A total of 351 cases of paraffin-embedded breast cancer tissues and clinicopathological data (Table I) were collected under Siriraj Institutional Review Board ethical approval (COA no. Si 580/2018). Following antigen retrieval in pH 6 citrate buffer, the samples were stained with anti-human MSLN antibody (cat. no. sc-271540; Santa Cruz Biotechnology, Inc.) at 4°C overnight. Then incubated with goat anti-mouse Envision (cat. no. K4001; Dako; Agilent Technologies, Inc.) for 30 min and HRP activity was detected by Dako-HRP detection kit (Dako; Agilent Technologies, Inc.) before counterstained with Mayer's hematoxylin. The samples were scanned at 400× by Scanscope slide scanner (Aperio Technologies, Inc.). MSLN expression was scored by multiplying the intensity [I; graded: 0 (negative), 1 (weak), 2 (moderate) and 3 (strong)], with the percentage of MSLN positive cells [P; graded: as 0 (no positive cells), 1 (1–25%), 2 (26–50%), 3 (51–75%) and 4 (76–100%)]. The MSLN expression scores calculated by multiplying I and P resulting in score between 0 to 12. The samples were classified into low if the score is less than median (which was 0) and rated as high if the score was equal or more than median.

MDA-MB-231, HCC70 and T2 cell lines were from American Type Culture Collection. Lenti-X™ 293T cells were from Takara Bio (Takara Bio, San Jose, CA). The Lenti-X™ 293T cells, MDA-MB-231 and MSLN-MDA-MB-231, produced by lentiviral transduction, were maintained in DMEM with 10% fetal bovine serum (FBS; Invitrogen; Thermo Fisher Scientific, Inc.). HCC70 cells were cultured in RPMI-1640 with 10% FBS (Invitrogen; Thermo Fisher Scientific, Inc.). The T2 cell line was cultured in RPMI-1640 with 10% FBS and 2 mM L-glutamine (Invitrogen; Thermo Fisher Scientific, Inc.). All cell lines were cultured at 37°C with 5% CO₂.

The construction of lentiviral vector was performed as previously described (8). Briefly, the full-length MSLN cDNA with restriction cut sites was amplified from HCC70 cells and cloned into pCDH1/GM-CSF/IL-4 plasmid. The sequence integrity of MSLN-SmartDC plasmid was confirmed by Sanger's sequencing. A mock control pCDH1/GM-CSF/IL-4 containing an irrelevant red fluorescence protein (IRFP), IRFP-SmartDC, was constructed (8). MSLN-SmartDC lentiviral particle was produced by transfecting the 30 µg of MSLN-SmartDC plasmid along with 21 µg of lentiviral envelop plasmid, 6 µg of pMD2.G and lentiviral packaging plasmid, psPAX2 into 8×10⁶ Lenti-X™ 293T using calcium phosphate method at room temperature (RT). The lentiviral particle in the culture medium was collected 3 days after transfection and concentrated by 20,000 × g ultracentrifugation at 4°C before measured for viral titer using the qualitative (q)PCR Lentivirus Titration kit (Applied Biological Materials, Inc.).

The peripheral blood mononuclear cells (PBMCs) were isolated from 30 ml of HLA-A2⁺ healthy donor blood with written consent under Siriraj Institutional Review Board ethical approval (COA no. Si 580/2018) by density centrifugation at 800 × g for 20 min at RT in lymphocyte separating medium (Corning Life Sciences). The monocytes were isolated and incubated for 1 h at 37°C. The non-adherence cells were collected and cryopreserved in human AB serum (MilliporeSigma) containing 10% dimethyl sulfoxide until use. The monocytes were transduced with IRFP-SmartDC or MSLN-SmartDC at 75 multiplicity of infection (MOI) together with 10 µg/ml protamine sulfate in AIM-V medium (Invitrogen; Thermo Fisher Scientific, Inc.). On day 5 post-transduction, 1 µg/ml of recombinant human RPS3 (cat. no. NBP2-90977; Novus Biologicals, LLC) was added. Monocytes were cultured in 100 ng/ml of rhGM-CSF (cat. no. 11343125; ImmunoTools GmbH) and 50 ng/ml of rhIL-4 (cat. no. 11340045; ImmunoTools GmbH) for five days and treated with 100 ng/ml of rhIFN-γ (cat. no. 11343536; ImmunoTools GmbH) and rhTNF-α (cat. no. 11343015; ImmunoTools GmbH) for additional 48 h served as positive control or conventional DC (conv. DC). All DCs were harvested on day 7 to check the activated characters.

The cryopreserved T cells were thawed and co-cultured with DCs for 48 h at a 10:1 ratio to support T cell activation and proliferation (27) in AIM-V medium at 37°C with 5% CO₂. T cells were expanded in AIM-V with 5% human AB serum (MilliporeSigma), 20 ng/ml of rhIL-2 (cat. no. 11340025, ImmunoTools GmbH), 10 ng/ml of rhIL-7 (cat. no. 11340075, ImmunoTools GmbH) and 20 ng/ml of rhIL-15 (cat. no. 11340155, ImmunoTools GmbH) for seven days at 37°C with 5% CO₂.

The immunophenotype of monocytes and DCs were assessed by anti-CD14 (cat. no. 21620143, ImmunoTools GmbH), anti-CD40 (cat. no. 21270403, ImmunoTools GmbH) and anti-human leukocyte antigen (HLA)-DR (cat. no. 21278993, ImmunoTools GmbH), anti-CD80 (cat. no. 12-0809-42, Thermo Fisher Scientific, Rockford, IL), anti-CD83 (cat. no. 12-0839-42, Thermo Fisher Scientific) and anti-CD86 antibodies (cat. no. 12-0869-42, Thermo Fisher Scientific). All antibodies were used at 1:50 dilution for 30 min at 4°C. Isotype antibodies were used as negative controls.

The memory T cell subsets were stained by anti-CD3 (cat. no. 21620033, ImmunoTools GmbH), anti-CD45RA (cat. no. 21819456, ImmunoTools GmbH) and anti-CD62L (cat. no. 21819624, ImmunoTools GmbH), anti-CD4 (cat. no. 56-0049-42, Thermo Fisher Scientific) and anti-CD8 (cat. no. A15448, Thermo Fisher Scientific). For intracellular cytokines, the activated T cells were re-stimulated with 10 µg of MSLN antigenic peptides (SLLFLLFSL and VLPLTVAEV) (GenScript, Jiangsu, China) (13,28) for 6 h in the presence of BD GolgiPlug (BD Biosciences, San Jose, CA) at 37°C with 5% CO₂. The cells were stained with anti-CD3, anti-CD4, anti-CD8 and anti-CD69 (cat. no. MA1-10277, Thermo Fisher Scientific). Cells were fixed in BD Cytofix/Cytoperm (BD Biosciences) for 20 min at 4°C, stained with anti-IFN-γ antibody (cat. no. 17-7319-82, Thermo Fisher Scientific). All antibodies were stained at 1:100 dilution for 30 min at 4°C.

The flow cytometry data of DCs and T cell immunophenotypes were acquired by CytoFLEX (Beckman Coulter, Inc.); the intracellular cytokine staining was acquired by BD LSRFortessa (BD Biosciences). The data were analyzed by FlowJo version 10.7 (FlowJo LLC) and shown as geometric mean fluorescence intensity (MFI) of each marker normalized by isotype control.

GM-CSF, IL-4 and IL-12p70 were measured using Human GM-CSF (cat. no. DGM00), IL-4 (cat. no. D4050) and IL-12p70 (cat. no. D1200) Quantikine ELISA kits (R&D systems, Inc.). IFN-γ was measured in medium collected from activated T cells co-cultured with target cancer cells using Human IFN-γ Quantikine ELISA kit (cat. no. DIF50, R&D systems, Inc.).

The IFN-γ ELISpot assay was performed using a Human IFN-γ ELISpot^BASIC kit (Mabtech AB). Briefly, 15 µg/ml of IFN-γ capture antibody was coated for overnight at 4°C. The 2×10⁵ activated T cells were then re-stimulated with 1×10⁴ MSLN peptides-pulsed HLA-A2⁺ T2 cells. T cells treated with 20 ng/ml of phorbol 12-myristate 13-acetate and 1 µg/ml of ionomycin (MilliporeSigma) served as positive controls. After removing these T cells, 1 µg/ml of biotinylated-IFN-γ antibody was incubated for 2 h and ALP-conjugated streptavidin for 1 h at RT. The IFN-γ spots were evaluated by 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium plus substrate (Mabtech AB) and captured by ELISpot plate reader (BIOREADER^® 5000 Fγ, BIOSyS). The numbers of spots were counted by CellCounter software (https://github.com/nghiaho12/CellCounter) version 0.2.1.

Lenti-X™ 293T and cancer cell lysates were prepared in RIPA Lysis Buffer System (Santa Cruz Biotechnology). The protein concentration was determined by Bradford assay and 30 µg of protein lysates were separated in 12% SDS-PAGE before transferred to PVDF membrane (GE Healthcare, Buckinghamshire, UK). The membrane was blocked in 5% skimmed milk (MilliporeSigma) for 30 min at RT. The membrane was incubated with 1:500 anti-MSLN and 1:5,000 anti-β-actin antibodies (cat. no. sc-47778; Santa Cruz Biotechnology, Inc.) at 4°C overnight. The 1:1,000 HRP-conjugated goat anti-mouse antibody (cat. no. 7076; Cell Signaling Technology, Inc.) was added and incubated for 1 h at RT. The signal was detected by Clarity™ western ECL substrate (Bio-Rad Laboratories, Inc.) using a Gel Doc instrument (G:Box Chemi XR5; Syngene Europe). The band intensity was analyzed by ImageJ software version 1.53 (NIH).

T cells were co-cultured with 5,000 cells of luciferase expressing HCC70, MDA-MB-231 and MSLN-MDA-MB-231 cells at Effector: Target ratio of 1:1, 5:1 and 10:1 for 24 h at 37°C with 5% CO₂. The luciferase activity was determined using Pierce Firefly Luciferase Glow Assay kit (Pierce; Thermo Fisher Scientific, Inc.) and Lumat LB 9507 Ultra-sensitive Luminometer (Berthold Technologies GmbH & Co. KG). After normalizing the luciferase activity of every condition with target cell alone condition, the percentage of cancer cell lysis was calculated. The luciferase activity of target cell alone was served as an internal control.

The 1×10³ target mWasabi-transduced cancer cells were formed into spheroid in 96-well ultra-low attachment plates (Corning Life Sciences) in 200 µl culture medium containing 2.5% cold Matrigel (BD Biosciences) by centrifugation at 300 × g for 3 min at 4°C. The spheroid was incubated for 4 days at 37°C with 5% CO₂ with CellTracker™ Orange CMRA Dye (Invitrogen; Thermo Fisher Scientific, Inc.)-labelled activated T cells at Effector: Target ratios of 1:1, 5:1, 10:1 and 20:1 for 48 h at 37°C with 5% CO₂. The mWasabi and CMRA fluorescence signals were detected by inverted fluorescence microscope and cellSens standard program version 1.15 (Olympus Corporation).

The association between MSLN score and clinicopathological data were accessed by Fisher's exact test. One-way ANOVA and Tukey's post-hoc test was used for all experiments except in the intracellular cytokine staining for comparison of control- and peptide-challenged T cells from the same condition in which Student's t-test was used. The Fisher's exact test was performed in SPSS 17.0 (SPSS, Inc.), whereas GraphPad prism V (GraphPad Software, Inc.) was used for one-way ANOVA and Student's t-test. All results were shown as mean ± standard deviation from at least three independent experiments. P<0.05 was considered to indicate a statistically significant difference.

All patient cases were female with mean age of 54±11.5 years diagnosed as luminal, HER2-positive and TNBC subtypes for 52 (15%), 154 (38%) and 165 (47%) cases, respectively (Table I). The cytoplasmic and membranous patterns of MSLN positive were detected in cancer cells with various levels (Fig. 1A-D). No MSLN expression was observed in the adjacent normal mammary cells (Fig. 1E), stromal cells (Fig. 1F) and immune cells (Fig. 1G). Among 351 total cases, 115 cases (32.8%) were positive for MSLN. When stratifying the patients according to the subtypes, MSLN expression was found in 94 of 165 cases (57%) for TNBC subtype, 20 of 154 cases (14.9%) in HER2-positive and 1 of 52 cases (1.9%) in luminal subtypes (Fig. 1H). The mean MSLN expression score was significantly higher in TNBC subtype compared with that in HER2-positive and luminal subtype (Fig. 1I). MSLN was significantly correlated with the absence of ER, PR and HER2 and TNBC subtype (Table 1). There was no association between MSLN and overall or disease-free survival in all samples, HER2-positive and TNBC subtypes (data not shown).

MSLN-SmartDC was generated by transducing the lentiviral vector containing GM-CSF, IL-4 and MSLN (Fig. 2A) into PBMCs-derived monocytes from four HLA-A2 positive healthy donors. The expression of MSLN protein from MSLN-SmartDC lentiviral vector was confirmed in Lenti-X™ 293T (Fig. 2B). All DCs demonstrated dendritic-like morphology (Fig. 2C). The GM-CSF was significantly detected higher levels in both MSLN-SmartDC and RPS3-MSLN-SmartDC compared with monocytes (Fig. 2D). MSLN-SmartDC and RPS3-MSLN-SmartDC also secreted IL-4 significantly higher than the monocytes but significantly lower compared with IRFP-SmartDC (Fig. 2E). Production of IL-12p70 was detected in significantly higher levels in RPS3-MSLN-SmartDC compared with other conditions (Fig. 2F). The significant reduction of CD14 monocyte marker in all DCs conditions (Fig. 2G) and mature DC markers including CD40, HLA-DR, CD80, CD83 and CD86 (Fig. 2H-L) were significantly increased compared with monocytes. The addition of RPS3 to MSLN-SmartDC could significantly increase CD40, CD80 and CD83 (Fig. 2H and J-K) compared with those of MSLN-SmartDC.

The results showed slight changes in CD4⁺ and CD8⁺ T cells frequencies and memory T cells subsets compared with those in PBMCs at day 0 (Fig. S1A and B). Re-stimulation with MSLN antigenic peptides in T cells activated by MSLN-SmartDC and RPS3-MSLN-SmartDC showed significantly higher number of IFN-γ than those without peptide treatment (Fig. 3A). The frequency of IFN-γ⁺ CD8⁺ T cells after MSLN peptides challenging were found at higher levels in MSLN-SmartDC- and RPS3-MSLN-activated T cells than unactivated and IRFP-activated T cells (Fig. 3B). However, only the RPS3-MSLN-activated T cells revealed significant increased IFN-γ⁺ CD8⁺ T cells (Fig. 3B). The MSLN-SmartDC-activated T cells result did not achieved a statistically significant level. The dual expressions of CD69 and IFN-γ were observed in MSLN- and RPS3-MSLN-activated T cells, but very low level in T cells treated with unactivated- and IRFP-activated T cells (Fig. 3C). The frequency of MSLN-specific CD69⁺ IFN-γ⁺ CD8⁺ T cells activated by RPS3-MSLN-SmartDC was clearly higher than unactivated and IRFP-activated T cells. Moreover, CD69⁺ IFN-γ⁺ CD8⁺ T cells in RPS3-MSLN-activated T cells was significantly higher compared with those without peptides re-stimulation.

The MSLN showed the highest level in MSLN-MDA-MB-231, followed by HCC70, whereas MDA-MB-231 had no MSLN (Fig. 4A). The IFN-γ releasing from RPS3-MSLN-activated T cells after co-culturing with MSLN-MDA-MB-231 and HCC70 cells was higher than MSLN-activated T cells and significantly higher than those of IRFP-activated T cells (Fig. 4B). No changes of IFN-γ secreted from T cells cocultured with MDA-MB-231 were observed.

The results exhibited no difference in MDA-MB-231 cell lysis co-cultured with both MSLN-activated T cells and RPS3-MSLN-activated T cells at Effector: Target ratios of 1:1, 5:1 and 10:1 (Fig. 4C). T cells activated by MSLN-SmartDC demonstrated significantly higher MSLN-MDA-MB-231 cell cytotoxicity compared with IRFP-activated T cells at 10:1 ratio (Fig. 4D). At 10:1, T cells activated by RPS3-MSLN-SmartDC could significantly induce MSLN-MDA-MB-231 cell lysis more than MSLN-activated T cells (Fig. 4D). In HCC70, T cells activated by MSLN-SmartDC and RPS3-MSLN-SmartDC demonstrated higher killing activity than IRFP-activated T cells (Fig. 4E). However, only RPS3-MSLN-activated T cells achieved statistical significance of cancer cell killing compared with that of IRFP-activated T cells at ratios of 5:1 and 10:1 (Fig. 4E).

MDA-MB-231 spheroids co-cultured with activated T cells or unactivated T cells for 48 h revealed no differences in mWasabi green fluorescence signals representing viable cancer cells (Fig. 5A). MSLN-MDA-MB-231 spheroid co-cultured with T cells activated by MSLN-SmartDC significantly exhibited lower viable cells compared with those of unactivated T cells at 20:1 (Fig. 5B). T cells activated by RPS3-MSLN-SmartDC had significant decrease viable cells compared with unactivated T cells (P<0.001), IRFP-activated and MSLN-activated T cells at 20:1 (Fig. 5B). RPS3 could enhance DCs activation that subsequently increase T cells capability to recognize MSLN-MDA-MB-231 cells. For HCC70 spheroid, the significant decreased fluorescence signal was observed in MSLN-SmartDC-activated T cells at ratio 20:1 compared with unactivated T cells. Moreover, T cells activated by RPS3-MSLN-SmartDC demonstrated significant reduction of mWasabi fluorescence signal compared with both unactivated and IRFP-activated T cells at 10:1 and 20:1 (Fig. 5C).