Memory T cell–driven differentiation of naive cells impairs adoptive immunotherapy

Abstract

Adoptive cell transfer (ACT) of purified naive, stem cell memory, and central memory T cell subsets results in superior persistence and antitumor immunity compared with ACT of populations containing more-differentiated effector memory and effector T cells. Despite a clear advantage of the less-differentiated populations, the majority of ACT trials utilize unfractionated T cell subsets. Here, we have challenged the notion that the mere presence of less-differentiated T cells in starting populations used to generate therapeutic T cells is sufficient to convey their desirable attributes. Using both mouse and human cells, we identified a T cell–T cell interaction whereby antigen-experienced subsets directly promote the phenotypic, functional, and metabolic differentiation of naive T cells. This process led to the loss of less-differentiated T cell subsets and resulted in impaired cellular persistence and tumor regression in mouse models following ACT. The T memory–induced conversion of naive T cells was mediated by a nonapoptotic Fas signal, resulting in Akt-driven cellular differentiation. Thus, induction of Fas signaling enhanced T cell differentiation and impaired antitumor immunity, while Fas signaling blockade preserved the antitumor efficacy of naive cells within mixed populations. These findings reveal that T cell subsets can synchronize their differentiation state in a process similar to quorum sensing in unicellular organisms and suggest that disruption of this quorum-like behavior among T cells has potential to enhance T cell–based immunotherapies.

Introduction

Adoptive cell transfer (ACT), the ex vivo expansion and reinfusion of antigen-specific (Ag-specific) T cells, represents a potentially curative treatment for patients with advanced cancer (1–4) and viral-reactivation syndromes (1, 5, 6). Recent progress in the ability to genetically redirect patient-derived peripheral blood T cells toward tumor and viral-associated antigens by modification with a T cell receptor (TCR) or chimeric antigen receptor (CAR) has greatly simplified the generation of therapeutic T cells (7–10). Given the clinical efficacy of T cell therapy combined with the ability of T cells to be manufactured according to standardized procedures, ACT is now poised to enter mainstream clinical practice. However, fundamental questions remain regarding the optimal source, expansion, and quality of therapeutic T cells used for transfer.

In mice, ACT of naive CD8⁺ T cell–derived cells (T_N-derived cells) exhibits a superior capacity to expand, persist, and treat cancer compared with normalized numbers of memory T cell–derived cells (T_Mem cells) (11, 12). Preclinical human studies have confirmed that T_N-derived cells maintain higher levels of the costimulatory marker CD27 and the lymphoid homing markers CD62L and CCR7; they also retain longer telomeres (12–15). Each of these parameters has correlated with the likelihood that patients will obtain an objective clinical response following ACT (15–17). Despite these findings, the majority of current T cell therapy clinical trials do not specifically enrich for defined T cell subsets, but rather utilize unfractionated T cell populations (2). As T_N cells are in the circulation of most cancer patients (13, 18), the following question arises: is the presence of T_N cells in the initial population used to generate therapeutic T cells sufficient to convey their desirable attributes, or is physical separation of T_N cells from antigen-experienced subsets required to unleash the full therapeutic potential of T_N-derived cells (19, 20)? Prior investigations revealed that T_N cells form homotypic clusters during T cell priming that can influence their subsequent maturation (21, 22). However, whether antigen-experienced populations directly interact with and influence naive cell differentiation is unknown.

Using human and mouse T cells, we describe here a previously unrecognized T cell–T cell interaction whereby T_Mem cells directly influence T_N cell differentiation during priming. This process, which we term precocious differentiation, synchronizes the behavior of T_N-derived cells with T_Mem cells, resulting in accelerated functional, transcriptional, and metabolic differentiation of T_N cell progeny. Precocious differentiation was cell-dose, contact, and activation dependent. Mechanistically, the phenomenon was mediated by nonapoptotic Fas signaling, resulting in activation of Akt and ribosomal S6 protein (S6), kinases responsible for cellular differentiation and metabolism (23). Consequently, induction of Fas signaling in the absence of T_Mem cells enhanced differentiation and impaired antitumor immunity, while isolation of T_N cells prior to priming or blockade of Fas signaling prevented T_Mem cell–induced precocious differentiation and preserved the antitumor efficacy of T_N-derived cells. Collectively, our results reveal that unleashing the therapeutic potential of T_N-derived cells for adoptive immunotherapy necessitates disruption of intercellular communication with T_Mem cells, a finding with direct implications for the design and execution of ACT clinical trials.

Results

T_Mem augment naive cell phenotypic maturation during ex vivo priming.

We sought to determine whether antigen-experienced CD8⁺ T cells influence the differentiation of T_N-derived progeny. To indelibly track the fate of T_N cells, we primed congenically distinguishable Thy1.1⁺ pmel-1 CD8⁺ T_N cells (CD44^loCD62L⁺), which recognize an epitope derived from the melanoma-associated Ag gp100 (24), alone or in a 1:1 mixture with Ly5.1⁺ T_Mem cells. To generate T_Mem cells, we adoptively transferred Ly5.1⁺ pmel-1 T cells into WT Ly5.2⁺ hosts and vaccinated recipient mice with a gp100-encoding recombinant vaccinia virus (rVV-gp100) to generate Ag-experienced CD8⁺ T cells in vivo. Twenty-eight days later, Ly5.1⁺CD8⁺ T central memory cell (T_CM cell; CD44^hiCD62L⁺) and T effector memory cell (T_EM cell; CD44^hiCD62L^–) subsets from vaccinated mice were isolated by FACS sorting (Figure 1A). Naive pmel-1 cells were subsequently expanded alone or in the presence of T_CM or T_EM cells using anti-CD3/CD28 antibodies and IL-2, which is similar to many current human ACT protocols (15, 25, 26).

Figure 1. T_Mem cells augment naive cell phenotypic maturation during priming.

(A) Experimental design and representative FACS plot showing the generation and isolation of vaccine-induced pmel-1 Ly5.1⁺CD8⁺ T_CM (CD44⁺CD62L⁺) and T_EM (CD44⁺CD62L^–) cell subsets by adoptive transfer of pmel-1 cells (10⁶) into Ly5.2⁺ WT hosts followed by rVV-gp100 vaccination (2 × 10⁷ pfu). (B) Representative FACS plots and (C) summary bar graphs demonstrating the distribution of Thy1.1⁺ T_N-derived T_SCM (CD44^loCD62L⁺Sca1⁺CD122⁺) and T_EM cell subsets 6 days following ex vivo expansion alone or in the presence of a 1:1 mixture with Ly5.2⁺ T_CM or T_EM cells using CD3/CD28-specific antibodies and IL-2. (D) Representative FACS plots and (E) summary graph demonstrating the distribution of Thy1.1⁺ T_N-derived subsets 6 days following ex vivo expansion alone or in the presence of titrated ratios with Ly5.1⁺ bulk (CD44⁺) T_Mem cells using CD3/CD28-specific antibodies and IL-2. Results in A–E are shown after gating on live⁺CD8⁺Thy1.1⁺ lymphocytes and are representative of 2 independently performed experiments. Results in C and E are presented as mean ± SEM with n = 3 per condition. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment. ***P < 0.001.

Following expansion, the progeny of isolated T_N cells had differentiated into all 3 antigen-experienced subsets, including T stem cell memory cells (T_SCM cells; CD44^loCD62L⁺CD122⁺Sca-1⁺), T_CM cells, and T_EM cells (refs. 27, 28, Figure 1B, and Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI81217DS1). Strikingly, we found that the presence of T_Mem cells during T_N cell priming caused a significant accumulation of T_EM cells and an attrition of T_SCM cells (Figure 1C). Both T_CM and T_EM cell subsets were equally capable of augmenting the phenotypic maturation of T_N-derived progeny, demonstrating that this prodifferentiation capacity was a property of T_Mem cell subsets in general. These findings were not attributable to differences in cell expansion, were not TCR specific, and occurred regardless of growth and cytokine conditions tested (Supplemental Figures 2 and 3). Augmented T_N cell differentiation was also observed when a CD4-containing bulk T_Mem cell population was used, although the magnitude of change was less than that observed using CD8⁺ T_Mem cells (Supplemental Figure 4).

As the frequency of circulating T_Mem cells to T_N cells can vary widely in patients (18, 29, 30), we evaluated the dose dependency of T_Mem cell–induced phenotypic maturation of T_N cells. We activated T_N cells alone or in titrated ratios with a bulk (CD44⁺) T_Mem cell population. Similar to results using fractionated T_CM and T_EM cell subsets, we found that priming T_N cells in a 1:1 mixture with bulk T_Mem cells caused an increase in the T_EM cell population and a corresponding loss in T_SCM and T_CM cells (Figure 1, D and E). As the ratio of T_Mem to T_N cells was increased, we measured a progressive enhancement in the differentiation of T_N-derived progeny. We conclude that the presence of T_Mem cells during T_N cell priming ex vivo caused dose-dependent phenotypic maturation of T_N-derived cells.

T_Mem cause precocious differentiation of naive cells.

Having established that T_Mem cells augment the phenotypic maturation of T_N cells during priming, we next evaluated whether T_Mem cells influence gene transcription and effector functions of T_N-derived progeny. We again primed Ly5.1⁺ T_N pmel-1 cells alone or in a 1:1 ratio with bulk in vitro–differentiated Ly5.2⁺ T_Mem cells and confirmed that the progeny of T_N cells expanded with T_Mem cells (T_N-derived, mix) were significantly skewed toward T_EM cells and away from T_SCM cells compared with T_N cells primed alone (T_N-derived, alone) (Figure 2, A and B). To determine whether CD62L loss on T_N-derived cells was attributable to genetic downregulation rather than proteolytic cleavage, we reisolated Ly5.1⁺ T_N-derived cells following expansion by FACS sorting and performed quantitative PCR (qPCR). We found that expression of selectin L (Sell), the gene encoding CD62L as well as other T_N cell–associated lymphoid-homing and costimulatory markers, including Ccr7 and Cd27, was significantly downregulated in T_N-derived cells expanded with T_Mem cells relative to T_N cells expanded alone (Figure 2C). Additionally, we found that T_N-derived cells primed with T_Mem cells were more functionally differentiated, as evidenced by increased granzyme B content (Figure 2D) and enhanced IFN-γ secretion following stimulation with either PMA/ionomycin or hpg100_25–33 peptide (Figure 2, E and F).

Figure 2. T_Mem cause precocious differentiation of naive cells.

(A) Representative FACS and (B) bar graph summarizing the distribution of Ly5.1⁺CD8⁺ T_N-derived cell subsets 6 days following priming with CD3/CD28-specific antibodies and IL-2 alone or with Ly5.2⁺CD8⁺ T_Mem cells. Data shown after gating on Ly5.1⁺CD8⁺cells. (C) qPCR analysis of Sell, Ccr7, and Cd27 expression in FACS-sorted reisolated T_N cells primed alone, T_N cells primed with T_Mem cells, or T_Mem cells primed alone. (D) Granzyme B and (E) IFN-γ intracellular staining in T_N-derived cells stimulated with PMA/ionomycin following expansion alone or with T_Mem cells. (F) IFN-γ ELISA of supernatants from reisolated T_N-derived cells expanded alone or with T_Mem cells 6 days prior to overnight stimulation with hgp100_25–33 peptide. (G) Heat maps of differentially expressed genes (1-way ANOVA, pFDR < 5%) among T_N-derived cells expanded alone or with T_Mem cells at 18 and 96 hours. (H) RMA-normalized intensity of selected T_N-associated genes. (I) Expression of effector cell–associated factors assessed by qPCR at 96 hours from FACS reisolated T_N-derived progeny expanded with or without T_Mem cells or T_Mem cells expanded alone. (J) In vivo expansion and (K) tumor regression following i.v. adoptive transfer of T_N-derived progeny expanded alone or with T_Mem cells, or T_Mem cells grown alone in combination with 6 Gy irradiation, i.v. rVV-hgp100, and 3 days of i.p. IL-2. n = 3 independently maintained cultures/condition or time point for experiments shown in B, C, and F–I. K was performed with n = 5 mice per group. All results shown as mean ± SEM. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by Bonferroni adjustment. *P < 0.05; **P < 0.01. Data are representative of 16 (A and B), 3 (D and E), and 2 (C, F, and I–K) independent experiments.

To globally assess the influence of T_Mem cells on the differentiation of T_N-derived progeny, we performed gene-expression analyses of reisolated cells at rest and at serial time points following activation. Hierarchical clustering demonstrated that within 18 hours, T_N cells primed with T_Mem cells had a gene-expression profile more closely related to T_Mem cells than to T_N cells (Figure 2G). We found that 139 and 259 transcripts were differentially expressed (fold change [FC] > 2, positive FDR [pFDR] < 0.05) between T_N cells primed alone or with T_Mem cells at 18 hours and 96 hours, respectively (Supplemental Tables 1 and 2). In contrast, only 0 and 10 transcripts were differentially expressed by these criteria in T_Mem cells expanded alone or with T_N cells at these time points (Supplemental Table 3), indicating that the predominant outcome of this T cell–T cell interaction was T_Mem cell–induced changes in T_N cells. Examination of specific genes revealed that key naive and memory-associated transcription factors (TFs), including transcription factor 7 (Tcf7) (27), Kruppel-like factor 2 (Klf2) (31), and forkhead box O1 (Foxo1) (32), were significantly downregulated in T_N cells primed with T_Mem cells compared with T_N cells primed alone (Figure 2H). Conversely, genes encoding effector-associated TFs and molecules, such as T-box 21 (Tbx21) (33), PR domain containing 1, with ZNF domain (Prdm1) (34), and granzyme B (Gzmb) were differentially overexpressed in T_N cells primed with T_Mem cells (Figure 2I). Collectively, these data revealed that T_Mem cells globally altered the differentiation program of T_N cells.

Finally, to assess the impact of T_Mem cell–induced differentiation of T_N-derived cells on antitumor efficacy, we adoptively transferred T_N cells primed alone or with T_Mem cells into mice bearing 10-day established s.c. B16 melanomas. We found that the expansion, persistence, and antitumor efficacy of T_N-derived cells were all significantly impaired in mice receiving T_N-derived cells primed with T_Mem cells compared with T_N-derived cells primed alone (P = 0.009) (Figure 2, J and K). We conclude that the presence of T_Mem cells during T_N cell priming ex vivo caused enhanced phenotypic, functional, and transcriptional differentiation of T_N-derived cells relative to T_N cells primed alone, resulting in impaired antitumor efficacy. Henceforth, we will refer to this process as precocious differentiation.

T_Mem cells cause precocious differentiation of T_N cells in vivo.

Having established that T_Mem augment naive cell differentiation during ex vivo expansion, we next asked whether T_Mem cells also influence T_N cell differentiation and antitumor efficacy in vivo after adoptive cotransfer. In prior studies, naive cell expansion and differentiation were impaired when limited numbers (5 × 10²) of T_N cells were transferred into hosts containing a 100- to 1,000-fold greater frequency of T_Mem cells (35). Under these conditions, T_N cells are numerically outcompeted by T_Mem cells for limited antigen presented by professional antigen-presenting cells, thereby limiting their priming and differentiation (36). However, whether T_N cell differentiation and antitumor efficacy is influenced when large numbers of cells typically used in adoptive immunotherapies are cotransferred in conjunction with T_Mem cells has not been previously evaluated. Therefore, we adoptively transferred (1 × 10⁵) Thy1.1⁺ T_N pmel-1 cells alone or in combination with (3 × 10⁵) FACS-sorted, vaccine-induced, Ly5.1⁺ pmel-1 T_Mem cells into sublethally irradiated Ly5.2⁺ hosts bearing 10-day established B16 melanoma tumors (Figure 3A). In addition to cells, recipient mice also received rVV-gp100 vaccination and exogenous IL-2. We found that both the expansion and persistence of T_N cell–derived progeny were significantly impaired when cotransferred with T_Mem cells (Figure 3B). In contrast, the recall response of cotransferred T_Mem cells was not affected compared with T_Mem cells transferred alone (Supplemental Figure 5). These findings were not due to a failure of T_N cells to become primed in vivo, as a significantly larger proportion of the T_N-derived population cotransferred with T_Mem cells lost CD27 expression and acquired the senescent marker killer cell lectin-like receptor subfamily G1 (KLRG1) at the conclusion of the immune response compared with T_N cells transferred alone (Figure 3C). Importantly, both tumor regression and animal survival were significantly impaired (P < 0.05) in mice receiving T_N cells and T_Mem cells together compared with T_N cells alone (Figure 3D). We conclude that T_Mem cells can cause precocious differentiation and impaired antitumor efficacy of T_N cell–derived progeny in vivo following adoptive cotransfer.

Figure 3. T_Mem cells cause precocious differentiation of T_N cells in vivo.

(A) Experimental schema showing the generation, isolation, and transfer of Thy1.1⁺ pmel-1 T_N cells (CD44^loCD62L⁺) alone or in combination with FACS-sorted, vaccine-induced Ly5.1⁺ pmel-1 T_Mem cells (CD44^hi) into Ly5.2⁺ hosts. (B) In vivo expansion and persistence of 1 × 10⁵ adoptively transferred Thy1.1⁺ T_N cells injected alone or in combination with 3 × 10⁵ Ly5.1⁺ T_Mem cells into Ly5.2⁺ WT mice bearing 10-day established B16 melanomas. (C) FACS analysis on day 17 of CD27 and KLRG1 expression on CD8⁺Thy1.1⁺ T_N-derived or CD8⁺Ly5.1⁺ T_Mem-derived cells. (D) Tumor regression and survival of mice bearing 10-day established B16 melanoma tumors who received 1 × 10⁵ T_N cells alone, in combination with 3 × 10⁵ T_Mem cells, or 3 × 10⁵ T_Mem cells alone. All treated mice received 6 Gy irradiation, i.v. rVV-gp100, and 3 days of i.p. IL-2. n = 3 mice/group/time point (B and C) or n = 5 mice per group (D). Results are displayed as mean ± SEM with statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment or log-rank test for animal survival. *P < 0.05; **P < 0.01. Data shown are representative of 2 independently performed experiments.

FasL-Fas interactions mediate precocious differentiation.

We next sought to elucidate what T_Mem cell factor or factors caused precocious differentiation. We determined the phenomenon was, in addition to being cell-dose dependent, also cell-contact dependent and activation dependent and could not be reproduced using supernatant transfers from restimulated T_Mem cells (Supplemental Figure 6). Accordingly, we hypothesized that an activation-induced, cell-surface molecule on T_Mem cells that can mediate costimulatory-like effects was responsible. Several members of the TNF superfamily satisfy these criteria (37). Therefore, we interrogated our microarray analysis comparing gene expression in T_Mem cells and T_N cells for TNF superfamily members uniquely overexpressed in activated T_Mem cells to generate a list of candidate ligands. Among the 19 known TNF superfamily ligands (37), only Fasl met criteria (log₂ FC > 2, P < 0.01) for being differentially expressed in restimulated T_Mem cells (Figure 4A and Supplemental Table 4). To determine whether T_CM and T_EM cells are similarly poised to express FasL following activation, we performed ChIP-seq analysis to assess histone H3 methylation dynamics at the Fasl locus within resting T_N, T_CM, and T_EM cell subsets. Consistent with previous reports (38), we found that T_N cells exhibited strong deposition of the repressive epigenetic modification trimethylation of histone 3 at lysine 27 (H3K27me3) and minimal deposition of the activating H3K4me3 mark (Figure 4B). In contrast, both T_CM and T_EM cells acquired permissive H3K4me3 marks and lost all detectable H3K27me3 repressive marks. Correlated with these changes, mRNA expression of Fasl was significantly increased in contemporaneously evaluated T_CM and T_EM cells compared with T_N cells (Figure 4C). Notably, there were no significant differences in Fasl gene expression between T_CM and T_EM cell subsets. We confirmed that T_Mem cells but not T_N cells are poised to express surface FasL at a protein level by performing FACS analysis for this molecule and its corresponding receptor, Fas, on both cell types at rest and 18 hours after activation (Figure 4, D and E).

Figure 4. FasL is poised for rapid surface expression in T_Mem cell subsets but not T_N cells.

(A) Volcano plot indicating differentially expressed genes in T_Mem cells versus T_N cells 18 hours after stimulation with CD3/CD28-specific antibodies and IL-2. Dashed lines, P < 0.01 and FC > 2. (B) Pattern of activating H3K4me3 and repressive H3K27me3 epigenetic marks within the promoter and gene body of Fasl and (C) RMA-normalized expression intensity of Fasl in resting FACS-sorted T_N, T_CM, and T_EM cell subsets. (D and E) Fas and FasL surface expression on T_N and T_Mem cells at rest or 18 hours after stimulation with CD3/CD28-specific antibodies. All bar graphs shown as mean ± SEM with n = 3 per indicated cell type or condition. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment. ***P < 0.001. Data shown in C–E are representative of 2 independently performed experiments.

FasL-Fas interactions can induce apoptosis through Fas-mediated activation of caspase-8 (39). However, Fas signaling can also perform nonapoptotic functions in a variety of tissues, including promotion of T cell costimulation (40), hepatocyte regeneration (41), tumor growth and invasiveness (42), and neuronal differentiation (43). To determine whether FasL can mediate precocious differentiation, we primed Ly5.1⁺ T_N cells alone or in a 1:1 mixture with Ly5.2⁺ T_Mem cells in the presence of a blocking antibody against FasL (αFasL) or isotype control (IgG) and measured changes in T cell differentiation. Provision of αFasL during T_N cell priming with T_Mem cells significantly but incompletely limited precocious differentiation (Figure 5, A–C). Consistent with FACS data demonstrating a paucity of FasL expression on recently activated T_N cells, we found no significant differences in the composition of T_N cell–derived progeny when isolated naive cells were primed with αFasL compared with IgG (Supplemental Figure 7). As antibody blockade may be incomplete, we complimented these findings by genetic means using CD8⁺ T_N cells derived from Fas-deficient lpr mice (44). Whereas WT T_N cells underwent augmented differentiation when primed with T_Mem cells, T_N cells from lpr mice were completely protected from this phenomenon (Figure 5, D and E). We conclude that precocious differentiation results from a contact-dependent interaction in which T_Mem cell–provided FasL engages Fas expressed on T_N cells.

Figure 5. FasL-Fas interactions mediate precocious differentiation.

(A) Representative FACS plots, (B) summary bar graph, and (C) scatter plot demonstrating T cell subset frequencies or percentage of IFN-γ⁺CD8⁺ T cells 6 days following priming of Ly5.1⁺ T_N alone or in a 1:1 mixture with Ly5.2⁺ T_Mem with CD3/CD28-specific antibodies, IL-2, and a blocking antibody against FasL (αFasL) or isotype (IgG) control. (D) FACS analysis and (E) bar graph summarizing the distribution of CD8⁺ T cell subsets 6 days following priming of Ly5.2⁺ WT T_N pmel-1 (T_N^WT/WT) or lpr/lpr T_N pmel-1 (T_N^lpr/lpr) cells alone or in the presence of a 1:1 mixture with WT Ly5.1⁺ T_Mem. All results shown as mean ± SEM with n = 2–3 per indicated condition or cell type. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment. *P < 0.05; **P < 0.01; ***P < 0.001. Data shown are representative of 14 (A and B), 4 (C), and 3 (D and E) independently performed experiments.

Nonapoptotic Fas signaling mediates precocious differentiation.

To evaluate whether FasL-delivered signals were sufficient to induce precocious differentiation, we primed T_N cells in the absence of T_Mem cells with titrated amounts of recombinant FasL oligomerized through a leucine zipper domain (lz-FasL) (45). We found that lz-FasL caused a dose-dependent increase in the differentiation of T_N cell–derived progeny, as evidenced by an increased frequency of T_EM cells and enhanced IFN-γ release upon restimulation (Figure 6, A and B). Moreover, we observed commensurate changes in the expression of key differentiation-associated genes, including Sell, Il7ra, Klf2, transferrin receptor (Tfrc), and Gzmb, as the concentration of lz-FasL was titrated up (Supplemental Figure 8). Consistent with the apoptosis-inducing function of FasL, exposure to lz-FasL resulted in reduced cell yields, particularly at concentrations greater than 33 ng/ml (Figure 6C). Despite this effect, we measured a significant increase in the relative numbers of T_EM cell phenotype/IFN-γ⁺ cells following lz-FasL exposure at 33 ng/ml compared with vehicle control (Figure 6D). Importantly, the increase in absolute numbers of T_EM cells occurred despite this subset’s enhanced susceptibly to lz-FasL–induced apoptosis relative to other subsets (Supplemental Figure 9). The differential death of T_EM cells was not attributable to differences in surface Fas expression between memory subsets (Supplemental Figure 10), a finding consistent with published studies of Fas expression on human T_Mem cell subsets (45). Collectively, these data suggested that skewing toward the T_EM cell phenotype could not be accounted for solely by selective culling of T_SCM and T_CM cells.

Figure 6. Precocious differentiation is mediated by nonapoptotic Fas signaling.

(A) Representative FACS plots and (B) bar graph summarizing the distribution of T cell subsets and IFN-γ production in isolated T_N cells primed with CD3/CD28-specific antibodies, IL-2, and indicated doses of lz-FasL for 6 days prior to analysis. Data shown are based on n = 3 independently maintained cultures/condition. (C) Relative overall cell yield and (D) relative cell yield of T_EM or IFN-γ⁺ cells, normalized to no lz-FasL treatment, of T_N-derived cells grown in the presence of titrated amounts of lz-FasL (C) or 33 ng/ml lz-FasL (D); data shown in C and D are based on n = 3–6 and n = 3 independently maintained cultures/condition, respectively. (E) Specific cell death of activated CD8⁺ T cells derived from WT (WT/WT), Lpr (lpr/lpr), and FasC194V^lpr/lpr mice following exposure to titrated amounts of lz-FasL or a vehicle control; n = 11 per condition/cell type. (F) Representative FACS plots demonstrating the frequency of T_N-derived cell subsets and (G) bar graph demonstrating fold increase in T_EM phenotype cells after WT/WT, lpr/lpr, or FasC194V^lpr/lpr CD8⁺ T cells were expanded with CD3/CD28-specific antibodies and IL-2 alone or with 40 ng/ml of lz-FasL for 6 days; n = 6–9 independently maintained cultures/cell type. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment. *P < 0.05; **P < 0.01. Results are representative of 6 (A), 2 (C–E), and 4 (F and G) independently performed experiments and are displayed as mean ± SEM.

We next sought to determine whether Fas-mediated cellular differentiation and cell death could be uncoupled using a genetic approach. Previous work established that Fas-induced apoptosis requires posttranslational palmitoylation of a cysteine residue located within the proximal cytoplasmic region of the receptor (46, 47). In cell lines, transfection of a Fas construct substituting a valine for a cysteine residue at the 194 position (C194V) prevented both Fas incorporation into membrane rafts and Fas-dependent apoptosis (46). Based on these data, a transgenic mouse expressing the FasC194V variant was generated, backcrossed to the lpr background (henceforth, FasC194V^lpr/lpr), and used to derive CD8⁺ T cells for further analyses. Whereas lz-FasL caused dose-dependent death of WT CD8⁺ T cells, FasC194V^lpr/lpr CD8⁺ T cells were significantly protected from apoptosis induction similarly to lpr CD8⁺ T cells (Figure 6E). However, unlike cells from the lpr mouse, which were also protected from lz-FasL–mediated precocious differentiation, FasC194V^lpr/lpr CD8⁺ T cells underwent enhanced conversion to a T_EM cell phenotype similar to that of WT cells following priming with 40 ng/ml lz-FasL (Figure 6, F and G). We conclude that a nonapoptotic Fas signal delivered by FasL is sufficient to induce the augmented differentiation of recently activated T_N cells.

Precocious differentiation is associated with induction of Akt signaling.

We next sought to determine what signals downstream of Fas contribute to precocious differentiation. Multiple signal transduction pathways have been implicated in nonapoptotic Fas signaling (39), including the phosphatidyl 3-kinase/protein kinase B (also known as Akt) pathway (42, 43). Given the established role of Akt in CD8⁺ T cell effector differentiation (48), we focused our attention on the induction of phospho-Akt (pAkt) and phospho-ribosomal protein S6 (pS6), a kinase downstream of Akt signaling in T cells (23). Compared with resting T_N cells, naive cells activated for 24 hours with CD3/CD28-specific antibodies exhibited augmented staining with phospho-specific antibodies against both the T308 and S473 activation residues on Akt as well as pS6 (Figure 7, A–C). Importantly, when Ly5.2⁺ T_N cells were primed in the presence of congenically distinguishable Ly5.1⁺ T_Mem cells, we observed a significant increase in the staining intensity of pAkt T308, pAkt S473, and pS6 in T_N-gated cells. To establish whether increased pAkt content in T_N cells primed with T_Mem cells leads to commensurate changes in gene expression, we returned to our microarray analyses. Within 18 hours of T_N cell activation in the presence of T_Mem cells, we observed significant alterations in the expression of known targets of Akt signaling, including Il7r (49), sphingosine-1-phosphate receptor 1 (S1pr1) (50), Tfrc (51), hexokinase-2 (Hk2) (52), and solute carrier family 2 (Slc2a1, also known as Glut1) (ref. 53 and Figure 7D). Based on these data, we next tested to determine whether provision of lz-FasL to activated T_N cells was sufficient to augment pAkt in the absence of T_Mem cells. Whereas pAkt T308 was undetectable in resting CD8⁺ T_N cells, we found that lz-FasL caused dose-dependent augmentation of this phosphorylated residue 24 hours following activation with CD3/CD28-specific antibodies (Figure 7E). Consistent with the augmented expression of Slc2a1 and Hk2 in T_N cells primed with T_Mem cells combined with the capacity of Akt activation to enhance Glut1 surface expression (53), we measured increased accumulation of the fluorescent d-glucose analog 2-(N-[7-nitrobenz-2-oxa-1,3-diazol-4-yl]amino)-2-deoxyglucose (2-NBDG) in T_N cells activated with lz-FasL (Figure 7, F and G). This indicated a relatively enhanced conversion to a glycolytic metabolism in T_N cells exposed to lz-FasL (54). Finally, to determine whether Akt signaling directly contributed to precocious differentiation, we primed CD8⁺ T_N cells in the presence of 50 ng/ml of lz-FasL with or without Akt inhibitor VIII (AktI), a well-validated allosteric Akt inhibitor (48, 55). We found that AktI significantly limited lz-FasL–mediated precocious differentiation, as evidenced by a reduced accumulation of T_EM cell phenotype and IFN-γ⁺ cells (Figure 7, H–J). We conclude that T_Mem cell and lz-FasL–induced precocious differentiation of T_N cells is associated with augmented Akt pathway activation. Pharmacologic inhibition of Akt can partially block the precocious differentiation phenotype.

Figure 7. Precocious differentiation is associated with augmented Akt signaling.

Representative FACS histograms and summary bar graphs demonstrating the mean fluorescence intensity of (A) pAkt T308, (B) pAkt S473, and (C) pS6 in Ly5.2⁺CD8⁺ T_N cells at rest or 24 hours after stimulation with CD3/CD28-specific antibodies alone or in a 1:1 mixture with Ly5.1⁺ T_Mem cells. Data shown after gating on live⁺Ly5.2⁺CD8⁺ cells. (D) RMA-normalized intensity showing expression of Il7ra, S1pr1, Tfrc, Hk2, and Slc2a1 in resting T_N cells, T_N cells primed alone, and T_N cells primed in a 1:1 mixture with T_Mem cells. (E) Western blot and densitometry analysis of pAkt T308 and GAPDH in resting T_N or T_N cells primed for 24 hours with CD3/CD28-specific antibodies and titrated doses of lz-FasL. (F) Representative FACS plot and (G) summary scatter plot showing 2-NBDG expression in T_N cells primed alone for 6 days with CD3/CD28-specific antibodies, IL-2, and lz-FasL (50 ng/ml) or vehicle control. Data shown after gating on live⁺CD8⁺ lymphocytes. (H) Representative FACS plots, (I) summary bar graph demonstrating the frequency of CD8⁺ T cell subsets, and (J) scatter plots showing IFN-γ production in T_N cells primed with CD3/CD28-specific antibodies, IL-2, and lz-FasL (50 ng/ml) in the presence or absence of an inhibitor of Akt1/2 (AktI) for 6 days. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment. *P < 0.05; **P < 0.01; ***P < 0.001. All data shown are displayed as mean ± SEM. Experiments performed with n = 3 per condition/time point in A–D and F–J. All data shown are representative of 2 independently conducted experiments.

Human T_Mem cells induce precocious differentiation of T_N-derived cells.

To evaluate the relevance of precocious differentiation for human ACT trials, we next tested to determine whether human T_Mem cells influence the differentiation of T_N cells during ex vivo cell expansion. First, we established the frequency of T_Mem cells compared with T_N cells in the circulation of healthy donors (HD) or patients with metastatic melanoma (Mel) and diffuse large B cell lymphoma (DLBCL), 2 malignancies where ex vivo–expanded adoptively transferred T cells have entered clinical trials (56–58). We found that the T_Mem cell to T_N cell ratio was 1 or greater in the majority of cases, with a median value of 1.5, 2.3, and 18.8 in HD, Mel patients, and DLBCL patients, respectively (Figure 8A). The increased ratio in DLBCL patients was likely due to the influence of prior lymphodepleting chemotherapy (30). To track the fate of human T_N cell–derived progeny in mixed cultures, we isolated T_N (CD8⁺CD45RA⁺CD45RO^–CCR7⁺) and T_Mem (CD8⁺CD45RA^–CD45RO⁺) cells from the same donor and labeled these subsets with alternative fluorescent membrane dyes (Figure 8B and Supplemental Figure 11A). Labeled T cells were expanded alone or together in titrated ratios using CD3/CD28-specific antibodies and IL-2. Although this approach did not allow for indelible cell tracking, we found we could reliably distinguish each population for up to 6 days.

Figure 8. Human T_Mem cells induce precocious differentiation of T_N-derived cells.

(A) Ratio of memory (CD3⁺CD8⁺CD45RO⁺) to naive (CD3⁺CD8⁺CD45RO^–CD45RA⁺CCR7⁺) CD8⁺ T cells in the peripheral circulation of HDs (n = 26), Mel patients (n = 31), or DLBCL patients (n = 8). Results shown as the median ± interquartile range. (B) Representative FACS plots demonstrating the surface phenotype of isolated T_N cells after membrane labeling with cell tracer eF450 but before stimulation. (C) Representative FACS plots demonstrating the gating strategy used to assess the division-normalized phenotype of human T_N-derived progeny 4 days following activation alone or in the presence of titrated ratios with T_Mem cells. Naive and T_Mem cells were labeled with cell tracer eF450 or eF670, respectively, and stimulated at indicated ratios using CD3/CD28-specific antibodies and IL-2. Naive-derived cells were subsequently analyzed for the coordinate expression of CD27, CCR7, and CD45RA using Boolean gating after gating on T_N cells that had diluted an equivalent amount of the cell tracer eF450 dye. (D) Summary graph demonstrating the frequency of T_N cells that coordinately express the markers CD27⁺CCR7⁺CD45RA⁺ after undergoing a normalized number of cell divisions plotted as a function of the ratio of T_Mem to T_N cells. Results shown as mean ± SEM for each condition for n = 3 independently evaluated donors. P = 0.0015 (repeated measures 1-way ANOVA).

Four days following activation, T_N cells and T_Mem cells lost membrane dye intensity, indicating both subsets had undergone cell division. To determine whether human T_N cells expand at the same rate in the presence or absence of memory cells, we calculated the proliferation index (PI) of T_N-derived cells primed alone or with different ratios of T_Mem cells. We discovered that human T_Mem cells caused a dose-dependent increase in T_N cell proliferation (Supplemental Figure 11B). To ascertain whether T_Mem cells influence T_N cell differentiation independently of cell proliferation, we gated on T_N-derived cells that had undergone an equivalent number of cell divisions (Figure 8C). We then evaluated the coordinate expression of naive-associated phenotypic markers (CD27, CCR7, CD45RA) by flow cytometry using a Boolean gating strategy. We found across all donors tested that, when normalized for division history, there was a significant dose-dependent loss of T_N cells (CD27⁺CCR7⁺CD45RA⁺) as the ratio of T_Mem cells was increased (P = 0.0015) (Figure 8D). Similar conclusions were reached regardless of the generation number evaluated (data not shown). We conclude that human T_Mem cells augmented both the proliferation and differentiation of T_N-derived cells. Further, we conclude that the rate of differentiation is enhanced even when division history is normalized.

In mice, we discovered that T_Mem cell–induced precocious differentiation of T_N cells was mediated by FasL. To determine whether enhanced Fas signaling augmented the differentiation of human CD8⁺ T cells, we primed peripheral blood CD8⁺ T cells from HDs with CD3/CD28-specific antibodies and IL-2 alone or in the presence of titrated concentrations of lz-FasL. After 9 to 10 days, we found that all expanded cells had assumed a CD45RA^lo/–CD45RO⁺ memory phenotype (data not shown). Using coexpression of CD27 and CCR7 (29), we characterized the ratio of T_CM (CD8⁺CD45RO⁺CCR7⁺CD27⁺) to T_EM cells (CD8⁺CD45RO⁺CCR7^–CD27^–). Consistent with experiments using mouse T cells, we found that augmenting Fas signaling using lz-FasL in HD cells caused a significant and dose-dependent accumulation of the more-differentiated T_EM cell subset and an attrition of T_CM cells (Figure 9, A and B). These changes were not attributable to differences in proliferation, as T cells expanded in lz-FasL and vehicle control diluted a membrane-associated dye similarly (Supplemental Figure 12A). Moreover, analysis of the surface phenotype of T cells that had undergone an identical number of divisions revealed that lz-FasL exposure resulted in a more-differentiated phenotype compared with cells expanded in vehicle control (Supplemental Figure 12B). Consistent with the apoptosis-inducing functions of Fas, lz-FasL resulted in a reduced cell yield at concentrations of 20 ng/ml or more (Supplemental Figure 12C). However, at lower concentrations of lz-FasL, we measured differences in the distribution of T cell subsets while expanding similar numbers of cells. To exclude that the observed changes in human CD8⁺ T cell subset composition were due to selective T_CM cell killing, we evaluated the induction of apoptosis (as measured by coordinate 7-AAD and annexin V staining) in T_CM and T_EM cell subsets following stimulation with titrated amounts of lz-FasL (Figure 9, C and D). Similar to both our mouse data and previously reported data in human CD4⁺ T cells (45), we found that T_EM cells were far more sensitive to Fas-induced apoptosis compared with the T_CM cell subset. Collectively, we conclude that human T_N cells, like mouse T_N cells, undergo accelerated differentiation when primed in the presence of T_Mem cells or augmented Fas signaling.

Figure 9. Fas signaling promotes the differentiation of human CD8⁺ T cells.

(A) Representative FACS plots and (B) summary graph showing the distribution of CD8⁺ T cell subsets 7 to 10 days after peripheral blood CD8⁺ T cells from HD were stimulated with CD3/CD28-specific antibodies, IL-2, and indicated concentrations of lz-FasL. Results shown after gating on viable CD8⁺ lymphocytes from n = 7 HD. T_CM = CD8⁺CD45RO⁺CCR7⁺CD27⁺, T_EM = CD8⁺CD45RO⁺CCR7^–CD27^–. (C) Representative FACS plots and (D) summary graph displaying the induction of lz-FasL specific cell death after gating on T_CM and T_EM cells stimulated with indicated concentrations of lz-FasL for 12 hours. Cells that coordinately stained for 7-AAD and annexin V were considered apoptotic. Statistical comparisons performed using an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment. *P < 0.05; **P < 0.01. Data shown are representative of 7 (A and B) and 2 (C and D) independently performed experiments with results displayed as mean ± SEM.

Fas signaling controls T cell differentiation and antitumor efficacy.

T_Mem–induced precocious differentiation of T_N-derived cells resulted in impaired antitumor efficacy in mice. Given our findings that precocious differentiation is mediated by FasL-Fas interaction, we next tested to determine whether modulation of Fas signaling could influence the therapeutic potential of adoptively transferred T_N-derived cells. We therefore primed pmel-1 T_N cells in the presence of congenically distinguishable T_Mem cells in combination with αFasL or IgG control. In addition, we also primed T_N cells alone with lz-FasL or vehicle control. Following expansion, equal numbers of viable CD8⁺ T cells obtained using a density separation media for each culture condition were adoptively transferred into sublethally irradiated mice bearing 10-day established B16 melanomas in combination with rVV-gp100 and IL-2. We found that FasL blockade during T_N cell expansion in the presence of T_Mem cells significantly improved tumor regression (P = 0.009) and animal survival (P < 0.05), nearly rescuing the in vivo antitumor potential of isolated T_N cells (Figure 10, A and B). In contrast, FasL blockade had no impact on the antitumor efficacy of isolated T_N cells expanded without T_Mem cells (Supplemental Figure 13). Conversely, augmenting Fas signaling using lz-FasL during T_N cell priming significantly impaired tumor regression (P < 0.01) and animal survival (P < 0.05). The differences in treatment efficacy of lz-FasL primed T_N cells were not attributable to impaired cell engraftment, as a similar frequency of cells was recovered 18 hours after transfer (Figure 10C). Overall, we found a significant inverse linear correlation between the state of T cell differentiation, as measured by CD62L expression on transferred T cells primed with FasL blockade or augmentation, and tumor growth (R² = 0.797, P < 0.0001; Figure 10D). These data suggest that FasL-mediated perturbations in T cell differentiation were highly correlated with the antitumor treatment efficacy of transferred CD8⁺ T cells. We conclude that the negative influence of T_Mem cell–induced precocious differentiation on the treatment efficacy of adoptively transferred T_N-derived cells in mice is mediated by a FasL-Fas interaction that can be rescued with FasL blockade during cell expansion.

Figure 10. Fas signaling controls T cell differentiation and influences adoptive immunotherapy efficacy.

(A) Tumor regression and (B) animal survival of mice bearing 10-day established s.c. B16 melanomas treated with 2.5 × 10⁵ T_N-derived pmel-1 cells primed alone, with lz-FasL (50 ng/ml), or with T_Mem in the presence of αFasL or IgG control. Viable cells were isolated using a density separation media before infusion. All treated mice received 6 Gy irradiation prior to cell infusion in addition to i.v. rVV-gp100 and 3 days of i.p. IL-2. Tumor treatment experiments performed with n = 5 mice/group. T_Mix, T_N cells primed with T_Mem cells in a 1:1 mixture. (C) Engraftment efficiency of Thy1.1⁺ T_N-derived pmel-1 cells (3 × 10⁵) primed alone or with lz-FasL 18 hours following i.v. adoptive transfer into nonirradiated Ly5.2⁺ hosts; n = 5 mice/group. (D) Correlation between the slope of tumor growth and CD62L expression at the time of cell transfer on T_N-derived pmel-1 cells primed alone with either lz-FasL (50 ng/ml), vehicle control, or with T_Mem and either αFasL or IgG. Pooled results from 2 independently performed experiments displayed using n = 4–10 mice per condition. Statistical comparisons performed using an unpaired 2-tailed Student’s t test or log-rank test for animal survival. *P < 0.05. Data shown are representative of 2 independent experiments with results displayed as mean ± SEM.

Discussion

Herein, we present evidence using both human and mouse cells that antigen-experienced CD8⁺ T cells directly influence naive cell differentiation both during cell expansion ex vivo and following adoptive cotransfer in vivo. This process, which we have termed precocious differentiation, serves to synchronize the functional, transcriptional, and metabolic state of T_N-derived progeny with that of T_Mem cells. Consequently, precocious differentiation leads to an attrition of the highly potent T_SCM and T_CM populations and an accumulation of T_EM cells, resulting in the impaired proliferation, persistence, and antitumor treatment efficacy of adoptively transferred T cells.

In multiple preclinical models, minimally differentiated T_N, T_SCM, and T_CM cell subsets exhibit superior persistence (12, 18, 27, 59–62) and enhanced antitumor (18, 27, 59), antibacterial (62), and antiviral responses following ACT compared with the highly differentiated T_EM cell and effector T cell (T_Eff) populations. Despite these data, the vast majority of current ACT trials utilize unfractionated T cell populations (2, 20). Largely, this is based on an assumption that having at least some representation of the less-differentiated subsets in the starting population used to generate therapeutic T cells is sufficient to convey their desirable attributes. Our data directly challenge this notion by demonstrating that T_N cells must either be isolated from T_Mem cells prior to cell expansion or have their capacity to respond to T_Mem cell–mediated signaling disrupted to preserve their full therapeutic potential. Accomplishing these goals at a clinical scale might be done using several strategies. These include isolating defined T cell subsets from peripheral blood using serial positive enrichments with reversible Fab streptamers (63), employing a combination of negative and positive magnetic bead isolations (64), and antagonizing Fas or Akt signaling (55). Based on the data in this manuscript, we have initiated a human clinical trial registered at ClinicalTrials.gov (NCT02062359) in which T_N cells are enriched relative to T_Mem cells using magnetic bead isolation of CD62L-expressing cells prior to gene engineering with a TCR recognizing the cancer-germline antigen NY-ESO-1. Results from this and other planned clinical trials will determine whether ACT of naive-enriched T cells results in superior clinical outcomes compared with transfer of unselected T cell populations.

Mechanistically, we identified the TNF superfamily member FasL, which is expressed on T_Mem cells but not T_N cells acutely following activation, as the mediator of precocious differentiation. FasL was both necessary and sufficient to induce the phenomenon. Provision of lz-FasL recapitulated precocious differentiation in the absence of T_Mem cells, while antibody blockade of FasL or use of T cells from Fas-deficient lpr mice prevented precocious differentiation. Reverse signaling by FasL in T cells has been observed under some conditions (65). Our findings that T cells derived from lpr mice or were protected from precocious differentiation combined with the ability of lz-FasL to mediate the phenomenon excluded reverse signaling as a mechanism in our system.

While Fas is conventionally known as a death receptor, its ability to mediate nonapoptotic functions has been documented in a variety of tissues (39), including T cells (40). We provide several lines of evidence that our findings were not attributable solely to Fas-induced apoptosis. First, we determined that T_EM cells, the population that accumulates in response to augmented Fas signaling, was most sensitive to Fas-mediated apoptosis relative to other T cell subsets in both mice and humans. This finding is in agreement with previously reported results (45). Thus, selective culling of T_SCM and T_CM cells could not account for the skewing toward the T_EM cell population. Second, using CD8⁺ T cells derived from a transgenic mouse expressing a FasC194V mutant receptor, we provide genetic evidence that Fas-mediated cell death and cellular differentiation can be uncoupled by impairing the ability of the receptor to undergo posttranslational lipid modification. Although further studies are required to resolve the precise signaling cascade necessary for Fas-induced precocious differentiation, we found that both T_Mem cells and lz-FasL augmented Akt signaling in T_N cells. Similarly to previously reported Akt-dependent functions of canonical costimulatory molecules such as CD28 (66), augmenting Fas signaling in T_N cells enhanced cellular differentiation and promoted a glycolytic metabolism. Conversely, inhibition of Akt using a well-validated allosteric inhibitor largely prevented lz-FasL–induced differentiation of T_N cells, further implicating Akt activation in Fas-mediated differentiation. These findings parallel recent reports of Akt-mediated nonapoptotic Fas signaling in glioblastoma (42) and neuronal stem cells (43).

We determined that the capacity to induce precocious differentiation was a generalized property of antigen-experienced T cells, regardless of whether they were T_CM or T_EM cells. These findings were associated with a permissive epigenetic signature at the FasL locus in both T_CM and T_EM cell subsets relative to T_N cells. Further, we resolved that precocious differentiation required both cell activation and direct cell contact between T cell subsets to occur. Separation of T_N and T_Mem cells by a semipermeable membrane prevented the phenomenon and transfer of T_Mem cell–derived supernatant to T_N cells could not reproduce the effect. Recent multiphoton microscopy studies have revealed that T_N and T_CM cells form stable clusters together in vivo around antigen-bearing targets (67, 68). Additionally, while lymphoid-trafficking of the CD62L^– T_EM/T_Eff cell subsets is limited in the steady state, inflamed lymph nodes become permissive to entry of these cells (69). Thus, there is evidence that T_N and T_Mem cells can physically interact with one another in vivo. Whether such T_N and T_Mem cell interactions might also influence the response to intracellular pathogens or vaccines under certain conditions is the subject of future work.

A prior study evaluated the influence of preexisting T cell memory on the recruitment and differentiation of T_N cells in vivo (35). Unlike current ACT clinical protocols in which large numbers of T cells are infused in an attempt to gain control over tumor replication, this study evaluated the transfer of a limiting number of T_N cells (500 cells) in hosts containing 100- to 1,000-fold greater numbers of T_Mem cells. Under these conditions, T_Mem cells numerically outcompeted T_N cells for access to limited antigen and cytokines, thus curtailing the differentiation of T_N cells. However, how T_Mem cells influence T_N cell differentiation when cells are coinfused in the relatively large numbers and ratios typically used in adoptive immunotherapies has previously not been addressed. Consistent with our ex vivo cell expansion data, we found that cotransfer of T_Mem cells caused accelerated T_N cell differentiation into a terminally differentiated KLRG1⁺CD27^lo phenotype despite an attenuated proliferative burst. The decreased expansion of antitumor T_N cells primed with T_Mem cells in vivo likely reflected their accelerated conversion to antigen-experienced subsets that are known to have a reduced proliferative capacity relative to T_N cells following poxvirus vaccination (70). Most importantly, cotransfer of T_Mem cells impaired the full in vivo antitumor efficacy of T_N cells.

The magnitude of accelerated T_N cell differentiation caused by T_Mem cells was remarkably sensitive to the ratio of the 2 subsets such that an exponential relationship existed between the extent to which T_N-derived progeny entered the T_EM/T_EFF pool and the starting proportion of T_Mem to T_N cells. In this manner, more-differentiated T cell subsets have the capacity to synchronize their phenotype, function, and gene expression with less-differentiated subsets. Some unicellular organisms exhibit collective decision making through cell-cell communication once a threshold concentration of members detect an environmental stress (71). This process, termed quorum sensing, allows individual members of a community to exhibit collectivist behaviors in order to coordinate expression of energetically expensive processes involved in virulence and differentiation (71). While quorum sensing has been theorized to be relevant to understanding lymphocyte behavior (72, 73), our data add to growing experimental data demonstrating that, under certain circumstances, T cells can also exhibit collectivist behaviors (74, 75).

In conclusion, we describe what we believe to be a previously unrecognized T cell–T cell interaction whereby antigen-experienced CD8⁺ T cells drive T_N cell differentiation during priming through a contact-dependent mechanism involving a nonapoptotic Fas-FasL interaction. The net influence of this process synchronizes T_N cell behavior with that of T_Mem cells, affecting the functional, transcriptional, and metabolic differentiation of T_N-derived progeny. For adoptive immunotherapies, where there is a strong inverse correlation between T cell differentiation and antitumor efficacy, the implications of our findings are clear: T_Mem cause significantly enhanced differentiation of T_N cells, impairing in vivo antitumor efficacy. Therefore, strategies that disrupt quorum-like behavior among T cell subsets might provide a new means of enhancing the effectiveness of T cell–based immunotherapies.

Methods

Mouse strains and animal studies.

Adult (6 to 12 weeks old) female C57BL/6 (B6; Ly5.2⁺), B6.SJL-Ptprc^a Pep3^b/BoyJ (Ly5.1⁺), B6.PL-Thy1^a/CyJ (Thy1.1⁺), B6.MRL-Fas^lpr/J (lpr) (44), B6.129S7-Rag1^tm1Mom/J (Rag), B6.Cg-Thy1^a/Cy Tg(TcraTcrb)8Rest/J (pmel-1) (24), and B6-Tg(TcraTcrb)1100Mjb/J (OT-1) (76) CD8⁺ TCR transgenic mice were all purchased from the Jackson Laboratory. Transgenic mice harboring the Fas C194V mutant receptor were generated via BAC containing the Fas locus, with the C194V generated by recombineering. These mice were backcrossed to lpr mice on a B6 background. FasC194V^lpr/lpr mice were backcrossed to homozygosity for both the lpr Fas allele and the C194V Fas transgene. Where indicated, pmel-1 and OT-1 mice were crossed to Thy1.1, Ly5.1, Rag, or Rag × lpr backgrounds. All mice were maintained under specific pathogen–free conditions.

Evaluation of vaccine-induced and antitumor immunity.

Adult female B6 mice were implanted by s.c. injection with 4 × 10⁵ B16 (H-2D^b) cells, a spontaneous gp100⁺ murine melanoma cell line obtained from the NCI tumor repository. Ten days later, tumor-bearing mice received 6 Gy total body irradiation. Treated mice received i.v. injection of indicated doses and subsets of pmel-1 CD8⁺ T cells in combination with 2 × 10⁷ pfu of a previously described recombinant vaccinia virus (rVV-gp100) (24) (rVV-gp100) and 12 μg IL-2 (Prometheus) administered twice daily by i.p. injection for a total of 6 doses. All tumor measurements were performed in a blinded fashion.

Antibodies and flow cytometry.

Mouse cells were stained with fluorochrome-conjugated antibodies against combinations of the following surface and intracellular antigens after Fc receptor blockade (2.4G2): CD8α (clone 53-6.7), CD27 (clone LG.3A10), CD44 (clone IM7), CD45.1(clone A20), CD45.2 (clone 104), CD62L (clone MEL-14), CD90.1 (clone OX-7), CD90.2 (clone 53-2.1), CD95 (clone Jo2), CD122 (clone TM–β1), CD178 (clone MFL3), IFN-γ (clone XMG1.2), KLRG-1 (clone 2F1), and Sca-1 (clone D7) (all purchased from BD Biosciences). Fluorochrome conjugates against pS6 S235/236 (D57.2.2E), pAkt T308 (C31E5E), and pAkt S473 (D9E) were obtained from Cell Signaling Technologies. Human cells were stained with combinations of the following fluorochrome-conjugated antibodies: CCR7 (150503), CD28 (CD28.2), CD45RO (UCHL1), CD8 (SK1) (BD Biosciences) or CD27 (M-T271), CD45RA (HI100), and CD62L (DREG-56) (BioLegend). Stimulation of T cells for intracellular cytokine staining was accomplished using Leukocyte Activation Cocktail containing phorbol myristate acetate and ionomycin in combination with brefeldin A and monensin solution (BD Biosciences). Apoptosis and specific cell death in defined T cell subsets was assessed using fluorochrome-conjugated annexin V and 7-AAD (both from BD Biosciences), as previously described (45). Where indicated, T_N-derived and T_Mem-derived subsets were reisolated for additional analyses by FACS sorting to greater than 92% purity using FACS sorting or magnetic bead isolation. Cell viability was determined using PI exclusion in FACS-sorting experiments or fixable live/dead cells (Invitrogen) for diagnostic experiments. For FACS–based glucose uptake assays, we incubated CD8⁺ T cells with 100 μM 2-NBDG (Invitrogen) for 2 hours before measuring by FACS, as previously described (54). Flow cytometric data were acquired using either BD FACSCanto II, LSR, or LSRFortessa cytometers (BD Biosciences). FACS data, including calculation of PI, were analyzed with FlowJo Version 9.7 software (TreeStar).

Isolation and generation of mouse and human CD8⁺ T cell subsets.

Mouse T_N cells were isolated either by using a MACS CD8⁺ negative selection kit (Miltenyi Biotech) in combination with a 1:300 dilution of biotin-conjugated anti-CD44 antibody or by FACS sorting CD44^loCD62L^hiCD8⁺ T cells using FACSAria (BD Biosciences). Mouse T_Mem cells were generated by in vitro differentiation (as previously described) or by adoptive transfer of congenically distinguishable Thy1.1⁺ or Ly5.1⁺ pmel-1 CD8⁺ T cells into Ly5.2⁺ WT mice, where recipient mice were vaccinated with rVV-gp100 (2 × 10⁷ PFU) and vaccine-induced T_CM (CD44^hiCD62L⁺) or T_EM (CD44^hiCD62L^–) adoptively transferred subsets were isolated by FACs sorting more than 28 days later (27, 77). Human T cells were obtained either by leukapheresis or venipuncture and prepared over Ficoll-Hypaque gradient (LSM; ICN Biomedicals Inc.). Human T_N cells and T_Mem cells were obtained by magnetic bead isolation using the EasySep Human Naive CD8⁺ T Cell and Memory CD8⁺ T Cell Enrichment Kits, respectively (STEMCELL Technologies). Fate tracking of T_N and T_Mem cells was accomplished by labeling cells with Cell Proliferation Dye eFluor 450 and Cell Proliferation Dye eFluor 670, respectively (eBioscience).

Activation and expansion of CD8⁺ T cells.

Both mouse and human T_N cells with or without T_Mem cells were activated and expanded at indicated ratios in 96-well round-bottom plates coated with 2 μg/ml of CD3-specific and 1 μg/ml of soluble CD28-specific antibodies (clones 145-2C11 and 37.51; BD Biosciences) in culture media containing 5 ng/ml (mouse) or 20 ng/ml (human) of IL-2 at a final density of 1 × 10⁵ cells/well. Where indicated, T cells were cultured with specified concentrations of lz-FasL, a previously described recombinant oligomerized form of FasL (45), or 10 μg ml^–1 of either a blocking antibody against FasL (MFL3; BD Biosciences) or IgG (A19-3; BioLegend). In some experiments, cells were also cultured with 1 μM of Akt Inhibitor VIII (Calbiochem) dissolved in DMSO (Sigma-Aldrich). Where indicated, T_N-derived and T_Mem-derived subsets were reisolated for additional analyses by FACS sorting to more than 95% purity using FACS sorting or magnetic bead isolation. Transwell experiments were conducted in 24-well plates using 0.4 μm inserts (Corning Costar) with T_N cells alone or in combination with T_Mem cells either in the bottom well or in the Transwell insert.

Cytokine release assays.

We pulsed B6 splenocytes with indicated concentrations of hgp100_25–33 peptide and incubated cells with T_N-derived pmel-1 CD8⁺ T cells at a 1:1 ratio overnight at 37°C. Supernatants were analyzed for mouse IFN-γ by ELISA (R&D Systems).

qPCR and Western blot.

For qPCR analysis, CD8⁺ T cell subsets were FACS sorted directly into RNAprotect Cell Reagent, and RNA was extracted using the QiaShredders and RNeasy Mini Kits (all from QIAGEN). cDNA was synthesized using the High Capacity RNA-to-cDNA kit (Applied Biosystems), and qPCR was performed on an ABI 7500 Fast Instrument (Applied Biosystems). Gene expression was quantified using probes targeting Ccr7, Cd27, Gzmb, Il7ra, Klf2, Sell, Tbx21, Tcf7, Tfrc, and Prdm1 (Applied Biosystems). For Western blot analysis, CD8⁺ T cells were sorted into FCS and subsequently lysed in protease inhibitor containing RIPA buffer (Cell Signaling Technology). Protein was quantified using the Bio-Rad protein assay. We separated 30 μg of total protein on a 4% to 12% SDS-PAGE gel followed by standard immunoblotting with antibodies to pAktT308 (C31E5E; Cell Signaling Technology), GAPDH (AB2302; EMD Millipore), and horseradish peroxidase–conjugated goat antibodies to mouse and rabbit IgG (sc-2005 and sc-2004; Santa Cruz Biotechnology Inc.). Blots were developed using chemiluminescence (Thermo Fisher Scientific) and acquired using the ChemiDoc system (Bio-Rad Laboratories).

Microarray analyses and ChIP-seq.

Gene-expression levels were determined using GeneChip Mouse Gene 1.0 ST arrays (Affymetrix) according to the manufacturer’s protocol. 300 ng of total RNA was used as starting material for cRNA amplification using the WT Expression Kit (Life Technologies) according to the manufacturer’s protocol. cDNA was reverse transcribed, fragmented, labeled using the GeneChip WT Terminal Labeling Kit (Affymetrix), and hybridized on the arrays for 18 hours according to the manufacturer’s directions. Arrays were stained and washed in the Fluidics Station 400 (Affymetrix) and scanned (Affymetrix 7G). Array data were imported into the Partek Genomic Suite using robust multichip analysis (RMA) normalization after background subtraction. One-way ANOVA was used to identify differentially expressed genes among the 4 T cell subtypes with a significant cut-off of pFDR < 0.01. Differentially expressed probe sets were selected using a FDR cut-off of 0.01 without specifying a FC criterion. ChIP-seq assays were performed as previously described (55). Briefly, 2 × 10⁷ T cell subsets isolated by FACS sorting were treated with MNase to generate approximately 20% dinucleosomes and 80% mononucleosomes. Antibodies against H3K4me3 (ab8580; Abcam) and H3K27me3 (07-449; Upstate) were used for immunoprecipitation. The ChIP DNA fragments were blunt ended, ligated to Solexa adaptors, and sequenced with the Illumina 1G Genome Analyzer with mapping to the mouse genome (build mm10) using Bowtie software. All original microarray data were deposited in the NCBI’s Gene Expression Omnibus (GEO GSE56344, GSE67825, and GSE67881).

Statistics.

The products of perpendicular tumor diameters were plotted as the mean ± SEM for each data point, and tumor treatment graphs were compared by using the Wilcoxon rank sum test and analysis of animal survival assessed using a log-rank test. Regression analysis of the slope of tumor regression as a function of CD62L⁺ cells among Fas-modulated CD8⁺ T cells was performed as described previously (78). For all other experiments, data were compared using either an unpaired 2-tailed Student’s t test corrected for multiple comparisons by a Bonferroni adjustment or repeated measures 1-way ANOVA, as indicated. In all cases, P values of less than 0.05 were considered significant. Statistics were calculated using Prism 5 GraphPad software (GraphPad Software Inc.).

Study approval.

Animal experiments were conducted with the approval of the NCI and NIAMS Animal Use and Care Committees. All anonymous NIH Blood Bank donors and cancer patients providing human samples were enrolled in clinical trials approved by the NIH Clinical Center and NCI institutional review boards. Each patient signed an informed consent form and received a patient information form prior to participation.

Author contributions

CAK, CDS, AJL, TNY, ACC, CO, MR, RR, YJ, RLE, MS, JGC, DCP, ZAB, DC, ZY, PM, AG, SKT, SP, HL, RMS, and NPR designed and/or performed experiments. CAK, CDS, TNY, MR, ACC, FMM, and EW analyzed data. JGC, LG, SAR, and RMS edited the manuscript. CAK and NPR wrote the manuscript.