The tumor-promoting actions of TNF-α involve TNFR1 and IL-17 in ovarian cancer in mice and humans (original) (raw)
The aim of our first experiments was to investigate the impact of TNF-α signaling on infiltrating leukocytes in the ID8 model of ovarian cancer. This syngeneic mouse model resembles advanced stage IV human ovarian cancer and is known to produce TNF-α, as well as a range of other inflammatory cytokines, in the peritoneal tumor microenvironment (28). WT mice were lethally irradiated and reconstituted with bone marrow transplants from WT, TNFR1–/–, or TNFR2–/– mice. We also generated reverse chimeras in which lethally irradiated TNFR1–/– mice received WT bone marrow. Chimeras were injected i.p. with luciferase-expressing ID8 ovarian cancer cells (ID8-luc), and tumor burden was monitored weekly in situ by bioluminescence. After 8 weeks growth in vivo, the tumor burden was significantly lower in chimeric mice reconstituted with TNFR1–/– bone marrow compared with mice reconstituted with WT bone marrow or reverse chimeras (TNFR1–/– mice reconstituted with WT bone marrow) (P < 0.01) (Figure 1A). There was no difference in tumor burden compared with WT controls or reverse chimeras when WT mice were reconstituted with TNFR2–/– bone marrow (Figure 1A). Additionally, we used TNFR1/2–/– double-knockout mice to see whether the effects observed were caused by driving all responses to TNF-α through TNFR2. The results from the double-knockout mice mirrored the TNFR1–/– chimera results (Supplemental Figure 1A; supplemental material available online with this article; doi:10.1172/JCI39065DS1).
Disease stabilization in TNFR1 bone marrow chimeras. (A) TNFR1–/– bone marrow, WT bone marrow, and reverse (WT bone marrow in TNFR1–/– mice) chimeras were i.p. injected with 107 ID8 cells/mouse. Tumor burden was monitored weekly in situ by bioluminescence. TNFR1–/– bone marrow chimeras had significantly lower disease burden, starting 4 weeks after tumor cell injection (P < 0.01). Data are represented as mean ± SEM of n = 12. Representative data are shown from 2 independent experiments. (B) Total cell counts in the malignant ascites at week 8. Cytospins were prepared from ascitic fluid, and cells were differentiated with Wright’s staining. TNFR1–/– bone marrow chimeras had a significantly lower neutrophil infiltrate in the peritoneal cavity (P < 0.05, n = 12). Data are represented as mean ± SD of n = 12. Representative data are shown from 2 independent experiments. Data for WT (control mice, no chimeras) and reverse chimeras are shown in Supplemental Figure 1. (C) FACS analysis of the ascitic leukocyte infiltrate. TNFR1–/– bone marrow chimeras had significantly fewer infiltrating Gr-1+ F4/80– neutrophils (P < 0.01, n = 6). Data are represented as percentage of the CD45+ population, including mean of n = 6. Representative data are shown from 2 independent experiments (week 8). (D) WT mice were injected with 107 ID8-luc cells i.p.; Gr-1–neutralizing antibody or IgG control antibody were commenced twice weekly i.p. (100 μg/mouse). Data are represented as mean + SEM of n = 6. Representative data are shown from 2 independent experiments (week 8).
Analysis of the cellular composition of the ascites after 8 weeks of tumor growth revealed that TNFR1–/– bone marrow chimeric mice had a significantly lower Gr-1+ F4/80– leukocyte infiltrate compared with WT chimeric mice (P < 0.05, Figure 1, B and C; Supplemental Figure 1B shows all groups). This Gr-1+ F4/80– population was IL-4Rα– (data not shown). We therefore concluded that TNFR1 mediated the tumor-promoting action of TNF-α through Gr-1+ F4/80– cells in this cancer model. To confirm this, we depleted Gr-1+ cells in tumor-bearing mice, which led to disease stabilization (Figure 1D).
Functional CD4 TNFR1 signaling is crucial for tumor progression. We next used a floxed TNFR1 knockin system (TNFR1flcneo) (37) to express TNFR1 in different leukocyte populations in C57BL/6 mice. TNFR1flcneo mice carry a conditional gain-of-function allele achieved by introducing a loxP-flanked neomycin-resistance cassette into intron 5 of the murine TNFRp55 gene (37). While TNFR1flcneo mice reflect TNFR1–/– mice, the respective TNFR1Δ mice gained TNFR1 function in different leukocyte populations.
Eight-week-old female WT, TNFR1ΔLysM, TNFR1ΔCD4, TNFR1ΔCD19, and TNFR1flcneo mice were injected i.p. with ID8-luc cells, and peritoneal tumor growth was monitored weekly by bioluminescence. Tumor burden was significantly lower in TNFR1ΔLysM mice and TNFR1flcneo mice compared with WT mice (Figure 2A). Similarly, tumor growth was diminished in TNFR1ΔCD19 mice (Figure 2B). However, gain of function for TNFR1 in CD4 cells restored tumor burden in TNFR1ΔCD4 mice to WT levels when compared with TNFR1flcneo (Figure 2C) or the TNFR1–/– chimeric mice shown in Figure 1. The successful TNFR1 knockin in TNFR1ΔCD4 mice is shown as a Western blot (Figure 2D).
Tissue-specific reconstitution of TNFR1. Data are represented as mean ± SEM of n = 6. TNFR1flcneo mice had significantly lowered disease burden (P < 0.01) compared with WT mice. (A) TNFR1 gain of function in the monocyte/macrophage lineage has (TNFR1DLysM) no effect on tumor growth. (B) TNFR1 gain of function in CD19+ cells (TNFR1ΔCD19) has no effect on tumor growth. (C) Gain of function of TNFR1 in the CD4 lineage of TNFR1ΔCD4 mice “rescued” the protective effect of TNFR1 depletion. (D) Western blot for TNFR1 on protein lysates from CD4+ cells shows knockin of TNFR1 in CD4+ cells of TNFR1ΔCD4+ mice but not CD19+ or CD11b+ cells. (E) ID8 tumor–bearing TNFR1flcneo mice have significantly fewer neutrophils in the malignant ascitic fluid (P < 0.05; compared with WT mice). This effect is reversed in TNFR1ΔCD4 mice (P < 0.05; compared with TNFR1flcneo). Data are represented as mean ± SD of n = 6. (F) FACS analysis of the ascitic CD4+ population. TNFR1flcneo tumor-bearing mice have significantly fewer ascitic Th17 cells (P < 0.01, n = 6) compared with WT and TNFR1ΔCD4 mice. (G) IL-17 mRNA expression in CD4+ cells isolated from ascites of tumor–bearing mice at 8 weeks. (H) Ascitic CD4+ cells were selected and pooled (n = 3) and ex vivo stimulated with PMA and ionomycin for 4 hours. CD4+ cells from TNFR1–/– chimeras or TNFR1flcneo mice secreted significantly lower amounts of IL-17 (P < 0.01). Data are represented as mean ± SD of n = 6. Representative data are shown from 2 independent experiments.
To further understand the impact of TNFR1 signaling in CD4 cells, we investigated cellular changes in the ascitic microenvironment. We found no impact on the recruitment of macrophages into the ascitic fluid (Figure 2E). However, TNFR1flcneo tumor-bearing mice had a significantly lower neutrophil infiltrate (Figure 2E) compared with WT or TNFR1ΔCD4 mice.
These experiments demonstrate that functional TNFR1 signaling on CD4+ cells is crucial for the tumor-promoting action of TNFR1 during peritoneal growth and spread in a model of advanced ovarian cancer and that Gr-1+ cells may be involved in the action of these cells.
TNF-α/TNFR1 signaling in CD4 cells is important for the Th17 cell subpopulation. Since TNF-α/TNFR1 signaling regulates many other cytokines and chemokines, we measured TNF-α, IL-4, IL-6, IL-13, IL-17, IL-23, TGF-β, CCL5, and KC in the malignant ascitic fluid of ID8 tumor–bearing mice. There was a significant increase of IL-17 at end point in the malignant ascites of WT (mean ± SD: 143 ± 64 pg/ml) or TNFR1ΔCD4 (mean ± SD: 176 ± 75 pg/ml) mice compared with TNFR1–/– (mean ± SD: 95 ± 36 pg/ml) or TNFR1flcneo mice (mean ± SD: 103 ± 38 pg/ml) (P < 0.05). We therefore focused on the CD4+ Th17 T cell subset producing IL-17, but not IFN-γ, because IL-17 has been identified as an important player in inflammatory responses and cancer (38).
We analyzed CD4+ cells from the ascites of tumor-bearing WT, TNFR1flcneo, and TNFR1ΔCD4 mice for IL-17 and IFN-γ expression by FACS (Figure 2F). CD4+ cells with intact TNFR1 signaling from ID8 tumor–bearing WT or TNFR1ΔCD4 mice had significantly more CD4+IL-17+IFN-γ– cells than CD4+ cells from TNFR1flcneo mice in which TNFR1 signaling was deficient (Figure 2F). There were no differences in Th17 cells in spleens of tumor-bearing mice or control non–tumor-bearing mice (data not shown); additionally, we did not see differences in CD4+CD25+ cells in these mice (Supplemental Figure 2).
Next, we isolated ascitic CD4+ cells from tumor-bearing mice for RNA analysis and for ex vivo stimulation with phorbol myristate acetate (PMA). TNFR1-deleted CD4+ cells from ID8 tumor–bearing mice expressed significantly lower amounts of IL-17 mRNA compared with WT mice and TNFR1ΔCD4 mice (Figure 2G). We also stimulated ascitic CD4+ cells ex vivo with PMA to measure IL-17 protein secretion. TNFR1– CD4+ cells from tumor-bearing mice secrete significantly lower amounts of IL-17 ex vivo than those from WT or TNFR1ΔCD4 tumor-bearing mice (Figure 2H). We conclude that CD4 TNFR1 signaling leads to an increase in IL-17 in tumor-bearing mice.
Plasma cytokine profile during tumor progression in mice. We then measured plasma levels of key cytokines involved in Th1, Th2, and Th17 subpopulation development and maintenance (IL-1, IL-4, IL-6, IL-17, IL-23, TGF-β) during 4 to 8 weeks of ID8 tumor growth, the time when growth was maximal in control mice and mice with TNFR1 on their CD4+ T cells. Four weeks after tumor cell injection, there were no differences in plasma levels of IL-6 and IL-17 between WT, TNFR1flcneo, and TNFR1ΔCD4 mice (Figure 3, A and B). Plasma levels of IL-1, IL-4, IL-23, TGF-β, and CCL5 did not differ among WT, TNFR1flcneo, and TNFR1ΔCD4 tumor-bearing mice (data not shown); however, at weeks 6 and 8 there was a significant decrease in IL-6 (Figure 3A) and IL-17 (Figure 3B) levels in the TNFR1flcneo mice compared with TNFR1ΔCD4 mice.
Analysis of cytokine expression during tumor progression in tumor-bearing mice at 4, 6, and 8 weeks. Mice were sacrificed and bled for plasma collection. (A) IL-6 analysis. (B) IL-17 analysis. Analysis showed no baseline difference (in non–tumor-bearing mice) for any of the analyzed cytokines (white bars in the 4-week graph). IL-6 and IL-17 levels are significantly lower in ID8 tumor–bearing mice 6 and 8 weeks after ID8 cell injection (P < 0.01, n = 6). Data are represented as mean + SD of n = 6. Representative data are shown from 2 independent experiments.
IL-17 mediates Gr-1+ F4/80– cell recruitment into the ascitic tumor microenvironment. Of particular interest was the significant decrease in Gr-1+ myeloid cells in the ascites from TNFR1flcneo mice (P < 0.05; Figure 2D). It had been demonstrated that IL-17 is capable of selectively recruiting neutrophils into the peritoneal cavity (39). To determine whether IL-17 is crucial for Gr-1+ F4/80– recruitment in our model of stage IV ovarian cancer, we treated the mice twice weekly with recombinant murine IL-17A (0.5 μg/mouse i.p.) (Figure 4A). Consistent with the previous results, TNFR1flcneo mice showed significantly decreased tumor burden compared with WT or TNFR1ΔCD4 mice (Figure 4A). Treatment with recombinant IL-17A rescued this effect and increased tumor growth in TNFR1flcneo mice (Figure 4A). Analysis of the ascitic microenvironment demonstrated that recombinant IL-17A injections into TNFR1flcneo mice increased the influx of Gr-1+ F4/80– cells (P < 0.01; Figure 4B).
Increased IL-17–dependent neutrophil recruitment. (A) Recombinant IL-17 rescued the antitumor effect in TNFR1flcneo mice (P < 0.01). Data are represented as mean ± SEM of n = 6. Representative data are shown from 2 independent experiments. The right-hand graph demonstrates the curves for only TNFR1cneo mice with or without recombinant IL-17. (B) Total cell counts of the leukocyte infiltrate in the malignant ascites at week 8 (after i.p. injection). Control (left): significantly lower neutrophils in TNFR1flcneo ID8 tumor–bearing mice compared with WT mice (P < 0.01). Treatment with recombinant murine IL-17A (0.5 μg/mouse/week; right): recombinant IL-17A rescued neutrophil infiltration in TNFR1flcneo ID8 tumor–bearing mice. Data are represented as mean + SD of n = 6. Representative data are shown from 2 independent experiments.
Anti–TNF-α treatment in the ID8 ovarian cancer model. We next investigated whether treatment of ID8 tumor–bearing mice with a neutralizing anti-mouse TNF-α antibody (50 μg/mouse/twice weekly) influenced IL-17 levels, Gr-1+ F4/80– cell recruitment into the peritoneal tumor microenvironment, and tumor growth. Anti–TNF-α–treated mice had a significantly lower disease burden compared with mice treated with IgG control antibody or PBS-treated mice (P < 0.01; Figure 5A). After 8 weeks of tumor growth, plasma and ascitic IL-17 levels were significantly lower in the anti–TNF-α–treated mice compared with PBS- or IgG control antibody–treated animals (P < 0.01; Figure 5B). Additionally we found that plasma IL-6 levels (week 8, end point) were significantly lower in the TNFR1flcneo mice (Figure 3A) and were also significantly (P < 0.05) reduced in the anti–TNF-α–treated group (mean ± SD: 95.5 ± 48.6 pg/ml) compared with the PBS- (142.11 ± 64.2 pg/ml) or IgG-treated group (129.2 ± 51.6 pg/ml). However, there was no difference in the cytokine levels of IL-1, IL-4, IL-23, TGF-β, or CCL5 (data not shown). The Gr-1+ F4/80– cell infiltrate in the ascitic microenvironment was also significantly lower in the anti–TNF-α–treated animals (P < 0.05; Figure 5, C and E) compared with control groups as was the number of ascitic CD4+IL-17+IFN-γ– Th17 cells (Figure 5D), although there was no difference in the splenic CD4+IL-17+IFN-γ– subset. Therefore, anti–TNF-α antibody treatment had effects similar to those of targeting TNFR1 in a cell-specific manner.
In vivo treatment with murine anti–TNF-α in a preclinical model of stage 4 ovarian cancer. (A) Neutralizing TNF-α antibody (50 μg/mouse) administered twice weekly led to disease stabilization in ID8 tumor–bearing mice (P < 0.01). Data are represented as mean ± SEM of n = 10. (B) IL-17 plasma and ascites levels (8-week time point). Anti–TNF-α–treated mice had significantly lower IL-17 levels (P < 0.01). Data are represented as mean ± SD of n = 10. (C) Ascitic leukocyte infiltration. Within the ascitic fluid of anti–TNF-α–treated mice were significantly fewer neutrophils (P < 0.05; anti–TNF-α vs. IgG control, anti–TNF-α vs. PBS). Data are represented as mean + SD of n = 10. (D) FACS analysis of the CD4+ subpopulation within ID8 tumor–bearing mice. Anti–TNF-α treatment led to a significant reduction in the CD4+IL-17+IFNγ– population compared with control IgG- or PBS-treated mice. Data are represented as mean + SD of n = 10. (E) FACS analysis of the Gr-1+ F4/80– population in the ascitic fluid. Significant reduction of neutrophils in the anti–TNF-α–treated group. Data are represented as percentage of the CD45+ population including mean ± SD of n = 10. (F) Neutralizing IL-17A antibody (100 μg/mouse) administered twice weekly led to neutrophil depletion (insert) and disease stabilization in ID8 tumor–bearing mice (P < 0.01). Data are represented as mean ± SEM of n = 10. Representative data are shown from 2 independent experiments.
To demonstrate the relevance of IL-17 for tumor promotion, we used a neutralizing antibody for IL-17. Mice treated with neutralizing anti–IL-17 but nor IgG control or PBS showed a significant reduction in disease progression (P < 0.01; Figure 5F). Sentinel mice were killed after 1 week of anti–IL-17 treatment, prior to tumor cell injection, to demonstrate efficient GR1+F4/80– cell depletion (Figure 5F).
Our results so far suggest that CD4+ TNFR1 signaling contributes to disease progression via IL-17 and recruitment of Gr-1+ F4/80– cells to the tumor microenvironment. An anti–TNF-α antibody inhibits the same pathway in this mouse model of ovarian cancer. To investigate the correlation between TNF-α and IL-17 further and its relevance to human cancer, we studied plasma samples from cancer patients who had been treated with an anti-human TNF-α antibody, infliximab.
Infliximab reduces IL-17 plasma levels in patients with advanced malignant disease. In a phase I clinical trial, we reported that infliximab anti–TNF-α antibody treatment of patients with advanced cancer significantly reduced plasma IL-6 levels (36). In view of the above animal data, we measured IL-1, IL-4, IL-17, IL-23, and TGF-β, and CCL5 by ELISA in the plasma samples from this trial. IL-1, IL-4, IL-23, TGF-β, and CCL5 levels were all above detection levels, but there were no significant differences between pretreatment and 24-hour posttreatment samples (data not shown). However, in all patients studied (n = 40), plasma IL-17 levels were significantly lower 24 hours after infliximab infusion (P < 0.05; Figure 6, A and B) compared with pretreatment samples.
Patients with advanced malignancies treated with anti–TNF-α in phase I clinical trial. (A) 40 patients with a variety of malignancies were treated in a phase I/II clinical trial with infliximab (36). Plasma analysis before and 24 hours after treatment with infliximab demonstrated that anti–TNF-α significantly reduced the plasma levels of IL-17 (P < 0.05). Data are represented as mean ± SD of n = 40. (B) Analysis of IL-17 levels in stable (SD) versus progressive disease (PD) patients (P < 0.05). In both groups, IL-17 levels dropped significantly 24 hours after anti–TNF-α treatment. Data are represented as mean ± SD of n = 40. (C) Cytokine release assay: whole-blood samples were collected from patients in the presence of PHA. IL-17 release was significantly lower 24 hours after anti–TNF-α treatment (P < 0.05). Similar results were obtained for D. IL-6 (P < 0.05).
In this trial, blood samples were also taken for the whole-blood cytokine release assay. This is a surrogate marker of biologic response to the antibody and involves short-term culture of whole blood with phytohemagglutinin (PHA) and analysis of cytokine production. Anti–TNF-α treatment significantly decreased production of IL-6 and IL-17 in cultures of the patients (Figure 6, C and D). There were no differences in IL-1, IL-4, IL-23, TGF-β, and CCL5 production (data not shown).
Infliximab treatment of patients with advanced ovarian cancer. We also examined samples from 15 additional patients with advanced ovarian cancer and ascites who had been treated with infliximab. Both blood and ascitic fluid were obtained prior to and during infliximab treatment, and paracentesis samples from 3 ovarian cancer patients who had no anti–TNF-α antibody treatment were used as controls for ascites samples. In all patients, infliximab treatment reduced the blood and ascitic TNF-α concentrations 1 hour and 24 hours after the end of the 2-hour infliximab infusion (Figure 7A). Paracentesis alone had no effect on ascitic cytokine levels (TNF-α in control patients, mean ± SD: before, 125 ± 51 pg/ml; 24 hours after, 142 ± 74 pg/ml).
Ovarian cancer patients with advanced disease treated with infliximab in a phase I/II clinical trial. Seventeen patients with advanced ovarian cancer received i.v. anti–TNF-α treatment in this clinical study. (A) TNF-α concentrations (P < 0.05) were significantly lower at 1 hour (serum) and 24 hours (ascitic fluid) after infliximab infusion. Data are presented as mean ± SD of n = 17. (B) IL-17 protein levels were significantly reduced in the ascitic fluid 24 hours after infliximab infusion (P < 0.05); however, there was no difference in IL-17 plasma levels. Data are represented as mean ± SD of n = 15. (C) TNF-α real-time analysis of ascitic cell total RNA. TNF-α gene expression was significantly reduced 24 hours after anti–TNF-α treatment (P < 0.01). Data are represented as mean ± SD of n = 15. (D) IL-17 real-time analysis of ascitic cell total RNA. IL-17 gene expression was significantly reduced 24 hours after anti–TNF-α treatment (P < 0.01). Data are represented as mean ± SD of n = 15. (E and F) Cytokine release assay: whole-blood samples were collected from patients in the presence of PHA. Cytokine release was measured by ELISA. IL-17 release was significantly lower 24 hours after anti–TNF-α treatment (E; P < 0.05); similar results were obtained for IL-6 (F; P < 0.05). (G) CD4+CD25– cells were purified by FACS sorting following MACS bead isolation from ascites samples before treatment and 24 hours after infliximab infusion. Control naive CD4+CD25– T cells were selected from peripheral blood of healthy blood donors. Data are represented as mean + SD of n = 10; *P < 0.01.
There was no difference in plasma IL-17 levels before and after treatment in the infliximab-treated patients (Figure 7B); however, ascitic IL-17 levels were significantly lower 24 hours after treatment with infliximab (P < 0.05; Figure 7B). Again, paracentesis alone did not affect IL-17 levels (IL-17 in control patients, mean ± SD: before, 143 ± 39 pg/ml; 24 hours after, 162 ± 42 pg/ml).
Ascites cell samples were available from 15 of the treated patients and 3 control untreated patients. We extracted RNA from these and observed downregulation of TNF-α mRNA expression by real-time PCR when pretreatment samples were compared with posttreatment samples (P < 0.01; Figure 7C). IL-17 levels were significantly lower in total RNA from ascitic cells from treated but not control patients (P < 0.01; Figure 7B) at the 24-hour time point. Since there was a presence of Th17 cells and expression of IL-17, we analyzed other Th17 cytokines such as IL-17F, IL-21, and IL-22 in these plasma and ascites samples (Supplemental Figure 3). There was no difference in plasma or ascitic IL-17F levels before and after treatment in the infliximab-treated patients (Supplemental Figure 3A); ascitic IL-21 levels were significantly lower 24 hours after treatment with infliximab (Supplemental Figure 3B); IL-22 levels were significantly reduced in ascites samples (Supplemental Figure 3C), showing that other Th17 cytokines were also regulated upon infliximab treatment.
As with the phase I trial, there were no differences in IL-1, IL-4, IL-23, TGF-β, or CCL5 levels after anti–TNF-α treatment in plasma or the whole-blood cytokine release assay (data not shown), but infliximab treatment significantly decreased production of IL-6 and IL-17 in the whole-blood cytokine release assay (Figure 7, E and F).
Next, we isolated RNA from sorted ascitic CD4+CD25– cells and analyzed the expression of characteristic Th17 genes before and 24 hours after infliximab. Analysis revealed a significant reduction of IL17, IL22, CCR4, and CCR6; additionally receptor tyrosine kinase–like orphan receptor 2 (ROR2) was downregulated (Figure 7G). Interestingly, IL-1Rα and IL-23R were also significantly downregulated in these CD4+CD25– cells after infliximab treatment (Figure 7G). There was no difference in CCL20, IL-6, or IFN-γ expression 24 hours after infliximab infusion. Infliximab treatment decreased IL-6 production by peripheral blood cells; however, anti–TNF-α had few effects on IL-6 mRNA expression by CD4+CD25– lymphocytes.
We postulated that TNF-α sensitizes T cells to Th17-promoting cytokines such as IL-1 and IL-23 by upregulation of the respective receptor. Therefore, we isolated CD4+CD25– T cells from the spleen of WT, TNFR1–/–, and TNFR2–/– tumor-bearing mice and stimulated them with recombinant TNF-α to measure IL-1Rα and IL-23R expression by real-time PCR. TNF-α did not upregulate IL-1Rα and IL-23R in TNFR1–/– CD4+CD25– T cells but did have an effect on WT and TNFR2–/– cells (Supplemental Figure 4).
TNF-α induces a network of IL-17–regulated genes in ovarian cancer patients. These results from 2 clinical trials of an anti–TNF-α antibody had parallels with the data obtained from the mouse model and suggest a role for IL-17 in the protumor actions of TNF-α. We used a bioinformatics approach to further investigate the link between TNF-α and IL-17 in the tumor microenvironment. We analyzed the publicly available gene expression data set GSE9899 (Gene Expression Omnibus accession number) containing 285 human ovarian cancer biopsies profiled on the Affymetrix Human Genome U133plus2 platform. The inflammatory gene set of the canonical TNF-α signaling pathway was extracted from the GeneGo Metacore pathway analysis tool. Gene symbols were mapped to representative probes on the microarray chip, which were used to test enrichment using the function “genesettest” within the limma package.
We ranked the 285 samples from this data set from low to high levels of gene expression in the TNF-α signaling pathway. As these biopsies may have variable proportions of malignant and stromal cells that might obscure/dilute any associations, we conducted our further analysis on the 50 samples with the highest expression of genes in the TNF-α signaling pathway and the 50 samples with the lowest expression of these genes. We obtained a list of genes that were statistically different between these 2 groups of samples (Student’s t test) and performed Gene Set Enrichment Analysis (GSEA; http://www.broadinstitute.org/gsea/) using Metacore pathway and process gene set analysis.
There was a significant association between high TNF-α signaling pathway gene expression and expression of genes in the “cytokine production in Th17 cells” pathway (P = 1.004–5), with the “immune suppression” pathway (P = 1.351–8) and “neutrophil activation” pathway (1.655–28). Network objects (or genes) that make up the pathway “cytokine production in Th17 cells” were used to draw the heat map shown in Figure 8. This clearly shows that the genes ascribed to the Th17 pathway map closely with high levels of genes in the TNF-α signaling pathway in ovarian cancer biopsy samples, and this is particularly striking for CD14, ICAM1, IL-8, IL-23, genes of the NF-κB system, and TGF-β1.
Gene array analysis. 285 samples of ovarian cancer data sets were ranked from low to high levels of gene expression in the TNF signaling pathway. The 50 samples with the highest expression of genes in this individual pathway and 50 samples with the lowest expression of these genes were analyzed. Network objects (or genes) that make up the pathway “cytokine production in Th17 cells” are represented as a heat map.