Figures
Abstract
Background
Currently, prognostication for pancreatic ductal adenocarcinoma (PDAC) is based upon a coarse clinical staging system. Thus, more accurate prognostic tests are needed for PDAC patients to aid treatment decisions.
Methods and Findings
Affymetrix gene expression profiling was carried out on 15 human PDAC tumors and from the data we identified a 13-gene expression signature (risk score) that correlated with patient survival. The gene expression risk score was then independently validated using published gene expression data and survival data for an additional 101 patients with pancreatic cancer. Patients with high-risk scores had significantly higher risk of death compared to patients with low-risk scores (HR 2.27, p = 0.002). When the 13-gene score was combined with lymph node status the risk-score further discriminated the length of patient survival time (p<0.001). Patients with a high-risk score had poor survival independent of nodal status; however, nodal status increased predictability for survival in patients with a low-risk gene signature score (low-risk N1 vs. low-risk N0: HR = 2.0, p = 0.002). While AJCC stage correlated with patient survival (p = 0.03), the 13-gene score was superior at predicting survival. Of the 13 genes comprising the predictive model, four have been shown to be important in PDAC, six are unreported in PDAC but important in other cancers, and three are unreported in any cancer.
Conclusions
We identified a 13-gene expression signature that predicts survival of PDAC patients and could prove useful for making treatment decisions. This risk score should be evaluated prospectively in clinical trials for prognostication and for predicting response to chemotherapy. Investigation of new genes identified in our model may lead to novel therapeutic targets.
Citation: Newhook TE, Blais EM, Lindberg JM, Adair SJ, Xin W, Lee JK, et al. (2014) A Thirteen-Gene Expression Signature Predicts Survival of Patients with Pancreatic Cancer and Identifies New Genes of Interest. PLoS ONE 9(9): e105631. https://doi.org/10.1371/journal.pone.0105631
Editor: Alfons Navarro, University of Barcelona, Spain
Received: April 15, 2014; Accepted: July 22, 2014; Published: September 2, 2014
Copyright: © 2014 Newhook et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The authors confirm that all data underlying the findings are fully available without restriction. The data presented in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE46385.
Funding: The authors have no funding or support to report.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Pancreatic ductal adenocarcinoma (PDAC) has the shortest survival duration of any solid organ malignancy [1], [2]. Currently, prognostication for patients with PDAC is based on the 7th edition of the American Joint Committee on Cancer (AJCC) staging system that takes into account size and invasive properties of the tumor and presence of nodal and distant metastatic disease [3]. This staging system remains the primary consideration by physicians to determine appropriate treatment as well as offer prognostic information for patients and families [3]. Significant ranges in survival exist within individual AJCC clinical stages [4], [5], [6]; for instance, stage IV patients may live only a few weeks after diagnosis or may live longer than one to two years with treatment. It is likely that this intra-stage variance is due to heterogeneous tumor gene expression resulting in differences in tumor biology.
We report the identification and validation of a 13-gene expression signature that predicts survival for patients with PDAC with stratification of patients into high- and low-risk groups based on the coordinate expression of genes defined by the gene expression signature. Assessment of lymph node status further added to the prognostic efficacy of the signature. The genes and pathways whose expression makes up the 13-gene signature represent possible targets for further research into the biology of PDAC tumors.
Methods
Ethics Statement
PDAC sample collection and processing were carried out with approval of the Institutional Review Board of the University of Virginia in coordination with the Biorepository and Tissue Research Facility. All patients provided written consent for participation. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health [7]. The protocol was approved by the Animal Care and Use Committee of the University of Virginia (PHS Assurance #A3245-01).
Propagation of Patient-Derived Tumors in Immunocompromised Mice for Gene Expression Profiling
The collection, pathologic examination, and propagation of human patient-derived PDAC specimens in immunocompromised mice were performed as previously described [8], [9]. Following surgical resection and pathological review of the patient tumor, residual tumor tissues were collected and placed in Roswell Park Memorial Institute media (RPMI) for surgical transplantation in mice. Six to eight week old, male, non-obese, diabetic, severe combined immunodeficient (NOD SCID) and athymic nude mice (National Cancer Institute, Fredricksburg, MD) were used. To achieve more efficient engraftment during initial establishment of the human PDAC tumor line, NOD SCID mice were used for the first two generations. For propagation of the tumor line beyond these first two generations, athymic nude mice were used, as they retain innate immunity (natural killer cells, B lymphocytes, antigen presenting cells, and complement activity), which is impaired in NOD SCID mice. Mice were housed in pathogen-free conditions, acclimated to their new surroundings for at least 48 hours prior to tumor engraftment, and maintained in accordance with institutional standards. All animal surgery was performed under 2,2,2-tribromoethanol anesthesia (4 mg/10 gm body weight). Post-surgery mice were administered ketoprofen 0.1 mg for pain control and were observed continuously for signs of pain or distress (hypoactivity, restlessness, vocalization, hiding, lack of grooming, abnormal posture, tremor, or respiratory distress) until they recovered from anesthesia, then monitored daily for 48 hr for signs of pain or distress. Humane endpoints were observed throughout experiments with mice being sacrificed when tumors reached a volume greater than 1500 mm3 by MRI assessment or when mice developed 15% weight loss. Mice were sacrificed via isofluorane anesthesia followed by cervical dislocation.
Human tumors were surgically implanted onto the pancreata of mice immediately following resection from either a patient or earlier generation xenograft. A 1.5-cm left flank incision was used to access the peritoneum of anesthetized mice, the pancreas was exteriorized using a sterile cotton swab and a small piece (∼25 mm3) of fresh patient tumor was sutured onto the pancreas using 5-0 Prolene (Ethicon, Cornelia, GA). The pancreas was repositioned and the wound closed using 4-0 Vicryl suture (Ethicon).
Human tumor tissue comprising tumor and associated stroma (without laser microdissection) was preserved after harvest from individual xenografts using AllProtect (Qiagen, Valencia, CA) for efficient RNA preservation. Tissue homogenization was performed utilizing the TissueLyzer LT (Qiagen) and RNA extraction was performed using the RNAeasy kit (Qiagen), according the manufacturer's instructions. Gene expression analysis using the Affymetrix GeneChip platform (Affymetrix, Santa Clara, CA) using the Human Genome U133 Plus 2.0 Arrays and the GeneChipt 3′ IVT Express Labeling Assay was carried out by the University of Virginia Biomolecular Research Facility.
Development of Prognostic Gene Signature and Statistical Analysis
Expression data sets from two distinctive patient cohorts were employed for gene prediction modeling and independent validation (Table 1). The first data set was derived from a cohort of 15 patients with pancreatic cancer at the University of Virginia (UVA-15; GSE46385) and was used for the gene expression biomarker discovery and prediction modeling for patient survival. To identify an appropriate external validation data set, we conducted a search for linked gene expression and overall survival data for patients with pancreatic cancer in the following publicly available databases: National Center for Biotechnology Information – Gene Expression Omnibus (NCBI GEO; http://www.ncbi.nlm.nih.gov/geo/), European Molecular Biology Laboratory – European Bioinformatics Institute Array Express (EBML EBI; https://www.ebi.ac.uk/arrayexpress/), The Cancer Genome Atlas (http://cancergenome.nih.gov/), and Oncomine (https://www.oncomine.org/resource/login.html). From this search, we found a data set derived from a cohort of 101 PDAC patients (Stratford-101; NCBI GEO Database: GSE21501). The second external data set reported appropriate survival information of 45 PDAC patient samples (GSE28735), so the larger external dataset (Stratford-101; GSE21501) was selected as the validation set. Raw expression data were downloaded from GEO, examined for quality control, and pre-processed using robust multi-array averaging (RMA) and quantile normalization methods in the R/Bioconductor programming environment. Patients were grouped by overall survival duration into short survival time (n = 7, survival range: 2.0–9.0 mo; median: 6.1 mo) and long survival time (n = 8, survival range: 10.6–32.8 mo; median: 13.7 mo, Fig. 1A). Using the UVA-15 data set, genes that were significantly differentially expressed between short and long survival groups were identified using both non-parametric Wilcoxon test and two-sample t-test to identify genes that were consistently associated with patient survival. The data were fitted to a Cox proportional hazard regression model using the metagene signatures (principal components) of 15 genes based on a statistical dimension reduction technique.
(A) Kaplan-Meier overall survival of patients comprising the UVA-15 derivation set grouped by short survival time (n = 7, survival range: 2.0–9.0 mo; median: 6.1 mo) and long survival time (n = 8, survival range: 10.6–32.8 mo; median: 13.7 mo; log-rank p<0.001). (B) Expression of 13-genes in the 15-tumor derivation set of patients with PDAC reveals clustering into high- (purple bar) and low-risk (yellow bar) populations. (C) Application of the 13-gene signature to an independent validation set of 101 patients with localized and resected PDAC reveals clustering into high- (purple bar) and low-risk (yellow bar) groups based on gene expression. (D) Kaplan-Meier overall survival of the independent validation set according to high- and low-risk groups as determined by 13-gene signature (log-rank p = 0.001).
Applying the fitted Cox regression model independently, survival times for patients in the Stratford-101 cohort were determined. The predicted survival times of the Stratford-101 patients were ranked and converted into percentiles - 1 for the patient with the shortest survival time and 100 for the patient with the longest survival time. The statistical significance of the predicted survival scores were evaluated compared to actual patient survival times using Student two-sample t-test at the optimal prediction (percentile) score cutoff that maximized the survival benefit with the highest positive predictive value for long-term survivors. Kaplan-Meier survival analysis was also performed at this cutoff point. Risk hazard ratios were also obtained from the Cox regression model for several contrasting conditions of interest on the Stratford-101 cohort.
Pathway-level gene expression changes between high risk and low risk patients from the Stratford-101 dataset were identified using gene set enrichment analysis. Annotated gene sets for Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were downloaded from the Molecular Signatures Database [10], [11], [12]. Expression changes for genes and KEGG pathways were evaluated using linear models for microarrays (limma) and gene set variability analysis [13], [14], [15]. A p-value cutoff of 0.05 was applied after false discovery rate (FDR) correction.
Results
Patient and tumor characteristics
Table 1 summarizes the information for the 15 patients comprising the derivation set all of whom underwent surgery for PDAC at the University of Virginia. Tumor stage ranged from I to IV, with the majority of patients having stage IIb disease with positive lymph nodes. The 40% of patients in the derivation set with stage IV disease underwent excision/biopsy of a metastasis (4 patients with liver metastases, one patient with peritoneal metastasis, and one with pleural metastasis), but did not undergo resection of their primary tumor. No patients in the derivation set received any form of neoadjuvant therapy, while all patients with localized disease who underwent resection received post-operative gemcitabine-based chemotherapy and 5 of 6 patients with metastatic disease received post-operative palliative gemcitabine-based chemotherapy (Table 1).
A total of 101 patients with localized, resected PDAC comprised the validation set for the gene expression signature [16]. The majority of patients within this group had stage IIb PDAC (72%) and none had stage IV disease, as all tumors were resectable (Table 1). Frequency of neoadjuvant and adjuvant therapies administered to this patient group was unavailable.
Identification and validation of a 13-gene prognostic signature
The 13 genes that comprise the 13-gene expression signature are described in Table 2. To evaluate the predictive ability of the initial candidate 13-gene prognostic signature we assessed an independent gene expression data set derived from 101 patients with localized primary PDAC [16]. The optimal cutoff point differentiating low risk score vs. high risk score was determined to be 70 by maximizing the Youden index (sensitivity+specificity−1) with the constraint that the proportion of high-risk patients was at least 10% or higher for practical clinical applications. The 13-gene prognostic score was then applied with a cutpoint of 70 (e.g. low risk: <70; high risk: >70), which provided the significant survival benefit and difference between patients with low vs. high risk scores in the derivation set of 15 patients (Fig. 1A). Heat maps of the gene expression signature for the UVA-15 derivation set and 101-tumor validation set are shown in Figure 1. This application of the gene signature effectively stratified patients into high- and low-risk groups with a median overall survival (MS) of 14.0 v. 21.0 months, respectively (Fig. 1D). Furthermore, patients in the high-risk group had a greater than two-fold increase risk of death as compared to those in the low-risk group (HR 2.27 [95% CI 1.34–3.85], p = 0.002; Fig. 1D).
A 13-gene expression signature more accurately predicts survival of PDAC patients when combined with nodal status
We next sought to further refine our prognostic gene score by incorporating lymph node status with the risk score. This effectively stratified patients into four groups – low-risk score, node-negative (n = 22); low-risk score, node-positive (n = 48); high-risk score, node-negative (n = 6); and high-risk score, node-positive (n = 25). As demonstrated in Figure 2A, patients in the low-risk, node-negative group had the best prognosis (MS: 41.0 mo), followed by low-risk, node-positive patients (MS: 18.0 mo). Patients with high-risk score had poor median survival, independent of lymph node status (node-negative: 15.5 mo; node-positive: 14.0 mo; p = NS; Fig. 2A). Compared to low risk, node-negative patients, the high-risk, node-positive patients had a near 4-fold increased risk of death (HR = 3.77 [95%CI: 1.75–8.10], p = 0.007), and high-risk, node-negative patients had 3-fold increase (HR = 3.09 [95% CI: 1.05–9.03], p = 0.007), whereas low-risk, node-positive patients had a 2-fold increased risk (HR = 1.95 [95% CI: 0.98–3.88], p = 0.007; Fig. 2A).
Kaplan-Meier overall survival of (A) a validation set of 101 patients with localized, resected PDAC according to 13-gene prognostic score combined with pathologic lymph node status at the time of surgery, and (B) the same 101 patients grouped according to either high-risk 13-gene prognostic score alone or low-risk 13-gene prognostic score plus pathologic nodal status at time of surgery.
Interestingly, what is apparent from the aforementioned results is that patients with high-risk tumors based on gene expression have a poor prognosis, regardless of nodal status; whereas, nodal status further refines prognosis for patients with low-risk tumors. Thus, there are three distinct prognostic groups: high-risk patients (n = 31; MS: 14.0 mo; Fig. 2B), low-risk, node-positive patients (n = 48; MS: 18.0 mo; Fig. 2B), and low-risk, node-negative patients (n = 22; MS: 41.0 mo; 2B).
Pathway analysis reveals key pathways differentially expressed between tumors with high- and low-risk prognostic signatures
To link the observed changes in gene expression with molecular and cellular pathways that may impact in the observed differential survival between high- and low-risk groups, we evaluated 5199 of 17623 genes and 97 of 186 KEGG pathways that were significantly differentially expressed between patients with high-risk and low-risk prognostic scores. Differentially expressed KEGG pathways between high- and low-risk patients included cancer cell signaling pathways (MAPK, VEGF, MTOR, and ERBB signaling pathways) and cancer pathways (acute myeloid leukemia, non-small cell lung cancer, chronic myeloid leukemia, and pancreatic cancer; Table 3). Additionally, three genes from the 13-gene prognostic signature, MDM2, PLCG1, and TGFA, were represented in 9 out of the top 20 significant pathways. These results revealed that genes involved in canonical cancer signaling pathways were most differentially expressed between high-risk and low-risk tumors and that the activity of these pathways may be responsible for the observed difference in survival between high- and low-risk patient groups.
Discussion
We report a 13-gene expression signature, derived from gene expression analysis of 15 patients with PDAC and externally validated on gene expression data from a separate cohort of 101 patients, which accurately predicts patient survival. Because this model was based on overall survival for patients with stage I to IV disease, we believe this is the most logical and accurate prognostic gene expression signature reported for patients with PDAC.
Due to variation in survival within AJCC clinical stages and to the large genomic heterogeneity within PDAC tumors, investigation into prognostic gene expression patterns has been increasingly reported [3], [16], [17], [18], [19], [20]. In a previously reported study of patients with metastatic versus non-metastatic PDAC, a 6-gene prognostic signature correlated with survival; however, this signature was derived from tumor stage at presentation and not from patient survival [16]. Moreover, no overlap exists between the candidate genes in the 6-gene signature and the 13-gene expression signature described herein, which was in fact based on patient survival. Thus, because of the selection of the derivation patient set, we believe this 13-gene expression signature outperforms others reported in the literature for patients with PDAC.
Commercially available mutational and gene expression profiling platforms are increasingly utilized as adjuncts to conventional clinical treatment algorithms for treatment of cancers including breast, prostate, and colon cancer [21], [22], [23], [24]. These expression analyses are arguably most robust in predicting outcomes for patients with breast cancer, including OncotypeDX and MammaPrint [25], [26]. These platforms are used to predict early outcomes and risk of metastasis in breast cancer; however, further applications of these tools helps tailor treatment based on predicting response to therapies [25], [27], [28], [29], [30], [31]. To date, no such prognostic tool is commercially available for patients with PDAC; however, predicting survival for patients with PDAC based on individual tumor biology would clearly benefit patients' and clinicians' therapeutic decisions.
The individual genes whose expression levels were used to derive this 13-gene prognostic signature reveal an intriguing network of pathways that impact PDAC patient survival (Table 2 and 3). Many of these genes have been implicated in various human cancers, including pancreatic cancer; however, some have not been reported to be associated with any cancers to date. Recognizable genes such as TGFA, ELAVL1 and MDM2, and less so MS4A3 are over-expressed in PDAC lesions or associated with patient prognosis [32], [33], [34], [35], [36]. Interestingly, genes such as CCDC88C, CD200R1, and CUL3 have been associated with prognosis or being highly expressed in other forms of cancer, however they have not been reported in PDAC to the best of our knowledge [37], [38], [39], [40]. Their involvement in the 13-gene prognostic signature is the first report of their expression being implicated in patient survival in PDAC. The identification of measureable differences in gene expression between PDAC tumors from patients with varying survival times supports the further application of our gene signature and investigation into these various pathways.
No patients within the derivation set received neoadjuvant therapy of any form, and therefore the gene expression analysis of these tumor samples represents the profile of the tumor prior to any systemic therapy (Table 1). However, within the prognostic calculation is the expression of ELAVL1, also known as Hu antigen-R (HuR), which has been implicated in PDAC response to chemotherapy [33]. In fact, PDAC patients with low levels of ELAVL1 expression have a 7-fold increase in mortality [33]. In our analysis, high-risk PDAC tumors have a decreased expression of ELAVL1 as compared to low-risk tumors and 93% of patients within the derivation set received standard-of-care adjuvant or palliative gemcitabine-based therapy. Unfortunately, clinical data on adjuvant therapy regimen of the 101 patients within validation set was unavailable; however, it stands to reason that the majority of patients also received standard of care gemcitabine adjuvant therapy. The 13-gene expression signature may predict patient response to adjuvant gemcitabine therapy and we plan to evaluate the ability of the 13-gene signature to predict response to therapy in future studies.
A particular strength of this study is that the patient tumors within the derivation set are from a good representative sample of AJCC disease stages (Table 1). A limitation of our study is that the validation set was comprised of 101 patients with localized, resectable PDAC and thus the great majority of patients had AJCC stage IIb or less (Table 1). We hypothesize that having an increased number of patients with stage III or IV disease in the derivation set would only serve to increase the stratification based on our gene signature with and without the addition of nodal status because this would add a larger number of patients with either high-risk disease, or, low-risk node-positive disease and push these survival stratifications toward greater statistical significance. Despite this, the fact that our signature was externally validated on a set of 101 patients with PDAC whose gene expression and clinical data was publically available adds to the unbiased nature of our study.
Due to the coarse nature of the current clinical PDAC staging paradigm, additional prognostic tools are needed to aid in therapeutic decision-making. The decision to undergo pancreatic resection, which is the only potentially curative option for PDAC patients, is a stressful one considering the 15–26% readmission rate and complications occurring in approximately 40% of patients after surgery even at centers of excellence [41], [42], [43], [44]. Additionally, systemic chemotherapy can be associated with significant toxicity and can negatively impact quality of life. Given the uncertain benefit of chemotherapy for a given patient, the concerns over treatment toxicity and quality of life are key factors facing patients. The ability to offer patients and clinicians accurate prognostic data about PDAC tumors based on gene expression measurement of individual tumor biology is invaluable and could influence the decision to offer (or forego) therapy. This prognostic instrument has the potential to aid patients and physicians in making treatment decisions, which ultimately may affect outcome and impact quality of life. Future evaluation of this gene signature will assess for the ability to predict response to chemotherapy.
Author Contributions
Conceived and designed the experiments: TWB JTP. Performed the experiments: TEN JML SJA. Analyzed the data: TEN EMB WX JKL JAP TWB. Contributed reagents/materials/analysis tools: TWB JTP. Wrote the paper: TEN JTP TWB.
References
- 1. Siegel R, Naishadham D, Jemal A (2012) Cancer statistics, 2012. CA: a cancer journal for clinicians 62: 10–29.
- 2. Cameron JL, Riall TS, Coleman J, Belcher KA (2006) One thousand consecutive pancreaticoduodenectomies. Annals of surgery 244: 10–15.
- 3. Edge SB, Compton CC (2010) The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 17: 1471–1474.
- 4. Geer RJ, Brennan MF (1993) Prognostic indicators for survival after resection of pancreatic adenocarcinoma. American journal of surgery 165: 68–72 discussion 72-63.
- 5. Sohn TA, Yeo CJ, Cameron JL, Koniaris L, Kaushal S, et al. (2000) Resected adenocarcinoma of the pancreas-616 patients: results, outcomes, and prognostic indicators. J Gastrointest Surg 4: 567–579.
- 6. Weber CE, Bock EA, Hurtuk MG, Abood GJ, Pickleman J, et al. (2014) Clinical and pathologic features influencing survival in patients undergoing pancreaticoduodenectomy for pancreatic adenocarcinoma. J Gastrointest Surg 18: 340–347.
- 7.
National Research Council (U.S.). Committee for the Update of the Guide for the Care and Use of Laboratory Animals., Institute for Laboratory Animal Research (U.S.), National Academies Press (U.S.) (2011) Guide for the care and use of laboratory animals. Washington, D.C.: National Academies Press. xxv, 220 p. p.
- 8. Stokes JB, Adair SJ, Slack-Davis JK, Walters DM, Tilghman RW, et al. (2011) Inhibition of focal adhesion kinase by PF-562,271 inhibits the growth and metastasis of pancreatic cancer concomitant with altering the tumor microenvironment. Molecular Cancer Therapeutics 10: 2135–2145.
- 9. Walters DM, Stokes JB, Adair SJ, Stelow EB, Borgman CA, et al. (2013) Clinical, molecular and genetic validation of a murine orthotopic xenograft model of pancreatic adenocarcinoma using fresh human specimens. PloS one 8: e77065.
- 10. Kanehisa M, Goto S, Sato Y, Kawashima M, Furumichi M, et al. (2014) Data, information, knowledge and principle: back to metabolism in KEGG. Nucleic Acids Res 42: D199–205.
- 11. Kanehisa M, Goto S (2000) KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28: 27–30.
- 12. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, et al. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A 102: 15545–15550.
- 13. Yamabuki T, Daigo Y, Kato T, Hayama S, Tsunoda T, et al. (2006) Genome-wide gene expression profile analysis of esophageal squamous cell carcinomas. Int J Oncol 28: 1375–1384.
- 14. Smyth GK (2004) Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3: Article3.
- 15. Hanzelmann S, Castelo R, Guinney J (2013) GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 14: 7.
- 16. Stratford JK, Bentrem DJ, Anderson JM, Fan C, Volmar KA, et al. (2010) A six-gene signature predicts survival of patients with localized pancreatic ductal adenocarcinoma. PLoS medicine 7: e1000307.
- 17. Bilimoria KY, Bentrem DJ, Ko CY, Ritchey J, Stewart AK, et al. (2007) Validation of the 6th edition AJCC Pancreatic Cancer Staging System: report from the National Cancer Database. Cancer 110: 738–744.
- 18. Jones S, Zhang X, Parsons DW, Lin JC, Leary RJ, et al. (2008) Core signaling pathways in human pancreatic cancers revealed by global genomic analyses. Science 321: 1801–1806.
- 19. Donahue TR, Tran LM, Hill R, Li Y, Kovochich A, et al. (2012) Integrative survival-based molecular profiling of human pancreatic cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 18: 1352–1363.
- 20. Kim HN, Choi DW, Lee KT, Lee JK, Heo JS, et al. (2007) Gene expression profiling in lymph node-positive and lymph node-negative pancreatic cancer. Pancreas 34: 325–334.
- 21. Knezevic D, Goddard AD, Natraj N, Cherbavaz DB, Clark-Langone KM, et al. (2013) Analytical validation of the Oncotype DX prostate cancer assay - a clinical RT-PCR assay optimized for prostate needle biopsies. BMC Genomics 14: 690.
- 22. Clark-Langone KM, Sangli C, Krishnakumar J, Watson D (2010) Translating tumor biology into personalized treatment planning: analytical performance characteristics of the Oncotype DX Colon Cancer Assay. BMC Cancer 10: 691.
- 23. Cronin M, Sangli C, Liu ML, Pho M, Dutta D, et al. (2007) Analytical validation of the Oncotype DX genomic diagnostic test for recurrence prognosis and therapeutic response prediction in node-negative, estrogen receptor-positive breast cancer. Clin Chem 53: 1084–1091.
- 24. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, et al. (2002) Gene expression profiling predicts clinical outcome of breast cancer. Nature 415: 530–536.
- 25. Paik S, Shak S, Tang G, Kim C, Baker J, et al. (2004) A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. The New England journal of medicine 351: 2817–2826.
- 26. Cardoso F, Piccart-Gebhart M, Van't Veer L, Rutgers E (2007) The MINDACT trial: the first prospective clinical validation of a genomic tool. Mol Oncol 1: 246–251.
- 27. Jansen MP, Foekens JA, van Staveren IL, Dirkzwager-Kiel MM, Ritstier K, et al. (2005) Molecular classification of tamoxifen-resistant breast carcinomas by gene expression profiling. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 23: 732–740.
- 28. Ayers M, Symmans WF, Stec J, Damokosh AI, Clark E, et al. (2004) Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 22: 2284–2293.
- 29. Folgueira MA, Carraro DM, Brentani H, Patrao DF, Barbosa EM, et al. (2005) Gene expression profile associated with response to doxorubicin-based therapy in breast cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 11: 7434–7443.
- 30. Hannemann J, Oosterkamp HM, Bosch CA, Velds A, Wessels LF, et al. (2005) Changes in gene expression associated with response to neoadjuvant chemotherapy in breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 23: 3331–3342.
- 31. Chang JC, Wooten EC, Tsimelzon A, Hilsenbeck SG, Gutierrez MC, et al. (2003) Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 362: 362–369.
- 32. Siveke JT, Einwachter H, Sipos B, Lubeseder-Martellato C, Kloppel G, et al. (2007) Concomitant pancreatic activation of Kras(G12D) and Tgfa results in cystic papillary neoplasms reminiscent of human IPMN. Cancer Cell 12: 266–279.
- 33. Costantino CL, Witkiewicz AK, Kuwano Y, Cozzitorto JA, Kennedy EP, et al. (2009) The role of HuR in gemcitabine efficacy in pancreatic cancer: HuR Up-regulates the expression of the gemcitabine metabolizing enzyme deoxycytidine kinase. Cancer Research 69: 4567–4572.
- 34. Ebert M, Yokoyama M, Kobrin M, Friess H, Buchler M, et al. (1994) Increased mdm2 expression and immunoreactivity in human pancreatic ductal adenocarcinoma. Int J Oncol 5: 1279–1284.
- 35. Kutok JL, Yang X, Folkerth R, Adra CN (2011) Characterization of the expression of HTm4 (MS4A3), a cell cycle regulator, in human peripheral blood cells and normal and malignant tissues. Journal of Cellular and Molecular Medicine 15: 86–93.
- 36. Sui X, Shin S, Zhang R, Firozi PF, Yang L, et al. (2009) Hdm2 is regulated by K-Ras and mediates p53-independent functions in pancreatic cancer cells. Oncogene 28: 709–720.
- 37. Long J, Zhang B, Signorello LB, Cai Q, Deming-Halverson S, et al. (2013) Evaluating genome-wide association study-identified breast cancer risk variants in african-american women. PLoS One 8: e58350.
- 38. Stumpfova M, Ratner D, Desciak EB, Eliezri YD, Owens DM (2010) The immunosuppressive surface ligand CD200 augments the metastatic capacity of squamous cell carcinoma. Cancer Res 70: 2962–2972.
- 39. Grau L, Luque-Garcia JL, Gonzalez-Peramato P, Theodorescu D, Palou J, et al. (2013) A quantitative proteomic analysis uncovers the relevance of CUL3 in bladder cancer aggressiveness. PloS one 8: e53328.
- 40. Hollstein PE, Cichowski K (2013) Identifying the ubiquitin ligase complex that regulates the NF1 tumor suppressor and Ras. Cancer Discov
- 41. Emick DM, Riall TS, Cameron JL, Winter JM, Lillemoe KD, et al. (2006) Hospital readmission after pancreaticoduodenectomy. J Gastrointest Surg 10: 1243–1252 discussion 1252-1243.
- 42. Yeo CJ, Cameron JL, Sohn TA, Lillemoe KD, Pitt HA, et al. (1997) Six hundred fifty consecutive pancreaticoduodenectomies in the 1990s: pathology, complications, and outcomes. Annals of surgery 226: 248–257 discussion 257-260.
- 43. Fong ZV, Ferrone CR, Thayer SP, Wargo JA, Sahora K, et al. (2013) Understanding Hospital Readmissions After Pancreaticoduodenectomy: Can We Prevent Them? : A 10-Year Contemporary Experience with 1,173 Patients at the Massachusetts General Hospital. J Gastrointest Surg
- 44. Valsangkar NP, Morales-Oyarvide V, Thayer SP, Ferrone CR, Wargo JA, et al. (2012) 851 resected cystic tumors of the pancreas: a 33-year experience at the Massachusetts General Hospital. Surgery 152: S4–12.
- 45. Tonks A, Hills R, White P, Rosie B, Mills KI, et al. (2007) CD200 as a prognostic factor in acute myeloid leukaemia. Leukemia 21: 566–568.
- 46. Moreaux J, Hose D, Reme T, Jourdan E, Hundemer M, et al. (2006) CD200 is a new prognostic factor in multiple myeloma. Blood 108: 4194–4197.
- 47. Speisky D, Duces A, Bieche I, Rebours V, Hammel P, et al. (2012) Molecular profiling of pancreatic neuroendocrine tumors in sporadic and Von Hippel-Lindau patients. Clinical cancer research : an official journal of the American Association for Cancer Research 18: 2838–2849.
- 48. Rezaei M, Friedrich K, Wielockx B, Kuzmanov A, Kettelhake A, et al. (2012) Interplay between neural-cadherin and vascular endothelial-cadherin in breast cancer progression. Breast Cancer Res 14: R154.
- 49. Craig DW, O'Shaughnessy JA, Kiefer JA, Aldrich J, Sinari S, et al. (2013) Genome and transcriptome sequencing in prospective metastatic triple-negative breast cancer uncovers therapeutic vulnerabilities. Molecular cancer therapeutics 12: 104–116.
- 50. Yuan Z, Sanders AJ, Ye L, Wang Y, Jiang WG (2011) Prognostic value of the human antigen R (HuR) in human breast cancer: high level predicts a favourable prognosis. Anticancer Res 31: 303–310.
- 51. Yuan Z, Sanders AJ, Ye L, Wang Y, Jiang WG (2011) Knockdown of human antigen R reduces the growth and invasion of breast cancer cells in vitro and affects expression of cyclin D1 and MMP-9. Oncol Rep 26: 237–245.
- 52. Lim SJ, Kim HJ, Kim JY, Park K, Lee CM (2007) Expression of HuR is associated with increased cyclooxygenase-2 expression in uterine cervical carcinoma. Int J Gynecol Pathol 26: 229–234.
- 53. Wang J, Wang B, Bi J, Zhang C (2011) Cytoplasmic HuR expression correlates with angiogenesis, lymphangiogenesis, and poor outcome in lung cancer. Med Oncol 28 Suppl 1: S577–585.
- 54. Wade M, Li YC, Wahl GM (2013) MDM2, MDMX and p53 in oncogenesis and cancer therapy. Nature reviews Cancer 13: 83–96.
- 55. Kassis J, Moellinger J, Lo H, Greenberg NM, Kim HG, et al. (1999) A role for phospholipase C-gamma-mediated signaling in tumor cell invasion. Clinical cancer research : an official journal of the American Association for Cancer Research 5: 2251–2260.
- 56. Arteaga CL, Johnson MD, Todderud G, Coffey RJ, Carpenter G, et al. (1991) Elevated content of the tyrosine kinase substrate phospholipase C-gamma 1 in primary human breast carcinomas. Proceedings of the National Academy of Sciences of the United States of America 88: 10435–10439.
- 57. Serao NV, Delfino KR, Southey BR, Beever JE, Rodriguez-Zas SL (2011) Cell cycle and aging, morphogenesis, and response to stimuli genes are individualized biomarkers of glioblastoma progression and survival. BMC Med Genomics 4: 49.
- 58. Poggi A, Venturino C, Catellani S, Clavio M, Miglino M, et al. (2004) Vdelta1 T lymphocytes from B-CLL patients recognize ULBP3 expressed on leukemic B cells and up-regulated by trans-retinoic acid. Cancer Res 64: 9172–9179.