Functional exhaustion of CD4⁺ T cells induced by co‐stimulatory signals from myeloid leukaemia cells

Summary

To cope with immune responses, tumour cells implement elaborate strategies such as adaptive resistance and induction of T‐cell exhaustion. T‐cell exhaustion has been identified as a state of hyporesponsiveness that arises under continuous antigenic stimulus. Nevertheless, contribution of co‐stimulatory molecules to T‐cell exhaustion in cancer remains to be better defined. This study explores the role of myeloid leukaemia‐derived co‐stimulatory signals on CD4⁺ T helper (Th) cell exhaustion, which may limit anti‐tumour immunity. Here, CD86 and inducible T‐cell co‐stimulator ligand (ICOS‐LG) co‐stimulatory molecules that are found on myeloid leukaemia cells supported Th cell activation and proliferation. However, under continuous stimulation, T cells co‐cultured with leukaemia cells, but not with peripheral blood monocytes, became functionally exhausted. These in vitro‐generated exhausted Th cells were defined by up‐regulation of programmed cell death 1 (PD‐1), cytotoxic T‐lymphocyte antigen 4 (CTLA‐4), lymphocyte activation gene 3 (LAG3) and T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3) inhibitory receptors. They were reluctant to proliferate upon re‐stimulation and produced reduced amounts of interleukin‐2 (IL‐2), tumour necrosis factor‐α (TNF‐α) and interferon‐γ (IFN‐γ). Nonetheless, IL‐2 supplementation restored the proliferation capacity of the exhausted Th cells. When the co‐stimulation supplied by the myeloid leukaemia cells were blocked, the amount of exhausted Th cells was significantly decreased. Moreover, in the bone marrow aspirates from patients with acute myeloid leukaemia (AML) or myelodysplastic syndrome (MDS), a subpopulation of Th cells expressing PD‐1, TIM‐3 and/or LAG3 was identified together with CD86⁺ and/or ICOS‐LG⁺ myeloid blasts. Collectively, co‐stimulatory signals derived from myeloid leukaemia cells possess the capacity to facilitate functional exhaustion in Th cells.

Abbreviations

aCD3

anti‐CD3 monoclonal antibody

AML

acute myeloid leukaemia

CD

cluster of differentiation

CTLA‐4

cytotoxic T‐lymphocyte antigen‐4

CTL

cytotoxic T lymphocyte

FBS

fetal bovine serum

FoxP3

forkhead box protein 3

ICOS

inducible T‐cell co‐stimulator

ICOS‐LG

inducible T‐cell co‐stimulator ligand

IFN

interferon

IL

interleukin

LAG3

lymphocyte activation gene

MDS

myelodysplastic syndrome

PD‐1

programmed cell death‐1

PD‐L

programmed‐cell death ligand

Th

T helper

TIM‐3

T‐cell immunoglobulin and mucin domain‐containing protein‐3

Treg cell

regulatory T cell

Introduction

Tumour cells implement various mechanisms to escape from immune‐mediated elimination.1 Immunogenic tumours, which frequently get in touch with the effector T cells, develop advanced immune regulation strategies.1, 2 Particularly, tumour‐infiltrating CD8⁺ cytotoxic T cells have been acknowledged to up‐regulate inhibitory receptors such as programmed cell death 1 (PD‐1), cytotoxic T‐lymphocyte antigen 4 (CTLA‐4), T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3) and lymphocyte activation gene 3 (LAG3). These cells have diminished capacity for proliferation and production of interleukin‐2 (IL‐2), tumour necrosis factor‐α (TNF‐α) and interferon‐γ (IFN‐γ) cytokines.1, 2, 3 These hypo‐responsive, functionally exhausted, T cells are generally observed in chronic inflammatory disorders including persistent infections, autoimmunity and cancer. Continuous exposure to antigens has been recognized as the main inducer of T‐cell exhaustion. Inhibitory receptors that are highly up‐regulated on activated T cells associate with this state.3, 4, 5 Hence, the effector phases of T‐cell responses can be followed by exhaustion. Expectedly, co‐stimulatory molecules, which are indispensable for T‐cell activation, proliferation and polarization, correlate with the development of exhaustion.6, 7

Antigen presentation and type 1 helper T (Th1) cell responses are essential for the stimulation of cytotoxic T cells.8 Their survival, expansion and memory generation are supported by the Th1 subset.8, 9 Even though underlying molecular mechanisms are not fully understood, similar to cytotoxic T cells, Th cells are also prone to exhaustion.10, 11, 12

Myeloid leukaemia blasts share certain properties with antigen‐presenting cells and possess the capacity to interact with Th cells.13, 14 Furthermore, potent co‐stimulatory molecules, i.e. CD86 and inducible T‐cell co‐stimulator ligand (ICOS‐LG), are constitutively expressed by certain subpopulations of leukaemia cells.15, 16 A recent study from our group demonstrated the capacity of leukaemia cells to stimulate Th cell activation, proliferation and secretion of Th1‐associated IL‐2, TNF‐α and IFN‐γ cytokines through the CD28‐mediated co‐stimulatory pathway.17 Intriguingly, upon engagement with effector Th cells, the leukaemia cells acquired immune suppression capacity, acknowledged as ‘adaptive resistance’.18, 19 Correspondingly, in myeloproliferative disorders, expression of CD86 and ICOS‐LG has been associated with poor clinical prognosis and disease severity.16, 20, 21 In haematological malignancies including acute myeloid leukaemia (AML), cytotoxic T cells have been identified with an exhaustion‐like phenotype; however, there is limited information about Th cells.22, 23, 24

Here, by using in vitro models established to observe Th cell exhaustion, we report the contribution of co‐stimulatory signals derived from myeloid leukaemia cells to Th exhaustion. Upon co‐culturing with myeloid leukaemia cells, Th responses were initially triggered; however, later, these cells displayed the features of functional exhaustion that was the result of the magnitude and persistence of co‐stimulatory signals.

Materials and methods

Patient and healthy donor samples

Healthy volunteers or patients newly diagnosed with AML [n = 6 (three female, three male), median age 52 years (minimum 22; maximum 65)] or with myelodysplastic syndrome (MDS) [n = 9 (four female, five male), median age 64 years (minimum 45; maximum 75)] were enrolled into the study (Hacettepe University Local Ethics Committee, Approval no.: LUT 12/153‐35 and GO 14/606‐31). Peripheral blood samples were collected from healthy donors. Leucocytes and the leukaemic blasts were isolated from freshly obtained bone marrow aspirates with density gradient centrifugation (Ficoll 1.119; Sigma, St Louis, MO) and used in further analyses.

Cell culture

Human myeloid leukaemia cell lines, KG‐1, Kasumi‐1, HL‐60, U937 and THP‐1 were either obtained from the American Type Culture Collection (ATCC, LGC Promochem, Rockville, MD) or received as kind gifts.17 The cell lines and the freshly isolated cells were maintained in RPMI‐1640 medium supplemented with 10% foetal bovine serum (Biological Industries, Kibbutz Beit Haemek, Israel), l‐glutamine (2 mm), penicillin (100 U/ml) and streptomycin (100 μg/ml) at 37° in a humidified 5% CO₂ incubator. Otherwise specified, all the reagents were obtained from Lonza (Allendale, NJ).

Flow cytometry and fluorescence‐activated cell sorting (FACS)

The monoclonal antibodies anti‐human‐CD4 (SK3), ‐CD3 (HIT3a), ‐CD69 (FN50), ‐CD25 (M‐A251), ‐CD14 (M5E2), ‐CD13 (L138), ‐CD274 (PD‐L1; MIH1) (Becton Dickinson, San Jose, CA); ‐LAG3 (FAB2319F) (R&D, Minneapolis, MN); ‐CD154 (24–31), ‐CD127 (hIL‐7r‐M21), ‐CD80 (2D10), ‐CD86 (IT2.2), ‐CD152 (CTLA‐4; L3D10), ‐CD275 (ICOS‐LG; 9F.8A4), ‐CD278 (ICOS; C398.4A), ‐CD279 (PD‐1; EH12.2H7), ‐FoxP3 (236A/E7), ‐TIM‐3 (F38‐2E2) (Biolegend, San Diego, CA) were used in immunophenotyping analyses. The percentage of positive cells was determined in comparison to the isotype‐matched antibody controls. Intracellular staining was performed with the Cytofix/Cytoperm^™ Kit (Becton Dickinson) and cell viability was measured with the Annexin V Apoptosis Detection Kit I (Becton Dickinson).

For cell isolation and purification by FACS, peripheral blood mononuclear cells (PBMCs) were separated by Ficoll 1.077 (Sigma) and labelled with anti‐CD4, ‐CD13 and ‐CD14 antibodies. Hence, Th cells were gated as CD4⁺ CD13⁻ lymphocytes (CD3‐untouched) sorted with a purity of ≥ 98%. Peripheral blood monocytes were sorted as CD13⁺ CD14⁺ cells. The gating strategy for FACS is shown in the Supplementary material (Fig. S1). For different assay set‐ups, CD4⁺ T cells were re‐collected from the co‐cultures and/or enriched according to their TIM‐3 expression levels (TIM‐3^−/low and TIM‐3^{moderate/high}). Briefly, the purified Th cells used in the co‐culture set‐ups were re‐sorted by FACS using a similar gating strategy to that described above. These cells were specified as ‘back‐sorted’ Th cells. Flow cytometric analyses and cell purification were performed on a FACSAria II sorter (Becton Dickinson).

Co‐cultures

Myeloid leukaemia cells were co‐cultured with purified CD4⁺ T cells (1·25 × 10⁵/ml) at different ratios. Co‐cultures were also established with peripheral blood monocytes and Th cells at a 2 : 1 ratio. Various concentrations of functional grade purified anti‐CD3 (aCD3, HIT3a) (Biolegend) were used in soluble form for mimicking primary antigenic stimulation. Incubations for 6, 12 and 24 hr were regarded as ‘transient‐stimuli co‐cultures’. To establish ‘continuous‐stimuli co‐cultures’, the incubation period was extended to 96 hr and the culture media were refreshed together with aCD3 (25 ng/ml) for every 24 hr (see Supplementary material, Fig. S2). At the end of 96 hr of co‐culture, the absolute number of cells was also counted. In the relevant co‐culture set‐ups, IgG isotype control (MOPC‐21, 10 μg/ml), aCD28 (CD28.2, 2 μg/ml) monoclonal antibodies (Biolegend), or recombinant human CTLA‐4‐Fc (0·5 µg/ml) and/or ICOS‐Fc (1·5 µg/ml) or PD‐1‐Fc (1·5 μg/ml) chimeras (R&D) were added and refreshed daily. The final concentrations of these blocking agents used were predetermined to not significantly change the proliferation capacity of co‐cultured Th cells (see Supplementary material, Fig. S3).

CFSE‐based proliferation assay

Purified CD4⁺ T cells were stained with carboxyfluorescein succinimidyl ester (CFSE, 5 μm; CellTrace^™; Invitrogen, Eugene, OR) or with the cell proliferation dye eFluor670 (5 μm; eBioscience, San Diego, CA) according to the protocols recommended by the manufacturers. These cells were cultured alone or co‐cultured with myeloid cells in the presence or absence of stimulatory antibodies or blocking chimeric proteins (as described in the co‐culture set‐ups). For the re‐stimulation of Th cells back‐sorted from the co‐cultures, following staining with CFSE, these cells were incubated with recombinant human IL‐2 (rhIL‐2, 250 pg/ml) and/or plate‐bound anti‐CD3 (HIT3a) with or without soluble anti‐CD28 (CD28.2, 2 μg/ml; Biolegend) or PBMCs. Following 96 hr of incubation, Th cells were gated and the dilution in CFSE fluorescence (initial mean fluorescence intensity/resulting mean fluorescence intensity) was analysed by flow cytometry.

ELISA and ELISA array

Supernatants were collected from the continuous‐stimuli co‐cultures established with CD4⁺ T cells and myeloid cells or from the TIM‐3^−/lo or TIM‐3^mo/hi CD4⁺ T cells back‐sorted from the co‐cultures (1·25 × 10⁵ cells/ml) that were further stimulated with 0·5 ng/ml PMA (Cell Signaling Technology, Danvers, MA) and ionomycin (0·5 μg/ml; Sigma) for 2 or 24 hr. Supernatants were also obtained from PBMCs (10⁶ cells/ml) stimulated with aCD3 (25 ng/ml) and aCD28 (1 μg/ml) for 24 hr. For ELISA arrays, equal volumes of supernatants obtained from at least three independent experiments were pooled. Arrays were performed with the Human Th1/Th2/Th17 Cytokines Multi‐Analyte ELISArray Kit (SABiosciences, Qiagen, Valencia, CA). In addition, IL‐2, IFN‐γ and TNF‐α levels were individually quantified by commercial ELISA kits (Biolegend). All assays were performed according to the manufacturers’ instructions.

RT‐PCR

RNA was extracted using a QIAamp RNA Mini Kit (Qiagen, Germantown, MD) and treated with RNase‐free DNase (RNA Clean & Concentrator^™, Zymo Research, Irvin, CA). The cDNA was synthesized using a RevertAid™ First Strand cDNA Synthesis Kit (Fermentas, Vilnius, Lithuania). Primer oligonucleotide sequences for PRDM1, forward 5′‐TGTGAATCCAGCACACTCTC‐3′, reverse 5′‐AGCCTTTCTGCAAAGTCCCGAC‐3′; PBX3, forward 5′‐CCAACCTCTATGCTGCAAAGACG‐3′, reverse 5′‐TTCCTCCAAGGCCATCACTG‐3′; TBX21, forward 5′‐GTCCAACAATGTGACCCAGATG‐3′, reverse 5′‐GCAGTCACGGCAATGAACTGGGT‐3′; and β‐actin, forward 5′‐CTGGAACGGTGAAGGTGACA‐3′, reverse 5′‐AAGGGACTTCCTGTAACAATGCA‐3′ were used. The reactions were terminated at a cycle corresponding to the logarithmic amplification phase predetermined. PCR products were separated by electrophoresis using a 2% agarose gel, stained with ethidium bromide, and UV‐excited densitometry analyses were performed with a Kodak Gel Logic 1500 Imaging System (Carestream Health, Rochester, NY).

Statistical analysis

Statistical analyses were performed using the statistical functions in excel 2013 (Microsoft Office Professional Plus, Redmond, WA). The data presented are from at least three independent experiments. Results are expressed as arithmetic mean ± standard deviation (SD). Statistical difference between experimental groups was assessed by analysis of variance followed by post hoc analyses or Student's paired or unpaired t‐test where applicable. An associated P value of < 0·05 was considered to be statistically significant.

Results

Exhaustion‐related inhibitory receptors are up‐regulated during Th responses stimulated by myeloid leukaemia cells

We have previously demonstrated the capacity of myeloid leukaemia cells, which harbour a CD86⁺ sub‐population, to stimulate CD4⁺ T‐cell responses and to induce Th1 differentiation.17 Accordingly, upon co‐culturing with the myeloid leukaemia cell lines HL‐60 or THP‐1 in the presence of stimulatory anti‐CD3 monoclonal antibody (aCD3), CD4⁺ T cells up‐regulated CD25, CD69 and ICOS, and down‐regulated CD127 as early indicators of activation. The inhibitory receptors PD‐1 and CTLA‐4, TIM‐3 (33·8 ± 7·4% with HL‐60, 42·2 ± 8·3% with THP‐1) and LAG3 (13·6 ± 3·3% with HL‐60, 18 ± 8·3% with THP‐1) were also up‐regulated on T cells in 24 hr (Fig. 1a). As the expression of TIM‐3 and LAG3 correlates with the exhaustion phenotype,3 we aimed to determine the effect of different aCD3 concentrations on leukaemia and Th cells co‐cultured for 96 hr. Notably, HL‐60 or THP‐1 cells induced comparable and high levels of T‐cell proliferation and did not hamper the viability of T cells (see Supplementary material, Figs S4 and S5). To provide continuous stimuli and to avoid nutrient restriction, aCD3 was refreshed together with the media every 24 hr (this co‐culture condition is referred to as ‘continuous‐stimuli co‐culture’). At the end of 96 hr, the majority (range 75·8–98·5%) of T cells up‐regulated CD25, ICOS, PD‐1, CTLA‐4 or TIM‐3 in the co‐cultures established with HL‐60 or THP‐1. LAG3 was detected on 38·5 ± 13·2% of T cells (Fig. 1b). CD127 and CD69 were down‐regulated. Critically, the percentage of T cells expressing these markers was not significantly changed among the co‐culture conditions maintained with different aCD3 concentrations (Fig. 1b).

Figure 1.

Expression of activation‐ and/or exhaustion‐associated molecules on CD4⁺ T cells incubated with myeloid leukaemia (ML) cells under aCD3 stimulation. (a) Purified CD4⁺ T cells were co‐cultured with HL‐60 or THP‐1 cells (at a 2 : 1 leukaemia cell : T‐cell ratio, 25 ng/ml aCD3) and the expression of CD25, CD127, inducible T‐cell co‐stimulator (ICOS), CD69, T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3), lymphocyte activation gene 3 (LAG3), programmed cell death 1 (PD‐1) and cytotoxic T‐lymphocyte antigen 4 (CTLA‐4) markers were determined on T cells at 0, 6, 12 and 24 hr of incubation (transient‐stimuli co‐culture). The co‐cultures were established (b) at a 2 : 1 leukaemia cell : T‐cell ratio in the presence of different concentrations of aCD3 or (c) at different leukaemia cell : T‐cell ratios in the presence of a constant aCD3 concentration (25 ng/ml), the antibody was refreshed together with the media for every 24 hr, and after 96 hr T cells were analysed by flow cytometry (continuous‐stimuli co‐culture), (n ≥ 3). (d) Expression of the inhibitory receptors on T cells from the 96 hr co‐cultures with different myeloid leukaemia ratios are shown. Representative overlay histograms out of three independent experiments are given. Iso. control, staining with isotype‐matched control antibody.

Nevertheless, increasing the ratio of leukaemia cells in the co‐cultures resulted in an increased percentage of T cells expressing TIM‐3 and LAG3, whereas the percentage of CD127⁺ or CD69⁺ T cells was reduced (Fig. 1c). Higher levels of the inhibitory receptors were detected on the surface of T cells that were incubated with greater numbers of myeloid leukaemia cells (Fig. 1d). Intriguingly, decreasing both the ratio of leukaemia cells (from 2 : 1 to 0·25 : 1) and the amount of aCD3 (from 25 ng/ml to 6·25 ng/ml) did not hamper T‐cell proliferation but drastically reduced TIM‐3 and/or LAG3 expression (see Supplementary material, Fig. S6).

Compared with the results obtained at 24 hr, at the end of 96 hr continuous‐stimuli co‐cultures, the percentages of T cells positive for TIM‐3, LAG3, PD‐1, CTLA‐4 and ICOS were significantly increased whereas CD127 and CD69 were down‐regulated on T cells (see Supplementary material, Fig. S7). Furthermore, a gating strategy (see Supplementary material, Fig. S8a) based on CFSE dilution was employed to identify T cells with low‐ and high‐proliferation. In terms of the exhaustion‐related markers studied, no statistically significant difference was detected between these low‐ and high‐proliferated T‐cell populations (see Supplementary material, Fig. S8b,c).

Collectively, in the presence of myeloid leukaemia cells providing co‐stimulatory signals for CD4⁺ T‐cell activation and proliferation, exhaustion‐related molecules were up‐regulated on T cells. These findings were also confirmed with the co‐cultures of CD4⁺ T cells and other co‐stimulation‐competent myeloid leukaemia cell lines (Table 1, and see Supplementary material, Table S1). Hence, not underscoring the essential role for aCD3 stimulation, the signals derived from leukaemia cells not only stimulated T‐cell responses but also induced the expression of surface molecules associated with exhaustion.

Table 1.

Up‐regulation of surface molecules related to activation and/or exhaustion on CD4⁺ T cells co‐cultured with different myeloid leukaemia cell lines for 96 hr under continuous stimuli conditions

	Expression on CD4+ T cells (% ± SD)
	TIM‐3	LAG3	PD‐1	CTLA‐4	ICOS	CD25	CD127
Th cells co‐cultured with
KG‐1	79·6 ± 12·9	24·7 ± 13·7	83·6 ± 12·9	74·5 ± 8·3	92·3 ± 4·5	79·0 ± 13·9	26·0 ± 4·6a
Kasu.‐1	91·0 ± 4·7	32·2 ± 4·0	60·2 ± 4·9a,b,c	83·7 ± 2·2	96·4 ± 1·7	98·8 ± 0·3	22·2 ± 7·9b
HL‐60	79·0 ± 12·2	30·1 ± 3·1	97·0 ± 3·8a	72·5 ± 7·2	99·2 ± 0·3	90·7 ± 2·0	25·6 ± 11·2
THP‐1	80·6 ± 10·1	27·8 ± 8·2	98·5 ± 0·8b	76·7 ± 6·5	98·5 ± 1·7	85·8 ± 8·8	17·1 ± 3·5c
U937	80·7 ± 1·9	20·9 ± 8·0	98·8 ± 0·8c	67·0 ± 14·4	96·2 ± 1·8	92·8 ± 5·1	39·7 ± 4·8a,b,c

Kasu.‐1, Kasumi‐1; ^a,b,c P < 0·05

CD4⁺ T‐cell exhaustion is induced in the presence of myeloid leukaemia cells

To check if the exhaustion markers were specifically induced in the presence of myeloid leukaemia cells, CD4⁺ T cells were co‐cultured with CD14⁺ peripheral blood monocytes under the continuous‐stimuli conditions. Upon co‐culturing with monocytes, TIM‐3 and LAG3 were only up‐regulated on 34·2 ± 20% and 9·7 ± 3·2% of T cells, respectively. Compared with the co‐culture with leukaemia cells, the level of CD127⁺ T cells (45 ± 21·3%) remained higher (Fig. 2a,b). The percentages of T cells positive for CD25, ICOS, PD‐1 and CTLA‐4 were comparable with that obtained with leukaemia co‐cultures (Fig. 2a). As expected, CD4⁺ T‐cell proliferation was highly induced with monocytes as well (Fig. 2c) but the absolute number of T cells expressing TIM‐3 and LAG3 did not increase (Fig. 2d).

Figure 2.

Comparative analysis of activation/exhaustion in Th cells co‐cultured with myeloid leukaemia cells or monocytes. (a) Percentage of Th cells carrying CD25, CD127, inducible T‐cell co‐stimulator (ICOS), CD69, T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3), lymphocyte activation gene 3 (LAG3), programmed cell death 1 (PD‐1) and cytotoxic T‐lymphocyte antigen 4 (CTLA‐4) were determined following continuous‐stimuli co‐culturing with representative myeloid leukaemia cell lines HL‐60 or THP‐1, or with CD14⁺ monocytes isolated from peripheral blood. For the specific marker, the mean percentage of CD4⁺ T cells cultured alone under continuous‐stimuli conditions is indicated with the dashed line. (b) Representative flow cytometry dot plots of LAG3/TIM‐3 and CD25/CD127 staining and (c) representative CFSE assay proliferation histograms of T cells obtained from continuous‐stimuli co‐culture experiments can be seen. Mean percentage ± SD values are given on the corresponding gates. (d) The number of TIM‐3⁺ LAG3⁺ T cells at the initiation (0 hr) and end (96 hr) of the co‐cultures are given. (e) interleukin‐2 (IL‐2), tumour necrosis factor‐α (TNF‐α) interferon‐γ and (IFN‐γ) levels were determined in the co‐culture supernatants with ELISA. (n ≥ 3, *P < 0·05, **P < 0·01). Mono., CD14⁺ monocytes; nd, not detected.

Secretion of IL‐2, TNF‐α and IFN‐γ was significantly lower in the co‐cultures with leukaemia cells than in the ones established with monocytes (Fig. 2e). Attenuation in IL‐2, TNF‐α, IFN‐γ and also IL‐17A production in the co‐cultures with leukaemia cells was also verified with an ELISA‐based array system (Fig. 3a). Furthermore, when TIM‐3^mo/hi T cells were purified from the THP‐1 co‐cultures and re‐stimulated with PMA and ionomycin, secretion of these cytokines was not restored. No IL‐10 or IL‐17A was detected (Fig. 3a). In support of these findings, expression of transcription factors, PRDM1, TBX21 and PBX3 that are associated with exhaustion25 was found to be elevated in the T cells isolated from the co‐cultures with myeloid leukaemia cells (Fig. 3b).

Figure 3.

Confirmation of leukaemia‐induced exhaustion in helper T (Th) cells. (a) Supernatants were collected from the continuous‐stimuli co‐cultures of CD4⁺ T cells with monocytes or with THP‐1 cells, or from the back‐sorted T‐cell immunoglobulin and mucin domain‐containing protein 3^{moderate/high} (TIM‐3^mo/hi) T cells stimulated with PMA/ionomycin for 24 hr, or from the peripheral blood mononculear cells (PBMCs) incubated with aCD3/aCD28 for 24 hr (as a native positive control). The samples pooled from three independent experiments were studied in an ELISA‐based array system to determine the cytokines associated with certain Th cell subsets. The graphical output of optical densities and corrected absorbance values higher than negative standard are shown. (b) Expression of exhaustion‐related PRDM1, PBX3 and TBX21 transcription factors in CD4⁺ T cells purified from continuous‐stimuli co‐cultures with HL‐60 or THP‐1 cells or from aCD3‐stimulated or unstimulated control PBMCs was studied by RT‐PCR (n = 3). Results are given as arbitrary units calculated relative to β‐actin gene expression.

Overall, in the presence of myeloid leukaemia cells that provide constitutive co‐stimulatory signals, CD4⁺ T cells specifically up‐regulated TIM‐3 and LAG3 molecules along with an attenuation in IL‐2, TNF‐α and IFN‐γ production. On the other hand, the effector responses of CD4⁺ T cells were not hampered when co‐cultured with peripheral blood monocytes under conditions of continuous stimulation.

Myeloid leukaemia‐induced exhausted Th cells do not have regulatory properties

The markers associated with exhaustion are also found on regulatory T (Treg) cells and the absence of CD127 on CD25⁺ T cells directly correlates with Treg cells.26, 27 Therefore, evidence was collected to distinguish between the exhausted T cells and the regulatory subsets. In the co‐cultures, the T cells were not Treg cells for the following reasons. (i) Only a small population of CD25⁺ T cells was identified with FoxP3 expression (range 1·5–7·6%) (see Supplementary material, Fig. S9). (ii) In contrast to Treg1 or Th3 regulatory subsets,28 CD69 was absent and CD25 was high on these cells (Fig. 2a). (iii) T‐cell proliferation was not suppressed in the co‐cultures (Fig. 2c). (iv) IL‐10 was not produced by TIM‐3^mo/hi T cells and similar levels of transforming growth factor‐β were secreted in the co‐cultures with either leukaemia cells or monocytes (Fig. 3a).

The reluctance of TIM‐3^mo/hi Th cells to proliferate is due to IL‐2 deprivation

T cells that have undergone exhaustion do not respond to re‐stimulation.3 Hence, CD4⁺ T cells recovered from the co‐cultures according to TIM‐3 expression levels were stimulated under various conditions (Fig. 4a). TIM‐3^mo/hi cells secreted significantly less IL‐2 than TIM‐3^−/lo cells (Fig. 4b). Moreover, the proliferative capacity of TIM‐3^mo/hi cells was marginally enhanced upon addition of aCD3 with freshly obtained PBMCs or with aCD28 monoclonal antibody (Fig. 4c). Their proliferation rate was increased (up to five CFSE dilutions) after the addition of exogenous IL‐2 (Fig. 4c,d). It should be noted that under all re‐stimulation conditions, TIM‐3^−/lo Th cells exhibited significantly higher proliferative activity than that of TIM‐3^mo/hi cells (Fig. 4c,d).

Figure 4.

Reactivation capacity of T helper (Th) cells purified from the co‐cultures with myeloid leukaemia cells. (a) CD4⁺ T cells were back‐sorted from the continuous‐stimuli co‐cultures with THP‐1 cells according to their T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3) surface expression levels. The resulting purified Th cells were enriched in TIM‐3^−/lo and TIM‐^3mo/hi populations. (b) Interleukin‐2 (IL‐2) production in purified TIM‐3^−/lo and TIM‐^3mo/hi Th cells under non‐specific stimulation with PMA/ionomycin for 2 hr (left panel) and for 24 hr (right panel) was determined by ELISA (n = 3). (c) Purified TIM‐3^−/lo and TIM‐^3mo/hi Th cells were stained with CFSE and re‐stimulated under various conditions. Overlay histograms are shown. The dashed line indicates the proliferation limit of these Th cells in the absence of re‐stimulation (control). (d) The proliferative activity of TIM‐3^−/lo and TIM‐3^mo/hi cells under various conditions was quantified as CFSE dilution. (n = 3, *P < 0·05, **P < 0·01). nd, not detected; rhIL‐2, recombinant human IL‐2.

Myeloid leukaemia‐derived co‐stimulatory signals, especially CD86, contribute to Th cell exhaustion

To define the effect of co‐stimulatory molecules on the exhaustion of Th cells, CD86 and/or ICOS‐LG were blocked during the co‐culture period. In the absence of CD86 ligation, the percentage of LAG3⁺ and/or TIM‐3⁺ T cells significantly decreased in a range of 25–40% (Fig. 5a,b). Blockade of ICOS‐LG reduced the amount of exhausted T cells, especially in the co‐cultures established with a 0·25 : 1 leukaemia:Th cell ratio (6·25 ng/ml aCD3) (Fig. 5a,b). Addition of PD‐1‐Fc protein into the co‐cultures did not significantly affect the emergence of exhausted T cells (Fig. 5a,b). Simultaneous blockade of both CD86‐ and ICOS‐LG‐mediated co‐stimulation led to 83·6 ± 4·6% reduction in the LAG3⁺/TIM‐3⁺ population (Fig. 5c).

Figure 5.

Effect of leukaemia‐derived co‐stimulatory signals on T helper (Th) cell exhaustion under continuous stimuli. Continuous‐stimuli co‐cultures of CD4⁺ T cells and myeloid leukaemia cells were performed in the presence of recombinant CTLA‐4‐Fc, inducible T‐cell co‐stimulator (ICOS)‐Fc, programmed cell death 1 (PD‐1)‐Fc or control mouse IgG (Iso. IgG). Together with aCD3 and culture media, these agents were also refreshed daily. At the end of 96 hr‐co‐culturing, the percentages of Th cells expressing lymphocyte activation gene 3 (LAG3)/ T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3) exhaustion markers were determined by flow cytometry. (a) Representative contour plots of LAG3/TIM‐3 staining in CD4⁺ T cells co‐cultured with THP‐1 cells are given out of three independent experiments. Mean percentage ± SD values are given on the corresponding gates. (b) Change in the amount of T cells carrying exhaustion markers were calculated in comparison with the results obtained from the control co‐cultures established in the presence of Iso. IgG. (c) The results of co‐cultures established in the presence of both CTLA‐4‐Fc and ICOS‐Fc can be seen. A typical contour plot of LAG3/TIM‐3 expression and a bar graph showing the change in the amount of T cells expressing LAG3 and/or TIM‐3 is given. (d) Representative results from the continuous‐stimuli co‐cultures between Th cells and HL‐60 or THP‐1 cells enriched into CD86⁻ or CD86⁺ sub‐populations [0·25 : 1 leukaemia cell : T‐cell ratio, 6·25 ng/ml (for HL‐60 co‐cultures), 25 ng/ml (for THP‐1 co‐cultures) aCD3]. (e) The percentage of CD4⁺ T cells expressing LAG3 and/or TIM3 in the continuous‐stimuli co‐cultures established with peripheral blood monocytes with or without aCD28 stimulation (2 : 1 leukaemia cell : T‐cell ratio, 25 ng/ml aCD3, 2 μg/ml aCD28), (n = 3, *P < 0·05).

Additionally, in another experimental setting, CD86⁺ or CD86⁻ sub‐populations from THP‐1 or HL‐60 were sorted by FACS and co‐cultured with Th cells. As a result, higher levels of Th cells were induced to express LAG3 and TIM‐3 exhaustion markers in the co‐cultures with CD86⁺ myeloid leukaemia sub‐populations than that obtained with CD86⁻ (Fig. 5d). Moreover, addition of stimulatory aCD28 into the co‐cultures established with peripheral blood monocytes significantly increased the frequency of Th cells expressing LAG‐3 and/or TIM‐3 (Fig. 5e).

Expression of TIM‐3, LAG3 and PD‐1 on CD4⁺ T cells from patients with myeloproliferative disorders

To determine the exhaustion status of Th cells in myeloproliferative disorders, bone marrow aspirates from patients with AML or MDS were studied. In accordance with the literature,15, 16, 20, 21 CD86 and ICOS‐LG expression was widely found on the myeloid blasts (Fig. 6a). Even though heterogeneous results were obtained from the patients (Fig 6b), a subpopulation of CD4⁺ T cells expressing LAG3, TIM‐3 or PD‐1 molecules was detected in the bone marrow (Fig. 6a). Compared with LAG3, PD‐1 and TIM‐3 were more frequently found on these bone marrow‐infiltrating Th cells (Fig. 6a,b). Between AML and MDS patients, there was no statistical difference in PD‐1, LAG3 or TIM‐3 expression on Th cells and in co‐stimulatory molecule expression on CD33⁺ myeloid cells (data not shown). On the other hand, the percentage of LAG3⁺ TIM‐3⁺ Th cells was significantly higher in AML than in MDS (Fig. 6c).

Figure 6.

Lymphocyte activation gene 3 (LAG3), T‐cell immunoglobulin and mucin domain‐containing protein 3 (TIM‐3) and programmed cell death‐1 (PD‐1) expression on T helper (Th) cells in the bone marrow of patients with myeloproliferative disease. (a) The percentages of CD4⁺ T cells positive for LAG3, TIM‐3 and PD‐1 and of CD33⁺ myeloid blasts positive for CD86 and inducible T‐cell co‐stimulator ligand (ICOS‐LG) are plotted for each patient. The short lines indicate mean percentage values for particular markers, [Empty circles, acute myeloid leukaemia (AML) patients; filled circles, myelodysplastic syndrome (MDS) patients]. (b) Flow cytometry histograms are given for representative patient samples with distinct expression levels of LAG3, TIM‐3 or PD‐1 on gated CD4⁺ T cells. (c) Co‐expression of TIM‐3 and LAG3 on AML or MDS patient‐derived Th cells, (*P < 0·05). Representative flow cytometry dot‐plots of two patients showing low and high percentages of TIM‐3⁺ LAG3⁺ Th cells are shown on the right.

Discussion

Upon chronic exposure to tumour antigens, T cells can undergo functional hypo‐responsiveness and fail to eliminate cancer cells.3, 4 As the overall magnitude of transduced signals is decisive in the fate of T cells, co‐stimulatory pathways possess the potential to contribute to exhaustion.3, 6, 7 Our findings showed that the co‐stimulatory molecules CD86 and ICOS‐LG found on myeloid leukaemia cells provoke Th responses; however, upon prolongation of the stimuli, these responder T cells were directed into an exhaustion‐related state. In a basic co‐culture model, we were able to induce Th cell exhaustion in a 96‐hr incubation period. These in vitro‐generated, exhausted Th cells were characterized by the up‐regulation of PD‐1, CTLA‐4, TIM‐3 and LAG3 receptors, the reluctance to proliferate upon re‐stimulation, and by the attenuation of IL‐2, TNF‐α and IFN‐γ production (as summarized in Fig. 7). Moreover, a sub‐population of Th expressing PD‐1, TIM‐3 and LAG3 was identified in the bone marrow aspirates obtained from patients with AML or MDS.

Figure 7.

Schematic demonstration of the major findings. (1) Potent co‐stimulatory molecules [CD86 and inducible T‐cell co‐stimulator ligand (ICOS‐LG)] expressed by myeloid leukaemia cells promote T helper (Th) cell responses (*Current findings and report by Dolen and Esendagli17). (2) Th cells are exposed to constitutive and potent co‐stimulatory signals derived from leukaemia cells. (3) Extended exposure to co‐stimulation facilitates the acquisition of an exhausted state, which is phenotypically and functionally demonstrated in Th cells.

In myeloproliferative disorders, ICOS‐LG and CD86, but not CD80, are frequently detected on dysplastic or leukaemic blasts.15, 16, 20, 21 Intriguingly, the presence of these potent co‐stimulatory molecules on myeloid blasts associates with poor prognosis and disease severity.16, 20, 21 Nevertheless, in many cancers, CD86 and ICOS‐LG have been linked to anti‐tumour immunity.29 A recent study from our group demonstrated the contribution of leukaemia‐derived co‐stimulatory signals to effector Th cell responses.17 Unfavourably, this interaction triggers ‘adaptive resistance’ in leukaemia cells and increases their immune suppression capacity through induction of PD‐1 ligands.18, 19 According to our current findings, the co‐stimulatory molecules expressed in the myeloproliferative disorders can contribute to the induction of T‐cell exhaustion, which serves as an immune dysregulation mechanism in cancer.

Even though exhaustion is largely studied on CD8⁺ cytotoxic T lymphocytes (CTLs), Th1 cells can also lose functional competence under chronic inflammation.11, 12, 30 Exhausted Th1 cells and CTLs share common properties. Similar to exhausted CTLs, when co‐cultured with myeloid leukaemia cells, Th1 cells reduced the production of IL‐2, IFN‐γ and TNF‐α, up‐regulated TIM‐3, LAG3 and PD‐1, and lost their proliferation capacity. Furthermore, the transcription factors PRDM1, TBX21 (T‐bet) and PBX3, which are associated with CTL exhaustion,25 were increased in Th cells purified from the co‐cultures. Cytotoxic responses are facilitated and reinforced through the actions of the Th1 subset; in addition, CD8⁺ T‐cell exhaustion is prevented where the functionality of the Th1 cells is preserved.8, 9 Therefore, the exhaustion of Th1 cells can impair the positive influence of this Th subset on CTL responses.

Exhaustion is preceded by effector immune responses. Exhausted T cells and the T cells at the late phases of activation are identified with the same markers.31 Following the receipt of stimulatory signals, the early activation phase of T cells (which generally occurs during the first 24 hr) is distinguished by the transient expression of surface molecules such as CD69 and CD154.32 Even though CD25, PD‐1, ICOS and CTLA‐4 are also induced as an early event, they remain up‐regulated through the advanced (late) phases of Th cell activation.31, 32 TIM‐3 and LAG3 can specifically indicate the late phase of activation and are used as putative markers of T‐cell exhaustion. Notably, co‐expression of these inhibitory receptors correlates with T‐cell exhaustion.3, 4 TIM‐3 and LAG3 are also found on Treg cells and Th1 cells.26, 33, 34 Therefore, the expression of activation/exhaustion‐related markers alone may not be sufficient to clearly distinguish exhaustion. Assays testing the functional response of T cells must be performed. Correspondingly, under the specific co‐culture conditions described in this study, CD4⁺ T‐cell exhaustion was both phenotypically and functionally demonstrated.

Interleukin‐2 and IL‐7 are critical mediators for T‐cell survival and activation.32, 35 Engagement with IL‐7 enhances T‐cell receptor‐mediated signalling, and primes IL‐2 production and proliferation.35 As observed in the co‐cultures, decrease in CD127 (IL‐7Rα) is accompanied by up‐regulation of CD25 (IL‐2Rα); hence, during effector phases, the dependence of T cells upon IL‐2 is elevated. Alternatively, prolonged T‐cell receptor/CD28 stimulation favours IL‐2‐independent lymphocyte proliferation.36 Nonetheless, under the continuous stimulation provided with aCD3 and leukaemia cell‐derived co‐stimulation, T cells with TIM‐3 and LAG3 expression stopped IL‐2 secretion and became reluctant to proliferate. These cells were CD25⁺; hence, they responded to exogenous IL‐2 and partially regained proliferative activity. Signalling through inhibitory receptors such as CTLA‐4 and PD‐1 interferes with the pathways mediating IL‐2 transcription.37 Similarly, TIM‐3 and LAG3 inhibit T‐cell expansion by blocking the entry of activated T cells into the growth phase.38, 39 The ligands for these inhibitory receptors can be found on myeloid leukaemia cells.17, 32, 38, 39, 40 Therefore, Th cells that are activated under the influence of myeloid leukaemia blasts are prone to inhibitory signals. Nevertheless, in our experimental setting, CD86 and ICOS‐LG, but not PD‐1 ligands, were responsible for the emergence of TIM‐3⁺ LAG3⁺ Th cells.

In mouse models of myeloid leukaemia, TIM‐3⁺ and/or PD‐1⁺ CTLs have been characterized with loss of effector functions.22, 24 Moreover, increased frequencies of PD‐1^hi TIM‐3⁺ T cells deficient in effector cytokines, especially IL‐2, can be predictive for disease relapse following allogeneic stem cell transplantation in AML.30 Here, CD4⁺ T cells were identified with PD‐1, TIM‐3 and/or LAG3 exhaustion markers in the bone marrow of patients with AML or MDS harbouring co‐stimulation‐competent blasts.

To cope with immune responses, tumour cells implement elaborate strategies such as adaptive resistance and induction of T‐cell exhaustion.18, 19 Due to constitutive expression of co‐stimulatory molecules, myeloproliferative diseases are distinguished with their capacity to provoke T‐cell‐mediated immunity.17, 18 Here, it was demonstrated that, in addition to adaptive resistance,17 myeloid leukaemia‐derived co‐stimulation confers functional exhaustion in Th1 cells. Therefore, check‐point blockade and reversal of the reduced functional response of T cells may be promising immunotherapy approaches for co‐stimulation‐competent myeloproliferative diseases.

Disclosures

None declared.

Supporting information

Table S1. Percentage of CD86 or ICOS‐LG positivity in myeloid leukemia cell lines and peripheral blood CD14⁺ monocytes (PB mono.) used (Mean ± SD%).

Figure S1. Gating strategy used for CD4⁺ T cell (CD3‐untouched) and CD14⁺ monocyte purification from PBMCs by FACS. PBMCs were labeled with anti‐CD4‐APC, anti‐CD14‐FITC and anti‐CD13‐PE antibodies and CD4⁺ CD13⁻ lymphocytes or CD13⁺ CD14⁺ monocytes were sorted.

Figure S2. Schematic demonstration of the co‐culture setups for transient‐ and continuous‐stimuli conditions.

Figure S3. Percentage of Th cells that proliferated in the 96 h continuous‐stimuli co‐cultures established with HL‐60 or THP‐1 myeloid leukemia cells (with myeloid cell:Th cell ratio 2:1) in the presence of different concentrations of aCD3 monoclonal antibody (left‐panel) or at different co‐culture ratios (with 25 ng/ml aCD3).

Figure S4. The viability of Th cells in the co‐cultures was determined with annexin V – PI staining.

Figure S5. Determination of Th cell proliferation and TIM‐3/LAG3 expression under two different 96 h continuous‐stimuli co‐culture conditions.

Figure S6. Comparison between the percentages of Th cells expressing activation‐ and/or exhaustion‐related surface markers that were determined in the 24 h transient‐stimuli and in the 96 h continuous‐stimuli co‐cultures (THP‐1 or HL‐60 cell:Th cell, 2:1; 25 ng/ml aCD3), *P < 0·05, **P < 0·01.

Figure S7. Expression of surface markers associated with activation and/or exhaustion on T cells that have undergone high or low proliferation in the continuous‐stimuli co‐cultures with HL‐60 or THP‐1.

Figure S8. Representative flow cytometry dot‐plots (upper panel) and percentage bar histograms (lower panel) showing CD25 and FoxP3 staining in CD4⁺ T cells co‐cultured with HL‐60, THP‐1 myeloid leukemia cells or with CD14⁺ monocytes obtained from healthy individuals (myeloid leukemia cell:Th cell 2:1, 25 ng/ml aCD3, continuous‐stimuli cultures).

Figure S9. Representative CFSE‐based proliferation assay flow cytometry histograms obtained from the continuous‐stimuli co‐cultures of THP‐1 and Th cells in the presence of isotype IgG (Iso. IgG), recombinant human CTLA‐4‐Fc, ICOS‐FC or PD‐1‐Fc proteins.