualizing the spatial gene expression organization in the brain through non-linear similarity embeddings.
Methods
73
:
79
89
.
27.
Huang
D.W.
,
B. T.
Sherman
,
R. A.
Lempicki
.
2009
.
Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources.
Nat. Protoc.
4
:
44
57
.
28.
Street
K.
,
D.
Risso
,
R. B.
Fletcher
,
D.
Das
,
J.
Ngai
,
N.
Yosef
,
E.
Purdom
,
S.
Dudoit
.
2018
.
Slingshot: cell lineage and pseudotime inference for single-cell transcriptomics.
BMC Genomics
19
:
477
.
29.
R Core Team
.
2018
.
R: A Language and Environment for Statistical Computing.
R Foundation for Statistical Computing
,
Vienna, Austria
.
30.
Stuart
T.
,
A.
Butler
,
P.
Hoffman
,
C.
Hafemeister
,
E.
Papalexi
,
W. M.
Mauck
3rd
,
Y.
Hao
,
M.
Stoeckius
,
P.
Smibert
,
R.
Satija
.
2019
.
Comprehensive integration of single-cell data.
Cell
177
:
1888
1902.e21
.
31.
Van Grembergen
O.
,
M.
Bizet
,
E. J.
de Bony
,
E.
Calonne
,
P.
Putmans
,
S.
Brohée
,
C.
Olsen
,
M.
Guo
,
G.
Bontempi
,
C.
Sotiriou
, et al
2016
.
Portraying breast cancers with long noncoding RNAs.
Sci. Adv.
2
:
e1600220
.
32.
Zhou
J.
,
X.
Zhi
,
L.
Wang
,
W.
Wang
,
Z.
Li
,
J.
Tang
,
J.
Wang
,
Q.
Zhang
,
Z.
Xu
.
2015
.
Linc00152 promotes proliferation in gastric cancer through the EGFR-dependent pathway. [Published erratum appears in 2016 J. Exp. Clin. Cancer Res. 35: 30.]
J. Exp. Clin. Cancer Res.
34
:
135
.
33.
Hayakawa
T.
,
K.
Yamashita
,
K.
Tanzawa
,
E.
Uchijima
,
K.
Iwata
.
1992
.
Growth-promoting activity of tissue inhibitor of metalloproteinases-1 (TIMP-1) for a wide range of cells. A possible new growth factor in serum.
FEBS Lett.
298
:
29
32
.
34.
Rogers
P. R.
,
J.
Song
,
I.
Gramaglia
,
N.
Killeen
,
M.
Croft
.
2001
.
OX40 promotes Bcl-xL and Bcl-2 expression and is essential for long-term survival of CD4 T cells.
Immunity
15
:
445
455
.
35.
Hashimoto
K.
,
T.
Kouno
,
T.
Ikawa
,
N.
Hayatsu
,
Y.
Miyajima
,
H.
Yabukami
,
T.
Terooatea
,
T.
Sasaki
,
T.
Suzuki
,
M.
Valentine
, et al
2019
.
Single-cell transcriptomics reveals expansion of cytotoxic CD4 T cells in supercentenarians.
Proc. Natl. Acad. Sci. USA
116
:
24242
24251
.
36.
Cano-Gamez
E.
,
B.
Soskic
,
T. I.
Roumeliotis
,
E.
So
,
D. J.
Smyth
,
M.
Baldrighi
,
D.
Willé
,
N.
Nakic
,
J.
Esparza-Gordillo
,
C. G. C.
Larminie
, et al
2020
.
Single-cell transcriptomics identifies an effectorness gradient shaping the response of CD4+ T cells to cytokines.
Nat. Commun.
11
:
1801
.
37.
Urata
S.
,
E.
Kenyon
,
D.
Nayak
,
B.
Cubitt
,
Y.
Kurosaki
,
J.
Yasuda
,
J. C.
de la Torre
,
D. B.
McGavern
.
2018
.
BST-2 controls T cell proliferation and exhaustion by shaping the early distribution of a persistent viral infection.
PLoS Pathog.
14
:
e1007172
.
38.
Oriol-Tordera
B.
,
M.
Berdasco
,
A.
Llano
,
B.
Mothe
,
C.
Gálvez
,
J.
Martinez-Picado
,
J.
Carrillo
,
J.
Blanco
,
C.
Duran-Castells
,
C.
Ganoza
, et al
2020
.
Methylation regulation of Antiviral host factors, Interferon Stimulated Genes (ISGs) and T-cell responses associated with natural HIV control.
PLoS Pathog.
16
:
e1008678
.
39.
Kisand
K.
,
M.
Link
,
A. S.
Wolff
,
A.
Meager
,
L.
Tserel
,
T.
Org
,
A.
Murumägi
,
R.
Uibo
,
N.
Willcox
,
K.
Trebusak Podkrajsek
, et al
2008
.
Interferon autoantibodies associated with AIRE deficiency decrease the expression of IFN-stimulated genes.
Blood
112
:
2657
2666
.
40.
Jameson
S. C.
,
D.
Masopust
.
2018
.
Understanding subset diversity in T cell memory.
Immunity
48
:
214
226
.
41.
Hoyer
S.
,
S.
Prommersberger
,
I. A.
Pfeiffer
,
B.
Schuler-Thurner
,
G.
Schuler
,
J.
Dörrie
,
N.
Schaft
.
2014
.
Concurrent interaction of DCs with CD4(+) and CD8(+) T cells improves secondary CTL expansion: It takes three to tango.
Eur. J. Immunol.
44
:
3543
3559
.

The authors have no financial conflicts of interest.

This article is distributed under The American Association of Immunologists, Inc., Reuse Terms and Conditions for Author Choice articles.

Supplementary data