Generation and application of human induced-stem cell memory T cells for adoptive immunotherapy.

Adoptive T-cell therapy is an effective strategy for cancer immunotherapy. However, infused T cells frequently become functionally exhausted, and consequently offer a poor prognosis after transplantation into patients. Adoptive transfer of tumor antigen-specific stem cell memory T (TSCM ) cells is expected to overcome this shortcoming as TSCM cells are close to naïve T cells, but are also highly proliferative, long-lived, and produce a large number of effector T cells in response to antigen stimulation. We previously reported that activated effector T cells can be converted into TSCM -like cells (iTSCM ) by coculturing with OP9 cells expressing Notch ligand, Delta-like 1 (OP9-hDLL1). Here we show the methodological parameters of human CD8+ iTSCM cell generation and their application to adoptive cancer immunotherapy. Regardless of the stimulation by anti-CD3/CD28 antibodies or by antigen-presenting cells, human iTSCM cells were more efficiently induced from central memory type T cells than from effector memory T cells. During the induction phase by coculture with OP9-hDLL1 cells, interleukin (IL)-7 and IL-15 (but not IL-2 or IL-21) could efficiently generate iTSCM cells. Epstein-Barr virus-specific iTSCM cells showed much stronger antitumor potentials than conventionally activated T cells in humanized Epstein-Barr virus transformed-tumor model mice. Thus, adoptive T-cell therapy with iTSCM offers a promising therapeutic strategy for cancer immunotherapy.

B‐cell lymphoma 6 protein

chimeric antigen receptor

C‐C chemokine receptor type 7

CD45 isoform RA

carboxyfluorescein succinimidyl ester

Epstein–Barr virus

interleukin

induced‐TSCM

lymphoblastoid cell line

MART‐1 peptide‐pulsed monocyte‐derived dendritic cell

melanoma antigen recognized by T cell‐1

tumor‐associated antigen

central memory T

T‐cell receptor

effector memory T

CD45RA‐positive effector memory T

tumor‐infiltrating lymphocyte

stem cell memory T

INTRODUCTION

Adoptive T‐cell therapy is a promising approach to cancer therapy.1 Tumor‐infiltrating lymphocyte infusion is a classic and effective method for treating melanoma patients.2, 3 Isolated tumor‐infiltrating lymphocytes (TILs) are restimulated by tumor‐associated antigen (TAA)‐loaded antigen‐presenting cells, and then expanded TILs are infused into patients. Gene transducing technologies have also recently been applied to adoptive T cell therapy. Artificially designed T‐cell receptor (TCR) or chimeric antigen receptor (CAR) against TAA and CAR target antigens are transduced into patients’ T cells with TCR stimulation and expansion.4 As in TIL therapy, these transduced T cells are also re‐infused into patients to attack tumor cells. However, the therapeutic effects of tumor‐specific T cell transfer have been limited because infused T cells frequently lose functionality after transplantation.1, 5 The TAA‐specific T cells become exhausted and dysfunctional in the tumor microenvironment and through the T cell culture procedures.6, 7 Therefore, restoration and enforcement of T cell function and persistency could improve long‐term outcomes for cancer patients.

Memory T cells have been classified into two subpopulations: peripheral tissue‐homing effector memory T (TEM) cells and lymphatic tissue‐homing central memory T (TCM) cells.8 Novel memory T cell subpopulations, called stem cell memory T (TSCM) cells, were recently identified as the “stem” of memory T cells. Stem cell memory T cells expressed naïve T‐cell surface markers, CD44lowCD62LhighCCR7+ in mice and CD45RA+CD45RO−CCR7+CD62L+ in humans.9, 10 These cells possess high proliferative and long‐term survival abilities in vivo, producing a large number of effector T cells with potent antitumor functions.11, 12 Similar naïve‐like memory T cells were discovered in human and called memory T cells with a naive phenotype.13 A recent paper by Ahmed et al14 indicated that human TSCM cells are highly proliferative but have long telomeres and high levels of telomerase activity.

Due to these superior features of TSCM cells, methods of generating TSCM cells in vitro for adoptive T cell therapy have been investigated. In 2009, Gattinoni et al15 first reported that a glycogen synthase kinase 3β inhibitor, TWS119, induces murine and human TSCM cells by activating Wnt signals. TWS119 treatment arrested the cell cycle during TCR stimulation and inhibited the differentiation to TCM and TEM cells. Therefore, this method is only applicable for naïve T cells, and the number of induced TSCM cells is limited. Alternative methods have been proposed by optimizing TCR strength, cytokine supplement (including interleukin [IL]‐7, IL‐15, and IL‐21), and drug treatment. These methods could generate TSCM cells in vitro, but these methods still generate TSCM cells from naïve T cells (Table 1).16, 17, 18, 19, 20 We have established a novel two‐step culture system for TSCM induction, which is constituted by a “prime” step and an “induction” step, and have named the induced‐TSCM “induced‐TSCM (iTSCM)” cells.12 In this paper, we describe a detailed methodology of iTSCM generation and report optimal conditions of priming and cytokine treatment. We also investigate the antitumor efficacy of human iTSCM cells in a humanized mouse model.

Table 1

Recently reported methods for stem cell memory T cell generation from naïve T cells

Publication	Year	TCR procedure	TCR strength	Cytokine	Signaling inhibitor
Gattinoni et al Nat Med	2011	CD3/CD28 beads	Full	IL‐2	GSK3‐β (TWS119)
Cieri et al Blood	2013	CD3/CD28 beads	Full	IL‐7 + IL‐15	None
Gomez‐Eerland et al Hum Gene Ther Methods	2014	CD3/CD28 beads	Full	IL‐7 + IL‐15	None
Sabatino et al Blood	2016	CD3/CD28 beads	Short (4 d)	IL‐7 + IL‐21	GSK3‐β (TWS119)
Scholz et al EBioMedicine	2016	CD3/CD28 beads	Full	IL‐2	mTOR (rapamycin)
Alvarez‐Fernandez et al J Trans Med	2016	CD3/CD28 beads	Short (2 d)	IL‐21	None
Hurton et al Proc Natl Acad Sci USA	2016	APC	Full	mbIL‐15	None
Kagoya et al JCI Insight	2017	CD3/CD28 beads, APC	Short (1 d)	IL‐2 + IL‐15	None
Jeza et al Pan Afr Med J	2017	APC	Full	IL‐21	mTOR (rapamycin)
Zanon et al Eur J Immunol	2017	CD3/CD28 beads, APC	Full	IL‐7 + IL‐15	None
Kaartinen et al Cytotherapy	2017	CD3/CD28 beads	Full	Low IL‐2	None

APC, antigen presenting cell; d, day; GSK3‐β, glycogen synthase kinase 3β; IL, interleukin; mb, membrane bound; TCR, T cell receptor.

MATERIALS AND METHODS

OP9‐hDLL1 cell coculture

Human T cells were activated using the methods mentioned above. To activate Notch signaling, activated T cells were cocultured with OP9‐hDLL1 cells. Human T cells and OP9‐hDLL1 cells were cocultured with human IL‐2 (20 ng/mL; PeproTech), human IL‐7 (10 ng/mL; PeproTech), human IL‐15 (20 ng/mL; Biolegend), or human IL‐21 (20 ng/mL; PeproTech) in Minimum Essential Medium Eagle‐alpha modification for 11 days.

Statistics

Statistical analysis was carried out using Student's t‐test, one‐way ANOVA and a long‐rank test, using GraphPad Prism version 6.05 software (GraphPad Software, La Jolla, CA, USA). The variance among the groups was estimated using the F‐test, and P‐values <.05 were considered statistically significant. All data are presented as the mean ± SEM. Mice were randomly assigned to experimental groups. The investigators were not blinded to allocation during experiments and outcome assessment.

Further information regarding materials and methods is included in Tables S1,S2.

RESULTS

Overview of iTSCM cell generation

Several methods for generating TSCM have recently been reported. The strength of TCR stimulation, cytokine effects, and signaling inhibitors were associated with TSCM generation from naïve T cells and are summarized in Table 1.11, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 In our two‐step culture system, T cells are first primed and expanded by strong TCR stimulation, and are then cocultured with OP9‐hDLL1 cells to convert into iTSCM cells (Figure 1A). To acquire a large number of antigen‐specific CD8α+ T cells, we first isolated peripheral human CD8+ T cells, labeled with cell trace dye and co‐cultured them with autologous Epstein–Barr virus (EBV)‐transformed lymphoblastoid cell lines (LCLs) for 7 days in the “prime” step. Next, cell trace dye‐diluted activated T cells were purified by a cell sorter and were then transferred onto and cocultured with DLL1‐expressing OP9 stromal cells (OP9‐hDLL1) for 11 days in the “induction” step (Figure 1B). More than 90% of singlet live cells expressed CD8α and more than 80% of CD8α+ T cells highly proliferated (dividing more than six times) and mainly showed CD45RA−CCR7− TEM (76.4%) and CD45RA−CCR7+ TCM (13.4%) phenotypes (Figure 2A). We then isolated phenotypical TEM and TCM cells and transferred them onto OP9‐hDLL1 layers with human IL‐7. Conversion efficiency from EBV‐specific activated human T cells to iTSCM is 10% from TEM cells and >85% from TCM cells (Figure 2B). In addition, the number of TEM‐derived CD8α+ T cells was not changed or was only slightly reduced, whereas that of TCM‐derived CD8α+ T cells was significantly increased (Figure 2C). To investigate the EBV‐specific recall response, we cocultured TEM, TCM, and iTSCM cells with autologous LCL for 60 hours. The EBV‐specific iTSCM cells recovered greater number of cells than the TEM and TCM cells did (Figure 2D). The same results could be acquired from several independent experiments using T cells from distinct healthy subjects. These data indicate that the coculture with OP9‐hDLL1 cells convert into iTSCM from both TEM and TCM cells. However, the efficiency of iTSCM generation from TCM cells is higher than that from TEM cells, and TCM cells do proliferate during iTSCM generation.

Figure 1

Experimental outline for induced stem cell memory T (iT) cell induction. A, Schematic for two‐step culture system for iT induction. Peripheral CD8α+ T cells are isolated and activated by T‐cell receptor stimulation for 7 d in the first step, called the prime step. Activated T cells are purified by cell sorting and transferred onto OP9‐hDLL1 cells for 11 d in the second step, the induction step. Evaluation of iT cells is carried out 11 d after OP9‐hDLL1 coculture. B, Detailed protocols for iT generation. Preparation of autologous lymphoblastoid cell lines (LCLs) and OP9‐hDLL1 feeder cells are necessary to induce human iT cells. Start with CD8α+ T cell isolation on Day −7 of the prime step (left). Peripheral CD8α+ T cells are negatively isolated from PBMCs and labeled by cell trace dye, followed by the addition of labeled‐CD8α+ T cells to irradiated autologous LCL and the start of coculture. Activated T cells, which are defined as cell trace dye‐diluted CD8α+ cells, are purified by cell sorting. Next, purified activated T cells (1 × 105 cells/mL) are cocultured with OP9‐hDLL1 cells in the presence of human interleukin‐7 (IL‐7; 10 ng/mL) (right). Harvesting CD8α+ T cells, adjusting cell density (1 × 105 cells/mL), and transferring onto a new OP9‐hDLL1 layer are carried out on Days 3 and 7. Coculture with OP9‐hDLL1 cells for 11 d induced iT cells, defined as CD8α+ CD45RA + C‐C chemokine receptor type 7 (CCR7)+ cells. Analysis earlier than Day 11 is possible, but the induction of iT cells was completed after more than 10 d of coculture. CFSE, carboxyfluorescein succinimidyl ester; CTFR, cell trace far red dye

Figure 2

Generation of Epstein–Barr virus (EBV)‐specific induced stem cell memory T (iT) cells from human peripheral blood T cells. A,B, Generating EBV‐specific CD8+ iT cells from human peripheral blood T cells. EBV‐specific activated T cells mainly showed effector memory T (T) (CD8α+ carboxyfluorescein succinimidyl ester [CFSE]low/− C‐C chemokine receptor type 7 [CCR7]−) cell phenotypes and central memory T (T) (CD8α+ CFSE low/− CCR7+) cell phenotypes (Day 0) (A). T and T cells were sorted, and then cocultured with OP9‐hDLL1 cells for 11 d. Flow cytometry analysis of CD8α+ cells 11 d after OP9‐hDLL1 cell coculture (B). C, Number of CD8α+ cells at 0, 3, 7, and 11 d after coculture with OP9‐hDLL1 cells. T or T cells (1 × 105) were cocultured with OP9‐hDLL1 cells on Day 0. D, Recall responses to EBV. Flow cytometry analysis of CFSE dilution in each cell population (left). Column graph shows the fold increase of recovered T cells (n = 3 per group) (right). *P < .05, **P < .01 (one‐way ANOVA). Data are representative of independent experiments using human samples provided by three healthy donors. Error bars show SEM. CD45RA, CD45 isoform RA; FSC, forward scatter; SSC, side scatter; SSC‐H, side scatter‐height; SSC‐W, side scatter‐width

T‐cell receptor stimulating procedures do not alter the efficiency of iTSCM generation

For targeting tumor cells, antigen specificity is critical for adoptive T cell transfer. Expansion of endogenous antigen‐specific T cells is used in TIL therapy. Patients’ peripheral T cells or tumor‐infiltrating T cells were expanded in vitro by culture with TAA‐loaded antigen‐presenting cells, tumor cells, or tumor tissues.3 Retroviral or lentiviral TCR or CAR transduction exogenously provides antigen specificity with pan‐TCR stimulation using anti‐CD3/CD28 beads.26 To compare the effects of TCR stimuli variations on iTSCM generation, we stimulated peripheral CD8α+ T cells by LCL, CD3/CD28 beads, and melanoma antigen recognized by T cell‐1 (MART‐1) peptide‐pulsed monocyte‐derived dendritic cell (MART‐1 DC). These activated T cells were cocultured with OP9‐hDLL1 cells for 11 days following TCR stimulation (Figure 3A). The CD8α+ T cells activated by LCL and anti‐CD3/CD28 beads were expanded, and approximately 13.9% and 30.6% of activated cells showed TCM phenotypes, respectively. Additionally, MART‐1 DC induced MART‐1‐specific TCM cells (Figure 3B). We next isolated these cells and cocultured them with OP9‐hDLL1 cells. The induction step efficiently converted the TCM cells into iTSCM cells (Figure 3C), and the efficiency and amplification rates of iTSCM from activated TCM cells are summarized in Table 2. The iTSCM cells from LCL and CD3/CD28 bead‐activated T cells were generated in greater number than from MART‐1‐specific T cells. These data indicate that activated T cells by any TCR stimulating procedure can be converted into iTSCM cells.

Figure 3

Generation of induced stem cell memory T (iT) cells from antigen‐specific and CD3/CD28 bead‐activated T cells. A, Schematic for generating human iT cells from Epstein–Barr virus and melanoma antigen recognized by T cell‐1 (MART‐1)‐specific and CD3/CD28 bead‐activated T cells. B,C, Sorting panels of each CD8α+ central memory T (T) cell population stimulated by lymphoblastoid cell lines (LCL), MART‐1 peptide‐pulsed monocyte‐derived dendritic cells (MART‐1 DC), and CD3/CD28 beads. MART‐1+ T cells were detected by MART‐1‐loaded MHC class I tetramer (B). CD45RA/C‐C chemokine receptor type 7 (CCR7) profiles of the CD8α+ T cells on Day 11 after coculture with OP9‐hDLL1 cells (C). Data are representative of independent experiments using human samples provided by two HLA‐A2+ healthy donors. CD45RA, CD45 isoform RA; CFSE, carboxyfluorescein succinimidyl ester; TCR, T‐cell receptor

Table 2

Percentages and number of recovered induced stem cell memory T (iTSCM) cells (mean ± SEM) after OP9‐hDLL1 coculture from activated T cells by various T‐cell receptor stimulation procedures

	iTSCM (%)	iTSCM (×105)
LCL	95.8 ± 0.36	5.44 ± 0.054
MART‐1 DC	92.3 ± 0.36	0.52 ± 0.044
CD3/CD28 beads	95.5 ± 0.071	3.86 ± 0.15

Lymphoblastoid cell line (LCL) and melanoma antigen recognized by T cell‐1 (MART‐1) peptide‐pulsed monocyte‐derived dendritic cell (DC), n = 3; CD3/CD28 beads, n = 6.

Cytokine effects in the prime step on iTSCM generation

Interleukin‐15 and IL‐21 have been reported to enhance TSCM inducing efficiency.17, 21 To investigate T cell stimulating and homeostatic cytokine effects in the prime step, we cocultured T cells with autologous LCL in the absence or presence of IL‐2, IL‐7, IL‐15, and IL‐21 (Figure S1A). IL‐7 and IL‐15 made CD8α+ T cells highly proliferate compared with other conditions (Figure S1B). Both TEM and TCM cells stimulated by IL‐2 and IL‐21 recovered comparable rates and counts of TEM and TCM cells that were activated in the absence of cytokines (Figure S1C,D). CD45RA‐positive effector memory T (TEMRA) cells, which are CD45RA‐positive effector memory T cells, are thought of as terminally differentiated effector T cells. We also observed low numbers of TSCM and TEMRA cells in IL‐2‐, IL‐21‐, and cytokine‐absent conditions, whereas large numbers of TSCM and TEMRA cells stimulated by IL‐7 and IL‐15 were observed (Figure S1C,D). We next cocultured IL‐7‐ and IL‐15‐stimulated T cells, which were individually isolated as TSCM, TCM, TEM, and TEMRA phenotypes, with OP9‐hDLL1 cells (Figure S1E). Both IL‐7‐ and IL‐15‐stimulated TSCM cells completely retained TSCM cell phenotypes and showed a 1‐ to 2‐fold increase in cell number after coculture with OP9‐hDLL1 cells (Figure S1F, Table 3). TCM cells activated in the absence of cytokines were efficiently converted into iTSCM cells and recovered more than 10‐fold the number of cells compared with the number before OP9‐hDLL1 coculture, whereas a partial conversion into iTSCM cells and 1‐ to 3‐fold increases in cell number compared with before the OP9‐hDLL1 coculture were observed in IL‐7 and IL‐15 stimulated TCM cells from 1 × 105 cells (Figure S1F, Table 3). The absence of cytokines and the presence of IL‐7 and IL‐15 in the prime step could more efficiently convert into iTSCM cells with a higher degree of proliferation than other T cells.

Table 3

Percentages and number of recovered induced stem cell memory T (iTSCM) cells (mean ± SEM) after OP9‐hDLL1 coculture from activated T cells with or without cytokines in the prime step (n = 3 per group)

	None	IL‐7	IL‐15
iTSCM (%)
TSCM	NA	99.6 ± 0.033	99.6 ± 0.033
TCM	89.20 ± 0.27	70.9 ± 0.290	39.6 ± 0.620
TEM	4.39 ± 0.32	17.9 ± 0.300	17.0 ± 0.220
TEMRA	NA	60.7 ± 0.660	55.3 ± 3.210
iTSCM (×105)
TSCM	NA	2.76 ± 0.0110	0.830 ± 0.0260
TCM	10.700 ± 0.4200	3.25 ± 0.1100	1.570 ± 0.0990
TEM	0.019 ± 0.0011	0.070 ± 0.0041	0.042 ± 0.0036
TEMRA	NA	0.43 ± 0.0370	0.260 ± 0.0360

IL, interleukin; NA, not applicable; TCM, central memory T; TEM, effector memory T; TEMRA, CD45RA‐positive effector memory T; TSCM, stem cell memory T.

Conversion of long‐term cultured T cells into iTSCM generation

In conventional T‐cell therapy, T cells are repeatedly stimulated for a long period to be expanded before infusion into patients; however, such long‐term TCR stimulation leads to T cell exhaustion or anergy.27 We assessed whether long‐term culture of T cells has any effects on iTSCM generation. We first cocultured carboxyfluorescein succinimidyl ester (CFSE)‐labelled T cells with autologous LCL in the presence of IL‐7 or IL‐15. Seven days after the coculture, we isolated EBV‐specific T cells and then restimulated these cells by coculture with LCL for an additional 21 days (Figure S2A,B). Long‐term culture strongly induced terminal differentiation of EBV‐specific T cells, whereas several T cells maintained TSCM and TCM phenotypes, as shown in Figure S2(C). Cytokine‐depleted conditions did not allow T cells to survive for 2 weeks (data no shown). We next isolated each memory T cell subset and then cocultured them with OP9‐hDLL1 cells. Both IL‐7‐ and IL‐15‐stimulated TSCM cells completely retained TSCM cell phenotypes and showed a 0.8‐fold increase in cell counts after coculture with OP9‐hDLL1 cells (Figure S2D, Table 4). Partial conversion into iTSCM cells and 0.8‐fold increases in the number of cells were observed in IL‐7‐stimulated TCM cells and IL‐7‐ or IL‐15‐stimulated TEMRA cells (Figure S2D, Table 4). Small percentage of interleukin‐15‐stimulated TCM cells and TEM cells were converted into iTSCM cells (Figure S2D, Table 4). These data indicate that the presence of IL‐7 in long‐term priming could efficiently generate iTSCM cells from TSCM, TCM, and TEMRA cells.

Table 4

Percentages and cell count of recovered induced stem cell memory T (iTSCM) cells (mean ± SEM) after OP9‐hDLL1 coculture from long‐term activated T cells (n = 3 per group)

	IL‐7	IL‐15
iTSCM (%)
TSCM	99.40 ± 0.133	98.90 ± 0.34
TCM	58.50 ± 3.920	0.27 ± 0.059
TEM	0.62 ± 0.110	0.63 ± 0.11
TEMRA	29.40 ± 0.150	47.40 ± 2.01
iTSCM (×105)
TSCM	0.830 ± 0.0440	0.7900 ± 0.00760
TCM	0.840 ± 0.0640	0.0097 ± 0.00220
TEM	0.006 ± 0.0010	0.0020 ± 0.00044
TEMRA	0.140 ± 0.0063	0.0900 ± 0.00370

IL, interleukin; TCM, central memory T; TEM, effector memory T; TEMRA, CD45RA‐positive effector memory T; TSCM, stem cell memory T.

Cytokine effects in the induction step on iTSCM generation

To assess cytokine effects in the induction step, we next cocultured T cells with autologous LCL for 7 days, and EBV‐specific T cells were then transferred onto OP9‐hDLL1 layers for 11 days in the presence of IL‐2, IL‐7, IL‐15, and IL‐21 (Figure 4A). The absence of cytokines induced T cell death in the induction step (data not shown). The conversion efficiency rates of iTSCM cells were 8% (IL‐2), 84% (IL‐7), 44% (IL‐15), and 21% (IL‐21) (Figure 4B,C). Although IL‐7 and IL‐15 generated a greater number of iTSCM cells than the initial cell number on Day 0, fewer IL‐2‐ and IL‐21‐induced iTSCM cells were recovered than the initial cell count (Figure 4D). These results indicate that both IL‐7 and IL‐15 generate a high number of iTSCM cells in the induction step, but IL‐7 generates iTSCM cells with high purity and certain cell expansion, whereas IL‐15 generates a larger number of iTSCM cells with lower purity than IL‐7 does. Next, to validate synergistic effects of cytokines, we induced iTSCM cells in the presence of IL‐7 with or without IL‐2, IL‐15, and IL‐21. The combination of IL‐7 and IL‐15 effectively converted into iTSCM cells and a larger number of the cells were recovered by the combination, compared with other conditions (Figure S3).

Figure 4

Effects of cytokines in the induction step for induced stem cell memory T (iT) cell generation. A, Schematic for the induction of human iT cells by human interleukin (IL)‐2, IL‐7, IL‐15, and IL‐21. Coculture with OP9‐hDLL1 layers was undertaken in the presence of human IL‐2 (20 ng/mL), human IL‐7 (10 ng/mL), human IL‐15 (20 ng/mL), or human IL‐21 (20 ng/mL). B, Flow cytometry analysis of CD8α+ T cells after OP9‐hDLL1 cell coculture with multiple cytokines. C,D, Percentages (C) and cell counts (D) of recovered iT cells after OP9‐hDLL1 coculture (n = 3 per group). **P < .01 (one‐way ANOVA). Data are representative of at least two independent experiments. Error bars show SEM

Characterization of iTSCM populations generated by different priming methods

It has been shown that the transcriptional program strictly controls effector and memory T cell fate and functions. Blimp‐1 (encoded by PRDM1) and T‐bet (encoded by TBX21) positively regulated terminal differentiation, whereas Eomes and B‐cell lymphoma 6 protein (Bcl‐6) promote memory formation and retain memory homeostasis.28 These transcriptional programs govern not only effector/memory formation but also T cell survival and proliferative and effector ability.

To characterize iTSCM cells derived from CD3/CD28 bead‐activated T cells (beads‐iTSCM) and those cells derived from LCL‐activated T cells (LCL‐iTSCM), we assessed these gene profiles and proliferative ability. Blimp‐1 and T‐bet were poorly expressed in all iTSCM populations compared with TEM and TCM cells (Figure 5A,B). High expression of EOMES and low expression of BCL6 were observed in beads‐iTSCM cells, whereas the opposite results were observed in LCL‐iTSCM cells either induced in the presence of IL‐7 (designated as “iTSCM (IL‐7)”) or IL‐15 (designated as “iTSCM (IL‐15)”) (Figure 5A,B). Beads‐iTSCM and iTSCM (IL‐7) cells showed strong proliferative ability after recall response, but weak proliferation was observed in iTSCM (IL‐15) cells (Figure 5C,D). Proliferation of iTSCM (IL‐7) cells was higher than beads‐iTSCM cells (Figure 5C,D). These results indicate that effector‐associated programs are suppressed in all iTSCM populations and iTSCM (IL‐7) cells have superior proliferative ability compared to other iTSCM cells.

Figure 5

Gene profile and proliferative ability of induced stem cell memory T (iT) cells. A,B, Gene expression in bead‐generated effector memory T (T), central memory T (T), and iT cells, and lymphoblastoid cell line‐generated T, T, and iT cells induced by interleukin (IL)‐7 (iT (IL‐7)) or IL‐15 (iT (IL‐15)) (n = 3 per group). Each gene expression was normalized by 18S rRNA expression level. C,D, Recall responses by T‐cell receptor stimulation. Each T cell population (5 × 104) was activated by CD3/CD28 beads for 60 h. Column graphs show the fold increase of recovered T cells (n = 3 per group). **P < .01 (one‐way ANOVA). Data are representative of at least two independent experiments. Error bars show SEM

We then tried to generate TAA‐specific iTSCM cells. In Figure 3, MART‐1‐specific iTSCM cells could be converted from MART‐1‐specific TCM cells, but the recovered cell number was much lower than LCL‐ and beads‐iTSCM cells. To overcome this low yield, we optimized culture conditions for generating MART‐1‐specific iTSCM cells. We first isolated naïve T cells from PBMC (endogenous TSCM cells were depleted by anti‐CD95 antibody) and cocultured with autologous MART‐1 DCs for 7 days (Figure 6A,B). Next, we purified MART‐1 tetramer+ T cells and restimulated them using CD3/CD28 beads for additional 7 days (Figure 6C). Fourteen days after stimulation, most of the activated MART‐1‐specific T cells, which showed TEM phenotypes, were transferred onto an OP9‐hDLL1 layer in the presence of IL‐7 alone or IL‐7 and IL‐15 (Figure 6D). The combination of IL‐7 and IL‐15 could effectively induce LCL‐iTSCM cells with more than 80‐fold expansion (Figure S3). The combination could effectively induce MART‐1‐specific iTSCM cells regardless of donors; IL‐7 alone could not effectively induce iTSCM cells from one donor (Figure 6E, Table 5). Gene profiles of the iTSCM cells resembled those of LCL‐iTSCM cells and showed strong proliferative ability compared with the cells before iTSCM induction (Figure 6F,G). These results confirm that this two‐step iTSCM induction system can be applied regardless of activation methods.

Figure 6

A, Schematic for generating human induced stem cell memory T (iT) cells from melanoma antigen recognized by T cell‐1 (MART‐1)‐specific activated T cells. Purified naïve CD8α+ T cells were activated by MART‐1 peptide‐pulsed monocyte‐derived dendritic cells (MART‐1 DC) for 7 d. MART‐1‐specific T cells were then purified and restimulated by CD3/CD28 beads for 7 d. After 14 d of stimulation, whole MART‐1‐specific T cells were cocultured with OP9‐hDLL1 cells for 11 d in the presence of interleukin (IL)‐7 or both IL‐7 and IL‐15. B‐E, Sorting panels of each step of CD8α+ T cell populations. Naïve T cells were purified from HLA‐A2+ PBMCs by cell sorting, gated as CD8α+ CD95− CD45RA + C‐C chemokine receptor type 7 (CCR7)+ cells (B). MART‐1+ central memory T (T) cells were detected by MART‐1‐loaded MHC class I tetramer (C). CD45RA/CCR7 profiles of the CD8α+ T cells on Day 0 before coculture with OP9‐hDLL1 cells (D) and on Day 11 after coculture with OP9‐hDLL1 cells (E). F, Gene expression in MART‐1‐specific T cells before (Day 0) or after (Day 11) OP9‐hDLL1 coculture. G, Recall responses by T‐cell receptor stimulation. MART‐1‐specific activated (Day 0) and iT cells (Day11) (5 × 104) were activated by CD3/CD28 beads for 60 h. Column graphs show the fold increase of recovered T cells (n = 3 per group). *P < .05, **P < .01 (Student's t‐test). Data are representative of independent experiments using human samples provided by two HLA‐A2+ healthy donors. Error bars show SEM. CD45RA, CD45 isoform RA; FSC, forward scatter; HS, healthy subject; SSC, side scatter; SSC‐H, side scatter‐height; SSC‐W, side scatter‐width

Table 5

Percentages and cell count of recovered induced stem cell memory T (iTSCM) cells (mean ± SEM) after OP9‐hDLL1 coculture from melanoma antigen recognized by T cell‐1‐specific T cells (n = 3 per group)

	HS1	HS2
iTSCM (%)
IL‐7	13.4 ± 0.42	68.1 ± 0.44
IL‐7 + IL‐15	76.4 ± 0.71	79.7 ± 0.26
iTSCM (×105)
IL‐7	0.004 ± 0.00058	1.16 ± 0.087
IL‐7 + IL‐15	2.250 ± 0.03900	11.9 ± 0.860

HS, healthy subject; IL, interleukin.

Application of human iTSCM cells for cancer immunotherapy

The iTSCM cells showed strong proliferation following recall response and long‐term persistence after adoptive transfer.29 Thus, we investigated the antitumor effects of human iTSCM cells using human LCL‐bearing mice. We s.c. inoculated LCL into NOD.Cg‐PrkDC Il2rg (NSG) mice. Eight days after tumor inoculation, we transferred EBV‐specific TEM, TCM, and iTSCM cells into autologous LCL‐bearing mice (Figure 7A). As shown in Figure 7(B), EBV‐specific iTSCM cells showed significantly stronger suppressive effects on LCL growth than EBV‐specific TEM and TCM cells. Consequently, EBV‐specific iTSCM cells improved the survival rates of the mice (Figure 7C). Tumor antigen‐specific human iTSCM cells are more likely to have potent antitumor effects and are appropriate for adoptive cancer immunotherapy.

Figure 7

Antitumor potential of human induced stem cell memory T (iT) cells. A, Schematic for generating a humanized tumor model mice for adoptive T‐cell therapy. Severe immunodeficient (NOD.Cg‐PrkDC Il2rg /Szj, NSG) mice were s.c. inoculated with 5 × 106 Epstein–Barr virus‐transformed lymphoblastoid cell line (LCL). Effector memory T (T), central memory T (T), and iT cells (5 × 105) were adoptively transferred into LCL‐bearing mice 12 d after LCL inoculation. B, Tumor volumes of LCL‐bearing mice. C, Survival rates of LCL‐bearing mice (no transfer and T, n = 7; T, n = 4; iT, n = 6) **P < .01 (one‐way ANOVA [B]; Long‐rank test [C]). Data are representative of at least two independent experiments. Error bars show SEM

DISCUSSION

Stem cell memory T cells have functional advantages for adoptive T‐cell therapy compared with other memory T cell populations.11 Thus, TSCM cells should play a significant role in cancer immunotherapy. We previously reported a novel TSCM generating method, converting memory and effector T cell subsets into TSCM cells. In this report, we optimized the conditions for generating iTSCM cells for potential adoptive immunotherapy. Our method constitutes two steps, the prime step and the induction step. This method can induce iTSCM cells regardless of the priming method: from activated T cells by non‐specific TCR stimulation (CD3/CD28 beads), expanded T cells from existing memory T cells with a specific antigen (EBV), or expanded antigen‐specific TEM cells from naive T cells (MART‐1). As T cells expand during both priming and induction steps, we achieved a high iTSCM cell yield. In our experiments, more than 1 × 106 MART‐1‐specific iTSCM cells were recovered from 100 mL whole blood. In addition, we showed the conceptual advantage of human tumor antigen‐specific iTSCM cells for antitumor adoptive T‐cell therapy.

Efficiency and expansion of antigen‐specific iTSCM cells are highly dependent on cytokines. For the induction of EBV‐specific iTSCM cells, IL‐7 was sufficient to generate a large number of EBV‐specific iTSCM cells (Figures 4,S1). Combined stimulation with IL‐7 and IL‐15 efficiently induced MART‐1‐specific iTSCM cells from MART‐1‐specific TEM cells during the induction step (Figure 6). The cultural settings provided high yield (15‐fold) of the iTSCM cells compared with only IL‐7 stimulation (Table 5). The reason why different cytokines are required for LCL‐induced iTSCM cells and MART‐1 DC‐induced iTSCM cells is not clear at present. This may be because of the difference in antigen‐presenting cells (B cells vs DCs) or the difference in the origin of primed T cell phenotypes (in vivo memory T cells in LCL‐iTSCM and naïve T cells in MART‐1 iTSCM).

Expression of key transcription factors for memory subset differentiation and functions was compared in iTSCM cells induced from different sources (Figures 5, 6). Blimp‐1 and T‐bet, which positively regulate terminal effector formation were low in three different types of iTSCM cells. Although Eomes and Bcl‐6 mRNA expression appear to be variable in iTSCM cells induced from CD3/CD28 bead‐activated T cells, reduced expression of EOMES and increased expression of BCL6 were observed in both MART‐1 DC‐induced iTSCM cells and LCL‐induced iTSCM cells, suggesting that iTSCM phenotypes are mostly conserved, regardless of the priming method.

One could argue that iTSCM cells might be a result of selective expansion of pre‐existing TSCM‐like cells. However, we generated MART‐1‐specific iTSCM cells from naïve T cells that excluded TEMRA, TEM, TCM, and TSCM cells, from healthy donors. Thus, the possibility of expanding pre‐existing TSCM cells is unlikely, although it is very difficult to completely exclude this possibility of contamination. In addition, it is hard to show a direct generation of iTSCM cells from pre‐existing TEM cells and TCM cells in vivo. We showed that iTSCM cells can be generated from activated T cells from immunized mice, which include TEM cells. However, it is difficult to show the direct conversion of human existing TEM cells to iTSCM cells from healthy donors without immunization. Nevertheless, it is a great advantage of our method for immunotherapy that iTSCM cells can be generated from TEM and TCM cells primed from any type of T cell, regardless of naive or memory.

The functional role of Notch signaling in iTSCM cells remains to be clarified. Previously, we showed that iTSCM cells can be induced by coculture with OP9‐DL1 but not with OP9 cells. In addition, Notch signaling inhibitors strongly suppressed generation of iTSCM cells.12 These data indicate that Notch signals are indispensable for the induction of iTSCM cells. Previous work by Maekawa et al30 also reported that Notch signaling plays a central role in maintaining CD4+ memory T cells. Therefore, we think that Notch signaling is important not only for induction but also for maintenance of iTSCM cells.

As a next step for cancer immunotherapy, establishing the method to generate iTSCM cells from exhausted T cells within the tumor. As we could not obtain TILs from patients at present, we have not addressed the question whether iTSCM cells can be generated directly from TILs. However, as TILs can be expanded in vitro by IL‐2 or TCR stimulation, we speculate that iTSCM cells will be induced from TILs after expansion by our methods, like LCL‐activated T cells or MART‐1 DC‐activated T cells. We also need to improve the method as the good manufacturing practice‐graded methods without the use of OP9‐hDLL1 feeder cells and stimulator LCLs.

CONFLICT OF INTEREST

The authors have no conflict of interest.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.