Adoptive Transfer of Anti-Nucleolin T Cells Combined with PD-L1 Inhibition against Triple-Negative Breast Cancer

Triple-negative breast cancer (TNBC) is a subtype of breast cancer characterized with no estrogen, progesterone or EGFRs leading to no targeted therapies. Patients with TNBC have a poor prognosis and high relapse rates (1). Chemotherapy and radiotherapy are the approved treatments for these patients (2). Currently, TNBC shows a high expression of programmed death-ligand 1 (PD-L1) influencing patient prognosis and highlighting the potential for anti–PD-L1 immunotherapy (3).

The conventional dendritic cells (ConvDC) produced from activated monocytes by exogenous cytokines have been shown to have T-cell activation and cancer killing ability, but have several limitations in their production, that is, high cost, poor consistency, and ex vivo low viability (4, 5). To solve these weaknesses, researchers have developed multifunctional DC precursor cells to produce GM-CSF/IL-4 and to present tumor antigens by themselves, called self-differentiated myeloid-derived antigen-presenting cells (APC) reactive against tumors (SmartDC) that can display greater cell viability and immune potency to enhance T-cell expansion and cytotoxicity against cancers than ConvDCs (5–7). Interestingly, a toll-like receptor 4–binding protein, 40S ribosomal protein S3 (RPS3) has recently been shown to induce DCs maturation and activation, followed by increasing IFNγ-producing CD8⁺ T-cell production and tumor destruction in the mouse model (8).

Nucleolin (NCL) is a 100–110-kDa protein that is ubiquitously expressed in exponentially growing eukaryotic cells (9). Although more than 90% of NCL is found in the nucleolus, NCL has been divided into nuclear, cytoplasmic, and cell surface NCL, depending on location in the cells (10). Reportedly, NCL has been targeted in preclinical and clinical trials of many cancers, including renal cell carcinoma (11), acute myelogenous leukemia (12), breast cancer (13), and advanced/refractory cancers with multiple metastatic sites (14). In TNBC, positive cases of NCL have been reported at around 80% (15). NCL is overexpressed and associated with the highly tumorigenic TNBC cells, metastasis, tumor relapse, poor disease prognosis, and evasion of success with standard chemotherapy (16). Inhibition of NCL using aptamers could reduce breast cancer cell aggressiveness both in vitro and in vivo (15). In a current review of literature, no NCL-specific T cells have revealed their effects on TNBC cell killing.

The alterations in autophagy are involved in tumor survival-promoting or tumor-suppressing mechanisms (17). Autophagy impairment such as low expression of BECLIN-1 was strongly associated with poor prognosis and an independent predictor of survival of basal-like and HER2-enriched breast cancers (18). The deficiency in the autophagy pathway has been associated with defects in T cell-mediated killing in TNBC tumors (19). Activation of autophagy significantly suppressed PD-L1 (20). The combination of anti–PD-L1 antibody with an autophagic stimulator bisdemethoxycurcumin showed the reduction of bladder cancer cell viability by the action of T cells (21–23).

In this present study, NCL-specific T cells were produced using SmartDCs and killing capability of these T cells on NCL^High/PD-L1^High TNBC cells was increased by direct targeting PD-L1 using anti–PD-L1 peptide or indirect controlling PD-L1 expression by curcumin, an autophagic stimulator. High NCL and PD-L1 in patients with TNBC revealed a significant correlation with short patient survival time. The production of monocyte-derived SmartDCs-NCL was performed by transduction of healthy donor peripheral blood mononuclear cells (PBMC)-derived monocytes with a GM-CSF/IL-4/NCL tricistronic lentiviral system. RPS3 was added and showed a significantly higher SmartDCs maturation leading to NCL-specific T cell activation and killing of NCL^High TNBC cells. The anti–PD-L1 peptide could sensitize NCL-specific T cells against NCL^High/PD-L1^High TNBC cells. Interestingly, curcumin could suppress PD-L1 production on cancer cells through p62-mediated autophagy leading to the sensitization of NCL-specific T cells to destroy NCL^High/PD-L1^High TNBC cells.

The primary antibodies used for IHC, immunofluorescence (IF), Western blotting (WB), and flow cytometry (FC) are listed: Rabbit polyclonal anti-NCL antibody (IHC 1:50; IF 1:50, WB 1:500, FC 1:50, #14574, Cell Signaling Technology, Inc.) or rabbit anti–PD-L1 antibody (IHC 1:50; IF 1:50, WB 1:1,000, FC 1:50, ab205921, Abcam) or rabbit anti-LC3 (IF 1:100; WB 1:2,000, L7543, Sigma-Aldrich Corporation) or mouse anti-SQSTM1/p62 (IF 1:50; WB 1:500, sc-28359, Santa Cruz Biotechnology, Inc.) or mouse anti-BAX (IF 1:100: WB 1:1,000, #610983, Becton Dickinson Holdings PTE. Ltd.) or rabbit anti–BCL-2 (WB 1:2,000, ab196495, Abcam) or β-actin (WB 1:10,000, sc-47778, Santa Cruz Biotechnology). Horseradish peroxidase (HRP)–conjugated goat anti-mouse (1:2,000) and goat anti-rabbit (1:2,000), both from Bio-Rad, were used as secondary antibodies for WB. Goat anti-rabbit IgG-FITC antibody (1:2,000, ab6717, Abcam) and goat anti-mouse IgG-Cy3-AffiniPure F(ab')2 Fragment (1:2,000, #115–116–071, Jackson Immuno Research Laboratories, Inc.) were used for IF. Hoechst33342, trihydrochloride trihydrate (IF 1:1,000, Invitrogen Corporation), from Thermo Fisher Scientific Inc. was added to stain the nucleus.

The antibodies for phenotypic analysis of monocytes and DCs were bought from ImmunoTools GmbH (Friesoythe, Germany): anti-CD11c APC (1:50, #21487116), anti-CD14 FITC (1:50, #21620143), anti-CD40 FITC (1:50, #21270403), and anti-HLA-DR FITC (1:50, #21278993); anti-CD80 PE (1:50, #12–0809–42), anti-CD83 PE (1:20, #12-0839-42) and anti-CD86 PE (1:20, #12-0869-42; Thermo Fisher Scientific, Inc.). The antibodies for T-cell subsets were purchased from Thermo Fisher Scientific Inc.: FITC-conjugated anti-CD3 (1:50, #48–0037–41), anti-CD4 Alexa Fluor 700 (1:50, #56–0049–42) and anti-CD8 APC-Cy7 (1:50, #A15448); anti-CD45RA APC (1:50, #21279456) and anti-CD62 L PE (1:50, #21819624) from ImmunoTools GmbH.

HLA typing of healthy donors was performed using anti-HLA-A2 APC (1:50, #17-9876-42, ImmunoTools GmbH). Anti–PD-L1 peptide (CLQKTPKQC; ref. 24) and NCL peptides (KMAPPPKEV and VLSNLSYSA) restricted to HLA-A*02 were synthesized from GenScript Biotech PTE. LTD. (Galaxis West Lobby). Curcumin (Merck-Schuchardt) was kindly provided by Prof. Somchai Pinlaor, Khon Kaen University. Spautin-1 (1262888–28–7, Sigma-Aldrich) was kindly provided by Prof. Ciro Isidoro, Università del Piemonte Orientale “A. Avogadro,” Italy.

The human embryonic kidney cells (HEK) 293T (Lenti-X 293T-cell, #632180; Takara Bio USA, Inc.) and human breast cancer cell lines, MDA-MB-231 (#HTB-26) and MCF-7 (#HTB-22) were purchased from the ATCC and cultured in DMEM (Gibco, Thermo Fisher Scientific Inc.) supplemented with 10% FBS (v/v; Gibco), 100 U/mL of penicillin and 100 µg/mL of streptomycin (Sigma-Aldrich). HCC70 (#CRL-2315, ATCC) was cultured in RPMI-1640 (Gibco). MCF-10A (#CRL-10317, ATCC) was cultured in DMEM supplemented with 5% horse serum (Invitrogen Corporation), 100 U/mL of penicillin and 100 µg/mL of streptomycin, 20 ng/mL EGF (PeproTech), 0.5 mg/mL hydrocortisone (Sigma-Aldrich), 100 ng/mL cholera toxin (Sigma-Aldrich), and 10 µg/mL insulin (Sigma-Aldrich). All cells were maintained at 37°C in 5% CO₂ incubator. The cell viability was measured at each passage using a standard trypsinization protocol. Cells were passaged every 3 to 4 days to maintain a confluence rate of 20%–80% and then passaged a maximum of 10 times when experiments were performed. All cell lines used in the study have been routinely tested every 6 months for negative mycoplasma as determined by the Mycoplasma nested PCR primer set (BioDesign Co., Ltd.) throughout the study. The HLA typing of two breast cancer cell lines were followed: MDA-MB-231 (A*02:01, 02:17; B*40:02,41:01; C*02:02,02:17; ref. 25); https://web.expasy.org/cellosaurus/CVCL_0062 and HCC70 (A*30:02,03:01; B*78:01,15:16; C*16:01,16:01; ref. 26); https://web.expasy.org/cellosaurus/CVCL_1270).

NCL and PD-L1 proteins were detected on the paraffin-embedded tumor microarray of 144 cases under the approval of sample collection and usage from Siriraj Institutional Review Board (COA no. Si 580/2018). In brief, 4-µm-thick sections were incubated with primary antibodies overnight in a humidified chamber at 4ºC and then with rabbit Envision⁺System HRP-labeled polymer IHC secondary antibody (K4003, DAKO) for 30 minutes at RT (room temperature). The peroxidase activity was visualized with diaminobenzidine (DAB) solution and counterstained by hematoxylin. The staining proteins were quantitatively scored by scanning the slides with 3DHistech Ltd. CaseViewer/QuantCenter software 2.4.0. (Sysmex) and scored on the basis of the percentages of positive cells: 0%, negative; 1%–25%, +1; 26%–50%, +2; 51%–75%, +3; and 76%–100%, +4. The intensity of protein staining was scored as weak, 1; moderate, 2; and strong, 3. The IHC scores were calculated by multiplying the positive cell score with the intensity score resulting to 0–12 IHC scores, which were used to classify the samples into two groups: low expression levels; scores <6 and high expression levels; sores ≥6.

MDA-MB-231, HCC70, MCF-7 and MCF-10A cell lines were cultured on sterile glass coverslips for 24 hours. Chloroquine (ClQ, C6628–25G, Sigma-Aldrich), an inhibitor of autophagosome degradation (27), was used at 30 µmol/L and added 8 hours before sample collection. The cells were fixed in ice-cold absolute methanol. Non-specific protein binding was blocked with 3% BSA/1X PBS, then incubated overnight at 4°C in a humidified chamber with the indicated primary listed in the above paragraph. After washing, the appropriate secondary fluorescence antibody was applied for 3 hours at RT. The nuclei were stained with Hoechst33342. Fluorescence was captured with a ZEISS LSM 800 confocal laser fluorescence scanning microscope (AxioObserver7, LSM 800, Zeiss).

Cells were lysed with radioimmuno-precipitation assay buffer. A protein assay was performed using the Bradford kit (Bio-Rad Laboratories). A total of 20 µg of homogenate was fractionated by SDS-PAGE and transferred to a polyvinylidene fluoride membrane (Whatman). The membranes were incubated overnight at 4°C with the primary antibody, followed by incubation with the appropriate secondary antibody at RT for 1 hour. β-Actin was used as an internal control. The immunoreactive signals were visualized by Enhanced Chemiluminescence Plus solution (Thermo Fisher Scientific) under Gel Document (Syngene). The bands were quantified by ImageJ version 1.48v (NIH, Bethesda, MD).

Lentivirus-harboring tri-cistronic (RRE-cPPT-EF-1α-mGM-CSF-P2A-mIL4-F2A-mNCL) complementary DNA sequences encoding human GM-CSF, human IL-4, and NCL protein (pCDH-GM-CSF/IL4/NCL; Fig. 2A) were produced as previously described (5, 6, 7). Unrelated protein, phytochrome-based near-infrared fluorescent protein (iRFP) was used as a control. Large-scale production was performed by transient co-transfection of Lenti-X 293T cells as previously described (5). Lentiviral titers were determined by quantifying the Lenti-X p24 Rapid Titer kit by RT-PCR (Clontech Laboratories, Inc.).

PBMCs from HLA-A*02–positive, were isolated with Lymphocyte separating medium (Corning). This present study was approved by Siriraj Institutional Review Board (COA no. Si 580/2018) and conducted in accordance with the Declaration of Helsinki. The written informed consent was obtained from all healthy donors. On day 0, PBMCs (1×10⁶/well) were plated in 12-well plates (Corning). For ConvDCs growth, the adherent monocytes were cultured in AIM-V media (Invitrogen Corporation) with GM-CSF (50 ng/m, ImmunoTools GmbH) and IL-4 (25 ng/mL, ImmunoTools GmbH; ref. 28). Transduction was carried out on day 5 of culture with LV (100 multiplicity of infection) and protamine sulfate (10 µg/mL) added to adherent monocytes and incubated for 16 hours at 37°C, the cells were replaced with AIM-V media to activate T cells. The immature DCs were cultured in AIM-V media with IFNα (50 ng/mL, ImmunoTools GmbH), IFNγ (50 ng/mL, ImmunoTools GmbH), and recombinant human RPS3 (1 µg/mL, Novus Biological) for 2 days to induce mature DCs. At day 7 of culture, the phenotypic analysis of DCs and SmartDCs was compared with monocytes performed by the CytoFLEX Flow cytometer (Beckman Coulter) and analyzed by FlowJo VX software (Tree star, Inc.).

For GM-CSF and IL4, culture supernatant was harvested on day 7 whereas on day 15 for IFNγ. The cytokine levels were measured by ELISA kits (R&D Systems, Inc.) followed the instruction.

SmartDCs-NCL or SmartDCs-iRFP or ConvDCs were used as stimulators for cryopreserved lymphocytes and resuspended in AIM-V medium supplemented with 5% human AB serum (Merck KGaA) at a ratio of lymphocytes to DCs as 10:1 for 3 days. Subsequently, the activated T cells were expanded in AIM-V medium supplemented with 5% human AB serum, IL-2 (20 ng/mL, ImmunoTools GmbH), IL-7 (10 ng/mL, ImmunoTools GmbH), and IL-15 (20 ng/mL, ImmunoTools GmbH) for 5 days. The culture media were replaced every other day.

Activated T cells were harvested on day 15. Immunostaining was performed with commercially available fluorescent antibodies against CD3, CD4, CD8, CD42RA, CD62 L, and respective isotype controls. FC was performed by CytoFLEX (Beckman Coulter).

Specific secretions of IFNγ from T cells in response to stimulator cells were assayed by Human IFNγ ELISpot^BASIC (Enzyme-linked immunospot technique) kit (Mabtech, Inc.) following the manufacturer's instructions. HLA-A*02 T2 cells (purchased from the ATCC) were pulsed with 10 µmol/L of NCL peptides (KMAPPPKEV_15–23 and VLSNLSYSN_488–496, GenScript). Cells were scanned with ELISpot plate reader (BIOREADER 5,000 Fγ, BIOSyS) and counted with CellCounter software (Nghia) version 0.2.1. Phorbol myristate acetate (PMA, P1585, Sigma-Aldrich) and ionomycin (I9657, Sigma-Aldrich) were used as positive controls.

The mWasabi and CMRA fluorescence target cells (1×10⁴) were seeded into 96-well plates (Corning). Effector cells were incubated with target cells at ratios of 1:1, 5:1, and 10:1. After 24 hours, target cells were measured by the Pierce Firefly Luciferase Glow Assay Kit (Thermo Fisher Scientific Inc.) and Lumat LB 9507 Ultra-sensitive Luminometer (Berthold Technologies GmbH & Co. KG.). Cancer cell lysis was calculated as:

The mWasabi and CMRA fluorescence target cells or CellTracker Green CMFDA Dye–labeled target cells were formed in Ultra Low Attachment (ULA) 96-well Round-Bottom plates (Corning) at 1×10⁴ cells/well in 2.5% of Matrigel (Corning). The plates were centrifuged for 3 minutes at 300 × g at 4°C to initiate the formation of spheroids. CellTracker Orange CMRA Dye (C2927, Thermo Fisher Scientific Inc.) or CellTracker Blue CMF2HC Dye (C12881, Thermo Fisher Scientific Inc.) labeled activated effector T cells (E) were added to target cancer cells (T) at (E:T) ratios of 1:1, 5:1, and 10:1. Fluorescence signals were detected by fluorescence inverted microscopy and CellSense Standard program version 1.15 (Olympus). The ImageJ (NIH) software was used to quantify the change in mean fluorescence intensity (MFI) relative to the spheroid inactivated T cells. Video-movie clips of cells killing up to an E:T ratio of 10:1 were performed through 30-frame images using a confocal microscope (Zeiss LSM 800, Carl Zeiss; ×63). The acquisition software was ZEN 2.3 software (Blue edition, 2002–2011).

MDA-MB-231 and MCF-7 (10,000 cells/well) were seeded in 96-well plates (Corning) and cultured overnight. T cells were added to the wells on the next day at the indicated E:T ratios for 24 hours. T cells were then removed and washed with 1X PBS, fixed with absolute methanol, stained with a 0.5% crystal violet solution. The colony number was counted by photometric measurements using CellCounter software (Nghia).

The relation between NCL and PD-L1 and clinicopathological factors were analyzed using the χ² test and Fisher's exact test. Survival curves were analyzed with the Kaplan–Meier method and compared using the log-rank test. Multivariate analysis of prognostic factors was evaluated in a Cox proportional hazards model. Data were presented as mean ± SD. Student t test and one-sided ANOVA followed by Tukey's post hoc test were performed for comparison of 2 samples and more than 2 samples. Statistical analyses were performed using SPSS 20.0 statistics software (SPSS, IBM) and GraphPad Prism version 7.04 software (GraphPad Software). A P value of less than 0.05 was considered statistically significant.

The data generated in this study are available within the article and its Supplementary Data Files. The DNA sequence data have been deposited in GenBank and are available under accession number BankIt2554073 SmartDC-NCL OM792960.

The expressions of NCL and PD-L1 were observed in 144 tumor microarray TNBC cases. The immunohistochemical stain for NCL showed staining pattern in both nucleolar and cytoplasmic parts whereas PD-L1 showed predominantly located in the cytoplasm with some cell membrane localization. Most of the cases had high NCL in the nucleus (109/144, 75%), whereas PD-L1 was detected at both cytoplasmic and outer membrane regions (83/144, 58%; Fig. 1A; Supplementary Fig. S1D). Four groups of samples were shown according to low or high of either NCL or PD-L1, including NCL^Low/PD-L1^Low, NCL^Low/PD-L1^High, NCL^High/PD-L1^Low and NCL^High/PD-L1^High. The percentage of NCL^Low/PD-L1^Low TNBC tissues was 8.3% (12/144), 13.9% (20/144) for NCL^Low/PD-L1^High, 14.6% (21/144) for NCL^High/PD-L1^Low, and the highest was 63.2% (91/144) for NCL^High/PD-L1^High (Supplementary Fig. S1A; Table 1). Consistent with the evidence in TNBC tissues, the TNBC cell lines, MDA-MB-231 and HCC70 revealed higher levels of both NCL and PD-L1 than MCF-7 luminal breast cancer cells and MCF-10A normal mammary epithelial cells with statistical significance (Fig. 1B; Supplementary Fig. S1B and S1C). FC analysis confirmed high membrane PD-L1 in both MDA-MB-231 and HCC70 cells (Supplementary Fig. S1C).

Survival analysis demonstrated that high NCL or high PD-L1 were significantly associated with a short overall survival (OS) time (Fig. 1C; P = 0.003 and Fig. 1D; P = 0.024). Most importantly, the combination pattern of NCL^Low/PD-L1^Low (Fig. 1E, blue line) was associated with a good OS whereas NCL^High/PD-L1^High (Fig. 1E, violet line) was related to poor prognostic clinical outcomes (P < 0.001). The hazard ratio (HR) for death based on level of NCL or PD-L1 was 0.270 [95% confidence interval (CI), 0.148–0.494; P < 0.001] for NCL^Low/PD-L1^Low and HR, 1.227; 95% CI, 1.299–3.341; P = 0.002 for NCL^High/PD-L1^High (Table 1).

The Fisher's exact test indicated significant inverse correlations of PD-L1 and pN stage (P = 0.034), and clinical staging (P = 0.010; Fig. 1F); however, no statistical significance of these variable parameters with the NCL level were found. No significant differences between NCL and PD-L1 staining patterns and clinicopathological features, including age, pT stage, pM stage, LN metastasis, perineural metastasis, and size were observed (Supplementary Tables S1 and S2). The univariate analyses revealed high NCL levels as an independent poor prognostic marker (Fig. 1G; HR, 2.091; 95% CI, 1.405–3.113; P = 0.000), as well as high PD-L1 (Fig. 1G; HR, 1.818; 95% CI, 1.290–2.564; P = 0.001).

Tri-cistronic LV containing GM-CSF, IL-4 and NCL (RRE-cPPT-EF-1α-mGM-CSF-P2A-mIL4-F2A-mNCL; Fig. 2A) was constructed and transfected into Lenti-X 293T cells. The LV system could successfully produce higher NCL expressions than those in the untransfected (Supplementary Fig. S2A). The immunophenotypic characteristics, of monocytes (day 0) or ConvDCs and cells transduced with LV-GMCSF-IL4-iRFP or LV-GMCSF-IL4-NCL (SmartDCs-NCL) or SmartDCs-NCL+RPS3 (day 7) were shown to have dendritic-like morphology (Supplementary Fig. S2B). All activated DCs had downregulation of CD14 and upregulation of DC markers, including CD11c, CD40, CD80, CD83, CD86, and HLA-DR compared with monocytes (Supplementary Fig. S2D). The significantly higher levels of CD40 and CD80 in SmartDCs-NCL treated with RPS3 in comparison with those of SmartDCs-NCL without RPS3 were detected (Fig. 2B). SmartDCs-NCL and RPS3-treated SmartDCs-NCL exhibited the expected significant increment levels of GM-CSF (Fig. 2C) and IL4 (Fig. 2D) compared with monocytes.

The established T-cell subsets after co-culture with ConvDCs, SmartDCs-iRFP, SmartDCs-NCL or SmartDCs-NCL plus RPS3 were analyzed (Fig. 2E; Supplementary Fig. S2C). There were no significant differences in the percentages of CD3⁺/CD4⁺ and CD3⁺/CD8⁺ T cells expanded after being co-cultured with SmartDCs-NCL and SmartDCs-NCL plus RPS3 that markedly increased CD3⁺/CD8⁺/CD45RA^–/CD62L^– effector memory T cells (T_EM; ref. 29) but were decreased in CD3⁺/CD8⁺/CD45RA^–/CD62L⁺ and CD3⁺/CD8⁺/CD45RA⁺/CD62L⁺ characterized by late CD45RA⁺ effector memory T cells (T_EMRA) and native T cells (T_N; ref. 29).

Subsequently, IFNγ was detected by ELISA assay in the culture supernatant when T cells were incubated with different DCs at a ratio 1:10. The T cells activated by SmartDCs-NCL and SmartDCs-NCL plus RPS3 showed significantly higher IFNγ reactivity against MDA-MB-231 (717.24 ± 84.48 pg/mL and 1,037.93±46.55 pg/mL) than that of unactivated T cells, ConvDCs-activated T cells, and SmartDCs-iRFP–activated T cells (Fig. 2F). The similar pattern of results was detected against HCC70 TNBC cells (798.28±31.03 pg/mL and 992.41 ± 89.65 pg/mL; Fig. 2F). No IFNγ secreted from T cells co-cultured with MCF-10A normal breast cells was observed.

To determine whether the expanded T cells were NCL-specific T cells, the HLA-A*02–restricted T2 cells were used as APCs loaded with pooled NCL peptides and then co-cultured with T cells. After 24 hours stimulations, high IFNγ spots were significantly detected in T cells activated by SmartDCs-NCL and SmartDCs-NCL plus RPS3, resulting in about 2.07-fold (P = 0.015) and 1.77-fold (P = 0.045) differences compared with those with no antigen peptide exposure (Fig. 2G and H). The highest number of IFNγ spots was observed in PMA/Ionomycin, a positive control condition.

The combination of NCL-specific T cells and anti–PD-L1 peptide on killing of NCL^High/PD-L1^High MDA-MB-231 and HCC70 TNBC cells was investigated. After incubation for 24 hours with target cancer cells, cell death was measured and calculated as an increased percentage of specific cell lysis of mWasabi and CMRA fluorescence-labeled viable cells (Fig. 3A–C). NCL-specific T cells activated with SmartDCs-NCL (red bar) or SmartDCs-NCL plus RPS3 (blue bar) were significantly increased in the percentage of specific cell lysis at ratios 1:1, 5:1, and 10:1 for MDA-MB-231 cell (Fig. 3A) whereas an only ratio 10:1 of HCC70 was revealed (Fig. 3B). Adding anti–PD-L1 peptide into the co-culture of cancer cells with NCL-specific T cells, the percentages of specific cell lysis of MDA-MB-231 by NCL-specific T cells activated with SmartDCs-NCL plus RPS3 (pink bar) were significantly greater than NCL-specific T cells without anti–PD-L1 peptide (red bar; Fig. 3A). A similar result was detected in the condition of using only NCL-specific T cells with anti–PD-L1 peptide (green bar) compared with SmartDCs-iRFP (black bar) for both MDA-MB-231 and HCC70 cells at ratio of E:T at 10:1. No killing activities of NCL-specific T cells were detected in MCF-10A having no NCL and PD-L1 expressions (Fig. 3C).

3-D cancer spheroids to evaluate the potency of NCL-specific T cells were developed (Fig. 3D–G). The overall sphere growth rate in 2.5% Matrigel over 96 hours is shown in MDA-MB-231, HCC70, and MCF-10A cells (Fig. 3D). The greater antitumor efficacy of T cells against MDA-MB-231 was associated with a decreased MFI at E:T equaled 5:1 of SmartDCs-NCL+anti–PD-L1 peptide (green line, P = 0.049) and SmartDCs-NCL+RPS3+anti–PD-L1 peptide (pink line, P = 0.023) compared with that of unactivated T cells (Fig. 3E). At a 10:1 ratio, MDA-MB-231 cells were killed by T cells activated by SmartDCs-iRFP (black bar) less than those of SmartDCs-NCL+anti–PD-L1 peptide (green line; P = 0.026), as well as T cells activated by SmartDCs-NCL+RPS3+anti–PD-L1 peptide (pink line; P = 0.003; Fig. 3E). Using NCL-specific T cells in combination with only anti–PD-L1 peptide showed better killing capability than T cells alone, but without statistical significance. Interestingly, T cells in combination with RPS3 and anti–PD-L1 peptide had higher killing capability against NCL-positive cancer cells than that of T cells alone (red line; P = 0.016) and of T cells+anti–PD-L1 peptide (P = 0.233; Fig. 3E). In addition, 3-D HCC70 spheroids showed a significant decrease in MFI at a ratio 10:1 of SmartDCs-iRFP (black bar) versus SmartDCs-NCL+anti–PD-L1 peptide (green line, P = 0.041; Fig. 3F). No treatment-associated toxicity was detected in MCF-10A normal breast cells (Fig. 3G). Furthermore, the tumor killing activities of T cells activated by SmartDCs-NCL and SmartDCs-NCL+RPS3 compared with SmartDCs-iRFP and tumor alone were recorded (Supplementary Movies S1–S5). T cells activated by SmartDCs-NCL and SmartDCs-NCL+RPS3 were specifically killed and infiltrated into NCL^High MDA-MB-231 cells.

To know how NCL-specific T cells would affect the death of NCL^High/PD-L1^High TNBC cells in the actions of anti–PD-L1 peptide and an autophagy modulator, the expressions of autophagy markers, LC3 and p62 in breast cancer cells treated with NCL-specific T cells were detected. The intrinsic autophagy levels were higher in the order of MDA-MB-231 > HCC70 > MCF-7 > MCF-10A (Fig. 4A and B). The cellular levels of LC3, p62, and PD-L1 were concomitantly accumulated in chloroquine (ClQ) treatment reflecting the degradation of PD-L1 in high autophagic MDA-MB-231 cells more than those in low autophagic MCF-7 cells (Fig. 4C; Supplementary Fig. S3A–S3C).

The mechanism underlying anti–PD-L1 or autophagy modulator-mediated NCL-specific T cells that induced TNBC death was explored. The induction of autophagy (increasing the LC3-II/I ratio) and apoptosis (increasing the BAX/BCL-2 ratio) was observed in MDA-MB-231 treated with NCL-specific T cells activated by SmartDCs-NCL+curcumin (Fig. 4D, lane 5) > SmartDCs-NCL+anti–PD-L1 peptide (Fig. 4D, lane 4) > SmartDCs-NCL (Fig. 4D, lane 3). The increased autophagy reduced BAX and restored expression of BAX in SmartDCs-NCL+RPS3+curcumin (Fig. 4D, lane 9) was detected more than those in SmartDCs-NCL+RPS3+anti–PD-L1 peptide (Fig. 4D, lane 8) and SmartDCs-NCL+RPS3 (Fig. 4D, lane 7) in both MDA-MB-231 and MCF-7. The quantitated results of these molecules were summarized (Fig. 4F and I). The curcumin stimulated NCL-specific T-cell killing against TNBC cells not only by BAX/BCL-2 induction, but also by PD-L1 degradation through the autophagy formation indicated by the decreased p62 and PD-L1 levels (Fig. 4G and H, lanes 5 and 9).

To test whether autophagy could involve NCL-specific T-cell toxicity, the treatment of spautin-1 that can induce the destabilization of BECLIN-1–dependent autophagy (30) was performed. Reduction of LC3 and BAX was detected that demonstrated the prevention of cell death by NCL-specific T cells along with increased PD-L1 (Fig. 4D and H, lanes 6 and 10) compared with that of NCL-specific T cells combined with anti–PD-L1 peptide (lanes 4 and 8) or curcumin (Fig. 4D and H, lanes 5 and 9). In MCF-7 cells, the levels of LC3-II (Fig. 4F, lanes 5 and 9) and BAX (Fig. 4I, lanes 5 and 9) were increased only in the cells cultivated in the combination of curcumin with NCL-specific T cells activated by SmartDCs-NCL (Fig. 4F, lane 5) and SmartDCs-NCL+RPS3 (Fig. 4F, lane 9). Of note, no significant differences of p62 expression between treatments were found (Fig. 4G, lanes 5 and 9).

The clonogenic assay confirmed the effect of autophagy modulators–mediated NCL-specific T-cell–induced cell death. Importantly, that spautin-1 could rescue a colony formation of cancer cells co-cultured with NCL-specific T cells demonstrated the reduction of cell availability in a dose-dependent manner to the E:T ratio (Fig. 5A and B; Supplementary Fig. S4A). The double staining demonstrated that apoptosis (BAX-positive) was ensured in the same cells in which autophagy (LC3-positive) was induced in NCL-specific T cells activated by SmartDCs-NCL or SmartDCs-NCL+RPS3 combined with either anti–PD-L1 peptide or curcumin (Fig. 5C). Spautin-1 limited the expression of the pro-apoptotic BAX protein induced by NCL-specific T cells and largely prevented autophagosome formation (Fig. 5C; Supplementary Fig. S4B–S4C). Moreover, on the basis of the CellTracker viability in 3-D spheroids, spautin-1 treatment reversed the killing effect of NCL-specific T cells (Fig. 5D; Supplementary Fig. S4D and S4E).

Many preclinical studies have demonstrated the use of antigen-specific T cells stimulated by DCs as promising immunotherapeutic approaches for the treatment of patients with cancers such as metastatic renal cell carcinoma (31), melanoma (32), cholangiocarcinoma (7), and breast cancer (7). ConvDCs generated from PBMC, however, require a standard method with clinical-grade laboratory cytokines that are costly (4–6). The limited effectiveness of poor consistency and low viability of ConvDCs leading to the insufficient numbers of CD8⁺ cytotoxic T lymphocytes was reported (33). To overcome these limitations, SmartDCs carrying cytokines needed for self-differentiation and antigen-encoding genes for antigen production and intracellular presentation onto the HLA molecules on DCs were introduced (5–7, 32, 34). NCL confirmed its impact as a tumor-associated antigen with its high expression in TNBC samples and revealed a significant correlation with short patient survival time. SmartDCs-NCL was performed and in combination with recombinant RPS3 could significantly induce NCL-specific T-cell activation to kill NCL^High TNBC cells. From the evidence, the anti–PD-L1 peptide or an autophagy modulator curcumin could notably sensitize NCL-specific T cells against NCL^High/PD-L1^High TNBC cells.

There is now growing evidence of the clinical perspective that self-differentiated DCs showed greater viability and immune potency in vitro and in vivo than traditionally generated DCs, in the absence of exogenous cytokines (35, 36). Importantly, T-cell stimulation was more consistent in immunogenic capability of MHC class I–restricted antigen presentation for SmartDCs (37). This evidence supports the currently used SmartDCs-NCL having features of reduction of CD14 and increments of CD11c, CD40, CD80, CD83, CD86, and HLA-DR. On the basis of the recent study that RPS3 was bound to TLR4 and activated DCs (38). To generate the maturated and activated DCs, LPS, a well-known TLR4 ligand, has a significant effect on inducing maturation and activation of DCs; however, LPS is not appropriate for use in humans because it is an endotoxin. Instead, the TLR9 ligand CpG, or the TLR3 ligand I:C, are commonly used, but large amounts are required to induce DC maturation and activation (8, 38). Therefore, to explore a new TLR4 ligand, RPS3 is needed, which is screened from human tumor cells. Moreover, DC vaccination with the newly discovered TLR4 ligand, RPS3, has recently been confirmed as an adjuvant to activate DC maturation (8). RPS3 as an exogenous adjuvant for further SmartDCs maturation showed up-expression of CD40 and CD80, and secretion of GM-CSF and IL-4. The numbers of the effector memory T cells (T_EM, CD3⁺/CD8⁺/CD45RA^–/CD62L^–) and tumor-specific CD8⁺IFNγ⁺ T cells were increased by SmartDCs-NCL or SmartDCs-NCL plus RPS3 activation. The specific effect of NCL-overexpressing tumor cell killing could be achieved using either SmartDCs-NCL or SmartDCs-NCL with RPS3 as an adjuvant.

As a 3-D cell culture, mimic solid tumors in the case of oncological studies (39), spheroids represent a promising technique to improve the function, differentiation and viability as compared with conventional 2-D cultures (40). Here, in vitro multicellular spheres adapted in tumor growth for T cells killing assays are focused on. The cell suspension was cultivated in low-adherent conditions and complete media with 10% FBS to promote extensive proliferation and was devoid of cancer stem cell enrichment (41).

Clinical trials for cancer immunotherapy, including immune checkpoint inhibitor therapy, have been conducted worldwide (42, 43). Therefore, it is expected that the better effect of using PD-L1 modulator together with NCL-specific T cells. Anti–PD-L1 peptide (CLQKTPKQC) has been screened by phage-displayed peptide libraries with a high affinity and efficiency blocks the interaction between PD-L1 and PD-1 in MDA-MB-231 and MCF-7 cells (24). Recently, autophagy has been reported to influence PD-L1 levels (17, 44). The significant overexpression of NCL in patients with TBNC and established the production of NCL-specific T cells aimed to use as a new treatment in patients with TNBC was demonstrated. Most of the NCL-positive cancer cells, however, expressed high PD-L1 that can reduce cancer cell destruction function of T cells. Hence, it is considered of great interest to show that an anti–PD-L1 peptide can help attenuate this effect. Results of this current study confirm the benefit of the combined anti–PD-L1 peptide and NCL-specific T cells over either NCL-specific T cells or anti–PD-L1 peptide alone (Supplementary Fig. S4A). The alteration of the autophagy pathway can stimulate or limit the potent immune responses that have been reported; however, the mechanism has remained unclear (45). Failure of autophagy contributed to a limitation of the T-lymphocyte attack on TNBC (19). Resveratrol, autophagy inducer, pre-treated MDA-MB-231 cells were more sensitive to T-cell–mediated cytotoxicity (19). Treatment with an autophagy inducer, the natural compound “curcumin,” or in combination with anti–PD-L1 antibody attenuated PD-L1 expression and enhanced T-cell responses against bladder cancer (21) and tongue squamous cell carcinoma (46). This evidence supports current findings that SmartDCs-NCL and SmartDCs-NCL plus RPS3-activated NCL-specific T cells could sensitize NCL^High/PD-L1^High MDA-MB-231 cell killing when used in combination with anti–PD-L1 peptide or curcumin.

Treatment with curcumin before SmartDCs-NCL and SmartDCs-NCL plus RPS3 application showed high LC3 and low p62 expression that could be interpreted as an efficient autophagy flux. Chloroquine or ClQ, is a well-established chemical able to interrupt the degradation of autophagosomes, as testified by several scientific reports (27). The degradation of PD-L1 via autophagy pathway has been reported (47, 48). It was herein noticed that the addition of ClQ could inhibit autophagosome degradation; thus, concerning the phenomena here described, the increased expression of PD-L1 ensures that PD-L1 was practically processed in autophagosome accumulation. Hence, PD-L1 can be degraded via the selective autophagy–lysosome pathway that facilitated recognition by p62/QSQTM1, autophagy substrate (47, 48). The ClQ treatment was performed to confirm the effect of PD-L1 degraded by autophagy. The inhibition of autophagy by ClQ was not expected to increase the activity of stimulated T cells; therefore, the condition of additive of ClQ was not required. Moreover, the hyper-induction of autophagy by SmartDCs-NCL, SmartDCs-NCL plus RPS3, SmartDCs-NCL/RPS3/anti–PD-L1 peptide, and SmartDCs-NCL/RPS3/curcumin led to apoptosis of breast cancer cells. In fact, this effect was abolished when autophagy was inhibited by spautin-1, through promoting the ubiquitin-mediated degradation of BECLIN-1 (30). Spautin-1 abrogated the induction of autophagy by SmartDCs-NCL as shown by the lack of formation of LC3-II. As predictable, spautin-1 inhibited autophagy and rescues PD-L1 expression. The current study results provide evidence supporting the potential mechanisms of autophagy-dependent apoptosis of TNBC cells that were obtained with treatment of NCL-specific T cells activated by SmartDCs-NCL. Enhancement of SmartDCs-NCL activity by RPS3 could be greatly increased NCL-specific T cells with high IFNγ production. The degradation of PD-L1–linked ubiquitination of p62 might be critical for a selective autophagy pathway to enhance NCL-specific T cells against NCL^High/PD-L1^High cells. Mechanistically, the reduction of p62 in cancer cells could attenuate PD-L1 induction by autophagy inhibition, implying that autophagy regulates PD-L1 expression through the p62 pathway (47). Notably, induction of autophagy sensitized TNBC cells to NCL-specific T-cell–mediated tumor killing similar to the improvement of anticancer effects of NCL-specific T cells by immune checkpoint inhibitor (Fig. 5E).

In conclusion, autophagy contributes to the anticancer ability of NCL-specific T cells activated by SmartDCs-NCL mechanistically through both apoptotic and autophagic cell death pathways. It is possible to provide clinical approaches by targeting NCL^High/PD-L1^High TNBC by NCL-specific T cells in combination with a PD-L1 inhibitor or autophagic stimulator. These approaches may highlight the potential of using NCL-specific T cells in patients with cancer featured with a high NCL level, and anti–PD-L1 molecules (anti–PD-L1 peptide or PD-L1 suppression through an autophagy stimulator) to sensitize T cells killing functions against NCL and PD-L1–expressing cancer cells.

No disclosures were reported.

S. Thongchot: Conceptualization, data curation, software, formal analysis, investigation, writing–original draft, writing–review and editing. N. Jirapongwattana: Data curation, investigation, methodology. P. Luangwattananun: Data curation, methodology. W. Chiraphapphaiboon: Data curation, methodology. N. Chuangchot: Data curation, methodology. D. Sa-nguanraksa: Resources. P. O-Charoenrat: Resources. P. Thuwajit: Conceptualization, methodology. P. Yenchitsomanus: Conceptualization, resources, methodology. C. Thuwajit: Conceptualization, resources, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

This research project was supported by Mid-Career Research Grant, The National Research Council of Thailand (grant no. RSA6280091), Ministry of Higher Education, Science, Research and Innovation, Thailand (to C. Thuwajit) and New Researcher Grant, Mahidol University (to S. Thongchot). The financial support from Siriraj Research Grant, Faculty of Medicine Siriraj Hospital, Mahidol University and The National Research Council of Thailand 2020. Thanks are due to Prof. Somchai Pinlaor (Khon Kaen University, Thailand) for kind offer of curcumin; Prof. Ciro Isidoro (Università del Piemonte Orientale “A. Avogadro,” Italy) for kind provision of spautin-1. Special thanks to Surat Phumphuang for clinical data collection. The authors would like to thank Prof. James A. Will, USA for the English edition.

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.