Abstract
Emerging evidence shows that the efficacy of chemotherapeutic drugs is reliant on their capability to induce immunogenic cell death (ICD), thus transforming dying tumor cells into antitumor vaccines. We wanted to uncover potential therapeutic strategies that target ovarian cancer by having a better understanding of the standard-of-care chemotherapy treatment. Here, we showed in ovarian cancer that paclitaxel induced ICD-associated damage-associated molecular patterns (DAMP, such as CALR exposure, ATP secretion, and HMGB1 release) in vitro and elicited significant antitumor responses in tumor vaccination assays in vivo. Paclitaxel-induced TLR4 signaling was essential to the release of DAMPs, which led to the activation of NF-κB–mediated CCL2 transcription and IkappaB kinase 2–mediated SNARE-dependent vesicle exocytosis, thus exposing CALR on the cell surface. Paclitaxel induced endoplasmic reticulum stress, which triggered protein kinase R–like ER kinase activation and eukaryotic translation initiation factor 2α phosphorylation independent of TLR4. Paclitaxel chemotherapy induced T-cell infiltration in ovarian tumors of the responsive patients; CALR expression in primary ovarian tumors also correlated with patients' survival and patient response to chemotherapy. These findings suggest that the effectiveness of paclitaxel relied upon the activation of antitumor immunity through ICD via TLR4 and highlighted the importance of CALR expression in cancer cells as an indicator of response to paclitaxel chemotherapy in ovarian cancer.
Introduction
Apoptosis triggered by certain types of chemotherapeutics agents causes dying tumor cells to elicit immune responses, a phenomenon referred to as “immunogenic cell death” (ICD); ICD contributes to the elimination of residual tumor cells (1, 2). Vaccination experiments in mice show a limited number of cytotoxic agents used in the clinic (including doxorubicin, mitoxantrone, oxaliplatin, and bortezomib) induce bona fide ICD in mouse tumor cells (3). Chemotherapy-driven ICD relies on the release or exposure of potential immunogenic signals known as “damage-associated molecular patterns” (DAMP) from dying cells to induce immune responses (4). ICD is characterized by multiple key DAMPs, including exposure of endoplasmic reticulum (ER) chaperones, such as calreticulin (CALR; ref. 5), ERp57 (protein disulfide isomerase family A member 3; also known as PD1A3; ref. 6), and HSP 70 kDa (HSP70) and HSP 90 kDa (HSP90; ref. 7) on the plasma membrane of dying tumor cells; secretion of ATP (8), high-mobility group box 1 (HMGB1; ref. 9), and Annexin A1 (ANXA1; ref. 10); and activation of the cancer cell–intrinsic type I IFN response and consequent secretion of the chemokine (C-X-C motif) ligand 10 (CXCL10; ref. 11). These DAMPs, upon binding to their own receptors on the surface of myeloid and lymphocytes cells, favor the recruitment, activation, homing, antigen uptake, and maturation of antigen-presenting cells, that is, dendritic cells. These processes eventually prime the activation of an adaptive immune response involving T cells, which can eliminate tumor cells in an IFNγ-dependent manner, in patients that survive chemotherapy (3).
With anthracycline treatment, CALR is translocated rapidly from ER to the cell surface of preapoptotic tumor cells (1). Exposure of CALR incudes recognition and phagocytosis of the anthracycline-treated tumor cells by dendritic cells, and enhances their immunogenicity in mice (5). Thus, the cell surface exposure of CALR in ICD serves as an important “eat me” signal for dendritic cells and a crucial inducer of antitumor immune response (12, 13). In addition, ERp57 is an ER-sessile protein that cotranslocates CALR to the cell surface of anthracycline-treated tumor cells (6).
The release of ATP, HMGB1, and ANXA1 by dying tumor cells is associated with chemotherapy-induced ICD, thus eliciting antitumor immunity driven by intratumoral dendritic cells (3). ATP secretion attracts and binds to P2X7 purinergic receptors on dendritic cells, which activates the NLR family pyrin domain containing 3 (NLRP3) inflammasome, allowing IL1β secretion, which subsequently primes IFNγ-producing CD8+ T cells (8). Anthracyclines induces tumor infiltration of myeloid cells via ATP release, which stimulates the local differentiation of inflammatory dendritic cells with a CD11c+CD11b+Ly6chi phenotype (14). These dendritic-like cells can present tumor antigens to T cells and induce antitumor immune responses (15). The immunogenicity of dying tumor cells after chemotherapy or radiotherapy depends on their release of the alarmin protein HMGB1, which activates tumor antigen–specific T-cell immunity via the innate receptor Toll-like receptor 4 (TLR4) and the adaptor myeloid differentiation primary response protein-88 (MyD88) expressed by the dendritic cells (9). In addition, ANXA1 released by dying tumor cells with anthracycline treatment requires interaction with the formyl peptide receptor 1 (FPR1) on dendritic cells to elicit T-cell–mediated antitumor immunity (10).
Anthracyclines activate the endosomal pattern recognition receptor TLR3, which in turn stimulates the rapid production of IFNs by tumor cells (11). By binding to IFNα and IFNβ receptors on tumor cells, type I IFNs trigger autocrine and paracrine circuits, and result in CXCL10 release (11). CXCL10 is considered an essential chemotactic factor for immune effectors that selectively attack the tumor (16).
In this study, we investigated whether the standard-of-care chemotherapy (carboplatin and paclitaxel) induced ICD in ovarian cancer. We found that paclitaxel, but not carboplatin, induced ICD in syngeneic murine models of ovarian cancer via TLR4-independent and -dependent pathways.
Materials and Methods
Patients and specimens
A total of 124 patients with epithelial ovarian cancer (EOC) after primary debulking surgery and 9 patients with advanced ovarian cancer after neoadjuvant chemotherapy (NACT) followed by interval debulking surgery (1 received carboplatin monotherapy and 8 received paclitaxel-carboplatin regimen) were recruited at the Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital (Hong Kong) and included in the study. Informed written consent was obtained from each patient prior surgery. Clinical parameters of the 124 patients with ovarian cancer with primary debulking surgery followed by paclitaxel chemotherapy are summarized in Supplementary Table S1. Ovarian tumor tissues were obtained during surgery and processed further for paraffin-embedded, formalin-fixed and optimal cutting temperature–embedded frozen tissues. Histologic typing of the ovarian tumors was classified by pathologists according to the World Health Organization criteria, whereas clinical staging was given by gynecologic oncologists following the Federation Internationale des Gynaecologistes et Obstetristes staging system. The research protocol was approved by the institution's clinical research ethics committee and the study was conducted in accordance to the Declaration of Helsinki.
Cell lines and cell culture conditions
ID8 cells and ID8F3 cells [a murine model of high-grade serous ovarian cancer (HGSOC) were made by generating novel ID8 derivatives that harbor Trp53−/− suppressor gene deletion; ref. 17] were cultured in DMEM supplemented with 4% FBS, 5 ng/mL insulin transferrin sodium selenite, and 100 U/mL penicillin and streptomycin. TKO cells (generated from triple-mutant mice that have developed metastatic HGSOC originating from fallopian tube; ref. 18) were cultured in RPMI medium supplemented with 10% fetal FBS and 100 U/mL penicillin and streptomycin. Unless otherwise specified, all cancer cell lines were maintained at 37°C under 5% CO2. ID8 cells were kindly provided by Prof. Fran Balkwill (Barts Cancer Institute, Queen Mary University of London, London, United Kingdom) in 2010. TKO cells were kindly provided by Prof. Samuel C. Mok (The University of Texas MD Anderson Cancer Center, Houston, TX) in 2017. ID8F3 cells were kindly provided by Prof. Iain McNeish (Wolfson Wohl Cancer Research Centre, Institute of Cancer Sciences, University of Glasgow, Glasgow, Scotland) in 2018. The cell lines were not authenticated in the past year. The cell lines were subjected to Mycoplasma, and passages 4–25 after thawing were used for the experiments.
Treatment with chemotherapeutic drugs and inhibitors
Chemotherapeutic drugs including paclitaxel (20 μmol/L; Greiner Bio-One), cisplatin (300 μmol/L; Greiner Bio-One), carboplatin (200 μmol/L; Greiner Bio-One), and mitoxantrone dihydrochloride (1 μmol/L; Sigma-Aldrich) were used. In functional blockade assays, cells were pretreated with TAK-242 (100 nmol/L; MedChem Express), BAY 11-7082 (1 nmol/L; Cayman Chemical), TPCA-1 (100 nmol/L; Cayman Chemical), or IKK2 inhibitor IV (100 nmol/L; Cayman Chemical) for 2 hours prior to paclitaxel treatment for 48 hours. Paclitaxel, cisplatin, carboplatin, mitoxantrone dihydrochloride, TAK-242, BAY 11-7082, TPCA-1, and IKK2 inhibitor IV were diluted in DMSO. Cell viability and apoptosis assays were carried out as described below to confirm that there was no alteration of cell growth or induction of cell death in cell lines under inhibitor treatment.
Generation of Tlr4-knockout cell line by CRISPR/Cas9
To generate Tlr4 knockout (KO) with ID8 and ID8F3 cells, murine Tlr4 CRISPR/Cas9-KO plasmid with a Tlr4 HDR plasmid (sc-423419 and sc-423419-HDR; Santa Cruz Biotechnology) was applied. Briefly, ID8 or ID8F3 cells (1.5 × 105 per transfection) were cultured in 6-well plates for 24 hours. The cells were then cotransfected with CRISPR/Cas9-KO plasmid (1 μg) and HDR plasmid (1 μg) by Lipofectamine 2000 (11668019; Invitrogen) for 24 hours. Culture media were replaced with complete medium with puromycin (2 μg/mL) post-transfection for selecting positive transfectants. Monoclonal cell populations of the transfectants were isolated by limiting dilution and expanding under puromycin treatment. The monoclonal cell population with Tlr4 KO were then verified by RT-PCR and Western blot analysis as described below.
Cell viability assay
Cell viability was measured by CellTiter-Blue Cell Viability Assay (G8081; Promega) according to the manufacturer's instruction. Briefly, 1 × 104 cells per well were seeded in 96-well plates for 24 hours. The cells were then treated with chemotherapeutic drugs in a dose-dependent manner (0.1–1,000 μmol/L) for 48 hours. After the indicated treatment, the cells were incubated in culture medium containing CellTiter-Blue Reagent (1:5) at 37°C for 4 hours. Cell viability was determined by recording the absorbance at 595 nm using Uquant MQX200 Microplate Reader Spectrophotometer (Bio-Tek). Percentage of growth inhibition was determined by comparing the difference between percentage of viable cells with and without treatment. Growth inhibition (%) and concentration (μmol/L in log) were arranged on vertical axis and horizontal axis, respectively. The IC50 is the concentration at which the curve passes through the 50% inhibition point.
Cell apoptosis by flow cytometry
Cell apoptosis was determined using a Dead Cell Apoptosis Kit with Annexin V (AnnV) Alexa Fluor 488 and propidium iodide (PI; V13242; Invitrogen) by flow cytometry according to the manufacturer's instruction. Briefly, after the indicated treatment, cells were resuspended in 100 μL of Annexin-binding buffer containing AnnV FITC and PI (1 × 106 cell/mL) and incubated at room temperature for 15 minutes in dark, followed by adding 400 μL of Annexin-binding buffer. Cell staining of AnnV and PI was detected by a flow cytometer FC500 (Beckman Coulter) and data were analyzed by a software FlowJo Ver.10 (Tree Star, Inc.). AnnV+PI− represents apoptotic cell death; AnnV−PI+ represents nonapoptotic (mitotic) cell death; and AnnV+PI+ represents necrotic cell death (19).
Cell-cycle analysis by flow cytometry
After the indicated treatment, cells were harvested by trypsinization and washed in PBS. The cells were then fixed in cold 70% ethanol for 30 minutes at 4°C. Cell-cycle analysis was done by the quantitation of DNA content. The fixed cells were further treated with RNase (5 μg per sample) and DNA in the cells was stained with PI (1 μg/mL) whose signal was detected by the flow cytometer FC500. Data were analyzed by FlowJo Ver.10.
Detection of cell surface protein by flow cytometry and immunofluorescence staining
Cells were first resuspended in complete medium after the indicated treatment and washed with FACS buffer (DMEM with 2% FBS). The cells were incubated in FACS buffer with a specific primary antibody: anti-CALR (ab22683; 1:200; Abcam), anti-ERp57 (ab10287; 1:200; Abcam), anti-ANXA1 (ab196830; 1:200; Abcam), anti-HSP70 (4873S; 1:200; Cell Signaling Technology), anti-HSP90 (4877S; 1:200; Cell Signaling Technology), or anti-F-actin (ab205; 1:200; Abcam) on ice for 1 hour. After washing, the cells were incubated with Alexa Fluor 488–conjugated isotype-matched anti-IgG (H+L) or isotype-matched anti-IgM (H+L) (A11008 and A21042; 1:200; Invitrogen) in FACS buffer on ice for 45 minutes. The cells were then resuspended in 400 μL of FACS buffer containing PI (1 μg/mL). Cell surface protein was detected by the flow cytometer FC500 and data were analyzed by FlowJo Ver.10. Difference of mean fluorescent intensity (Diff. MFI) values was obtained by subtracting CALR, ERp57, HSP90, HSP70, ANXA1, or F-actin staining with respective IgG staining by FlowJo Ver.10.
For immunofluorescence staining of F-actin, cells cultured in chamber slide were fixed with 4% paraformaldehyde at room temperature for 15 minutes, followed by blocking with FASC buffer and incubation with anti-F-actin (ab205; 1:100; Abcam) at 4°C overnight. After washing, the cells were incubated with Alexa Fluor 488–conjugated isotype-matched anti-IgM (H+L) (A21042; 1:200; Invitrogen) in FACS buffer at room temperature for 45 minutes. After washing, the cells were incubated with DAPI (1 μg/mL in PBS; D3571; Invitrogen) for 5 minutes at room temperature in the dark and mounted with ProLong Diamond Antifade Mountant (P36970; Invitrogen). The labeled cells were then examined by fluorescence microscopy (Zeiss Axio Observer Z1, software Zen 2012). At least three images were taken for each sample and the most representative image was shown.
ATP and HMGB1 assays
Extracellular ATP in conditioned medium and intracellular ATP in cell lysate following the indicated treatment were measured by a luciferin-based ENLITEN ATP Assay System Bioluminescence Detection Kit (FF2000; Promega) and an ATP Assay Kit (119107; Merck) according to the manufacturer's instruction, respectively. HMGB1 concentrations in the conditioned medium following the indicated treatment were measured by an HMGB1 ELISA Kit (ST51011; Shino-Test Corporation) according to the manufacturer's instruction. Luminescence and optical density (OD) were measured using Perkin Elmer 2030 Multiable Reader VICTOR X4 (PerkinElmer) and Uquant MQX200 Microplate Reader Spectrophotometer (Bio-Tek), respectively. Luminescence (RLU) of the ATP standards and OD (450 nm) of the HMGB1 standards were plotted against their corresponding concentrations, and the concentrations of extracellular ATP and HMGB1 in conditioned medium, and intracellular ATP in cell lysates were read directly from the standard curve.
Quantitative real-time RT-PCR
Total RNA was extracted using TRizol Reagent (15596-018; Invitrogen) and RNA concentration was measure by Nano Drop 1000 (Thermo Fisher Scientific). Amount of 2 μg total RNA was transcribed into cDNA using a High Capacity cDNA Kit (4374966; Applied Biosystems) in final volume of 20 μL according to the manufacturer's protocol. Amount of 1 μL cDNA was used in real-time quantitative reverse transcription-PCR (qRT-PCR) performed with TaqMan probes for Ccl2 (Mm00441242_m1; Applied Biosystems), Cxcl10 (Mm00445235_m1; Applied Biosystems), Ifnb1 (Mm00439552_s1; Applied Biosystems), Nfkbiz (Mm00600522_m1; Applied Biosystems), Tlr4 (Mm00445273_m1; Applied Biosystems), or β-actin (4352933E; Applied Biosystems) using an ABI Prism 7900HT Sequence Detection System (Applied Biosystems) according to the manufacturer's protocol. Relative mRNA gene expression was calculated using a 2−ΔCt method (Applied Biosystems User Bulletin N°2, P/N 4303859). All samples were performed in triplicates in qRT-PCR.
Cytoplasmic and nuclear fractionation
Cells from 100-mm culture plate were trypsinized and resuspended in complete culture medium after the indicated treatment. The cell pellet was collected by centrifugation at 1,000 rpm at 4°C and then subjected to cytoplasmic and nuclear protein fractionation by NE-PER Nuclear and Cytoplasmic Extraction Reagent (78833; Pierce Biotechnology) according to the manufacturer's protocol. Protein concentration was measured by a Pierce BCA Protein Assay Kit (23225; Pierce Biotechnology). The cytoplasmic and nuclear proteins were stored at −80°C prior to Western blot analysis.
Phos-tag SDS-PAGE
Detection of SNAP23 phosphorylation was done by Phos-tag SDS-PAGE as described previously (20). Briefly, total cell lysates were extracted using RIPA Buffer (89900; Pierce Biotechnology). One microgram of extracted protein from each sample was used in Phos-tag SDS-PAGE [7.5% polyacrylamide gels containing 50 μmol/L Phos-tag acrylamide and 100 μmol/L MnCl2 (304-93526; Wako Pure Chemical Industries, Ltd.)]. After electrophoresis, the Phos-tag acrylamide gels were washed with transfer buffer (50 mmol/L Tris, 384 mmol/L glycine, 0.1% SDS, and 20% methanol) containing 1 mmol/L EDTA for 10 minutes with gentle shaking prior to Western blot analysis.
Western blot analysis
Total cell lysates were extracted using RIPA Buffer (89900; Pierce Biotechnology) and 10 μg of extracted protein was used for Western blot analysis. Specific primary antibodies were applied to probe the target proteins, including anti-HMGB1 (3935S; 1:2,000; Cell Signaling Technology), anti-PERK (3192S; 1:2,000; Cell Signaling Technology), anti-eIF2α (9722S; 1:2,000; Cell Signaling Technology), anti-phospho-eIF2α (9721S; Ser51; 1:2,000; Cell Signaling Technology), anti-TLR4 (AF1478; 1:2,000; R&D Systems), anti-IKKβ (2678T; 1:2,000; Cell Signaling Technology), anti-phospho-IKKβ (2078T; 1:2,000; Cell Signaling Technology), anti-p105/p50 (13568S; 1:2,000; Cell Signaling Technology), anti-p65 (8242S; 1:2,000; Cell Signaling Technology), anti-RelB (4922S; 1:2,000; Cell Signaling Technology), anti-SNAP23 (ab133703; 1:2,000; Abcam), anti-F-actin (ab205; 1:2,000; Abcam), and anti-β-actin (A5441; 1:10,000; Sigma-Aldrich). After probing with a horseradish peroxidase (HRP)–linked secondary antibody (NA934; 1:2,000, GE Healthcare), the signals were detected with ECL Reagent (NEL104001EA; PerkinElmer). Blot was scanned (V700 Photo; EPSON) and band intensity was quantified using ImageJ software. β-actin served as a loading control.
ATP-containing vesicles staining
ATP-containing vesicles were detected using quinacrine staining (21). Briefly, cells were cultured in 8-well chamber slides, 1 × 104 cells per well (177402; Nalge Nunc International) for 24 hours. After indicated treatment, the cells were washed with PBS and fixed with 4% paraformaldehyde, followed by incubation with 5 μmol/L quinacrine (sc-204222; Santa Cruz Biotechnology) for 30 minutes to label ATP-containing vesicles. The cells were then washed with PBS and examined by fluorescence microscopy as described above.
In vivo model of tumor vaccination assay
All mice were maintained in pathogen-free conditions and the in vivo experiments were performed according to (22) under the guidelines of Animal Experimentation Ethics Committee at The Chinese University of Hong Kong (Hong Kong). Briefly, ID8F3 cells (wild-type or TLR4 KO) were incubated with paclitaxel (20 μmol/L) for 48 hours, and those treated with DMSO served as the control group. After drug treatments, the cells were harvested and resuspended in PBS at a concentration of 1 × 107 cells/mL. For vaccination, 50 μL of the resuspended cells in saline (5 × 105 cells per mouse) were injected subcutaneously to the left flank of immunocompetent C57BL/6 mice (8-week-old female; n = 5 per group). After 7 days, the vaccinated mice were rechallenged with 300 μL live untreated wild-type ID8F3 cells in saline (5 × 106 cells per mouse) by intraperitoneal injection. Tumor growth was monitored regularly by observation of ascitic fluid development for the following weeks; the experiment would be stopped as soon as the mice become unmanageable (duration averaged about 50–60 days), that is, with the development of ascitic fluid that caused movement problems or when the weight of the mouse became over 30 g. The absence of ascitic fluid development in the mice after 120 days after intraperitoneal injection was an indication of the absence of tumors and thus efficient antitumor vaccinations. All mice were eventually sacrificed by cervical dislocation for examination to confirm the absence or presence of peritoneal tumor development. The experiment was repeated twice.
Microarray gene expression data analysis
The microarray gene expression data (accession no. GSE15622; including data from 20 patients with ovarian cancer randomized to paclitaxel monotherapy in CTCR-OV01) from Gene Expression Omnibus database (23) was adopted. For immune cell profiling analysis, data of the 14 patients who had a matched biopsy performed before and after three cycles of paclitaxel monotherapy (with 11 patients responded and 3 patients resistant to the chemotherapy) were selected. The immune cell type of each sample was inferred using gene expression enrichment analysis by xCell pipeline (a novel gene signature–based method; http://xCell.ucsf.edu/; ref. 24), and the amount of each immune cells' type were represented as enrichment scores (ES). Differential expression of immune cell types was determined by comparing the ES between pre- and postchemotherapy samples using DChip analysis (DChip; http://www.dchip.org). Correlations between the mRNA expression of TLR4 and ES of immune cells' types in postchemotherapy samples of 11 patients who responded to the chemotherapy were determined by Pearson correlation analysis. In addition, gene expression of CALR in the biopsies (prepaclitaxel monotherapy) from patients responded (n = 13) and resistant (n = 7) to the chemotherapy were extracted from the database and analyzed by Mann–Whitney U test.
IHC and image analysis of TLR4 and CALR
Ovarian tumor specimens after collection from surgery were fixed in 10% neutral-buffered formalin for 24 hours and embedded with paraffin. Paraffin sections of 4 μm thickness were cut and mounted on slides. The slides were deparaffinized in two washes of xylene for 10 minutes each, and rehydrated from 100%, 95%, 80%, and 70% ethanol to distilled water for 5 minutes each. Protein expression of TLR4 in ovarian tumor tissues were examined by routine IHC as described previously (25) using TLR4-specific antibody (AF1478; 1:25; R&D Systems). Protein expression of CALR was detected by IHC using a Ventana BenchMark XT System (a fully automated IHC slide staining instrument; Roche). In brief, after deparaffinization and rehydration, paraffin-embedded tissue sections were subject to Ventana BenchMark XT platform. Sections were blocked with 3% hydrogen peroxide at room temperature for 4 minutes, and then treated with heat-induced antigen retrieval CC1 (Cell Conditioning 1; Roche) solution using the optimized antigen retrial condition, followed by incubation with CALR-specific antibody (ab22683; 1:500; Abcam) at 37°C for 30 minutes. Then the tissue sections were incubated with OptiView HRP Linker for 12 minutes and OptiView HRP multimer for 12 minutes, and finally developed with 3,3′-diaminobenziding (DAB) for 4 minutes. Nuclei were then counterstained with hematoxylin.
Images at 200× fields were captured by Mantra Quantitative Pathology Workstation (PerkinElmer) using a brightfield protocol. Ovarian tumor tissues stained with hematoxylin or DAB alone were prepared to develop a spectral library that contains the spectral peaks of individual dyes for spectral unmixing and identification of signals using the inform 2.4.2 Image Analysis Software (PerkinElmer). All spectrally unmixed images were subject to a trainable tissue segmentation tool of the inForm software that distinguishes tumoral area from stromal and blank areas. A cell segmentation tool was subsequently applied to classify the associated cellular compartments (nuclei and cytoplasm for TLR4, and membrane for CALR) in the segmented tumoral areas on the basis of an object-based approach, in which hematoxylin was selected for nuclear segmentation, while cytoplasm and membrane were selected as the cellular compartments for pixel validation of TLR4 and CALR, respectively. Protein expression in tumoral areas was scored from 0 to 3+ to bin spectrally unmixed signals into four bins. H-score for each image was computed by the software using the percentages in each bin that ranges from 0 to 300.
Opal multiplex IHC and multispectral imaging analysis of CD8, CD208, and IFNγ
Ovarian tumor specimens were fixed in 10% neutral-buffered formalin for 24 hours and embedded with paraffin. Paraffin sections of 4 μm thickness were cut. The tissue sections were deparaffinized in two washes of xylene for 10 minutes each, rehydrated from 100% ethanol for 10 minutes, 95% ethanol for 10 minutes, and 70% ethanol for 5 minutes to distilled water for 5 minutes, and fixed in 10% neutral-buffered formalin for 20 minutes. Antigen retrieval was performed using AR6 Buffer (AR6001KT; PerkinElmer) and peroxidase activity was blocked by Dako REAL Peroxidase-Blocking Solution (S202386-2; Agilent). The opal multiplexed assay was carried out with each section subject to three successive rounds of antibody staining, each of which consists of (i) protein blocking, (ii) incubation with primary antibodies: CD8 (ab17147; 1:25, Abcam), CD208 (DDX0191; 1:250, Dendritics), and IFNγ (sc-8308; 1:500, Santa Cruz Biotechnology) were stained sequentially; (iii) incubation with HRP-conjugated secondary antibodies: Opal Polymer HRP Ms + Rb (ARH1001A; PerkinElmer) was applied to the primary antibodies against CD8 and IFNγ, while Rabbit Anti-Rat IgG H&L (HRP) (ab6734; Abcam) was applied to the primary antibody against CD208; and (iv) tyramide signal amplification (TSA)-conjugated fluorophore (PerkinElmer). CD8, CD208, and IFNγ were visualized using Opal 520 (FP1487001KT; 1:50 dilution; PerkinElmer), Opal 570 (FP1488001KT; 1:150 dilution; PerkinElmer), and Opal 690 (FP1497001KT; 1:100 dilution; PerkinElmer), respectively. Microwave treatment in AR6 buffer was repeated before next round of staining to remove the bound antibody being detected by the TSA reagent. After incubation with the last antibody, nuclei were counterstained with DAPI (FP1490A; Perkin Elmer) and mounted with ProLong Diamond Antifade Mountant (P36970; Thermo Fisher Scientific). Tumor tissue without primary antibodies served as negative controls.
The positively labeled CD8+ T cells, CD208+ dendritic cells, and IFNγ+ cells in the specimens were then quantified. All five filter cubes in the Mantra Quantitative Pathology Workstation associated with spectral bands for DAPI, FITC, CY3, Texas Red, and CY5 were employed for multispectral imaging. One 200× representative image of each marker that emits single or multiple spectral bands was selected to set the exposure times for image acquisition. Spectral unmixing was performed using the inForm 2.4.2 image analysis software with a spectral library containing spectral peak information of all fluorophores. Images from the negative control slide were acquired to assess the autofluorescence signal. All spectrally unmixed images were subject to the trainable tissue segmentation tool of the inForm software that distinguishes the tissue areas of interest from necrotic or blank areas, followed by the cell segmentation tool that classifies the associated cellular compartments (nuclei, cytoplasm, and membrane) in the segmented tissue areas on the basis of a DAPI counterstain–based approach. The cell phenotyping recognition learning algorithm tool was then applied to identify the cells positively stained for individual markers. Tissue area and cell counts in the segmented areas in each image were quantitated for the evaluation of cell density for each tumor specimen. Cell density (in megapixel) per 200× field image was accounted by the number of cells per tissue area.
Statistical analysis
All data were described as mean ± SEM and analyzed using GraphPad Prism 5.0 Software (GraphPad Inc.) from at least three independent experiments. Statistical difference between experimental groups was determined by Student t test and Mann–Whitney U test. Correlation between experimental groups was determined by Pearson correlation. Survival analysis was performed by the Kaplan–Meier curves and log-rank test were generated using GraphPad Prism 5.0 software. χ2 test was performed by SPSS version 18.0 software. *, P < 0.05 was considered as statistically significant; **, P < 0.01 as highly significant; and ***, P < 0.001 as extremely significant.
Results
Paclitaxel, but not carboplatin, induced ICD in mouse models of ovarian tumor
We investigated whether the standard-of-care chemotherapeutic drugs, including carboplatin and paclitaxel, induced ICD in ovarian cancer by examining the associated DAMPs in ID8 murine–derived ovarian cancer cell line (3, 26). Mitoxantrone was used as positive controls as it induces ICD (5, 27). Cisplatin was used as negative controls, as it does not induce ICD or surface expression of CALR in a syngeneic murine model of colon cancer (28). We treated ID8 cells with serial dilutions (0.1–1,000 μmol/L) of paclitaxel, cisplatin, carboplatin, or mitoxantrone for 48 hours, and determined their IC50 by growth inhibitory assays as follows: 20 μmol/L (paclitaxel), 200 μmol/L (carboplatin), and 1 μmol/L (mitoxantrone). The IC50 of each drug was confirmed by AnnV/PI flow cytometry which showed more than 70% of cell death (Fig. 1A). This IC50 was applied to the following experiments unless otherwise specified. Nevertheless, we applied cisplatin at a lower concentration of 300 μmol/L instead of an IC50 for cisplatin (500 μmol/L).
Paclitaxel, but not carboplatin, induced ICD in a mouse model of ovarian tumor. A, Cell viability assay demonstrating the cell proliferation of ID8 cells treated with paclitaxel, cisplatin, carboplatin, or mitoxantrone. Cell death was determined by treating the ID8 cells with respective IC50 of paclitaxel, carboplatin, or mitoxantrone, whereas cisplatin was minimized at a concentration of 300 mmol/L. B, Representative surface expression of CALR and ERp57 on ID8 cells after treatment with paclitaxel, cisplatin, carboplatin, or mitoxantrone. C, Extracellular ATP (left) and intracellular ATP (right) concentrations in ID8 cells after drug treatments. D, Soluble HMGB1 in conditioned media of ID8 cells after drug treatments. Representative surface expressions of ANXA1 (E), HSP70 (F), and HSP90 (G) on ID8 cells after drug treatments. H, mRNA expressions of Ifnb1 and Cxcl10 in ID8 cells after drug treatments. Data shown are expressed as mean ± SEM from three independent replicates (*** P < 0.001; A–H). I, Tumor-free survival analysis by Kaplan–Meier curves indicating the antitumor response of paclitaxel-induced ICD on a syngeneic mouse model in vivo (mice immunized with paclitaxel-treated ID8F3 cells vs. mice immunized with DMSO-treated ID8F3 cells; n = 10 per group). CBDCA, carboplatin; CDDP, cisplatin; i.p., intraperitoneal; MTX, mitoxantrone; PTX, paclitaxel.
We examined the surface expressions of CALR and ERp57 on dying (but not dead) ID8 cells after drug treatments (6). Paclitaxel and mitoxantrone, but not carboplatin and cisplatin, increased CALR and ERp57 exposure on the cell surface of dying ID8 cells (Fig. 1B; Supplementary Fig. S1A). There was increased expression of extracellular ATP in ID8 cells treated with paclitaxel and mitoxantrone, whereas there was significantly lower intracellular ATP in ID8 cells treated with mitoxantrone (Fig. 1C). Both ELISA and Western blot analyses demonstrated high HMGB1 in the conditioned media of ID8 cells after the treatment with paclitaxel, carboplatin, or mitoxantrone (Fig. 1D). Only paclitaxel induced an increase in ANXA1 exposure in dying ID8 cells (Fig. 1E; Supplementary Fig. S1A). Paclitaxel also induced Cxcl10 upregulation (Fig. 1H). Although surface exposure of HSPs (HSP70 and HSP90) was found in anthracyclines-treated human tumor cells (7), we could not detect any significant alterations after treatment with paclitaxel or carboplatin (Fig. 1F and G; Supplementary Fig. S1A). Paclitaxel or carboplatin did not alter Ifnb1 (Fig. 1H). These results demonstrated that paclitaxel, but not carboplatin, induced CALR and ERp57 exposure, ATP secretion, HMGB1 release, AXNA1 exposure, and Cxcl10 upregulation in ID8 murine ovarian cancer cells in vitro. Similar observations were found in two other murine ovarian cancer cell lines including TKO (Supplementary Fig. S1B–S1G) and ID8F3 cells (Supplementary Fig. S1H–S1L).
We validated the paclitaxel-induced ICD in a syngeneic model using a “tumor vaccination” assay. Our results demonstrated a significant increase in tumor-free survival among mice immunized with paclitaxel-treated ID8F3 murine ovarian tumor cells when compared with the controls (immunized with DMSO-treated ID8F3 tumor cells; Fig. 1I), indicating that paclitaxel was a bona fide inducer of ICD in ovarian cancer.
Paclitaxel stabilizes microtubules, resulting in a mitotic cell death and G2–M-phase cell-cycle arrest (29–31). We set up a time-course experiment (0–48 hours) to study the correlation between paclitaxel-induced cell death/cell-cycle arrest and paclitaxel-induced DAMPs in ID8 cells. Paclitaxel induced cell death (AnnV+PI− cells, AnnV+PI+ cells, and AnnV−PI+ cells) and G2–M-phase arrest (Supplementary Fig. S2A). CALR and ERp57 exposure, ATP and HMGB1 release, ANXA1 exposure, and Cxcl10 were significantly increased (Supplementary Fig. S2B–S2F) after paclitaxel treatment. These data indicated that paclitaxel-induced DAMPs are correlated with the cell death and mitotic arrest.
Paclitaxel required cancer cell–autonomous TLR4 to induce ICD in ovarian tumors
We explored the molecular mechanism underlying the paclitaxel-induced ICD in our syngeneic mouse models of ovarian cancer by investigating the role of TLR4 (32–34), on the surface of ovarian cancer cells, by generating isogenic derivatives of ID8 and ID8F3 cells that lack Tlr4 expression (Tlr4−/−) with CRISPR/Cas9. There was high endogenous TLR4 in wild-type ID8 and ID8F3 cells, but absence of Tlr4 gene transcription and protein expression in our isogenic derivatives of ID8 (Tlr4-KO 3F, 8A, 8H, 10G, and 11F) and ID8F3 (Tlr4-KO 1A1) Tlr4-KO clones (Supplementary Fig. S3A). There was no difference in cell proliferation or endogenous cytokine expression (Ccl2 and Cxcl10; Supplementary Fig. S3B and S3C) nor in paclitaxel-induced cell death and G2–M-phase arrest (Supplementary Fig. S3D and S3E) between the wild-type and Tlr4-KO clones. These data indicated the absence of TLR4 did not affect the cell proliferation and endogenous cytokine expression of ovarian cancer cells, and that the paclitaxel-induced cell death and mitotic arrest were TLR4 independent.
We then investigated the immunogenicity of cell death induced by paclitaxel in the ID8 Tlr4-KO clones, and found reduced ICD-associated DAMPs compared with the wild-type ID8 cells (Fig. 2A–E; Supplementary Fig. S3F). Nevertheless, mitoxantrone-induced DAMPs in the KO clones have no significant difference compared with the wild-type (Supplementary Fig. S3G and S3H). Inhibiting the endogenous TLR4 activity of wild-type ID8 cells by pretreatment with antagonist TAK-242 (35) reduced paclitaxel-induced DAMPs (Fig. 2F–J; Supplementary Fig. S3I). Our in vivo tumor vaccination assays confirmed these findings by showing no significant difference in tumor-free survival of the mice immunized with paclitaxel-treated Tlr4−/− ID8F3 cells compared with controls (DMSO-treated Tlr4−/− ID8F3 cells; Fig. 2K). These results indicated paclitaxel required cancer cell–autonomous TLR4 to induce ICD in murine ovarian cancer.
Paclitaxel required cancer cell–autonomous TLR4 to induce ICD in ovarian tumors. A, Representative surface expression of CALR (left) and ERp57 (right) on ID8 wild-type (WT) cells and isogenic derivatives of ID8 Tlr4−/− (Tlr4-KO 3F/8H/10G) clones after paclitaxel treatment. Extracellular ATP concentration (B), soluble HMGB1 (C), representative surface expression of ANXA1 (D), and mRNA expression of Cxcl10 (E) in ID8 WT cells and Tlr4-KO clones after paclitaxel treatment. Representative surface expression of CALR (left) and ERp57 (right; F), extracellular ATP concentration (G), soluble HMGB1 (H), representative surface expression of ANXA1 (I), and mRNA expression of Cxcl10 (J) on ID8 cells with TAK-242 pretreatment prior to paclitaxel treatment. K, Antitumor immunity of ID8F3 Tlr4-KO clones after paclitaxel treatment (paclitaxel vs. DMSO; n = 5 per group) was examined by tumor vaccination in vivo. **, P < 0.01 and ***, P < 0.001. i.p., intraperitoneal; PTX, paclitaxel.
Paclitaxel activated canonical NF-κB signaling via the TLR4/MyD88 pathway
Because the abo paclitaxel-induced ICD required cell-autonomous TLR4 in ovarian cancer, we investigated whether the MyD88-dependent TLR4 signaling pathway, which activates canonical NF-κB signaling (36), was involved. We confirmed two major adaptors that bind to the intracellular domain of TLR4, MyD88, and TRIF [TIR (toll/IL1 receptor) domain-containing adapter-inducing IFNβ; ref. 37] were expressed in wild-type ID8 murine cells (Fig. 3A; Supplementary Fig. S4A). Paclitaxel induced nuclear translocation of NF-κB p50 and p65 subunits in the ID8 cells (Fig. 3B; Supplementary Fig. S4B). The paclitaxel-induced NF-κB activation (i.e., nuclear translocation of p65) was diminished in the ID8 Tlr4-KO clones and TAK-242–treated wild-type ID8 cells (Fig. 3C and D; Supplementary Fig. S4C and S4D). These results indicated that paclitaxel induced the activation of canonical NF-κB signaling in murine ovarian cancer cells in a TLR4/MyD88-dependent manner.
Paclitaxel activated canonical NF-κB signaling via the TLR4/MyD88 pathway and induced Ccl2 transcription via TLR4/MyD88/NF-κB signaling. A, Representative protein expression of TLR4, MyD88, and TRIF in ID8 cells after paclitaxel treatment. B, Representative expression of p105, p50, RelA (p65), and RelB in the cytoplasmic or nuclear proteins of ID8 cells after paclitaxel treatment. Representative protein expression of p65 in the cytoplasmic or nuclear proteins of ID8 wild-type (WT) cells and Tlr4-KO clones after paclitaxel treatment (C) and of ID8 cells with TAK-242 pretreatment prior to paclitaxel treatment (D). mRNA expression of Ccl2 in ID8 cells after treatment with paclitaxel, cisplatin, carboplatin, or mitoxantrone (left), after paclitaxel treatment (right; E), and in ID8 WT cells and Tlr4-KO clones after paclitaxel treatment (F). G, mRNA expression of Tlr4 in ID8 cells with treatment of Tlr4-specific siRNA (siTlr4; left). mRNA expression of Ccl2 in ID8 cells with pretreatment of siTlr4 prior to paclitaxel treatment (middle). Ccl2 promoter activity in ID8 cells with pretreatment of siTlr4 prior to paclitaxel treatment (right). H, mRNA expression of Ccl2 in ID8 cells with pretreatment of TAK-242 prior to paclitaxel treatment. mRNA expression of Nfkbiz (IκBζ) in ID8 cells after paclitaxel treatment (I); in ID8 WT cells and ID8 Tlr4-KO clones after paclitaxel treatment (J); and in ID8 cells with pretreatment of siTlr4 (K) and TAK-242 prior to paclitaxel treatment (L). M, Representative expression of p65 in the cytoplasmic and nuclear proteins of ID8 cells with pretreatment of BAY 11–7082 prior to paclitaxel treatment. N, mRNA expressions of Ccl2 (left) and Nfkbiz (right) in ID8 cells with pretreatment of BAY 11-7082 prior to paclitaxel treatment. O, Ccl2 promoter activity in ID8 cells with pretreatment of BAY 11-7082 prior to paclitaxel treatment. ***, P < 0.001. CBDCA, carboplatin; CDDP, cisplatin; MTX, mitoxantrone; PTX, paclitaxel.
Paclitaxel induced Ccl2 transcription via TLR4/MyD88/NF-κB signaling
Chemokine CCL2/chemokine receptor CCR2 signaling axis is essential for the recruitment of functional antigen-presenting cells into tumors upon anthracycline-based chemotherapy (14). We investigated whether CCL2 was involved in the paclitaxel-induced ICD, thus driving infiltration of dendritic cells into tumors. Paclitaxel (but not carboplatin, cisplatin, or mitoxantrone) induced upregulation of Ccl2 mRNA expression in ID8 cells (Fig. 3E). Paclitaxel-induced Ccl2 upregulation was reduced in the ID8 Tlr4-KO clones (Fig. 3F) and wild-type ID8 cells with the pretreatment of Tlr4-specific siRNA (Fig. 3G) or TAK-242 (Fig. 3H). These data indicated that paclitaxel induced Ccl2 upregulation in murine ovarian cancer cells in a TLR4-dependent manner.
We investigated the role of a transcriptional key regulator of CCL2, namely IκBζ (Nfkbiz, nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor zeta) in paclitaxel-induced Ccl2 transcription in ovarian cancer cells (38). The transcription of IκBζ is regulated by TLR4/MyD88/NFκB pathway, which forms a complex with p50 and p65 on the promoter of target genes and activates their expression (39). Paclitaxel induced significant upregulation of Nfkbiz mRNA in wild-type ID8 cells (Fig. 3I), which was reduced in the ID8 Tlr4-KO clones (Fig. 3J) and wild-type ID8 cells with the pretreatment of Tlr4-specific siRNA (Fig. 3K) or TAK-242 (Fig. 3L). These data indicated that paclitaxel-induced Nfkbiz upregulation in murine ovarian cancer cells was dependent on TLR4. We confirmed these data by blocking the NF-κB p65 activation of wild-type ID8 cells with BAY 11-7082 (ref. 40; Fig. 3M; Supplementary Fig. S4E), and showed that this suppressed the paclitaxel-induced Nfkbiz, Ccl2 mRNA transcription (Fig. 3N), and Ccl2 promoter activity (Fig. 3O). Paclitaxel thus induced Ccl2 transcription in murine ovarian cancer cells via the TLR4/NFκB signaling pathway.
Paclitaxel activated protein kinase R–like ER kinase and eukaryotic translation initiation factor 2α in a TLR4-independent manner
We also investigated the upstream signaling pathway of CALR and ERp57 exposure in the paclitaxel-induced ICD. CALR exposure is reliant on ER stress that triggers protein kinase R–like ER kinase (PERK) activation and eukaryotic translation initiation factor 2α (eIF2α) phosphorylation (1, 27) to induce ICD. Paclitaxel and mitoxantrone, but not carboplatin and cisplatin, induced PERK activation and eIF2α phosphorylation (41, 42), in both ID8 and TKO murine ovarian cancer cell lines (Fig. 4A–C; Supplementary Fig. S4F–S4H). Paclitaxel and mitoxantrone induced PERK activation and eIF2α phosphorylation in the ID8 Tlr4-KO clones when compared with wild-type ID8 cells (Fig. 4D and E; Supplementary Fig S4I and S4J). In addition, the pretreatment with TAK-242 did not affect paclitaxel-induced PERK activation or eIF2α phosphorylation in the ID8 cells (Fig. 4F; Supplementary Fig. S4K). Our results, therefore, indicated paclitaxel-induced PERK activation and eIF2α phosphorylation in murine ovarian cancer cells independent of TLR4.
Paclitaxel activated PERK and eIF2α in a TLR4-independent manner. Representative protein expression of PERK, eIF2α, and phosphorylated-eIF2α in ID8 cells after treatments with paclitaxel, cisplatin, carboplatin, or mitoxantrone (A), in ID8 and TKO cells after paclitaxel treatment (B and C), in ID8 wild-type (WT) cells and Tlr4-KO clones after paclitaxel treatment (D), in ID8 WT cells and Tlr4-KO clones with mitoxantrone treatment (E), and in ID8 cells with TAK-242 pretreatment prior to paclitaxel treatment (F). CBDCA, carboplatin; CDDP, cisplatin; MTX, mitoxantrone; p-eIF2α, phosphorylated-eIF2α; PTX, paclitaxel.
Paclitaxel induced SNARE-dependent exocytosis of ATP-containing vesicles
We demonstrated that in the absence of TLR4, paclitaxel induced ER stress (PERK activation and eIF2α phosphorylation; Fig. 4D) but not CALR exposure (Fig. 2A) in murine ovarian cancer cells. Here, we determined the mechanism underlying CALR exposure based on the “translocation module,” which involves the anterograde transport of CALR by actin cytoskeleton from ER to Golgi apparatus, followed by active exocytosis of CALR-containing vesicles to plasma membrane surface via molecular interactions between vesicle-associated SNAREs (soluble N-ethylmaleimide-sensitive factor attachment protein receptors, such as VAMP1) and plasma membrane–associated SNAREs (such as SNAP23/25; refs. 1, 43, 44). F-actin was exposed on the cell surface of ID8 wild-type cells treated with paclitaxel or mitoxantrone (Supplementary Fig. S5A and S5B). Paclitaxel-induced F-actin exposure was reduced in the ID8 Tlr4-KO clones (Supplementary Fig. S5C), suggesting paclitaxel induced the actin-cytoskeleton–mediated anterograde transport of CALR from ER to Golgi via a TLR4-dependent pathway.
We determined the role of paclitaxel-induced SNARE-dependent vesicle exocytosis in murine ovarian cancer cells via IkappaB kinase 2 (IKK2) activation. SNARE-dependent vesicle exocytosis is the final step of CALR exposure to the cell surface, in which depletion of the plasma membrane–associated SNARE, namely SNAP23, abolishes the anthracycline-induced ICD in mouse colon tumor (1, 27). The TLR/MyD88 signaling pathway transduces the IKK2-mediated phosphorylation of SNAP23 in dendritic cells (45), and leads to mast cell degranulation and controls platelet secretion (46, 47). Paclitaxel induced IKK2 phosphorylation in wild-type ID8 cells (Fig. 5A; Supplementary Fig. S5D), which was diminished in the Tlr4-KO clones and TAK-242–treated wild-type ID8 cells (Fig. 5B and C; Supplementary Fig. S5E and S5F). These data indicated that paclitaxel induced IKK2 activation in ovarian cancer cells via TLR4/MyD88 signaling. Paclitaxel also induced SNAP23 phosphorylation in wild-type (Fig. 5D; Supplementary Fig. S5G) but not in the ID8 Tlr4-KO clones or TAK-242–treated wild-type ID8 cells (Fig. 5E and F; Supplementary Fig. S5H and S5I). Blocking IKK2 phosphorylation with TPCA-1, but not an IKK2 inhibitor IV, abrogated the paclitaxel-induced IKK2 phosphorylation in ID8 cells (Fig. 5G; Supplementary Fig. S5J). Paclitaxel-induced SNAP23 phosphorylation was reduced in the TPCA-1–treated wild-type ID8 cells (Fig. 5H; Supplementary Fig. S5K). These results indicated that paclitaxel induced IKK2-mediated phosphorylation of SNAP23 in murine ovarian cancer cells via the TLR4 pathway. Paclitaxel-induced CALR, ERp57, and ANXA1 exposure were also reduced in the wild-type ID8 cells pretreated with TPCA-1 (Fig. 5I; Supplementary Fig. S5L), suggesting IKK2-mediated SNAP23 phosphorylation played a key role in the ICD-associated DAMPs induced by paclitaxel.
Paclitaxel induced SNARE-dependent exocytosis of ATP-containing vesicles. Representative protein expression of IKK2 and phosphorylated-IKK2 (p-IKK2) in ID8 cells after paclitaxel treatment (A), in ID8 wild-type (WT) cells and ID8 Tlr4-KO clones after paclitaxel treatment (B), and in ID8 cells with pretreatment of TAK-242 prior to paclitaxel treatment (C). D, Representative protein expression of SNAP23 and phosphorylated-SNAP23 (p-SNAP23) in ID8 cells after paclitaxel treatment. Representative protein expression of SNAP23 and p-SNAP23 in ID8 WT cells and Tlr4-KO clones after paclitaxel treatment (E) and in ID8 cells given TAK-242 prior to paclitaxel treatment (F). G, Representative protein expression of IKK2 and p-IKK2 in ID8 cells given TPCA-1 or IKK2 inhibitor V prior to paclitaxel treatment. H, Representative protein expression of SNAP23 and p-SNAP23 in ID8 cells given TPCA-1 or IKK2 inhibitor IV prior to paclitaxel treatment. I, Representative surface expression of CALR (left), ERp57 (middle), and ANXA-1 (right) in ID8 cells given TPCA-1 prior to paclitaxel treatment. Representative images showing the detection of ATP-containing vesicles (green) in ID8 WT cells and Tlr4-KO clones after paclitaxel treatment (J) and in ID8 cells with pretreatment of TAK-242 or TPCA-1 prior to paclitaxel treatment (K). Scale bar, 25 μm. PTX, paclitaxel.
Next, we studied the effect of paclitaxel on vesicle exocytosis in ovarian cancer cells by monitoring ATP-containing vesicles using quinacrine staining (21). Paclitaxel induced quinacrine-positive ATP-containing vesicles in the wild-type ID8 cells, suggesting paclitaxel induced ATP synthesis for secretion (Fig. 5J). A substantially higher number of ATP-containing vesicles induced by paclitaxel were observed in the Tlr4-KO clones and the TAK-242- or TPCA-1–treated wild-type ID8 cells (Fig. 5J and K). The increased number of ATP-containing vesicles may have correlated with the reduced activities in SNARE-dependent vesicle exocytosis associated with the absence of TLR4 activities (Fig. 5E and F) or IKK2 phosphorylation (Fig. 5H), thus retaining ATP inside the cell. These results indicated that paclitaxel induced SNARE-dependent vesicle exocytosis in ovarian cancer cells via TLR4-mediated IKK2 activation.
Paclitaxel induced ICD and antitumor immune responses in human ovarian cancer
We wanted to determine whether paclitaxel induced ICD and subsequent antitumor responses in human ovarian cancer. Anthracyclines induce the expression of several ICD-associated DAMPs and antitumor immunity in ovarian cancer (7). In human ovarian cancer cell line SKOV3, paclitaxel induced CALR and ERp57 exposure, ATP secretion, and HMGB1 release in vitro (Fig. 6A; Supplementary Fig. S5M). We also investigated the immune cell profiling of ovarian tumor microenvironment from the advanced ovarian cancer biopsy expression dataset GSE15622 that contains both responsive and resistant cases to paclitaxel monotherapy (23). In the GSE15622 dataset, there were 14 patients (of 20) with ovarian cancer who had a biopsy performed before and after three cycles of paclitaxel monotherapy. We have analyzed 33 immune cell types in these 14 cases by gene expression enrichment analysis using xCell pipeline (a novel gene signature–based method; ref. 24) in pre- and postpaclitaxel monotherapy samples. ESs of various types of CD4+ memory T cells, CD8+ naive central memory T (Tcm) cells, CD4+ T cells, CD8+ T cells, and CD4+ naïve T cells were significantly higher in posttreatment than in the pretreatment samples of responsive patients (Fig. 6B). Such differences in T-cell induction were not observed between the post- and pretreatment samples of resistant patients (Fig. 6B). We also analyzed T-cell infiltration in nine cases of advanced ovarian tumors from patients receiving NACT (eight cases with paclitaxel–carboplatin regimen and one case with carboplatin monotherapy) followed by interval debulking surgery and nine cases of HGSOC from patients receiving primary debulking surgery (at Department of Obstetrics and Gynaecology, The Chinese University of Hong Kong, Prince of Wales Hospital, Hong Kong) by multiple IHC. The number of CD8+ T cells, IFNγ+ cells, and IFNγ+CD8+ T cells were significantly higher in post-NACT ovarian tumors compared with those in primary ovarian tumors (Fig. 6C; Supplementary Fig. S6A–S6C). The number of positively labeled infiltrating CD8+ T cells, IFNγ+ cells, and IFNγ+CD8+ T cells were smaller in the ovarian tumors of patients with carboplatin monotherapy (Fig. 6C; Supplementary Fig. S6A–S6C) than those with combinatorial therapy. These data suggested that paclitaxel chemotherapy induced T-cell infiltration in ovarian tumors of the responsive patients.
Paclitaxel induced ICD and antitumor immunity in human ovarian cancer. A, Representative surface expression of CALR (top left) and ERp57 (top right), extracellular ATP concentration (bottom left) and intracellular ATP concentration (bottom middle), and soluble HMGB1 (bottom right) in SKOV3 human ovarian cancer cells after paclitaxel treatment. B, Immune cell profiling in the ovarian tumor microenvironment of pre- and postpaclitaxel monotherapy samples in the GSE15622 dataset. Volcano plots showed the gene expression ESs of a total of 33 immune cell types in the post- versus prepaclitaxel biopsies of the resistant (n = 3; left) and responsive (n = 11; right) patients. Red circles represent significant upregulated immune cell types after paclitaxel monotherapy; blue circles represent significant downregulated immune cell types after paclitaxel monotherapy; and gray circles represent the immune cell types with no significant difference after paclitaxel monotherapy. P < 0.05 (orange line) and fold change < −2 (blue line) or > +2 (red line) were regarded as significant. C, Representative images of multiple IHC of CD8, IFNγ, and CD208 in primary ovarian tumors without NACT (top left), post-NACT tumors with carboplatin monotherapy (top middle), and post-NACT tumors with paclitaxel–carboplatin regimen (top right) are shown. Scale bar, 100 μm. Cell densities of CD8+ T cells, IFNγ+ cells, IFNγ+CD8+ T cells, and CD208+ dendritic cells (DC) in primary and post-NACT ovarian tumors were analyzed. Green squares represent the data of a post-NACT sample from a patient with NACT carboplatin monotherapy. D, Correlations between the gene expression ESs of different T-cell subtypes and mRNA expression of TLR4 in post-paclitaxel monotherapy samples of responsive patients (n = 11) by Spearman correlation. E, Representative protein expression of TLR4 (left) and CALR (right) in the primary tumors of patients with EOC (n = 124) by IHC. Representative images of the tumors with low, medium, and high TLR4 cytoplasmic expression (top left) and CALR surface expression (top right) are shown (magnification, 200×; scale bar, 100 μm). Overall survival (OS) and disease-free survival (DFS) of patients with high TLR4 (n = 43) compared with those with low/medium TLR4 (n = 81; bottom left) and patients with high CALR (n = 42) compared with those with low/medium CALR (n = 82; bottom right) were analyzed by Kaplan–Meier survival analysis and log-rank test. F, Comparison of mRNA expression of CALR (log2 RMA-signal) between the prechemotherapy samples of patients who were resistant to paclitaxel monotherapy (n = 7 in the GSE15622 study) and those who responded to paclitaxel monotherapy (n = 13 in the GSE15622 study). G, Donut charts showing chemotherapy responses (responsive in red; partial responsive in blue; and resistant in gray) of patients with CALRlow/medium EOC (n = 82; top left) and patients with CALRhigh EOC (n = 42; top right). Correlations between chemotherapy responses (responsive, partial responsive, or resistant) and CALR surface expression (CALRlow/medium or CALRhigh) in the cancer cells of patients with EOC (n = 124) were analyzed by χ2test (bottom). *, P < 0.05; **, P < 0.01; and ***, P < 0.001. CBDCA, carboplatin; OvCa, ovarian cancer; PTX, paclitaxel.
We then tested whether the T-cell infiltration after paclitaxel correlated with TLR4 expression in ovarian cancer. In the postpaclitaxel monotherapy samples of responsive patients (n = 11; GSE15622), TLR4 mRNA expression positively correlated with T-cell gene expression ESs (including CD8+ T cells, CD8+ effector memory T (Tem) cells, CD4+ Tem cells, CD8+ Tcm cells, and CD4+ naïve T cells; Fig. 6D), indicating that high TLR4 expression in ovarian tumor favor higher amounts of T-cell infiltration after paclitaxel chemotherapy. Next, we studied the prognostic impact of TLR4 and CALR expression in ovarian cancer by IHC in the tumor samples of 124 patients with EOC (including 45 high-grade serous, 42 clear cell, 22 endometrioid, 13 mucinous, 1 adenocarcinoma, and 1 mixed serous and clear cell; Supplementary Table S1) who received primary debulking surgery. Positive signals of TLR4 expression in the cytoplasm and CALR expression on the surface membrane of ovarian epithelial tumor cells were evaluated by H-score using an imaging software inForm, in which patients with EOC were divided into three groups based on their expression levels of TLR4 cytoplasmic expression and CALR surface expression according to their respective H-score (low: 0–33 percentile; medium: >33–67 percentile; high: >67–100; Supplementary Figs. S7A and S7B and S8A and S8B). Patients with high CALR surface expression had better overall survival and disease-free survival by Kaplan–Meier survival analyses (Fig. 6E), indicating CALR expressions in primary HGSOC tumors were correlated with patients' survival. These data also suggested that the paclitaxel-induced ICD (via CALR exposure) contributed to the chemotherapy treatment outcomes.
We studied the CALR expression in primary tumors from patients receiving paclitaxel chemotherapy. Among the prepaclitaxel monotherapy samples from the GSE15622 dataset, significantly higher CALR mRNA expression was found in responsive patients when compared with those of resistant patients (Fig. 6F). Our IHC results also showed high CALR surface expression in primary EOC tumors were correlated with better responses to paclitaxel chemotherapy, that is, a significantly higher proportion of responsive patients were found among the group of patients with high CALR surface expression in the EOC tumor cells (Fig. 6G). Our results therefore suggested that the paclitaxel-induced ICD-associated DAMPs were correlated with chemotherapy response.
Discussion
Paclitaxel induced CALR exposure, ATP secretion, HMGM1 release, ANXA1 exposure, and CXCL10 upregulation in murine ovarian cancer cells in vitro, all ICD markers. Our tumor vaccination assays validated that the paclitaxel-induced ICD elicits antitumor immunity in a murine model of ovarian tumor in vivo. Treatment of U2OS human osteosarcoma cells with a chemical library showed that paclitaxel induces significant eIF2α phosphorylation, HMGB1 release, and ATP secretion in the cell line (48). Hyperploid cancer cells become immunogenic because of a constitutive ER stress response resulting in aberrant CALR surface exposure, and that the treatment of tetraploidizing agents (including 300 nmol/L paclitaxel) induces CALR exposure in U2OS cells (49). We found that paclitaxel induced G2–M-phase arrest and >4N (i.e., multinucleated) cell populations in both the wild-type ID8 cells and Tlr4-KO clones, and that paclitaxel-induced CALR exposure was reduced in the Tlr4-KO clones. These data thus suggested that tetraploidization did not correlate with the paclitaxel-induced ICD in murine ovarian cancer models.
The antitumor immunity induced by paclitaxel modulated tumor-associated macrophages from a M2-like profile toward an immune stimulatory M1-like phenotype via TLR4 signaling pathway in the murine models of breast and melanoma tumors (50). Here, cancer cell–autonomous TLR4 was essential for the paclitaxel-induced CALR exposure in murine ovarian cancer cells. There are three hierarchical modules of CALR exposure in response to ICD inducers such as anthracyclines (43), including: (i) “ER stress module,” which involves eIF2α phosphorylation due to PERK activation (the kinase) or PP1/GADD34 inhibition (the phosphatase), ROS and/or NO production, and possible SERCA inhibition; (ii) “translocation module,” in which CALR is anterogradely transported from ER lumen to Golgi facilitated by actin cytoskeleton rearrangement, and the vesicle-associated SNARE (VAMP1) interacts with the plasma membrane–associated SNARE (SNAP23) and mediates membrane fusion by allowing luminal CALR to reach its final destination, the cell surface; and (iii) “apoptotic module,” which might mediate an interorganellar cross-talk between the ER and mitochondria via apical caspase-8–dependent cleavage of Bap31 and activation of the Bcl-2 family proteins Bax and Bak. Here, paclitaxel induced PERK activation and eIF2α phosphorylation in the Tlr4-KO clones, indicating the paclitaxel-induced “ER stress module” of CALR exposure pathway was independent of TLR4. Paclitaxel-induced F-actin exposure in murine ovarian cancer cells was TLR4 dependent, suggesting actin cytoskeleton mediated the anterograde transport of CALR from ER to Golgi may have been involved in paclitaxel-induced CALR exposure via TLR4 signaling pathway. Other studies have regarded F-actin as a DAMP recognized by DNGR-1 (also known as CLEC9A; refs. 51–53), a dendritic cell receptor that couples sensing of necrosis to immunity (54). Our discovery of F-actin exposure in paclitaxel/mitoxantrone-induced ICD supports F-actin as a new ICD-associated DAMP in chemotherapy. In addition, paclitaxel induced IKK2 phosphorylation, SNAP23 phosphorylation, and ATP-containing vesicle exocytosis in wild-type murine ovarian cancer cells, with reduction in the Tlr4-KO clones. These results indicated that paclitaxel induced the “translocation module” (i.e., SNARE-dependent vesicle exocytosis) of CALR exposure pathway via TLR4/IKK2 signaling.
Paclitaxel induced ATP secretion in murine ovarian cancer cells via TLR4 signaling pathway. Autophagy is essential for the immunogenic ATP release from dying tumor cells (1, 55). In response to anthracycline chemotherapy, autophagy-competent cancers attracted dendritic cells and T lymphocytes into the tumor bed, whereas suppression of autophagy inhibited the release of ATP from the dying cells. There was only a weak induction of autophagy detected in murine ovarian cancer cells after paclitaxel treatment when compared with that of the positive control using thapsigargin, indicating that paclitaxel-induced ATP release in ovarian cancer cells did not depend on autophagy.
In conclusion, our results demonstrated that paclitaxel induced ICD in ovarian cancer via TLR4-independent and TLR4-dependent pathways as summarized in Fig. 7. Our results therefore provided new evidence that the antitumor effects of paclitaxel are in part mediated by the activation of antitumor immunity via ICD and opens up the possibility of a combination of paclitaxel and immunotherapies as a new treatment for ovarian cancer.
Paclitaxel induced ICD in ovarian cancer through TLR4-independent and -dependent pathways. A schematic diagram showing how (1) paclitaxel induced mitotic arrest, cell death, and ER stress via TLR4-independent pathways; (2) paclitaxel induced Ccl2 (likewise Cxcl10) upregulation via TLR4/MyD88-NFκB pathways; and (3) paclitaxel induced SNARE-dependent vesicle exocytosis via TLR4-mediated IKK2 activation. The paclitaxel-induced vesicle exocytosis was essential to the release of ICD-associated DAMPs. PTX, paclitaxel.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: T.S. Lau, L.K.Y. Chan, J. Kwong
Development of methodology: T.S. Lau, L.K.Y. Chan, G.C.W. Man, J. Kwong
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.S. Lau, L.K.Y. Chan, G.C.W. Man, C.H. Wong, J.H.S. Lee, T.H. Cheung, I.A. McNeish, J. Kwong
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.S. Lau, L.K.Y. Chan, G.C.W. Man, J. Kwong
Writing, review, and/or revision of the manuscript: T.S. Lau, L.K.Y. Chan, T.H. Cheung, I.A. McNeish, J. Kwong
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.S. Lau, L.K.Y. Chan, S.F. Yim, J. Kwong
Study supervision: T.S. Lau, J. Kwong
Acknowledgments
The authors give special thanks to Prof. Samuel C. Mok, Department of Gynecologic Oncology and Reproductive Medicine, Division of Surgery, The University of Texas MD Anderson Cancer Center (Houston, TX) for the TKO cell line. The authors also give thanks to Dr. Amy Kit Ying Chung for revising the article. This research was supported by Hong Kong Research Grant Council General Research Fund (467713 and 14109515) and The Chinese University of Hong Kong (CUHK) Research Committee Funding (Direct Grants for Research 4054185, 4054352, and 4054410) to J. Kwong.
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.
Footnotes
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).
Cancer Immunol Res 2020;8:1099–111
- Received August 13, 2019.
- Revision received February 25, 2020.
- Accepted April 17, 2020.
- Published first April 30, 2020.
- ©2020 American Association for Cancer Research.
References
Volume 8, Issue 8
