WO2013052480A1 - Marker-based prognostic risk score in colon cancer - Google Patents

Abstract

Methods and compositions for the prognosis and classification of cancer, especially colon cancer, are provided. For example, in certain aspects methods for colon cancer prognosis using expression analysis of selected biomarkers are described. In particular, a risk score may be developed to provide a cancer prognosis.

Description

DESCRIPTION

MARKER-BASED PROGNOSTIC RISK SCORE IN COLON CANCER BACKGROUND OF THE INVENTION

This application claims the benefit of United States Provisional Patent Application No 61/542,480, filed October 3, 2011, the entirety of which is incorporated herein by reference.

1. Field of the Invention

[0001] The present invention relates generally to the fields of oncology, molecular biology, cell biology, and cancer. More particularly, it concerns cancer prognosis or classification using molecular markers.

2. Description of Related Art

[0002] Colorectal cancer is one of the most common cancers in the United States and the rest of the world, accounting for an estimated 146,900 new cases and 49,920 deaths in 2009 in the United States alone (Parkin et al., 2005; Jemal et al., 2009). Although surgical resection is highly effective for patients with early-stage colon cancers, a high proportion of patients have relapse after complete surgical resection, with 40% to 50% of patients with stage III disease experiencing such relapse within 5 years (Carlsson et al., 1987; Midgley and Kerr, 1999). Development of various chemotherapy regimens, including 5-fluorouracil, oxaliplatin, and irinotecan, as first-line treatment has considerably improved tumor response rate and median overall survival (OS) (Midgley et ah, 2009; Kopetz et ah, 2009). The use of recently developed targeted drugs, such as cetuximab and bevacizumab, in combination with chemotherapy has further improved the survival of patients with advanced colon cancer (Cunningham et ah, 2004; Kabbinavar et ah, 2003). However, there is considerable clinicopathological heterogeneity among the tumors. In addition, tumors with similar histopathological appearance can follow significantly different clinical courses. Approximately 50% of patients with advanced colon cancer have a radiographic response after systemic chemotherapy (Midgley et ah, 2009). Thus, for better patient care and management, it is important to understand any molecular heterogeneity significantly associated with this differential response to chemotherapy and to develop models to predict those patients who would benefit the most or least. Clinicopathological staging systems (both AJCC and Dukes staging) have been the gold standard for prognostication (Mamounas et al., 1999). However, they offer little information about response to treatment in individual patients or about potential therapeutic targets. Clearly, there exists a need to discover novel prognostic markers for cancer patients, especially colon cancer patients. SUMMARY OF THE INVENTION

[0003] The present invention overcomes limitations in the prior art by providing biomarker genes or gene expression signatures that may be used to detect or predict the prognosis of a colon cancer. More specifically, a genome-wide survey of gene expression data was applied to distinguish subtypes of colon cancer that have distinct biological characteristics associated with prognosis and to identify potential biomarker genes or a gene expression signature that reflect the biological or clinical characteristics of each subtype. A prediction model was established and may be used to help guide treatment strategies for colon cancer patients, e.g., after surgery. For example, detection of biomarkers or expression patterns may be used to select or identify colon cancer patients who may need further treatment due to the aggressive biological characteristics of their disease. A limited number of genes whose expression patterns can predict the survival of patients as well as their response to chemotherapy are provided herein.

[0004] An aspect of the present invention relates to a method of providing a prognosis or prediction for a subject determined to have a colorectal cancer, comprising: (a) obtaining expression information of biomarkers in a colorectal cancer sample of a subject by testing said sample, the biomarkers being at least ten genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRES1, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, VNN1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALB1, ZIC2, CXCL5, TCN1, AGR3, TACSTD2, DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGF1; and (b) providing a prognosis or prediction for the subject based on the expression information, wherein, as compared with a reference expression level, increased expression of one or more genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRESl, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, VNN1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALB1, ZIC2, CXCL5, TCN1, AGR3, and TACSTD2 indicates a poor survival, a high risk of recurrence, or a favorable response to an adjuvant therapy, and increased expression of one or more genes selected from the group consisting of DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGFl indicates a favorable survival, a low risk of recurrence or a poor response to an adjuvant therapy. Said obtaining expression information may comprise obtaining or receiving the sample. The sample may be paraffin-embedded or frozen. Said obtaining expression information may comprise RNA quantification, such as, e.g., cDNA microarray, quantitative RT-PCR, in situ hybridization, Northern blotting, or nuclease protection. Said obtaining expression information may comprise protein quantification, such as, e.g., immunohistochemistry, an ELISA, a radioimmunoassay (RIA), an immunoradiometric assay, a fluoroimmunoassay, a chemiluminescent assay, a bioluminescent assay, a gel electrophoresis, or a Western blot analysis. Providing the prognosis or prediction may comprise generating a classifier based on the expression, wherein the classifier is defined as a weighted sum of expression levels of the biomarkers. The classifier may be generated on a computer. The classifier may be generated by a computer readable medium comprising machine executable instructions suitable for generating a classifier. Providing the prognosis or prediction may comprise classifying a group of subjects based on the classifier associated with individual subjects in the group with a reference value. The method may further comprise reporting said prognosis or prediction. The method may further comprise prescribing or administering an adjuvant therapy to said subject based on said prediction. The cancer may be a stage I cancer, a stage II cancer, a stage III cancer, or a stage IV cancer. In some embodiments, the cancer is not a stage IV cancer. [0005] Another aspect of the present invention relates to an array comprising a plurality of antigen-binding fragments that bind to expression products of biomarkers or a plurality of primers or probes that bind to transcripts of the biomarkers to assess expression levels, the biomarkers comprising at least ten genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRES1, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, VNN1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALB1, ZIC2, CXCL5, TCN1, AGR3, TACSTD2, DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGF1. The array may be a microchip, such as, e.g., a cDNA microarray.

[0006] Yet another aspect of the present invention relates to a kit comprising a plurality of antigen-binding fragments that bind to expression products of biomarkers or a plurality of primers or probes that bind to transcripts of the biomarkers to assess expression levels, the biomarkers comprising at least ten genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRESl, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, VNN1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALB1, ZIC2, CXCL5, TCN1, AGR3, TACSTD2, DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGF1, wherein said kit is housed in a container. [0007] The biomarkers may be measured in a sample either directly or indirectly. For example, in some embodiments, a cancer sample is directly obtained from a subject at or near the laboratory or location where the biological sample will be analyzed. In other embodiments, the cancer sample may be obtained by a third party and then transferred, e.g., to a separate entity or location for analysis. In other embodiments, the sample may be obtained and tested in the same location using a point-of-care test. In these embodiments, said obtaining refers to receiving the sample, e.g., from the patient, from a laboratory, from a doctor's office, from the mail, courier, or post office, etc. In some further aspects, the method may further comprise reporting the determination or test results to the subject, a health care payer, an attending clinician, a pharmacist, a pharmacy benefits manager, or any person that the determination or test results may be of interest.

[0008] Embodiments discussed in the context of methods and/or compositions of the invention may be employed with respect to any other method or composition described herein. Thus, an embodiment pertaining to one method or composition may be applied to other methods and compositions of the invention as well.

[0009] As used herein the terms "encode" or "encoding" with reference to a nucleic acid are used to make the invention readily understandable by the skilled artisan; however, these terms may be used interchangeably with "comprise" or "comprising," respectively.

[0010] As used herein the specification, "a" or "an" may mean one or more. As used herein in the claim(s), when used in conjunction with the word "comprising," the words "a" or "an" may mean one or more than one.

[0011] The use of the term "or" in the claims is used to mean "and/or" unless explicitly indicated to refer to alternatives only or the alternatives are mutually exclusive, although the disclosure supports a definition that refers to only alternatives and "and/or." As used herein "another" may mean at least a second or more.

[0012] Throughout this application, the term "about" is used to indicate that a value includes the inherent variation of error for the device, the method being employed to determine the value, or the variation that exists among the study subjects.

[0013] Other objects, features, and advantages of the present invention will become apparent from the following detailed description. It should be understood, however, that the detailed description and the specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The following drawings form part of the present specification and are included to further demonstrate certain aspects of the present invention. The invention may be better understood by reference to one or more of these drawings in combination with the detailed description of specific embodiments presented herein.

[0015] FIG. 1: Kaplan-Meier plots of the prognosis of patients with colon cancer in the Moffit cohort. Patients were stratified according to AJCC stage or gene expression patterns (2 clusters). Recurrence free survival data are not available from 32 patients.

[0016] FIGS. 2A-B: Construction of prediction model in the test cohort according to gene expression signatures from the Moffit cohort. FIG. 2A, Schematic overview of the strategy used for the construction of prediction models and evaluation of predicted outcomes based on gene expression signatures. FIG. 2B, Kaplan-Meier plots of OS. Patients were stratified according to AJCC stage or 2 subgroups predicted by compound covariate predictor (CCP). P values were obtained from the log-rank test. The + symbols in the panels indicate censored data.

[0017] FIGS. 3A-D: Significant association of two subtypes with adjuvant chemotherapy. FIG. 3A, Kaplan-Meier plots of DFS of colorectal cancer patients in the combined cohort. Patients were plotted according to prognostic expression signature of 114 genes (2 subtypes). Patients in stage I, II, and III with available adjuvant chemotherapy data were included for analysis (n = 266). FIG. 3B, Kaplan-Meier plots of DFS of colorectal cancer patients in the combined cohort (Patients in stage III, n = 109). FIG. 3C, Kaplan-Meier plots of patients in subtype B with stage III disease (n = 56). Patients were plotted according to presence and absence of adjuvant chemotherapy (CTX). FIG. 3D, Kaplan-Meier plots of patients in subtype A with stage III disease (n = 53). Patients were plotted according to presence and absence of adjuvant chemotherapy (CTX).

[0018] FIGS. 4A-B: Subtype-specific gene expression patterns conserved in all 3 cohorts of colorectal cancer patients. FIG. 4A, Venn diagram of genes with expression that differed significantly between subtype A and B colorectal cancer patients in the 3 different cohorts. Univariate test (2-sample ?-test) with multivariate permutation test (10,000 random permutations) was applied. In each comparison, a cut-off P value of less than .001 was applied to retain genes with expression that differed significantly between the 2 groups of tissues examined. FIG. 4B, Expression patterns of selected genes shared in the 3 colon cancer cohorts. The expressions of only 755 genes were commonly upregulated or downregulated in all 3 cohorts. Colored bars at the top of the heat map represent samples as indicated. [0019] FIG. 5: Gene set enrichment analysis of genes in prognostic gene expression signature. Fisher's exact test was applied to gene sets defined in Ingenuity Pathway Analysis database to identify enriched biological characteristics in prognostic gene expression signature.

[0020] FIGS. 6A-B: Kaplan-Meier plots of OS colon cancer patients in VMP cohort. In subset analysis of gene expression signature, patients in stage II (FIG. 6A) and III (FIG. 6B) were independently stratified by the signature. P values were obtained from the log-rank test. The + symbols in the panel indicate censored data.

[0021] FIGS. 7A-B: Construction of prediction model in 2nd test (Melbourne) cohort according to gene expression signatures from Moffit cohort. FIG. 7A, Schematic overview of the strategy used for the construction of prediction models and evaluation of predicted outcomes based on gene expression signatures. FIG. 7B, Kaplan-Meier plots of DFS. Patients were stratified according to AJCC stage or two subgroups predicted by compound covariate predictor (CCP). P values were obtained from the log-rank test. The + symbols in the panel indicate censored data. [0022] FIG. 8: Interaction of subgroups with adjuvant chemotherapy in patients with stage III colorectal cancer. Cox proportional hazard regression model was used to analyze interaction between subgroups and adjuvant chemotherapy treatment. Dotted lines represent 95% confidence interval of hazard ratios.

[0023] FIG. 9: TGF-β networks from Ingenuity® pathway analysis. Gene networks from Ingenuity® pathway analysis, showing networks of inter-connection among genes with expression significantly associated with the TGF-β pathway in conserved gene expression data from the 3 cohorts. Upregulated and downregulated genes in the H subgroup are indicated by red and green, respectively. The lines and arrows represent functional and physical interactions and directions of regulation, as demonstrated in the literature. Interactions with the TGF- β pathway are highlighted in bold light grey lines. [0024] FIG. 10: NFkB networks from Ingenuity® pathway analysis. Gene networks of interconnection among genes whose expression is significantly associated with NF-kappa B and SRC kinase family (LYN and FYN) in conserved gene expression data from the 3 cohorts. Up- and down-regulated genes in the H subgroup are indicated by red and green, respectively. Each line and arrow represent functional and physical interaction and direction of regulation demonstrated in the literature.

DESCRIPTION OF ILLUSTRATIVE EMBODIMENTS

[0025] The instant invention overcomes several major problems with current cancer prognosis in providing methods and compositions using novel combinations of biomarkers identified by expression profiling and survival analysis of colon cancer patients. In various embodiments, particular biomarkers have been identified that may be used predict response to chemotherapy and clinical outcome, e.g., overall survival or disease-free survival, in colorectal cancer patients.

I. Definitions

[0026] "Prognosis" refers to a prediction of how a patient will progress, and whether there is a chance of recovery. "Cancer prognosis" generally refers to a forecast or prediction of the probable course or outcome of the cancer. As used herein, cancer prognosis includes the forecast or prediction of any one or more of the following: duration of survival of a patient susceptible to or diagnosed with a cancer, duration of recurrence-free survival, duration of progression free survival of a patient susceptible to or diagnosed with a cancer, response rate in a group of patients susceptible to or diagnosed with a cancer, duration of response in a patient or a group of patients susceptible to or diagnosed with a cancer, and/or likelihood of metastasis in a patient susceptible to or diagnosed with a cancer. Prognosis also includes prediction of favorable responses to cancer treatments, such as a conventional cancer therapy. [0027] By "subject" or "patient" is meant any single subject for which therapy is desired, including humans, cattle, dogs, guinea pigs, rabbits, chickens, and so on. Also intended to be included as a subject are any subjects involved in clinical research trials not showing any clinical sign of disease, or subjects involved in epidemiological studies, or subjects used as controls. [0028] As used herein, "increased expression" refers to an elevated or increased level of expression in a cancer sample relative to a suitable control (e.g., a non-cancerous tissue or cell sample, a reference standard), wherein the elevation or increase in the level of gene expression is statistically significant (p < 0.05). Whether an increase in the expression of a gene in a cancer sample relative to a control is statistically significant can be determined using an appropriate t-test (e.g., one-sample t-test, two-sample t-test, Welch's t-test) or other statistical test known to those of skill in the art. Genes that are overexpressed in a cancer can be, for example, genes that are known, or have been previously determined, to be overexpressed in a cancer.

[0029] As used herein, "decreased expression" refers to a reduced or decreased level of expression in a cancer sample relative to a suitable control (e.g., a non-cancerous tissue or cell sample, a reference standard), wherein the reduction or decrease in the level of gene expression is statistically significant (p < 0.05). In some embodiments, the reduced or decreased level of gene expression can be a complete absence of gene expression, or an expression level of zero. Whether a decrease in the expression of a gene in a cancer sample relative to a control is statistically significant can be determined using an appropriate ?-test (e.g., one-sample ?-test, two-sample t-test, Welch's ?-test) or other statistical test known to those of skill in the art. Genes that are underexpressed in a cancer can be, for example, genes that are known, or have been previously determined, to be underexpressed in a cancer.

[0030] In a further embodiment, the marker level may be compared to the level of the marker from a control, wherein the control may comprise one or more tumor samples (e.g., colon cancer samples) taken from one or more patients determined as having a good prognosis ("good prognosis" control) or a poor prognosis ("poor prognosis" control), or both.

[0031] The control may comprise data obtained at the same time (e.g., in the same hybridization experiment) as the patient's individual data, or may be a stored value or set of values, e.g. stored on a computer, or on computer-readable media. If the latter is used, new patient data for the selected marker(s), obtained from initial or follow-up samples, can be compared to the stored data for the same marker(s) without the need for additional control experiments.

[0032] A good or bad prognosis may, for example, be assessed in terms of patient survival, likelihood of disease recurrence or disease metastasis (patient survival, disease recurrence and metastasis may for example be assessed in relation to a defined timepoint, e.g. at a given number of years after cancer surgery (e.g. surgery to remove one or more tumors) or after initial diagnosis. In one embodiment, a good or bad prognosis may be assessed in terms of overall survival or disease-free survival.

[0033] For example, a "good prognosis" may refer to an increased likelihood that a patient afflicted with cancer, particularly colon cancer, will remain disease-free (i.e., cancer-free). "Poor prognosis" may refer to an increased likelihood of a relapse or recurrence of the underlying cancer or tumor, metastasis, or death. Cancer patients classified as having a "good prognosis" may have an increased likelihood of remaining free of the underlying cancer or tumor. In contrast, "bad prognosis" cancer patients may have an increased likelihood of experiencing disease relapse, tumor recurrence, metastasis, or death. In particular embodiments, the time frame for assessing prognosis and outcome is, for example, less than one year, one, two, three, four, five, six, seven, eight, nine, ten, fifteen, twenty, or more years. In certain aspects, the relevant time for assessing prognosis or disease-free survival time may begin with the surgical removal of the tumor or suppression, mitigation, or inhibition of tumor growth. Thus, for example, in particular embodiments, a "good prognosis" refers to the likelihood that a colon cancer patient will remain free of the underlying cancer or tumor for a period of at least five years, such as for a period of at least ten years. In further aspects of the invention, a "poor prognosis" refers to the likelihood that a colon cancer patient will experience disease relapse, tumor recurrence, metastasis, or death within less than ten years, such as less than five years. Time frames for assessing prognosis and outcome provided herein are illustrative and are not intended to be limiting.

[0034] The term "high risk" means the patient is expected to have a relapse in a shorter period less than a predetermined value (for example, from a control), for example in less than 5 years, preferably in less than 3 years. The term "low risk" means the patient is expected to have a relapse in a shorter period more than a predetermined value, for example, after 5 years, or in more than 3 years. Time frames for assessing risks provided herein are illustrative and are not intended to be limiting.

[0035] The term "antigen binding fragment" herein is used in the broadest sense and specifically covers intact monoclonal antibodies, polyclonal antibodies, multispecific antibodies (e.g. bispecific antibodies) formed from at least two intact antibodies, and antibody fragments. [0036] The term "primer," as used herein, is meant to encompass any nucleic acid that is capable of priming the synthesis of a nascent nucleic acid in a template-dependent process. Primers may be oligonucleotides from ten to twenty and/or thirty base pairs in length, but longer sequences can be employed. Primers may be provided in double-stranded and/or single-stranded form, although the single-stranded form is preferred.

II. Risk Scores

[0037] The biomarkers as used herein may be related to cancer prognosis, for example, prediction of survival, recurrence, or therapy response. In a particular embodiment, the differential patterns of expression of a plurality of these biomarkers may be used to predict the survival outcome of a subject with cancer. Certain biomarkers tend to be over-expressed in long-term survivors, whereas other biomarkers tend to be over-expressed in short-term survivors. The unique pattern of expression of a plurality of biomarkers in a subject (i.e., the gene signature) may be used to generate a risk score of survival. Subjects with a high risk score may have a short survival time (e.g., less than about 2 years) after surgical resection. Subjects with a low risk score may have a longer survival time (e.g., more than about 3 years) after resection.

[0038] Regardless of the technique used to measure the differential expression of a plurality of biomarkers, the expression of each biomarker may be converted into an expression value. These expression values then will be used to calculate a risk score of survival for a subject with cancer using statistical methods well known in the art. The risk scores may be calculated using a principal components analysis. The risk scores may also be calculated using a partial Cox regression analysis. In a preferred embodiment, the risk scores may be calculated using a univariate Cox regression analysis.

[0039] The scores generated may be used to classify patients into high or low risk score, wherein a high risk score is associated with a poor prognosis, such as a short survival time or a poorer survival, and a low risk score is associated with a good prognosis, such as a long survival time or a better survival. The cut-off value may be derived from a control group of cancer patients as a median risk score. In further aspects, there may be two or more major prognostic subgroups. Although two-group classification (e.g., high vs. low risk) of cancer prognosis is largely supported by the results of previous studies (Hoshida et al., 2008; Lee et al, 2004a; Lee et al., 2004b; Lee et al., 2006; Woo et al., 2008) it may be contemplated that there are more than two prognostic groups of colon cancer patients, given the genetic heterogeneity of the disease. Because these methods provided herein generate continuous risk scores, it is easy to adjust cutoff criteria to restratify colon cancer patients according to the degree of genetic heterogeneity.

[0040] In addition, the risk score might be developed by incorporating genomic data from surrounding tissues that does not overlap with but is complementary to those from tumor tissues. The risk score may also be combined with other clinical characteristics or demographic information.

[0041] In various embodiments, a tissue sample may be collected from a subject with a cancer, for example, a colon cancer. The collection step may comprise surgical resection. The sample of tissue may be stored in RNAlater or flash frozen, such that RNA may be isolated at a later date. RNA may be isolated from the tissue and used to generate labeled probes for a nucleic acid microarray analysis. The RNA may also be used as a template for qRT-PCR in which the expression of a plurality of biomarkers is analyzed. The expression data generated may be used to derive a risk score, e.g., using the Cox regression classification method to obtain regression coefficients as the weight of each corresponding biomarker gene expression. The risk score may be used to predict whether the subject will be a short-term or a long-term cancer survivor.

[0042] Biomarker genes that may be used in cancer prognosis or risk score generation may be one or more selected from Table 1 below.

III. Cancer and Cancer Samples

[0043] The expression of a plurality of biomarkers may be measured in a sample of cells from a subject with cancer. The type and classification of the cancer can and will vary. The cancer may be an early stage cancer, i.e., stage I or stage II, or it may be a late stage cancer, i.e., stage III or stage IV. In a particular aspect, the cancer may be a cancer of the colon.

[0044] Colon cancer is properly considered to be a cancer which starts in the colon, as opposed to a cancer which originates in another organ and migrates to the colon, known as a colon metastasis. The most frequent colon cancer is adenocarcinoma. Although surgical resection for colon cancer may provide the best chance for cure, the prognosis after surgery differs considerably among patients. Because of this clinical heterogeneity, predicting the recurrence or survival of colon cancer patients after surgical resection or chemotherapy remains challenging.

[0045] The present invention may also be used to predict prognosis, disease-free survival, or overall survival after treatment of a colon cancer. For example, a subject with a colon cancer may be treated via a surgery, a chemotherapeutic (e.g., leucovorin, leucovorin, 5-FU, capecitabine, irinotecan, oxaliplatin, bevacizumab, cetuximab, or panitumumab), or a radiation therapy. Some subjects may require treatment of a metastatic tumor, such as a colon cancer metastasis, during treatment. [0046] In further aspects, the colon cancer may comprise an adenocarcinoma (e.g., a mucinous (colloid) adenocarcinoma or a signet ring adenocarcinoma), a scirrhous tumor, or a neuroendocrine turor. Tumors with neuroendocrine differentiation typically have a poorer prognosis than pure adenocarcinoma variants (Saclarides et al., 1994).

[0047] Other types of cancer that may be evaluated in some aspects of the present invention include, but are not limited to, anal cancer, bladder cancer, bone cancer, brain cancer, breast cancer, cervical cancer, liver cancer, duodenal cancer, endometrial cancer, eye cancer, gallbladder cancer, head and neck cancer, larynx cancer, non-small cell lung cancer, small cell lung cancer, lymphomas, melanoma, mouth cancer, ovarian cancer, pancreatic cancer, penal cancer, prostate cancer, rectal cancer, renal cancer, skin cancer, testicular cancer, thyroid cancer, and vaginal cancer.

[0048] In certain aspects, the sample of cells or tissue sample may be obtained from the subject with cancer by biopsy or surgical resection. The type of biopsy can and will vary, depending upon the location and nature of the cancer. A sample of cells, tissue, or fluid may be removed by needle aspiration biopsy. For this, a fine needle attached to a syringe is inserted through the skin and into the organ or tissue of interest.

[0049] The needle may be guided to the region of interest using ultrasound or computed tomography (CT) imaging. Once the needle is inserted into the tissue, a vacuum is created with the syringe such that cells or fluid may be sucked through the needle and collected in the syringe. A sample of cells or tissue may also be removed by incisional or core biopsy. For this, a cone, a cylinder, or a tiny bit of tissue is removed from the region of interest. CT imaging, ultrasound, or an endoscope is generally used to guide this type of biopsy. Lastly, the entire cancerous lesion may be removed by excisional biopsy or surgical resection.

[0050] Once a sample of cells or sample of tissue is removed from the subject with cancer, it may be processed for the isolation of RNA or protein using techniques well known in the art and disclosed in standard molecular biology reference books, such as Ausubel et al. (2003). A sample of tissue may also be stored in RNAlater (Ambion; Austin TX) or flash frozen and stored at -80 °C for later use. The biopsied tissue sample may also be fixed with a fixative, such as formaldehyde, paraformaldehyde, or acetic acid/ethanol. The fixed tissue sample may be embedded in wax (paraffin) or a plastic resin. The embedded tissue sample (or frozen tissue sample) may be cut into thin sections. RNA or protein may also be extracted from a fixed or wax-embedded tissue sample.

[0051] The subject with cancer may be a mammalian subject. Mammals may include primates, livestock animals, and companion animals. Primates may include humans, New World monkeys, Old World monkeys, gibbons, and great apes. Livestock animals may include horses, cows, goats, sheep, deer (including reindeer), and pigs. Companion animals may include dogs, cats, rabbits, and rodents (including mice, rats, and guinea pigs). In an exemplary embodiment, the subject is a human.

IV. Expression Assessment

[0052] In certain aspects, this invention entails measuring expression of one or more prognostic biomarkers in a sample of cells from a subject with cancer. The expression information may be obtained by testing cancer samples by a lab, a technician, a device, or a clinician.

[0053] The pattern or signature of expression in each cancer sample may then be used to generate a risk score for cancer prognosis or classification, such as predicting cancer survival or recurrence. The level of expression of a biomarker may be increased or decreased in a subject relative to other subjects with cancer. The expression of a biomarker may be higher in long-term survivors than in short-term survivors. Alternatively, the expression of a biomarker may be higher in short-term survivors than in long-term survivors.

[0054] Expression of one or more of biomarkers identified by the inventors may be assessed to predict or report prognosis or prescribe treatment options for cancer patients, especially colon cancer patients.

[0055] The expression of one or more biomarkers may be measured by a variety of techniques that are well known in the art. Quantifying the levels of the messenger RNA (mRNA) of a biomarker may be used to measure the expression of the biomarker. Alternatively, quantifying the levels of the protein product of a biomarker may be used to measure the expression of the biomarker. Additional information regarding the methods discussed below may be found in Ausubel et al. (2003) or Sambrook et al. (1989). As would be recognized by one of skill in the art, various parameters may be manipulated to optimize detection of the mRNA or protein of interest. [0056] A nucleic acid microarray may be used to quantify the differential expression of a plurality of biomarkers. Microarray analysis may be performed using commercially available equipment, following manufacturer's protocols, such as by using the Affymetrix GeneChip® technology (Santa Clara, CA) or the Microarray System from Incyte (Fremont, CA). For example, single-stranded nucleic acids (e.g., cDNAs or oligonucleotides) may be plated, or arrayed, on a microchip substrate. The arrayed sequences are then hybridized with specific nucleic acid probes from the cells of interest. Fluorescently labeled cDNA probes may be generated through incorporation of fluorescently labeled deoxynucleotides by reverse transcription of RNA extracted from the cells of interest. Alternatively, the R A may be amplified by in vitro transcription and labeled with a marker, such as biotin. The labeled probes are then hybridized to the immobilized nucleic acids on the microchip under highly stringent conditions. After stringent washing to remove the non-specifically bound probes, the chip is scanned by confocal laser microscopy or by another detection method, such as a CCD camera. The raw fluorescence intensity data in the hybridization files are generally preprocessed with the robust multichip average (RMA) algorithm to generate expression values.

[0057] Quantitative real-time PCR (qRT-PCR) may also be used to measure the differential expression of a plurality of biomarkers. In qRT-PCR, the RNA template is generally reverse transcribed into cDNA, which is then amplified via a PCR reaction. The amount of PCR product is followed cycle-by-cycle in real time, which allows for determination of the initial concentrations of mRNA. To measure the amount of PCR product, the reaction may be performed in the presence of a fluorescent dye, such as SYBR Green, which binds to double- stranded DNA. The reaction may also be performed with a fluorescent reporter probe that is specific for the DNA being amplified.

[0058] A non-limiting example of a fluorescent reporter probe is a TaqMan® probe (Applied Biosystems, Foster City, CA). The fluorescent reporter probe fluoresces when the quencher is removed during the PCR extension cycle. Multiplex qRT-PCR may be performed by using multiple gene-specific reporter probes, each of which contains a different fluorophore.

Fluorescence values are recorded during each cycle and represent the amount of product amplified to that point in the amplification reaction. To minimize errors and reduce any sample-to-sample variation, qRT-PCR may be performed using a reference standard. The ideal reference standard is expressed at a constant level among different tissues, and is unaffected by the experimental treatment.

[0059] Suitable reference standards include, but are not limited to, mRNAs for the housekeeping genes glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) and β-actin. The level of mRNA in the original sample or the fold change in expression of each biomarker may be determined using calculations well known in the art. [0060] Immunohistochemical staining may also be used to measure the differential expression of a plurality of biomarkers. This method enables the localization of a protein in the cells of a tissue section by interaction of the protein with a specific antibody. For this, the tissue may be fixed in formaldehyde or another suitable fixative, embedded in wax or plastic, and cut into thin sections (from about 0.1 mm to several mm thick) using a microtome. Alternatively, the tissue may be frozen and cut into thin sections using a cryostat. The sections of tissue may be arrayed onto and affixed to a solid surface (i.e., a tissue microarray). The sections of tissue are incubated with a primary antibody against the antigen of interest, followed by washes to remove the unbound antibodies. The primary antibody may be coupled to a detection system, or the primary antibody may be detected with a secondary antibody that is coupled to a detection system. The detection system may be a fluorophore or it may be an enzyme, such as horseradish peroxidase or alkaline phosphatase, which can convert a substrate into a colorimetric, fluorescent, or chemiluminescent product. The stained tissue sections are generally scanned under a microscope. Because a sample of tissue from a subject with cancer may be heterogeneous, i.e., some cells may be normal and other cells may be cancerous, the percentage of positively stained cells in the tissue may be determined. This measurement, along with a quantification of the intensity of staining, may be used to generate an expression value for the biomarker.

[0061] An enzyme-linked immunosorbent assay, or ELISA, may be used to measure the differential expression of a plurality of biomarkers. There are many variations of an ELISA assay. All are based on the immobilization of an antigen or antibody on a solid surface, generally a microtiter plate. The original ELISA method comprises preparing a sample containing the biomarker proteins of interest, coating the wells of a microtiter plate with the sample, incubating each well with a primary antibody that recognizes a specific antigen, washing away the unbound antibody, and then detecting the antibody-antigen complexes. The antibody-antibody complexes may be detected directly. For this, the primary antibodies are conjugated to a detection system, such as an enzyme that produces a detectable product. The antibody-antibody complexes may be detected indirectly. For this, the primary antibody is detected by a secondary antibody that is conjugated to a detection system, as described above. The microtiter plate is then scanned and the raw intensity data may be converted into expression values using means known in the art. [0062] An antibody microarray may also be used to measure the differential expression of a plurality of biomarkers. For this, a plurality of antibodies is arrayed and covalently attached to the surface of the microarray or biochip. A protein extract containing the biomarker proteins of interest is generally labeled with a fluorescent dye. [0063] The labeled biomarker proteins are incubated with the antibody microarray. After washes to remove the unbound proteins, the microarray is scanned. The raw fluorescent intensity data may be converted into expression values using means known in the art.

[0064] Luminex multiplexing microspheres may also be used to measure the differential expression of a plurality of biomarkers. These microscopic polystyrene beads are internally color-coded with fluorescent dyes, such that each bead has a unique spectral signature (of which there are up to 100). Beads with the same signature are tagged with a specific oligonucleotide or specific antibody that will bind the target of interest (i.e., biomarker mRNA or protein, respectively). The target, in turn, is also tagged with a fluorescent reporter. Hence, there are two sources of color, one from the bead and the other from the reporter molecule on the target. The beads are then incubated with the sample containing the targets, of which up 100 may be detected in one well. The small size/surface area of the beads and the three dimensional exposure of the beads to the targets allows for nearly solution-phase kinetics during the binding reaction. The captured targets are detected by high-tech fluidics based upon flow cytometry in which lasers excite the internal dyes that identify each bead and also any reporter dye captured during the assay. The data from the acquisition files may be converted into expression values using means known in the art.

[0065] In situ hybridization may also be used to measure the differential expression of a plurality of biomarkers. This method permits the localization of mRNAs of interest in the cells of a tissue section. For this method, the tissue may be frozen, or fixed and embedded, and then cut into thin sections, which are arrayed and affixed on a solid surface. The tissue sections are incubated with a labeled antisense probe that will hybridize with an mRNA of interest. The hybridization and washing steps are generally performed under highly stringent conditions. The probe may be labeled with a fluorophore or a small tag (such as biotin or digoxigenin) that may be detected by another protein or antibody, such that the labeled hybrid may be detected and visualized under a microscope. Multiple mRNAs may be detected simultaneously, provided each antisense probe has a distinguishable label. The hybridized tissue array is generally scanned under a microscope. Because a sample of tissue from a subject with cancer may be heterogeneous, i.e., some cells may be normal and other cells may be cancerous, the percentage of positively stained cells in the tissue may be determined. This measurement, along with a quantification of the intensity of staining, may be used to generate an expression value for each biomarker.

[0066] The number of biomarkers whose expression is measured in a sample of cells from a subject with cancer may vary. Since the risk score is based upon the differential expression of the biomarkers, a higher degree of accuracy should be attained when the expression of more biomarkers is measured; however, a large number of biomarkers in the gene signature would hamper the clinical usefulness. In a certain embodiment, the differential expression of a selected number of biomarkers may be measured.

V. Statistical and Informatical Methods

[0067] In certain aspects of the invention, expression information of the biomarkers may be analyzed by statistical and informatical methods to help provide prognosis prediction and treatment prescription. Those methods may comprise processing the test data stored on a data storage device by using a tangible computer readable medium having computer usable program code executable to perform operations for the statistic analysis and prediction/prescription output, or for assisting risk score generation as described above.

A. Kaplan-Meier method

[0068] The Kaplan-Meier method (also known as the product limit estimator) estimates the survival function from life-time data. In medical research, it might be used to measure the fraction of patients living for a certain amount of time after treatment.

[0069] A plot of the Kaplan-Meier method of the survival function is a series of horizontal steps of declining magnitude which, when a large enough sample is taken, approaches the true survival function for that population. The value of the survival function between successive distinct sampled observations ("clicks") is assumed to be constant.

[0070] An important advantage of the Kaplan-Meier curve is that the method can take into account "censored" data— losses from the sample before the final outcome is observed (for instance, if a patient withdraws from a study). On the plot, small vertical tick-marks indicate losses, where patient data has been censored. When no truncation or censoring occurs, the Kaplan-Meier curve is equivalent to the empirical distribution. [0071] A method might involve grouping patients into categories, for instance, those with Gene A profile and those with Gene B profile. In the graph, patients with Gene B die much more quickly than those with gene A. After two years, about 80% of the Gene A patients still survive, but less than half of Gene B patients still survive. B. Log-rank test

[0072] In statistics, the log-rank test (sometimes called the Mantel-Cox test) is a hypothesis test to compare the survival distributions of two samples. It is a nonparametric test and appropriate to use when the data are right censored (technically, the censoring must be non- informative). It is widely used in clinical trials to establish the efficacy of new drugs compared to a control group (often a placebo) when the measurement is the time to event (such as a heart attack).

[0073] The log-rank test statistic compares estimates of the hazard functions of the two groups at each observed event time. It is constructed by computing the observed and expected number of events in one of the groups at each observed event time and then adding these to obtain an overall summary across all time points where there is an event.

[0074] The log-rank statistic can be derived as the score test for the Cox proportional hazards model comparing two groups. It is therefore asymptotically equivalent to the likelihood ratio test statistic based from that model.

C. Proportional hazards analysis

[0075] Proportional hazards models are a sub-class of survival models in statistics. For the purposes of simplification, consider survival models to consist of two parts: the underlying hazard function, describing how hazard (risk) changes over time, and the effect parameters, describing how hazard relates to other factors - such as the choice of treatment, in a medical example. The proportional hazards assumption is the assumption that effect parameters multiply hazard: for example, if taking drug X halves a hazard at time 0, it also halves the hazard at time 1, or time 0.5, or time t for any value of t. The effect parameter(s) estimated by any proportional hazards model can be reported as hazard ratios.

[0076] It was observed that if the proportional hazards assumption holds (or, is assumed to hold) then it is possible to estimate the effect parameter(s) without any consideration of the hazard function. This approach to survival data is called application of the Cox proportional hazards model, sometimes abbreviated to Cox model or to proportional hazards model.

[0077] Other proportional hazards models exist. Another approach to survival data is to assume that the proportional hazards assumption holds, but in addition to assume that the hazard function follows a known form. For example, assuming the hazard function to be the Weibull hazard function gives the Weibull proportional hazards model (in which the survival times follow a Weibull distribution).

D. Hierarchical clustering

[0078] Clustering is the assignment of objects into groups (called clusters) so that objects from the same cluster are more similar to each other than objects from different clusters. Often similarity is assessed according to a distance measure. Clustering is a common technique for statistical data analysis, which is used in many fields, including machine learning, data mining, pattern recognition, image analysis, and bioinformatics.

[0079] Besides the term data clustering (or just clustering), there are a number of terms with similar meanings, including cluster analysis, automatic classification, numerical taxonomy, botryology, and typological analysis.

[0080] Hierarchical clustering builds (agglomerative), or breaks up (divisive), a hierarchy of clusters. The traditional representation of this hierarchy is a tree (called a dendrogram), with individual elements at one end and a single cluster containing every element at the other. Agglomerative algorithms begin at the leaves of the tree, whereas divisive algorithms begin at the root. The former method builds the hierarchy from the individual elements by progressively merging clusters.

[0081] In transcriptomics, clustering is used to build groups of genes with related expression patterns (also known as coexpressed genes). Often such groups contain functionally related proteins, such as enzymes for a specific pathway, or genes that are co-regulated. High throughput experiments using expressed sequence tags (ESTs) or DNA microarrays can be a powerful tool for genome annotation, a general aspect of genomics.

VI. Cancer Treatments

[0082] Biomarkers and a new "risk score" system that can predict the likelihood of recurrence or overall survival in colon cancer patients, as disclosed herein, may be used to identify patients who can benefit from a conventional single or combined modality therapy, prior to the treatment of the cancer. Further, methods provided herein may be used to identify patients who may not substantially benefit from a conventional single or combined modality therapy; such patients may benefit more from alternative treatment(s). [0083] In certain aspects of the present invention, a conventional cancer therapy may be administered to a subject wherein the subject is identified or reported as having a good prognosis based on the assessment of a group of biomarkers as described herein. An alternative cancer therapy may be administered to a subject, e.g., alone or in combination with a conventional cancer therapy, if the subject is identified as having a poor prognosis via a method or kit disclosed herein.

[0084] Conventional cancer therapies include one or more therapy, such as a chemotherapy, radiotherapy, and/or a surgery. For example, a subject with a colon cancer may be treated via a surgery, a chemotherapeutic, e.g., leucovorin, leucovorin, 5-FU, capecitabine, irinotecan, oxaliplatin, bevacizumab, cetuximab, or panitumumab, or one or more combination of the foregoing. Chemotherapies include, for example, cisplatin (CDDP), carboplatin, procarbazine, mechlorethamine, cyclophosphamide, camptothecin, ifosfamide, melphalan, chlorambucil, busulfan, nitrosurea, dactinomycin, daunorubicin, doxorubicin, bleomycin, plicomycin, mitomycin, etoposide (VP 16), tamoxifen, raloxifene, estrogen receptor binding agents, taxol, gemcitabien, navelbine, farnesyl-protein tansferase inhibitors, transplatinum, 5- fluorouracil, vincristin, vinblastine, and methotrexate, or any analog or derivative variant of the foregoing.

[0085] Radiation therapies generally cause DNA damage and have been used extensively. A radiation therapy may include administrationof γ-rays, X-rays, and/or the directed delivery of a radioisotope to tumor cells. Other forms of DNA damaging factors are also contemplated, such as microwaves and UV-irradiation. These factors may effect a broad range of damage on DNA, on the precursors of DNA, on the replication and repair of DNA, and on the assembly and maintenance of chromosomes. Dosage ranges for X-rays range from daily doses of 50 to 200 roentgens for prolonged periods of time (3 to 4 wk), to single doses of 2000 to 6000 roentgens. Dosage ranges for radioisotopes vary widely, and depend on the half-life of the isotope, the strength and type of radiation emitted, and the uptake by the neoplastic cells. [0086] The terms "contacted" and "exposed," when applied to a cell, are used herein to describe the process by which a therapeutic construct and a chemotherapeutic or radiotherapeutic agent are delivered to a target cell or are placed in direct juxtaposition with the target cell. To achieve cell killing or stasis, both agents are delivered to a cell in a combined amount effective to kill the cell or prevent it from dividing.

[0087] Approximately 60% of persons with cancer will undergo surgery of some type, which includes preventative, diagnostic or staging, curative, and palliative surgery. Curative surgery is a cancer treatment that may be used in conjunction with other therapies, such as a chemotherapy, radiotherapy, hormonal therapy, gene therapy, immunotherapy, and/or alternative therapies.

[0088] Curative surgery includes resection in which all or part of cancerous tissue is physically removed, excised, and/or destroyed. Tumor resection refers to physical removal of at least part of a tumor. In addition to tumor resection, treatment by surgery includes laser surgery, cryosurgery, electrosurgery, and microscopically-controlled surgery (Mohs' surgery). It is further contemplated that the present invention may be used in conjunction with removal of superficial cancers, precancers, or incidental amounts of normal tissue.

[0089] Laser therapy is the use of high-intensity light to destroy tumor cells. Laser therapy affects the cells only in the treated area. Laser therapy may be used to destroy cancerous tissue and relieve a blockage in the esophagus when the cancer cannot be removed by surgery. The relief of a blockage can help to reduce symptoms, especially swallowing problems. Photodynamic therapy (PDT), a type of laser therapy, involves the use of drugs that are absorbed by cancer cells; when exposed to a special light, the drugs become active and destroy the cancer cells.

[0090] Upon excision of part of all of cancerous cells, tissue, or tumor, a cavity may be formed in the body. Treatment may be accomplished by perfusion, direct injection or local application of the area with an additional anti-cancer therapy. Such treatment may be repeated, for example, every 1, 2, 3, 4, 5, 6, or 7 days, or every 1, 2, 3, 4, and 5 weeks or every 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months. These treatments may be of varying dosages as well. [0091] Alternative cancer therapies generally include any cancer therapy other than surgery, chemotherapy and radiation therapy. Alternative cancer therapies include immunotherapy, gene therapy, hormonal therapy or a combination thereof. Subjects identified with a poor prognosis as described herein may not respond favorably to conventional treatment(s) alone and may be prescribed or administered one or more alternative cancer therapies alone, or in combination with one or more conventional treatments. [0092] Immunotherapeutics, generally, rely on the use of immune effector cells and molecules to target and destroy cancer cells. The immune effector may be, for example, an antibody specific for some marker on the surface of a tumor cell. The antibody alone may serve as an effector of therapy or it may recruit other cells to actually affect cell killing. The antibody also may be conjugated to a drug or toxin (chemotherapeutic, radionuclide, ricin A chain, cholera toxin, pertussis toxin, etc.) and serve merely as a targeting agent. Alternatively, the effector may be a lymphocyte carrying a surface molecule that interacts, either directly or indirectly, with a tumor cell target. Various effector cells include cytotoxic T cells and NK cells.

[0093] Gene therapy generally involves the insertion of polynucleotides, including DNA or RNA, into an individual's cells and tissues to treat a disease. Antisense therapy is also a form of gene therapy. A therapeutic polynucleotide may be administered before, after, or at the same time as a first cancer therapy. Delivery of a vector encoding a variety of proteins is encompassed within the invention. For example, cellular expression of the exogenous tumor suppressor oncogenes may exert their function to inhibit excessive cellular proliferation, such as p53, pi 6, and C-CAM.

[0094] Additional agents to be used to improve the therapeutic efficacy of treatment include immunomodulatory agents, agents that affect the upregulation of cell surface receptors and GAP junctions, cytostatic and differentiation agents, inhibitors of cell adhesion, or agents that increase the sensitivity of the hyperproliferative cells to apoptotic inducers. Immunomodulatory agents include tumor necrosis factor; interferon alpha, beta, and gamma; IL-2 and other cytokines; F42K and other cytokine analogs; or MIP-1, MIP-lbeta, MCP-1, RANTES, and other chemokines. It is further contemplated that the upregulation of cell surface receptors or their ligands, such as Fas / Fas ligand, DR4 or DR5 / TRAIL, would potentiate the apoptotic inducing abilities of the present invention by establishment of an autocrine or paracrine effect on hyperproliferative cells. Increases intercellular signaling by elevating the number of GAP junctions would increase the anti-hyperproliferative effects on the neighboring hyperproliferative cell population. In other embodiments, cytostatic or differentiation agents can be used in combination with the present invention to improve the anti-hyperproliferative efficacy of the treatments. Inhibitors of cell adhesion are contemplated to improve the efficacy of the present invention. Examples of cell adhesion inhibitors are focal adhesion kinase (FAKs) inhibitors and Lovastatin. It is further contemplated that other agents that increase the sensitivity of a hyperproliferative cell to apoptosis, such as the antibody c225, could be used in combination with the present invention to improve the treatment efficacy.

[0095] Hormonal therapy may also be used in the present invention or in combination with any other cancer therapy previously described. The use of hormones may be employed in the treatment of certain cancers, such as breast, prostate, ovarian, or cervical cancer, to lower the level or block the effects of certain hormones such as testosterone or estrogen. This treatment is often used in combination with at least one other cancer therapy as a treatment option or to reduce the risk of metastases.

VII. Kits and Nucleic Acid Arrays

[0096] The present invention also encompasses kits for performing the diagnostic and prognostic methods of the invention. Such kits can be prepared from readily available materials and reagents. For example, such kits can comprise any one or more of the following materials: enzymes, reaction tubes, buffers, detergent, primers, and probes. In a preferred embodiment, these kits allow a practitioner to obtain samples of neoplastic cells in blood, tears, semen, saliva, urine, tissue, serum, stool, sputum, cerebrospinal fluid, and supernatant from cell lysate. In another preferred embodiment these kits include the needed apparatus for performing R A extraction, RT-PCR, and gel electrophoresis. Instructions for performing the assays can also be included in the kits.

[0097] In a particular aspect, these kits may comprise a plurality of agents for assessing the differential expression of a plurality of biomarkers, wherein the kit is housed in a container. The kits may further comprise instructions for using the kit for assessing expression, means for converting the expression data into expression values and/or means for analyzing the expression values to generate scores that predict survival or prognosis. The agents in the kit for measuring biomarker expression may comprise a plurality of PCR probes and/or primers for qRT-PCR and/or a plurality of antibody or fragments thereof for assessing expression of the biomarkers. In another embodiment, the agents in the kit for measuring biomarker expression may comprise an array of polynucleotides complementary to the mRNAs of the biomarkers of the invention. Possible means for converting the expression data into expression values and for analyzing the expression values to generate scores that predict survival or prognosis may be also included.

VIII. Examples

[0098] The following examples are included to demonstrate preferred embodiments of the invention. It should be appreciated by those of skill in the art that the techniques disclosed in the examples which follow represent techniques discovered by the inventor to function well in the practice of the invention, and thus can be considered to constitute preferred modes for its practice. However, those of skill in the art should, in light of the present disclosure, appreciate that many changes can be made in the specific embodiments which are disclosed and still obtain a like or similar result without departing from the spirit and scope of the invention.

Example 1 - Methods

Patients and gene expression data

[0099] All clinical and gene expression data are available from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (www.ncbi.nlm.nih.gov/geo). Gene expression data from Moffit Cancer Center (Moffit cohort, GSE17536, n = 177) were used as the exploration data set (Smith et al., 2010). Gene expression data from Vanderbilt Medical Center (GSE17537, n = 55) and Max Planck Institute (GSE12945, n = 62) were pooled and used as the first validation data set (VMP cohort, n = 117) (Smith et al., 2010; Staub et al., 2009). Gene expression data from Royal Melbourne Hospital that is part of GSE14333 (n = 96) were used as the second validation data set (Jorissen et al., 2009). Gene expression data of these patients were re-deposited as independent data set to GEO (GSE29971). To test the prognostic significance of gene expression signatures, only gene expression data with patient survival data available was used. Although 3 prognostic variables (OS, disease-specific survival, and disease-free survival [DFS]) were available for the Moffit cohort, only OS and DFS data were available for the VMP and Melbourne cohorts, respectively.

[0100] Adjuvant chemotherapy data were available only for the Moffit, Vanderbilt Medical Center, and Melbourne cohorts. Of the 328 patients in the Moffit, Vanderbilt, and Melbourne cohorts, 147 (2 in AJCC stage I, 28 in stage II, 81 in stage III, 36 in stage IV) had received standard adjuvant chemotherapy (either single-treatment 5-fluorouracil/capecitabine or combination of 5-fluorouracil and oxaliplatin). The remaining patients did not receive chemotherapy (n = 168) or treatment data were not available (n = 13). DFS was defined in a previous study as the time from surgery to the first confirmed relapse, and censored when a patient died or was alive without recurrence at last contact (Jorissen et al., 2009).

Statistical analysis of microarray data

[0101] BRB-ArrayTools (linus.nci.nih.gov/BRB-ArrayTools.html) were used primarily for statistical analysis of gene expression data (Simon et ah, 2007), and all other statistical analyses were performed in the R language environment (www.r-project.org). All gene expression data were generated by using the Affymetrix U133 version 2.0 platform except for the Max Planck Institute cohort. Data from the Max Planck Institute were generated by using the Affymetrix U133A platform. Raw data were downloaded from public databases and normalized using a robust multi-array averaging method (Irizarry et al., 2003). Genes that were differentially expressed among the 2 classes were identified using a random-variance t- test. Differences of gene expression between 2 classes were considered statistically significant if their P value was less than 0.001. A stringent significance threshold was used to limit the number of false-positive findings. A global test of whether the expression profiles differed between the classes was performed by permuting the labels of which arrays corresponded to which classes. For each permutation, the P values were re-computed and the number of genes significant at the 0.001 level was noted. The proportion of the permutations that gave at least as many significant genes as with the actual data was the significance level of the global test. Cluster analysis was performed with Cluster and Treeview (Eisen et al., 1998).

[0102] To predict the class of the independent patient cohort, a previously developed model using different algorithms was used (Lee et al., 2006; Lee et al., 2004a; Lee et al., 2004b). Gene expression data from different cohorts was independently centralized by subtracting mean expression values across samples before pooling them together for building prediction models. Then, gene expression data in the training set (Moffit cohort) were combined to form a classifier according to a given algorithm (compound covariate predictor [CCP] (Radmacher et al., 2002), linear discriminant analysis [LDA] (Dudoit et al., 2002), or Bayesian compound covariate predictor [BCCP]) (Wright et ah, 2003). The robustness of the classifier was estimated by misclassification rate determined during the leave-one-out cross-validation (LOOCV) in the training set. For each prediction, training of the classifier was done independently and the misclassification rate was calculated during each training. When applied to the independent validation sets (VMP and Melbourne cohorts), prognostic significance was estimated by Kaplan-Meier plots and log-rank tests between 2 predicted subgroups of patients. After LOOCV, sensitivity and specificity of prediction models were estimated by the fraction of samples correctly predicted. Multivariate Cox proportional hazard regression analysis was used to evaluate independent prognostic factors associated with survival, and gene signature, tumor stage, and pathologic characteristics were used as covariates. Cox proportional hazard regression model was also used to analyze interaction between subgroups and adjuvant chemotherapy treatment. A P value of less than 0.05 was considered to indicate statistical significance, and all tests were 2-tailed.

Gene set and gene network analysis

[0103] Ingenuity™ Pathways Analysis (IPA, Ingenuity Systems®, Redwood City, CA) was used for gene set enrichment analysis and gene network analysis. Gene set enrichment analysis was carried out to identify the most significant gene sets associated with disease process, molecular and cellular functions, and normal physiological and development condition in 114 prognostic genes as described in instruction from Ingenuity Systems. The significance of over-represented gene sets was estimated by the right-tailed Fisher's Exact Test. Gene network analysis was carried out by using a global molecular network developed from information contained in the Ingenuity™ Knowledge Base. Seven hundred fifty-five gene features were mapped to the Ingenuity Knowledge Base. Identified gene networks were ranked according to scores provided by IPA. The score is the likelihood of a set of genes being found in the networks due to random chance. For example, a score of 3 indicates that there is a 1/1000 chance that the focus genes are in a network due to random chance. [0104] The following paragraphs describe the calculations used in BRB-ArrayTools for predictive classification using the compound covariate predictor (CCP), diagonal linear discriminant analysis, and the Bayesian CCP. Cross-validation or bootstrap re-sampling is performed in order to provide a proper estimate of the prediction accuracy of the classifiers. A completely specified classifier is developed on the training set and used to classify the cases in the test set. The paragraphs below describe the classifier development algorithms that are applied in each training set. Compound covariate predictor (CCP)

[0105] The CCP was introduced by MD Radmacher, LM McShane, and R Simon (A paradigm for class prediction using gene expression profiles, Radmacher et al., 2002). Let t_j denote the t statistic for gene j for comparing class 1 to class 2 in the training set.

where denotes the mean of log expression j for training samples in class m (m = 1,2), S_j

is an estimate of the intra-class variance for gene j, and n_m is the number of cases in the training set in class m (m = 1,2). Let X_ij denote the log expression for gene j in sample i of the training set. For each sample of the training set we compute the compound covariate value

Let denote the mean of the C_i values for samples in class 1 in the training set and let denote the mean for samples in class 2 of the training set. [0106] To classify a case not in the training set with log expression values (y₁, y₂, ..., y_p), one computes the compound covariate

where Selected denotes the pre-defined prognostic genes for that training set. If C is closer to than to then the case is predicted to be class 1 and the reverse if it is closer to .

This is using as the threshold of classification.

[0107] For each loop of the leave-one-out-cross-validation (or each bootstrap re-sampling), the training set changes, the t_j values change, and the threshold of classification changes. Diagonal Linear Discriminant Analysis (DLDA)

[0108] The DLDA predictor is similar to the CCP but the weight for the importance of gene j which is t_j for the CCP replaced by t_j/s_j where is the pooled estimate of intra-class variance

for gene j as described for (1). The predictive index C_i is computed for all of the training set samples i

and then the class me

ans in the training set and

are computed from these values.

[0109] To classify a case not in the training set with log expression values (yi, y₂, ..., y_p), computes the predictive index

If C is closer to

than to

then the case is predicted to be class 1 and the reverse if it is closer to . This is using as the threshold of classification. For further

description of the relationship between the CCP and DLDA classifiers, see Simon et al., Chapter 8 of Design and Analysis of DNA Microarray Investigations, 2003, available on-line at brb.nci.nih.gov.

Bayesian CCP [0110] For a given training set one computes the C_i compound covariate values for all training set samples i and the class means and as described above for the CCP. One

also computes the pooled (between classes) estimator V of the variance of C_i values within each class of the training set.

[0111] To classify a case not in the training set with log expression values Y = (yi, y₂, y_p), one computes the compound covariate value C using expression (3). The probability that the sample Y came from class 1 is estimated as

where denotes the Gaussian density of a value C when the mean is and the

variance is V. π₁ denotes the prior probability that the class is 1. The user has the option of using prior probability of 0.5, or of using the prevalence of class 1 samples in the training set to determine the prior probability of class 1. The probability that the new sample Y is from class 2 is 1 minus the value given in (4). The class with the greater computed probability is predicted for the new sample. The Bayesian Compound Covariate method was developed by GW Wright et al. (A gene expression-based method to diagnose clinically distinct subgroups of diffuse large B cell lymphoma, Wright et al., 2003). The implementation in BRB- ArrayTools uses a pooled-sample variance estimate V in order to improve stability of classification when the number of samples is small.

Example 2 - Prognostic Gene Expression Signature Associated with Two Molecularly

Distinct Subtypes of Colorectal Cancer

Significant Association of Prognosis with 2 Subgroups Found by Hierarchical Clustering

[0112] To uncover potential subgroups of colorectal cancer, applied hierarchical clustering analysis was applied to gene expression data (Table 2, Moffit cohort, n = 177) and found 2 major subgroups of colorectal cancer. Hierarchical clustering of gene expression data of 177 colon cancer tissues from Moffit Cancer Center. Genes with an expression level that had at least a 2-fold difference relative to median value across tissues in at least 10 tissues were selected for hierarchical clustering analysis (6,630 gene features). The data are presented in matrix format in which rows represent individual genes and columns represent each tissue. Each cell in the matrix represents the expression level of a gene feature in an individual tissue. The red and green color in cells reflect relative high and low expression levels, respectively, as indicated in the scale bar (log2 transformed scale). [0113] When association of the 2 subgroups with clinical variables were examined, significant association with the 2 subgroups was apparent only for patient survival (Table 3, 53% vs. 29%, P = 0.0007, by X²-test). As expected, Kaplan-Meier plots and the log-rank test showed significant differences in all prognostic variables, including OS and DFS (P = 5.6 x 10^-4 and P = 0.01, respectively, by log-rank test; FIG. 1).

[0114] A limited number of genes were next identified whose expression is tightly associated with the 2 subgroups. By applying a stringent threshold cut-off (P < 0.001 and 2-fold difference), 114 gene features were identified (Table 4). Genes were selected by univariate test (two-sample i-test) with multivariate permutation test and stringent cut-off (P < 0.001 and 2-fold difference) was applied to retain genes whose expression is significantly different between the two groups of tissues examined (114 genes). The data are presented in matrix format in which rows represent individual genes and columns represent each tissue. Each cell in the matrix represents the expression level of a gene feature in an individual tissue. The red and green color in cells reflect relative high and low expression levels, respectively, as indicated in the scale bar (log2 transformed scale). [0115] Interestingly, expressions of SPP1 and POSTN, previously reported as metastasis genes and associated with poor prognosis in colon cancer (Bao et al., 2004; Yeatman and Chambers, 2003), were much higher in the poor prognosis subgroup B. To uncover biological characteristics enriched in 114 genes, gene set enrichment analysis was carried out with Ingenuity Pathway Analysis™. As expectedly, it revealed enrichment of genes whose function is highly associated with cancer, cell growth and proliferation, and tumor morphology (FIG. 5), indicating that selected genes might well reflect biological characteristics of the 2 subgroups of colon cancer.

The prediction rule is defined by the inner sum of the weights (w_i) and expression (x_i) of genes. A sample is classified to the class A if the sum is greater than the threshold; that is, ∑_iw_i x_i > threshold. The threshold for the Compound Covariate predictor is -90.282

.

Validation of the Prognostic Gene Expression Signature in Independent Cohorts

[0116] Having in hand a distinct gene expression signature (114 genes) that well reflects the prognosis of colorectal cancer patients, the inventors next sought to validate the association of the signature with prognosis in independent patient cohorts. For this validation, gene expression data from Vanderbilt Medical Center (n = 55) and Max Planck Institute (n = 62) were pooled (Vanderbilt and Max Planck [VMP] cohort, n = 117), and previously established data training and prediction methods were applied and gene expression data from Moffit cohort were used for training of classifiers. When colon cancer patients in the VMP cohort were stratified according to a cluster-specific gene expression signature, Kaplan-Meier plots showed significant differences (P = 0.032 by log-rank test) in OS of patients between the 2 subtypes that were predicted by compound covariate predictor (CCP) (FIGS. 2A-B). Specificity and sensitivity for correctly predicting subtype B in test set (Moffit cohort) during LOOCV were 0.826 and 0.868 respectively. When CCP was replaced by different prediction algorithms such as linear discriminant analysis (LDA) and Bayesian compound corvariate predictor (BCCP), the prognostic difference between the 2 subgroups remained significant (P = 0.04 and P = 0.005 by log-rank test for LDA and BCCP, respectively).

[0117] Subset analysis was next performed with only patients with stage II or III cancer. In subset analysis, the gene expression signature successfully identified poor survival patients among both stage II and III patients (FIGS. 6A-B), supporting that the gene expression signature is independent of the current staging systems. To evaluate the prognostic value of the gene expression signature in combination with other clinical variables including patient age at diagnosis, AJCC stage, sex, and grade, univariate and multivariate Cox proportional hazards regression analyses were undertaken in the VMP cohort. In the univariate analysis, the gene signature and AJCC stage were significantly associated with OS (P = 0.036 and P = 1.16 x 10^-7, respectively). In the multivariate analysis, AJCC stage and gene signature still maintained the significance (P = 1.59 x 10^"5 and P = 0.008), and the grade showed marginal significance (P = 0.04) (Table 5).

[0118] To further test the robustness of the gene expression signature, it was applied to a second validation cohort (Melbourne cohort, n = 96). Consistent with the results for the 2 previous cohorts, this 114-gene expression signature distinguished patients with poor prognosis from those with a better prognosis (FIGS. 7A-B). In the univariate analysis, the gene signature and AJCC stage were significantly associated with DFS (P = 0.048 and P = 0.002 respectively) and significance remains similar in the multivariate analysis (P = 0.049 and 0.002 respectively) (Table 6).

Significant Association of the Signature with Disease-free Survival of Patients after Adjuvant Chemotherapy

[0119] Adjuvant chemotherapy data were available for 328 of the 390 patients from the 3 cohorts. The inventors next determined the association of the new prognostic gene expression signature with response to chemotherapy. As expected, the difference in DFS between the 2 subtypes remained significant in the combined subset of patients (3-year rate, 88.2% [A] vs. 73.7% [B]; P = 6.6 x 10^"4 by log-rank test, FIG. 3A) after excluding stage IV patients. The hazard ratio (HR) of subtype B for relapse was 2.6 (95% CI, 1.47-4.57; P = 0.001). To examine the association of the signature with response to adjuvant chemotherapy, subset analysis was performed with patients in AJCC stage III (n = 111), a stage for which the benefit of adjuvant chemotherapy has been well established (Laurie et al., 1989; Moertel et al, 1990; Moertel et al, 1995). Patients with stage III disease were subdivided into 2 subtypes (A or B), and the difference in DFS was independently assessed. Adjuvant chemotherapy significantly affected DFS in patients in subtype B (3-year rate, 70.1% [CTX] vs. 39.7% [no CTX]; P = 0.004 by log-rank test, FIG. 3C). However, such benefit of adjuvant chemotherapy was not significant for patients in subtype A (3-year rate, 85.9% [CTX] vs. 76.2% [no CTX]; P = 0.5 by log-rank test, FIG. 3D). When Cox regression model was applied, the interaction of subtypes with adjuvant chemotherapy reached significance level of 0.36 (FIG. 8). However, consistent with Kaplan-Meier plot and log-rank test, the estimated HR for adjuvant chemotherapy in subgroup B was 0.31 (95% CI, 0.13-0.73; P = 0.007), while HR for relapse for adjuvant chemotherapy in subtype A was 0.65 (95% CI, 0.18-2.28; = 0.5).

Biological Insight of Prognostic and Predictive Signature

[0120] Although the 114-gene expression signature was robust enough to discriminate between patients in all 3 different cohorts, the number of genes was too small to develop a gene network analysis, because an extremely stringent cut-off (P < 0.001 and 2-fold) was applied to avoid any potential false-positive problem during signature-based prediction. Thus, to explore biological characteristics of subgroups with poorer prognosis (subtype B patients), genes were identified whose expression patterns were conserved in all 3 cohorts. To maximize the compatibility of the 3 data sets, the inventors included only gene expression data generated by using the Affymetrix U133 v2.0. Gene lists X, Y, and Z (FIG. 4A) represent the 755 genes that were differentially expressed between A and B subgroups in all three cohorts.

[0121] Pathway analysis was next performed on the 755 genes in the extended gene list (FIG. 4B) using the Ingenuity™ Pathway Analysis tool (Table 7). This analysis revealed a series of putative networks. Functional connectivity of the top network revealed a strong over- representation of TGF-β pathways, suggesting that its activation might be a key genetic determinant associated with poorer survival of colon cancer patients in subtype B. In addition, genes involved in activation of NF-kappa B pathway, a key survival pathway in many cancers, are over-expressed in subtype B colon cancer, suggesting that activation of NF-K B might be, in part, accountable for poorer survival of patients in subtype B. Interestingly, two of genes in the NF-κ B network are FYN and LYN, SRC tyrosine kinase family, whose activity are frequently increased and well associated with metastasis potential and poorer outcome in colon cancer (Yeatman, 2004; Kopetz et al., 2007).

[0122] By analyzing gene expression data from colorectal cancer tissues, a limited number of genes (114 genes) were identified that are significantly associated with prognosis. Robustness of the signature was validated in 2 independent cohorts. Subset analysis with only stage III patients indicated that the signature might be associated with benefit from adjuvant chemotherapy. Current staging systems and biomarkers are very limited in providing therapeutic guidance to patients with colorectal cancer. This gene expression signature may be used to develop molecular stratification of patients and provide treatment guidance.

[0123] Two independent but complimentary methods were applied to construct the signature, test its robustness, and validate its association with clinical outcomes. In the first approach, unsupervised clustering were used to identify subgroups that differed with respect to gene expression patterns. Subsequent analysis with clinical data revealed that the 2 subgroups differed significantly in OS and DFS, indicating that biological characteristics reflected in gene expression patterns may well represent clinical heterogeneity that has not been properly addressed in the current staging systems. In a second approach, supervised prediction models were applied to validate the association of the signature with clinical outcomes in 2 independent patient cohorts. The robustness of the 114-gene signature that was studied was supported by the high sensitivity (>0.8) and specificity (>0.8) during training of prediction models within the Moffit cohort and a significant association of predicted outcome with patient prognosis in both test cohorts (FIGS. 2A-B and FIGS. 7A-B). [0124] Subset analysis of patients with available chemotherapy data suggested that the 114- gene signature might be able to predict which patients would have benefit from adjuvant chemotherapy. In patients with stage III disease, 5-fluorouracil-based chemotherapy was significantly associated with improved outcome for patients in subtype B (HR, 0.31; 95% CI, 0.13-0.73; P = 0.007), whereas its benefit was not statistically significant for patients in subtype A. Thus, this newly identified gene signature showed a mixed prognostic and predictive association in the current study. This mixed association is in good agreement with previous findings in breast cancer showing that a strong prognostic recurrence score based on the Oncotype DX™ assay is highly predictive for response to chemotherapy as well (Paik et al, 2004; Sparano and Paik, 2008). [0125] Although an overall survival benefit for 5-fluorouracil-based adjuvant chemotherapy has been established for patients with AJCC stage III cancer (Laurie et al., 1989; Moertel et al, 1990; Wolmark et al., 1993), the use of adjuvant chemotherapy remains controversial for patients with AJCC stage II disease. The Quick And Simple And Reliable (QUASAR) trial reported a modest but statistically significant improvement in DFS in stage II patients treated with 5-fluorouracil-based adjuvant chemotherapy compared with observation (Quasar Collaborative Group, 2007). However, no additional trials have been performed appropriately to validate it in stage II patients, owing in part to the large sample size that would be required. It has been suggested that at least 9,680 patients per group would be required to detect a 2% survival difference between the treatment and control arms, with 90% power and a 0.05 significance level (Benson et al., 2004). The findings of QUASAR and other pooled analyses suggest that there might be a subset of stage II patients at high risk of recurrence whom might benefit from 5-fluorouracil-based adjuvant chemotherapy. This analysis was limited to patients with stage III cancer because the number of patients with stage II disease who received chemotherapy was too small in the current study cohort. The use of new predictive gene signatures may help reduce the number of patients needed in prospective clinical trials to estimate the benefit of chemotherapy in stage II patients by identifying patients at higher risk in advance of treatment.

[0126] Molecular functions of genes upregulated in subtype B were in good agreement with clinical characteristics of that subtype. Interestingly, network analysis revealed that many of these genes were under the control of the TGF-β pathway, which is frequently associated with metastasis in many types of cancers by promoting epithelial-to-mesenchymal transition (Wendt et ah, 2009; Robson et ah, 1996). Activation of SRC family kinases in poorer prognostic subtype is in good agreement with previous studies showing that SRC or its related kinase activity increases in colorectal tumors relative to adjacent mucosa, with the highest activity observed in metastases and correlates inversely with patient survival (Yeatman, 2004; Lieu and Kopetz, 2010). Therefore, these genes overexpressed in patients with subtype B well reflect the aggressiveness of colorectal cancer cells. SRC -targeted agents, that are now in advanced clinical development for patients with solid tumors, might be good candidates for targeted therapy for patients in subtype B (Saad and Lipton, 2010).

[0127] Several previous studies have tried to identify prognostic gene expression signatures. Twenty-three-gene and 30-gene prognostic signatures were independently developed to predict recurrence in patients with AJCC stage II disease (Wang et al., 2004; Barrier et al., 2006). However, these signatures have not been validated in independent patient groups by other investigators and cannot predict the response to adjuvant chemotherapy. A recent study developed a risk score of recurrence based on 34 evolutionarily conserved genes (Smith et al, 2010). Although its independence over the use of stage has not been firmly established, a 128-gene signature was identified as a marker for the genomic stage of colorectal cancer and was well associated with prognosis (Jorissen et al., 2009). A new multi-gene expression assay for colon cancer, known as Oncotype DX™, has been introduced with the aim of improving treatment decisions, especially for patients with stage II disease (Kerr et al., 2009). Although this 7-gene prognostic marker was validated in the QUASAR cohort, the chemotherapy- benefit gene signature was not validated in the same cohort. Interestingly, of the 7 genes in the prognostic Oncotype DX, 3 genes (FAP, INHBA, and BGN) are present in the extended gene list (FIGS. 4A-B), suggesting that there might be partial overlap between the prognostic subtypes (A and B) and the high and low risk groups identified by the Oncotype DX recurrence score.

[0128] This new signature may overcome current limitations of biomarkers of colorectal cancer. Among several interesting biomarkers linked with clinical outcomes of patients with colorectal cancer (Walther et al , 2009), microsatellite instability (MSI) is the only prognostic marker validated in multiple studies and independently of stage (Popat et al, 2005; Thibodeau et al., 1993). Although the predictive values of MSI to adjuvant chemotherapy and of KRAS mutations to the use of EGFR inhibitors have been established, these markers are only useful as negative markers for treatments (Karapetis et al., 2008; Ribic et al., 2003). Thus, these markers fail to predict which patients will benefit from treatments.

[0129] While it is interesting to see the association of the signature with the potential benefit of adjuvant chemotherapy in patients with stage III colorectal cancer, the predictive nature of the signature is not firmly established yet since interaction of subgroup with adjuvant chemotherapy (or heterogeneity of two subgroups over adjuvant chemotherapy) did not reach significance level (FIG. 8). However, due to the small number of patients used in analysis, it would be too premature to draw strong conclusions for the predictive nature of the signature. Although significant, the multivariate analysis (Table 5) has also some limitation since other known predictors of prognosis in colorectal cancer, such as the refined TN substages, MSI status, and number of examined nodes, are not included. Thus, the significance and robustness of the signature as prognostic markers and predictive markers for adjuvant chemotherapy remains to be determined in future studies. [0130] In conclusion, 2 new prognostic subtypes of colorectal cancer that showed a significant difference in survival of patients were identified. The 114-gene signature can predict the response to adjuvant chemotherapy. These results clearly demonstrated that this gene signature reflects the molecular characteristics of colorectal cancer patients and may be used for the rational design of future clinical trials for testing the benefit of adjuvant chemotherapy for patients with stage II and ultimately stage III colorectal cancer. These methods may be used to improve patient care by providing additional practical guidance for different treatments.

* * *

[0131] All of the methods disclosed and claimed herein can be made and executed without undue experimentation in light of the present disclosure. While the compositions and methods of this invention have been described in terms of preferred embodiments, it will be apparent to those of skill in the art that variations may be applied to the methods and in the steps or in the sequence of steps of the method described herein without departing from the concept, spirit, and scope of the invention. More specifically, it will be apparent that certain agents which are both chemically and physiologically related may be substituted for the agents described herein while the same or similar results would be achieved. All such similar substitutes and modifications apparent to those skilled in the art are deemed to be within the spirit, scope, and concept of the invention as defined by the appended claims.

REFERENCES

The following references, to the extent that they provide exemplary procedural or other details supplementary to those set forth herein, are specifically incorporated herein by reference.

Ausubel et al., In: Current Protocols in Molecular Biology, John Wiley & Sons, NY, 631- 636, 2003.

Bao et al., Cancer Cell, 5:329-39, 2004.

Barrier et al., J. Clin. Oncol, 24:4685-91, 2006.

Benson et al., J. Clin. Oncol., 22:3408-19, 2004.

Carlsson et al., Dis. Colon Rectum, 30:431-4, 1987.

Cunningham et al., N. Engl. J. Med., 351 :337-45, 2004.

Dudoit et al., J. Am. Stat. Assoc., 97:77-87, 2002.

Eisen et al., Proc. Natl. Acad. Sci. USA, 95: 14863-8, 1998.

Hoshida et al., N. Engl. J. Med., 359: 1995-2004, 2008.

Irizarry et al., Biostatistics., 4:249-64, 2003.

Jemal et al., CA Cancer J. Clin., 59:225-49, 2009.

Jorissen et al., Clin. Cancer Res., 15:7642-51, 2009.

Kabbinavar et al., J. Clin. Oncol. , 21 :60-5, 2003.

Karapetis et al., N. Engl. J. Med, 359: 1757-65, 2008.

Kerr et al., J. Clin. Oncol, 27(15s):4000, 2009.

Kopetz et al., Clin. Cancer Res., 13:7232-6, 2007.

Kopetz et al., J. Clin. Oncol, 27:3677-83, 2009.

Laurie et al., J. Clin. Oncol, 7: 1447-56, 1989.

Lee et al., Hepatology, 40:667-76, 2004b.

Lee et al., Nat. Genet., 36: 1306-11, 2004a.

Lee et al., Nat. Med, 12:410-6, 2006.

Lieu and Kopetz, Clin. Colorectal Cancer, 9:89-94, 2010.

Mamounas et al., J. Clin. Oncol, 17: 1349-55, 1999.

Midgley and Kerr, Lancet, 353:391-9, 1999.

Midgley et al., Nat. Clin. Pract. Gastroenterol. Hepatol, 6: 108-20, 2009.

Moertel et al., Ann. Intern. Med., 122:321-6, 1995.

Moertel et al., N. Engl. J. Med., 322:352-8, 1990. Paik et al., N. Engl. J. Med., 351 :2817-26, 2004.

Parkin et al., CA Cancer J. Clin., 55:74-108, 2002.

Popat et al., J. Clin. Oncol, 23:609-18, 2005.

Quasar Collaborative Group, Gray et al (Eds.), Lancet, 370:2020-9, 2007.

Radmacher et al., J. Comput. Biol., 9(3):505-l 1, 2002.

Ribic et al., N. Engl. J. Med, 349:247-57, 2003.

Robson et al., Br. J. Cancer, 74:753-8, 1996.

Saad and Lipton, Cancer Treat. Rev., 36: 177-84, 2010.

Saclarides et al., Dis. Colon Rectum, 37(7):635-42, 1994.

Sambrook et al., In: Molecular cloning: a laboratory manual, 2^nd Ed., Cold Spring Harbor Laboratory Press, Cold Spring Harbor, ΝΥ, 1989.

Simon et al., Cancer Inform., 3: 11-7, 2007.

Simon et al., In: Design and Analysis of DNA Microarray Investigations, Chapter 8, Springer, 2003.

Smith et al., Gastroenterology, 138:958-68, 2010.

Smith et al., Gastroenterology, 138:958-68, 2010.

Sparano and Paik, J. Clin. Oncol, 26:721-8, 2008.

Staub et al., J. Mol. Med., 87:633-44, 2009.

Thibodeau et al., Science, 260:816-9, 1993.

Walther et al., Nat. Rev. Cancer, 9:489-99, 2009.

Wang et al., J. Clin. Oncol, 22: 1564-71, 2004.

Wendt et al., Future Oncol, 5: 1145-68, 2009.

Wolmark et al., J. Clin. Oncol, 11 : 1879-87, 1993.

Woo et al., Nature, 394:700-703, 1998.

Wright et al., Proc. Natl. Acad. Sci. USA, 100(17):9991-6, 2003.

Yeatman and Chambers, Clin. Exp. Metastasis, 20:85-90, 2003.

Yeatman, Nat. Rev. Cancer, 4:470-80, 2004.

Claims

CLAIMS;

1. An in vitro method of providing a prognosis or prediction for a subject determined to have a colorectal cancer, comprising:

a) obtaining expression information of biomarkers in a colorectal cancer sample of a subject by in vitro testing said sample, the biomarkers being at least ten genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD 109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRESl, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, V N1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALBl, ZIC2, CXCL5, TCNl, AGR3, TACSTD2, DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGF1 and

b) providing a prognosis or prediction for the subject based on the expression information, wherein, as compared with a reference expression level, increased expression of one or more genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRESl, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, VNN1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALBl, ZIC2, CXCL5, TCNl, AGR3, and TACSTD2 indicates a poor survival, a high risk of recurrence or a favorable response to an adjuvant therapy, and increased expression of one or more genes selected from the group consisting of DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGF1 indicates a favorable survival, a low risk of recurrence or a poor response to an adjuvant therapy.

2. The method of claim 1, wherein said obtaining expression information comprises obtaining or receiving said sample.

3. The method of claim 2, wherein said sample is paraffin-embedded.

4. The method of claim 2, wherein said sample is frozen.

5. The method of claim 1, wherein said obtaining expression information comprises RNA quantification.

6. The method of claim 5, wherein the RNA quantification comprises cDNA microarray, quantitative RT-PCR, in situ hybridization, Northern blotting or nuclease protection.

7. The method of claim 1, wherein said obtaining expression information comprises protein quantification.

8. The method of claim 7, wherein said protein quantification comprises immunohistochemistry, an ELISA, a radioimmunoassay (RIA), an immunoradiometric assay, a fluoroimmunoassay, a chemiluminescent assay, a bioluminescent assay, a gel electrophoresis, or a Western blot analysis.

9. The method of claim 1, wherein providing the prognosis or prediction comprises generating a classifier based on the expression, wherein the classifier is defined as a weighted sum of expression levels of the biomarkers.

10. The method of claim 9, wherein the classifier is generated on a computer.

11. The method of claim 9, wherein the classifier is generated by a computer readable medium comprising machine executable instructions suitable for generating a classifier.

12. The method of claim 9, wherein providing the prognosis or prediction comprises classifying a group of subjects based on the classifier associated with individual subjects in the group with a reference value.

13. The method of claim 1, further comprising reporting said prognosis or prediction.

14. The method of claim 1, further comprising prescribing or administering an adjuvant therapy to said subject based on said prediction.

15. The method of claim 1, wherein the cancer is a stage II cancer.

16. The method of claim 1, wherein the cancer is a stage III cancer.

17. The method of claim 1, wherein the cancer is not a stage IV cancer.

18. An array comprising a plurality of antigen-binding fragments that bind to expression products of biomarkers or a plurality of primers or probes that bind to transcripts of the biomarkers to assess expression levels, the biomarkers comprising at least ten genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD 109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRESl, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, V N1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALB1, ZIC2, CXCL5, TCN1, AGR3, TACSTD2, DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGFl.

19. The array of claim 18, wherein the array is a microchip.

20. The array of claim 18, wherein the array is a cDNA microarray.

21. A kit comprising a plurality of antigen-binding fragments that bind to expression products of biomarkers or a plurality of primers or probes that bind to transcripts of the biomarkers to assess expression levels, the biomarkers comprising at least ten genes selected from the group consisting of TMEM45A, ALOX5, DUSP4, ANOl, LOX, CD 109, LOX, OLR1, FAP, CXCR4, CD55, SPON1, ANXA1, CD55, GAS1, MEIS2, BCAT1, ADAM 12, PLK2, TNFAIP6, POSTN, RGS2, PTGS2, DUSP4, RARRESl, COL11A1, SPP1, FCGR3B, RARRESl, SERPINB5, GPNMB, CRIPl, TNFAIP6, VGLL3, RARRESl, SFRP2, CTSE, COL10A1, KLK10, VNN1, AHNAK2, PTGS2, CYP1B1, BAG2, COL10A1, SDR16C5, COL11A1, PLA2G4A, RAB27B, IL8, REG4, GCNT3, LYZ, REG4, CCL18, IGF2BP3, CALB1, ZIC2, CXCL5, TCN1, AGR3, TACSTD2, DEFA6, SLC26A3, KRT23, FABP1, EREG, DACH1, MUC12, ACE2, DPEP1, PTPRO, UGT2A3, APCDD1, VAV3, LRRC19, NOX1, PLCB4, SLC3A1, CHN2, ACSL6, ACE2 PLCB4, ENPP3, MEP1A LOC100288092, ABAT, AMACR, ΑΧΓΝ2, CTTNBP2, CFTR, C13orfl8, CEL, QPRT, VAV3, C10orf99, LY6G6D, CELP, VAV3, QPRT, ASCL2, and TDGFl, wherein said kit is housed in a container.