Immunoregulation and Clinical Implications of ANGPT2/TIE2⁺ M-MDSC Signature in Non–Small Cell Lung Cancer

Myeloid-derived suppressor cells (MDSC) are heterogeneous populations of immature myeloid cells that pathologically accumulate in many cancers (1). MDSCs are classified into polymorphonuclear (PMN-MDSCs) and monocytic MDSCs (M-MDSC) that are phenotypically and morphologically similar to neutrophils and monocytes, respectively (2). MDSCs promote tumor progression and metastases, leading to a poor prognosis associated with their accumulation in various cancers (3, 4). In addition, MDSCs coordinate cancer cell plasticity and activate signaling pathways associated with drug resistance and angiogenesis (5–7). Although MDSCs are implicated in the inhibition of various immune effector cells, their main targets are antitumor T cells (8, 9). Indeed, immunosuppressive mechanisms reported, among others, include l-arginine deprivation–induced CD3 ζ chain downregulation and prevention of IL2R expression as well as production of reactive oxygen species (ROS), IL10, and TGFβ (10, 11).

The tyrosine kinase receptor TIE2 is primarily expressed on endothelial cells and is capable of binding to angiopoietin-2 (ANGPT2), a proangiogenic factor overexpressed in the tumor microenvironment (12, 13). TIE2-expressing monocytes/macrophages (TEM) have been described as subsets of peripheral and tumor-infiltrating myeloid cells and are found both in humans and mice (14–16). TEM exposure to ANGPT2 augments their ability to stimulate angiogenesis and to promote tumor growth (17, 18). In murine models, activation of the ANGPT2/TIE2 axis promotes immunosuppression in tumor microenvironment through multiple mechanisms (19, 20). However, expression of TIE2 on MDSC phenotypes from patients with cancer has not yet been investigated. Furthermore, the immunomodulatory roles of ANGPT2 are not as well understood as the immunomodulatory roles of VEGF, another regulator of tumor angiogenesis (21).

In this study, we analyzed the expression, distribution, and immune-suppressive role of different types of peripheral M-MDSCs in a cohort of patients with non–small cell lung cancer (NSCLC). We found an excess of circulating M-MDSCs overexpressing TIE2 related to the concentration of ANGPT2 in blood. We showed that the upregulation of the ANGPT2/TIE2⁺ M-MDSC signature in peripheral blood negatively affected patients' clinical outcome. Our results demonstrated the ability of an ANGPT2-rich environment to suppress preexisting T-cell responses against tumor-associated antigens (TAA) through a mechanism involving TIE2⁺ M-MDSCs. Thus, high concentration of ANGPT2 combined with increased numbers of TIE2⁺ M-MDSCs in blood signal a poor prognosis for patients with NSCLC through mechanisms in which antitumor T-cell responses are suppressed.

Patients with NSCLC were enrolled from 2010 to 2014 at the University Hospital Georges Pompidou (Paris, France) and University Hospital of Besançon (Besançon, France) in the TELOCAP 01 cohort, a prospective study of antitumor T-cell immunity in lung cancer (N°EUDRACT: 2009-A00642-55). Blood samples from patients with stages I to IV cancer were included before any anticancer therapy. All patients were enrolled with informed consent in accordance with French laws and after approval by the local and national ethics committees. The main clinical characteristics of the patients are summarized in Supplementary Table S1. PBMCs were isolated by density gradient separation on Ficoll Unisep tubes (Eurobio) and frozen until use. Information about patients' outcome was collected 1 and 2 years after their inclusion. Blood cells were also collected from 34 anonymous healthy donors (HD) at the Etablissement Français du Sang (EFS, Besançon France) with informed consent signature and following EFS guidelines.

To discriminate live from dead cells, PBMCs were first washed in 1× PBS (Gibco) and stained with eFluor 506 viability dye (eBiosciences) according to the manufacturer's instructions. For MDSC analysis, samples were surface-stained in the dark for 30 minutes at 4°C with a mixture of the following antibodies: PerCP-Cy5.5 anti-human HLA-DR (BioLegend, clone L243), BV421 anti-human CD14 (BioLegend, clone HCD14), APC anti-human CD33 (Becton Dickinson, clone WM53), PE-Cy7 anti-human CD11b (Becton Dickinson, clone ICRF44), and PE anti-human TIE2 (BioLegend, clone 33.1 Ab33). Lineage cocktail (lin) was composed of anti-human CD19 APC Alexa Fluor 750 (Beckman Coulter, clone J3-119), CD56 APC Alexa Fluor 750 (Beckman Coulter, clone N901), and CD3 APC Alexa Fluor 750 (Beckman Coulter, clone A94680). The following isotype controls were used for anti-CD11b: PE-Cy7 mouse IgG1, (Beckton Dikinson, clone MOPC-21), for anti-CD33: APC mouse IgG1 (Beckton Dikinson, clone MOPC-1), and for anti-TIE2: PE mouse IgG1 BioLegend, clone MOPC-1) antibodies (see also Supplementary Table S2).

For intracellular cytokine secretion assay (ICS), CD4⁺ T-cell clones were cultured in the presence of TERT peptide (UCP2, 5 μg/mL) and GolgiPlug (BD Biosciences) for 12 hours at 37 °C in 5% CO₂. Then, CD4⁺ T cells were surface-stained with BV421 anti-human CD4 (Becton Dickinson, clone RPA-T4) and APC Fire 750 anti-human CD3 (BioLegend, clone UCHT1). Then, intracellular cytokine labeling was performed following fixation and permeabilization using the BD Cytofix/Cytoperm kit, according to the manufacturer's instructions. After permeabilization, PE anti-human TNFα (BioLegend, clone MAb11) and APC anti-human IFNγ (Becton Dickinson, clone B27) antibodies were added for 30 minutes at 4°C and washed. Samples were acquired on a FACS BD Canto II (BD Biosciences) and analyzed with FACSDiva software (BD Biosciences).

To estimate the absolute number of TIE2 receptors expressed by each M-MDSC phenotype, PE-conjugated Quantibrite beads were used (BD Biosciences). Quantibrite tubes contain lyophilized beads conjugated with four determined concentrations of PE fluorochrome (low: 474 PE, medium-low: 5359, medium-high: 23843 and high: 62 336 PE molecules/beads). After reconstitution, Quantibrite tubes were loaded on the FACS BD Canto II cytometer, FSC/SSC were adjusted and a calibration curve was plotted and fitted by a linear regression according to the manufacturer's instructions. Then, circulating M-MDSCs from 5 healthy volunteers and 8 patients with NSCLC were stained as previously described and acquired. Mean fluorescence intensity values of TIE2 staining were converted in the number of TIE2 receptors per MDSC phenotype. Samples were acquired on a FACS BD Canto II (BD Biosciences) and analyzed with FACSDiva software (BD Biosciences).

Cells were collected in RLT buffer (Qiagen) and total mRNAs were extracted using RNAeasy Mini Kit according to the manufacturer's instructions (Qiagen). Total mRNA was reverse transcribed using the TaqMan gene expression assay Hs00945142_m1 (Thermo Fisher Scientific) and the CFX96 Real-Time PCR Detection System (Bio-Rad). Colo320 colon cancer cell line (RRID:CVCL_0219) and HT29 colorectal carcinoma cells (RRID:CVCL_0320) were used as positive controls for TIE2 expression. A549 lung carcinoma cells (RRID:CVCL_0023) were used as negative controls. All these cell lines were purchased at ATCC, cryopreserved, and routinely tested for Mycoplasma with Mycoalert detection kit (Lonza). Upon receipt, each cell line was expanded and stock vials were frozen. Each cell line was cultured no longer than one week before thawing a cryotube from the original stock.

A previously described mixture of eight pan MHC class II-restricted peptides derived from telomerase (22, 23) as well as a mixture of overlapping 15-mer peptides derived from NY-ESO1 and from Wilms tumor 1 (WT1) were used to monitor circulating T-cell responses against TAAs. To assess antiviral T-cell immunity, peptide mixtures derived from influenza virus, Epstein–Barr virus and cytomegalovirus were used (PA-CEF-001). TERT-derived peptides were purchased from JPT (purity >80%) and all other peptides were purchased from CTL (Cellular Technology Ltd).

T-cell responses were assessed by IFNγ ELISpot assay after a short in vitro stimulation as described previously (22, 23). Ficoll-isolated PBMCs were plated at 4 × 10⁶ cells/well for 6 days in 24-well plates containing 5 μg/mL of the peptide mixtures derived from TERT, NY-ESO-1, WT1, and CEF. Recombinant interleukins, IL7 (5 ng/mL; PeproTech) and IL2 (20 UI/mL; Novartis) were added at days 1 and 3, respectively. At day 7, specific T-cell responses were measured by IFNγ ELISpot according to the manufacturer's instructions. Briefly, cells were incubated at 1 × 10⁵ cells/well in X-Vivo 15 medium (Lonza) in a 96-well ELISpot plate with the relevant peptides for 15 hours. Cells cultured with medium or phorbol myristate acetate (PMA, 1 ng/mL)/ionomycin (500 ng/mL) were used as negative and positive controls respectively. Spots were revealed and spots forming cells were counted using the C.T.L Immunospot System (Cellular Technology Ltd). Responses were considered as positive when IFNγ spot numbers were twice higher than medium control and > 10.

IL6, IL10, and TGFβ, production was measured in patients' serum using Cytometric Bead Array (CBA) Flex Set (BD Biosciences). ANGPT2, VEGF-A, and prostaglandin E2 (PGE2) serum concentrations were assessed by ELISA assay (R&D Systems and Thermo Fisher Scientific, respectively).

PBMCs from patients with treatment-naïve metastatic NSCLC with TIE2⁺ M-MDSCs were assessed for T-cell responses in the absence or presence of exogenous ANGPT2. Briefly, PBMCs were cultivated with or without 300 ng/mL of recombinant ANGPT2 (R&D Systems) and 5 μg/mL of TERT and NY-ESO-1–derived peptides. T-cell responses were assessed by IFNγ ELISpot as described above. In some experiments, PBMCs from untreated patients with metastatic NSCLC were depleted of TIE2-expressing CD11b⁺ cells using CD11b⁺ Macs Magnetic Microbeads Kit and separator according to the manufacturer's instructions (Miltenyi Biotec). Cells recovered from both depleted and undepleted fractions were stimulated in vitro with TERT-derived peptides (5 μg/mL). Antigen-specific T-cell responses were measured by IFNγ ELISpot as described above.

CD14⁺ HLA-DR⁻ M-MDSCs were magnetically sorted from HD PBMCs by using CD14⁺ negative selection and HLA-DR positive selection kits according to the manufacturer's protocols (Miltenyi Biotec). Briefly, CD14⁺ monocytes were first isolated by using a pan monocyte CD14⁺ negative selection kit. The negative faction containing CD14⁺ monocytes was collected and a positive HLA-DR selection was performed. The negative fraction containing CD14⁺ HLA-DR⁻ MDSCs was then collected. Next, TIE2 expression was assessed on sorted M-MDSCs as described above and coculture experiments were performed on TIE2^low and TIE2^high M-MDSCs. Then, TIE2 low or high CD14⁺ HLA-DR⁻ M-MDSCs were incubated for 3 hours in the presence or absence of recombinant ANGPT2 (300 ng/mL) and cocultured for 12 hours with 3 × 10⁴ anti-TERT CD4⁺ T-cell clones (ratio 1:1) in a 96-well plate. Anti-TERT T-cell clones were then assessed for their intracellular production of IFNγ and TNFα. Percentages of inhibition were calculated as:

[(% IFNγ⁺ CD4⁺ T-cell or TNFα⁺ CD4⁺ T-cell clones stimulated with TERT-derived peptide alone) − (% IFNγ⁺ CD4⁺ T-cell or TNFα⁺ CD4⁺ T-cell clones in the other tested wells) × 100]/(% IFNγ⁺ CD4⁺ T-cell or TNFα⁺ CD4⁺ T-cell clones stimulated with TERT-derived peptide alone).

Statistical analysis was performed using Prism 6 GraphPad Software. For two-group comparisons, the nonparametric Student t test (Mann–Whitney U test) was used. Frequency (percentage) was provided for the description of categorical variables. Proportions were compared using the χ² test (or Fisher exact test, if appropriate). Hierarchical clustering analysis was performed using the online Morpheus software (https://software.broadinstitute.org/morpheus/). Briefly, percentages of circulating M-MDSC and ANGPT2 concentrations were tabulated in a (.txt) file and loaded in Morpheus software. Data were normalized with log₂ and robust Z-score according to the software guidelines. Clustering analysis was performed in row and columns according to the intensity of the scores calculated. For survival analysis, according to the amounts of TIE2⁺ M-MDSCs and ANGPT2, we determined thresholds using the upper quartile of the median rate of each TIE2⁺ M-MDSC phenotype (M-MDSC1: 16.7%, M-MDSC2: 6.7%, M-MDSC3: 5.9, M-MDSC4: 7.1% and M-MDSC5: 9.4%) or of ANGPT2: 3 ng/mL. Overall survival (OS) was calculated from the date of study enrollment to the date of death from any cause. Patients known to be alive were censored at the time of their last follow-up assessment. OS was estimated using the Kaplan–Meier method, described using median or rate at specific time points with 95% confidence intervals (95% CI), and compared among the groups with the log-rank test. All tests were two sided, and P values lower than 0.05 were considered statistically significant (*, P < 0.05, **, P < 0.01, ***, P < 0.001, ****, P < 0.0001).

In this study, we analyzed five phenotypes of circulating M-MDSCs in patients with treatment-naïve NSCLC (n = 122) and in healthy, control, donors (n = 34). These M-MDSC phenotypes were described previously (24): M-MDSC 1 (HLA DR^low CD14⁺), M-MDSC 2 (lin⁻ HLA-DR^low CD33⁺), M-MDSC 3 (lin⁻ HLA-DR^low CD11b⁺), M-MDSC 4 (lin⁻ HLA-DR^low CD33⁺ CD11b⁺), M-MDSC 5 (lin⁻ HLA DR^low CD11b⁺ CD33⁺ CD14⁺; Fig. 1A). We showed that all the five phenotypes of circulating M-MDSCs from patients with NSCLC and HDs expressed TIE2 (Fig. 1B). However, the percentage of circulating TIE2⁺ M-MDSCs was 2 to 3 times higher in patients than HD so that only patients with NSCLC had greater than 20% of these cells in the peripheral blood (Fig. 1B). The intensity of TIE2 expression was significantly higher in M-MDSCs from NSCLC patients than on those from healthy subjects (Fig. 1C). To confirm the overexpression of TIE2 on patients' M-MDSCs, we quantified TIE2 mRNA expression in sorted HLA-DR⁻ CD14⁺ M-MDSCs (corresponding to M-MDSC 1) and showed that M-MDSC1 from patients express 1.5–2 times more TIE2 mRNA than those from HD. In addition, TIE2 receptor density on M-MDSCs measured by Quantibrite assay also showed that the number of TIE2 receptors on M-MDSCs was higher in patients with NSCLC compared with HD especially on M-MDSC 4 and M-MDSC 5, the more phenotypically characterized subsets (Supplementary Fig. S1).

Unsupervised hierarchical clustering analysis showed that these circulating TIE2⁺ M-MDSCs overlapped in patients with NSCLC and segregated in three categories of low, intermediate, and high TIE2⁺ M-MDSC expression (Fig. 1D). Healthy subjects presented low and intermediate profiles but the cluster containing high percentages of TIE2⁺ MDSC was lacking (Supplementary Fig. S1). Thus, overexpression of TIE2 represents a feature of circulating M-MDSCs in patients with NSCLC.

Next, we investigated the relationship between ANGPT2, the ligand of TIE2, and M-MSDC accumulation. In line with the literature (25, 26), concentration of ANGPT2 in serum was higher in patients with NSCLC than in healthy volunteers; amounts were significantly elevated in metastatic patients (Fig. 2A and B). As shown in Fig. 2C, patients segregated into three distinct profiles: those with ANGPT2^low/TIE2^low M-MDSCs, ANGPT2^low/TIE2^high M-MDSCs, or ANGPT2^high/TIE2^high M-MDSCs. Upregulation of the ANGPT2^high/TIE2^high M-MDSC signature characterized patients with metastatic disease (Fig. 2C and D). These observations suggest that ANGPT2 might be involved in TIE2⁺ M-MDSC expansion in advanced stages of NSCLC.

Antitumor T cells are a target for MDSCs in the tumor microenvironment (10, 11). Therefore, we investigated the impact of the ANGPT2/TIE2⁺ M-MDSC axis on preexisting T-cell responses against TAA in patients. For this purpose, we evaluated blood T-cell reactivity against two TAAs, NY-ESO-1, and telomerase (TERT), by IFNγ ELISpot assay. The proportion of patients displaying preexisting T-cell responses against either TERT or NY-ESO-1 averaged about 35% in this cohort (Fig. 3A). In contrast, we observed that concentration of ANGPT2 was significantly lower in patients exhibiting functional anti-TAA T-cell responses, and mainly in patients with both TERT and NY-ESO-1–specific T cells (Fig. 3B and C). Similar results were obtained using a third TAA model, Wilms tumor-1 (WT1; Supplementary Fig. S2). When ANGPT2 serum concentrations increased, the percentage of patients responding against TAAs decreased, dropping down to 4% in patients exhibiting concurrent T-cell responses against TERT and NY-ESO-1 (Fig. 3D and E). No relationship was found between ANGPT2 serum concentrations and antiviral T-cell responses, which were concurrently measured in all patients (Fig. 3F).

Next, we addressed the immunomodulatory effect of ANGPT2 against anti-TAA T-cell responses in vitro. To this end, we selected patients with untreated metastatic NSCLC presenting high percentages of TIE2-expressing M-MDSCs and cultured their PBMCs with TERT or NY-ESO-1–derived peptides in the presence or absence of exogenous ANGPT2 (Fig. 4A). Addition of recombinant ANGPT2 during in vitro stimulation abolished IFNγ production in response to TAAs (Fig. 4B and C; Supplementary Table S3). This result was confirmed in melanoma and ovarian cancers (Supplementary Fig. S3).

To assess whether this inhibitory effect involved TIE2⁺ M-MDSCs, we added GW3965 LXR agonist to inactivate M-MDSCs during in vitro stimulation. LXR agonists selectively deplete MDSCs both in vitro and in vivo (27). We showed that the ANGPT2-inhibitory effect was impaired when MDSCs were inactivated by the addition of the GW3965 LXR agonist (Fig. 4D and E). To further evaluate the involvement of TIE2⁺ M-MDSCs in anti-TAA T-cell inhibition, we removed these cells before blood lymphocyte stimulation with TAA-derived peptides. Because TIE2 is mainly expressed on CD11b⁺ cells in peripheral blood, CD11b microbeads were used for TIE2⁺ M-MDSC depletion (Fig. 4F). We found that removal of TIE2-expressing M-MDSCs before in vitro stimulation restored or significantly increased anti-TAA T-cell responses in 6 of 13 patients with untreated metastatic NSCLC (Fig. 4G). On the other hand, we detected in the sera of patients exhibiting an ANGPT2^high/TIE2^high M-MDSC signature, high concentrations of immune-suppressive cytokines such as TGFβ, VEGF-A, IL6, and PGE2, but not IL10 (Supplementary Fig. S4). Collectively, our results indicate that the activated ANGPT2/TIE2⁺ M-MDSC axis inhibits T-cell responses directed against TAAs.

To scrutinize the interaction between the ANGPT2/TIE2⁺ M-MDSC axis and the anti-TAA T-cell responses, we cocultured T cells and TIE2⁺ M-MDSCs. We selected HDs with low or high percentages of TIE2⁺ M-MDSCs and sorted HLA-DR⁻CD14⁺ M-MDSCs (referred to as phenotype 1) from their PBMCs according to a previously described method (28). Some of these sorted M-MDSC 1s were then preexposed to recombinant ANGPT2 before coculture with an anti-TERT CD4⁺ T-cell clone (Fig. 5). This CD4⁺ clone is a high avidity Th1-polarized CD4⁺ clone reactive against TERT (22, 29) that did not express TIE2 even during coculture with M-MDSCs (Fig. 5A). As depicted in Fig. 5B, in the presence of TIE2^high M-MDSCs, we observed a decrease of IFNγ and TNFα production by the CD4⁺ clone. When TIE2⁺ M-MDSCs were pretreated with ANGPT2, their ability to inhibit the IFNγ and TNFα production in response to cognate peptide doubled (Fig. 5B and C). In contrast, the effect of ANGPT2 on M-MDSCs was impaired in the context of low TIE2 expression. Recombinant ANGPT2 had no effect on the production of cytokines by the CD4⁺ clone (Fig. 5D–F). Thus, an ANGPT2-rich environment increases the inhibitory functions of M-MDSCs overexpressing TIE2.

In this cohort of patients, median OS in advanced NSCLC previously treated with chemoradiation or chemotherapy was around 13 months. For the 5 advanced patients who received tyrosine kinase inhibitor, their median OS was 23 months. For patients with localized disease treated by surgery, median OS at 2 years was not reached. These observations support the clinical outcomes already reported in NSCLC (30).

In line of the literature (26, 31), we found that a higher amount of serum ANGPT2 was significantly associated with poor prognosis both in patients with localized and metastatic NSCLC (P < 10⁻⁴; Fig. 6A).

Our results showed that high numbers of MDSCs of all TIE2⁺ phenotypes negatively affected patients' OS, especially in metastatic NSCLC. The median OS was 9 months for patients with TIE2^high M-MDSCs, as compared with 16 months for patients with TIE2^low M-MDSCs (Fig. 6B; Supplementary Fig. S5). In contrast, when the association with clinical outcome was analyzed regardless of TIE2 expression, only high numbers of MDSCs with phenotypes of M-MDSC 1 and M-MDSC 5 were associated with poor prognosis (Supplementary Fig. S5).

Next, we addressed the clinical significance of the ANGPT2/TIE2⁺ M-MDSC axis in this cohort. We found that patients exhibiting the ANGPT2^low/TIE2^low M-MDSC signature in blood (who did not reach median OS) had better clinical outcomes than patients belonging to the group with ANGPT2^high/TIE2^high M-MDSCs (median survival of 8 months; Fig. 6C). We also distinguished an intermediate prognostic group of patients (median OS around 14 months) with ANGPT2^high/TIE2^low M-MDSC profile supporting the requirement of an active ANGPT2/TIE2⁺ MDSC axis.

Collectively, our results highlight the ANGPT2/TIE2⁺ M-MDSC signature as a participant in tumor immune evasion in which release of high amounts of ANGPT2 from the tumor microenvironment targets circulating TIE2^high M-MDSCs and sensitizes them to suppress anti-TAA T cells (Fig. 6D).

The presence of MDSCs with immunosuppressive activity is common in human cancers (10, 32). There is growing interest for monitoring circulating MDSCs in patients with cancer and efforts have been made to harmonize the parameters that characterize these heterogeneous immune-suppressive cells (24, 33). Here, we analyzed in NSCLC five phenotypes of circulating M-MDSCs and studied their expression of TIE2, the ANGPT2 receptor. A subset of TIE2-expressing monocytes/macrophages called TEMs has been described in various human cancers (15, 16, 18). In this report, we found that all the five phenotypes of M-MDSCs studied overexpressed TIE2 and overlapped both in healthy subjects and patients with NSCLC. Patients exhibited higher circulating rates of TIE2⁺ M-MDSCs as compared with healthy individuals. The frequency of TIE2⁺ M-MDSCs was related to the amount of the ligand ANGPT2 in peripheral blood. The ANGPT2^high/TIE2^high M-MDSC signature was preferentially detected in metastatic patients, correlating tumor dissemination with accumulation of these cells.

We showed that the prognostic value of M-MDSCs with or without TIE2 expression appeared different in NSCLC. Without TIE2 expression, only the phenotypes M-MDSC 1 and 5 affected NSCLC patients' survival, in accordance with previous findings (34, 35). However, with TIE2 expression, all phenotypes of M-MDSCs were associated with poor prognosis, especially in metastatic patients. These observations suggest that TIE2 expression could worsen the prognostic value of circulating M-MDSCs. Indeed, high percentages of TEMs in blood or in the tumor microenvironment are associated to a poor prognosis in various patients with cancer (36, 37). Thus, we believe that overexpression of TIE2 characterizes circulating M-MDSCs regardless of the expression of other markers used for M-MDSC characterization (24).

By combining the percentages of TIE2⁺ M-MDSCs with ANGPT2 concentrations, we stratified patients into three prognostic groups (best, intermediate, and poor). Patients in the group with the best prognosis were characterized by the ANGPT2^low/TIE2^low M-MDSC profile and did not reach median OS. Patients in the group with the worst prognosis were characterized by the ANGPT2^high/TIE2^high M-MDSC profile and has median OS ≈ 8 months. Patients in the group characterized by intermediate survival exhibited the ANGPT2^high/TIE2^low M-MDSC signature. We suggest that ANGPT2 may enhance the protumor function of TIE2⁺ M-MDSCs. Thus, combining both ANGPT2 concentration and TIE2⁺ M-MDSC frequency may predict patient survival in NSCLC.

The ANGPT2/TIE2 axis is involved in blood vessel maturation and has been associated with progression and poor prognosis in various cancers, including lung cancer (17, 18, 25, 32). Because ANGPT2 stimulates the immunosuppressive functions of TEMs (16, 18, 19, 38), we proposed that the poor prognosis associated with high signature of ANGPT2/TIE2⁺ M-MDSCs in blood could be related to the inherent inhibition of antitumor T-cell immunity. To test this assumption, we studied the effect of the ANGPT2/TIE2⁺ M-MDSC axis on preexisting T-cell responses against three shared TAAs: NY-ESO-1, TERT, and WT-1 (39). Our results revealed that the presence of an ANGPT2/TIE2⁺ M-MDSC–rich blood environment was associated with an impaired T-cell response against TAAs in patients with NSCLC.

Previous work shows that T-cell responses against certain TAAs in peripheral blood may influence cancer outcome. For instance, we and others have reported that increased numbers, in peripheral blood, of IFNγ-secreting T cells targeted to NY-ESO-1 and TERT was associated with good prognosis (40, 41), suggesting that a preexisting antitumor Th1 signature in blood may confer tumor protection. In this study, we found that patients exhibiting T-cell responses against all the three TAAs simultaneously had a better prognosis. Hence, one explanation of the poor prognosis associated with the signature of excess ANGPT2/TIE2⁺ M-MDSCs may be related to the inhibitory effect exerted by this signature against TAA-specific T-cell immunity. Our in vitro experiments confirmed this assumption and showed the ability of ANGPT2 to promote inhibition of TAA-specific T cells via TIE2⁺ M-MDSCs.

The immunomodulatory roles in human cancers of ANGPT2 have been less studied than those of VEGF. In mouse models, ANGPT2 promotes recruitment of immune-suppressive cells such as MDSCs and regulatory T cells (Treg), as well as TEMs (19, 42). ANGPT2 also facilitates the migration of immune cells out of the vasculature and into the tumor microenvironment (43). Human TEMs exposed to ANGPT2 can promote Treg expansion via IL10-dependent mechanisms (19). In this study, we observed that an ANGPT2-rich blood environment was related to increased concentrations of immunosuppressive factors such as VEGF, IL6, TGFβ, and PGE2, which could explain the inhibition of T-cell responses against TAA (21, 44–46). However, further studies are needed to determine whether the suppressive mechanisms of the ANGPT2/TIE2⁺ M-MDSC axis depends on direct cell–cell contact or soluble mediators (11).

M-MDSC frequency could be used to predict response to anti–PD-1/PD-L1 therapies (47–49). A report in patients with melanoma showed that ANGPT2 could represent a predictive biomarker for anti–PD-1 treatment (50). Here, by linking MDSC and ANGPT2 through the tyrosine kinase receptor TIE2, we raise the question of the involvement of ANGPT2/TIE2⁺ M-MDSC immunosuppressive signature in the resistance to immune checkpoint inhibitors. Whether this signature is also involved in the resistance to treatments such as chemotherapy and radiation remains to be explored.

In summary, our results indicate that ANGPT2/TIE2 axis activation can potentiate the inhibitory activity exerted by M-MDSCs against antitumor T cells. This study supports the use of TIE2 expression as a phenotypical feature to identify suppressive M-MDSCs and also prompts us to explore the ANGPT2/TIE2⁺ M-MDSC axis as biomarker for cancer immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: E. Lauret Marie Joseph, O. Adotévi

Development of methodology: E. Lauret Marie Joseph, C. Laheurte, K. Asgarov, A. Bouard, M. Dosset, O. Adotévi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Lauret Marie Joseph, L. Boullerot, O. Adotévi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Lauret Marie Joseph, C. Laheurte, M. Jary, M. Dosset

Writing, review, and/or revision of the manuscript: E. Lauret Marie Joseph, C. Laheurte, M. Jary, L. Rangan, M. Dosset, O. Adotévi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Lauret Marie Joseph, E. Gravelin, A. Bouard, L. Rangan, C. Borg

Study supervision: O. Adotévi

We sincerely thank all the patients, the medical doctors, and the nurses from the Thoracic Oncologic Department of University Hospital of Besançon and European Georges Pompidou hospital in Paris for their contributions. The authors also thank the Biomonitoring platform of CIC-1431 and Pr. De Carvalho Marcelo for their technical support. This work was supported by grants from Fonds de Dotation de Recherche en Santé Respiratoire, La Ligue Contre le Cancer, the Conseil Regional de Franche-Comte, and Canceropôle Grand Est.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.