Toll-Like Receptor 3 Expressing Tumor Parenchyma and Infiltrating Natural Killer Cells in Hepatocellular Carcinoma Patients

JNCI Journal of the National Cancer Institute, Dec 2012

Background Hepatocellular carcinoma (HCC) is a highly aggressive cancer that is linked to chronically dysregulated liver inflammation. However, appropriate immune responses can control HCC progression. Here we investigated the role and underlying mechanism of toll-like receptor 3 (TLR3) in HCC.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

https://jnci.oxfordjournals.org/content/104/23/1796.full.pdf

Toll-Like Receptor 3 Expressing Tumor Parenchyma and Infiltrating Natural Killer Cells in Hepatocellular Carcinoma Patients

Valerie Chew Charlene Tow Caleb Huang Emilie Bard-Chapeau Neal G. Copeland Nancy A. Jenkins Achim Weber Kiat Hon Lim Han Chong Toh Mathias Heikenwalder Irene Oi-Lin Ng Alessandra Nardin Jean-Pierre Abastado ) Hepatocellular carcinoma (HCC) is a highly aggressive cancer that is linked to chronically dysregulated liver inflammation. However, appropriate immune responses can control HCC progression. Here we investigated the role and underlying mechanism of toll-like receptor 3 (TLR3) in HCC. HCC cell death, and natural killer (NK) cell activation and cytotoxicity were assessed in vitro after treatment with the TLR3 ligand poly(I:C). The effect of TLR3 on the tumor parenchyma and infiltrating immune cells was investigated in a spontaneous liver tumor mouse model and a transplanted tumor mouse model (n=3-9 mice per group). Immunohistochemistry and quantitative polymerase chain reaction were used to analyze tumor samples from 172 HCC patients. Paired t-tests and analysis of variance tests were used to calculate P-values.The relationship betweenTLR3 expression and survival was determined by the Kaplan-Meier univariate survival analysis and a log-rank test. All statistical tests were two-sided. TLR3 activation increased cell death in the TLR3+ SNU182 HCC cell line (30.5% vs 8.5%, P=.03) and promoted NK-cell activation (32.6% vs 19.4%, P < .001) and cytotoxicity (relative fourfold increase, P = .03) in vitro. In vivo, poly(I:C) treatment increased intratumoral chemokine expression, NK-cell activation and tumor infiltration, and proliferation of tumor-infiltrating T and NK cells. Proliferation of tumor parenchyma cells was decreased. Also, expression of chemokines or treatment with poly(I:C) decreased tumor growth. TLR3 expression in patient samples correlated with NK-cell activation, NK- and T-cell tumor infiltration, and inversely correlated with tumor parenchyma cell viability.TLR3 expression was also associated with longer survival in HCC patients (hazard ratio of survival=2.1, 95% confidence interval=1.3 to 3.4, P=.002). Background Methods Results Conclusions TLR3 is an important modulator of HCC progression and is a potential target for novel immunotherapy. Hepatocellular carcinoma (HCC) is the fifth most common cancer and the third leading cause of cancer-related death worldwide (1). Resection and transplantation are the most successful treatments for HCC; however, most patients relapse and overall survival remains poor. HCC is thought to result from persistent, nonspecific activation of the immune system within the chronically inflamed liver, causing repeated cycles of tissue damage, repair and regeneration, and eventually carcinogenesis (2,3). However, we have previously reported that expression of specific proinflammatory genes within the tumor microenvironment is associated with improved patient survival (4), suggesting a more complex role of the immune system inHCC. Toll-like receptor 3 (TLR3) is one of the key proinflammatory genes associated with good prognosis in HCC (4). TLR3 is an endosomal receptor for double-stranded RNA and is expressed in several subsets of immune cells, including dendritic cells (5) and natural killer (NK) cells (6). TLR3 is also expressed by fibroblasts (7), lung epithelial cells (8), hepatocytes (9), and several types of tumor cells, including breast cancer and melanoma cells (10,11). TLR3 is involved in antiviral responses and the production of type Iinterferons (IFNs) (12). RNA from damaged tissues can also serve as a ligand for TLR3 (13,14). Synthetic TLR3 ligands, such as polyinosinic:polycytidylic acid [poly(I:C)] acts as potent immune adjuvants by enhancing dendritic cell cross-presentation and promoting CD8+ T-cell responses (15,16). Such ligands have thus been used to treat a range of malignancies in a variety of clinical settings (17). In vitro studies indicate that TLR3 activation triggers apoptosis of several tumor cell lines, including breast cancer (10) and melanoma (11). Moreover, transfection of TLR3 into HCC cell lines renders them sensitive to poly(I:C)-induced killing (18). TLR3expressing NK cells can also be directly activated by poly(I:C) (6,19). However, the role of TLR3 in HCC remains to be evaluated in human patients. We explored the role played by TLR3 in the apoptosis of HCC cells and assessed the contribution of this receptor to NK-cell activation and cytotoxicity using HCC cell lines, two in vivo mouse models, and patient samples and survival data. Materials and Methods Cell Lines and Reagents The methods by which the HCC cell lines (Hepa1-6, SNU182, SNU368, SNU387, SNU-423, SNU449, SNU475, PLC/PR5, HuH-7, HepG2, Hep3B, and SK-HEP-1) were maintained are described in detail in the Supplementary Methods (available online). Endotoxin-free poly (I:C) and polyadenylicpolyuridylic acid [poly(A:U)] were obtained from InvivoGen (San Diego, CA). We assumed the authenticity of the cell lines was verified by thesource. All the antibodies used for immunohistochemistry, flow cytometry, or NK-cell depletion were listed in Supplementary Table 1 (available online). TLR3 was knocked down using predesigned stealth select RNAi (Invitrogen) according to the manufacturers instructions (Supplementary Methods, available online). Patient Samples Resected tumor samples were obtained from patients who underwent curative resection for HCC (n=172) between 1991 and 2009 at the National Cancer Centre (Singapore), Queen Mary Hospital (Hong Kong, China), or the University Hospital of Zurich (Zurich, Switzerland). All samples were collected with written informed consent in compliance with the requirements of the local ethical committee at each institution. The patient demographics and clinical characteristics are summarized in Supplementary Table2 (available online) (20). RNA isolation, cDNA conversion, quantitative polymerase chain reaction (qPCR), and gene expression analysis were performed as described in the Supplementary Methods and Supplementary Table 3 (available online). Because of the small number of women in the study population, we did not report subanalyses by sex, but analysis by major racial/ethnic group was done. In Vivo Studies Animal care and experimental procedures were approved by the Institutional Animal Care and Use Committee, Biological Resource Centre, Agency for Science, Technology and Research (A*STAR), Singapore. The spontaneous liver tumor mouse model and transplanted tumor mouse model are described in the Supplementary Methods (available online). Immunohistochemistry Paraffin-embedded tissue sections of liver and tumor from HCC patients (n=40) or from mice used in the above experiments were processed as previously described (4). All the antibodies used are listed in Supplementary Table1 (available online). Aterminal deoxynucleotidyl transferase dUTP nick end labeling kit from Millipore (Billerica, MA) was used as per the manufacturers protocol. Immunohistochemistry (IHC) images were captured with a DP20 camera attached to a CX31 microscope (Olympus, Center Valley, PA). For immunoflourescence images, an Olympus FlourView FV1000 confocal microscope was used. The number of positively stained cells in 1012 random fields were counted at 100 magnification using Image-Pro Analyzer version 6.2 (Media Cybernetics, Inc, Rockville, MD). Flow Cytometry Apoptotic HCC cells were identified by staining with Annexin V (BD Pharmingen, San Diego, CA) and Topro-3 (Invitrogen). Activation of CD56+ human NK cells was determined by assessing the protein expression of the early activation marker CD69 using antihuman CD69 antibody (eBiosciences, San Diego, CA). Immune cells were isolated from mouse liver tumors or healthy livers by digesting the tissues with Collagenase A (1 mg/mL) and DNase I(0.1 mg/mL; Roche, Indianapolis, IN) and were analyzed using antibodies listed in Supplementary Table1 (available online). All flow cytometry analyses were done on a BD LSRII flow cytometer equipped with five lasers (BD Biosciences, San Jose, CA), and the data were processed using FlowJo version 7.5 software (Tree Star, Inc, Ashland, OR). NK-Cell Activation and CytotoxicityAssay Blood samples from 15 healthy donors were obtained from the Singapore Health Science Authority blood bank following Ethics Committee approval. NK cells were isolated by negative selection using the NK-cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany), treated, and analyzed by using the DELFIA EuTDA-based cytotoxicity assay kit (Perkin-Elmer, Boston, MA) by following the manufacturers instructions (Supplementary Methods, available online). IFN- present in supernatants from activated NK cells was measured by enzyme-linked immunosorbent assay (Biolegend, San Diego, CA) by following the manufacturers instructions. Statistical Analysis KaplanMeier univariate survival analysis was performed using data from HCC patients and hazard ratios (HRs), 95% confidence intervals (CIs), and P-values were calculated using the log-rank (MantelCox) test. Paired or unpaired t-tests, MannWhitney tests, or one- or two-way analysis of variance tests were also used in various experiments as indicated. A P-value less than .05 was considered statistically significant. All statistical tests were twosided. For more detailed information on the statistical analyses, see the Supplementary Methods (available online). TLR3 Expression in Tumor Parenchyma and Infiltrating NK Cells in HCC Patient Samples We previously reported that TLR3 gene expression in tumors was associated with overall survival among a cohort of 61 patients from Singapore (4). Here, we extended our previous study to include 111 additional patients from Hong Kong and Zurich. First, we performed qPCR on patient samples and confirmed that expression of TLR3 is associated with overall survival (HR of survival = 2.1, 95% CI = 1.3 to 3.4, P = .002) (Figure 1, A). We also observed Figure 1. The association of intratumor toll-like receptor 3 (TLR3) expression by tumor parenchyma and natural killer (NK) cells with patient survival. KaplanMeier analyses were done to investigate the association between A) TLR3 gene expression (by quantitative polymerase chain reaction, n=172) and B)TLR3 protein levels (by immunohistochemistry [IHC], n=40, median=33 cells per field) and patient survival. Log-rank MantelCox test was used to calculate two-sided P-values. Representative IHC images at 400 magnification of C) TLR3+ tumor parenchyma (red indicates TLR3 staining and blue indicates CD3 staining) and D)TLR3+ NK cells (red indicatesTLR3 staining; NK cells are identified by granzyme-B expression in blue; and costaining appears purple) are shown. The insets show magnified areas. Scale bar = 30 m. E) KaplanMeier analyses of TLR3+ tumor cell (median=21 cells per field) and tumor-infiltrating NK cells densities (median = 13 cells per field) and patient survival (n=40). The log-rank MantelCox test was used to calculate two-sided P-values. CI=confidence interval, HR=hazard ratio. a statistically significant association using IHC to analyze TLR3 protein expression in 40 representative tumor samples (HR of survival=17.3, 95% CI=6.4 to 47.1, P < .001) (Figure1, B). TLR3 was either expressed by tumor parenchyma (Figure1, C) or small round immune cells infiltrating the tumor nest (Figure1, D). TLR3 was coexpressed with granzyme B but not with CD3 or CD20, suggesting that the cells that stained positively for TLR3 were NK cells (Figure1, D) and not T or B cells (Supplementary Figure1, available online). We next quantified TLR3 expression separately in tumor parenchyma (hepatocytes) and NK cells to determine if these two cell populations were associated with overall survival. As shown in Figure1, E, both TLR3-expressing cell populations were associated with overall survival in KaplanMeier analyses (tumor parenchyma: HR of survival=7.3, 95% CI=2.8 to 19.1, P < .001; tumor-infiltrating NK cells: HR of survival=8.7, 95% CI=3.2 to 23.4, P <.001). Effect of TLR3 Activation on HCC CellDeath To study the effects of endogenous expression of TLR3 in HCC, we screened 11 human HCC cell lines for TLR3 RNA expression by qPCR and identified sevenSNU182, SNU387, SNU449, SNU475, SK-HEP-1, HepG2, and Hep3Bthat expressed the receptor (data not shown). Treatment of SNU182 cells, which exhibited high expression of TLR3, with increasing concentration of poly(I:C) for 24 hours at 37C, resulted in a dose-dependent increase in cell death (no treatment vs 50g/mL poly(I:C): 8.5% vs 30.5%, P=.03; no treatment vs 100g/mL poly(I:C): 8.5% vs 30.9%, P = .03) (Figure 2, A). Similar results were observed for SNU449 and SNU475 cells that express moderate levels of TLR3 (Supplementary Figure 2, A, available online). TLR3 expression was increased in a dose-dependent manner by treatment with poly(I:C) in SNU182 cells (Supplementary Figure 2, B, available online). Ligands for other TLRs (lipopolysaccharide and CpG) did not induce cell death in SNU182 cells (Supplementary Figure 2, C, available online), and no effect was observed in PLC/PR5 cells that do not express TLR3 (Supplementary Figure 2, D, available online). In addition, knock-down of TLR3 in SNU182 cells reduced poly(I:C)-induced cell death (scrambled vs TLR3 siRNA: 39.9% vs 27.2%, P=.03) (Figure2, B, and Supplementary Figure2, E, available online). As poly(I:C) is a ligand for both TLR3 and MDA5 (21), we treated the cells with the TLR3-specific ligand poly(A:U) and observed similar results (Supplementary Figure2, F, available online). These data indicate that TLR3 activation induces cell death in HCC cells that express endogenousTLR3. Effect of TLR3 Activation on NK-Cell Activation and Cytotoxicity Against HCCCells We next evaluated the antitumor properties of NK cells following TLR3 activation. NK-cell treatment with poly(I:C) led to a marked increase in expression of the early activation marker CD69 [percentage of CD56+CD69+ cells relative to all live CD45+ cells after poly(I:C) treatment vs control = 32.6% vs 19.4%, P < .001] (Figure 3, A) and in expression of cytotoxic mediators, such as IFN- [ninefold increase in IFNG gene expression after treatment with poly(I:C), P < .001; mean secreted IFN- after treatment with poly(I:C) vs control = 415 vs 179 pg/mL, P = .03] (Figure 3, B and C) and granzyme-B [threefold increase in granzyme-B gene expression after treatment with poly(I:C), P < .001] (Figure3, B). Activated NK cells displayed up to fourfold higher cytotoxic activity (E:T ratio=5:1, P=.03) against SNU182 cells (Figure3, D). Thus, in vitro activation of TLR3 promotes NK-cell activation, secretion of IFN-, and cytotoxicity against HCC cells. This result is consistent with previous studies using lymphoma target cells (19). Effect of TLR3 Activation on Intratumoral Chemokine Expression and NK-Cell Infiltration To investigate the effect of TLR3 activation in vivo, we treated mice that spontaneously develop liver tumors with poly(I:C) (22,23). In the tumors of treated mice compared with mice injected with PBS or livers from age-matched, tumor-free mice, increase in relative expression of the chemokine genes Ccl5 (HCC of tumor-bearing mice treated with poly(I:C) vs PBS=19-fold increase, P=.01; vs livers of tumor-free mice = 189-fold increase, P = .02) and Cxcl9 (HCC of tumor-bearing mice treated with poly(I:C) vs PBS=fivefold increase, P = .04; vs livers of tumor-free mice = 12-fold increase, P=.04) was observed (Figure4, A). CCL5 and CXCL9 proteins were expressed by both the tumor-infiltrating leukocytes (TILs) and tumor parenchymal cells (Supplementary Figure 3, A, available online). Further studies on TILs revealed expression of CCL5 by T cells and macrophages and CXCL9 expression by macrophages (Supplementary Figure3, B, available online). These results are consistent with our previous data from human HCC samples (20). Furthermore, poly(I:C) treatment induced NK-cell infiltration [percentage of NK 1.1+ cells relative to all live CD45+ cells in HCC of tumor-bearing mice treated with poly(I:C) vs PBS or livers of tumor-free mice = 14.6% vs 4%, P = .002 or 4.2%, P = .02, respectively] (Figure 4, B) and stimulated expression of the NK-cell effector genes Prf and Gzmb [13-fold increase in mean relative expression of Prf in poly(I:C) vs PBS-treated mice, P=.03, and 29-fold increase in mean relative expression of Gzmb in poly(I:C) vs PBS-treated mice, P < .001] (Figure4, C). Increased intratumor expression of Tlr3 was also observed in these mice (Supplementary Figure 3, C, available online), in agreement with the in vitro data presented in Supplementary Figure 2, C (available online). No statistically significant difference was observed in the numbers of tumor-infiltrating T cells or macrophages, or in the expression of Ccl2 or Cxcl10 genes (Supplementary Figure 3, D and E, available online). Treating the mice with poly(A:U) also increased NK-cell infiltration and chemokine expression in the liver tumors (Supplementary Figure4, A and B, available online). The effect of intratumor expression of CCL5 and CXCL9 (24) was further studied in C57BL/6J mice transplanted subcutaneously with the mouse HCC cell line Hepa1-6. Expression of these two chemokines increased intratumor lymphocyte infiltration (fourfold increase in mean relative expression of Cd3g in tumor cells transfected with Ccl5/Cxcl9 vs control (Ctl), P = .009) (Supplementary Figure 5, A, available online) and slowed down tumor growth (mean tumor area of Ccl5/Cxcl9 transfected vs control groups=9.2 vs 22.0 mm2, P=.02 on day 7; 11.7 vs 26.5 mm2, P=.006 on day 10; and 15.8 vs 39.5 mm2, P < .001 on day 14)(Supplementary Figure5, B, available online). When NK cells were depleted using PK136 antibody, tumor growth was enhanced (mean tumor area of PK136 vs isotype control treated groups=42.3 vs 18.7 mm2, P=.005 on day 7; 44.3 vs 16.0 mm2, P < .001 on day 10; and 56.3 vs 23.3 mm2, Figure2. The effect of toll-like receptor 3 (TLR3) activation by poly(I:C) on hepatocellular carcinoma cell line death. A) Dose-dependent induction of cell death in SNU182 cells was measured as Annexin V+ Topro 3+ cells by flow cytometry 24 hours after treatment with 0100 g/mL poly(I:C). The graph shows the means and SD from three independent experiments. Aone-way analysis of variance test with Tukeys multiple comparison post test was used to calculate two-sided P-values. In the right panel, representative dot plots of cell death 24 hours post-treatment with 50g/mL poly(I:C) are shown. B) Poly(I:C)-induced cell death (Annexin V+ Topro 3+ cells by flow cytometry) in SNU182 cells withTLR3 or scrambled siRNA knockdown was measured 24 hours after treatment with 50g/mL poly(I:C). The graphs show the means and SD from three independent experiments. Unpaired t-test was used to calculate two-sided P-values. The right panel shows representative dot plots of poly(I:C)-induced cell death with TLR3 knockdown. Scr=scrambled siRNA. P < .001 on day 14)(Figure4, D), further demonstrating the critical role of NK cells in controlling HCC tumor growth. Effect of TLR3 Activation In Vivo on Intratumor Proliferation ofT Cells and NK Cells and Viability of Tumor Parenchyma We next characterized the functional status of TILs in poly(I:C)treated mice. As shown in Figure5, A, the frequency of Ki67+CD45+ proliferating TILs was ninefold higher in liver tumors from poly(I:C)- vs PBS-treated mice (mean number of Ki67+CD45+ TILs per field = 167 vs 19, respectively, P < .001). Furthermore, double-IHC using antibodies against Ki67 with either CD3 or granzyme B identified these proliferating TILs as T cells and NK cells (Figure5, B). Conversely, proliferation of tumor parenchyma cells (identified as Ki67+CD45 cells with distinct morphology) was fourfold lower in poly(I:C)- vs PBS-treated mice (mean=41 vs 166 cells per field, respectively, P=.002) (Figure5, C), whereas apoptosis of tumor cells was enhanced by 11-fold (mean=32 vs 3 cells per field, respectively, P < .001) (Figure5, D). Similar results were observed when mice were treated with poly(A:U) (Supplementary Figure4, C and D, available online). These data show that TLR3 activation in tumor-bearing mice increased the proliferation of tumor-infiltrating T and NK cells whereas tumor-cell proliferation and viability were decreased. Figure 3. The effect of toll-like receptor 3 (TLR3) activation on natural killer (NK)-cell activation and cytotoxicity against hepatocellular carcinoma cells in vitro. A) The percentage of activated NK cells coexpressing CD56 with the early activation marker CD69 was measured by flow cytometry 24 hours after treatment with 50 g/mL poly(I:C). Data are shown usingTukey box-and-whiskers representation from 15 independent experiments. A paired t-test was used to calculate the two-sided P-value. B) Expression of the effector genes interferon- (IFNG) and granzyme-B (GZB) was determined by quantitative polymerase chain reaction 24 hours after NK cells were treated with 50 g/mL poly(I:C). The graph shows the mean fold change relative to the untreated control from 12 independent experiments. Whisker bars indicate the SD. A Wilcoxon signed-rank test was used to calculate the two-sided P-values. C)The mean interferon- (IFN-) secretion by NK cells 24 hours after treatment with 50g/mL poly(I:C) was measured by enzyme-linked immunosorbent assay of the culture supernatant. The data are shown using Tukey box-and-whiskers representation from eight independent experiments. Apaired t-test was used to calculate the two-sided P-value. D) Cytotoxicity of poly(I:C)-activated NK cells at various ratios of effector (E) to target cells (T, SNU182 cells) was measured by determining the percentage of specific lysis in vitro. Apaired t-test was used to calculate the two-sided P-value at E:T ratio of 5:1. The graph shows the means from poly(I:C)-treated and untreated NK cells and the SD (whisker bars) from three independent experiments. To further show the effect of TLR3 in controlling tumor growth, we treated C57BL/6J mice transplanted with Hepa1-6 cells with poly(I:C) and observed a statistically significant reduction in tumor growth compared with PBS-treated mice (mean tumor area of poly(I:C)- vs PBS-treated mice=14.5 vs 39.5 mm2, P < .001, day 14)(Figure5, E). Infiltration of T cells and NK cells was also increased in transplanted tumors from the poly(I:C)-treated mice (T-cell infiltration: threefold increase in mean relative expression of Cd3g, P=.01; NK-cell infiltration: twofold increase in mean relative expression of Nkp46, P=.04) (Figure5, F). Figure4. The effect of toll-like receptor 3 (TLR3) activation on intratumor chemokine expression and natural killer (NK)-cell infiltration in hepatocellular carcinoma in vivo. A)The mean RNA expression of the Ccl5 and Cxcl9 chemokines in liver tumors from mice treated with poly(I:C) (n = 6), or phosphate buffered saline (PBS)-treated mice (n=4), or livers from nontumor-bearing littermates (n=2), is shown. Whisker bars indicate the SD. Aone-way analysis of variance test with Bonferroni correction was used to calculate the two-sided P-values. B) In the left panel, the mean percentage of NK1.1+ CD45+ NK-cell infiltration of liver tumors harvested from poly(I:C)-treated mice (n=6), or PBS-treated mice (n=4), or livers from non-tumor-bearing littermates (n=2), is shown. Whisker bars indicate the SD. A one-way analysis of variance test with Bonferroni correction was used to calculate the two-sided P-values. In the right panel, representative dot plots showing the percentage of NK1.1+ CD45+ NK cells (box) infiltrating the liver tumors harvested from poly(I:C)- and PBS-treated mice. C) The mean expression of the NK-cell effector genes perforin (Prf) and granzyme-B (Gzmb) was measured in liver tumors harvested from poly(I:C)and PBS-treated mice using quantitative polymerase chain reaction. Graphs show means and SD. The MannWhitney test was used to calculate the two-sided P-values. D)Transplanted tumor growth was measured after NK cells were depleted using PK136 antibody or isotype control antibody (n=3 per group).The graphs show the means and SD (whisker bars). Atwo-way analysis of variance test with Bonferroni correction was used to calculate two-sided P-values. The mean tumor area was statistically significantly different between PK136-treated and isotype control treated mice at days 7, 10, and 14 (P=.005, <.001, and <.001, respectively). Figure 5. The effect of toll-like receptor 3 (TLR3) activation on antitumor activity in vivo. A) The mean number of Ki67+CD45+ proliferating tumor-infiltrating leukocytes per field was assessed by immunohistochemistry in liver tumors from poly(I:C)-treated mice compared with phosphate buffered saline (PBS)-treated mice. Whisker bars indicate the SD. An unpaired t-test was used to calculate the two-sided P-value. B) Representative immunohistochemical images of CD3 or granzyme-B (GZB) staining (red) and Ki67 (blue) at 400 magnification. Membrane expression of CD3, cytoplasmic expression of GZB, and nuclear expression of Ki-67 were observed. The insets show a representative magnified area. Scale bar=30m. C) The mean number of Ki67+CD45proliferating tumor parenchyma cells per field in liver tumors harvested from poly(I:C)-treated mice compared with PBS-treated mice is shown. Whisker bars indicate the SD. The MannWhitney test was used to calculate the two-sided P-value. D) The mean number of apoptotic tumor parenchyma cells per field detected by terminal deoxynucleotidyl transferase dUTP nick end labeling in liver tumors harvested from poly(I:C)treated or PBS-treated mice is shown. Whisker bars indicate the SD. The MannWhitney test was used to calculate the two-sided P-value. Representative immunohistochemical images are also shown at 200 magnification. Scale bar = 50 m. E) Transplanted tumor growth was determined in mice treated with poly(I:C) vs PBS (n=6 mice per group). On day 14, the mean tumor area of poly(I:C)- vs PBS-treated mice is 14.5 vs 39.5 mm2, P < .001. Whisker bars indicate the SD. Atwo-way analysis of variance test with Bonferroni correction was used to calculate the two-sided P-value. F) Relative gene expression of T (threefold increase in Cd3g) and NK (twofold increase in Nkp46) cell markers in transplanted tumors treated with poly(I:C) vs PBS (n=6 mice per group) is shown.The graphs show the means and SD (whisker bars). An unpaired t-test was used to calculate the two-sided P-values. Correlation Between TLR3, NK-Cell Infiltration, and Antitumor Activity in Patient Samples In tumor tissue from HCC patients, a higher density of TLR3expressing cells correlated with increased tumor infiltration by CD56+ NK cells (Pearsons r = 0.7, P < .001) and CD8+ T cells (Pearsons r = 0.6, P <.001) (Figure 6, A). This is consistent with our previous finding that TLR3 RNA expression is associated with expression of chemokines able to attract NK cells and T cells (20). Furthermore, TLR3 protein expression correlated with NK-cell activation marker IFNG expression when measured by qPCR (Pearsons r=0.5, P=.001), and a similar, albeit weaker, correlation was observed with granzyme-B expression when measured by IHC (Pearsons r=0.4, P=.03) (Figure6, B). Correlations between TLR3 and NCR3, CD8A, and IFNG (NK- and T-cell marker) were also observed at the RNA level in patient samples (n = 154164) (Supplementary Table 4, available online). In addition, the density of both TLR3+ tumor parenchyma and NK cells correlated positively with tumor cell apoptosis (TLR3+ tumor parenchyma, Pearsons r = 0.5, P = .006; TLR3+ NK cells, Pearsons r = 0.6, P = .002) and negatively with tumor cell proliferation (TLR3+ tumor parenchyma, Pearsons r = 0.4, P = .03; TLR3+ NK cells, Pearsons r = 0.5, P = .005) (Figure 6, C) and were decreased in samples from patients with stage III and IV advanced HCC (Supplementary Figure 6, available online). These data from HCC patients are consistent with our in vitro and in vivo findings that TLR3 activation is associated with intratumor expression of chemokines, NK- and T-cell infiltration, NK-cell activation, and antitumor activity, as well as enhanced apoptosis and reduced proliferation of tumor parenchymacells. In the past few decades, TLR agonists have emerged as promising drugs for the treatment of cancer (25). Several TLR ligands have since been used to treat cancer patients either as a monotherapy or in combination with other therapies (17). These ligands include the TLR7 ligand imiquimod for superficial basal cell carcinoma (26) and the TLR9 ligand CpG ODN for a variety of carcinomas (27). In particular, TLR3 ligands were used to treat patients with various malignancies, including breast cancer (28) and melanoma (29). We previously showed that TLR3 expression is associated with superior HCC patient survival (4). Here, we confirmed this finding using a larger set of patient samples and explored the underlying mechanisms of the beneficial role of TLR3 activation inHCC. By integrating in vitro data from human cell lines with in vivo treatment of mice that developed spontaneous HCC or were injected with HCC cell line, we demonstrated that the prolonged survival associated with higher expression of TLR3 may result from three independent effects: induction of tumor parenchyma (hepatocytes) cell death; induction of intratumor expression of chemokines that attract NK cells or T cells to the tumor microenvironment; and activation of tumor-infiltrating NK cells that promote cytotoxic activity. These effects were consistent with transcriptomal and histopathological analyses of HCC patient samples. TLR3 is expressed by tumor parenchyma and infiltrating NK cells. We showed that expression in both compartments correlated with reduced tumor cell proliferation, increased tumor cell death, and prolonged patient survival. TLR3 expression by tumor parenchyma was associated with improved survival even in patients with low NCR3 expression (KapleinMeier analysis, HR of survival = 7.1, 95% CI = 1.7 to 30.5, P = .008) (unpublished data), indicating that TLR3 expression has independent effects on the tumor parenchyma and infiltrating NK cells. Moreover, during disease progression (from stages I and II to stages III and IV), TLR3 expression is downregulated in both compartments, suggesting that expression of TLR3 in both cell populations impacts on patient survival. These findings are consistent with our in vitro data; TLR3 activation induces apoptosis and chemokine expression in tumor and also stimulates the functional activation of NK cells. Patients with high TLR3 expression are found in both early and late stage groups (P=.56) (unpublished data), suggesting that earlier diagnosis of HCC in these patients is unlikely to explain their prolonged survival. In vitro activation of TLR3 was previously reported to trigger apoptosis in several tumor cell lines, including breast cancer (10), melanoma (11), and even HCC cell lines transfected with TLR3 (18,30). TLR3+ NK cells are directly activated by treatment with poly(I:C) (6,19,31). Previous studies from prostate and glioma tumor models showed that TLR3 activation enhanced chemokine expression and immune effector cell infiltration. Both studies suggest that TLR3-induced immune infiltration promotes antitumor activity. Our study confirms and expands on these previous reports by providing in vitro and in vivo evidence on the multiple mechanisms of TLR3-mediated HCC progression. The effects of TLR3 activation observed in our study are consistent with a previous study in breast cancer showing that only patients with TLR3+ tumors benefited from poly(A:U) treatment (28). We showed that TLR3 activation enhances intratumor expression of chemokines that attract NK cells and stimulates IFN- production by NK cells. IFN- is known to synergize with inflammatory cytokines, such as TNF-, to further induce expression of NK cellattracting chemokines (20,32). This synergy could lead to further tumor infiltration by NK cells through a positive paracrine loop. In addition, HCC cells killed by NK cells may release TLR ligands that further activate NK cells to trigger more parenchyma cell death, thus amplifying antitumor activity. These putative feedback loops will require further investigation to establish whether they can account for the strong association between TLR3 expression and prolonged HCC patient survival. In HCC patients, we observed a correlation between intratumor expression of TLR3 and the density of tumor-infiltrating CD8+ T cells, whereas treating tumor-bearing mice with poly(I:C) failed to increase T-cell infiltration in liver tumors. This discrepancy could be attributed to the analysis of murine tumors only 3days after treatment, which could be an insufficient time period for T cells to expand and accumulate. In patients, however, sustained signaling through TLR3 may lead to increased recruitment and proliferation of T cells. In the transplanted HCC tumor model, poly(I:C) promoted tumor infiltration by both NK and T cells at day16. Our study has several limitations. First, the effect of poly(I:C) on tumor growth was only analyzed in a transplanted, nonorthotopic model of HCC. In the spontaneous orthotopic model, in which Figure6. The correlation between toll-like receptor (TLR3) expression and tumor infiltration by natural killer (NK) cells and antitumor activity in hepatocellular carcinoma patients. A) The density of TLR3+ cells correlates with the density of CD56+ NK and CD8+ T cells in patient samples as determined by immunohistochemistry (n = 31). B) The density of TLR3+ cells correlates with RNA expression of interferon- (IFNG, n = 35) and with the density of granzyme-B (GZB)-positive cells by immunohistochemistry (n=40) in hepatocellular carcinoma patient samples. C) The density of TLR3+ tumor parenchyma and tumor-infiltrating NK cells correlates positively with the density of Ki-67+ tumor parenchyma cells, and negatively with the density of activated caspase 3+ tumor parenchyma cells (n = 30). The Pearson correlation test was used to calculate the correlation coefficient (r) and two-sided P-values. tumors take months to develop, we only treated the tumors with poly(I:C) for a short time. In addition, the effect of poly(I:C) on human NK cells was assessed only with cells from healthy donors, and future studies are required to show that patient NK cells are functionally identical. Finally, not all HCC cell lines undergo apoptosis after TLR3 triggering, and the molecular basis of this variation is notknown. In summary, our study shows that TLR3 expression by tumor parenchyma and infiltrating NK cells was associated with superior HCC patient survival. TLR3 is likely to act directly on tumor parenchyma cells and promote the recruitment and activation of NK cells. Several synergistic pathways involving chemokines and IFN- or release of TLR ligands by dying cells may reinforce these effects. These findings emphasize the need for further clinical studies aimed at evaluating the potential of TLR3 ligands to treat HCC. r eferences This research was funded by Singapore Immunology Network (SIgN) of the Biomedical Research Council, A*STAR, Singapore. This study was partly funded by a Hong Kong Research Grants Council Collaborative Research Grant (HKU 7/CRG/09 to IOLN). Mathias Heikenwalder was funded by a European Research Council starting grant. The funders did not have a role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication. We wish to thank Dr Ksenija Slankamenac (University Hospital of Zurich, Zurich, Switzerland) for providing the patient samples and information; Dr Peter Schraml (University Hospital of Zurich, Zurich, Switzerland) for providing the patient formalin-fixed paraffinembedded sections from the Zurich cohort; and Joyce Lee (University of Hong Kong, Hong Kong) for preparing the patient RNA and formalin-fixed paraffinembedded samples from the Hong Kong cohort. We thank Dr Neil McCarthy (Insight Editing, London, UK) for manuscript editing. Our thanks are due to Dr Benjamin Toh, Dr Muly Tham, and Dr Lu-En Wai (Singapore Immunology Network, Singapore) for their help in harvesting mouse liver tumors and to Dr Michael Poidinger, Dr Bernett Lee, and Pavandip Singh Wasan (Singapore Immunology Network, Singapore) for help with the statistical analyses.


This is a preview of a remote PDF: https://jnci.oxfordjournals.org/content/104/23/1796.full.pdf

Valerie Chew, Charlene Tow, Caleb Huang, Emilie Bard-Chapeau, Neal G. Copeland, Nancy A. Jenkins, Achim Weber, Kiat Hon Lim, Han Chong Toh, Mathias Heikenwalder, Irene Oi-Lin Ng, Alessandra Nardin, Jean-Pierre Abastado. Toll-Like Receptor 3 Expressing Tumor Parenchyma and Infiltrating Natural Killer Cells in Hepatocellular Carcinoma Patients, JNCI Journal of the National Cancer Institute, 2012, 1796-1807, DOI: 10.1093/jnci/djs436