Heterogeneous yet stable Vδ2⁽⁺⁾ T-cell profiles define distinct cytotoxic effector potentials in healthy human individuals

Significance

A type of human white blood cell, known as the Vδ2⁽⁺⁾ T cell, has shown promise in immunotherapies against a range of tumors. However, in recent clinical trials patient responses and clinical outcomes have been variable and unpredictable. To address this, we here reveal a significant variability in Vδ2⁽⁺⁾ T-cell functional potential between individuals in the general population, which develops shortly after birth, is stable over time, and is manifested by differential mechanistic capacities to kill tumor targets. These results support personalized clinical approaches to identify patients with “Vδ2 profiles” that are compatible with killing of their particular tumor and suggest that tailored Vδ2-profile–specific activation protocols may maximize the chances of future treatment success.

Abstract

Human γδ T cells display potent responses to pathogens and malignancies. Of particular interest are those expressing a γδ T-cell receptor (TCR) incorporating TCRδ-chain variable-region-2 [Vδ2⁽⁺⁾], which are activated by pathogen-derived phosphoantigens (pAgs), or host-derived pAgs that accumulate in transformed cells or in cells exposed to aminobisphosphonates. Once activated, Vδ2⁽⁺⁾ T cells exhibit multiple effector functions that have made them attractive candidates for immunotherapy. Despite this, clinical trials have reported mixed patient responses, highlighting a need for better understanding of Vδ2⁽⁺⁾ T-cell biology. Here, we reveal previously unappreciated functional heterogeneity between the Vδ2⁽⁺⁾ T-cell compartments of 63 healthy individuals. In this cohort, we identify distinct “Vδ2 profiles” that are stable over time; that do not correlate with age, gender, or history of phosphoantigen activation; and that develop after leaving the thymus. Multiple analyses suggest these Vδ2 profiles consist of variable proportions of two dominant but contrasting Vδ2⁽⁺⁾ T-cell subsets that have divergent transcriptional programs and that display mechanistically distinct cytotoxic potentials. Importantly, an individual’s Vδ2 profile predicts defined effector capacities, demonstrated by contrasting mechanisms and efficiencies of killing of a range of tumor cell lines. In short, these data support patient stratification to identify individuals with Vδ2 profiles that have effector mechanisms compatible with tumor killing and suggest that tailored Vδ2-profile–specific activation protocols may maximize the chances of future treatment success.

Human γδ T cells display potent responses to pathogens and malignancies (1–3). Of particular interest are those that express TCRδ-chain variable-region-2 [Vδ2⁽⁺⁾] (4). These cells are uniquely activated by low-molecular-weight nonpeptide phosphoantigens (pAgs), such as microbial-derived (E)-4-hydroxy-3-methyl-but-2-enyl pyrophosphate (HMB-PP) (5, 6). Vδ2⁽⁺⁾ T cells are also activated by isopentyl pyrophosphate, a phosphoantigen (pAg) that accumulates in transformed eukaryotic cells or cells exposed to aminobisphosphonates such as zoledronate (7). The mechanism of pAg-mediated activation of Vδ2⁽⁺⁾ T cells is still unclear, but likely involves pAg association with butyrophilin 3A1 (BTN3A1) (8, 9).

Subsequent to pAg stimulation, Vδ2⁽⁺⁾ T cells can display significant cytokine-dependent functional plasticity. Indeed, Th1-like, Th2-like, Th17-like, and Treg-like effector characteristics have all been reported (10–14). Moreover, Vδ2⁽⁺⁾ T cells demonstrate efficient cytolytic capacity (15–17) and function as antigen-presenting cells if activated under appropriate conditions (18). These functional potentials underpin the role of Vδ2⁽⁺⁾ T cells in diverse immune scenarios. Vδ2⁽⁺⁾ T cells respond vigorously to Mycobacterium tuberculosis (19) and Plasmodium falciparum (20), often expanding to >50% of total blood T cells (21), and show responses to both HIV (22) and influenza (23). Vδ2⁽⁺⁾ T cells also kill a spectrum of malignant cells that includes leukemias and lymphomas and solid tumors such as renal cell, breast, prostate, and colorectal carcinomas (24). Indeed, meta-analysis of gene expression signatures from ∼18,000 human tumors across 39 malignancies indicated a tumor-associated γδ T-cell profile as the best predicator of patient survival (25). Thus, there appears enormous potential to harness these antipathogen and antitumor effector functions for clinical applications.

Despite this therapeutic promise, results from phase I/II clinical trials that have activated Vδ2⁽⁺⁾ T cells with aminobisphosphonates are mixed. Although objective clinical outcomes were observed in some patients with relapsed/refractory low-grade non-Hodgkin’s lymphoma, multiple myeloma, metastatic hormone-refractory prostate cancer, or advanced metastatic breast cancer (26–28), numerous patients failed to demonstrate effective Vδ2⁽⁺⁾ T-cell responses. Clearly, understanding this person-to-person heterogeneity in Vδ2⁽⁺⁾ T-cell responsiveness, correlated with subsequent clinical outcome, is critical not only for optimization of Vδ2⁽⁺⁾ T-cell–related therapies, but also for predicting disease progression where Vδ2⁽⁺⁾ T-cell responses are involved.

In this study, we reveal functional Vδ2⁽⁺⁾ T-cell heterogeneity between individuals in a large cohort of healthy volunteers. The effector potentials of these “Vδ2 profiles” are characterized by two dominant but qualitatively contrasting phenotypes. At one extreme, Vδ2⁽⁺⁾ T cells display high proliferative capacity, express several cytokine and chemokine receptors, and demonstrate unusual granzyme K-mediated target cell killing. At the other extreme, Vδ2⁽⁺⁾ T cells have lower expansion potential but possess a dominantly cytotoxic nature characterized by granzyme B-mediated cytotoxicity. This interindividual Vδ2⁽⁺⁾ T-cell heterogeneity develops after birth although acquisition of a particular Vδ2 profile does not correlate with gender, age, country of birth, or chronic Vδ2⁽⁺⁾ T-cell stimulation in vivo. Moreover, these Vδ2 profiles were stable in individuals over the 3-y study period, suggesting an active homeostatic maintenance. Importantly, an individual’s Vδ2 profile predicts functional potential that we demonstrate by differential killing of various tumor cell lines. Thus, these data highlight a phenotypic and functional heterogeneity in the human Vδ2⁽⁺⁾ T-cell pool that has profound clinical implications such that individuals with different Vδ2 profiles would be predicted to respond differently to Vδ2⁽⁺⁾ T-cell–targeted immunotherapies or in response to infections.

Results

Significant Functional Heterogeneity in Vδ2⁽⁺⁾ T-Cell Subsets Between Healthy Individuals.

We had regularly observed phenotypic heterogeneity when using the commonly used markers CD27 and CD45RA to assess human Vδ2⁽⁺⁾ T cells from small numbers of healthy volunteers (Fig. 1A). As this compromised our interpretation of Vδ2⁽⁺⁾ T-cell involvement in disease, the nature and extent of this heterogeneity was investigated in a much larger cohort of healthy individuals (n = 63). In our hands, CD45RA staining of Vδ2⁽⁺⁾ T cells (but not other T-cell subsets) does not give distinct demarcation of positive and negative subsets (Fig. 1A). Thus, we instead assessed Vδ2⁽⁺⁾ T cells using CD27, CD28, and CD16 that consistently identified four distinct Vδ2⁽⁺⁾ populations: γδ⁽²⁸⁺⁾ cells [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ cells [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ cells [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ cells [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾] (Fig. 1B). The γδ⁽²⁸⁺⁾ population was most prevalent across the cohort (Fig. 1C) comprising on average ∼54% of total Vδ2⁽⁺⁾ T cells, followed by the γδ⁽²⁸⁻⁾ subset (∼22%), γδ⁽¹⁶⁺⁾ subset (∼11%), and γδ⁽¹⁶⁻⁾ subset (∼8%). However, dominance of a particular subset did not correlate with age, gender, or developmental index of country of birth (Fig. S1). The four Vδ2⁽⁺⁾ T-cell subsets were functionally assessed after stimulation with bisphosphonate (zoledronate) and IL-2. Proliferative potential segregated with CD27 expression, as the γδ⁽²⁸⁺⁾ and γδ⁽²⁸⁻⁾ subsets, but less so the γδ⁽¹⁶⁻⁾ and γδ⁽¹⁶⁺⁾ subsets, divided multiply after activation (Fig. 1D). Consistent with this, γδ⁽¹⁶⁺⁾ cells expressed the highest level of CD57 that has been reported to correlate with replicative senescence (29) although other markers of exhaustion, such as PD-1, were similar across all populations (Fig. S2). All subsets had potential to secrete both IFNγ and TNFα (Fig. S3). However, cytolytic capacity, assessed by intracellular staining for perforin and granzyme B (Fig. 1E), and surface staining of CD56 (Fig. S4), was greater in the γδ⁽¹⁶⁻⁾ and γδ⁽¹⁶⁺⁾ subsets. Finally, the γδ⁽²⁸⁺⁾ subset demonstrated a reduced capacity to degranulate, judged by surface CD107a expression after 8 h of activation (Fig. 1F). Thus, CD27, CD28, and CD16 unambiguously identify four subsets of Vδ2⁽⁺⁾ T cells in normal healthy individuals. Importantly, as these subsets have distinct functional capacities, the overall effector potential of the Vδ2⁽⁺⁾ T-cell compartment will reflect the relative contributions of these subsets in any given individual.

Fig. 1.

Interindividual Vδ2⁽⁺⁾ T-cell phenotypic variation. (A) CD27/CD45RA plots of blood Vδ2⁽⁺⁾ cells from five healthy individuals. Quadrant percentages are indicated. (B) Nonnaive CCR7⁽⁻⁾ Vδ2⁽⁺⁾ cells comprise four subsets: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾]. (C) Vδ2⁽⁺⁾ subset distribution for 63 healthy individuals (mean ± SD). (D) Representative plots of eFluor670 dilution in sorted Vδ2⁽⁺⁾ subsets after culture with Vδ2-depleted mononuclear accessory cells (1:1) 7 d after stimulation with zoledronate (1 μM) + IL-2 (100 U/mL). P, proliferating cells; UD, undivided cells. (E) Intracellular Perforin and granzyme B in Vδ2⁽⁺⁾ subsets after 4 h stimulation of PBMC ex vivo with phorbol 12-myristate 13-acetate (PMA)/ionomycin. Summary chart for percentage of Vδ2⁽⁺⁾ subset producing. Pfn⁽⁺⁾, perforin; GzmB⁽⁺⁾, granzyme B; both, Pfn⁽⁺⁾GzmB⁽⁺⁾ (n = 3). (F) Mean fluorescence intensity (MFI) of CD107a on Vδ2⁽⁺⁾ subsets 8 h after activation of PBMCs with HMB-PP (20 nM) + IL-2 (100 U/mL). Multiple comparison testing using one-way ANOVA with Tukey’s post hoc test used in E and F. *P < 0.05, **P < 0.01, and ***P < 0.001.

Fig. S1.

Distribution of Vδ2⁽⁺⁾ T-cell subsets in peripheral blood is unaffected by age, gender, or country of birth. Vδ2⁽⁺⁾ T-cell subsets expressed as a percentage of total Vδ2⁽⁺⁾ T cells according to (A) age (years), (B) gender (male/female), and (C) Human Development Index of country of birth (HDI). HDI is a United Nations summary statistic accounting for life expectancy, income, and education of a country in terms of human development. Vδ2⁽⁺⁾ T-cell subsets are defined as the following: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾]. Multiple comparison testing using one-way ANOVA with Tukey’s post hoc test used in A–C. ns, not significant.

Fig. S2.

CD57 and PD-1 expression on Vδ2⁽⁺⁾ T-cell subsets. (A) Representative plots from peripheral blood of a healthy individual showing CD57 expression in Vδ2⁽⁺⁾ subsets. Percentages of gated cells are indicated. Summary graph (n = 4) shows mean percentage of CD57⁽⁺⁾ cells within each indicated Vδ2⁽⁺⁾ T-cell subset. Error bars are SD. (B) Representative plots from peripheral blood of a healthy individual showing PD-1 expression in a Vδ2⁽⁺⁾ T-cell subset. Percentages of gated cells are indicated. Summary graph (n = 4) shows mean percentage of PD-1⁽⁺⁾ cells within each indicated Vδ2 subset. Error bars are SD. Vδ2⁽⁺⁾ T-cell subsets are defined as the following: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾].

Fig. S3.

Vδ2⁽⁺⁾ T-cell subsets express IFNγ and TNFα. (A) Intracellular staining for IFNγ and TNFα in gated Vδ2⁽⁺⁾ subsets following 4 h stimulation of PBMCs ex vivo with IL-2 (100 U/mL), zoledronate (20 μM) + IL-2 (100 U/mL), HMB-PP (1 nM) + IL-2 (100 U/mL), or PMA (50 ng/mL) + Ionomycin (1 μg/mL). (B) Stimulation conditions as described in A but for a 24-h period. (C) Summary graph (n = 3) for 4 h stimulation with PMA/Ionomycin (as in A) showing mean percentage of TNFα⁽⁺⁾IFNγ⁽⁺⁾ Vδ2⁽⁺⁾ T cells in each subset. Error bars are SD. Multiple comparison testing using one-way ANOVA with Tukey's post hoc test demonstrated no significant (ns) differences between Vδ2⁽⁺⁾ T-cell subsets. Vδ2⁽⁺⁾ T-cell subsets are defined as the following: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ^(16-) [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾]. Percentages of gated cells are indicated for each quadrant.

Fig. S4.

CD56 expression in Vδ2⁽⁺⁾ T-cell subsets. (A) Representative plots from peripheral blood of a healthy individual showing CD56 expression against size (FSC) in Vδ2⁽⁺⁾ T-cell subsets compared with CD3⁽⁻⁾ lymphocytes (largely NK cells). Percentages of gated cells are indicated. (B) CD56 MFI is shown for Vδ2⁽⁺⁾ subsets (n = 5). Vδ2⁽⁺⁾ T-cell subsets are defined as the following: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ^(28-) [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾]. Multiple comparison testing using one-way ANOVA with Tukey’s posttest used in B. *P < 0.05, **P < 0.01.

Individuals Possess Stable Vδ2 Profiles.

The 63 healthy individuals could be stratified into six Vδ2 profiles defined by relative distribution of the γδ⁽²⁸⁺⁾, γδ⁽²⁸⁻⁾, γδ⁽¹⁶⁻⁾, and γδ⁽¹⁶⁺⁾ subsets (Fig. 2A), which also did not correlate with gender, with developmental index of country of birth, or, as yet, with a limited assessment of ethnicity (Fig. S5). Profile #2 (50–70% γδ⁽²⁸⁺⁾ cells with 10–35% γδ⁽²⁸⁻⁾ cells) was observed most frequently (n = 28), and only two profiles featured a single dominant subset; γδ⁽²⁸⁺⁾ cells were dominant in profile #1 whereas γδ⁽¹⁶⁺⁾ cells dominated profile #6 (Fig. 2A). This is consistent with γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ cells being at opposite ends of a differentiation pathway, with γδ⁽²⁸⁻⁾ and γδ⁽¹⁶⁻⁾ cells in-between. However, the prevalence of any particular profile (e.g., profile #6) did not increase with age (Fig. 2B), suggesting that regular and cumulative exposure to phosphoantigen-producing pathogens is not sufficient to drive differentiation of Vδ2⁽⁺⁾ T cells from one subset phenotype to another. This stable nature of an individual’s Vδ2 profile over time was supported by a longitudinal analysis of Vδ2⁽⁺⁾ T cells from three individuals over 36 mo, in which each person’s Vδ2 profile remained largely constant (Fig. 2C). Moreover, analysis of blood from osteoporotic patients on long-term oral or i.v. bisphosphonates (essentially the chronic activation of Vδ2⁽⁺⁾ T cells in vivo) did not reveal a selective accumulation of any Vδ2⁽⁺⁾ T-cell subset or profile (Fig. 2D). In vitro monitoring over 12 d of Vδ2⁽⁺⁾ T cells after phosphoantigen/IL-2 activation of peripheral blood mononuclear cells (PBMCs) from healthy individuals demonstrated significant recovery of the initial Vδ2 profiles (Fig. S6). And even short-term stimulation of sorted Vδ2⁽⁺⁾ T-cell subsets by phosphoantigen/IL-2 for 3 d, in the presence of sorted CD14⁽⁺⁾ autologous accessory cells, demonstrated little change to the initial Vδ2⁽⁺⁾ T-cell phenotypes (Fig. S7). Thus, the Vδ2⁽⁺⁾ T-cell compartments of normal healthy adult individuals can be stratified into distinct Vδ2 profiles that are stable over time and that do not essentially change in response to acute or chronic Vδ2⁽⁺⁾ T-cell activation. To investigate whether these Vδ2 profiles are “set” during early Vδ2⁽⁺⁾ T-cell development, Vδ2 profiles were assessed from 10 neonatal (4–22 mo) thymuses obtained from cardiac surgeries. Only the CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾ phenotype of profile #1 was evident in all samples (Fig. 2E). Such initial uniformity of Vδ2⁽⁺⁾ T-cell phenotype suggests that the range of distinct Vδ2 profiles is established postthymically, presumably in response to early interactions with the neonatal and/or infant environment.

Fig. 2.

Healthy individuals have stable Vδ2 profiles. (A, Top rows) Representative individuals possess distinct Vδ2 profiles. (Bottom charts) Individuals (n = 63) were assigned to a Vδ2 profile. (B) Chart of mean age (years) for Vδ2 profiles (#1 to #6). (C) Longitudinal analysis of Vδ2 profiles #3, #4, and #6 at t = 0 mo and 36 mo. (D) Plots of Vδ2 profiles from osteoporotic patients on weekly oral (oral BPs; n = 3) or annual (>3 y) i.v. (iv BPs; n = 3) bisphosphonates. (E) Plots of Vδ2⁽⁺⁾ T cells from four infant thymuses (4–22 mo old). Percentage of gated cells is indicated.

Fig. S5.

Vδ2 profiles do not correlate with gender, developmental index of country of birth, or with a limited assessment of ethnicity. Distribution of Vδ2 profiles #1 through #6 is displayed according to (A) gender (male/female), (B) HDI of country of birth, and (C) ethnicity [white (European), Asian (Indian), and others]. HDI is a United Nations summary statistic accounting for life expectancy, income, and education of a country in terms of human development. HDI is divided into low (0.3–0.499), medium (0.5–0.799), high (0.8–0.899), and very high (0.9–0.999) depending on level of a country’s development. Multiple comparison testing using one-way ANOVA with Tukey’s posttest (A–C). ns, not significant.

Fig. S6.

Vδ2 profiles remain stable after stimulation in vitro. Plots of Vδ2⁽⁺⁾ T cells from three individuals with Vδ2 profiles (#1, #3, and #6) during 12 d of HMB-PP (1 nM) + IL-2 (100 U/mL) stimulation in vitro (plots shown at t = 0 and then on day 3, day 7, and day 12). Vδ2⁽⁺⁾ T-cell subsets are defined as the following: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾]. Percentages are indicated for each gate.

Fig. S7.

Phenotypic analysis of sorted Vδ2⁽⁺⁾ T-cell subsets after 3 d of coculture with CD14⁽⁺⁾ monocytes (1:5 Vδ2:monocyte ratio) and activation with HMB-PP (1 nM) + IL-2 (100 U/mL). (A) (Left) Initial Vδ2-profile presort (profile #3). (Middle) Postsort reruns of the four Vδ2⁽⁺⁾ T-cell subsets. (Right) Reanalysis of Vδ2⁽⁺⁾ T-cell subsets after 3 d of coculture/activation as described above. (B) Summary graph from two experiments (n = 2) indicating starting Vδ2⁽⁺⁾ T-cell subset (along horizontal axis) and phenotype of cells after 3 d of coculture/activation (shaded bars). Mean values with error bars (SD) are shown.

Vδ2 Profiles Are Polarized Toward Two Dominant Phenotypes.

Each Vδ2 profile consists of different proportions of γδ⁽²⁸⁺⁾, γδ⁽²⁸⁻⁾, γδ⁽¹⁶⁻⁾, and γδ⁽¹⁶⁺⁾ cells, the combination of which will differentially dictate overall Vδ2⁽⁺⁾ T-cell responses. To explore these functional potentials further, we sorted γδ⁽²⁸⁺⁾, γδ⁽²⁸⁻⁾, γδ⁽¹⁶⁻⁾, and γδ⁽¹⁶⁺⁾ cells to high purity from three different individuals to perform microarray expression analysis (using Illumina’s HumanHT-12 v4 beadchip array). After multiple comparisons for genes expressed at least twofold higher or lower between any two subsets the γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ populations appeared most different (469 genes), consistent with these subsets having distinct functional potentials (Fig. 3A). By contrast, only 78 genes differed between γδ⁽²⁸⁻⁾ and γδ⁽¹⁶⁻⁾ cells.

Fig. 3.

Vδ2 profiles polarize toward two dominant phenotypes. (A) Number of differentially expressed genes (two- or more fold difference and P < 0.05) between Vδ2⁽⁺⁾ subsets. Multiple comparison tests were with false discovery rate Benjamini–Hochberg P-value correction. (B) Average probability state model (Gemstone software) for CD28, CD27, CD16, CD45RA, and CD11a on Vδ2⁽⁺⁾CD3⁽⁺⁾ cells (n = 63). Lines show average relative intensity of marker along a cumulative progression axis (x axis). (C) Differentially expressed immune-related genes with two- or more fold expression difference (P < 0.05) between γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ subsets. Dark shaded bars indicate in which subset gene expression is highest.

We also reanalyzed our flow cytometry data using Gemstone software that assesses multiple flow parameters from multiple flow cytometry data files simultaneously (www.vsh.com/products/gemstone/). This is used to order or group cells by phenotypic similarity. Using assumptions that CD27, CD28, and CD16 expression can be high or low on any cell, the software analyzed all collected events from all 63 individuals to generate a summary representation of common cell phenotypes from all samples (Fig. 3B). This analysis supported the idea of dominant γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ subsets and again suggested less well-defined distinctions between γδ⁽²⁸⁻⁾ and γδ⁽¹⁶⁻⁾ cells with several additional intermediate phenotypes. Taken together with our data indicating that Vδ2 profiles are stable over time, it suggests that the Vδ2⁽⁺⁾ T-cell compartments of healthy adult individuals are polarized, to different extents, toward one of two major effector potentials, dominated, respectively, by either the γδ⁽²⁸⁺⁾ or γδ⁽¹⁶⁺⁾ subset.

Demarcation of Dominant Vδ2⁽⁺⁾ T-Cell Phenotypes by CCR6 and CX3CR1.

To explore Vδ2 profile effector potentials further, a table of immune-related genes that differed in expression by ≥twofold between γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ cells was compiled (Fig. 3C). This highlighted differential expression in γδ⁽¹⁶⁺⁾ cells of genes related to cytotoxic function, including perforin, granzyme B, granzyme H, granulysin, and eight members of the killer cell Ig-like receptor (KIR) family. By contrast, γδ⁽²⁸⁺⁾ cells expressed higher levels of granzyme K (30) and cytokine and chemokine receptors such as IL-7Rα, IL-18Rα, IL-23R, CCR2, CCR7, and CCR6 (see below).

Of particular interest was differential expression of CCR6 in γδ⁽²⁸⁺⁾ cells (∼4.4-fold higher) and of CX3CR1 (fractalkine receptor) in γδ⁽¹⁶⁺⁾ cells (∼10-fold higher), which was validated in 12 individuals (Fig. 4A). CX3CR1 was expressed at high levels by all γδ⁽¹⁶⁺⁾ cells, but only by a fraction of the γδ⁽²⁸⁺⁾ subset and at a lower per-cell level. Conversely, CCR6 was not detected on γδ⁽¹⁶⁺⁾ cells, but identified a distinct population of CCR6⁽⁺⁾CX3CR1⁽⁻⁾ γδ⁽²⁸⁺⁾ cells. An identifiable subset of CCR6⁽⁺⁾CX3CR1⁽⁺⁾ double-positive cells was never truly evident, suggesting a mutually exclusive relationship that may have functional implications. Consistent with this, the intensity of CX3CR1 expression correlated with cytolytic potential as judged by coexpression of granzyme B and perforin (Fig. 4B). By contrast, CCR6⁽⁺⁾ γδ⁽²⁸⁺⁾ cells were GzmB⁽⁻⁾Pfn⁽⁻⁾ but expressed granzyme K (Fig. 4C). CCR6⁽⁺⁾ γδ⁽²⁸⁺⁾ cells also expressed higher surface levels of both CCR2 and CCR5 (Fig. 4D) and the IL-18Rα chain (Fig. 4E), compared with CX3CR1⁽⁺⁾ cells. CCR6 and granzyme K were also expressed (in the absence of CX3CR1 and granzyme B) in a sizable proportion of Vδ2⁽⁺⁾ T cells from all our neonatal thymus samples (Fig. 4F), supporting the idea that freshly generated Vδ2⁽⁺⁾ T cells initially adopt a profile #1 phenotype.

Fig. 4.

Demarcation of Vδ2⁽⁺⁾ T-cell phenotypes by CCR6 and CX3CR1. (A) Representative plots (Left) and summary charts (Right) showing (A) CCR6 and CX3CR1 on γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ T cells in healthy individuals (n = 12), gating CCR6⁽⁺⁾ γδ⁽²⁸⁺⁾ cells, CX3CR1⁽⁺⁾ γδ⁽²⁸⁺⁾ cells, and CX3CR1⁽⁺⁾ γδ⁽¹⁶⁺⁾ cells. Subsets from A, further stained for intracellular perforin and granzyme B (B), granzyme K and granzyme B (C), or surface CCR2 and CCR5 (D), following 4 h PMA/Ionomycin stimulation (n = 6–9). (E) MFI for IL-18Rα for subsets described in A (n = 5). (F) Plots from four infant thymuses (4–22 mo old) for CCR6, CX3CR1, and intracellular granzyme K and granzyme B. (G) CCR6 and intracellular IL-17A expression in Vδ2⁽⁺⁾ and Vδ2⁽⁻⁾ (predominantly αβ T-cell) subsets from blood (n = 2). (H) MFI for CD161 for Vδ2⁽⁺⁾ subsets described in A (n = 6). Mean values with error bars (SD). Multiple comparisons using one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

CCR6, along with RORγt and IL-23R that are also differentially up-regulated in γδ⁽²⁸⁺⁾ cells (Fig. 3C), are associated with an IL-17A–secreting phenotype (31, 32). Unlike in mice, human γδ cells with IL-17A–secreting potential are not abundant in adult peripheral blood (12). However, they are present in cord blood (33) and in inflammatory lesions of immunopathologies such as meningitis and psoriasis (12, 34); thus, their origin is of particular importance. As expected, very few IL-17A–secreting Vδ2⁽⁺⁾ T cells were observed in peripheral blood of any individual tested. Nonetheless, when these cells were observed they expressed CCR6 (Fig. 4G) and resided in the γδ⁽²⁸⁺⁾ subset (Fig. S8). Moreover, CCR6⁽⁺⁾ γδ⁽²⁸⁺⁾ cells expressed higher levels of CD161, the up-regulation of which was recently demonstrated on IL-17A–secreting Vδ2⁽⁺⁾ T cells in vitro (12) (Fig. 4H).

Fig. S8.

IL-17A–producing Vδ2⁽⁺⁾ T cells reside within the CCR6⁽⁺⁾ fraction of the γδ⁽²⁸⁺⁾ subset. Plots from two different healthy individuals showing expression of IL-17A and IFNγ in CCR6⁽⁺⁾ and CCR6⁽⁻⁾ Vδ2⁽⁺⁾ T-cell subsets after 4 h stimulation with PMA (50 ng/mL) and Ionomycin (1 μg/mL). Vδ2⁽⁺⁾ T-cell subsets are defined as the following: γδ⁽²⁸⁺⁾ [CD28⁽⁺⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽²⁸⁻⁾ [CD28⁽⁻⁾CD27⁽⁺⁾CD16⁽⁻⁾], γδ⁽¹⁶⁻⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁻⁾], and γδ⁽¹⁶⁺⁾ [CD28⁽⁻⁾CD27⁽⁻⁾CD16⁽⁺⁾]. Percentages of gated cells are indicated.

Distinct Vδ2 Profiles Display Distinct Cytotoxic Effector Potentials.

The data above predict that healthy adult individuals with distinct Vδ2 profiles will display qualitatively different effector responses to phosphoantigen challenge. To explore this, PBMC were isolated from individuals with either Vδ2 profile #1 or #2, in which γδ⁽²⁸⁺⁾ cells dominate, or Vδ2 profile #5 or #6, in which γδ⁽¹⁶⁺⁾ cells dominate. As expected, total Vδ2⁽⁺⁾ T cells from profiles #1 and 2 expanded on average >10-fold more than those from profiles #5 and 6 in response to HMB-PP/IL-2 (Fig. 5A). Next, we assessed cytotoxic potential by first confirming, in five further subjects, that Vδ2⁽⁺⁾ T cells from profile #6 individuals express high levels of granzyme B and granulysin, whereas Vδ2⁽⁺⁾ T cells from profile #1 individuals instead express granzyme K (Fig. 5B). Consistent with this contrast in cytotoxic potential, HMB-PP/IL-2–expanded Vδ2⁽⁺⁾ T-cell lines from profile #6 subjects efficiently lysed MOLT-4 cells (acute lymphoblastic leukemia), DOHH-2 cells (B-cell lymphoma), and HL-60 cells (acute promyelocytic leukemia) to a much greater extent than expanded Vδ2⁽⁺⁾ T-cell lines from profile #1 individuals (Fig. 5C). Nonetheless, Vδ2⁽⁺⁾ T-cell lines from both profiles efficiently killed Jurkat cells (acute T-cell leukemia), whereas Vδ2⁽⁺⁾ T-cell lines from profile #1 individuals killed HCT-116 cells (colon carcinoma) more effectively than those from profile #6 (Fig. 5C). HCT-116 cells are reported to express proteinase inhibitor 9 (PI-9), a potent inhibitor of granzyme B (but not granzyme K) (35), possibly explaining the reduced cell killing in profile #6 cultures. This differential killing was confirmed using sorted Vδ2⁽⁺⁾ T cells from profile #1 and #6 individuals activated with HMB-PP/IL-2 for 4 h and then cocultured with HCT-116 cells (Fig. 5D). Importantly, the greater killing of HCT-116 cells in profile #1 cultures could be inhibited by preincubation of the Vδ2⁽⁺⁾ T cells with nafomostat mesylate (NM) that inhibits granzyme K (36) (Fig. 5D). NM also reduced killing of MOLT-4 cells by freshly isolated HMB-PP/IL-2–activated Vδ2⁽⁺⁾ T cells from profile #1 individuals, whereas granzyme B inhibitor Z-Ala-Ala-Asp(OMe)-chloromethyl ketone (Z-AAD) (37) had minimal effect (Fig. 5E). Conversely however, Z-AAD, but not NM, was able to inhibit MOLT-4 cell killing in similar cultures of activated Vδ2⁽⁺⁾ T cells from profile #6 individuals, confirming a granzyme B-mediated mode of cytotoxicity (Fig. 5E). Vδ2⁽⁺⁾ T cells from profile #6 subjects also express abundant CD16 that can mediate antibody-dependent cellular cytotoxicity. Consistent with this, DAUDI cells (Burkitt’s lymphoma) that are killed to some degree by the anti-CD20 antibody rituximab, show almost complete cell lysis in the presence of activated Vδ2⁽⁺⁾ T cells from profile #6 individuals (Fig. 5F). By contrast, Vδ2⁽⁺⁾ T cells from profile #1 subjects in the presence of rituximab, and Vδ2⁽⁺⁾ T cells from either profile in the absence of rituximab, displayed no significant DAUDI cell killing above control (Fig. 5F). In sum, our data demonstrate that individuals with distinct Vδ2 profiles display qualitatively distinct Vδ2⁽⁺⁾ T-cell–mediated cytotoxic effector function that is mediated in large part by differential expression of cytotoxic mediators such as granzyme B and K. Importantly, this suggests that only individuals possessing Vδ2 profiles that are compatible with particular features of their tumor targets are likely to show promising clinical outcomes.

Fig. 5.

Distinct Vδ2 profiles display distinct cytotoxic effector potentials. (A) Fold-change in Vδ2⁽⁺⁾ cells as percentage of CD3⁽⁺⁾ cells 7 d after stimulation of PBMCs with zoledronate (1 μM) and IL-2 (100 U/mL), comparing Vδ2 profiles #1 and #2 (n = 10) with profiles #5 and #6 (n = 4). (B) Representative plots and summary MFIs (n = 6) for intracellular granzyme B, granzyme K, and granulysin from sorted Vδ2⁽⁺⁾ cells from Vδ2 profiles #1 or #6 after 4 h stimulation with PMA/ionomycin. (C) Chart showing percentage of apoptosis of tumor lines after 4 h coculture with sorted Vδ2⁽⁺⁾ T cells (Vδ2 profiles #1 or #6) from PBMC cultures stimulated for 12 d with 1 nM HMB-PP and 100 U/mL IL-2. Effector:target ratios were 5:1. Error bars are SD. (D–F) Charts showing percentage apoptosis of PI-9⁽⁺⁾ HCT-116 (D), MOLT-4 (E), or DAUDI (F) cells after 4 h coculture with sorted and activated (1 nM HMB-PP and 100 U/mL IL-2) Vδ2⁽⁺⁾ T cells from Vδ2 profiles #1 or #6 (n = 3) ex vivo. Pretreatment (16 h) of Vδ2⁽⁺⁾ cells with 100 μM NM was used to inhibit granzyme K (D and E). Granzyme B was inhibited by 100 μM Z-AAD (E). Rituximab (anti-CD20 mAb) was included to induce antibody-dependent cellular cytotoxicity (F). Effector:target ratios were 5:1. Error bars are SD. Differences compared with two-tailed, unpaired Student’s t tests. *P < 0.05, **P < 0.01, and ***P < 0.001.

Discussion

Current understanding of Vδ2⁽⁺⁾ T-cell biology does not adequately explain the variability in objective clinical responses in Vδ2⁽⁺⁾ T-cell–based immunotherapies against cancer (26–28). In addressing this, we demonstrate significant interindividual heterogeneity in Vδ2⁽⁺⁾ T-cell phenotype that translates to distinct Vδ2⁽⁺⁾ T-cell effector potentials. Our analysis of 63 healthy individuals revealed several Vδ2 profiles, each defined by differing proportions of Vδ2⁽⁺⁾ T-cell subsets with distinct functional characteristics. Using multiple analyses, our data suggest that these Vδ2 profiles consist of varying contributions from two dominant but contrasting Vδ2⁽⁺⁾ T-cell populations identified as γδ⁽²⁸⁺⁾ and γδ⁽¹⁶⁺⁾ cells. Moreover, these opposing phenotypes segregate by mutually exclusive expression of chemokine receptors CCR6 and CX3CR1. CX3CR1 correlates with increased granzyme B/perforin-associated cytotoxic capacity but decreased proliferative potential, whereas CCR6 characterizes γδ⁽²⁸⁺⁾ cells with increased expression of cytokine receptors (e.g., IL-18Rα) and chemokine receptors (e.g., CCR2 and CCR5) and expression of granzyme K. CCR6⁽⁺⁾ γδ⁽²⁸⁺⁾ cells also include the few IL-17A–secreting Vδ2⁽⁺⁾ T cells found in peripheral blood.

A key finding is that distinct Vδ2 profiles appear stable over time, as highlighted in three individuals over 36 mo. It had been suggested from studies in vitro that phosphoantigen stimulation of Vδ2⁽⁺⁾ T cells resulted in differentiation from a CD27⁽⁺⁾CD45RA⁽⁺⁾ phenotype to either a CD27⁽⁻⁾CD45RA⁽⁻⁾ or CD27⁽⁻⁾CD45RA⁽⁺⁾ phenotype via a CD27⁽⁺⁾CD45RA⁽⁻⁾ intermediate (15). However, in our cohort, older individuals, who inevitably will have been exposed to more phosphoantigen-producing pathogens than younger individuals, did not favor any particular Vδ2 profile. Moreover, patients on long-term (>36 mo) bisphosphonate therapies, either orally or intravenously, did not present with any specific Vδ2 profile, as would be expected if repeated stimulation drove Vδ2⁽⁺⁾ T-cell differentiation. Together, these findings suggest Vδ2 profiles change little over time despite repeated stimulation. This evokes a model in which Vδ2⁽⁺⁾ T-cell activation expands short-lived effector cells, but preserves proportions of longer-lived “memory” cells to maintain the Vδ2 profile. The underlying mechanism for this requires further investigation but may involve “memory stem cell” activity, akin to that described for memory CD8⁽⁺⁾ αβ T cells (38), although whether Vδ2⁽⁺⁾ T cells demonstrate true memory is still uncertain (39).

The stable yet heterogeneous nature of Vδ2 profiles between healthy individuals raises the question of how they develop. The neonatal (4–22 mo) thymus data demonstrate a uniformity of phenotype that resembles profile #1, with a predominance of γδ⁽²⁸⁺⁾ cells, and expression of CCR6 and granzyme K. This suggests that, presumably early in life (40), an individual’s Vδ2⁽⁺⁾ T-cell compartment transitions from a uniform thymic profile to a particular peripheral profile that constitutes a stable set-point. The drivers of this transition may be genetic. Alternatively, it may result from childhood exposure to specific pathogens, such as M. tuberculosis (19), P. falciparum (20), or CMV (41). Such infections may cause aggressive, or qualitatively different, Vδ2⁽⁺⁾ T-cell responses that force a change in Vδ2 profile. Such profile-changing responses could also depend on certain Vγ9Vδ2TCR specificities that are restricted to certain individuals. These ideas are currently under investigation.

A corollary to identification of distinct stable Vδ2 profiles across healthy subjects is that these individuals should demonstrate distinct Vδ2⁽⁺⁾ T-cell–mediated effector responses. Indeed, we demonstrate differences in both efficiency and mechanism by which Vδ2⁽⁺⁾ T cells, from profile #1 versus profile #6 individuals, kill a range of tumor lines. Profile #6 Vδ2⁽⁺⁾ T cells show granzyme B-mediated cytotoxicity, but are not efficient at killing tumor lines that express the granzyme B inhibitor PI-9. However, these cells can additionally deploy ADCC to kill tumor targets in the presence of tumor-binding antibodies. By contrast, Vδ2⁽⁺⁾ T cells from profile #1 individuals favor granzyme K-mediated cytotoxicity and appear more effective at killing PI-9⁽⁺⁾ tumor cells than profile #6 Vδ2⁽⁺⁾ T cells. These data thus have important clinical implications for patient stratification in Vδ2⁽⁺⁾ T-cell–based therapies, as there is likely an optimal Vδ2⁽⁺⁾ T-cell response for any given malignancy. Although the Vδ2⁽⁺⁾ T-cell compartment is in theory capable of making a range of responses, a particular patient may be restricted in the responses that they can actually make. If this effector potential is mismatched to that required to target a particular tumor, treatment is likely to fail, which may have contributed to inconsistent patient responses in recent clinical trials (26–28, 42). By contrast, selecting patients with defined Vδ2⁽⁺⁾ T-cell effector potentials that complement features of a tumor, and tailoring activation protocols to optimize specific responses, may maximize the chances of future treatment success.

Materials and Methods

Informed consent and ethical approval was obtained for blood [London City and East Research Ethics Committee (REC) 13/LO/0548] and thymus samples (London Queen Square REC 14/LO/2132).

The isolation of PBMCs, flow cytometry, proliferation assays, cytotoxicity assays, degranulation assay, Gemstone probability state modeling, microarray analysis, and statistical analyses are described in SI Materials and Methods.

SI Materials and Methods

Study Population.

Peripheral blood came from 63 healthy volunteers (31 male, 32 female, median age 33 y, range 3–69 y) and 10 osteoporotic patients (2 male, 8 female, median age 70 y, range 44–81 y) with informed consent and ethical approval (London City and East REC 13/LO/0548). Thymus samples were from neonatal (<24 mo) cardiac surgery with informed consent and ethical approval (London Queen Square REC 14/LO/2132).

Isolation of PBMCs.

PBMCs were isolated from peripheral blood by Ficoll-Paque (GE Healthcare) gradient centrifugation at 400 × g (brake off) for 35 min at 20 °C. Interface cells were harvested and washed in PBS supplemented with 10% (vol/vol) FBS and 5 mM EDTA (Gibco and Life Technologies).

Flow Cytometry.

Fluorochrome-conjugated antibodies were from eBioscience, Becton Dickinson, or BioLegend: CD3ε (HIT3a), CD11a (HI111), CD16 (3G8), CD27 (O323), CD28 (CD28.2), CD45RA (HI100), CD56 (HCD56), CD57 (NK-1), CD62L (DREG-56), CD107a (H4A3), CD161 (HP-3G10), CCR2 (K036C2), CCR5 (HEK/1/85a), CCR6 (G034E3), CCR7 (G043H7), CX3CR1 (K0124E1), Vδ2 (B6), granzyme B (GB11), granzyme K (G3H69), granulysin (DH2), IFNγ (4S.B3), IL-17A (BL-168), IL18Rα (H44), PD-1 (EH12.2H7), perforin (dG9), and TNFα (MAb11). For chemokine receptors, cells were stained at 20 °C with fluorochrome-conjugated antibodies diluted in FACS buffer [PBS, 2% (vol/vol) FBS, 5 mM EDTA]. For other surface markers PBMCs were stained on ice, washed, and resuspended in FACS buffer containing 0.5 μg/mL DAPI (Invitrogen) for dead cell exclusion before analysis. For cytokines, cells were either stimulated with 50 ng/mL PMA (Sigma) and 1 μg/mL ionomycin (Sigma), or 1 nM HMB-PP (Sigma) and 100 U/mL rhIL-2 (PeproTech), or 20 μM zoledronate (Sigma) and 100 U/mL rhIL-2 for 4, 6, or 24 h at 37 °C; 10 μg/mL Brefeldin A (eBioscience) and 2 μM Monensin (eBioscience) were added for at least the last 2 h, or 8 h if stimulating cells for 24 h. Cells were stained for surface markers, fixed with IC fixation buffer (eBioscience) for 15 min on ice, permeabilized, and stained with intracellular cytokine-specific antibodies diluted in permeabilization buffer (eBioscience). Samples were acquired using a Canto-II, LSR-II, or FACSAria-II flow cytometer (BD) and analyzed using FlowJo software (Tree Star, Inc.).

Proliferation Assays.

First γδ T cells were enriched from PBMCs by magnetic separation with a biotin-conjugated TCRδ antibody (Miltenyi Biotec). Subsequently, enriched γδ T cells were FAC-sorted into Vδ2⁽⁺⁾ subsets on a FACSAria-II. Sorted cells were labeled with 5 μM eFluor670 (eBioscience) for 10 min at 37 °C with intermittent mixing and quenched with ice-cold RPMI-1640 (Life Technologies) containing 10% (vol/vol) FBS. Labeled cells were cultured with Vδ2-depleted mononuclear accessory cells at a 1:1 ratio at 37 °C in RPMI-1640 complete media [RPMI-1640 supplemented with 10% (vol/vol) FBS, 100 U/mL penicillin and 100 μg/mL streptomycin] in 96-well round-bottom plates in the presence of 1 μM zoledronate and 100 U/mL rhIL-2. After 7 d, cells were analyzed for eFluor670 dilution. FAC-sorted Vδ2⁽⁺⁾ subsets were also cultured for 3 d with autologous CD14⁽⁺⁾ monocytes, isolated by magnetic separation with a biotin-conjugated anti-CD14 antibody (BioLegend) at a 1:5 (Vδ2:monocyte) ratio with 1 nM HMB-PP and 100 U/mL rhIL-2 at 37 °C in RPMI-1640 complete media.

Cytotoxicity Assays.

Vδ2⁽⁺⁾ T cells were isolated to high purity with magnetic-activated cell sorting (MAC-sorted) using a biotin-conjugated anti-human Vδ2 antibody (Miltenyi Biotec) from either freshly isolated PBMCs or Vδ2⁽⁺⁾ T-cell lines (from profile #1 and #6 donors) that had been expanded from PBMCs over 12 d with 1 nM HMB-PP and 100 U/mL rhIL-2 with replenishment of rhIL-2 every 3 d. Sorted Vδ2⁽⁺⁾ T cells were then rested for 16 h in RPMI-1640 media supplemented with 100 U/mL rhIL-2 ± 100 μM NM (granzyme K inhibitor, Cayman Chemical) before use. MOLT-4, Jurkat, HL-60, DOHH-2, Daudi, and HCT-116 cell lines were labeled with 1 μM CellTrace Far Red DDAO-SE (Molecular Probes). Vδ2⁽⁺⁾ T cells that were MAC-sorted from freshly isolated PBMCs were preactivated for 4 h with 1 nM HMB-PP and 100 U/mL rhIL-2 before coculture with target cell lines. All coculture cytotoxicity assays were performed in 96-well round-bottom plates at a 5:1 effector:target ratio for 4 h in RPMI-1640 supplemented with 100 U/mL rhIL-2, ± 100 μM Z-AAD-CMK (granzyme B inhibitor, Abcam). For ADCC assays, Daudi cells were rested or pretreated with 1 μg/mL anti–hCD20-hIgG1 (InvivoGen) for 30 min at 37 °C. Vδ2⁽⁺⁾ T cells were then added and activated with 1 nM HMB-PP and 100 U/mL rhIL-2. Cells were stained with Annexin V-PE (BioLegend) and Zombie aqua (BioLegend) and analyzed on a NovoCyte Flow Cytometer (ACEA Biosciences).

Degranulation Assay.

Vδ2⁽⁺⁾ T-cell degranulation was determined by CD107a surface expression. PBMCs (1 × 10⁶) were cultured for 8 h in 96-well round-bottom plates in RPMI-1640 complete media with 20 nM HMB-PP, 100 U/mL of rhIL-2, 1 μM Monensin and anti-human CD107a at 37 °C. PBMCs were then stained for cell-surface markers as described.

Probability State Modeling.

GemStone version 1.0.69 (Verity Software House) was used to model extracellular marker similarity of flow-cytometry data of Vδ2⁽⁺⁾ T cells as per manufacturer’s instructions. Modeling used simple rules/assumptions about directional changes in marker expression with data from all 63 samples integrated into a final model.

Microarray Analysis.

Total RNA was extracted using RNeasy Micro Kit (Qiagen) according to the manufacturer’s instructions. After confirming concentration, purity, and integrity, total RNA was reverse-transcribed and then converted to cRNA using the Ambion WT Expression Kit (Life Technologies). The HumanHT-12 v4 Expression BeadChip (Illumina) was used for the microarray [GEO database (www.ncbi.nlm.nih.gov/bioproject/) accession no. GSE75519]. Data were analyzed using Genome Studio (Illumina) with a threshold for differential gene expression of two- or more fold and P value < 0.05 (Benjamini–Hochberg false discovery rate).

Statistical Analysis.

Statistical analysis was performed by Prism 6.0 (GraphPad). Data are mean ± SD.