Donor peripheral blood mononuclear cells (PBMCs) were collected after written, informed consent from the University of Pennsylvania’s Center for AIDS Research Human Immunology Core (IRB #705906) in compliance with IRB guidelines. PBMCs were cryopreserved in fetal bovine serum (FBS; Hyclone) containing 10% dimethyl sulfoxide (DMSO; Fisher Scientific) and stored at −140°C until further use.

Flow cytometry analysis was performed as previously described (10) using PBMCs from at least eight normal donors. Where appropriate, statistical analyses were performed using GraphPad Prism software (Version 5.0a). For these studies, non-parametric Wilcoxon matched paired t tests were used where Gaussian distribution is not assumed because we analyzed <25 subjects.

To identify CD4⁺, CD8⁺, and T-regulatory (T_reg) T-cells, the following antibodies were used: α–CD3-BV570 (Biolegend), α–CD4-PE Cy5.5 (Invitrogen), α–CD8-BV605 (Biolegend), α–CD14/α–CD16/α–CD19-APC Cy7 (BD Bioscience), α–CCR7-BV711 (Biolegend), α–CD45RO-PE Texas Red (Beckman Coulter), α–CD27-FITC (eBioscience), α–CD25-PE Cy5 (Invitrogen), α–CD127-PE Cy7 (eBioscience), α–T-bet-PE (eBioscience), α–Eomes-Alexa647 (eBioscience), and α–Foxp3 Alexa700 (eBioscience).

To identify natural killer (NK), invariant natural killer (iNKT), and γδ T-cells, the following antibodies were used: α–CD3-BV570 (Biolegend), α–CD4-PE Cy5.5 (Invitrogen), α–CD8-BV605 (Biolegend), α–CD16-PE Cy5 (Biolegend), α–CD56-Alexa700 (Biolegend), α–CD19-BV785 (Biolegend), α–CD45RO-PE Texas Red (Beckman Coulter), α–γδ TCR-FITC (Biolegend), α–Vα24Jα18-PE Cy7 (Biolegend), α–T-bet-PE (eBioscience), and α–Eomes-eF660 (eBioscience).

To identify B-cell subsets, the following antibodies were used: α–CD3-BV570 (Biolegend), α–CD14/α–CD16-APC Cy7 (BD Bioscience), α–CD19-BV785 (Biolegend), α–CD20-PE Texas Red (BD Bioscience), α–IgD-Alexa700 (BD Bioscience), α–IgM-BV605 (Biolegend), α–CD38-BV421 (Biolegend), α–CD10-PE Cy5 (BD Bioscience), α–CD21-PE Cy7 (Biolegend), α–CD27-BV650 (Biolegend), α–IgG1-Alexa488 (Invitrogen), α–T-bet-PE (eBioscience), and α–Eomes-eF660 (eBioscience).

In addition to previously mentioned antibodies, the following were used to identify dendritic cell subsets: α–CD14-QD655 (Invitrogen), α–CD11c-FITC (BD Bioscience), α–HLADR-V450 (BC Bioscience), α–CD56-PE Texas Red (Invitrogen), and α–CD123 PE Cy5 (Biolegend).

T-bet and Eomes have been extensively studied in murine CD8⁺ T-cells and are critical for effector function and long-term memory formation, respectively (13, 20, 23–25). Recently, we described T-bet and/or Eomes in human CD8⁺ T-cell memory populations using various combinations of the memory markers CD27, CCR7, and CD45RO (9, 10, 26). While a fair amount is known about the relationship between T-bet and Eomes and the individual memory markers, the breakdown of their combined expression is not well defined.

Based upon coordinate expression of CD27, CD45RO, and CCR7, we delineated peripheral blood CD8⁺ T-cells into six different populations, including naïve (CCR7⁺CD45RO⁻CD27⁺), central memory (T_CM, CCR7⁺CD45RO⁺CD27⁺), transitional memory (CCR7⁻CD45RO⁺CD27⁺), effector memory (T_EM, CCR7⁻CD45RO⁺CD27⁻), intermediate (CCR7⁻CD45RO⁻CD27⁺), and effector (CCR7⁻CD45RO⁻CD27⁻) CD8⁺ T-cells using a Boolean gating strategy (Figure 1A; Figure S1 in Supplementary Material). For simplicity, we focused on naïve, central memory, effector memory, and effector populations in Figure 1 (refer to Figures S1A–D in Supplementary Material for the more detailed breakdown of these populations). Representative gating for T-bet and Eomes expression within effector CD8⁺ T-cells from a normal human donor is shown in Figure 1A. As has been previously reported (10, 23, 26), T-bet has a graded expression pattern including distinct T-bet^hi and T-bet^lo populations (Figure 1A, right panel). Approximately 60% of total CD8 T-cells expressed T-bet (Figure 1B). While less than 20% of naïve CD8⁺ T-cells were T-bet⁺, significantly more T_CM, T_EM, and effector T-cells expressed T-bet (Figure 1B, data not shown). As shown previously (10, 26), the majority of T-bet⁺ cells within memory CD8⁺ T-cell populations were T-bet^lo; however, as cells progress toward a more terminally differentiated phenotype, the frequency of T-bet^hi cells significantly increased (Figure 1C). Taken together, these data suggest that high levels of T-bet are likely crucial for function in effector and some effector memory CD8⁺ T-cells, whereas T-bet may play a less definitive role in naïve and T_CM differentiation and function.

We next investigated Eomes expression within memory CD8⁺ T-cell memory subsets. Less than half of total CD8⁺ T-cells expressed Eomes (Figure 1D). Like T-bet, Eomes is expressed most infrequently in naïve CD8⁺ T-cells. While the frequency of Eomes⁺ T_CM, T_EM, and effector CD8⁺ T-cells increased significantly compared to naïve T-cells (data not shown), we found no significant difference in the frequency of T_CM T-bet⁺ cells compared to T_EM or effector cells. Additionally, there is more Eomes per cell in memory CD8⁺ T-cells compared to naïve T-cells (when it is expressed) as measured by median fluorescence intensity (MFI) analysis (Figure 1E). Taken together, these data indicate that Eomes is expressed in the highest frequency in effector cells but, on average, these cells do not express more Eomes per cell compared to other CD8⁺ T-cell memory subsets.

Because T-bet and Eomes likely have both redundant and unique roles in CD8⁺ T-cells (20, 27), we next investigated the co-expression of these factors in the CD8⁺ T-cell memory populations (Figure 1F). Naïve cells were almost exclusively T-bet⁻Eomes⁻ (Figure 1F, black bars). T_CM CD8⁺ T-cells were mainly T-bet^lo Eomes^+/−, or T-bet⁻ Eomes⁻ (dark gray bars). Like T_CM, T_EM T-cells were mainly T-bet^lo Eomes^+/−; however, T_EM had a significantly higher frequency of T-bet^hi Eomes⁺ compared to T_CM. Further, effector CD8⁺ T-cells had the highest frequency of T-bet^hi Eomes⁺ CD8⁺ T-cells. Interestingly, T-bet^hi T_EM and effector CD8⁺ T-cells were almost exclusively Eomes⁺ suggesting high levels of T-bet correlate with Eomes expression and these factors could be cooperating to promote critical functions in these CD8⁺ T-cell subpopulations.

Initial work in mouse CD4⁺ T-cells first defined T-bet as the master regulator of T_H1 development, with direct regulation of T_H1-specific genes such as IFNγ (4). Early reports demonstrated expression of T-bet also within human CD4⁺ T-cells, but the patterns of T-bet expression within CD4⁺ T-cell memory subsets have remained unclear. We therefore sought to characterize T-bet expression in human CD4⁺ memory T-cell subsets. As with CD8⁺ T-cells, memory subpopulations were defined using the markers CD27, CD45RO, and CCR7 (Figure 2; Figures S1E–H in Supplementary Material). Representative flow plots for effector CD4⁺ T-cells are shown in Figure 2A. Overall, within our cohort, ~25% of total CD4⁺ T-cells expressed T-bet (Figure 2B). Few naïve human CD4⁺ T-cells were T-bet⁺, with a significant majority of T-bet expression found within the non-naïve CD4⁺ T-cell population (Figure 2B, data not shown). Approximately 20% of T_CM CD4⁺ T-cells expressed T-bet and that T-bet is almost exclusively T-bet^lo (Figures 2B,C). Significantly more T_EM and effector CD4⁺ T-cells expressed T-bet compared to T_CM, however there was no difference between T_EM and effector CD4⁺ T-cells. While T_EM cells had the highest frequency of T-bet^lo cells, the frequency of T-bet^hi CD4⁺ T-cells increased significantly as cells progress toward a more effector-like phenotype suggesting that T-bet is likely more critical to the functions of CD4⁺ T_EM and effector cells than T_CM function.

Like T-bet, Eomes also plays a role in murine CD4⁺ T-cell differentiation. Eomes can induce both T_H1 differentiation as well as expression of IFNγ and perforin in mouse CD4⁺ T-cells (13, 20, 21, 28). Eomes can also compensate for the loss of T-bet in CD4⁺ effector T-cells and drive polyfunctionality in human CD4⁺ T-cells (29); however, there are few ex vivo studies addressing Eomes expression in human CD4⁺ T-cells. In contrast to T-bet, very few human peripheral blood CD4⁺ T-cells express Eomes (Figure 2D). As cells become more effector-like, the frequency of Eomes⁺ cells significantly increases however the frequency only reaches ~25% in effector cells. Compared to naïve cells, we found a significantly higher Eomes MFI within each Eomes⁺ memory CD4⁺ T-cell subpopulation; however, while there was a trend toward higher Eomes MFI as cells become more effector-like, these differences were not significant (Figure 2E).

Because both T-bet and Eomes are known to induce IFNγ expression and multiple cytolytic functions in CD4⁺ T-cells (13, 20, 21, 30), we next examined co-expression of these factors within CD4⁺ T-cell subsets. We found significant increases in the frequency of all Eomes⁺ populations regardless of T-bet levels as cells progress to being more effector-like (Figure 2F). Like CD8⁺ T-cells, there were very few T-bet^hi cells that did not express Eomes, suggesting that high levels of these factors could cooperate to drive T_EM and effector differentiation and function; however, in contrast to murine studies, the majority of resting CD4⁺ T-cells were either T-bet⁻ Eomes⁻ or T-bet^lo Eomes⁻ suggesting that, at least in the context of resting peripheral blood CD4⁺ T-cells, T-bet, and Eomes may not contribute to resting CD4⁺ T-cell function.

T-regulatory cells function to suppress immune responses from other cell types to prevent hyperactivity or autoimmune disease. T_reg cells have been reported to upregulate T-bet in vivo during type-1 inflammatory responses (31); however, little is known about T-bet in circulating human T_reg cells. We next characterized T-bet and Eomes expression in resting CD8⁺ and CD4⁺ T_reg cells. We identified T_reg cells within both CD8⁺ and CD4⁺ populations using the markers CD25 and FoxP3 (Figure 3A). Only a small fraction of CD8⁺ T-cells were CD25⁺ Foxp3⁺, whereas CD4⁺ CD25⁺ Foxp3⁺ cells comprise about 1.5% of total PBMCs (Figure 3B). Within each of these T_reg populations, ~8% of cells expressed T-bet (Figure 3C) and this T-bet was almost exclusively T-bet^lo (Figure 3D). Eomes expression was not detected in circulating T_reg cells (data not shown). Taken together, these data indicate that neither T-bet nor Eomes likely contribute to resting human T_reg function.

We next investigated the relationship between long-term memory formation and expression of T-bet and Eomes in CD8⁺ and CD4⁺ T-cells through examination of IL-7 receptor (CD127) expression. T-cells are dependent upon IL-7 signaling for survival (32–36). In mice, naïve T-cells express CD127 and following T-cell receptor stimulation, CD127 is downregulated (32, 33, 37–39). Recent studies have suggested that the upregulation of T-bet in CD8⁺ T-cells results in the downregulation of CD127 (14, 23); however, a study in Leishmania-specific CD4⁺ T-cells suggest that the expression of T-bet did not inhibit CD127 expression, nor did the loss of T-bet result in upregulation of CD127 (40). Based on these results, we examined the relationship between T-bet and Eomes expression with CD127 in the context of human CD8⁺ and CD4⁺ T-cells. Correlating with previously published data (41, 42), CD127 was expressed in upwards of 85% of both CD8⁺ and CD4⁺ T-cells and the frequency of cells expressing CD127 decreased as cells became more effector-like (data not shown). Representative flow plots displaying the relationship between T-bet or Eomes and CD127 from one normal donor are shown in Figure 4A. In CD4⁺ T-cells, CD127 was expressed regardless of the presence or absence of T-bet or Eomes (Figure 4B, white bars). In contrast, as T-bet or Eomes expression increased in memory CD8⁺ T-cells, the frequency of CD127⁺ cells significantly decreased compared to CD4⁺ T-cells (Figure 4B, black bars). These results suggest that T-bet and Eomes may play a different role in controlling CD127 expression in CD8⁺ T-cells compared to CD4⁺ T-cells.

γδ T-cells and invariant natural killer T-cells (iNKT cells) are innate-like T-cell family members expressing T-cell receptors with restricted antigen recognition potential compared to classical αβ T-cells (43, 44). T-bet mRNA and protein, as well as Eomes mRNA, are detectable in mouse γδ T-cells, where it is suggested that they cooperate to control IFNγ production (45, 46). In mouse iNKT cells, T-bet is required for iNKT developmental progression and the acquisition of effector functions (21, 47). Expression of these transcription factors has not been comprehensively demonstrated in human γδ T and iNKT cells; therefore, we sought to characterize T-bet and Eomes expression in human γδ T and iNKT cells. We defined γδ T-cells as γδ TCR⁺ within CD3⁺CD4⁻CD8⁻PBMCs (Figure 5A); iNKT cells were identified within CD3⁺ PBMCs as TCR Vα24Jα18⁺ (Figure 5B). We further subdivided iNKT cells into CD4⁻CD8⁻ and CD4⁺ subgroups, the best described subpopulations of human iNKT cells (43). While we could detect CD4⁻CD8⁺ iNKT cells, they were infrequent and only detectable in 4/8 donors (data not shown).

As found in mice (21, 45–47), we observed T-bet expression in both resting γδ T and the major subsets of iNKT cells (Figure 5C). Approximately 60% of γδ T-cells and 50% of iNKT cells expressed T-bet, although we found considerable variation between donors within the iNKT population (Figures 5C,D, white bars). After subdividing iNKT cells into CD4⁻CD8⁻ and CD4⁺ subsets, we found that the frequency of T-bet-expressing cells was significantly higher in the CD4⁻CD8⁻ population compared to the CD4⁺ subset (Figure 5C). Additionally, significantly more CD4⁻CD8⁻ iNKT cells were T-bet^hi compared to the CD4⁺ population (Figure 5D, gray bars). These findings indicate that CD4⁻CD8⁻ iNKT cells generally express T-bet, while CD4⁺ iNKT cells express T-bet at lower, more variable levels, suggesting that T-bet plays a particularly important role in the function of CD4⁻CD8⁻ iNKT cells.

Eomesodermin was detectable in ~40% of γδ T-cells. In contrast to previous murine studies that were unable to detect Eomes mRNA (21), we found that ~30% of human iNKT cells expressed Eomes protein, although there was significant variation between subjects (Figure 5E). CD4⁻CD8⁻ iNKT cells account for the majority of Eomes expression within the total iNKT-cell population, as a significantly higher frequency of CD4⁻CD8⁻ cells expressed Eomes compared to CD4⁺ iNKT cells (Figure 5E). Despite differences in the frequency of Eomes⁺ cells between iNKT subsets, both subgroups expressed similar levels of Eomes on a per cell basis (Figure 5F). Taken together, these data indicate that Eomes is differentially expressed in human iNKT cells compared to murine iNKT cells and suggest a role for Eomes in the context of human iNKT cells.

In both total γδ T-cells and total iNKT cells, the majority of Eomes⁺ cells co-expressed T-bet, whereas Eomes⁻ cells were either T-bet^lo or T-bet⁻ (Figure 5G). Greater than 60% of CD4⁺ iNKT cells did not express T-bet or Eomes, and the remainder were T-bet^lo Eomes⁻. CD4⁻CD8⁻ iNKT cells contained substantial T-bet^hi Eomes⁺ and T-bet^lo Eomes⁺ populations, both of which occurred at a significantly higher frequency than in CD4⁺ iNKT cells. These findings indicate that T-bet and Eomes are highly co-expressed in the CD4⁻CD8⁻, but not in CD4⁺ iNKT cells, suggesting T-bet and Eomes could cooperatively function in CD4⁻CD8⁻ NKT cells.

In mice, T-bet and Eomes modulate many NK cell effector functions, including cytotoxicity and cytokine production (21, 48). Additionally, their expression is crucial for murine NK developmental regulation where they cooperate to influence several key developmental checkpoints (49). T-bet and Eomes have been highly studied in mouse models, but there are few studies investigating the expression patterns of T-bet and Eomes within human NK cell populations; therefore, we next assessed T-bet and Eomes in human NK subsets. We identified two mature NK cell populations within CD14⁻CD19⁻CD3⁻PBMCs based upon CD56 and CD16 expression (Figure 6A), here referred to as CD56^bright (CD56^hi CD16⁻) and CD56^dim (CD56^lo CD16⁺) cells. We observed a gradient of T-bet expression with the CD56^bright population (Figure 6A, right panel). While virtually all mature NK cells expressed T-bet, both the frequency of T-bet⁺ cells and the amount of T-bet per cell was significantly greater in the CD56^dim population compared to the CD56^bright population (Figures 6B,C, gray bars). Conversely, the T-bet^lo population was significantly larger in CD56^bright cells (white bars). Taken together, these data suggest that there may be an association between T-bet expression levels and functional capacity in NK cells: CD56^dim NK cells highly express T-bet and are highly cytotoxic, while poorly cytotoxic CD56^bright NK cells express less T-bet and function mainly to produce cytokines (50).

While Eomes was expressed in both CD56^bright and CD56^dim NK cells, a significantly higher frequency of CD56^bright cells were Eomes⁺ compared to CD56^dim cells (Figure 6D). Additionally, Eomes⁺ CD56^bright cells expressed significantly more Eomes on a per cell basis than Eomes⁺ CD56^dim cells (Figure 6E), suggesting that Eomes is likely crucial for CD56^bright function.

We next investigated the co-expression of T-bet and Eomes within the CD56^dim and CD56^bright NK populations. The majority of both NK cell subpopulations were T-bet^hi Eomes⁺ (Figure 6F); however, in the remaining cells we found unique co-expression patterns of these transcription factors between the NK subsets. Approximately 15% of CD56^bright NK cells were T-bet^lo Eomes⁺, and this subpopulation was virtually non-existent in the CD56^dim cells. Conversely, ~35% of CD56^dim NK cells were T-bet^hi Eomes⁻, compared to CD56^bright cells which did not have this population. Taken together, these results suggest that while T-bet and Eomes likely play complementary or cooperative roles in the majority of NK cells, there may be distinct subpopulations of NK cells where T-bet and Eomes differentially regulate NK cell function.

Murine studies have revealed that T-bet is expressed in lymphoid tissue B-cells, where it regulates functional processes such as class switching (51, 52) and homing (53). A recent study of T-bet in B-cells revealed that T-bet is initially expressed during the primary immune response and expression is maintained in IgG2a⁺ memory B-cells, where it is necessary for cell survival and secondary immune responses (54). Additionally, there is evidence to suggest that expression of T-bet in B-cells is necessary for clearance of gHV68, a murine herpes virus (55). While it is appreciated that T-bet is necessary for appropriate B-cell function and antibody responses in mice, T-bet expression is not well-characterized in human B-cells.

To identify B-cell subpopulations, CD19⁺ PBMCs were phenotyped using several B-cell markers: IgD, CD10, CD38, and CD27 (Figure S2 in Supplementary Material). Our analysis focused specifically on transitional/immature B-cells (IgD⁺CD10⁺ CD38⁺CD27⁻), naïve B-cells (IgD⁺CD10⁻CD38^+/−CD27⁻), memory B-cells (IgD⁻CD10⁻CD38^lo/−), and plasmablasts (IgD⁻CD10⁻CD38^hiCD27⁺). Representative flow plots of T-bet expression (contour plot) within each subpopulation are shown in Figure 7A. While Eomes was undetectable in B-cells (data not shown), we found T-bet in ~15% of B-cells (Figure 7B). This T-bet expression was largely relegated to memory B-cells and plasmablasts, with significantly lower amounts observed in transitional/immature and naïve B-cells (Figure 7B). Greater than 40% of plasmablasts expressed T-bet, a significantly higher frequency than that of all other B-cell populations, suggesting that T-bet may play a particularly important role in plasmablast function.

We next compared T-bet expression levels within different B-cell populations to other known T-bet-expressing cell types. Plasmablasts expressed the highest amount of T-bet within the B-cell subsets (Figure 7C, shaded gray), while naïve B-cells displayed the lowest T-bet levels (thick black line). Memory B-cells (thin black line) generally expressed intermediate levels of T-bet; however, a small fraction expressed higher T-bet levels compared to plasmablasts. T-bet expression in T-bet⁺ B-cells was relatively low compared to other T-bet⁺ cells, including CD56^dim NK cells (shaded black), representing some of the brightest T-bet-expressing cells, and CD56^bright NK cells (thin gray line), which express lower levels of T-bet compared to CD56^dim cells. Taken together, these data suggest a key role for T-bet in plasmablasts and memory B-cell subsets and further indicate that T-bet is expressed at a lower level in B-cells compared to other T-bet⁺ lymphocytes.

Previous studies have shown that several non-lymphocyte populations, including myeloid dendritic cells (mDCs) and plasmacytoid dendritic cells (pDCs), can express T-bet transcript following IFNγ stimulation (56–58); however, it is unclear if human resting cells can also express T-bet. Additionally, Eomes has not been investigated in the context of these populations in humans. To characterize the expression of T-bet and Eomes in resting human dendritic cell populations, we defined mDCs as CD3⁻CD14⁻CD19⁻CD11c⁺ and pDCs as CD3⁻CD14⁻CD19⁻CD123⁺CD11c⁻. Neither T-bet nor Eomes protein was detectable in circulating dendritic cell populations (data not shown), suggesting that T-bet may be upregulated only under specific conditions.