Article Text

Original research
Combination of cancer vaccine with CD122-biased IL-2/anti-IL-2 Ab complex shapes the stem-like effector NK and CD8+ T cells against tumor
  1. Kanako Shimizu1,2,
  2. Shogo Ueda1,
  3. Masami Kawamura1,
  4. Honoka Aoshima1,
  5. Mikiko Satoh1,
  6. Jun Nakabayashi3 and
  7. Shin-ichiro Fujii1,2
  1. 1Laboratory for Immunotherapy, RIKEN Center for Integrative Medical Sciences, Yokohama, Japan
  2. 2Program for Drug Discovery and Medical Technology Platforms, RIKEN, Yokohama, Japan
  3. 3Department of Mathematics, Tokyo Medical and Dental University, Ichikawa, Japan
  1. Correspondence to Dr Shin-ichiro Fujii; shin-ichiro.fujii{at}


Background A key to success of cancer immunotherapy is the amplification and sustenance of various effector cells. The hallmark of prominent antitumor T cells is their long-term effector function. Although interleukin (IL)-2 is an attractive cytokine, several attempts have been made towards developing IL-2 modalities with improved effectiveness and safety that enhance natural killer (NK) cells or T cells in cancer models. However, whether such IL-2 modalities can simultaneously support long-term innate and adaptive immunity, particularly stem-like memory, has not been shown. To resolve this issue, we compared the antitumor cellular mechanism with two IL-2/anti-IL-2 complexes (IL-2Cxs) administered in combination with a therapeutic cancer vaccine, which we had previously established as an in vivo dendritic cell-targeting therapy.

Methods Two types of IL-2Cxs, CD25-biased IL-2Cx and CD122-biased IL-2Cx, together with a Wilms’ tumor 1-expressing vaccine, were evaluated in a leukemic model. The immunological response and synergistic antitumor efficacy of these IL-2Cxs were then evaluated.

Results When CD25-biased or CD122-biased IL-2Cxs in combination with the vaccine were assessed in an advanced-leukemia model, the CD122-biased IL-2Cx combination showed 100% survival, but the CD25-biased IL-2Cx did not. We first showed that invariant natural killer T (NKT) 1 cells are predominantly activated by CD122-biased IL-2Cx. In addition, in-depth analysis of immune responses by CD122-biased IL-2Cx in lymphoid tissues and the tumor microenvironment revealed a dramatic increase in the distinct subsets of NK and CD8+ T cells with stem-like phenotype (CD27+Sca-1hi, CXCR3hi, CD127+TCF-1+T-bet+ Eomes+). Moreover, CD122-biased IL-2Cx combination therapy maintained long-term memory CD8+ T cells capable of potent antitumor protection. After the high dimensional profiling analysis of NK and CD8+T cells, principal component analysis revealed that the stem-like-NK cell and stem-like-CD8+T cell state in the combination were integrated in the same group.

Conclusions CD122-biased IL-2Cx combined with the vaccine can induce a series of reactions in the immune cascade, including activation of not only NKT1 cells, but also NK and CD8+ T cells with a stem-like memory phenotype. Since it can also lead to a long-term, strong antitumor response, the combination of CD122-biased IL-2Cx with a vaccine may serve as a potential and competent strategy for patients with advanced cancer.

  • Cytokines
  • Natural Killer T-Cells
  • Killer Cells, Natural
  • CD8-Positive T-Lymphocytes
  • Immunologic Memory

Data availability statement

Data are available upon reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Natural killer (NK) and CD8+ T cells are decreased in the advanced stage of cancer. To enhance the efficacy and avoid adverse effects, several modalities for interleukin (IL)-2, which can enhance the antitumor effector cells, have been developed.


  • When CD122-biased IL-2 complex (IL-2Cx) is used in combination with our cancer vaccine, which can potentiate both innate and adaptive immunity, CD8+ T and NK cells with stem-like phenotype, revealing the prominent antitumor effect, were simultaneously elicited in an advanced-leukemia model.


  • Our results provide the design of future combination therapy of cancer vaccine plus CD122-biased IL-2Cx to maximize and maintain the long-term effect of CD8+ T and NK cells against cancer.


Interleukin (IL)-2 is a pleiotropic cytokine that exerts immunostimulatory or immunoinhibitory activity depending on the target cells. IL-2 has three receptors, IL-2Rα (CD25), IL-2Rβ (CD122), and common γ chain (CD132) and two types of heterodimers, IL-2Rαβγc and IL-2Rβγc. The high-affinity heterodimer, IL-2Rαβγc is constitutively expressed on regulatory T cells (Tregs) and transiently expressed on effector T cells but not on naïve/memory T cells, whereas the intermediate-affinity heterodimer, IL-2Rβγ is expressed by memory CD4+ and CD8+ T cells and constitutively expressed by natural killer (NK) cells.1 2 High dose IL-2 therapy has been approved as an immunotherapy for melanoma and renal carcinoma.1 2 However, IL-2 therapy exhibited not only limited clinical efficacy (approximately 5–20%), but also several toxicities, such as vascular leak syndrome due to CD25 expression on endothelial cells, cytokine storm, and a shortened half-life in vivo.3 4 To solve these problems, several modalities of IL-2 have been developed, such as engineered IL-2, PEGylated IL-2 agonists, IL-2 fusion protein, and IL-2/anti-IL-2 mAb complex (referred to as IL-2Cx).1 2 IL-2Cxs have been designed for redirecting IL-2 toward either IL-2Rαβγ+ Tregs or IL-2Rβγ+ NK and memory CD8+ T cells.5 In fact, IL-2Cx(JES6) predominantly expands IL-2Rαhigh Treg by preventing IL-2 from binding to CD122 or CD132, whereas IL-2Cx(S4B6) expands memory CD8+ T and NK cells by binding CD122, but not CD25.6

For cancer immunotherapies, it is worthwhile to generate long-term effective T (ie, memory T) and prolonged effective NK cells with the abilities to self-renew and proliferate. T-cell memory has been studied in terms of long-term persistence and recall response by measuring interferon (IFN)-γ, tumor necrosis factor (TNF)-α, and IL-2 using the enzyme-linked immunospot (ELISpot) assay and/or intracellular cytokine staining with several memory phenotypes using four or five color flow cytometric analysis. However, recently, T cells have been further; for example, not only central memory T (TCM), effector memory T (TEM), and stem-like memory T (TSCM), but also pre-exhausted T or terminal exhausted T cells.7–10 Therefore, to classify memory T or NK cells simultaneously in detail cannot be done by conventional assays alone. To resolve this issue, high dimensional analysis of T and NK cell populations simultaneously, using multiparameters is required.

The innate and acquired immune systems are linked by antigen-presenting cells, such as macrophages and dendritic cells (DC). It is particularly important to determine how the innate immune system drives the immune cycle in cancer.11 Furthermore, it is difficult to determine the types of NK cells that cooperate with CD8+ cytotoxic T cells (CTL) against cancer. Since most studies focus only on either NK or T cells under the specific conditions in tumor models, the link between them cannot be understood. As an approach to augment both NK and antigen-specific T-cell immunity, we previously established an in vivo DC-targeting therapeutic cellular vaccine system, named an artificial adjuvant vector cell (aAVC).12–15 aAVCs are composed of allogeneic cells, that is, NIH3T3 cells for mice or HEK293 cells for human, expressing the CD1d/invariant natural killer T (iNKT) ligand complex on the cell surface and the target antigen protein inside. We previously detected long-term memory T cells using this vaccine in mouse models.14 Recently, we observed an NK and T-cell activation by treatment using Wilms’ tumor 1 (WT1)-expressing aAVC (aAVC-WT1) and safety in a phase I investigator-initiated clinical trial for patients with relapse and refractory acute myeloid leukemia (RR-AML).16 Therefore, by using the aAVC model, we can follow the series of the immune circuit of one tumor-bearing host from innate to adaptive and memory immunity. Therefore, aAVC therapy would be useful to solve the different kinds of effector NK or memory T cells that are generated and enhanced by IL-2Cx.

In the current study, we investigated the efficacy of IL-2Cxs after administering the vaccine in an advanced-leukemia model. We studied the combination of IL-2Cxs at the contraction phase of T cells after immunization of vaccine and characterized T cells and other effector cells, such as iNKT and NK cells, in terms of amplification and functions. In particular, considering recent reports of stem-like features of effector cells,9 17 we characterized the stem-like features of NK and CD8+ T cells in the spleen and bone marrow (BM) as tumor tissues using high dimensional mass cytometry; these features may be related to the strong synergistic antitumor efficacy of aAVCs.


Mice and cell lines

Specific pathogen-free 6–8 weeks old C57BL/6 mice were purchased from Charles River, Japan. Mice bearing a conditional allele of Irf8 (B6(Cg)-Irf8<tm1.1Hm>/J)18 and mice bearing Cre recombinase driven by the CD11c promoter (B6.Cg-Tg(Itgax-cre)1-1Reiz/J)19 were purchased from the Jackson Laboratory. Irf8flox/flox mice were bred with CD11c-cre mice and then interbred to produce CD11c-cre-Irf8flox/flox mice (Irf8 cKO mice). All mice were maintained under specific pathogen-free conditions at the RIKEN animal facility, and all procedures were performed in compliance with protocols approved by the Institutional Animal Care Committee at RIKEN (AEY2022-020). C1498 and NIH3T3 cells were purchased from American Type Culture Collection and the Riken Cell Bank, respectively. EL4 cells were a gift from Dr R M Steinman (Rockefeller University, New York, New York, USA, 2004). C1498-WT1 cells were established by transfecting a human WT1 complementary DNA expression vector into C1498 cells (major histocompatibility complex (MHC) class I+ class II).

In vitro transcription of RNAs

Murine CD1d, OVA, and WT1 plasmids used in this study have been previously described.14 In vitro transcription was carried out using the mMESSAGE mMACHINE T7 Ultra Kit (Ambion) according to the manufacturer’s instructions. RNA was purified using the RNeasy Mini/Midi Kit (Qiagen), and the integrity was verified by denaturing agarose gel electrophoresis.

Preparation of aAVC-WT1 and aAVC-OVA

To load α-GalCer, NIH3T3 cells were cultured for 48 hours in the presence of 500 ng/mL α-GalCer and then washed before electroporation with OVA or WT1 messenger RNA (mRNA) together with murine CD1d mRNA. RNA electroporation of NIH3T3 cells was performed as follows. Briefly, cells were resuspended in Opti-MEM, RNA was transferred to a 4 mm cuvette (Harvard Apparatus), and the samples were pulsed in an ECM 830 Square Wave Electroporation System (Harvard Apparatus). Immediately after electroporation, the cells were transferred to culture medium and cultured in the presence of 500 ng/mL α-GalCer. Transfected cells were analyzed by flow cytometry for CD1d, ELISA for OVA protein, and western blot analysis for WT1 protein.


Human and murine IL-2 were purchased from Shionogi and PeproTech, respectively; anti-IL-2 Ab(S4B6) and anti-IL-2 Ab(JES6), Bio X Cell; α-GalCer, Funakoshi, OVA257-264 peptide (SIINFEKL), Toray Research Center α-GalCer, and a vehicle (0.4% DMSO) were diluted in phosphate buffered saline (Nacalai Tesque). To generate IL-2Cx, 2.5 µg of murine IL-2 with anti-IL-2 Ab (S4B6) (7.5 µg) or anti-IL-2 Ab(JES6) (7.5 µg) were incubated at 37°C for 30 min. A library of 129 15-mer peptides spanning the WT1 protein with 11 overlapping residues was synthesized at the Proteomics Resource Center at the Rockefeller University.14 The peptides were reconstituted in dimethyl sulfoxide at 50 mg/mL (as a stock solution) and then stored at −80°C until use.

Tumor experiments

C57BL/6 mice were inoculated intravenously with 2×104 or 1×105 WT1-expressing C1498 cells (C1498-WT1 cells) at day 0 and treated with 5×105 aAVC-WT1 cells on day 7. The mice were evaluated for survival. In some experiments, the mice were treated with soluble human IL-2 (2.5×104 U/mouse), soluble murine IL-2 (2.5 µg/mouse), IL-2Cx(S4B6) (IL-2:Ab; 2.5 µg:7.5 µg/mouse), or IL-2Cx(JES6) (IL-2:Ab; 2.5 µg:7.5 µg/mouse) from days 14 to 20. For depletion experiments, CD8+ and NK cell depletion was performed using anti-CD8 Ab (clone 53.6.72; Bio X Cell) (200 µg/mouse) and anti-asialoGM1 Ab (WAKO Chemicals) (50 µL/mouse), respectively, twice a week from day 12 until the end of the experiments.


Mass cytometry antibodies

Antibodies were either purchased from Fluidigm (formerly DVS Sciences) or purified and conjugated in-house using Maxpar X8 Polymer Kits (Fluidigm), according to the manufacturer’s instructions. Online supplemental table S1 contains detailed cytometry by time of flight (CyTOF) panel information, including antibody clones and suppliers.

Supplemental material

Mass cytometry staining and measurement

Mononuclear cells from the spleen and BM of mice were stained using 1 µM cisplatin viability staining reagent (Fluidigm) for 3 min, washed using cell staining buffer, and incubated with antibodies (5 µg/mL) against CD16/32 (BioLegend) for 10 min at room temperature (RT). The samples were washed and stained with surface marker antibodies for 30 min at RT. After washing, the cells were fixed and permeabilized using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher, #00-5523-00), according to the manufacturer’s protocols. The cells were then washed using Fix and Perm wash buffers and stained with intracellular antibodies for 30 min at RT. Cells were washed twice using wash buffer and incubated overnight at 4°C with 500 nM iridium intercalator (Fluidigm) in Cytofix/Perm solution for DNA staining, washed, reconstituted in Milli-Q filtered distilled water in the presence of EQ Four Element Calibration beads (Fluidigm) at a concentration of 1×106 cells/mL, and acquired on a Helios CyTOF mass cytometer (Fluidigm).

CyTOF data analysis

The mass cytometry data was randomized using the Fluidigm acquisition software V.7.0.5189 and normalized using bead-based normalization in the CyTOF software. The data was extracted from Cytobank as new .fsc files and uploaded to Cytobank Premium ( for further analysis. Individual samples were manually gated using Cytobank for excluding normalization beads, cell debris, dead cells, and doublets on the basis of 103 Rh intensity and dead cell removal by 194 cisplatin intensity for identifying CD45+ cisplatin-negative cells for further downstream analysis.

Principal component analysis of mass cytometry data

Median values from BM CD8+ T cells and NK cells from naïve, D21 C1498-WT1-bearing, D21 C1498-WT1-bearing and aAVC-WT1 treated, or the combination therapy with aAVC-WT1 and IL-2Cx(S4B6) mice were normalized using the ‘Preprocessing’ module in the ‘Sklearn’ library of Python. A total of 30 proteins were detected from 46 samples. Values obtained were converted into logarithmic values by using the ‘FunctionTransformer’ function when a sufficiently small value (0.001) was added to avoid reaching a minus infinite value. These logarithmic values were then normalized using the ‘Normalizer’ function. These normalized values were used in further analysis. Data points of samples were projected onto a two-dimensional space using principal component analysis (PCA). Samples were clustered into two or three clusters using the ‘KMeans’ function.


Statistical significance was determined at p<0.05 using StatMate (3B Scientific, Hamburg, Germany). The number of animals (n), median values, and statistical comparison groups are described in the figure legends. All p values were calculated using Tukey’s test for three or more groups or the Mann-Whitney U test for two groups. Log-rank tests were used for survival calculations.


Immunological effects of a mRNA-transfected cellular vaccine

aAVC-WT1 as a tumor antigen mRNA-transfected cellular vaccine expresses the CD1d/α-GalCer complex on the surface and the WT1 antigen inside the vector cells (figure 1A, left).14 The immune response and antitumor effects were evaluated using aAVC-WT1 treatment in a leukemia-bearing mouse model. C57BL/6 mice were administered with C1498-WT1 intravenously and treated with aAVC-WT1 1 week later (figure 1A, right). First, we assessed the innate (NK/iNKT) and adaptive (T cell) immune responses. We observed the activation of iNKT and NK cells on day 10 (3 days after aAVC-WT1) in the spleen and BM of leukemic mice. The frequency and absolute number of iNKT cells increased on day 10 and returned to basal levels on day 14 (figure 1B, online supplemental figure S1A). NK cells were apparently increased in the BM and returned to a steady state 1 week later (figure 1C, online supplemental figure S1B). IFN-γ ELISpot assays20 showed that IFN-γ spot-forming cells (SFCs) were not detected in the spleen of non-immunized and C1498-WT1-bearing mice, even after culture with α-GalCer in vitro. However, we detected significantly high levels of IFN-γ SFCs in the spleen and BM of aAVC-WT1-treated C1498-WT1 leukemia mice (figure 1D).

Figure 1

Immunological response elicited by aAVC-WT1 treatment. (A) aAVC-WT1 treatment and assay protocol. WT1-expressing C1498 leukemic cell (C1498-WT1)-bearing mice were treated with aAVC-WT1 and monitored for immunological response and survival. (B) iNKT cell responses. The absolute number of iNKT cells (CD3dimCD1d-tetramer/Gal+ cells) was analyzed on days 10 and 14 using flow cytometry (mean±SEM, n=5; aAVC-WT1-treated day 10 vs others, ***p<0.001; Tukey’s test). (C) NK cell responses. The absolute number of NK cells (CD3NK1.1+ cells) in the spleen and BM were analyzed (n=5/group; **p<0.01, ***p<0.001, Tukey’s test) (D) IFN-γ production by iNKT and NK cells in ELISpot assays (mean±SEM, n=5/group; ***p<0.001; Tukey’s test). (E) CD8/Treg ratio in the BM at the indicated time points (n=5/group; ***p<0.001; Tukey’s test). (F) WT1-specific CD8+ T-cell responses in IFN-γ ELISpot assays in C1498-WT1-bearing mice treated with or without aAVC-WT1. In some experiments, Irf8 cKO mice were used as recipients. (mean±SEM, n=5/group; ***p<0.001; Tukey’s test). aAVC, artificial adjuvant vector cell; BM, bone marrow; ELISpot, Enzyme-Linked Immunospot; IFN, interferon; iNKT, invariant natural killer T; i.v., intravenous; MNCs, mononuclear cells; NK, natural killer; SFC, spot-forming cell; WT1, Wilms’ tumor 1.

Among T-cell subsets, CD4+ and CD8+ effector T/TEM cell subsets, rather than naïve T or TCM cells, were increased in the spleen and BM on days 10 and 14 (online supplemental figures S1C,D). Elevation of the CD8/Treg ratio in tumor-infiltrating lymphocytes is a favorable prognostic marker in cancer as previously reported.21 22 Surprisingly, in our study, the CD8/Treg ratio increased in the BM (figure 1E).

To demonstrate the in vivo DC-mediated response, we performed two experiments. When carboxyfluorescein succinimidyl ester (CFSE)-labeled aAVC-WT1 was intravenously injected into C57/BL6 mice, they were taken up by in situ splenic conventional DC (cDC)1 (CD11c+CD8+ DCs), leading to DC maturation, as assessed by CD86 upregulation (online supplemental figure S1E). The expression of CD86 in aAVC-phagocytosed CD8+DCs is significantly higher than that by non-phagocytosed CD8+DCs (online supplemental figure S1E). Additionally, using a human WT1 peptide library, we verified WT1-specific T-cell responses in IFN-γ ELISpot assays 7 days after the administration of aAVC-WT1 (figure 1F). However, we could not detect a WT1-specific CD8+ T-cell response in CD11c-cre-Irf8flox/flox (Irf8 cKO) mice, which lacked the cDC1 subset. As expected, we detected a strong T-cell response in WT mice, but not in Irf8 cKO mice (figure 1F). This data suggested that WT1 presentation by the cDC1 subset is required for the increased WT-1-specific CD8+T cells by treatment with aAVC-WT1. In summary, aAVC-WT1 activated both iNKT and NK cells, and the debris of aAVC-WT1 were efficiently captured by DCs in situ. Subsequently, the DCs matured in vivo. Finally, antigen-captured DCs exhibited cross-presentation with CD8+ T cells. Collectively, both innate and adaptive T-cell immunity were induced during the effector phase in tumor-bearing hosts.

Antitumor effects of aAVC-WT1 in combination with soluble IL-2 and IL-2Cxs in a leukemic model

We assessed the antitumor effects of aAVC-WT1 in C1498-WT1-bearing mice. In the tumor model (C1498-WT1, 2×104 cells/mouse), 80% of aAVC-WT1-treated mice survived for more than 3 months when compared with the untreated group (figure 2A). When these surviving mice were rechallenged using C1498-WT1, all the mice survived. However, when rechallenged using WT1-negative EL4 tumor cells, the mice did not survive (figure 2B), thereby indicating that aAVC-WT1 therapy induced specific memory protection against WT1-expressing leukemia.

Figure 2.

Antitumor effects of aAVC and IL-2 or IL-2/anti-IL-2 complex. (A). The antitumor effects of aAVC-WT1 in a leukemia model. As described in figure 1A, mice were challenged using 2×104 C1498-WT1 cells intravenously and then treated with aAVC-WT1 on day 7. Survival was monitored. (n=10/group; ***p<0.001; log-rank test). (B) Recall response to antitumor effects. Three to 6 months later, surviving mice from aAVC-WT1 traeted group were rechallenged using C1498-WT1 cells or irrelevant EL4 cells, and survival was assessed (n=4–5/group; ***p<0.001; log-rank test). (C) As in (A); however, high doses of C1498-WT1 cells (1×105 cells) were administered to mice. Mice were treated with or without aAVC-WT1 on day 7 and then injected i.p. with or without soluble human IL-2 (2.5×104 U/day) every day from days 14 to 20. (n=9–10/group, non-treated vs aAVC-WT1 or aAVC-WT1+IL-2, ***p<0.001; log-rank test). (D) As in (C), mice were challenged using high doses of C1498-WT1 (1×105 cells) and treated with aAVC-WT1 on day 7. With regard to the IL-2/anti-IL-2 complex (IL-2Cxs) combination, IL-2Cx(S4B6) or IL-2Cx(JES6) was injected into the mice every day from days 14 to 20. (n=7–9/group, non-treated vs others; ***p<0.001, aAVC-WT1 vs aAVC-WT1+IL-2Cx(S4B6) or IL-2Cx(JES6); ***p<0.001 and *p<0.05, respectively; aAVC-WT1+IL-2Cx(S4B6) vs aAVC-WT1+IL-2Cx(JES6); *p<0.05, log-rank test). (E) Depletion of CD8+ cell and NK cells on C1498-WT1 bearing mice treated with a combination of aAVC-WT1 and IL-2Cx(S4B6). Survival was monitored. (n=6/group; ns, not significant; log-rank test). aAVC, artificial adjuvant vector cell; IL, interleukin; i.p., intraperitoneal; i.v., intravenous; NK, natural killer; WT1, Wilms’ tumor 1.

We examined the therapeutic effect of aAVC-WT1 in mice that had been administered with C1498-WT1 at a five times higher dose (1×105 /mouse), a mouse model of advanced leukemia. The survival of leukemia-bearing mice treated with aAVC-WT1 was reduced to approximately 10–20% (figure 2C,D). Therefore, we examined whether a combination of IL-2 or IL-2Cxs could improve treatment efficacy. iNKT cells expressed CD122 and CD132 at a steady state and upregulated all IL-2 receptors on day 2; however, the levels returned to the basal level on day 7 (online supplemental figures S2A(i),S2B). To avoid IL-2 toxicity, we initiated IL-2 treatment 7 days after aAVC immunization. The combination of aAVC-WT1 and IL-2 elevated the survival rate as a partial synergistic effect, whereas human IL-2 monotherapy alone did not show any anti-leukemic effect (figure 2C). Next, we compared the antitumor efficacy of the combination of the two types of IL-2Cxs, that is, IL-2Cx(S4B6) and IL-2Cx(JES6). The combination of aAVC-WT1 and IL-2Cx(JES6) showed greater efficacy than aAVC therapy or IL-2Cx(S4B6) alone (figure 2D). Surprisingly, all mice treated with the combination of aAVC-WT1 and IL-2Cx(S4B6) survived (figure 2D), thereby indicating that it was the most effective antitumor treatment. To demonstrate whether the antitumor effect of aAVC with IL-2Cx(S4B6) is dependent on NK and CD8+ T cells, these cells were depleted in mice treated with the combination of aAVC-WT1 and IL-2Cx(S4B6). We used the anti-asialo GM1 antibody since it can deplete NK cells but not iNKT cells as we previously reported.23 24 The antitumor effects by this therapy were canceled by the depletion of either of them (figure 2E). This indicated that both NK and CD8+ T cells are important in this therapy.

iNKT cell response to the vaccine with or without IL-2Cxs

To evaluate the cellular mechanism elicited by different combination therapies, that is, the aAVC system and IL-2Cx(S4B6) or IL-2Cx(JES6), we analyzed innate (iNKT and NK cell response) and adaptive immunity using OVA antigen-expressing aAVC (aAVC-OVA) immunization. We immunized mice using aAVC-OVA intravenously on day 0 and injected IL-2Cx(S4B6) or IL-2Cx(JES6) intraperitoneally from days 7 to 13 (figure 3A). First, we assessed whether the combination of IL-2Cxs increased the number of iNKT cells in the spleen and BM on day 14. As in figure 1B and online supplemental figure S1B, the frequency and absolute number of iNKT cells were returned to a basal level 7 days after aAVC treatment. Compared with aAVC monotherapy, the combination of both IL-2Cxs increased the absolute number of iNKT cells in the spleen and BM at day 14 (figure 3B and online supplemental figure S3A). In particular, the number of T-bet+Ki-67+ iNKT cells as proliferating NKT1 cells was more robustly expanded in the group administered using the IL-2Cx(S4B6) combination than in other groups (figure 3C,D). Next, no difference was observed in IFN-γ production by iNKT cells in the spleen and BM between the groups administered with the vaccine and with or without IL-2Cxs (figure 3E and online supplemental Figures S3B,S3C). Thus, after aAVC vaccination, the NKT1 subset showed preferential proliferation capacity with IL-2Cx(S4B6) treatment.

Figure 3

aAVC/IL-2Cx combination therapies dramatically enhance iNKT cells. (A) Mice were immunized using aAVC-OVA intravenously at day 0 and treated with IL-2Cx(S4B6) or IL-2Cx(JES6) from days 7 to 13. (B) The absolute number of iNKT cells in the spleen was analyzed on day 14 using flow cytometry (mean±SEM, n=4–5/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). (C) Gating strategy of iNKT cells (left) and representative dot plots showing % T-bet+Ki-67+ among splenic iNKT cells (right). Numbers indicate the frequency of each quadrant. (D) The frequency of T-bet+Ki-67+ among iNKT cells was summarized. (mean±SEM, n=3/group; *p<0.05, **p<0.001***p<0.001; Tukey’s test). (E) The frequency of IFN-γ-positive cells among splenic iNKT cells stimulated with PMA/ionomycin. (mean±SEM, n=3/group). aAVC, artificial adjuvant vector cell; IFN, interferon; IL-2Cx, interleukin 2 complex; iNKT, invariant natural killer T; i.p., intraperitoneal; i.v., intravenous.

Proliferation of CD27+CD11b+Klrg1+ subset in NK cells after vaccination plus IL-2Cx(S4B6)

NK cell proliferation in vaccinated mice with or without IL-2Cxs was assessed on day 14. The number of NK cells returned to basal levels 7 days after aAVC treatment (figure 1C and online supplemental figure S1C). In contrast, the combination with IL-2Cx(S4B6) increased the number of NK cells in the spleen when compared with monotherapy or combination with IL-2Cxs(JES6) (figure 4A). Ki-67+ NK cells increased in the combination of both the IL-2Cxs (figure 4B). With regards to maturation stages and subsets of NK cells, the combination with IL-2Cx(S4B6) led to the expansion of not only CD27+CD11b+NK cells (semi-mature NK), but also CD27+CD11b NK cells (immature NK) or CD27CD11b+ NK cells (terminal mature NK) (figure 4C and online supplemental figure S4A upper). In contrast, only CD27+CD11bNK cells preferentially proliferated in mice following treatment with the vaccine and IL2Cx(JES6) (figure 4C and online supplemental figure S4A upper). Although Klrg1 expression in NK cells is a functional marker of terminal maturation, CD27+CD11b+Klrg1+ NK cells increased in the group administered with the IL-2Cx(S4B6) combination treatment preferentially, whereas CD27+CD11bKlrg1 NK cells were dominant in the IL-2Cx(JES6) combination treatment (figure 4D and online supplemental figure S4A lower). Additionally, we evaluated the expression of the transcription factors T-bet, Eomes, and TCF-1 in NK cells. As previously reported, TCF-1 is downregulated as NK cells mature.25 The semi-mature CD27+CD11b+Klrg1+ NK cells exhibited T-bet+TCF-1+Eomeshi (online supplemental Fig. S4B–S4D). The CD27+CD11bNK cell subset still expressed CD25 (IL-2Rα) on day 7 after aAVC immunization, while both CD27+CD11b and CD27+CD11b+ subsets showed higher CD122 (IL-2Rβ) expression than the CD27CD11b+ subset (online supplemental figure S2A(i)(ii),S2B). This implied that IL-2Cx(S4B6) stimulated the cell cycle of NK cell subsets at the immature and semi-mature stages, thereby resulting in the expansion of immature NK cells (CD27+CD11bKlrg1), semi-mature NK cells (CD27+CD11b+Klrg1+), and terminal mature NK cells (CD27CD11b+Klrg1+). Thus, based on the IL-2 receptor expression pattern of NK subsets, the CD122-biased IL-2/anti-IL-2Cx(S4B6) may be potent for both CD27+CD11b and CD27+CD11b+ NK subsets.

Figure 4

aAVC/IL-2Cx(S4B6) combination therapy leads to expansion of NK cells with semi-mature state. As shown in figure 3; however, spleen NK cells were analyzed on day 14. (A) The gating strategy of NK cells (left) and the absolute number of NK cells in the spleen (right) were analyzed (mean±SEM, n=4–5/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). (B, C) Gating strategy of the NK subset (left) and the absolute number of each NK subset (right), using expression of CD27 and CD11b (B) and CD27 and Klrg1 (C) (mean±SEM, n=3–5/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). (D) Frequency of Ki-67 expression in NK cells (mean±SEM, n=3/group; *p<0.05, **p<0.001; Tukey’s test). (E) Mass cytometry results for splenic NK cell compartments in leukemic mice. As shown in Figure 2D, spleen cells were harvested on day 21. t-SNE plots showing annotated clusters (upper), heat map (right upper), and frequency (lower) of each cluster in the splenic NK cells. (mean±SEM, n=3–6/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). aAVC, artificial adjuvant vector cell; IL-2Cx, interleukin 2 complex; NK, natural killer; WT1, Wilms’ tumor 1; t-SNE, t-distributed Stochastic Neighbor Embedding.

Potent effect of IL-2Cx(S4B6) combination treatment on the activation of CD27+CD11b+Klrg1+ NK cell phenotypes in a leukemia model

We then analyzed NK cell subsets in tumor-bearing mice. We initially examined splenic NK cell status in C1498-WT1-bearing mice after treatment with aAVC-WT-1 alone or together with IL-2Cx(S4B6) using mass cytometry analysis. Using manual gating (online supplemental figure S5), we defined the NK cell population and performed unsupervised, hierarchical clustering of NK cells. The combination therapy preferentially induced CD11bloLy6CloKlrg1+Ki67+Tbet+TCF-1dimEomeshi NK cells in cluster 4 (figure 4E), suggesting that the IL-2Cx(S4B6) combination therapy led to a semi-mature NK cell status in tumor-bearing hosts (figure 4E).

IL-2Cx(S4B6) promotes stem-like features in CD8+ T cells

The previous reports showed that aAVC immunization can induce long-term memory CD8+ T cells and that TEM/TEF (CD44hiCD62L) is dominant in the effector phase, whereas the TCM population CD44hiCD62L+ is generated in the memory phase in a murine model.13 26 In this study, the combination of vaccine plus IL-2Cx(S4B6) dramatically increased CD8+ T cells in the spleen and BM when compared with the IL-2Cx(JES6) combination treatment (figure 5A, online supplemental figure S6A). With regards to transcription factors, CD8+ T cells induced by the IL-2Cx(S4B6) combination treatment expressed TCF-1hi, Tbethi, and EomeshiBlimp1lo (figure 5B). Analysis of the subset and phenotype of CD8+ T cells showed that the IL-2Cx(S4B6) combination promoted CD62L+CXCR3hiCD127+Sca-1hi CD27+CD122hi T cells, which are similar to phenotypes of stem-like memory T cells, or the CD44hi CD62L+ (TCM) CD8+ T-cell subset (figure 5C,D). In contrast, the aAVC monotherapy and IL-2Cx(JES6) combination immunization induced CD44hi CD62L (TE/TEM) cells (figure 5C,D).

Figure 5

aAVC/IL-2Cx(S4B6) combination therapy induces CD8+ T cells with stem-like features. As shown in figures 3 and 4; however, splenic CD4+ and CD8+ T cells were analyzed on day 14. (A) Gating strategy of natural killer cells (left) and absolute number of total CD8+ T cells in the spleen (right panel) was analyzed using flow cytometry (mean±SEM, n=4–5/group; ***p<0.001; Tukey’s test). (B) Representative histogram of expression of the transcription factors TCF-1, T-bet, Eomes, and Blimp-1 in total CD8+ T cells in the spleen. (C) Representative dot plots showing phenotypic profiles of splenic CD8+ T cells. (D) The frequencies of central memory T (CD44+CD62L+) and effector memory T (CD44+CD62L) subsets are summarized (mean±SEM, n=4/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). (I) Gating strategy of OVA-SIINFEKL-tetramer+ (OVA-tet+) CD8+ T cells in the spleen (left) and summary of absolute number of OVA-tet+ CD8+ T cells (right) (mean±SEM, n=4–5/group; ***p<0.001; Tukey’s test). (F) Representative histogram of expression of the transcription factors TCF-1, T-bet, Eomes, and Blimp-1 in OVA-tet+ CD8+ T cells in the spleen. (G) Kinetics of the frequency of OVA-tet+ CD8+ T cells in the blood (lower panel). Representative flow cytometry data of OVA-tet+ CD8+ T cells in the blood at day 14 are shown (upper panel) (mean±SEM, n=6–7/group; **p<0.001, ***p<0.001; Tukey’s test). (H) Gating strategy of Foxp3+ CD4+ T cells in splenic CD4+ T cells (left) and summary of absolute number of Foxp3+ CD4+ T cells (right panel) (mean±SEM, n=4–5/group; **p<0.001, ***p<0.001; Tukey’s test). (I) CD8/Treg ratio in spleen (mean±SEM, n=4–5/group; ***p<0.001; Tukey’s test). (J) Mass cytometry results for splenic CD8+ T-cell compartment. As shown in figure 2D, spleen cells were harvested on day 21. t-SNE plots showing annotated clusters (upper panel), heatmap (upper right panel), and frequency (lower) of each cluster in splenic CD8+ T cells (mean±SEM, n=4–7/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). aAVC, artificial adjuvant vector cell; IL-2Cx, interleukin 2 complex; Treg, regulatory T cell; WT1, Wilms’ tumor 1; t-SNE, t-distributed Stochastic Neighbor Embedding; TCR, T-cell receptor.

Furthermore, the IL-2Cx(S4B6) combination significantly increased the absolute number of Kb-OVA tet(SIINFEKL)-specific CD8+ T cells in the spleen and BM when compared with the IL2Cx(JES6) combination (figure 5E and online supplemental figure S6B,S6C). The TE/TEM (CD44hi CD62L) subset in OVA-tet+ CD8 T cells increased similarly in all the groups (online supplemental figure S6D). However, OVA-tet+ CD8 T cells expressing stem-like features increased in the IL2Cx(S4B6) (TCF1hi CXCR3hi Sca-1hi CD27+CD122hi; figure 5F and online supplemental figure S6D). Therefore, we monitored the levels of OVA-tet+ CD8+ T cells in the blood. We detected OVA-specific memory CD8+ T cells even 6 months after aAVC-OVA immunization (figure 5G). The IL-2Cx(S4B6) combination therapy led to a robust expansion of antigen-specific memory CD8+ T cells not only in the effector phase, but also in the memory maintenance phase for a 6-month period.

In contrast, the number of CD4+ T cells was almost the same among all groups (online supplemental figure S6E). When we analyzed the phenotypic profiles of CD4+ T cells, it was observed that CD4+ T cells exhibited TEM/TEF(CD44hiCD62L) dominantly in the combination group of vaccine plus IL-2Cx(S4B6) (online supplemental figure S6F), which are different from CD8+ T cells. In addition, CD122 expression on CD4+ T cells was much lower than those of CD8+ T, NK, and iNKT cells (online supplemental figure S2). Therefore, CD4+ T-cell response to IL-2Cx(S4B6) is different from CD8+ T-cell response.

Furthermore, the IL-2Cx(S4B6) combination did not promote Treg cell numbers, whereas the IL-2Cx(JES6) combination increased Tregs in the spleen (figure 5H). Thus, the CD8/Treg ratio in the spleen was higher in the IL-2Cx(S4B6) combination, but lower in the IL-2Cx(JES6) combination treatment group (figure 5I).

Cell-type-specific contribution of NK and CD8+ T cells to treatment responses

To understand whether the transition of T cells can occur from the early phase to the memory phase in tumor-bearing mice, we traced the T-cell response in the tumor and lymphoid tissues. We defined the CD8+ T-cell population (online supplemental figure S5) and performed unsupervised clustering of CD8+ T cells.

In the hierarchical clustering from CD8+ T cells, we verified that CD8+ T-cell clusters (clusters 3 and 6) with stem-like memory features were detected in the spleen by the combination of aAVC-WT-1 with IL-2Cx(S4B6) using mass cytometry analysis (figure 5J). Then, we analyzed tumor (BM)-infiltrating immune cells. We performed further BM CD8+ T-cell subset analyses using an unbiased t-distributed Stochastic Neighbor Embedding (t-SNE) analysis (figure 6A). The combination therapy led to the formation of C5 and C10 clusters (CD44dim−hi CD62L+Klrg1TCF-1+Eomes+TbetdimKi-67hi); these features were similar to the stem-like memory features of splenic CD8 T cells (figure 5J), suggesting that the IL-2Cx(S4B6) combination therapy led to the stem-like memory features of CD8+ T cells in tumor-bearing hosts systemically (figure 6A).

Figure 6

Tumor-site-infiltrating CD8+ T and NK cells induced by the combination therapy. As shown in figure 2D, the mice were challenged using C1498-WT1 and then treated with aAVC-WT1 or aAVC-WT1+IL-2Cx(S4B6). On day 21, the bone marrow cells from these mice were analyzed using mass cytometry. (A) Mass cytometry results of the BM CD8+ T-cell compartment. t-SNE plots showing annotated clusters (upper), heatmap (right upper), and frequency (lower) of each cluster in BM CD8+T cells (mean±SEM, n=4–7/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). (B) Mass cytometry results for the BM NK cell compartment. T-SNE plots showing annotated clusters (upper panel), heatmap (right panel), and frequency (lower panel) of each cluster in the splenic NK cells. (mean±SEM, n=3–6/group; *p<0.05, **p<0.001, ***p<0.001; Tukey’s test). (C) PCA and clustering using the KMeans of BM CD8+ T cells and NK cells. Left: colored dots indicate the samples from naïve (red), D21 C1498-WT1 bearing (light green), D21 C1498-WT1 bearing, aAVC-WT1 treated (green) and D21 C1498-WT1 bearing, aAVC-WT1 and IL-2Cx(S4B6) treated (blue). Middle: colored dots indicate the samples of CD8+ T cells (red) and NK cells (blue). Right: Data shows clustering into three clusters using the KMeans. Cluster 1 (green), Cluster 2 (yellow) and Cluster 3 (violet). aAVC, artificial adjuvant vector cell; BM, bone marrow; IL-2Cx, interleukin 2 complex; NK, natural killer; PCA, principal component analysis; WT1, Wilms’ tumor 1; t-SNE, t-distributed Stochastic Neighbor Embedding.

Additionally, we focused on BM NK cells (figure 6B). BM NK cells in the IL-2Cx(S4B6) combination group exhibited an increase in semi-mature C2 (CD11bloLy6ChiKlrg1T-betloTCF-1loEomeshiCD62L+Ki-67hi) and mature C5 (CD11bloLy6ChiKlrg1+T-betloTCF-1EomeshiCD62L+Ki-67) clusters, and a decrease in immature C3 clusters (figure 6B). As we observed the vaccination results in non-tumor-bearing models (figure 4B,C), the IL-2Cx(S4B6) combination therapy led to a semi-mature NK cell status in spleen and tumor sites (BM) of tumor-bearing hosts (figure 4D and figure 6B).

To resolve the complex relationships between several clusters based on the proteome of NK and CD8+T cells in tumor sites, BM derived from each group, and therapeutic response, we leveraged a PC-based unsupervised approach. Finally, we conducted the PCA of BM CD8+ T and NK cells using the median values of multiple parameters (figure 6C left and middle) and clustered them into three clusters using the KMeans function (figure 6C right). Intriguingly, we found that CD8+ T and NK cells from the combination therapy group, but not others, were clustered and converged to the same category (figure 6C). These findings indicated that expression profiles of the marker proteins of both CD8+ T and NK cells were close to each other both phenotypically and functionally. Together our results confirmed the critical role of NK and CD8 T cells with stem-like features for the establishment of long-term therapeutic effects and suggests a stem-like T and NK cell-type-specific contribution to treatment, which may be informative.


Considering that the key to the success of cancer immunotherapy is the amplification of multiple effector cells and the suppression of suppressor cells, current immunotherapy research is shifting towards combination therapies. IL-2Cx(S4B6) has been reported to enhance antitumor effects particularly in combination with anti-programmed death receptor-1 (PD-1), radiation therapy, or cancer vaccines.27–29 The success of this combination was discussed due to enhanced CD8+ T cells or NK cells. In this study, we first demonstrated the ability of IL-2Cx(S4B6) to amplify iNKT cells and strengthen the chain immune reaction from innate to adaptive immunity in a single mouse organism. Furthermore, IL-2Cx(S4B6) therapy at the contraction phase of the aAVC vaccine revealed not only robust amplification and activation of multiple effector cells (iNKT, NK, and T cells), but also maintained NK and CD8+ T cells with stem-like features, thereby resulting in marked antitumor effects against advanced leukemia.

More and prolonged immune amplification of effector T cells is desired for the treatment of advanced patients. Studies have focused on the development of a new type of T-cell receptor (TCR) gene-transduced or chimeric antigen recptor (CAR)-T cells toward the long-term memory T cells by modifying these checkpoint blockade or inhibitory molecules. For example, T cells deleting NR4 molecule can act as stem-like memory T cells.30 31 However, TSCM may not be easily generated in vivo in tumor-bearing hosts by the vaccine. Historically, cancer vaccine research began with the search for various antigens and adjuvants to induce T cells. Since short peptides used as cancer vaccines induced a transient CD8+T-cell induction, resulting in a limited clinical response,32 vaccines using mRNA, DNA, or viral vector along with adjuvants have been developed to ensure long-term efficacy.33 34 As a form of antigen plus adjuvant, we showed that the aAVC vaccine generates long-term T-cell memory in mice in addition to innate and adaptive immunity.12–16 In humans, we recently verified that long-lasting CD8+T cells for more than 12 months, were detected in the phase I clinical trial of aAVC-WT1 for patients with RR-AML. We also found that the long-term functional CD8+T-cell clones comprised of not only clones of newly memory CD8+T cells, but also long-term reinvigorated CD8+T cells.16 Patients with advanced cancer such as those of RR-AML usually exhibit reduced frequency and dysfunction of T cells due to the disease progression and/or the treatment.35 Nevertheless, this immunotherapy is surprizing because it restored the immunocompromised state. Currently, our next goal to improve the efficacy of immunotherapy is expected to expand more long-lasting tumor antigen-specific CD8+T cells by IL-2Cx. The current results in an advanced leukemia model encourage us to expand them, because stem-like memory CD8+ T cells have a potential to robustly expand.

The mechanism of the IL-2Cx effect would probably depend on the following two factors. First, it depends on the expression pattern of IL-2R heterodimers on T cells. An interaction of IL-2R and IL-2 agonist may depend on the affinity, strength, and duration. However, the expression of IL-2Rs on effector cells can be upregulated by TCR and cytokine stimulation. In this study, we found that the upregulation of IL-2Rβ on T cells was higher on day 7 after aAVC immunization although it was also expressed on NK cells at a steady state. Another factor is the signaling in the lymphocytes after an interaction of IL-2R and IL-2 agonist. On forming IL-2Cx to IL-2R, intracellular Janus kinase (JAK) proteins are tightly associated with IL-2Rβ and IL-2Rγ phosphorylate tyrosine residues in the receptor intracellular domains. Then, the signal recruits and activates the signal transducer and activator of transcription 5 (STAT5) to coordinate immune-related gene expression.36 37 The IL-2Cx produces secondary signals through the mitogen-activated protein kinase (MAPK) and phosphatidylinositol 3-kinase (PI3K) pathways.36 37 Therefore, to modify the binding of IL-2 agonist to IL-2Rβ and IL-2Rγc can control the signal transduction, thereby altering the phenotype and function of T cells and NK cells. For example, H9T, a partial agonist of H9, which is known as IL-2Rβ-biased superkine, reduces the binding of H9 to IL-2Rγc and lowers the phosphorylated STAT5, resulting in CD8+ T-cell stem-like characteristics.38 In addition, it has recently been demonstrated that the engineering of ligands for the modification of IL-2Rβ and IL-2Rγc alter these signal pathways.39 Taken together, the effector and memory cell phenotypes may be affected by the binding pattern of IL-2 agonists and their IL-2R subunits, and the stimulatory duration of IL-2 agonists. Such an IL2-induced signal modulation could lead to an imbalance in the JAK-STAT5, MAPK, and PI3K pathways, thereby resulting in the generation of stem-like memory T cells.

NK cells, which have recently gained attention as the other main effector cells, can maintain their antitumor effector function for a prolonged time.40 Previously, the generation of cytokine-induced memory NK cells or cytomegalovirus (CMV)-specific NK cells has been demonstrated.41 42 However, it was unknown whether NK cells express stem-like memory phenotypes. Recently, NK memory stem cells (NKscm) have been reported to be involved in Zika viral infection; NKscm exhibit CD27+Klrg1+TCF1hi and great antiviral potential.43 We additionally found that the aAVC and IL-2Cx(S4B6) combination therapy induced a similar NK cell population (CD27+CD11b+Klrg1+TCF-1+Eomes+T-bet+). The CD27+CD11b+ Klrg1+NK subset showed a semi-maturation status but not terminal maturation. Generally, the expression of Ly49 C/I, which recognizes self, increases as NK cells mature, thus resulting in a higher cytotoxicity against MHC class I-negative tumor cells. In contrast, CD27+CD11b+ NK cells exhibit higher cytotoxicity against tumors expressing MHC class I and NKG2D ligand.44 In fact, our finding that the anti-leukemic effect was canceled when we depleted NK cells from day 12 (figure 2E) implicated that MHC class I+ leukemic cells were probably killed by CD27+CD11b+ NK cells and CTLs. Furthermore, as CD27CD11b+ NK cells also proliferated in the combination therapy with IL-2Cx(S4B6) (figure 4C), it could block the escape from MHC class I-positive to negative leukemic cells. Since a mixture of MHC class I-negative and MHC class I-positive tumor cells sometimes emerge in the progression of tumors, it would be worthwhile to generate stem-like NK clinically.

Finally, the development of combination therapy, a treatment strategy that combines two or more therapeutic agents, is a cornerstone of sustainable developmental goals in cancer immunotherapy, wherein we suggested aAVC and IL-2Cx to be more effective than a single agent therapy. One main synergy that has not been revealed so far is that the PCA after high dimensional analysis using mass cytometry revealed that both CD27+CD11b+ Klrg1+NK cells and CD62L+CXCR3hiCD127+Sca-1hi CD27+CD122hi CD8+ T cells have similar phenotypes of stem-like memory T cells. This implicates that the sharing of stem-like feature by NK and CD8 T cells leads to a long-term synergized antitumor effect. This would encourage the investigation of combination therapy of CD122-biased IL-2Cx and therapeutic cancer vaccines for patients with advanced cancer through clinical trials.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.


We thank An Sanpei, Jun Shinga, and Hiroshi Nakazato for technical assistance and support in this study.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors SF and KS conceptualized the work and strategy, planned, and analyzed data, and wrote the paper. SU, MK, HA and MS conducted experiments and analyzed data. JN conducted statistical analysis of mass cytometry data. SF is responsible for the overall content as guarantor.

  • Funding This study was supported by funding from the RIKEN (RIKEN President’s Discretionary Fund).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.