Chemotherapy agents stimulate dendritic cells against human colon cancer cells through upregulation of the transporter associated with antigen processing

Single immunotherapy fails to demonstrate efficacy in patients with microsatellite stable (MSS) metastatic colorectal cancer (mCRC). Research on immune reactions before and after systemic agents for mCRC is warranted. Our study examined cell line models to compare the expression of immune surface markers on colon cancer cells before and after chemotherapy agents. We also elucidated mechanisms underlying the effects of chemotherapy agents on immune surface markers. We used real-world clinical samples with NanoString analysis and the Perkin-Elmer Opal multiplex system. We established that chemotherapy agents, particularly 7-ethyl-10-hydroxycamptothecin (SN-38), the active metabolite of irinotecan, stimulated the expression of stimulatory MHC class I alleles through stimulation the pathway of transporters associated with antigen processing 1 and 2 (TAP1 and TAP2) in cell line models. Application of infected cell protein 47 (ICP-47), a specific inhibitor of the TAP1/TAP2, significantly inhibited expression of TAP1/TAP2 and also inhibited the expression of the downstream MHC class I. In the functional assay, SN-38 significantly promoted the phagocytosis of colon cancer cells by monocyte-derived dendritic cells (MoDCs). We confirmed that the expression of major histocompatibility complex (MHC) class I, significantly increased after first-line chemotherapy and targeted therapy in the samples of real-world patients with de novo mCRC. Our study provides new insights for novel immunotherapy combinations.

www.nature.com/scientificreports/ (HLA) class I. Both actions enhance the efficacy of cytotoxic T cells 19,20 . Moreover, oxaliplatin induces the expression of calreticulin in tumor cells and thus enhances ICD 16,20 . Fluorouacil (5-FU) stimulates the cytotoxicity of natural killer (NK) cell and also stimulates the expression of MHC class I and PD-L1. Cetuximab, an antiepidermal growth factor receptor monoclonal antibody, can stimulate immune effector cells and induce antibody-dependent cell-mediated cytotoxicity (ADCC) 24,25 . Bevacizumab, an antivascular endothelial growth factor monoclonal antibody, targets endothelial cells in peritumor parts and exerts immune modulating effects by influencing the tumor microenvironment 26,27 . The aforementioned agents are all standard systemic agents for the treatment of mCRC. That is, research on immune reactions before and after systemic agents for mCRC is warranted for new immunotherapy targets.
Our study plans to incorporate data from real-world clinical samples into cell line models. We compared the expression of immune surface markers in colon cancer cells before and after chemotherapy agents and elucidated mechanisms underlying the effects of chemotherapy agents on immune surface markers. We would focus on the interactions between chemotherapy agents and antigen processing pathway and the subsequent dynamic change of MHC class I [28][29][30] .

Results
IFN-γ specifically stimulated the expression of stimulatory MHC class I alleles. First, we tested three colon cancer cell lines, SW480, COLO 320 and HT29, through flow cytometry. These cell lines all initially exhibited low expression of MHC class I and NK cell ligands. The expression of MHC class I significantly increased after IFN-γ stimulation (Fig. 1A,B). By contrast, NK cell ligands including MIC A/B and ULBP-1 were both irresponsive to IFN-γ (Fig. 1C). We then tested more allele expressions specifically on SW480. SW480 demonstrated low expression of all MHC class I alleles, including pan-MHC class I, HLA-A, HLA-C, HLA-E, HLA-F, and HLA-G. NK cell ligands, including MIC A/B and ULBP, both exhibited low expression for SW480 cells (Fig. 1B,D). IFN-γ specifically stimulated the expression of MHC class I, particularly HLA-A, but NK cell ligands were responsive to IFN-γ stimulation. In summary, IFN-γ significantly stimulates the expression of MHC class I, particularly HLA-A without stimulating NK cell ligands. Although it was non-significant, a mild trend occurred whereby the stimulatory effect of MHC class I and HLA-A in response to IFN-γ stimulation was positively correlated with IFN-γ dosage and incubation time.
Chemotherapy agents, especially SN-38, which was the active metabolite of irinotecan, stimulated the expression of stimulatory MHC class I alleles and NK cell ligands. The chemotherapy backbone of mCRC consisted of 5-FU, oxaliplatin, and irinotecan. The cytotoxicity effect of irinotecan was exerted by the active metabolite SN-38. We firstly tested these three chemotherapy agents at the level of human pharmacokinetic studies (PK studies) 31 . Notably, all three of these chemotherapy agents increased the expression of MHC class I alleles, albeit at different levels ( Fig. 2A,B). Unlike IFN-γ, SN-38, particularly, not only increased the expression of MHC class I but also increased the expression of NK cell ligands. Not only HLA-A, SN-38 also increased the expression of all MHC class I alleles. The dosage effect for these three agents greatly differed. SN-38 significantly increased the expression of surface markers even at low doses, and the stimulation declined gradually with higher doses (Fig. 2C,D). The peak stimulation effect on the SN-38 level of 0.1 μM demonstrated a high expression level comparable to the level stimulated by IFN-γ. Although the absolute fluorescence intensity (MFI) was relatively low after exposure to SN-38, the fold change of NK cell ligands was relatively high without the correlation to SN-38 concentration. By contrast, 5-FU and oxaliplatin only moderately stimulated the expression of MHC class I alleles at a peak level of 15 μM for 5-FU. The peak stimulation of pan-MHC class I for oxaliplatin was 10 μM, and the expression of HLA-A increased in combination with increasing oxaliplatin doses (Fig. 2E,F). However, extremely few tumor cells were viable with higher doses of oxaliplatin. Moreover, 5-FU and oxaliplatin stimulated the expression of NK cell ligands, although the expression remained extremely low (Fig. 2G,H).
Chemotherapy agents, particularly SN-38, stimulated the expression of TAP1, TAP2, and tapasin, namely, the TAP pathway, and the subsequent MHC class I expression. IFN-γ stimulated MHC class I expression by triggering the downstream pathway of STAT1, but none of these three chemotherapy agents stimulated the STAT1 pathway (Fig. 3A). We thus tested all other downstream antigen processing pathway molecules. Only TAP1 and TAP2 were significantly upregulated by chemotherapy agents, particularly SN-38. None of other molecules was stimulated by chemotherapy agents (Fig. 3B,C). ICP47, a specific TAP inhibitor, significantly inhibited expression of TAP1 and TAP2 as well as the expression of downstream MHC class I. ICP47, which was a herpes simplex virus 1 (HSV-1) product, could specifically inhibit human TAP 32,33 . ICP47 had to be transfected into cells with Xfect. Initially we tested the transfection efficacy through X-gal staining (Fig. 4A). After transfection with ICP47, the stimulation effects of SN-38 on MHC class I and HLA-A were diminished by ICP47, which blocked the upregulation of TAP1 and TAP2 (Fig. 4B). These results confirmed that the stimulation effect on the expression of MHC class I mainly came from the upregulation of TAP1/TAP2 (Fig. 4C).  MHC class I, PD-1, and programmed death ligand 1 (PD-L1) expression increased after first-line chemotherapy and targeted therapy from real-world patient's samples. We enrolled patients with de novo mCRC. The 62 year-old man who had stage-IV sigmoid colon cancer initially presented with multiple liver and lung metastases. The expanded RAS and BRAF V600E status were all wild type. The mismatch repair proteins were all preserved. The patient received first-line therapy of cetuximab in combination with irinotecan and infusional 5-FU. Almost complete remission was achieved (Fig. 6A), and the patient thus received laparoscopic low anterior resection for sigmoid tumor and radiofrequency ablation for liver metastases after eight cycles of systemic therapies. We obtained tumor tissues from primary sigmoid tumors before and after treatment and subjected these samples to NanoString analysis. Except for HLA-C, all MHC class I alleles were significantly upregulated (Fig. 6B). MHC class II alleles were also significantly upregulated (Supplementary data). Notably, the expression of programmed death-1 (PD-1), programmed death-1 ligand (PD-L1), and programmed death-2 ligand (PD-L2) also increased after treatment. By contrast, natural killer (NK) cell ligands, including MHC class I-related chains A and B (MIC A, MIC-B) and UL16 binding protein (ULBP), demonstrated no significant www.nature.com/scientificreports/ change in expression. We reconfirmed these results with the Perkin-Elmer Opal multiplex system of pathological samples, which combined all IHC staining signals within one image. We observed that, comparing the image before and after treatment, the expression of HLA-A significantly increased (Fig. 6C). Using computerized scanning to differentiate tumor parts (both pan-CK positive & DAPI-positive cells) and nontumor parts (pan-CK negative & DAPI-positive cells, Fig. 6D), the results were similar to those from NanoString analysis. The HLA-A were significantly upregulated after treatment both in tumor and nontumor parts. HLA-G were upregulated only in the tumor parts ( Fig. 6E).

Discussion
In our study, we indicated that the expression of MHC class I, PD-1, and PD-L1 significantly increased after first-line chemotherapy and targeted therapy from the samples of real-world patients with de novo mCRC. Using cell line models, we confirmed that chemotherapy agents, particularly SN-38, the active metabolite of irinotecan, stimulated the expression of stimulatory MHC class I alleles through stimulation of the TAP1 and TAP2 pathway expression. Application of ICP47, a specific TAP inhibitor, significantly inhibited expression of TAP1 and TAP2 as well as the expression of downstream MHC class I. Currently, clinical trials have failed to demonstrate the efficacy of immunotherapy with single ICIs for patients with microsatellite stable (MSS) mCRC. In cohort 2 of the MODUL study, reported at the 2018 European Society for Medical Oncology (ESMO) annual meeting, enrolled patients were randomized into two arms. Patients in the control arm, who received bevacizumab plus 5-FU, had a median overall survival (mOS) of 22.05 months. Patients in the study arm who received bevacizumab plus 5-FU in combination with immunotherapy atezolizumab had mOS of 21.91 months 34 . Although this is a negative study, a long "tail" characterizes the survival curve for patients who received atezolizumab, which usually signifies the efficacy for immunotherapy. Our results indicate that although all three chemotherapy agents stimulated MHC class I expression, SN-38 provided the greatest efficacy, particularly starting from low doses. Chemotherapy agents also stimulated the expression of PD-1 and PD-L1. Our results provided the rationale and a real combination strategy for further immunotherapy for mCRC.
Our study focused on chemotherapy agents for mCRC. The current standard of care for first-line treatment of mCRC is the use of single targeted agents plus doublet chemotherapy regimens 35,36 . Although targeted therapy increased overall response in combination with chemotherapy agents, monotherapy of single targeted agents demonstrated extremely low response rates, which implies that chemotherapy agents exerted greater effects on mCRC treatments. In the KEYNOTE-048 study, pembrolizumab plus platinum and 5-fluorouracil were considered a superior first-line treatment for recurrent or metastatic head and neck cancer compared with the prior standard treatment of cetuximab plus platinum and 5-fluorouracil 37 . This study also implied that immunotherapy in combination with chemotherapy, rather than with targeted therapy, would provide an alternative further clinical trial design.
Our study nonetheless has some limitations: First, our data supported only some but not all mCRC treatments. Metastatic colorectal cancer was so heterogeneous that no single regimen could be applied for all mCRCs. Currently, molecular subtypes of mCRC have been classified into at least four to six categories 38,39 . More combination strategies are required to overcome various molecular subtypes and mCRC immune microenvironment. Second, combination doublet chemotherapy with targeted therapy was a standard of care as mentioned. Our study simplified the treatment agents one by one, but the complexity of interactions among plentiful combination regimens required more research. Third, we also required more clinical samples to confirm and validate details of our hypotheses. Finally, clinical trials to apply these results in routine practice are crucial.
Our study might illuminate the subject of immunotherapy in mCRC treatment and provide new rationale for novel immunotherapy combinations. Chemotherapy agents were the backbone of treatment in the targeted therapy era and continue to be so in the immunotherapy era.   Western blot for SW480 STAT1 pathway in response to chemotherapy agents. The full blot was displayed in Supplement figures. Photoshop was applied for only modifying brightness. ImageJ was applied for each signals to quantification. The results of quantification and statistical analysis were shown in the right side of this figure. Quantification of western blot for SW480 STAT1 pathway in response to chemotherapy agents by using ImageJ and statistical analysis by using SPSS with at least three or more independent tests. (*P < 0.05, **P < 0.01, ***P < 0.001). (B) Western blot for the SW480 antigen processing pathway in response to chemotherapy agents. The full blot was displayed in Supplement figures. Photoshop was applied for only modifying brightness. ImageJ was applied for each signals to quantification. The results of quantification and statistical analysis were shown in the right side of this figures. Quantification of western blot for the SW480 antigen processing pathway in response to chemotherapy agents by using ImageJ and statistical analysis by using SPSS with at least three or more independent tests. (*P < 0.05, **P < 0.01, ***P < 0.001). (C) Western blot for the SW480 antigen processing pathway in response to chemotherapy agents. The full blot was displayed in Supplement figures. Photoshop was applied for only modifying brightness. ImageJ was applied for each signals to quantification. The results of quantification and statistical analysis were shown in the right side of this figures. Quantification of western blot for the SW480 antigen processing pathway in response to chemotherapy agents by using ImageJ and statistical analysis by using SPSS with at least three or more independent tests. (*P < 0.05, **P < 0.01, ***P < 0.001).   (3 times). After washing, membranes were incubated with enhanced chemiluminescence Western blotting detection reagents (Millipore, P90720, Burlington, MA, U.S.) and exposed to film for 1 to 10 min. The interferon-γ was applied as positive control 40,41 . Protein transfection. We first seeded 4 × 10 5 cells with 3 mL of culture on a 6-well plate. The next day, TAPinhibitor ICP47 (5 μg) was added to the tube containing the Xfect protein transfection reagent and incubated at room temperature for 30 min. The mixture was then applied to cells with 400 μL of serum-free medium (RPMI-1640 only) in each well and incubated at 37 °C for 60 min. We added 3 mL of the culture medium to each well and performed incubation at 37 °C for another 2 h. SN-38 (0.1 μM) was treated for 24 or 48 h. As a control, some other cells were stained with X-gal to determine the transfection efficiency through beta-galactosidase control (Takara Bio, 631,326, Mountain View, CA, U.S.).

Isolation of PBMCs.
PBMCs were isolated from peripheral blood of the health human volunteers. We applied Ficoll-Paque Plus and diluted with PBS by density gradient centrifugation (400 g, 20 min, without break). Then, the monocytes were purified by positive selection with human CD14 + magnetic MicroBeads by manual MACS cell separation system (Miltenyi Biotec). The monocytes were incubated for 6 days into RPMI-1640 and supplemented with 10% FBS, 1% Antibiotic-Antimycotic, 50 ng/mL GM-CSF and 20 ng/mL IL-4 in an atmosphere of 95% O 2 and 5% CO 2 at 37 °C to harvest MoDCs. The induction medium would be renewed at the third day.
MoDCs-based phagocytosis. After   Patient enrollment. We enrolled one patient who had been treated at National Taiwan University Hospital (NTUH). Primary sigmoid tumor tissues were obtained at diagnosis and after curative surgery. We obtained complete medical records including all treatment flows of systemic therapies and regular computed tomography (CT) scan follow-up reports at NTUH. This study was approved by the Institutional Review Board of NTUH  www.nature.com/scientificreports/ (NTUH#201612146RINB), and informed consent was obtained from the patient. All methods were carried out in accordance with relevant guidelines and regulations.
NanoString assay of patient samples. The NanoString nCounter PanCancer Immune Profiling Panel (NanoString Technologies) is a commercialized multiplexed gene expression panel, which can simultaneously quantitate 770 immune-related genes. This was used for further mRNA quantification of two 150-ng ribonucleic acid extracted from patient's both tissue samples. All procedures, including preparation, hybridization, detection, scanning, and normalization, were performed according to the manufacturer's instructions. Detailed information and the gene list are available on the official website at https:// www. nanos tring. com/ produ cts/ gene-expre ssion-panels/ gene-expre ssion-panels-overv iew/ hallm arks-cancer-gene-expre ssion-panel-colle ction/ panca ncerimmune-profi ling-panel.
Immunohistochemical (IHC) staining of patient samples. We applied the Perkin-Elmer Opal multiplex system to simultaneously detect multiple biomarkers plus nuclear counterstain within a single image. Initially, the 5-μm FFPE pathology slides were incubated at 70 °C for 1.5 h, deparaffinized with xylene, and then hydrated through an ethanol gradient ending with distilled water wash. The slides were fixed using 10% neutral buffered formalin for 20 min. Antigen retrieval was subsequently performed with a microwave for 15 min in AR6 (

Statistical analysis.
All results were collected with at least three or more independent tests. The testing results were presented in average and standard deviations were also demonstrated within figures. We also applied the ImageJ for quantification of western blot (by National Institute of Health, USA). We also performed the paired samples t test or ANOVA test as indicated. All P-values were two-tailed and value < 0.05 were considered statistically significant. All data analyses were performed using SPSS version 20.0 software (Chicago, IL, USA).