Widespread parainflammation in human cancer
- Dvir Aran†1,
- Audrey Lasry†2,
- Adar Zinger2,
- Moshe Biton2,
- Eli Pikarsky2,
- Asaf Hellman3,
- Atul J. Butte1Email author and
- Yinon Ben-Neriah2Email author
© The Author(s). 2016
Received: 25 January 2016
Accepted: 2 June 2016
Published: 8 July 2016
Chronic inflammation has been recognized as one of the hallmarks of cancer. We recently showed that parainflammation, a unique variant of inflammation between homeostasis and chronic inflammation, strongly promotes mouse gut tumorigenesis upon p53 loss. Here we explore the prevalence of parainflammation in human cancer and determine its relationship to certain molecular and clinical parameters affecting treatment and prognosis.
We generated a transcriptome signature to identify parainflammation in many primary human tumors and carcinoma cell lines as distinct from their normal tissue counterparts and the tumor microenvironment and show that parainflammation-positive tumors are enriched for p53 mutations and associated with poor prognosis. Non-steroidal anti-inflammatory drug (NSAID) treatment suppresses parainflammation in both murine and human cancers, possibly explaining a protective effect of NSAIDs against cancer.
We conclude that parainflammation, a low-grade form of inflammation, is widely prevalent in human cancer, particularly in cancer types commonly harboring p53 mutations. Our data suggest that parainflammation may be a driver for p53 mutagenesis and a guide for cancer prevention by NSAID treatment.
KeywordsInflammation Parainflammation p53 mutations NSAID treatment Cancer prevention Genomics
Inflammation is one of the enabling hallmarks of cancer  and it has been estimated that approximately 20 % of cancers are caused by chronic inflammation [2, 3]. Tumor-promoting inflammation can contribute to various stages of tumor development, from tumor initiation to metastasis . We recently developed and characterized a mouse model of intestinal cancer based on tissue-specific ablation of CKIα [5, 6]. Inducible ablation of CKIα in the gut epithelium has several immediate consequences, Wnt activation due to stabilization of β-catenin, induction of DNA damage response with robust p53 activation, and elicitation of a low-grade inflammatory response in the epithelium. This inflammatory reaction, which is detected by mRNA and protein analysis of the gut epithelial cells, is an atypical one, having a relatively low secretory component and, conspicuously, is not accompanied by a typical inflammatory cell infiltrate in CKIα-deficient mucosa. We coined this low-grade inflammatory reaction “parainflammation” (PI) based on its relationship to a term defined by Medzhitov: a low-grade inflammatory response at an intermediate state between tissue homeostasis and classic inflammation which can be induced by persistent tissue stress, serving to restore tissue homeostasis . It has been proposed that PI could play a role in several conditions, such as aging and obesity . In contrast to “classic” inflammation, often ignited by an extrinsic assault such as bacterial infection, PI may erupt due to tissue-intrinsic assaults, such as DNA damage [6, 7], and cooperate with the tumor suppressor p53, contributing to tissue protective senescence and counteracting tumor progression. Upon p53 ablation, however, PI loses its beneficial role and, instead, contributes to carcinogenesis . Here we constructed a PI gene signature based on analysis of the CKIα and CKIα-p53-deficient gut epithelium and, using existing databases, examined this signature and its implications in thousands of human tumors and cell lines. We noticed a striking occurrence of PI in several cancer types distinct from their normal tissue counterparts. Our study indicates that PI is an important factor in tumor development, with a significant influence on prognosis. Notably, we found that PI can be markedly attenuated in human samples by non-steroidal anti-inflammatory drug (NSAID) treatment, possibly alluding to a mechanism of cancer prevention.
Deriving the PI gene signature from CKIα-deficient mice
PI is suppressed by NSAID treatment in mouse tumors
It is not yet known why NSAID treatment has an impact on cancer prevalence, progression, and mortality in human solid tumors [15–18]. We have previously shown that gut epithelium PI can be suppressed by NSAID treatment in the CKIα-deficient mouse model . We therefore hypothesized that cancer cells may harbor tumor-promoting PI subject to suppression by NSAID treatment. To test this hypothesis, we generated a tumor model for examining NSAID treatment effects. We prepared organoids from normal gut and adenomatous polyps of APCmin/+ mice, treated them with the NSAID sulindac, and performed RNA-seq on three biological replicates (Additional file 6). Strikingly, of the 17 upregulated PI genes observed in the tumor organoids, 11 (64.7 %) were significantly downregulated in the sulindac-treated samples, compared with 33.2 % of all genes and 38.3 % of all inflammatory response genes (p value = 0.0053) (Fig. 2a). Immunofluorescence staining shows that the protein expression of PI gene Ifitm2/3 (Fig. 2b) is markedly elevated in the tumor organoids and is mostly suppressed by sulindac treatment. These data validate the PI signature in a new mouse tumor model, underscoring both the epithelial origin and the tumor specificity of this new signature, and show that NSAID treatment can reduce PI.
Analogously to the procedure described above, we also found 75 inflammatory context genes which are downregulated in the PI mouse models (Additional file 3: Table S2). This list of genes is again highly enriched with LPS and IFN γ response genes but also contains interleukin (IL)2/IL4 response genes. However, we didn’t observe expression changes of those 75 genes in the organoids model, suggesting that the downregulated genes are not part of the general PI mechanism. Therefore, while it is well established that activation of innate inflammation pathways leads to both up- and downregulation of genes involved in activation and inhibition of this highly regulated network [19–21], we chose to derive the PI signature only from the upregulated genes in the mouse and organoid PI models.
Human cancers display the PI signature discovered in mice
We next computed a score for each cell line by performing single-sample gene set enrichment analysis (ssGSEA)  for the PI gene signature. We defined this score, which is an enrichment measure of the overexpression of the PI genes, as the PI score (Additional file 7). The PI score revealed major differences in PI both between cancer types and within cancer types, varying from high levels of PI in head and neck and pancreatic cancer to low levels in samples originating from prostate and liver (Fig. 3c). Thus, the PI expression and PI score patterns suggest that PI occurs across most cancer types, although not uniformly. It is important to note that while the PI score is based on a signature of 40 genes, the PI phenomenon is not restricted to these genes alone but has a wide effect on numerous genes over many different molecular functions (Additional file 8). Finally, we observed a high correlation between the PI score and scores derived from the downregulated PI genes (Spearman R = 0.654, p value <1e-20; Additional file 3: Figure S3). This once again confirms the validity of our gene sets in human and the relevance of the PI mouse models to a phenomenon that is also observed in human.
Next, we explored the representation of the PI signature in primary human cancers. As differences in PI expression between individual samples may be explained solely by differences in purity levels of the samples , we designed a simple adjustment procedure for removing inflammatory gene expression originating from immune infiltrations (Additional file 3: Figure S4). This adjustment procedure consists of two steps: first, utilizing expression data of normal tissues from the Genotype-Tissue Expression (GTEx) project , we learned the normal association of gene expression with immune infiltrations in each tissue type. We then used the expression level of PTPRC (CD45), a pan-hematopoietic exclusive marker expressed in all leukocyte cell types but not in other tissues, as an estimate for immune infiltrations. Finally, we applied the gene-specific, tissue-specific learned slopes to adjust the expression levels of The Cancer Genome Atlas (TCGA) tumor samples. This procedure diminishes expression differences among samples which are most likely explained by differences in purity.
Notably, certain PI genes are members of the Toll-like receptor (TLR) activation pathway (TLR2, CD14, and TIRAP) and, when upregulated, could have mediated an innate immune response to tissue-associated microbiota, igniting conventional inflammation with inflammatory infiltrate, secondary to PI. We therefore hypothesized that PI may enhance the recruitment of certain immune cell subsets to the tumors. To this end, we utilized hematoxylin and eosin (H&E) estimations provided by TCGA and gene signature enrichments of immune subset types from Rooney et al.  (Additional file 3: Table S5) and associated them with PI scores in tumor samples across cancer types. H&E estimations of major immune subtypes (lymphocytes, monocytes, and neutrophils) did not, however, show significant associations with PI scores across different cancer types (Additional file 3: Table S7). We further correlated PI scores of individual tumors with specific immune subsets based on gene signature enrichments in both TCGA and CCLE samples (Fig. 5b; Additional file 3: Table S8). Among the immune subsets the PI score demonstrated highest correlations across tumor types with the macrophage signature (average Spearman coefficient = 0.362). However, we observed the same trend of correlation between the PI score and macrophages in cell lines (average R = 0.407). This observation rules out the possibility that PI is dependent on macrophages infiltrating the tumor. Moreover, whereas the role of macrophages in orchestrating PI has been previously suggested , our finding supports this inference yet suggests that the tumors themselves may fulfill the macrophage inflammatory function in PI by expressing macrophage-relevant genes. Importantly, we did not observe any correlation with CD8+ enrichment, which is the main component of the “immunoscore” , thus suggesting that PI may represent a different immunotype, which, similarly to the immunoscore, may serve as a clinical parameter in evaluating tumorigenesis.
Thus, the PI signature is widely expressed in human tumors, distinguishing the tumor cells from adjacent normal tissue and the tumor microenvironment, with certain cancer types having stronger PI signatures than others. PI appears to be primarily a cancer cell-autonomous phenomenon, distinct from all other well-established cancer-promoting immune inflammatory responses.
PI is associated with poor prognosis and p53 mutations
PI response to NSAID treatment in human tumors
To confirm these computational findings, we treated two human cancer cell lines, predicted to have high PI scores from the CCLE dataset analysis, with sulindac: the BxPC3 cell line, a pancreatic cancer cell line with a PI score which ranks it ninth out of the 634 CCLE carcinomas, and the T84 cell line, a colorectal cancer cell line ranked 18th. Sulindac treatment led to a decrease of more than 50 % in the expression of all tested PI genes which are highly expressed in BxPC3 and T84 cells, respectively (Fig. 8b, c). This, together with the analyses of the SCC-25 cell line and the mouse tumor organoids (Fig. 2), confirms the repressive effect of NSAID treatment on PI, suggesting that part of the cancer-prevention mechanism of NSAIDs may be attributed to PI suppression.
Inflammation is emerging as one of the hallmarks of cancer  yet histologically detectable chronic inflammation characterizes early tumor development in only a minority of solid tumors . In this respect it is surprising that NSAID treatment is remarkably effective in reducing mortality rates associated with major human solid tumors, albeit their mechanism of action in cancer is controversial and there is no common property distinguishing cancers in which NSAID treatment is beneficial. NSAIDs are non-selective Cox2 inhibitors and it is thus not surprising that they are effective in cancers where Cox2 is indeed elevated. However, NSAIDs are also effective in cancers where Cox2 is not activated, occurring through hereditary or stochastic mutations and not preceded by prolonged inflammation. This indicates that inflammation may have a covert course that plays a major unappreciated role in cancer. Accordingly, in spite of many studies addressing this question, the mechanism of NSAID action in cancer prevention remains mostly elusive. Medzhitov coined the term parainflammation for a low-grade inflammatory response, referring to an adaptive response due to tissue stress or malfunction . Here we defined PI on the basis of analysis of existing databases and experimental data: an epithelial-autonomous inflammatory response observed in genetically modified mouse models. We then investigated the occurrence of PI in human cancer, asking (a) whether PI is a universal cancer phenomenon and (b) does PI have clinical/prognostic implications in human cancer. Using several large sets of human cancer samples, we detected frequent occurrence of PI in human cancers with several interesting features: PI is cancer cell-autonomous, readily detectable in human carcinoma cell lines; its repertoire is distinct from other common inflammatory signatures; and it doesn’t bear the phenotypic hallmarks nor some of the regulatory characteristics of common inflammatory reactions (e.g., NF-κB). PI is very common in certain types of cancer, including bladder, head and neck, cervical, and colorectal cancer, and conspicuously absent in other types, such as hepatocellular carcinoma, prostate and endometrial adenocarcinoma, and low grade glioma.
In the CKIα-deficient mouse model, PI promotes tumorigenesis only after loss of p53. We therefore investigated whether p53 loss/mutation in human tumors is associated with occurrence of PI. Indeed, the p53 status of the tumor displays a high association with PI in a variety of cancer types. A similar association of p53 mutation or loss with PI occurs across the entire spectrum of cancer types: those with low PI, like prostate, liver, thyroid, and melanoma, have few p53 mutations. Why certain cancer types have low p53 mutation frequencies with no evidence of p53 pathway suppression is an enigma. The strong association between PI and p53 mutations in cancer suggests that PI may be one of the major driving forces for inactivating the p53 pathway. This corresponds to the relationship between PI and p53 deficiency in mouse tumor models, showing a sharp PI switch from a tumor suppressor to a tumor promoter mechanism upon p53 loss . Such a tumorigenesis switch mechanism may be a particularly powerful cancer driver mechanism and is thus possibly one of the main mechanisms enforcing p53 mutations in cancer, which cannot be substituted by loss of other tumor suppressor genes. While we are not aware yet of other means of switching PI from a cancer suppressor to a promoter, it is possible that certain other tumor-specific genetic aberrations may fulfill a function similar to p53 inactivation, pushing cancer progression. Indeed, PI is associated with worse prognosis, beyond the p53 status of the tumor. These retrospective analyses could indicate the value of screening individual tumors for PI manifestation as a prognostic or treatment/prevention tool. Particularly relevant in this aspect is NSAID treatment. Retrospective analysis of cancer patients who were routinely treated with aspirin, including low dose aspirin, for non-cancer indications showed a surprisingly beneficial effect in reducing cancer mortality following surgical removal of the tumor . Nearly all the cancer types in which beneficial aspirin effects were noted (e.g., colorectal, pancreatic, lung, and breast) [17, 18, 33] are characterized by high PI, either throughout the entire cancer type or a significant subtype, calling for implementation of PI screening of tumor biopsy or resection samples. Supporting our hypothesis implicating PI suppression as an important anti-tumorigenic mechanism of NSAID is our study of a mouse tumor model and human cancer cell lines displaying high PI: we demonstrated effective sulindac repression of the majority of the tested PI genes upregulated in the tumor cells. Thus, whereas we are still far from understanding the origin and mechanisms of PI emergence in tumors, its recognition and monitoring may be of great value in the clinical care of cancer.
In this work, we have characterized a novel PI signature present in 25.9 % of all human tumors and in the vast majority of certain types of cancer, such as gastrointestinal, lung, bladder, and head and neck tumors. PI has a distinct signature, originating endogenously from the tumor cells, and does not coincide with canonical inflammation. Our data indicate that PI is linked to p53 mutations, suggesting it might be a major driving force for p53 mutation. As PI is suppressible by NSAID treatment in vitro and is particularly prominent in cancers in which aspirin treatment is beneficial, we propose that a tumor PI signature should be tested as a potential biomarker/clinical indication for NSAID chemoprevention and treatment of cancer.
PI gene signature
We obtained raw RNA-seq counts from Pribluda et al.  for WT, CKIα-deficient and CKIα-p53 doubly deficient gut epithelium, each with two replicates (Additional file 2). Using the DESeq2 package for R , we calculated p values and fold ratio changes against the WT counts. Upregulated genes were chosen as genes with fold ratio >2 and adjusted p value <0.01. Genes upregulated in the mouse models were then intersected with our list of inflammatory response genes. The downregulated gene signature was defined in the same manner but was more than twofold lower in the deficient mouse models compared with WT.
Upstream regulator analysis
The analysis was performed through the use of QIAGEN’s Ingenuity® Pathway Analysis (IPA®, QIAGEN Redwood City, http://www.qiagen.com/ingenuity). Figure 1c shows the top 18 non-chemical regulators; Additional file 4 presents all data.
Adjustment for blood admixture
Infiltration of immune cells into tumor samples may influence the analysis of gene expression profiles of the tumor cells and these should therefore be adjusted. The level of blood cells in tumor samples may be efficiently analyzed using leukocyte-specific expression profiles, e.g., by the ESTIMATE score . We found tight correlation between the ESTIMATE measurement and CD45 expression levels over tumor samples across all cancer types (Additional file 3: Figure S4a). The ESTIMATE score, which is calculated using a gene signature of 141 genes, might be perturbed by our notion of PI; thus, we adjusted the expression levels of each gene according to CD45 expression alone. Using expression data of normal samples from the Genotype-Tissue Expression Project (GTEx) , we first fit a linear curve for each gene according to its expression level and the expression level of CD45 in each tissue type (Additional file 3: Figure S4b). The slope learned from this section represents the normal tissue-specific association between the gene expression levels and the immune infiltration levels. Using this slope we then adjust the expression levels of the gene for all TCGA samples in the corresponding cancer type (Additional file 3: Figure S4c). Finally, the expression of all samples is shifted so the minimal level for each gene is 0.
The PI score is the ssGSEA score of the 40 PI genes (Additional file 3: Table S1). A TCGA tumor sample was considered PI positive (PI+) if its PI score is over 0.2951. This threshold is found in less than 5 % of TCGA adjacent normal samples and corresponds to 25.9 % of the tumor samples overall. In the CCLE dataset only 9.5 % of the carcinoma cell lines were over this threshold. However, the ssGSEA method is highly sensitive to different expression platforms with different numbers of genes and is thus incomparable between these. Of the carcinoma cell lines, 25.9 % have a PI score over 0.1859. This notion is true for other datasets used in this article for calculating the PI scores: the scores should only be compared between samples from the same expression-measuring platform in the species. To allow some degree of comparison between the different datasets, for visualization the scores were shifted in each data set such that PI+ samples have positive scores.
Organoids were produced from WT and APCmin/+ mice as previously described [13, 36]. For adenoma and APCmin/+ organoids, adenomas and normal adjacent tissue were separated and processed accordingly. All organoids were grown in Advanced DMEM/F12 culture medium (Gibco) supplemented with Noggin, EGF, bFGF (Peprotech, 1:1000), R-spondin1 (Peprotech, 1:500) and B27 (Gibco, 1:50). For staining, organoids were fixed in 4 % paraformaldehyde and incubated overnight with the primary antibody (IFITM2/3 - 1:200, Abcam). Secondary antibody was Alexa fluor-647 donkey anti-rabbit (Molecular Probes, 1:1000). Hoechst (1 μg/ml, Molecular Probes) was used to stain nuclei.
NSAID treatment of human cancer cell lines
BxPC3 cells were grown in RPMI medium. T84 cells were grown in 1:1 DMEM/F12 medium. For sulindac treatment, cells were incubated with 400 μM sulindac (Sigma) for 48 h before harvesting. Controls were treated with DMSO, the NSAID vehicle.
RNA extraction and qPCR analysis
RNA was extracted using the miRNeasy kit (Qiagen). cDNA was produced using qScript cDNA kit (Quanta). Primers used for qPCR analysis can be found in Additional file 3: Table S11.
CCLE, Cancer Cell Line Encyclopedia; GEO, Gene Expression Omnibus; GTEx, Genotype-Tissue Expression; H&E, hematoxylin and eosin; IFN, interferon; IL, interleukin; KIRC, kidney renal clear cell carcinoma; LPS, lipopolysaccharide; NSAID, non-steroidal anti-inflammatory drug; PAAD, pancreatic adenocarcinoma; PI, parainflammation; qPCR, quantitative PCR; ssGSEA, single-sample gene set enrichment analysis; TCGA, The Cancer Genome Atlas; WT, wild type.
We thank Haya Hamza for technical assistance.
This work was supported by the Gruss Lipper Postdoctoral Fellowship to D.A., grants from the Israel Science Foundation (ISF)—Centers of Excellence and the European Research Council within the FP-7 to YB-N (294390 PICHO) and EP (281738 LIVERMICROENV), the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (AMRF), the I-CORE program of ISF (grant number 41/11) and the Israel Cancer Research fund to Y.B.N., and Israel Science Foundation grant number 1162/12, I-CORE Program of the Planning and Budgeting Committee and the Israel Science Foundation grant number 41/11, and the Rosetrees Trust grant to A.H and the National Cancer Institute of the National Institutes of Health U24 CA195858 to A.J.B.
Availability of data and materials
The results published here are in part based on data generated by TCGA Research Network (http://cancergenome.nih.gov). TCGA datasets were downloaded on 15 January 2015. The organoid RNA-seq data are available at the Gene Expression Omnibus (GEO) under accession GSE81836.
Tumor and adjacent normal samples: Level 3 RNA-seq data (RNAseqV2 normalized RSEM), clinical data, and Mutation Annotation Format (MAF) files for multiple human cancers and adjacent normal samples were downloaded from TCGA data portal on 15 January 2015 (http://cancergenome.nih.gov; Additional file 3: Table S3). Only primary tumors were included in the analyses.
Normal samples: Processed RNA-seq data were downloaded from the Genotype-Tissue Expression (GTEx) portal (http://www.gtexportal.org/home/).
Cancer cell lines: Normalized Affymetrix U133 + 2 array data were downloaded from the Cancer Cell Line Encyclopedia (CCLE) website .
Aspirin treatment: Illumina HumanHT-12 V4.0 expression beadchip of three replicates obtained from oral cancer cell line SCC25 treated with 2.0 mM of aspirin and untreated samples used as a control were downloaded from the NCBI GEO (accession GSE58162).
Software: All statistical analyses were performed using MATLAB®. The Deseq2 package for R  was used for differential expression re-analysis of the RNA-seq of the CKIα-deficient mice. The RSEM (RNA-Seq by Expectation-Maximization) software package  was used for the analysis of the organoid RNA-seq experiments. The GSVA package  was used for the single-set gene set enrichment analysis (ssGSEA).
Conception and design: DA, AL, MB, EP, AH, YB-N. Development of methodology: DA, AL. Acquisition of data: DA, AL, AZ. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): DA. Writing, review, and/or revision of the manuscript: DA, AL, EP, AH, YB-N. Study supervision: YB-N, EP, AH, AJB. All authors read and approved the final manuscript.
The authors declare that they have no competing interests.
Ethics approval and consent to participate
All mouse experiments were conducted in accordance with guidelines of the Hebrew University animal ethics committee, accredited by AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International).
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144:646–74.View ArticlePubMedGoogle Scholar
- Aggarwal BB, Vijayalekshmi RV, Sung B. Targeting inflammatory pathways for prevention and therapy of cancer: short-term friend, long-term foe. Clin Cancer Res. 2009;15:425–30.View ArticlePubMedGoogle Scholar
- Balkwill FR, Mantovani A. Cancer-related inflammation: common themes and therapeutic opportunities. Semin Cancer Biol. 2012;22:33–40.View ArticlePubMedGoogle Scholar
- Grivennikov SI, Greten FR, Karin M. Immunity, inflammation, and cancer. Cell. 2010;140:883–99.View ArticlePubMedPubMed CentralGoogle Scholar
- Elyada E, Pribluda A, Goldstein RE, Morgenstern Y, Brachya G, Cojocaru G, Snir-Alkalay I, Burstain I, Haffner-Krausz R, Jung S, et al. CKIalpha ablation highlights a critical role for p53 in invasiveness control. Nature. 2011;470:409–13.Google Scholar
- Pribluda A, Elyada E, Wiener Z, Hamza H, Goldstein RE, Biton M, Burstain I, Morgenstern Y, Brachya G, Billauer H, et al. A senescence-inflammatory switch from cancer-inhibitory to cancer-promoting mechanism. Cancer Cell. 2013;24:242–56.Google Scholar
- Medzhitov R. Origin and physiological roles of inflammation. Nature. 2008;454:428–35.View ArticlePubMedGoogle Scholar
- Chovatiya R, Medzhitov R. Stress, inflammation, and defense of homeostasis. Mol Cell. 2014;54:281–8.View ArticlePubMedPubMed CentralGoogle Scholar
- Breuer K, Foroushani AK, Laird MR, Chen C, Sribnaia A, Lo R, Winsor GL, Hancock RE, Brinkman FS, Lynn DJ. InnateDB: systems biology of innate immunity and beyond--recent updates and continuing curation. Nucleic Acids Res. 2013;41:D1228–33.Google Scholar
- Kelley J, de Bono B, Trowsdale J. IRIS: a database surveying known human immune system genes. Genomics. 2005;85:503–11.View ArticlePubMedGoogle Scholar
- Kramer A, Green J, Pollard Jr J, Tugendreich S. Causal analysis approaches in Ingenuity Pathway Analysis. Bioinformatics. 2014;30:523–30.View ArticlePubMedGoogle Scholar
- Doyle S, Vaidya S, O'Connell R, Dadgostar H, Dempsey P, Wu T, Rao G, Sun R, Haberland M, Modlin R, Cheng G . IRF3 mediates a TLR3/TLR4-specific antiviral gene program. Immunity. 2002;17:251–63.Google Scholar
- Sato T, Vries RG, Snippert HJ, van de Wetering M, Barker N, Stange DE, van Es JH, Abo A, Kujala P, Peters PJ, Clevers H . Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche. Nature. 2009;459:262–5.Google Scholar
- Wolff WI, Shinya H. Definitive treatment of "malignant" polyps of the colon. Ann Surg. 1975;182:516–25.View ArticlePubMedPubMed CentralGoogle Scholar
- Burn J, Gerdes AM, Macrae F, Mecklin JP, Moeslein G, Olschwang S, Eccles D, Evans DG, Maher ER, Bertario L, et al. Long-term effect of aspirin on cancer risk in carriers of hereditary colorectal cancer: an analysis from the CAPP2 randomised controlled trial. Lancet. 2011;378:2081–7.Google Scholar
- Lasry A, Zinger A, Ben-Neriah Y. Inflammatory networks underlying colorectal cancer. Nature Immunology. 2016;17:230–240.View ArticlePubMedGoogle Scholar
- Rothwell PM, Fowkes FG, Belch JF, Ogawa H, Warlow CP, Meade TW. Effect of daily aspirin on long-term risk of death due to cancer: analysis of individual patient data from randomised trials. Lancet. 2011;377:31–41.View ArticlePubMedGoogle Scholar
- Streicher SA, Yu H, Lu L, Kidd MS, Risch HA. Case–control study of aspirin use and risk of pancreatic cancer. Cancer Epidemiol Biomarkers Prev. 2014;23:1254–63.View ArticlePubMedPubMed CentralGoogle Scholar
- Hawiger J. Innate immunity and inflammation: a transcriptional paradigm. Immunol Res. 2001;23:99–109.View ArticlePubMedGoogle Scholar
- Hashimoto S, Suzuki T, Dong HY, Nagai S, Yamazaki N, Matsushima K. Serial analysis of gene expression in human monocyte-derived dendritic cells. Blood. 1999;94:845–52.PubMedGoogle Scholar
- Hashimoto S, Suzuki T, Dong HY, Yamazaki N, Matsushima K. Serial analysis of gene expression in human monocytes and macrophages. Blood. 1999;94:837–44.PubMedGoogle Scholar
- Barretina J, Caponigro G, Stransky N, Venkatesan K, Margolin AA, Kim S, Wilson CJ, Lehar J, Kryukov GV, Sonkin D, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature. 2012;483:603–7.Google Scholar
- Barbie DA, Tamayo P, Boehm JS, Kim SY, Moody SE, Dunn IF, Schinzel AC, Sandy P, Meylan E, Scholl C, et al. Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1. Nature. 2009;462:108–12.Google Scholar
- Aran D, Sirota M, Butte AJ. Systematic pan-cancer analysis of tumour purity. Nat Commun. 2015;6:8971.View ArticlePubMedPubMed CentralGoogle Scholar
- Consortium GT. The Genotype-Tissue Expression (GTEx) project. Nat Genet. 2013;45:580–5.View ArticleGoogle Scholar
- Auerbach RK, Chen B, Butte AJ. Relating genes to function: identifying enriched transcription factors using the ENCODE ChIP-Seq significance tool. Bioinformatics. 2013;29:1922–4.View ArticlePubMedPubMed CentralGoogle Scholar
- Rooney MS, Shukla SA, Wu CJ, Getz G, Hacohen N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell. 2015;160:48–61.View ArticlePubMedPubMed CentralGoogle Scholar
- Galon J, Mlecnik B, Bindea G, Angell HK, Berger A, Lagorce C, Lugli A, Zlobec I, Hartmann A, Bifulco C, et al. Towards the introduction of the 'Immunoscore' in the classification of malignant tumours. J Pathol. 2014;232:199–209.Google Scholar
- Taylor BS, Schultz N, Hieronymus H, Gopalan A, Xiao Y, Carver BS, Arora VK, Kaushik P, Cerami E, Reva B, et al. Integrative genomic profiling of human prostate cancer. Cancer Cell. 2010;18:11–22.Google Scholar
- Petitjean A, Achatz MI, Borresen-Dale AL, Hainaut P, Olivier M. TP53 mutations in human cancers: functional selection and impact on cancer prognosis and outcomes. Oncogene. 2007;26:2157–65.View ArticlePubMedGoogle Scholar
- Lamb J, Crawford ED, Peck D, Modell JW, Blat IC, Wrobel MJ, Lerner J, Brunet JP, Subramanian A, Ross KN, et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Science. 2006;313:1929–35.Google Scholar
- Coussens LM, Werb Z. Inflammation and cancer. Nature. 2002;420:860–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Fraser DM, Sullivan FM, Thompson AM, McCowan C. Aspirin use and survival after the diagnosis of breast cancer: a population-based cohort study. Br J Cancer. 2014;111(3):623–7.View ArticlePubMedPubMed CentralGoogle Scholar
- Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 2014;15:550.View ArticlePubMedPubMed CentralGoogle Scholar
- Yoshihara K, Shahmoradgoli M, Martinez E, Vegesna R, Kim H, Torres-Garcia W, Trevino V, Shen H, Laird PW, Levine DA, Van Houdt WJ, Pronk A, Van Gorp J, Siersema PD, Clevers H. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612.Google Scholar
- Sato T, Stange DE, Ferrante M, Vries RG, Van Es JH, Van den Brink S, Van Houdt WJ, Pronk A, Van Gorp J, Siersema PD, Clevers H. Long-term expansion of epithelial organoids from human colon, adenoma, adenocarcinoma, and Barrett's epithelium. Gastroenterology. 2011;141:1762–72.Google Scholar
- Li B, Dewey CN. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinformatics. 2011;12:323.View ArticlePubMedPubMed CentralGoogle Scholar
- Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7.View ArticlePubMedPubMed CentralGoogle Scholar