WO2012142330A1 - Micro rnas as diagnostic biomarkers and therapeutics for ovarian cancer and metastatic tumors that disseminate within the peritoneal cavity - Google Patents

Abstract

The invention provides compositions and methods for predicting metastases of cancer.

Description

MICRO RNAS AS DIAGNOSTIC BIOMARKERS AND THERAPEUTICS FOR OVARIAN CANCER AND METASTATIC TUMORS THAT DISSEMINATE WITHIN THE PERITONEAL CAVITY

RELATED APPLICATIONS

This application claims the benefit of priority under 35 U.S. C. § 119(e) to U.S.

Provisional Application No: 61/476,056, filed April 15, 2011, which is incorporated herein by reference in its entirety.

STATEMENT AS TO FEDERALLY SPONSORED RESEARCH

This invention was made with Government support under NIH/NCR grant

5P41R 001395. The Government has certain rights in the invention.

FIELD OF THE INVENTION

This invention relates generally to the field of cancer.

BACKGROUND OF THE INVENTION

A biomarker is an identifying or distinguishing characteristic that is objectively measured and evaluated as an indicator of a normal biologic process, pathogenic process, or pharmacologic response to a therapeutic intervention. For example in the context of cancer, a biomarker often refers to a substance or process that is indicative of the presence of cancer in the body. A biomarker might be either a molecule secreted by a tumor or it can be a specific response of the body to the presence of cancer. Genetic, epigenetic, proteomic, glycomic, and imaging biomarkers can be used for cancer diagnosis, prognosis and epidemiology.

Epithelial ovarian cancer (EOC) is one of the most lethal gynecologic malignancy, affecting nearly twenty-two thousand women in the United States in 2010, and accounting for nearly fourteen thousand deaths during that year. Serous adenocarcinomas represent the majority of new EOC diagnoses, with most patients presenting symptoms during advanced stages of the disease. Despite removal of the ovaries, surgical debulking, and chemotherapy, most patients with serous EOC will suffer recurrences and ultimately succumb to the disease. Due to the limited efficacy of currently available treatment options for advanced EOC, there is a pressing need to develop new strategies to diagnose and treat ovarian cancer and other metastatic tumors. SUMMARY OF THE INVENTION

The invention represents a major advance in the diagnosis, prognosis, and treatment of tumors of the peritoneal cavity by providing micro-ribonucleic acids

(microRNAs/miRNAs) that are differentially expressed in primary and metastatic tumors (e.g. , serous epithelial ovarian cancer (EOC)), and which act as biomarkers to predict the severity of cancer and how patients respond to treatment. Specifically, the compositions and methods of the invention predict time to disease progression and overall survival. As such, described herein is the role of post-transcriptional regulation in controlling the spread of cancerous tissue in a subject. The subject is preferably a mammal in need of such treatment, e.g., a subject that has been diagnosed with a primary tumor in an organ or tissue of the peritoneal cavity. The mammal can be, e.g. , any mammal, e.g. , a human, a primate, a mouse, a rat, a dog, a cat, a cow, a horse, or a pig. In a preferred embodiment, the mammal is a human.

For example, a composition for predicting presence of secondary site metastases or a predisposition thereto in a subject diagnosed as having a primary tumor comprises a detection reagent specific for at least one microRNA sequence selected from the group consisting of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa- let-7d, hsa-miR-29a, hsa-miR-508-5p, hsa-miR-152, hsa-miR-509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR-886-3p. Optionally, the composition further comprises a second detection reagent specific for at least one mRNA selected from the group consisting of INTS4, NARS2, SNORD31, INTS2, TRIP10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1, ITGA11, IGHM, COL8A1, NNMT, COL1A1, BGN, INHBA, and COL11 Al (Group I). Alternatively, the composition comprises a detection reagent specific for at least one mRNA selected from the group consisting of FNTS4, NARS2, SNORD31, INTS2, TRIP10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1, ITGA11, IGHM, COL8A1, NNMT, COL1A1, BGN, FNHBA, and COL11A1 (i.e., in the absence of a detection reagent specific for a microRNA described above). In another example, the composition optionally comprises a second detection reagent specific for at least one mRNA selected from the group consisting of COPZ2, NUCB1, LPL, CCDC49, GFPT2, LOX, NNMT, RGS1, ASNA1, FXYD5, SERPINE1, KIF26B, S100A10, ALDH1A3, CALB2, and PLAUR (Group II). Exemplary compositions for miRNA detection/quantification include those in which the miRNA is selected from the group consisting of hsa-let-7d, hsa-miR-146a, hsa-miR-29a, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-708, hsa-miR-152, hsa-miR-214, and hsa-miR-150. In another example, the miRNA is selected from the group consisting of hsa- let-7d, hsa-miR-146a, hsa-miR-29a, hsa-miR-193a-5p, hsa-miR-31, and hsa-miR-150. In yet another example, the miRNA is selected from the group consisting of hsa-miR-146a, hsa- miR-193a-5p, hsa-miR-31, and hss-miR-150.

Combinations of miRNAs are particularly useful. For example, the microRNA comprises a combination of two microRNAS selected from the combinations listed in Table 10, a combination of three microRNAS selected from the combinations listed in Table 11, or a combination of four microRNAS selected from the combinations listed in Table 12.

mRNAs are useful as independent prognostic tools and together with miRNAs offer improved prognostic capability. Thus, detection/quantifiction of microRNAs is combined with detection/quantification of one or more mRNAs selected from the group consisting of mRNAs listed in Table 1 A or IB.

A method for predicting the presence of secondary site metastases or a predisposition thereto in a subject diagnosed as having a primary tumor is carried out by providing a tissue sample obtained from the primary tumor; detecting in the tissue sample at least two biomarkers selected from the group consisting of hsa-miR-146a, hsa-miR-150, hsa-miR- 193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa-miR-29a, hsa-miR-508-5p, hsa-miR-152, hsa-miR-509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR-886-3p; comparing the levels of at least one or two biomarkers in the tissue sample to a control level of the biomarkers, wherein a higher level of at least two biomarkers compared to the control level of the biomarkers indicates that the subject is suffering from or at an increased risk of developing secondary site metastases. RNA is detected/measured directly or processed into a cDNA before quantification.

Following quantification, the data is calculated, correlated, and compared, e.g., using a computer, to yield a prognosis for metastases or survival.

The method optionally further comprises detecting in the tissue sample a messenger ribonucleic acid (mRNA) transcript encoding INTS4, NARS2, SNORD31, INTS2, TRIP 10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1, ITGA1 1, IGHM, COL8A1, NNMT, COL1A1, BGN, INHBA, and

COL11 Al or a mRNA transcript in Table 1 or Table 2; comparing the levels of the mRNA transcript in the tissue sample to a control level of the mRNA. A lower level of the mRNA transcript compared to the control level of the mRNA indicates that the subject is suffering from or at an increased risk of developing secondary site metastases.

Specifically, described herein are compositions and methods of identifying and inhibiting or mimicking (or augmenting/replacing) microR As that promote the

dissemination of primary tumors and the formation of cancer metastasis. If a miR is upregulated/higher in a metastatic state, the target miR is inhibit to achieve a therapeutic effect. If a miR is downregulated/lower in a metastatic state, a miR or miR mimic agent is administered to achieve a therapeutic benefit. For example, the primary tumor is a tumor located within the peritoneal cavity, and the metastasis is intra-abdominal metastases or lymph node metastases. The primary tumor comprises serous, clear cell, endometrioid, or mucinous carcinoma cells. The compositions and methods described herein are useful in the diagnosis, prognosis, and treatment of various primary tumors and metastases, e.g., epithelial ovarian cancer (EOC), pancreatic cancer, gastric, kidney, colorectal cancer, hepatic cancer, bladder cancer, or breast cancer. Preferably the tumor is within the peritoneal cavity or the abdominal-pelvic cavity. Most preferably, the primary tumor comprises serous epithelial ovarian cancer (EOC), the most aggressive and lethal form of EOC.

The invention provides methods for predicting an increased risk of developing distant metastatic neoplastic disease {i.e., secondary site metastases) in a subject diagnosed as comprising a primary tumor. By "increased risk" is meant a risk greater than that typically associated with individuals comprising a primary tumor. First, the level of at least one biomarker {e.g., at least one, at least two, at least three, at least four, or at least five biomarkers) is detected in a tissue sample obtained from a primary tumor, tissue/organ from which the tumor was derived, or bodily fluid from the affected subject, e.g., a subject diagnosed as having an ovarian tumor or a lump that is diagnosed as being suspect of a malignant condition. For example, the tissue sample is obtained from a bodily fluid, e.g., blood, serum, lymphatic fluid, plasma, urine, saliva, semen, or breast milk from a subject. A difference in level miR or mRNA levels comprises at least 10%, 20%, 50%>, 2-fold, 10- fold, 20-fold or more compared to a control or reference value. For example, a control or reference value comprises an average of measurements of at least 100 primary ovarian tumors. For example, mir-146a and mir-150 are separate from the average by 50% on the Agilent arrays in the TCGA data, and there is a two-fold difference in expression for mir- 146a and mir-150 in the high and low risk groups.

The invention also encompasses testing expression of the above-described microRNAs and mRNAs in metastatic tumors using the same methods. Optionally, the change in expression of the microRNA or mRNA between the primary tumor and metastatic tumor is measured and quantified. The difference in expression level is indicative of patient outcome, i.e., a greater difference corresponds to poor prognosis. If upregulation of a given microRNA or mRNA is correlated with a higher risk of developing metastatic disease, an even higher level of expression of the microRNA or mRNA in a sample of a metastatic tumor from the same individual indicates a poor outcome (more severe disease) in that individual, and vice versa.

Provided herein are miRNA expression signatures (i.e., biomarkers) that are differentially expressed between primary and metastatic tumors and predict time to disease progression and overall survival. The biomarkers are selected from the group consisting of micro ribonucleic acid-31 (miR-31 ; GenBank Accession Number NR 029505.1

(GL262205714), incorporated herein by reference), miR-124 (GenBank Accession Number NR 029670.1 (GL262205258), NR 029669.1 (GL262205253), or NR 029668.1

(GL262205247), incorporated herein by reference), miR-150 (GenBank Accession Number NR 029703.1 (GL262205410), incorporated herein by reference), miR-508-3p and miR-508- 5p (GenBank Accession Number NR 030235.1 (GL262205536), incorporated herein by reference), miR-21 (GenBank Accession Number NR 029493.1 (GL262205659), incorporated herein by reference), miR- 185 (GenBank Accession Number NR 029706.1 (GL262205427), incorporated herein by reference), miR-708 (Gene ID: 100126333, incorporated herein by reference), miR-193a-5p (GenBank Accession Number NR 029710.1 (GL262205449), incorporated herein by reference), miR-29a (GenBank Accession Number NR 029503.1 (GL262205706), incorporated herein by reference), let-7d (GenBank

Accession Number NR 029481.1 (GL262205605), incorporated herein by reference), miR- 886-3p (5' - cgcgggugcuuacugacccuu - 3'; SEQ ID NO: 1), miR-370 (GenBank Accession Number NR 029863.1 (GL262206207), incorporated herein by reference), miR-509-3-5p (GenBank Accession Number NR 030629.1 (GL262206205), NR 030586.1

(GL262205998), or NR 030236.1 (GL262205543), incorporated herein by reference), miR- 152 (GenBank Accession Number NR 029687.1 (GL262205334), incorporated herein by reference), miR-146a (GenBank Accession Number NR 029701.1 (GL262205399), incorporated herein by reference), miR-431 (Gene ID: 574038, incorporated herein by reference), and miR-214 (GenBank Accession Number NR 029627.1 (GL262206305), incorporated herein by reference). Preferably, due to their unique ability to predict patient outcome, the miRNA biomarkers are selected from the group consisting of hsa-miR-193a-5p, hsa-miR-708, hsa- miR-31, hsa-miR-508-3p, hsa-miR-124, and hsa-miR-185.

In one aspect, the at least two biomarkers are detected via quantitative real time polymerase chain reaction (qPCR). The level of the at least two biomarkers in the

tissue/tumor sample is compared to a control level obtained from a normal tissue. By

"control level" is meant a level previously determined to be associated with non-metastatic tumors. For example, the control level is obtained from a normal tissue sample. By "normal tissue" is meant non-cancerous tissue. An increase in the level of at least two biomarkers in the tumor sample indicates that the subject is suffering from or at an increased risk of developing a malignant tumor at an anatomical site distant from the primary tumor, thereby predicting an increased risk of developing distant metastatic neoplastic disease. For example, a distant metastasis from ovarian cancer is a tumor that occurs in an organ or tissue other than ovarian tissue. The tumor is an endocrine tumor such as breast cancer, ovarian cancer, colon cancer, prostate cancer and endometrial cancer. In ovarian cancer, metastases to the lymph nodes are typically used for initial staging; however, the methods of the invention are also useful to predict lymph node involvement such as axillary lymph node involvement. The methods are also useful to predict time course of disease progression and to predict patient survival.

After determining the level of at least two biomarkers as described above, a messenger ribonucleic acid (mRNA) transcript is optionally detected in the tissue sample. Alternatively, methods for predicting an increased risk of developing distant metastatic neoplastic disease (i.e., secondary site metastases) in a subject diagnosed as comprising a primary tumor include detecting an mRNA transcript in a tissue sample without prior identification of miRNA biomarkers, as described above. A tissue sample is obtained from a primary tumor, tissue/organ from which the tumor was derived, or bodily fluid from the affected subject, e.g., a subject diagnosed as having an ovarian tumor or a lump that is diagnosed as being suspect of a malignant condition. For example, the tissue sample is obtained from a bodily fluid, e.g., blood, serum, lymphatic fluid, plasma, urine, saliva, semen, or breast milk from a subject. For example, at least one mRNA transcript is detected, e.g., at least two, at least three, at least four, or at least five mRNA transcripts are detected. In one aspect, the mRNA transcript is detected via quantitative real time reverse transcription polymerase polymerase chain reaction (qRT-PCR). Provided herein are mRNA expression signatures (i.e., biomarkers) that are differentially expressed between primary and metastatic tumors and predict time to disease progression and overall survival. Suitable mRNA transcripts (i.e., miRNA targets) include those encoding programmed cell death protein 4 (PDCD4; GenBank Accession Number CAI40095.1 (GI:57162254), incorporated herein by reference), serine/threonine-protein kinase 38-like (STK38L; GenBank Accession Number NP_055815.1 (GL24307971), incorporated herein by reference), reversion-inducing-cysteine-rich protein with kazal motifs (RECK; GenBank Accession Number AAH60806.1 (GL38174247), incorporated herein by reference), ELAV-like protein 1 (HuR; GenBank Accession Number NP 001410.2

(GL38201714) , incorporated herein by reference), protein numb homolog (NUMB;

GenBank Accession Number AAD01548.1 (GL4102705), incorporated herein by reference), 14-3-3 protein epsilon (YWHAE; GenBank Accession Number NP 006752.1 (GL5803225), incorporated herein by reference), and AT-rich interactive domain-containing protein 1 A (ARID 1 A; GenBank Accession Number AC049597.1 (GL226346315), incorporated herein by reference). The mRNA transcripts listed in Tables 1 and 2 are also suitable miRNA targets.

The level of the at least one mRNA transcript in the tissue/tumor sample is compared to a control level obtained from a normal tissue. For example, a control level is obtained from a non-cancerous tissue sample. A decrease (i.e., downregulation) in the level of mRNA in the tumor sample compared to the control level indicates that the subject is suffering from or at an increased risk of developing distant metastatic neoplastic disease. Alternatively, an increase (i.e., upregulation) in the level of mRNA in the tumor sample compared to the control level indicates that the subject is suffering from or at an increased risk of developing distant metastatic neoplastic disease. The primary tumor comprises serous, clear cell, endometrioid, or mucinous carcinoma cells. Preferably, the tumor is within the peritoneal cavity. These methods are useful in the diagnosis, prognosis, and treatment of various primary tumors and metastases, e.g. , epithelial ovarian cancer (EOC), pancreatic cancer, colorectal cancer, hepatic cancer, bladder cancer, and breast cancer.

Also provided are methods of treating a peritoneal cavity tumor in a subject in need thereof comprising administering to the subject a therapeutically effective amount of a composition comprising an inhibitor of a miRNA selected from the group consisting of miR-31, miR-21, miR-124, miR-150, miR-185, miR-708, miR-193-5p, miR-29a, let-7d, miR- 886-3p, miR-270, miR-152, miR-146a, miR-431, and miR-214. Suitable inhibitors include anti-miRs, antagomiRs, peptide nucleic acid (PNA) locked nucleic acid (LNA), or small molecule inhibitors.

Also described herein are methods of treating a peritoneal cavity tumor in a subject in need thereof comprising administering to a subject a therapeutically effective amount of a composition comprising miR-508-3p, miR-509-3-5p, or miR-508-5p. Alternatively, synthetic miRNA mimics that have the same sequence as the depleted, naturally occurring miRNA are administered. Suitable miRNA mimics include a miR-509-3-5p mimic, a miR-508-3p mimic, and a miR-508-5p mimic.

Also provided is a method of inhibiting cancer metastases in the peritoneal cavity of a subject comprising modulating the level of miR-31, miR-21, miR-124, miR-150, miR-185, miR-708, miR-193-5p, miR-29a, let-7d, miR-886-3p, miR-270, miR-152, miR-146a, miR-431, miR-214, miR-508-3p, miR-509-3-5p, or miR-508-5p to change mRNA transcript levels. Preferably, the mRNA encodes PDCD4, STK38L, RECK, HuR, NUMB, YWHAE, or ARID 1 A. By changing the level of miRNA, the level of mRNA and encoded protein would also be modulated. In this manner, proteins are utilized as markers to evaluate the therapeutic efficacy of miRNA administration.

A method of prognosis for ovarian cancer patients includes the steps of detecting various miRNAs (described above) in a sample of ovarian tissue or bodily fluid of the patient following excision of a primary tumor, wherein an elevation in the level of various miRNA compared to a normal control level or over time indicates recurrence of malignancy. A method for predicting survival time of a cancer patient is carried out by detecting various miRNAs (described above) in a tissue biopsy in which an increase in various miRNAs levels is correlated with a decrease in survival time. The assay for prediction of survival of the individual or recurrence of a tumor is carried out before or after any treatment of the patient for the cancer.

All polynucleotides and polypeptides of the invention are purified and/or isolated.

Purified defines a degree of sterility that is safe for administration to a human subject, e.g., lacking infectious or toxic agents. Specifically, as used herein, an "isolated" or "purified" nucleic acid molecule, polynucleotide, polypeptide, or protein, is substantially free of other cellular material, or culture medium when produced by recombinant techniques, or chemical precursors or other chemicals when chemically synthesized. Purified compounds are at least 60% by weight (dry weight) the compound of interest. Preferably, the preparation is at least 75%, more preferably at least 90%>, and most preferably at least 99%>, by weight the compound of interest. Purity is measured by any appropriate standard method, for example, by column chromatography, polyacrylamide gel electrophoresis, or HPLC analysis. A pure polynucleotide (ribonucleic acid (RNA) or deoxyribonucleic acid (DNA)) is free of the genes or sequences that flank it in its naturally-occurring state.

Similarly, by "substantially pure" is meant a nucleic acid, polypeptide, or other molecule that has been separated from the components that naturally accompany it.

Typically, the polynucleotide, polypeptide, or other molecule is substantially pure when it is at least 60%, 70%, 80%, 90%, 95%, or even 99%, by weight, free from the proteins and naturally-occurring organic molecules with which it is naturally associated. For example, a substantially pure polypeptide may be obtained by extraction from a natural source, by expression of a recombinant nucleic acid in a cell that does not normally express that protein, or by chemical synthesis. A substantially pure nucleic acid (ribonucleic acid (RNA) or deoxyribonucleic acid (DNA)) is free of the genes or sequences that flank it in its naturally- occurring state.

By the terms "effective amount" and "therapeutically effective amount" of a formulation or formulation component is meant a nontoxic but sufficient amount of the formulation or component to provide the desired effect. For example, by an "effective amount" is meant an amount of a compound, alone or in a combination, required to reduce or prevent the growth or invasiveness of a tumor of the peritoneal cavity in a mammal. For example, the tumor is of an organ of the female reproductive system such as a breast or an ovarian tumor. The effective amount of active compound(s) varies depending upon the route of administration, age, body weight, and general health of the subject. Ultimately, the attending physician or veterinarian will decide the appropriate amount and dosage regimen.

The methods are also useful to inform choice of therapy, because gene expression signatures and levels of expression indicate whether or not the patient being treated will respond to Platinum/taxane therapy. The chemotherapy drugs used to treat ovarian cancer are fairly standard. Typically doctors combine a platinum-based drug such as carboplatin

(Paraplatin) or cisplatin with a taxane such as paclitaxel (Taxol) or docetaxel (Taxotere). Accordingly, a method of determining whether a tumor is resistant to a chemotherapeutic agent is carried out by detecting in a tissue sample from a subject at least one biomarker selected from the group consisting of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa- miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa-miR-29a, hsa-miR-508-5p, hsa-miR-152, hsa-miR-509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR-886-3p and comparing the levels of the biomarker in the tissue sample to a control level of the biomarkers. A higher level of hsa-miR-146a, hsa-miR-150, hsa-miR- 193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa-miR-29a, hsa-miR-152, hsa- miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, or hsa-miR-886-3p compared to the control level of the biomarkers indicates that the subject comprises a drug resistant tumor and wherein a lower level of has- miR-508-5p, hsa-miR-509-3-5p, or hsa- miR-508-3p indicates that the subject comprises a drug-resistant tumor. mRNA markers are also suitable for this purpose used either alone or in combination with an evaluation of microRNAs. For example, a method of determining whether a tumor is resistant to a chemotherapeutic agent is carried out by detecting in a tissue sample from a subject a messenger ribonucleic acid (mRNA) transcript encoding at least one of INTS4, NARS2, SNORD31, INTS2, TRIP10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1 , ITGA11, IGHM, COL8A1, NNMT, COL1A1, BGN, INHBA, and COL11A1 or a mRNA transcript in Table 1 or Table 2 and comparing the levels of the mRNA transcript in the tissue sample to a control level of the mRNA, wherein a lower level of the INTS4, NARS2, SNORD31, INTS2 mRNA transcript compared to the control level of the mRNA indicates that the subject comprises a drug resistant tumor and wherein a higher level of TRIP 10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1, ITGA11, IGHM, COL8A1, NNMT, COL1A1, BGN, INHBA, and COL11A1 level indicates that the subject comprises a drug resistant tumor. In another example, a method of determining whether a tumor is resistant to a chemotherapeutic agent comprises the steps of detecting in a tissue sample from a subject a messenger ribonucleic acid (mRNA) transcript encoding

encoding at least one of COPZ2, NUCB1, LPL, CCDC49, GFPT2, LOX, NNMT, RGS1, ASNA1, FXYD5, SERPINE1, KIF26B, S100A10, ALDH1A3, CALB2, and PLAUR and comparing the levels of the mRNA transcript in a tissue sample to a control level of the mRNA, wherein a higher level of mRNA indicates that subject comprises a drug resistant tumor. The information generated by these methods is useful to a physician. If the microRNA and/or mRNA profile indicates that the tumor(s) are drug resistant to platinum and/or taxane drugs, the physician would choose an alternative therapeutic strategy.

Other features and advantages of the invention will be apparent from the following description of the preferred embodiments thereof, and from the claims. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, suitable methods and materials are described below. All publications, patent applications, patents, Genbank/NCBI accession numbers, and other references mentioned herein are incorporated by reference in their entirety. In the case of conflict, the present specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and not intended to be limiting.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a schematic showing that metastatic tumors are clonal expansions with phenotypic diversity. Figure 1 A is a schematic showing hierarchical clustering of array comparative genomic hybridization (aCGH) data for primary and metastatic tumors. The results show that each matched pair of primary (P) and omental metastases (M) are more similar to each other than tumors from other patients. Segments were mapped to genes before clustering. Figure IB is a schematic showing hierarchical clustering of micro- ribonucleic acids (microRNAs/miRNAs). The results suggests that more primary and metastatic pairs are not alike, which implies phenotypic diversity.

Figure 2 is a series of graphs and photomicrographs showing that metastases exhibit lower expression of cell cycle checkpoints and are more proliferative compared to primary tumors. Figure 2A is a Gene Set Enrichment Analysis (GSEA) curve indicating that G2/M cell cycle checkpoints are exhibit higher expression in primary tumors compared to metastases. Figure 2B is a photomicrograph demonstrating that metastases have higher Ki-67 proliferation indexes than primary tumors, which is consistent with cell cycle checkpoint expression. Blue indicates hematoxylin and eosin (H&E) staining of tumor cells consistent with lower expression of cell cycle checkpoints in metastases. Brown indicates

immunohistochemistry (IHC) staining of Ki-67. p=0.01 by paired t-test, n=19.

Figure 3 is a series of bar charts and photomicrographs showing that miR-21 is overexpressed in metastases compared to primary tumors. Figure 3 A is a bar chart showing the results of a top 10 (p<0.03) Taqman qPCR screen from bulk tumor. Figure 3B is a bar chart showing the results of RNA Taqman qPCR on tumor cells isolated by Laser Capture Microdissection (LCM). Figure 3C is a series of photomicrographs of in situ hybridization (ISH) of two cases illustrating increased expression of miR-21 in metastases. Blue staining is ISH signal. Red staining is nuclear Red staining of tumor cells. Arrows indicate Blue staining of miR-21 signal. Increased miR-21 signal originates from tumor cells and not just stroma.

Figure 4 is a series of photomicrographs and bar charts demonstrating that metastatic miRNAs promote anchorage independent growth in ovarian cancer cells. Figure 4 A demonstrates that inhibiting miR-21 reduces spheroid growth, formed by hanging drop, over 4 days in soft agar in ES-2 cells (n=5, p=6e-4). Figure 4B shows that inhibiting miR-21 reduces colony formation in soft agar in ES-2 cells (n=3, p=0.05). Figure 4C shows that inhibition of miR-21 and miR-31 by peptide nucleic acids (PNAs) reduces spheroid size in soft agar in ovarian carcinoma cells (OVCAR-8). Figure 4D shows that that inhibition of miR-21 and miR-31 by peptide nucleic acids (PNAs) reduces the number of colonies in soft agar in OVCAR-8 cells. Error bars are standard deviation. P-values were calculated by Student's t-test. miR-31 did not significantly affect ES-2 cells in either spheroid or soft agar assays. Figure 4E is a bar chart showing that free miRNA expression levels in OVCAR-8 spheroids were reduced after inhibition, as determined by Taqman after non-organic RNA purification preserving the miRNA-PNA interaction.

Figure 5 is a schematic and series of line graphs showing that a metastasis expression signature (both mRNAs and miRNAs) predicts survival. The number of patients in each curve is noted. Figure 5 A is a flow-chart of classification using Cox univariate analysis, support vector machine (SVM) analysis, and Kaplan-Meier analysis. Figure 5B is a line graph showing that the 32 gene metastasis signature distinguishes patients by their overall survival. The lower curve is strongly biased towards higher expression of up-regulated metastasis genes. Figure 5C is a line graph showing that out of the top 10 differentially expressed miRNAs, 6 contribute significantly to Kaplan-Meier analysis of survival including miR-31. miR-21 was relatively uniformly expressed in primary tumors in this dataset.

miRNA classification was performed using a combination SVM and logistic model with leave one out validation.

Figure 6 is a schematic showing the experimental approach for sampling tumors. Specimens are sampled from each quadrant and compared to each other.

Figure 7 is a series of graphs and a chart demonstrating that miR-21 and mir-31 mRNA-predicted TargetScan targets are globally repressed in metastases compared to primary tumors. Figure 7A is a line graph showing global distribution of the Spearman correlation coefficients between mRNA targets and miR-31 and miR-21. miRNA mRNA targets predicted by PITA (red line) are TargetScan (blue line) are enriched for negatively correlated transcripts consistent with being down-regulated compared to randomly selected sets of transcripts permuted 1,000 times (grey lines). Figure 7B is a dot plot showing miR-21 and the literature validated miR-21 target, programmed cell death protein 4 (PDCD4), are negatively linearly correlated. Each point represents a primary/metastases matched pair. Figure 7C is a chart showing that genes with Spearman coefficients <-0.5 are significantly enriched for specific pathways and functions as determined by IPA. P-values are multiple hypothesis corrected using Benjamini-Hochberg. Selected genes for each pathway are listed, thereby representing high quality candidates.

Figure 8 is a series of line graphs, bar graphs, and photomicrographs showing that metastases are more proliferative and include less apoptotic cells than matched primary tumors. Figure 8 A is a line graphs showing a GSEA enrichment plot that suggests Reactome G2/M cell cycle checkpoints including Chekl, CCNB2, and BUB1 are expressed higher in primary tumors. Figure 8B is a bar graph showing a qPCR validation of cell cycle checkpoints. Figure 8C is a series of photomicrographs showing representative

immunohistochemical staining of Ki-67 from three cases, and suggests higher Ki-67 staining in omental metastases. Figure 8D is a line graph showing a GSEA enrichment plot that suggests that repressors of apoptosis from Gene Ontology are expressed higher in metastases than primary tumors. Figure 8E is a bar graph showing qPCR validation of selected genes. Figure 8F is a series of photomicrographs showing that representative TUNEL staining reveals more apoptotic cells in primary tumors compared their matched omental metastases.

Figure 9 is a series of line graphs showing Kaplan-Meier analysis of metastasis gene expression signature (Group II list of genes). Black dotted lines indicates high risk group and the grey line indicates the low risk group. Figure 9 A is a line graph showing the top 100 up- regulated genes in metastases applied to the TCGA Affymetrix dataset. Figure 9B is a line graph showing the 32 genes significantly up-regulated in omental metastases and with HR>1, p<0.05 by a Cox proportional hazards model in the TCGA Affymetrix U133 Plus 2.0 microarray dataset. Figure 9C is a line graph showing the 28 genes significantly up-regulated in omental metastases and with HR>1, p<0.05 in the TCGA Agilent 244K expression dataset. The 16 genes that overlap between the Affymetrix and Agilent platforms applied to the Figure 9D Affymetrix dataset and to the Figure 9E Agilent dataset. Figure 9F is a line graph showing the 16 overlap genes applied to the Tothill data (Tothill RW, et al. (2008) Clin Cancer Res 14 (16):5198-5208). Figure 9G-9I are a series of line graphs showing the combination of residual disease after debulking leads to robust discrimination, even for patients with sub-optimal debulking in Figure 91, on the TCGA Affymetrix U133 Plus 2.0 platform.

Figure 10 is a series of charts showing hierarchical clustering of copy number aberrations and mRNA expression. Figure 10A is a chart showing that most patients cluster by their copy number aberrations (CNAs). The circles indicate matched pairs of primary and metastatic tumors that cluster together with similar patterns of CNAs. Clustering by mRNA expression levels in primary and metastatic tumors reveals a less ordered linkage. Figure 1 OB is a series of representative heat maps of the copy number data from two cases. Many CNAs are found in both primary and metastatic tumors.

Figure 11 is a series of photomicrographs showing that tumors are of ovarian origin and are serous epithelial as indicated from examination of H&E and cytokeratin staining. Fiogure 11 shows H&E staining of two representative cases. CA125 and cytokeratin staining of one case is consistent with ovarian tumor origins.

Figure 12 is a series of bar charts showing microarray and qPCR measurements generally agree. Figure 12 shows qPCR validation of genes primary and metastatic tumors from two cases.

Figure 13 is a series of line graphs and charts showing that enriched pathways reveal common features of metastases. GSEA enrichment plot suggests that WNT/p-Catenin signaling from the Signaling Transduction database and double strand break repair gene ontology are up-regulated in metastases compared to primary tumors.

Figure 14 is a series of heat maps showing the distribution of expression levels in all TCGA patients ordered by survival. Green indicates higher expression and the gene expression signature is expressed higher in shorter survival patients.

Figure 15 is a series of line graphs showing Kaplan-Meier analysis of the 16 gene high-risk signature (Group II list of genes) applied to the TCGA Agilent microarray and Tothill datasets in combination with residual disease. Black dotted line indicates high risk group and grey line indicates low risk group.

Figure 16 is a schematic showing a network analysis of the 16 gene expression signature (Group II list of genes). 11 of the 16 genes are included in the network and are shaded in gray. Central regulatory nodes include NFKB, AKT, PDGF, and β-estradiol.

Figure 17 is a schematic showing a summary of all CNAs in all 12 cases. Most of the genome is subject to copy number aberrations consistent with extensive genomic instability of serous ovarian cancer. Green is amplification, red is copy loss. These data pertain to the Group II list of genes/gene expression signature.

Figure 18 is a schematic showing top IPA identified network in the 467genes up- regulated in metastases. Pathway analysis identified increased TFGB1 signaling in metastases using Ingenuity software analysis. Ingenuity pathway analysis (IPA) network representation of the most highly rated network. The shaded genes in red were selected to Fc>0.4, p<0.05. Solid lines represent a direct interaction between the gene products and a dotted line indicates an indirect interaction. Figure 19 is a line graph demonstrating the proportion survival of patients from 0 to 120 months after diagnosis. (Table 1A list of genes)

Figure 20 is a schematic demonstrating that 24 mR As significantly differentially upregulated in primary versus omental metastatic tumors distinguish primary and metastatic tumors.

Figure 21 is a series of graphs showing that a combination of mRNAs and miRNAs improve patient survival prognostic predictions. (Group I list of genes).

Figure 22 is a series of graphs showing that a combination of mRNAs and miRNAs improve patient survival prognostic predictions. (Group I list of genes).

Figure 23 is a series of graphs showing that a combination of mRNAs and 4 miRNAs improve patient survival prognostic predictions. (Group I list of genes).

Figure 24 is a series of graphs showing that an mRNA and miRNA signature combined with residual disease clear stratifies in high and low risk groups. (Group I list of genes).

Figure 25 is a series of line graphs demonstrating that miR-150 induces resistance and stimulates growth in SKOV-3 and IGROV-1 cells (miR-150 and miR-146a were key contributors to the survival curves described above). (Group I list of genes).

DETAILED DESCRIPTION

Described herein are compositions and methods for the diagnosis, prognosis, and treatment of primary tumors and metastases, e.g., metastatic ovarian cancer. Epithelial ovarian cancer (EOC) is a deadly disease affecting thousands of women each year. Serous adenocarcinomas represent the majority of new EOC diagnoses, with most present in advanced stages. Most are treated by removal of the ovaries and debulking surgery followed by platinum-taxane-based therapy, with 5-year survival under 50%, as patents often suffer from extensive metastatic disease. While the 5 year survival rates are 94% for localized tumors, the survival rates drop to 28% after distant metastasis. Only 15% of cases are diagnosed while the cancer is still confined to the primary site, and 62% of ovarian cancers are not detected until after distant metastasis.

Ovarian cancer is a lethal disease, typically diagnosed in late stages. To treat ovarian cancer, it is commonly thought that early diagnosis through imaging and/or the development of a screening test could be a panacea (Moore RG, et al. Am J Obstet Gynecol.

2010;203(3):228 el-6; Bast RC, Jr., et al. Int J Gynecol Cancer. 2005;15 Suppl 3:274-81). The cure rates and occurrence for cervical cancer have significantly improved from early detection. However, recent analysis suggests that early screening for prostate and breast cancer may not significantly reduce the occurrence of severe disease (Welch HG and

Albertsen PC J Natl Cancer Inst. 2009;101(19): 1325-9; Esserman L, et al. JAMA.

2009;302(15): 1685-92). One implication of these observations is that an improved understanding of cancers at early and late stages are needed. This includes identification of biomarkers to determine which tumors need to be treated. A second implication is that even with improvements in early detection strategies, many patients will still develop advanced disease and will require treatment. Thus, as described herein, the increased molecular characterization of primary and metastatic ovarian tumors improves understanding of malignant disease, ultimately allowing for personalized cancer treatment. The methods address some of the most critical issues in treating ovarian cancer today by focusing on the key features of advanced metastatic disease using a combination of clinical, in vitro, and animal models. As described below, this combination of approaches identifies novel features of metastasis and gains insight into the mechanisms of advanced disease. In addition, intra- tumor heterogeneity is examined to develop miRNA markers and gene expression profiles of ovarian cancer.

Locoregional metastases and medical significance

Patients suffer greatly from metastatic tumors infiltrating various organs (Nguyen DX, et al, Nat Rev Cancer. 2009;9(4):274-84). Understanding the genetic and expression changes in tumor cells that mediate metastasis is of paramount importance to treat ovarian cancer, which is rarely diagnosed before metastasis (Hoskins W et al. , Principles and

Practices of Gynecologic Oncology. 2nd ed. Philadelphia, PA: JB Lippincott; 1997). Prior to the invention described herein, methods did not enable effective detection of ovarian cancer in its early and potentially curable stages or at low volume relapse. Moreover, therapy for patients presenting with advanced stages (i.e., with intra-abdominal or lymphatic metastases) or relapsed disease does not exist. Synchronous intra-abdominal metastases (and lymphatic metastases) are critical to this disease because prognosis (and staging) is based on their presence or absence. When patients develop metastasis, their entire abdominopelvic cavity is at risk. Molecular characterization of these metastatic tumors guide application of current therapy options, as well as reveal new factors that can be targeted therapeutically.

Metastasis in model systems vs. patients

Comparing primary and metastatic tumors, in mouse models and in patients, generates insights into the genetic evolution of the disease and identifies key barriers to metastasis. Mouse models have provided important insights into the basis of metastasis and provide some insight into how the genetic make-up of the primary tumor leads to higher probability of metastasis including critical roles of specific miRNAs (Valastyan S, et al. Cell.

2009;137(6): 1032-46; Ma L, et al. Nature. 2007;449(7163):682-8; Tavazoie SF, et al. Nature. 2008;451(7175): 147-52). Prior to the invention described herein, most human studies have focused on comparing expression profiles from primary tumors to determine if patterns emerge that may predict metastasis (Minn AJ, et al. Nature. 2005;436(7050):518-24) with some successes including a commercialized expression chip (Martin M, et al. Clin Transl Oncol. 2009;11(10):634-42). These data support a "primary tumor predisposition" model where the innate primary tumor properties determine the metastatic potential (Hynes RO Cell. 2003;113(7):821-3). Recent sequencing data supports a rare metastatic variant model (Yachida S et al. Nature. 2010;467(7319): 1114-7).

Comparing genetically matched primary and metastatic tumors by copy number and mRNA expression are consistent with a "primary tumor predisposition" model (Ramaswamy S, et al. Nat Genet. 2003;33(l):49-54; Colella S, et al. Head Neck. 2008;30(10): 1273-83; Liu CJ, et al. J Pathol. 2008;214(4):489-97; Paris PL, et al. Hum Mol Genet. 2004;13(13): 1303- 13). In ovarian cancer, early studies suggest that metastases form by clonal expansion based on 22 microsatellite markers (Khalique L, et al. Int J Cancer. 2009;124(7): 1579-86). mRNA expression data using early generation microarrays suggest that few significant express changes occur (Lancaster JM, et al. Int J Gynecol Cancer. 2006; 16(5): 1733-45), consistent with a "primary tumor predisposition" model. Similar copy numbers are observed in primary and metastases; however, as described herein, deeper sequencing reveals many genetic differences as recently found in pancreatic cancer (Yachida S, et al. Nature.

2010;467(7319): 1114-7). As described in detail below, using modern approaches, many significant differences were observed at the expression level (Figure 2). The data described herein explores miRNAs from matched primary and metastases (Figure 3), revealing significant expression differences.

Most of what is known about metastasis mechanisms originates from breast cancer studies (Nguyen DX and Massague J. Nat Rev Genet. 2007;8(5):341-52). Other cancers may use some of the same principles, but the specifics almost certainly differ. Ovarian and breast cancer expression profiles clearly differ (Schaner ME, et al. Mol Biol Cell.

2003; 14(11):4376-86. PMCID: 266758). Ovarian cancer likely has a different etiology, and appears to be particularly heterogeneous at the mRNA expression and DNA copy levels (Gorringe KL Mol Oncol. 2009;3(2): 157-64). Intra-tumor hetero eneity, sampling and metastasis

Intra-tumor heterogeneity provides a possible bottleneck in understanding and characterization of tumors, especially with gene expression where cell mixtures complicates interpretation (Marusyk A and Polyak K Biochim Biophys Acta. 2010;1805(1): 105-17.

PMCID: 2814927). Genetic divergence between primary and metastases has been observed by monitoring DNA copy number (Navin N, et al. Genome Res. 2009) and DNA sequencing identified mutations (Yachida S, et al. Nature. 2010;467(7319): 1114-7). These data are consistent with metastases originating from rare tumor cells within a primary tumor and clonally expanding into more genetically homogeneous metastases. However, prior to the invention described herein, why particular regions of a tumor are more likely to metastasize, and if metastases are always more homogeneous was unclear (Marusyk A and Polyak K. Biochim Biophys Acta. 2010;1805(1): 105-17. PMCID: 2814927). Genetic heterogeneity may manifest into a range of miRNA and mRNA expression profiles within each tumor, complicating characterization. Transiently stable epigenetic events including DNA methylation and miRNA expression may also contribute to driving progression (Visvader JE. Nature. 2011;469(7330):314-22). Specific changes may be selected during metastasis resulting in genomic remodeling accompanied by expression changes (Shah SP, et al. Nature. 2009;461(7265):809-13; Ding L, et al. Nature. 2010;464(7291):999-1005. PMCID:

2872544). Specific tumor cells with higher tumorigenicity, perhaps cancer stem cells, have been suggested to exist in ovarian tumors (Visvader JE Nature. 2011;469(7330):314-22; Alvero AB, et al. Cell Cycle. 2009;8(1): 158-66) and these cells may be the most important to drive disease progression. Deeper sampling of tumors is required to address whether minor cell populations may be the source of metastasis (Marusyk A and Polyak K. Biochim

Biophys Acta. 2010;1805(1): 105-17. PMCID: 2814927).

microRNAs as regulators of metastasis

MicroRNAs (miRNA) are small, noncoding single-stranded RNAs that comprise a class of gene regulators (Bader, et al. Cancer Research. 2010;70(18):7027-30, incorporated herein by reference). They are highly conserved from plants to humans and are encoded by their respective genes. miRNAs are transcribed from the genome as longer precursor molecules that are cleaved by the nuclear ribonuclease Drosha into approximately 70- to 100- nt-long oligonucleotides that form a distinct hairpin structure. Following nuclear export, this precursor is further cleaved by the RNAse Dicer, which yields a 17- to 25-nt double-stranded oligonucleotide that enters the RNA-induced silencing complex (RISC), a multiprotein complex that separates the mature strand from the passenger strand and facilitates the interaction of mR As with sequences that are complementary to the mature miRNA. RISC loaded with miRNA and the target mRNA inhibits the translation of the mRNA by either a silencing mechanismor by degradation of the mRNA. Inmost cases, the miRNA and mRNA sequences are merely partially complementary, which enables miRNAs to target a broad, but nevertheless a specific, set of mRNAs. To date, more than 900 human miRNA sequences have been annotated and may regulate at least 20 to 30% of all protein-encoding genes. The uncapitalized "mir-" refers to the pre-miRNA, while a capitalized "miR-" refers to the mature form.

miRNAs that affect metastasis in model systems have differential expression in primary and metastatic tumors, further justifying the approach described herein (Valastyan S, et al. Cell. 2009;137(6): 1032-46). miRNAs regulate gene expression post-transcriptionally affecting both mRNA stability and translation often repressing gene expression by 3 ' UTR binding (Filipowicz W, et al. Nat Rev Genet. 2008;9(2): 102-14). miRNAs are critical regulators of metastasis in a number of cancers (Hurst DR, et al. Cancer Res.

2009;69(19):7495-8). Some miRNAs may be pleiotropic and affect multiple steps of the metastatic process (Valastyan S, et al. Cell. 2009;137(6): 1032-46; Valastyan S, et al. Genes Dev. 2009; Valastyan S, et al. Cancer Res. 2010;70(12):5147-54. PMCID: 2891350). Prior to the invention described herein, determining which mRNAs are targeted by miRNAs was difficult, as the many computational methods have very high false positive rates in any given condition as recently reviewed, even with genome -wide expression data (Thomas M, et al. Nat Struct Mol Biol. 2010; 17( 10): 1169-74). The importance of miRNAs in driving tumor progression is highlighted by the suggestion that tumor cells become addicted to a specific miRNA for growth and survival (Medina PP, et al. Nature. 2010;467(7311):86-90). From in vitro and animal models, a number of miRNAs have been linked to metastasis, including miR-31 (Valastyan S, et al. Cell. 2009; 137(6): 1032-46; Dykxhoorn DM Cancer Res.

2010;70(16):6401-6. PMCID: 2922433). miRNA inhibitors and mimics as possible therapeutics to target miRNAs have generated significant interest as well as targeting putative miRNA-mRNA target genes (Garzon R, et al. Nat Rev Drug Discov. 2010;9(10):775-89; Cohen EE, et al. Cancer Res. 2009;69(l):65-74. PMCID: 2746005).

Detection and quantification of miRNAs and mRNAs

A qualitative analysis of the miRNAs (differentially) expressed in a particular sample is performed using known methods, e.g., the Agilent miRNA microarray platform (Agilent Technologies, Santa Clara, Calif, USA) according to the manufacturer's instructions.

Quantitative analysis (verification) of the miRNA expression data obtained is typically performed via real-time quantitative RT-PCR employing a TaqMan MicroRNA assay (Applied Biosystems, Foster City, Calif, USA) according to the manufacturer's instructions. Alternatively, the quantification of the miRNAs may be performed by using real-time quantitative RT-PCR employing SYBR Green I (Sigma Aldrich Corporation, St. Louis, Mo., USA), an asymmetrical cyanine dye binding to double-stranded DNA. Next generation sequencing methods and machines, e.g., Illumina's sequencing by synthesis (SBS) technology (HiSeq, TruSeq, MiSeq), Life Technologies SOLiD Sequencing products are also useful detection/quantification. For RT-PCR based methods, primers are used to amplify the microRNA or mRNA target and process it into cDNA. Such primers are at least 10 nucleotides in length (typically 17-25 nucleotides in length) and are complementary to the miRNA or mRNA sought to be evaluated. Primer design to amplify a known sequence is well known in the art. All of these methods and devices are suitable for analysis of both microRNAs and mRNAs.

Kits or systems for evaluating miRNAs and/or mRNAs include a) a oligonucleotide complementary to an miRNA; and b) optionally, reagents for the formation of the

hybridization between the oligonucleotide and the miRNA. For example, an apparatus or composition for diagnosing or prognosing metastases or survival in a patient with a cancer, comprises a solid support, in which a surface of the solid support is linked to an

oligonucleotide complementary to an miRNA. The level of an miRNA in a sample can be measured using any technique that is suitable for detecting RNA expression levels in a biological sample. Suitable techniques for determining RNA expression levels in cells from a biological sample are well known in the art. Examples of such techniques include, but are not limited to, Northern blot analysis, RT-PCR, microarrays, in situ hybridization. In a particular embodiment, a high-throughput system, for example, a microarray, is used to measure the expression level of a plurality of genes.

Accordingly, the plurality of nucleic acid molecules encoding a miRNA sequence that are comprised in a diagnostic kit of the present invention may include one or more "sense nucleic acid molecules" and/or one or more "anti-sense nucleic acid molecules". In case, the diagnostic kit includes one or more "sense nucleic acid molecules" (i.e. the miRNA sequences as such). In another example, a diagnostic kit includes one or more "anti-sense nucleic acid molecules" (i.e. sequences complementary to the miRNA sequences), in which the molecules may comprise probe molecules (for performing hybridization assays) and/or oligonucleotide primers (e.g., for reverse transcription or PCR applications) that are suitable for detecting and/or quantifying one or more particular (complementary) miRNA sequences in a given sample.

The relative number of miRNA gene transcripts in cells can also be determined by reverse transcription of miRNA gene transcripts, followed by amplification of the reverse- transcribed transcripts by polymerase chain reaction (RT-PCR), i.e., the miRNA is processed into cDNA and then analyzed. The levels of miRNA gene transcripts can be quantified in comparison with an internal standard, for example, the level of mRNA from a

"housekeeping" gene present in the same sample. A suitable "housekeeping" gene for use as an internal standard includes, e.g., myosin or glyceraldehyde-3 -phosphate dehydrogenase (G3PDH). The methods for quantitative RT-PCR and variations thereof are within the skill in the art. In another embodiment, a high throughput stem loop real-time quantitative polymerase chain reaction (RT-qPCR) is used to detect miRNA expression.

In some instances, it may be desirable to simultaneously determine the expression level of a plurality of different miRNA gene products in a sample. In other instances, it may be desirable to determine the expression level of the transcripts of all known miRNAs correlated with a cancer. In another example, a microchip (i.e., a microarray), may be constructed containing a set of probe oligodeoxynucleotides that are specific for a set of miRNA genes. Using such a microarray, the expression level of or or more miRNAs in a biological sample is determined by reverse transcribing the RNAs to generate a set of target oligodeoxynucleotides, and hybridizing them to probe oligodeoxynucleotides on the microarray to generate a hybridization, or expression, profile. The hybridization profile of the test sample is then compared to the pre-determined expression level of a control sample to determine which miRNAs have an altered expression level in cancer cells. As used herein, "probe oligonucleotide" or "probe oligodeoxynucleotide" refers to an oligonucleotide that is capable of hybridizing to a target oligonucleotide. "Target oligonucleotide" or "target oligodeoxynucleotide" refers to a molecule to be detected (e.g., via hybridization). By "miRNA-specific probe oligonucleotide" or "probe oligonucleotide specific for an miRNA" is meant a probe oligonucleotide that has a sequence selected to hybridize to a specific miRNA gene product, or to a reverse transcript of the specific miRNA gene product.

An "expression profile" or "hybridization profile" is essentially a fingerprint of the state of the sample; while two states may have any particular gene similarly expressed, the evaluation of a number of genes simultaneously allows the generation of a gene expression profile that is unique to the state of the cell. That is, normal tissue may be distinguished from a cancer tissue, and within a cancer tissue, different prognosis states (good or poor long term survival prospects, for example) may be determined. By comparing expression profiles of a cancer tissue in different states, information regarding which genes are important (including both up- and down-regulation of genes) in each of these states is obtained. The identification of sequences that are differentially expressed in a cancer tissue or normal tissue, as well as differential expression resulting in different prognostic outcomes, allows the use of this information in a number of ways. For example, a particular treatment regime may be evaluated (e.g., to determine whether a chemotherapeutic drug act to improve the long-term prognosis in a particular patient). Similarly, diagnosis may be done or confirmed by comparing patient samples with the known expression profiles. Furthermore, these gene expression profiles (or individual genes) allow screening of drug candidates that suppress the cancer expression profile or convert a poor prognosis profile to a better prognosis profile.

The microarray can be prepared from gene-specific oligonucleotide probes generated from known miR A sequences. In one example, the array contains one, two, three, four or more different oligonucleotide probes for each miRNA, one containing the active, mature sequence and the other being specific for the precursor of the miRNA. The array may also contain controls, such as one or more mouse sequences differing from human orthologs by only a few bases, which can serve as controls for hybridization stringency conditions. tRNAs from both species may also be printed on the microchip, providing an internal, relatively stable, positive control for specific hybridization. One or more appropriate controls for nonspecific hybridization may also be included on the microchip. For this purpose, sequences are selected based upon the absence of any homology with any known miRNAs.

A microarray may be fabricated using techniques known in the art and printed using commercially available microarray systems, e.g., the GENEMACHINE, OMNIGRID 100 MICRO ARRAYER and AMERSHAM CODELINK activated slides. Labeled cDNA oligomer corresponding to the target RNAs is prepared by reverse transcribing the target RNA with labeled primer. Following first strand synthesis, the RNA/DNA hybrids are denatured to degrade the RNA templates. The labeled target cDNAs thus prepared are then hybridized to the microarray chip under hybridizing conditions. At positions on the array where the immobilized probe DNA recognizes a complementary target cDNA in the sample, hybridization occurs. The labeled target cDNA marks the exact position on the array where binding occurs, allowing automatic detection and quantification. The output consists of a list of hybridization events, indicating the relative abundance of specific cDNA sequences, and therefore the relative abundance of the corresponding complementary miRNA, in the patient sample. For example, the labeled cDNA oligomer is a biotin-labeled cDNA, prepared from a biotin-labeled primer. The microarray is then processed by direct detection of the biotin- containing transcripts using, e.g., streptavidin-alexa647 conjugate, and scanned utilizing conventional scanning methods. Image intensities of each spot on the array are proportional to the abundance of the corresponding miRNA in the patient sample.

For detection of miRNAs or mRNAs, primers (Taqman MicroRNA reagents, e.g., primer pairs, panels, cards are available from Life Technologies/ Applied Biosystems (Carlsbad, California). Other reagents for detection of microRNAs, e.g., miScript miRNA PCR Array, miScript II RT Kit) are available from Qiagen (Valencia, California). Assay kits for measuring microRNA expression are also available from Nano String (Seattle, WA), e.g., nCounter miRNA Expression Assay Kits. miRCURY LNA™ Universal RT microRNA PCR is available from Exiqon (Vedbaek Denmark; Woburn, MA). High throughput methods relating to microRNA expression analysis are described in U.S. Patent No. 7,635,563, herein incorporated by reference. Other approaches include tiny hydrogel particles, about 200 micrometers in length, to rapidly detect microRNA patterns in RNA taken from four individuals with four different types of cancer (Chapin et al., 2011 Angew. Chem Int. Ed. 50:2289-2293; Chapin et al, 2011, Anal. Chem. 83;7179-7185; herein incorporated by reference).

miRNAs, mRNAs, and ovarian cancer

Down-regulation of miRNAs has been associated with more aggressive ovarian cancer and poorer patient outcomes through the down-regulation of Dicer in primary tumors (Merritt WM, et al. N Engl J Med. 2008;359(25):2641-50). Many miRNAs have been found to be up- or down-regulated in primary tumors compared to normal tissue (Dahiya N, et al. PLoS One. 2008;3(6):e2436. PMCID: 2410296; Corney DC, et al. Clin Cancer Res.

2010;16(4): 1119-28. PMCID: 2822884; Iorio MV, et al. Cancer Res. 2007;67(18):8699-707) including down-regulation of miR-34 and up-regulation of miR-200 families. However, specific miRNAs may not be as sensitive to Dicer, including not requiring Dicer for biogenesis (Cheloufi S, et al. Nature. 2010;465(7298):584-9) or up-regulated to regulate Dicer (Martello G, et al. Cell. 2010;141(7): 1195-207), highlighting the range of possibilities of miRNA processing and regulation of gene expression.

mRNA expression signatures can characterize different subtypes and predict prognosis of ovarian cancer (Tothill RW, et al. Clin Cancer Res. 2008;14(16):5198-208). miRNA expression signatures in metastasis and survival have not been investigated. The data presented herein is also applicable to large primary tumor studies such as TCGA to develop a possible prognosis expression signature. The integrated genomics approach of the national TCGA project identifies many mutations and candidate factors from primary tumors; however, prior to the invention described herein, their role in metastasis was unclear.

miR-21 and miR-31 in cancer

miR-21 usually promotes aggressive tumors, and is currently being targeted for therapeutic development (Medina PP, et al. Nature. 2010;467(7311):86-90; Connolly EC, et al. Mol Cancer Res. 2010;8(5):691-700; Bonci D. Recent Pat Cardiovasc Drug Discov.

2010;5(3): 156-61; Krichevsky AM and Gabriely G. J Cell Mol Med. 2009;13(l):39-53). Prior to the invention described herein, the role of miR-21 in metastasis and in ovarian cancer was unclear. Like many miRNAs, miR-21 has the potential to regulate hundreds of transcripts, which suggests miR-21 acts through a range of mechanisms. Described herein are results that indicate that some "literature-validated" targets are likely being repressed by miR-21 during ovarian metastasis, while others are not. On the other hand, miR-31 inhibits breast cancer progression, and appears to act as a suppressor of metastasis (Valastyan S, et al. Cell. 2009; 137(6): 1032-46). However, miR-31 has higher expression in metastatic tumors demonstrating that it promotes metastasis. The in vitro data described herein indicate that both miR-21 and miR-31 promote anchorage independent growth, consistent with the clinical observations.

miRNA mimics or mimetics

MicroRNAs (miRNA have also been identified as targets for therapeutic intervention (Bader, et al. Cancer Research. 2010;70(18):7027-30, incorporated herein by reference). The rationale for developing miRNA therapeutics is based on the premise that aberrantly expressed miRNAs play key roles in the development of human disease, and that correcting these miRNA deficiencies by either antagonizing or restoring miRNA function may provide a therapeutic benefit. For example, synthetic miRNA mimics that have the same sequence as the depleted, naturally occurring miRNA are administered to restore miRNA function.

Alternatively, miRNA function is inhibited by, e.g., anti-miRs, antagomiRs, peptide nucleic acid (PNA) locked nucleic acid (LNA), or small molecule inhibitors.

Mechanism based expression signatures to predict patient outcomes

mRNA expression signatures have been investigated to define subtypes and predict prognosis of ovarian cancer patients (Tothill RW, et al. Clin Cancer Res. 2008;14(16):5198- 208). Most of these signatures are developed by identifying a set of miRNAs or mRNAs that correlates with survival or some other clinical attribute. Sometimes mechanisms can be found suggesting that these factors are important. This approach often runs into statistical bias problems (Bair E and Tibshirani R. PLoS Biol. 2004;2(4):E108. PMCID: 387275; Shi L, et al. Nat Biotechnol. 2010;28(8):827-38; Majewski IJ and Bernards R. Nat Med.

2011;17(3):304-12). As described herein, a mechanism-based approach where putative genes are identified independently of the survival dataset identifes key general features of disease progression (Ramaswamy S, et al. Nat Genet. 2003;33(l):49-54). Described herein is the identification of miRNAs for use as biomarkers and therapeutic targets.

As key, and perhaps master, regulators of gene expression programs, miRNAs provide a practical tool to investigate metastasis, as they are relatively straightforward to measure and genetically manipulate. The results described herein identify miR-21 and miR-31 as critical regulators of ovarian cancer metastasis. Similar to ovarian tumors, a number of common cancers metastasize to the omentum and within the abdominopelvic cavity including pancreatic, colo-rectal, and liver cancers. Thus, the results described below are also relevant general features important for dissemination within the abdominopelvic cavity (Lengyel E. Am J Pathol. 2010;177(3): 1053-64. PMCID: 2928939). Identifying key features of advanced disease also helps discriminate the most aggressive early lesions as early detection strategies improve.

In epithelial ovarian cancer (EOC), it is common to remove both primary and metastatic {i.e., debulking) tumors during initial surgery allowing convenient access and collection of this material. For this reason, EOC provides an excellent clinical model system to probe metastasis. Characterization of metastases within the peritoneal cavity may be relevant to other abdominal cancers including tumors from the liver, colon and pancreas that also disseminate within the peritoneal cavity (Lengyel E. Ovarian cancer development and metastasis. Am J Pathol. 2010;177(3): 1053-64. PMCID: 2928939).

Described herein is a quadrant sampling approach to refine the characterization of ovarian tumors to gain insight into intra-tumor heterogeneity and to identify differentially expressed miRNAs regulating metastasis. The often large serous ovarian tumors allow for extensive intra-tumor sampling, while the large dynamic range of Taqman qPCR is utilized to accurately and specifically identify miRNAs with expression differences between primary tumors and metastases. As described in detail below, the results are validated using in situ hybridization and other qPCR methods to gain high confidence in the genomic screening data. Also described below is the function and expression of miRNAs and mRNA targets in a panel of ovarian cancer cell lines to reduce the dependence on any one genetic background. Finally, the clinical data is validated with multiple methods {e.g., in vitro and in vivo models), combined with functional testing of candidate miRNAs. Moreover, as described below, miR-21 and miR-31 are important to reduce apoptosis and to promote proliferation, respectively, leading to the more aggressive nature of cancer cells in metastases. The data presented herein indicate an interesting role of these miRNAs in driving 3D spheroid formation and growth, but not proliferation in 2D, adherent culture. As described in detail below, the function of miR-21 and miR-31 in metastasis is determined, and mRNA targets of miR-21 and miR-31 that drive metastasis are identified. The utility of 2D vs. 3D culture conditions for ovarian cancer cells is characterized by linking laboratory and clinical observations, and an approach is established to link human tumor observations with mechanisms and regulation of metastasis in xenograft mouse models.

Provided herein are miRNA expression signatures (i.e., biomarkers) that are differentially expressed between primary and metastatic tumors and predict time to disease progression and overall survival. miR-31 and miR-21 are examined herein because both have been previously associated with cancer, and because their changing expression levels was validated by multiple assays, as described below. Indeed, miR-21 and miR-31 are expressed higher in metastases in 89% and 78% of patients, respectively. As described herein, miR-31 is the top-ranked miRNA in the primary vs. metastases comparison, and miR-31 is significantly associated with more aggressive disease in a miRNA expression signature. miR-21 is one of the first miRNAs identified to promote growth and robustness of cancer cells, but has not been extensively examined in ovarian cancer. A large number of miR-21 targets have been previously identified. As described in detail below, miRNAs (e.g. , miR-31 and miR-21) are metastatic regulators. As such, described herein is the role of miRNA in promoting metastasis.

miR-150 is intergenic on chromosome 19 and thus is not obviously co-regulated with gene. miR-150 regulates B-cell differentiation and the timing of expression is critical for its proper role in promoting B cell development. Prior to the invention, there was little known about the role of miR-150 in cancer. Previous reports suggested that miR-150 can either promote or inhibit tumors, highlighting the common theme of context dependent functions of miRNAs. Using primary/metastatic tumor data, the results described herein does not indicate an inverse correlation with the expression of previously identified miR-150 targets P2RX7 or EGR2. Interestingly, miR-150 is modestly inversely correlated (Pearson = -0.21) in the primary and metastatic tumors with miR-21, a strong anti-apoptotic, oncogenic miRNA, suggesting a possible function for miR-150 in compensating for lower miR-21 activity in some tumors and/or selection of low PDCD4 expression in omental lesions. PDCD4 is significantly lower in most omental lesions compared to primary tumors (p=0.01, paired t- test).

Also provided herein are mRNA expression signatures (i.e., biomarkers) that are differentially expressed between primary and metastatic tumors and predict time to disease progression and overall survival. mRNA transcripts that are upregulated between primary and metastatic tumors include those set forth in Table 1.

TABLE 1 A: Top Upregulated niRNAs

(ie Hi¹ Symbol Gen liank Accession Number

I ^'propitiated mRNA (incorporated herein by reference)

FABP4 NM_001442 (NM_001442.2 GI: 168480125)

SFRP2 NM_003013 (NM_003013.2 GL52630413)

EPYC NM_004950 (NM_004950.4 GL223941903)

LUM NM_002345 (NM_002345.3 GL61742794)

FAP NM_004460 (NM_004460.2 GI: 16933539)

ΓΝΗΒΑ NM_002192 (NM_002192.2 GL62953137)

COL11A1 NM_001854 (NM_001854.3 GL98985805)

POSTN NM_006475 (NM_006475.2 GL209862906)

DCN NM_001920 (NM_001920.3 GL47419925)

VCAN NM_004385 (NM_004385.4 GL255918074)

COL1A2 NM_000089 (NM_000089.3 GL48762933)

TIMP3 NM_000362 (NM_000362.4 GL75905820)

COL3A1 NM_000090 (NM_000090.3 GI: 110224482)

CTSK NM_000396 (NM_000396.3 GI:315075295)

COL6A3 NM_004369 (NM_004369.3 GL190343014)

ITGBL1 NM_004791 (NM_004791.1 GL4758613)

THBS1 NM_003246 (NM_003246.2 GL40317625)

COL1A1 NM_000088 (NM_000088.3 GI:110349771)

THBS2 NM_003247 (NM_003247.2 GL40317627)

BGN NM_001711 (NM_001711.4 GL268607602)

ACTA2 NM_001141945 (NM_001141945.1 GL213688374)

NNMT NM_006169 (NM_006169.2 GL62953139)

SULF1 NM_001128205 (NM_001128205.1 GL189571640)

FN1 NM_212482 (NM_212482.1 GL47132556)

LOC100293211 ENST00000390601

AEBP1 NM_001129 (NM_001129.3 GL53692188)

MMP2 NM_004530 (NM_004530.4 GL189217851)

MXRA5 NM_015419 (NM_015419.3 GL201860266)

CCDC80 NM_199511 (NM_199511.1 GI:41152073)

KRT7 NM_005556 (NM_005556.3 GL67782364)

COL5A2 NM_000393 (NM_000393.3 GL89363016)

SERPINF1 NM_002615 (NM_002615.5 GL318037587)

IGLJ3 AB001736 (AB001736.1 GL2094748)

IGHD BC021276 (BC021276.2 GL33989177) Gene S\ in hoi ( ien liank Accession N umber

I |)rc»uliilc(l iiiR.NA (incorporated herein by reference)

ODZ3 NM_001080477 (NM_001080477.1 GL 122937399) IGHA1 AK128476 (AK128476.1 GL34535863)

TIMP3 NM_000362 (NM_000362.4 GL75905820) FGF7 NM_002009 (NM_002009.3 GL219842354)

ST6GALNAC2 NM_006456 (NM_006456.2 GI: 192448439) RAR ES2 NM_002889 (NM_002889.3 GL218931208) LGALS1 NM_002305 (NM_002305.3 GL85815826) CALB2 NM_001740 (NM_001740.4 GL297632392) COL8A1 NM_001850 (NM_001850.4 GL366392934) PRRX1 NM_006902 (NM_006902.3 GL56699461) FBN1 NM_000138 (NM_000138.4 GL281485549) ADAM 12 NM_003474 (NM_003474.4 GI: 194733766) FBX032 NM_058229 (NM_058229.3 GL335057517) LOX NM_002317 (NM_002317.5 GL296010938) TAGLN NM_001001522 (NM_001001522.1 GL48255906) CI S NM_201442 (NM_201442.2 GL224967098) HAS2 NM_005328 (NM_005328.2 GL 169791020) ITGA11 NM_001004439 (NM_001004439.1 GL52485852) MOXD1 NM_015529 (NM_015529.2 GI: 118421086) ETV1 NM_004956 (NM_004956.4 GL253683425) CNN2 NM_004368 (NM_004368.2 GL41327728) THY1 NM_006288 (NM_006288.3 GL221136764) RSAD2 NM_080657 (NM_080657.4 GL90186265) CYR61 NM_001554 (NM_001554.4 GL 197313774) ANXA1 NM_000700 (NM_000700.1 GL4502100)

SEMA3C NM_006379 (NM_006379.3 GL335057525) IGHA1 AF067420 (AF067420.1 GL3201899)

COMP NM_000095 (NM_000095.2 GL40217842) FNDC1 NM_032532 (NM_032532.2 GI: 148806907) IL7R NM_002185 (NM_002185.2 GL28610150) PDLIM3 M_014476 (NM_014476.4 GL302393571) IGHM BC020240 (BC020240.1 GL18044958)

SPARC NM_003118 (NM_003118.3 GL365777426) BCAT1 NM_005504 (NM_005504.6 GL296010898) SERPINE1 NM_000602 (NM_000602.4 GL383286745) CDH11 NM_001797 (NM_001797.2 GL 16306531) ANTXR1 NM_032208 (NM_032208.2 GL208022654) IGHA1 AF067420 (AF067420.1 GL3201899)

CYP1B1 NM_000104 (NM_000104.3 GI: 189491762) PLAU NM_002658 (NM_002658.3 GL222537757) CDCP1 NM_022842 (NM_022842.3 GL30410804) TLR3 NM_003265 (NM_003265.2 GL 19718735) SVEP1 NM_153366 (NMJ53366.3 GL 148886653) MFAP4 NM_002404 (NM_002404.2 GI:310923209) Gene S\ in hoi en liank Accession N umber

I |)rc»uliilc(l in NA ( incorporated herein In reference)

ADH1B NM_000668 (NM_000668.4 GL 160298141)

S100A10 NM_002966 (NM_002966.2 GL 115298655)

MMP13 NM_002427 (NM_002427.3 GL296010793)

DPYSL3 NM_001387 (NM_001387.2 GL45505175)

NTM NM_016522 (NM_016522.2 GL38045920)

IGHA1 BC073771 (BC073771.1 GL49257463)

ADIPOQ NM_004797 (NM_004797.3 GL295317373)

OLFML2B NM_015441 (NM_015441.1 GL46195718)

TRIM22 NM_006074 (NM_006074.4 GL313760627)

EFEMP1 NM_004105 (NM_004105.3 GL86787911)

TMEM139 NM_153345 (NM_153345.2 GL338222095)

IL1RAP NM_002182 (NM_002182.3 GL268839967)

GJB2 NM_004004 (NM_004004.5 GI: 195539329)

MT2A NM_005953 (NM_005953.3 GL205277384)

LY6E NM_002346 (NM_002346.2 GL 187827163)

KGFLP1 NR_003674 (NR_003674.2 GL383276551)

CRISPLD2 NM_031476 (NM_031476.3 GL215422418)

LOC100290146 ENST00000390603

HLA-A NM_002116 (NM_002116.7 GL337752171)

MMP14 NM_004995 (NM_004995.2 GL 13027797)

CRIPl NM_001311 (NM_001311.4 GL 188595726)

RAB31 NM_006868 (NM_006868.3 GL 170295841)

ABCC9 NM_005691 (NM_005691.2 GI: 110832834)

CXCL14 NM_004887 (NM_004887.4 GL208022628)

FYB NM_001465 (NM_001465.4 GL340545551)

BHLHE40 NM_003670 (NM_003670.2 GL 197276631)

FCGR2C NM_201563 (NM_201563.4 GL226874954)

KCNE4 NM_080671 (NM_080671.2 GL 154448883)

NEXN NM_144573 (NM_144573.3 GL 148839338)

TNFAIP6 NM_007115 (NM_007115.3 GL315139000)

C19orf33 NM_033520 (NM_033520.1 GL41281602)

RGS4 NM_001102445 (NM_001102445.2 GL 164664485)

S100A11 NM_005620 (NM_005620.1 GL5032056)

GREM1 NM_013372 (NM_013372.6 GL300795276)

HLA-A NM_002116 (NM_002116.7 GL337752171)

GADD45B NM_015675 (NM_015675.3 GL299782594)

PLXNCl NM_005761 (NM_005761.2 GL317008579)

PTRF NM_012232 (NM_012232.5 GL266457267)

LAMB1 NM_002291 (NM_002291.2 GI: 167614503)

COPZ2 NM_016429 (NM_016429.2 GI: 116734720)

MYH9 NM_002473 (NM_002473.4 GL225703132)

CALD1 NM_033138 (NM_033138.3 GL261490655)

ISM1 NM_080826 (NM_080826.1 GL 153791191)

C1R NM_001733 (NM_001733.4 GL66347874) Gene S\ in hoi en liank Accession N umber

I |)rc»uliilc(l in NA ( incorporated herein In reference)

COL6A1 NM_001848 (NM_001848.2 GL87196338)

GMFG NM_004877 (NM_004877.2 GL209954784)

BICC1 NM_001080512 (NM_001080512.1 GL 122937471)

LTBP2 NM_000428 (NM_000428.2 GL46389563)

RRP7A NM_015703 (NM_015703.4 GL296317288)

FCGR3A NM_000569 (NM_000569.6 GL51593094)

HLA-E NM_005516 (NM_005516.5 GL301171456)

ELF4 NM_001421 (NM_001421.3 GI: 187608760)

TGFBI NM_000358 (NM_000358.2 GL 170650698)

PRDM1 NM_001198 (NM_001198.3 GL 172072683)

PTGIS NM_000961 (NM_000961.3 GL61676177)

LIMK2 NM_016733 (NM_016733.2 GL73390131)

RGS1 NM_002922 (NM_002922.3 GL56682943)

HLA-F NM_018950 (NM_018950.2 GL 149158697)

LPL NM_000237 (NM_000237.2 GL 145275217)

RUNX2 NM_001024630 (NM_001024630.3 GL226442782)

CD68 NM_001251 (NM_001251.2 GL91199547)

PMEPA1 NM_020182 (NM_020182.4 GL364023807)

TNFRSF12A NM_016639 (NM_016639.2 GL 187828666)

IL2RG NM_000206 (NM_000206.2 GL291045209)

IBSP NM_004967 (NM_004967.3 GL 167466186)

HLA-F NM_018950 (NM_018950.2 GL 149158697)

FPR3 NM_002030 (NM_002030.3 GL38455412)

C1QA NM_015991 (NM_015991.2 GL87298824)

KLK6 NM_002774 (NM_002774.3 GL61744422)

PODXL NM_001018111 (NM_001018111.2 GL 144094253)

MS4A6A NM_152852 (NM_152852.2 GL351720864)

FGF9 NM_002010 (NM_002010.2 GL209529671)

PPP1R15A NM_014330 (NM_014330.3 GL 157674362)

RASGEF1B NM_152545 (NM_152545.1 GL22749128)

IRF1 NM_002198 (NM_002198.2 GL 196049386)

MMP11 NM_005940 (NM_005940.3 GL58331147)

C1QC NM 001114101 (NM 001114101.1 GL 166235902)

HLA-C NM_002117 (NM_002117.5 GL339882742)

COL5A1 NM_000093 (NM_000093.3 GL89276750)

HLA-B NM_005514 (NM_005514.6 GL 170650640)

HLA-G NM_002127 (NM_002127.5 GL269914083)

KGFLP1 NR_003674 (NR_003674.2 GL383276551) LRP1 NM_002332 (NM_002332.2 GI: 126012561) HLA-F NM_001098479 (NM_001098479.1 GL 149158701) ALCAM NM_001627 (NM_001627.3 GL343168768) Gene S\ in hoi en liank Accession N umber

I |)rc»uliilc(l in NA ( incorporated herein In reference)

SCARNA17 NR_003003 (NR_003003.2 GL93352543)

NFIX NM_002501 (NM_002501.2 GL56549648)

FRMD6 NM_001042481 (NM_001042481.1 GL 109715865)

PDGFRL NM_006207 (NM_006207.2 GL300068919)

ARPC1B NM_005720 (NM_005720.3 GL325197176)

CCL5 NM_002985 (NM_002985.2 GL22538813)

ISG15 NM_005101 (NM_005101.3 GL 193083170)

C1QTNF1 NM_030968 (NM_030968.2 GL38372915)

PRKG1 NM_001098512 (NM_001098512.2 GL345842460)

STAT2 NM_005419 (NM_005419.3 GL291219920)

FXYD5 NM_144779 (NM_144779.2 GL257195179)

LBA1 NM_014831 (NM_014831.2 GL257467635)

CTGF NM_001901 (NM_001901.2 GL98986335)

ITGA3 NM_002204 (NM_002204.2 GL 171846266)

FCGR1A NM_000566 (NM_000566.3 GL 167621452)

GFPT2 NM_005110 (NM_005110.2 GL 142360336)

ACSL1 NM_001995 (NM_001995.2 GL40807490)

ADAMTS12 NM_030955 (NM_030955.2 GL51558723)

SNCA NM_000345 (NM_000345.3 GL225690599)

NPR1 NM_000906 (NM_000906.3 GL 167830410)

LCP2 NM_005565 (NM_005565.3 GL47078282)

PLEK NM_002664 (NM_002664.2 GL 156616272)

WNT7A NM_004625 (NM_004625.3 GL34328912)

PPAPDC1A NM_001030059 (NM_001030059.1 GL73611919)

NUPR1 NM_001042483 (NM_001042483.1 GL 109948306)

FSTL3 NM_005860 (NM_005860.2 GL95104789)

KRT17 NM_000422 (NM_000422.2 GL 197383031)

C9orf31 AF220263 (AF220263.1 GL9587087)

LAPTM5 NM_006762 (NM_006762.2 GL222418609)

ZFHX4 NM_024721 (NM_024721.4 GL291167748)

C19orf24 NM_017914 (NM_017914.3 GL 189339238)

FYB NM_001465 (NM_001465.4 GL340545551)

MICAL2 NM_014632 (NM_014632.2 GL41281417)

FLNA NM_001456 (NM_001456.3 GL 160420313)

HLA-G NM_002127 (NM_002127.5 GL269914083)

FCGR2A NM_001136219 (NM_001136219.1 GL210031821)

RASSF2 NM_014737 (NM_014737.2 GL 169259774)

MMP16 AL136588 (AL136588.1 GL 13276678)

KHDRBS3 NM_006558 (NM_006558.1 GL5730072)

AK5 NM_174858 (NM_174858.2 GL320118914)

KRT18 NM_000224 (NM_000224.2 GL40354193)

RUNX1 NM_001001890 (NM_001001890.2 GL 169790826)

NUCB1 NM_006184 (NM_006184.5 GL297374833) Gene S\ in hoi ( ien liank Accession N umber

I |)rc»uliilc(l iiiR.NA ( incorporated herein In reference)

COLEC12 NM_130386 (NM_130386.2 GL 197245360)

FBLN1 NM_006486 (NM_006486.2 GL34734065)

PDE4B NM_002600 (NM_002600.3 GL82799480)

COL10A1 NM_000493 (NM_000493.3 GL98985802)

LHFP NM_005780 (NM_005780.2 GL 108773775)

MNDA NM_002432 (NM_002432.1 GL4505226)

SYK NM_003177 (NM_003177.5 GL293332490)

NUAK1 NM_014840 (NM_014840.2 GL48374438)

VDR NM_001017535 (NM_001017535.1 GL63054844)

GAS 6 NM_ 000820 (NM_ _000820.2 GL221316718)

GADD45A NM_ 001924 (NM_ 001924.3 GL315075321)

ODZ1 NM_ 014253 (NM_ _014253.3 GL253970447)

HCLS1 NM_ 005335 (NM_ 005335.4 GI: 167234421)

HLA-DQB1 NM_ 002123 (NM_ 002123.4 GL345461082)

PRKCDBP NM_ 145040 (NM_ 145040.2 GL47132586)

PLAUR NM_ 002659 (NM_ _002659.3 GL318037601)

TFAP2A NM_ 003220 (NM_ 003220.2 GL 109389359)

FOSL2 NM_ 005253 (NM_ 005253.3 GL44680151)

KIAA0040 NM_ 014656 (NM_ _014656.2 GL242332495)

LOXL2 NM_ 002318 (NM_ 002318.2 GL67782347)

FCGR1B NM_001017986 (NM_001017986.3 GL349732137)

LBH NM_030915 (NM_030915.3 GL 149274623)

MSRB3 NM_001031679 (NM_001031679.2 GL301336160)

GALNT5 NM_014568 (NM_014568.1 GL32698685)

AXL NM_021913 (NM_021913.3 GL 157779132)

LI CAM NM_000425 (NM_000425.3 GL221316755)

BCAM NM_005581 (NM 005581.3 GL61742795)

LRRC15 NM_001135057 (NM_001135057.2 GL288541294)

MT2A NM_005953 (NM_005953.3 GL205277384)

CD248 NM_020404 (NM_020404.2 GL45387956)

OSTF1 NM_012383 (NM_012383.4 GL 166235147)

DCBLD1 NM_173674 (NM_173674.1 GL27735142)

LRRFIP1 NM_001137550 (NM_001137550.1 GL212276077)

MRVI1 NM_130385 (NMJ30385.3 GL332634577)

CD2 NM_001767 (NM_001767.3 GL 156071471)

CD99 NM_002414 (NM_002414.3 GL34147599)

RGS16 NM_002928 (NM_002928.3 GL 156416008)

FN1 NM_212482 (NM_212482.1 GL47132556)

RCN3 NM_020650 (NM_020650.2 GL28626509)

SDC4 NM_002999 (NM_002999.3 GI:318037226)

FAM38B AK302764 (AK302764.1 GL221044241)

BASP1 NM_006317 (NM_006317.3 GL30795230)

CCDC102B NM_001093729 (NM_001093729.1 GL 147905591) Gene Symbol GenBank Accession Number I |)re<*uliile(l iiiR.NA ( incorporated herein In reference)

AHNAK2 138420 (NM_ 138420.2 GI: : 156766049)

LCN1 002297 (NM_ _002297.3 GI: :357933613)

CTLA4 005214 (NM_ 005214.4 GI: :339276048)

SLFN5 144975 (NM_ _144975.3 GI: : 145580599)

COL4A1 001845 (NM_ _001845.4 GI: : 148536824)

OAS3 006187 (NM_ _006187.2 GI: :45007006)

TRIP 10 004240 (NM_ 004240.2 GI: :42516568)

BIK 001197 (NM_ _001197.4 GI: :347658925)

LASP1 006148 (NM_ _006148.2 GI: :219803666)

MFGE8 005928 (NM_ _005928.2 GI: : 167830474)

FCGR1A 000566 (NM_ _000566.3 GI: : 167621452)

C8orf73 NM_001100878 (NM_001100878.1 GL 154937379)

CFD NM_001928 (NM_001928.2 GL42544238)

CD14 NM_000591 (NM_000591.3 GL291575160)

C17orf49 NM_001142798 (NM_001142798.2 GL320089584)

PARP12 NM_022750 (NM_022750.2 GL52345407)

TAX1BP3 NM_014604 (NM_014604.3 GL325296950)

GPX3 NM_002084 (NM_002084.3 GL89903006)

CYTH2 NM_017457 (NM_017457.5 GL335883182)

CLTB NM_007097 (NM_007097.3 GL365906278)

VGLL3 NM_016206 (NM_016206.2 GL56550034)

KLF6 NM_001300 (NM_001300.5 GL236460447)

ERBB2 NM_001005862 (NM_001005862.1 GL54792097)

ACTA2 NM_001141945 (NM_001141945.1 GL213688374)

PDPN M_006474 (NM_006474.4 GL54792055)

CAMK4 NM_001744 (NM_001744.4 GL312596937)

GPNMB NM_001005340 (NM_001005340.1 GL52694751)

TNFAIP2 NM_006291 (NM_006291.2 GL26051239)

HEG1 NM_020733 (NM_020733.1 GL 153792109)

C10orf26 NM_017787 (NM_017787.4 GI: 144953879)

APBB1IP NM_019043 (NM_019043.3 GL56118220)

MS4A7 NM_021201 (NM_021201.4 GL46249344)

GGTA1 NR_003191 (NR_003191.1 GI: 115334680)

ITGA5 NM_002205 (NM_002205.2 GL56237028)

ALDH1B1 NM_000692 (NM_000692.4 GL346644808)

KRT5 NM_000424 (NM_000424.3 GI: 119395753)

PDLIM5 NM_006457 (NM_006457.4 GL374092019)

LOC732275 NR_024406 (NR_024406.1 GI:212276062)

CTSB NMJ47780 (NM_147780.2 GL66346647)

SH3PXD2B NM_001017995 (NM_001017995.2 GL219521929)

CAPG NM_001747 (NM_001747.3 GL371502123)

GSTK1 NM_001143679 (NM_001143679.1 GL219842316)

PEA 15 NM_003768 (NM_003768.3 GL208431812)

KIAA1618 NM_020954 (NM_020954.3 GL366039977) Gene S\ in hoi en liank Accession N umber

I |)rc»uliilc(l in NA ( incorporated herein In reference)

FAS NM_000043 (NM_000043.4 GL325652050)

IFITM2 NM_006435 (NM_006435.2 GL 151101190)

DOCK2 NM_004946 (NM_004946.2 GL205277317)

ALDH1A3 NM_000693 (NM_000693.2 GL 153266821)

DOK2 NM_003974 (NM_003974.2 GL41406049)

TYROBP NM_003332 (NM_003332.3 GL291045269)

RUNX1T1 NM_175634 (NM_175634.2 GL310923087)

EIF5AL1 NM_001099692 (NM_001099692.1 GL 153791631)

LAMP3 NM_014398 (NM_014398.3 GL 156627583)

IER2 NM_004907 (NM_004907.2 GL48675810)

SERPING1 NM_000062 (NM_000062.2 GL73858567)

PSMB10 NM_002801 (NM 002801.3 GL325652084)

C19orf66 NM_018381 (NM_018381.2 GL 154350197)

VAMP5 NM_006634 (NM_006634.2 GL31543930)

ARHGDIB NM_001175 (NM_001175.4 GL56676392)

HRH1 NM_001098213 (NM_001098213.1 GL 149158708)

CST1 NM_001898 (NM_001898.2 GL 19882250)

GIMAP4 NM_018326 (NM_018326.2 GL28416432)

MVP NM_017458 (NM_017458.3 GL312261257)

TNIP1 NM_006058 (NM_006058.4 GL356874794)

BCL2L1 NMJ38578 (NMJ38578.1 GL20336334)

CTSD NM_001909 (NM_001909.4 GL332078524)

TAP1 NM_000593 (NM_000593.5 GL53759115)

ALPP NM_001632 (NM_001632.3 GL94721245)

KIF26B NM_018012 (NM_018012.3 GL 142370197)

TOM1 NR_024194 (NR_024194.1 GL209180451)

ANXA2P2 // ANXA2P2 // ANXA2P2 NR_003573 (NR_003573.1 GL 148833516)

NEK6 NM_001145001 (NM_001145001.2 GL261244918)

ECM1 NM_004425 (NM_004425.3 GL221316613)

CLEC2B NM_005127 (NM_005127.2 GL37577106)

ARHGAP23 NM_020876

MPRIP NM_201274 (NM_201274.3 GL338968846)

GSN NM_000177 (NM_000177.4 GI: 89276753)

HTRA1 NM_002775 (NM_002775.4 GL 190014575)

EPHA1 NM_005232 (NM_005232.4 GL221316649)

SH3GL1 NM_003025 (NM_003025.3 GL317108189)

SDF4 NM_016547 (NM_016547.2 GI: 170763494)

CERCAM NM_016174 (NM_016174.4 GL 193788559)

AQP1 NM_198098 (NMJ98098.2 GL297307114)

PKM2 NM_182470 (NM_182470.2 GL332164772)

CD81 NM_004356 (NM_004356.3 GL62240999)

AIF1 M_004847 (NM_004847.3 GL201861714)

AIF1 M_004847 (NM_004847.3 GL201861714) Gene S\ in hoi en liank Accession N umber

I |)rc»uliilc(l in NA ( incorporated herein In reference)

AIF1 NM_004847 (NM_004847.3 GL201861714)

CASP1 NM_033292 (NM_033292.3 GL380254451)

LOC654342 NR_027238 (NR_027238.1 GL224549025)

GAS 7 NM_201433 (NM_201433.1 GL41406079)

SPHK1 NM_182965 (NM_182965.2 GL217272879)

POLD4 NM_021173 (NM_021173.4 GL379046943)

CHSY1 NM_014918 (NM_014918.4 GL258613868)

FAM108A1 NM_031213 (NM_031213.3 GL 194306561)

EMR2 NM_013447 (NM_013447.2 GL23397680)

CD86 NM_175862 (NM_175862.4 GL332634933)

C8orf73 NM_001100878 (NM_001100878.1 GL 154937379)

ADAMTSL3 NM_207517 (NM_207517.2 GL 145275197)

AQP7P1 NR_002817 (NR_002817.2 GL219555684)

SRPX NM_006307 (NM_006307.4 GL282721073)

CTSZ NM_001336 (NM_001336.3 GL209364552)

PDLIM7 NM_005451 (NM_005451.3 GL42741673)

DOCK8 NM_203447 (NM_203447.3 GL254750683)

TIMP2 NM_003255 (NM_003255.4 GL73858577)

PDGFA NM_002607 (NM_002607.5 GL 197333758)

GPX4 NM_002085 (NM_002085.3 GL90903236)

ACTN1 NM_001130004 (NM_001130004.1 GL 194097349)

C22orf9 NM_001009880 (NM_001009880.1 GL57863294)

HSPB1 NM_001540 (NM_001540.3 GL209969817)

SMARCD3 NM_001003802 (NM_001003802.1 GL51477703)

ACADVL NM_000018 (NM_000018.2 GL76496473)

KIAA1949 NM_133471 (NMJ33471.3 GL213417747)

IRF4 NM_002460 (NM_002460.3 GL305410879)

CCDC49 NM_017748 (NM_017748.3 GL34147582)

ASNA1 NM_004317 (NM_004317.2 GL50428937)

HABP4 NM_014282 (NM_014282.2 GL 157676330)

COL8A2 NM_005202 (NM_005202.2 GL281182524)

CSF1R NM_005211 (NM_005211.3 GL 195947380)

AQP7P1 NR_002817 (NR_002817.2 GL219555684)

LOXL1 NM_005576 (NM_005576.2 GL67782345)

HOXA4 NM_002141 (NM_002141.4 GL 148613881)

MYOIC NM_001080779 (NM_001080779.1 GL 124494237)

ASAM NM_024769 (NM_024769.2 GL41393588)

MYD88 NM_002468 (NM_002468.4 GL 197276653)

IGKC ENST00000390273

GNAL NM_182978 (NM_182978.2 GL215276940)

PIP4K2B NM_003559 (NM_003559.4 GI: 117938278)

TGM2 NM_004613 (NM_004613.2 GL39777596)

DDIT4L NM_145244 (NM_145244.3 GL307691170)

PTGFRN NM_020440 (NM_020440.2 GL41152505) Gene S\ in hoi en liank Accession N umber

I |)rc»uliilc(l in NA ( incorporated herein In reference)

1-Sep NM_052838 (NM_052838.4 GL348041284)

LMNA NM_170707 (NM_170707.3 GL383792147)

CTHRC1 NM_138455 (NMJ38455.3 GL368711290)

PCGF2 NM_007144 (NM_007144.2 GL37595566)

SH3PXD2A NM_014631 (NM_014631.2 GL55749543)

KIAA0415 NM_014855 (NM_014855.2 GL319996625)

ARRB2 NM_004313 (NM_004313.3 GL39812034)

SLITRK2 NM_032539 NM_032539.4 GL288856235

HSPA6 NM_002155 (NM_002155.3 GL42822885)

BIN2 NM_016293 (NM_016293.2 GL49472834)

MAFB NM_005461(NM_005461.3 GL31652256)

GBP1 NM_002053 (NM_002053.2 GL 166706902)

PLAT NM_000930 (NM_000930.3 GI: 132626665)

FLJ14100 AK024162 (AK024162.1 GL10436473)

MST150 NM_032947 (NM_032947.4 GL296317235)

C19orf21 BC052236 (BC052236.1 GL30353976)

CSF2RA NM_001161531 (NM_001161531.1 GL238908518)

CSF2RA NM_001161531 (NM_001161531.1 GL238908518)

AQP7P1 NR_002817 (NR_002817.2 GL219555684)

ITGB4 NM_000213 (NM_000213.3 GL54607034)

CST7 NM_003650 (NM_003650.3 GL262263313)

KIAA0125 NR_026800 (NR_026800.1 GL223453018)

ZFP92 NM_001136273 (NM_001136273.1 GL210147474)

INE1 NR_024616 (NM_001136273.1 GL210147474)

LOC441208 NR_003502 (NR_003502.2 GL225007645)

COL6A2 NM_001849 (NM_001849.3 GI: 115527061)

NCF1 NM_000265 (NM_000265.4 GI: 115298671)

ARHGAP15 NM_018460 (NM_018460.3 GL 188497641)

MLLT6 NM_005937 (NM_005937.3 GL 133908639)

GK3P NR_026575 (NR_026575.1 GL219803770)

ACER2 NM_001010887 (NM_001010887.2 GL71043497)

IRF8 NM_002163 (NM_002163.2 GL55953136)

TMEM200A NM_052913(NM_052913.2 GL40538800)

C1QTNF5 NM_015645 (NM_015645.3 GL223633881)

PODN NMJ53703 (NMJ53703.4 GL312283619)

CYTH4 NM_013385 (NM_013385.3 GL 197100456)

HIP1 NM_005338 (NM_005338.5 GL342307094)

TGFB1 NM_000660 (NM_000660.4 GL260655621)

SNX20 NM_153337 (NM_153337.2 GL222352100)

NCF1 NM_000265 (NM_000265.4 GI: 115298671)

C7orf50 NM_032350 (NM_032350.5 GL 197304706)

IFF02 NM_001136265 (NM_001136265.1 GL210147456)

LOC554223 AK128290 (AK128290.1 GL34535592)

GIMAP8 NM_175571 (NMJ75571.2 GL55953077) Gene S\ in hoi ( ien liank Accession N umber

I |)rc»uliilc(l iiiR.NA ( incorporated herein In reference)

MEF2B NM_001145785 (NM_001145785.1 GL224809411)

ADAP2 NM_018404 (NM_018404.2 GL93102369)

HSPG2 NM_005529 (NM_005529.5 GL 140972288)

ENST00000302079 Gene ID: 63895, updated on 4-Feb-

C18orf58 2012

GNAI2 NM_002070 (NM_002070.2 GL49574535)

PPP1R9B NM_032595 (NM_032595.3 GL 140972062)

GPR56 NM_201524 (NM_201524.2 GL224809311)

SLC02B1 NM_001145211 (NM_001145211.2 GL312176373)

PTGIR NM_000960 (NM_000960.3 GL39995095)

STK10 NM_005990 (NM_000960.3 GL39995095)

PLCD3 NM_133373 (NMJ33373.3 GL 116284409)

APOE NM_000041 (NM_000041.2 GL48762938)

ITGAL NM_002209 (NM_002209.2 GL 167466214)

FSCN1 NM_003088 (NM_003088.3 GL347360903)

SORBS3 NM_005775 (NM_005775.4 GL 155030229)

LOC613206 AY567967 (AY567967.1 GL50236398)

NCF1 NM_000265 (NM_000265.4 GI: 115298671)

SLITRK4 NM_173078 (NM_173078.3 GL296080771)

COL4A2 NM_001846 (NM_001846.2 GI: 116256353)

MYOIF NM_012335 (NM_012335.3 GL296278250)

WIPF1 NM_003387 (NM_003387.4 GI: 116284399)

MANIAI NM_005907 (NM_005907.2 GL24497518)

FAM38B2 ENST00000383408

GNA15 NM_002068 (NM_002068.2 GL 156104882)

HSD17B1 NM_000413 (NM_000413.2 GL223718073)

SFMBT2 NM_001029880 (NM_001029880.2 GL311771655)

LMOD1 NM_012134 (NM_012134.2 GI: 112380629)

SLC19A3 NM_025243 (NM_025243.3 GL93352566)

IGHV4-31 AK301335 (AK301335.1 GL194376247)

PCOLCE NM_002593 (NM_002593.3 GL 157653328)

GPX1 NM_201397 (NM_201397.1 GL41406081)

KIAA1671 NM_001145206 (NM_001145206.1 GL223633987)

HOXA2 NM_006735 (NM_006735.3 GL37596298)

CD1E NM_030893 (NM_030893.3 GL303304961)

INPP5D NM_001017915 (NM_001017915.1 GL64085166)

4-Sep NM_080415 (NM_080415.2 GL311201832)

FBXL7 NM_012304 (NM_012304.3 GL21071079)

AXIN2 NM_004655 (NM_004655.3 GL 195927058)

ITGAM NM_001145808 (NM_001145808.1 GL224831238)

MYOIE NM_004998 (NM_004998.3 GL368711281)

SLAMF7 NM_021181 (NM_021181.3 GL 19923571)

PDGFRB NM_002609 (NM_002609.3 GL68216043)

DAB1 AF263547 (AF263547.1 GL8118614) S\ in hoi GenBank Accession Number

I prejiiikiled mRNA ( incorporated herein In reference)

CLDN11 NM_005602 (NM_005602.5 GI: 195947370) EPAS1 NM_001430 (NM_001430.4 GL262527236)

Table IB: Reduced list of upregulated mRNAs Gene Symbol mRNA Accession

INTS4 NM 033547

NARS2 NM 024678

SNORD31 NR 002560

INTS2 NR 026641

TRIP 10 NM 004240

RASSF2 NM ^"014737

GADD45B NM 015675

LTBP2 NM 000428

MYH9 NM 002473

MMP14 NM 004995

PLAU NM 002658

OLFML2B NM 015441

THY1 NM ^"006288

CRISPLD2 NM 031476

COMP NM 000095

FNDC1 NM 032532

ITGA11 NM 001004439

IGHM BC020240

COL8A1 NM 001850

NNMT NM 006169

COL1A1 NM 000088

BGN NM 001711

INHBA NM 002192

COL11A1 NM ^"001854

Table 1C shows Group II mRNA expression signature: COPZ2, NUCBl, LPL, CCDC49, GFPT2, LOX, NNMT, RGSl, ASNAl, FXYD5, SERPP E1, KIF26B, SIOOAIO, ALDH1A3, CALB2, and PLAUR.

Table 1C:

Gene Refseq ID Known Gene Name

ALDH1A3 NM_000693 aldehyde dehydrogenase 1 family, member A3

ASNAl NM_004317 arsA arsenite transporter, ATP-binding, homolog 1

(bacterial)

CALB2 NM_001740 calbindin 2, transcript variant CALB2

CCDC49 NM_017748 coiled-coil domain containing 49 (CWC25) COPZ2 NM_. _016429 coatomer protein complex, subunit zeta 2

FXYD5 NM_. _144779 FXYD domain containing ion transport regulator 5

GFPT2 NM_. _005110 glutamine-fructose-6-phosphate transaminase 2

KIF26B NM_. _018012 kinesin family member 26B

LOX NM_. _002317 lysyl oxidase

LPL NM_. _000237 lipoprotein lipase

NNMT NM_. _006169 nicotinamide N-methyltransferase

NUCB1 NM_. _006184 nucleobindin 1

PLAUR NM_. _002659 plasminogen activator, urokinase receptor

RGS1 NM_. _002922 regulator of G-protein signaling 1

S100A10 NM_. _002966 SI 00 calcium binding protein A10

SERPINE1 NM_. _000602 serpin peptidase inhibitor, clade E (nexin,

plasminogen activator inhibitor type 1)

The mRNAs listed above, e.g., those listed in Table IB, distinguish primary and metastatic tumors and have prognostic significance on their own. When combined with the miRNA list, particularly strong patient stratification was observed. Additional exemplary upregulated mRNA transcripts from Table 1 include FAM38B, COLEC12, GFPT2, LOX, KIF26B, CALB2, RGS4, FSTL3, PDGFA, KRT5, PTGIS, RGS1, SERPINE1, NUCB1, ADAM 12, MMP16, LPL, NNMT, ASNA1, APBB1IP, FXYD5, S100A10, ALDH1A3, CD1E, ZFHX4, C10orf26, CCDC49, EMR2, FAS, ERBB2, PLAUR, and CLASS.

Suitable mRNA transcripts that are downregulated between primary and metastatic tumors include those set forth in Table 2.

TABLE 2: Top 500 Downregulated mRNAs

( ieno S.Mii bol Con Bank Accession N umber

(Dow nroyulalod ni KMA ) (incorporated herein by reference)

MTA3 NM_020744 (NM_020744.2 GL50878291)

ASCC1 NM_015947

NARG1L NM_024561 (NM 024561.4 GI: 160707983)

INTS4 NM_033547 (NM_033547.3 GL50086623)

UGGT2 NM_020121 (NM 020121.3 GL238859592)

CTH NM_001902 (NM_001902.5 GL299473757)

INTS4L2 NR_027392 (NR_027392.1 GL225637470)

LGR4 NM_018490 (NM_018490.2 GI: 157694512) Gene Symbol Gen Bank Accession Number ( Oow nrejiiihiled niR.NA) (incorporated herein by reference)

SNORD26 NR_002564 (NR_002564.1 GL74315931)

PIBF1 NM_006346 (NM_006346.2 GL55769582)

RBM26 NM_022118 (NM_022118.3 GL31652263)

INTS4L1 NR_027393 (NR_027393.1 GL225637471)

NCAPD3 NM_015261 (NM_015261.2 GL76880473)

ZNF75A NM_153028 (NM_153028.2 GI: 188528679)

CA14 NM_012113 (NM_012113.1 GL6912283)

CLASP 1 NM_015282 (NM_015282.2 GL214010171)

BLM NM_000057 (NM_000057.2 GI: 105990533)

TFDP2 NM_006286 (NM_006286.4 GL296278213)

LOC389834 // LOC389834 // LOC389834 //

LOC389834 NR_027420 (NR_027420.1 GL225903387)

PDE8B NM_003719 (NM_003719.3 GL300244572)

NARS2 NM_024678 (NM_024678.5 GL343098475)

USP14 NM_005151 (NM_005151.3 GL82880646)

RBM45 NM_152945 (NM_152945.2 GL93277069)

DPY19L3 NM_207325 (NM_207325.2 GL289666757)

N4BP2 NM_018177 (NM_018177.4 GL339275991)

AGL NM_000028 (NM_000028.2 GI: 116734846)

STXBP4 NM_178509 (NM_178509.5 GL63999047)

TMEM135 NM_022918 (NM_022918.3 GL281182570)

NM_001001415 (NM_001001415.2

ZNF429 GI: 116256454)

NM_001113378 (NM_001113378.1

FANCI GL 164607123)

NM_001001346 (NM_001001346.3

CLDN20 GL297591810)

BBS4 NM_033028 (NM_033028.4 GL358051116)

SULT1C4 NM_006588 (NM_006588.2 GL28830307)

MYEF2 NM_016132 (NM_016132.3 GI: 154146212)

ILIA NM_000575 (NM_000575.3 GL27894329)

C10orfl 8 NM_017782 (NM_017782.4 GL296011009)

NM_001128620 (NM_001128620.1

PAK1 GI: 190886456)

EFTUD1 NM_024580 (NM_024580.5 GI: 111120335) TOM1L1 AB065085 (AB065085.1 GI:21104503) STAR NM_000349 (NM_000349.2 GL56243550)

NM_001080541 (NM_001080541.2

MGA GL256017158)

DZIP1 NM_198968 (NMJ98968.2 GI: 112789558)

IP05 NM_002271 (NM_002271.4 GL51873051)

C3orf33 AK289890 (AK289890.1 GI: 158260802)

BTBD3 NM_014962 (NM_014962.2 GL31317210)

OTUD3 NM_015207 (NM_015207.1 GI: 149192870)

CIRH1A NM_032830 (NM_032830.2 GI: 186928846)

C2orf56 M_144736 (NM_144736.4 GL 145701026)

YES1 NM_005433 (NM_005433.3 GL51702529)

SMARCA1 NM_003069 (NM_003069.3 GL 164419747)

ATE1 NM_001001976 (NM_001001976.1 GL50345876)

NFXL1 NM_152995 (NM_152995.4 GL89363019)

PTPN4 NM_002830 (NM_002830.2 GI: 18104987)

LOC100129038 ENST00000391432

TTC14 NM_133462 (NM_133462.3 GL 197333813)

BXDC1 NM_032194 (NM_032194.1 GL39930468)

TUBD1 NM_016261 (NM_016261.3 GL302191698)

CCDC112 NM_152549 (NM_152549.2 GI: 145580594)

CCNT2 NM_058241 (NM_058241.2 GL315075318)

PHIP NM_017934 (NM_017934.5 GL224589121)

ATF2 NM_001880 (NM_001880.3 GL368711267)

ZNF431 NM_133473 (NM_133473.2 GL 194239710)

LRRC28 M_144598 (NMJ44598.2 GL33285014)

BLZF1 NM_003666 (NM_003666.2 GL37622354)

C13orf34 NM_024808 (NM_024808.2 GL38505206)

SERF 1 A NM_021967 (NM_021967.2 GI: 196162706)

SERF 1 A NM_021967 (NM_021967.2 GI: 196162706)

SERF 1 A NM_021967 (NM_021967.2 GI: 196162706)

INTS2 NR 026641 (NR 026641.1 GL221136797)

NM_001127202 (NM_001127202.1

PCID2 GL224451000)

PAAF1 NM_025155 (NM_025155.1 GI: 13376750)

HDAC1 M_004964 (NM_004964.2 GL 13128859)

THUMPD1 NM_017736 (NM_017736.3 GL62865619)

SNORD47 NR_002746 (NR_002746.1 GI: 84872026)

TIPIN NM_017858 (NM_017858.2 GL 157388909)

MBTD1 NM_017643 (NM_017643.2 GL 158508475)

XP07 NM_001100161

FDX1 NM_004109 (NM_004109.4 GL342307102)

ZNF529 NM_020951 (NM_020951.4 GL224549027)

MIR17HG NR_027350 (NR_027350.1 GL224994261)

SULT1C2 NM_001056 (NM_001056.3 GL45935387)

MFSD11 NM_024311 (NM_024311.3 GL335892876)

SESN3 M_144665 (NM_144665.2 GL31377590)

GPR180 NMJ 80989 (NM_180989.5 GL324710989)

PRKAB2 NM_005399 (NM_005399.3 GL46877069)

ZDHHC20 NM_153251 (NMJ53251.3 GL312147398)

KLHL8 NM_020803 (NM_020803.3 GL34101267)

SFRS12IP1 NM_173829 (NM_173829.3 GI: 115529440)

LACTB2 NM_016027 (NM_016027.2 GL208973264)

ZNF26 NM_019591 (NM_019591.3 GL372266176)

BUB IB NM_001211 (NM_001211.5 GI: 168229167)

NM 001136156 (NM 001136156.1

ZNF507 GL209954810)

NUSAP1 NM_016359 (NM_016359.4 GL341865557)

COMMD8 NM_017845 (NM_017845.3 GL 141801729)

PACRGL NMJ45048 (NM_145048.3 GL 195539373)

DBR1 NM 016216 (NM 016216.3 GL302129646)

NM 001039724 (NM 001039724.3

NOSTRIN GL284172480)

NOL11 NM 015462 (NM 015462.3 GL 142378677)

NM 001104647 (NM 001104647.1

SLC25A36 GL 157388988)

NM 001006607 (NM 001006607.2

LRRC37A2 GI: 116325992)

THOC1 NM_005131 (NM_005131.2 GI: 154448889)

FAM169A NM_015566 (NM_015566.2 GL380861664)

MAP4K3 NM_003618 (NM_003618.2 GI: 15451901)

SLC24A5 NM_205850 (NM_205850.2 GL84697035)

HIST1H2BD NM_021063 (NM_021063.3 GL342837697)

ZNF141 NM_003441 (NM_003441.2 GI: 187608430)

SUPV3L1 NM_003171 (NM_003171.3 GL 142366966)

TIAL1 NM_001033925 (NM_001033925.1 GL77695911)

ORC4L NM_002552 (NM_002552.4 GL308818135)

ANKRD36 AK304740 (AK304740.1 GL194388231)

KIAA0564 NM 015058 (NM 015058.1 GL57863270)

NM 001042683 (NM 001042683.2

SHPRH GL289547540)

LRRC37B NM_052888 (NM_052888.2 GL53829384)

BIRC6 NM_016252 (NM_016252.3 GL 153792693)

SNORA38 NR_002971 (NR_002971.1 GL91680838)

KIAA1919 NM_153369 (NM_153369.2 GI: 154937329)

SSBP2 NM_012446 (NM_012446.3 GL376319225)

CKMT2 NM_001825 (NM_001825.2 GL 153251981)

TRIP 13 NM_004237 (NM_004237.3 GL261337159)

MANSC1 NM_018050 (NM_018050.2 GL31542649)

GPR52 NM_005684 (NM_005684.4 GI: 187960054)

MAP3K7 NM_145331 (NMJ45331.1 GL21735561)

NSUN7 NM_024677 (NM_024677.4 GI: 187828634)

SCARNA9L NR_023358 (NR_023358.1 GL 189181650)

NIPBL NM_015384 (NM_015384.4 GL 189163520)

TBP NM_003194 (NM_003194.4 GL285026518)

UBR3 NM_172070 (NM_172070.3 GI: 160948609)

FLVCR1 NM_014053 (NM_014053.3 GL321116986)

PGM5 NM_021965 (NM_021965.3 GL68299819)

C2orf64 BC047722 (BC047722.1 GL28839513)

MTF2 NM_007358 (NM_007358.3 GI: 166706893)

RAD17 NM_133338 (NMJ33338.1 GL 19718783) Gene Symbol GenBank Accessions Number

(Downregulated mRNA) (incorporated herein by reference)

PIGU NM_080476 (NM_080476.4 GL52426746)

C2orf3 NM_003203 (NM_003203.4 GL256017127)

LANCL1 NM_006055 (NM_006055.2 GL212276218)

ALDH9A1 NM_000696 (NM_000696.3 GI: 115387103)

ACSM3 NM_005622 (NM_005622.3 GL47458816)

KIF6 NM_145027 (NMJ45027.4 GL 157502183)

CTDSPL2 NM_016396 (NM_016396.2 GL 100815974) ARHGAP1 IB NM_001039841 (NM_001039841.1 GI:89886349) TAFIA NM_005681 (NM_001039841.1 GL89886349) EIF2A NM_032025 (NM_032025.3 GL83656780) TMEM128 NM_032927 (NM_032927.2 GL39725660)

NM_001079515 (NM_001079515.1

TBCE GI: 118442827)

RP1 -199H16.1 AK301819 (AK301819.1 GL221043769)

FGFRI OP NM_007045 (NM_007045.2 GL36287087)

PRDX3 NM_006793 (NM_006793.2 GL32483378)

KBTBD10 NM_006063 (NM_006063.2 GL42741668)

AQR NM_014691 (NM_014691.2 GL58374127)

USP6NL NM_014688 (NM_014688.2 GL221219073)

NM_001 130862 (NM_001130862.1

RAD51AP1 GI: 195947390)

LRRC37A3 NMJ99340 (NM_199340.2 GL75677611)

RSF1 NM_016578 (NM_016578.3 GL38788332)

CREM NM_183013 (NMJ 83013.2 GL359465528)

GGH NM_003878 (NM_003878.2 GL215422332)

UNQ1829 AY358798 (AY358798.1 GL37182714)

TIGD2 NMJ45715 (NM_145715.2 GI: 124256528)

UBR1 NM_174916 (NM_174916.2 GL83656781)

SNORA29 NR_002965 (NR_002965.1 GI:91680832)

SASS6 NM_194292 (NMJ94292.1 GL35038600)

ELOVL7 NM_024930 (NM_024930.2 GL 157388946)

PDSS1 NM_014317 (NM_014317.3 GL50659085)

HACL1 NM_012260 (NM_012260.2 GL93004077)

TMEM67 NM_153704 (NMJ53704.5 GL214830731)

EHHADH NM_001966 (NM_153704.5 GL214830731)

PTCD3 NM_017952 (NM_017952.5 GL203097228)

EPB41L5 NM_020909 (NM_020909.3 GL296531460)

LYRM7 NM_181705 (NM_181705.2 GI:91 176332)

SF3B3 NM_012426 (NM_012426.4 GL225579068)

RAD17 NMJ33338 (NMJ33338.1 GL 19718783)

TMED4 NM_182547 (NM_182547.2 GL33457307)

MEAF6 BX538212 (BX538212.1 GL31874678)

FAM165B NM_058182 (NM_058182.4 GL 149363652)

SLC30A5 NM_022902 (NM_022902.4 GL354548831)

SEC23IP NM_007190 (NM_007190.3 GL321400046) Gene S\ mbol Gen Bank Accession Number

« iirejiiilaled iii RN.A ) (incorporated herein by reference)

TRMT11 NM_001031712 (NM_001031712.2 GL94420682)

EIF2C3 NM_024852 (NM_024852.3 GL324073530)

BCKDHB NM_183050 (NM_183050.2 GI: 168480068)

ITFG1 NM_030790 (NM_030790.3 GI: 141803256)

TOMM22 NM_020243 (NM_020243.4 GL56788357)

MFN1 NM_033540 (NM_033540.2 GL45269136)

PRPS1 NM_002764 (NM_002764.3 GI: 164663865)

PSME4 NM_014614 (NM_014614.2 GI: 163644282)

KATNAL2 NM_031303 (NM_031303.2 GL226371753)

ITLN1 NM_017625 (NM_017625.2 GL31542985)

MIPOL1 NM_138731 (NMJ38731.6 GL305682559)

FSIP2 AK092099 (AK092099.1 GI:21750609)

C10orfl04 NMJ73473 (NM_173473.3 GL335892799)

UCHL5 NM_015984 (NM_015984.3 GL312922357)

ZNF639 NM_016331 (NM_016331.1 GL7705934)

TFB2M NM_022366 (NM_022366.2 GL210147473)

TM9SF2 NM_004800 (NM_004800.1 GI.-4758873)

HLTF NM_003071 (NM_003071.3 GI: 1 17968550)

ABCA11P NR_002451 (NR_002451.2 GI: 187608333)

Clorfl24 NM_001010984 (NM_001010984.1 GI:58331106)

NM_001039083 (NM_001039083.3

ARL17 GL 156938236)

SNORD28 NR_002562 (NR_002562.1 GL74315928)

WNK3 NM_020922 (NM_020922.4 GL296080744)

HSN2 NM_213655 (NM_213655.4 GL300797779)

MANEA NM_024641 (NM_024641.3 GL335057564)

SNRPG NM_003096 (NM_003096.2 GI:21359839)

PLBD1 NM_024829 (NM_024829.5 GI: 1 10227597)

CACYBP NM_014412 (NM_014412.2 GL55925645)

LRRC34 NM_153353 (NM_153353.4 GL289666777)

TXNRD1 NM_003330 (NM_003330.2 GL33519431)

NM_001048166 (NM_001048166.1

STIL GI: 115298662)

SNRPA1 BC067846 (BC067846.1 GL45709272)

MIB1 NM_020774 (NM_020774.2 GL62868229)

SNORD31 NR_002560 (NR_002560.1 GL74315926)

TET1 NM_030625 (NM_030625.2 GI: 156139121)

TIPARP NM_015508 (NM_015508.4 GL296080689)

SLC38A9 NM_173514 (NMJ73514.2 GL222418628)

FLJ32742 AK057304 (AK057304.1 GI: 16552940)

NM_001079839 (NM_001079839.2

OCIAD1 GL269914124)

GPR89B NM_016334 (NM_016334.3 GL 148230342) ACBD5 NM_145698 (NM_145698.3 GL203098628) CCDC76 NM_019083 (NM_019083.2 GL226493202) in S in bo I Gen Bank Accession Number w iircnulaled III R.NA ) (incorporated herein by reference)

NM 001006607 ( NM 001006607.2

LRRC37A2 GI: 116325992)

UBXN8 NM_005671 (NM_005671.2 GL87080808)

SLC44A5 NM_152697 (NM_152697.4 GI: 194239630)

Clorf25 NM_030934 (NM_030934.4 GI: 163792193)

GPR89B NM_016334 (NM_016334.3 GL 148230342)

ENPP5 NM_021572 (NM_021572.4 GL38455386)

DZIP3 M_014648 (NM_014648.3 GL 197304766)

NM_001098783 (NM_001098783.2

RPP14 GL 153791326)

MGC57346 NR_026680 (NR_026680.2 GL224593261)

ESSPL NM_183375 (NM_183375.2 GI: 148806899)

EMG1 NM_006331 (NM_006331.7 GL296010913)

SNORD50B NR_003044 (NR_003044.3 GI: 157057556)

6-Mar NM_005885 (NM_005885.2 GL33589845)

TUBE1 NM_016262 (NM_016262.4 GL223278354)

IPO 11 NM_016338 (NM_016338.4 GL 198041774)

LRRC37A NM_014834 (NM_014834.4 GL289547511)

SACM1L NM_014016 (NM_014016.3 GI: 190014577)

DDHD2 NM_015214 (NM_015214.2 GL256017244)

FUSIP1 NM_006625 (NM_006625.4 GL300360542)

DTNA NM_001390 (NM_001390.4 GL 189571584)

SLC30A9 NM_006345 (NM_006345.3 GL57164947)

SNORA4 NR_002588 (NR_002588.1 GL77993294)

LYAR NM_017816 (NM_017816.2 GL224593253)

SMCHD1 NM_015295 (NM_015295.2 GL 148839304)

NM_001097612 (NM_001097612.1

GPR89A GL 148228818)

PMCHL2 NR_003922(NR_003922.1 GI: 157278030)

RAVER2 NM_018211 (NM_01821 1.3 GL 197927)

C8orf38 NM_152416 (NM_152416.3 GL356582231 )

TXNRD3IT1 NM_001039783

RNF7 NM_014245 (NM_014245.4 GL319004059)

CCDC138 M_144978 (NM_144978.1 GL21450664)

ATP 13 A3 NM_024524 (NM_024524.3 GI: 148839291)

ENPP4 NM_014936 (NM_014936.4 GL 194688140)

NM_001080449 (NM_001080449.2

DNA2 GL320461727)

Clorf27 AY854248 (AY854248.1 GL56790004)

SSR3 NM_007107 (NM_007107.3 GI:302129655)

NM_001144013 (NM_001 144013.1

RGPD3 GL221307606)

NM_001100624 (NM_001100624.1

CENPN GI: 154800482)

ZW10 NM_004724 (NM_004724.3 GL342837718)

0SGEPL1 NM_022353 (NM_022353.2 GI: 116812635) (icue S\ mbol GenBank Accessions Number wnregulated III RNA) (incorporated herein by reference)

PMCHL2 NR_003922 (NR_003922.1 GI: 157278030)

DHX29 NM_019030 (NM_019030.2 GL67782361)

APH1B NM_031301 (NM_031301.3 GL224548985)

WDR76 NM_024908 (NM_024908.3 GL269847397)

SNORD25 NR_002565 (NR_002565.1 GL74315927)

CALML4 NM_033429 (NM_033429.2 GI: 1 10227593)

CDC40 NM_015891 (NM_015891.2 GI:38570087)

CLDN12 NM_012129 (NM_012129.4 GI:313851094)

HSPC159 NM_014181 (NM_014181.2 GI: 156151365)

ZCCHC17 NM_016505 (NM_016505.2 GL21361575)

SNORD80 NR_003940 (NR_003940.1 GL 157738632)

DYNC2LI1 NM_016008 (NM_016008.3 GL301500718)

CRYZL1 NM_145858 (NM_145858.2 GL90669510)

UBE2V2 NM_003350 (NM_003350.2 GI: 12025664)

RNF2 NM_007212 (NM_007212.3 GL54792140)

GTPBP8 NM_014170 (NM_014170.2 GL56549684)

SNORA56 NR_002984(NR_002984.1 GI:91680851)

UBR2 NM_015255 (NM_015255.2 GL296317266)

NM_001080481 (NM_001080481.1

USP45 GI: 122937411)

SNORD30 NR_002561 (NR_002561.1 GL74315923)

HEATR1 NM_018072 (NM_018072.5 GL301069317)

IRAKI BP 1 NM_001010844 (NM_001010844.1 GL59797062)

KLHL15 NM_030624 (NM_030624.2 GL226442728)

ATPBD4 NM_080650 (NM_080650.3 GI:213972613)

CCDC68 NM_025214 (NM_025214.2 GI:219689134)

NM_001024599 (NM_001024599.3

HIST2H2BF GL238624108)

SLC15A2 NM_021082 (NM_021082.3 GL226371745)

FBXL4 NM_012160 (NM_012160.3 GL21536437)

ANAPC1 NM_022662 (NM_022662.3 GL359338995)

C16orf80 NM_013242 (NM_013242.2 GL42716281)

STRADB NM_018571 (NM_018571.5 GL 130979227)

UBE2E1 NM_003341 (NM_003341.4 GL321 1 17162)

ZNF850P BC052603 (BC052603.1 GL30851223)

NUP35 NM_138285 (NM_138285.3 GI:56788372)

TMEM57 NM_018202 (NM_018202.4 GL209977071)

SAAL1 NM_138421 (NMJ38421.2 GI: 1 16235443)

TAS2R50 NM_176890 (NM_176890.2 GI: 154937356)

SELI NM_033505 (NM_033505.2 GI: 144094257)

NM_001093755 (NM_001093755.1

C5orf44 GI: 148277001)

TRMT61B NM_017910 (NM_017910.3 GL222831586) MRS2 NM_020662 (NM_020662.2 GL93204868) CEP57 NM_014679 (NM_014679.4 GL344925821) Gene Symbol Gen Bank Accession Number (Downregulated mRNA) (incorporated herein by reference)

LIFR NM_002310 (NM_002310.5 GI: 189095278) C16orf63 AY507846 (AY507846.1 GL41057618) FLJ44048 BX648733 (BX648733.1 GL34367897) ZNF66 BC067843 (BC067843.1 GL45709646) MEDIO NM_032286 (NM 032286.2 GL49227853) MAPK6 NM_002748 (NM_002748.3 GL213972644) EXOC6 NM_019053 (NM_019053.4 GI:219842254) POLR2G NM_002696 (NM_002696.2 GI:219879812) MRPS23 NM_016070 (NM_016070.3 GL312222785) CEP70 NM_024491 (NM_024491.2 GL50658081) CCDC52 NMJ44718 (NMJ44718.3 GL 193211455) NUPL2 NM_007342 (NM_007342.2 GI: 197245407) CDC123 NM_006023 (NM 006023.2 GL221316619) ARP1 1 AB039791 (AB039791.1 GI: 12583651) FLJ44048 AK126104 (AK126104.1 GL34532477) TFB1M NM_016020 (NM_016020.3 GL315013580) MTMR2 NR_023356

CUL5 NM_003478 (NM_003478.3 GL67514034) ATF7IP2 NM_024997 (NM_024997.3 GL371872959) NUP155 NM_153485 (NMJ 53485.1 GL24430148) PSMD14 NM_005805 (NM 005805.5 GL375268695) TMEM62 NM_024956 (NM_024956.3 GL52851436)

OR1F2P // OR1F2P // OR1F2P // OR1F2P NR_002169 (NR_002169.1 GL52421330)

SNORD58A NR_002571 (NR_002571.1 GL77735353)

HSDL1 NM_031463 (NM_031463.4 GL226371730)

PCBD1 NM_000281 (NM_000281.2 GL50086629)

SNORD38A NR_001456 (NR_001456.1 GL32526880)

SNORD75 NR_003941 (NR_003941.1 GL 157738634)

TMEM69 NM_016486 (NM_016486.3 GI: 118918408)

GSTA1 NM_145740 (NMJ45740.3 GL215276985)

NBN NM_002485 (NM_002485.4 GL67189763)

LRIG3 NM_153377 (NM_153377.3 GL40255156)

ZFAND6 NM_019006 (NM_019006.3 GL339276057)

ACAT2 NM_005891 (NM_005891.2 GI: 148539871)

TAF1B NM_005680 (NM 005680.2 GL205360986)

NM_001085411 (NM_001085411.1

C5orf33 GL 146134340)

SMCHD1 NM_015295 (NM_015295.2 GL 148839304)

CAPS2 NM_032606 (NM_032606.3 GI: 166795237)

ANKAR NMJ44708 (NM_144708.3 GL284413733)

HOOK1 NM_015888 (NM_015888.4 GL95007031)

RAB4A NM_004578 (NM_004578.2 GI: 19923259)

SNORD27 NR_002563 (NR_002563.1 GL74315925)

ARL5B NM_178815 ( M_178815.3 GL59858805)

SLC33A1 NM_004733 (NM_004733.3 GL300360494)

GPR125 NMJ45290 (NM_145290.2 GL59823630)

XRN1 NM_019001 (NM_019001.3 GI: 110624791)

NM_001042416 (NM_001042416.1

ZNF596 GI: 109240543)

FAM126B NMJ73822 (NMJ73822.3 GL214832287)

HEATR5B NM_019024 (NM_019024.1 GL55749741)

MBNL3 NM_018388 (NM_018388.3 GL282720999)

SKP2 NM_005983 (NM_005983.3 GL340805878)

ANKRD31 AK097510 (AK097510.1 GI:21757297)

NBR1 NM_031858 (NM_031858.2 GI: 112382229)

LONRF2 NMJ98461 (NM_198461.3 GL 148528974)

ATR NM_001184 (NM_001184.3 GI: 157266316)

CA8 NM_004056 (NM_004056.4 GL56121820)

CPSF3 NM_016207 (NM_016207.2 GL21314666)

PBK NM_018492 (NM_018492.2 GI: 18490990)

NM_001017975 (NM_001017975.3

HFM1 GI: 130484566)

ANAPC1 NM_022662 (NM_022662.3 GL359338995)

C6orf70 NM_018341 (NM_018341.1 GL62988330)

APPBP2 NM_006380 (NM_006380.2 GI: 18104961)

HIST1H4H NM_003543 (NM_003543.3 GL21264599)

PFN2 NM_053024 (NM_053024.3 GL94538348)

Clorfl l4 BC026073 (BC026073.2 GL34190387)

CCDC14 NM_022757 (NM_022757.4 GI: 169646280)

SNORD79 NR_003939 (NR_003939.1 GL 157738631)

KIF18A NM_031217 (NM_031217.3 GL 148612830)

SNORA1 NR_003026 (NR_003026.1 GL93352542)

NEK2 NM_002497 (NM_002497.3 GL323510686)

SRD5A3 NM_024592 (NM_024592.4 GL325651871)

PEX3 NM_003630 (NM_003630.2 GL 196259812)

GPN3 NM_016301 (NM_016301.3 GL256818741)

TBL1XR1 NM_024665 (NM_024665.4 GL 166091478)

ATG5 NM_004849 (NM_004849.2 GL92859692)

CHST9 NM_031422 (NM_031422.5 GL372622387)

ORC6L NM_014321 (NM_014321.3 GL313661351)

PDCD4 NMJ45341 (NMJ45341.3 GL313760535)

SKA2 NM_182620 (NM_182620.3 GI: 154689645)

C3orf67 BC050317 (BC050317.1 GL30046309)

CEP97 NM_024548 (NM_024548.2 GI:31377704)

GNPAT NM_014236 (NM_014236.3 GL 170650722)

REV3L NM_002912 (NM_002912.3 GL 153792011)

KPNA1 NR_026698 (NR_026698.1 GL222144296)

SNRPB2 NM_003092 (NM_003092.4 GL300795278)

STT3A NMJ52713 (NMJ52713.3 GI: 194097383)

RCHY1 NM_015436 (NM_015436.3 GL324120885)

SNORD45B NR_002748 (NR_002748.1 GL84872028)

PIGT ENST00000425680

COPS2 NM_004236 (NM_004236.3 GL219842260)

ALG8 NM_024079 9 NM_024079.4 GL91984776)

SEC62 NM_003262 (NM_003262.3 GL61744441)

VPS54 NM_016516 (NM_016516.2 GL54234033)

PIGN NM_176787 (NM_176787.4 GI: 134244279)

SCML2 NM_006089 (NM_006089.2 GI: 193794845)

NM_001080463 (NM_001080463.1

DYNC2H1 GI: 122937397)

NM_001040153 (NM_001040153.3

SLAIN1 GL338968857)

NDUFB3 NM_002491 (NM_002491.2 GI: 161169037)

NDUFB5 NM_002492 (NM_002492.3 GL316659417)

ARV1 NM_022786 (NM_022786.1 GI: 12232478)

LBR NM_002296 (NM_002296.3 GL328447199)

RARS2 NM_020320 (NM_020320.3 GI: 197100772)

SFT2D1 NM_145169 (NMJ45169.1 GL21553316)

NM_001040448 (NM_001040448.2

DEFB131 GI: 149944550)

SNORA33 NR_002436 (NR_002436.1 GL71361645)

MPP6 M_016447 (NM_016447.2 GL21361597)

AKR1C3 NM_003739 (NM_003739.5 GL359806986)

PREPL NM_006036 (NM_006036.4 GL284172362)

E2F5 NM_001951 (NM_001951.3 GI: 134142810)

DENND5B NMJ44973 (NM_144973.3 GI: 122891861)

TROVE2 NM_004600 (NM_004600.5 GL291084622)

TCP1 NM_030752 (NM_030752.2 GL57863256)

NM_001142725 (NM_001142725.1

ERI2 GL218505679)

SPC25 NM_020675 (NM_020675.3 GL23510353)

KLHL7 NM_018846 (NM_018846.4 GL289063408)

NOL10 NM_024894 (NM_024894.2 GI: 171460957)

DPMI NM_003859 (NM_003859.1 GL4503362)

GEN1 NMJ 82625 ( M_182625.3 GL 194018530)

Cl lorflO NM_014206 (NM_014206.3 GI: 194018428)

NMD3 NM_015938 (NM_015938.3 GL 142359942)

HERC4 NM_022079 (NM_022079.2 GI: 145386556)

TARBP1 NM_005646 (NM_005646.3 GI: 110825987)

MLLT10 NM_004641 (NM_004641.3 GL307548830)

DEFB109P1B NR_003668 (NR_003668.2 GL222446630)

COG2 NM_007357 (NM_007357.2 GL223029377)

SNORD14C NR_001453 (NR_001453.2 GL255308891)

SH3BGRL2 NM_031469 (NM_031469.2 GI: 141803116)

RICTOR NMJ52756 (NMJ52756.3 GL71043496)

NM_001004317 (NM_001004317.2

LIN28B GL72535166)

B3GALNT1 NM_001038628 (NM_001038628.1 GL84452145) ZNF486 NM_052852 (NM_052852.3 GL375065864) ACTL6A NM_178042 (NM_178042.2 GL98985781)

NM_001145345 (NM_001145345.1

ZNF566 GL223890214)

C20orfi0 NM_001009924 (NM_001009924.1 GL58331121)

SLC35F5 NM_025181 (NM_025181.2 GL21361958)

RANBP2 NM_006267 (NM_006267.4 GI: 150418006)

PRKRIR NM_004705 (NM_004705.2 GI: 19923267)

C7orf44 NM_018224 (NM_018224.2 GL90855770)

SLC35A1 NM_006416 (NM_006416.4 GI: 194578912)

DERL2 NM_016041 (NM_016041.3 GI: 141802997)

AMNl NR_004854 (NR_004854.1 GI: 164663799)

DNAH6 NM_001370 (NM_001370.1 GL 194353965)

BCCIP NM_016567 (NM_016567.3 GI: 169790844)

HIST1H4E NM_003545 (NM_003545.3 GL21264600)

TBPL1 NM_004865 (NM_004865.3 GL358439416)

ATP13A4 NM_032279 (NM_032279.2 GL66932948)

STOX1 NM_152709 (NMJ52709.4 GL239735514)

NM_001145432 (NM_001145432.1

LOC389203 GL224177527)

RANBP17 NM_022897 (NM_022897.3 GI:313569857)

CYCS NM_018947 (NM_018947.5 GL300863084)

YME1L1 NM_139312 (NMJ39312.2 GL359718984)

TUBB4 NM_006087 (NM_006087.2 GL21361321)

NM_001145154 (NM_001145154.1

DNAH14 GL223555932)

UBE2K NM_005339 (NM_005339.4 GL 163660383)

NM_001099676 (NM_001099676.2

C12orf56 GL218505691)

WDR75 NM_032168 (NM_032168.1 GL29789282)

GSTM3 NM_000849 (NM_000849.4 GI:215277001)

SERPINA3 NM_001085 (NM_001085.4 GL73858562)

C6orfl62 NM_001042493 (NM_001042493.1 GI: 1099483)

ADAM32 NMJ45004 (NM_145004.5 GI: 148664237)

NME7 NM_013330 (NM_013330.3 GL37574616)

ARL6IP6 NM_152522 (NM_152522.5 GL343966413)

FBX016 NM_172366 (NM_172366.2 GL30089920)

CNOT7 NM_013354 (NM_013354.5 GL85067506)

NDUFAB1 NM_005003 (NM_005003.2 GL39753951)

C6orfl30 AJ420538 (AJ420538.1 GL 17066402)

NSUN6 NM_182543 9 NM_182543.2 GL 197382604)

ANK3 NM_020987 (NM_020987.3 GL325053663)

POPDC3 NM_022361 (NM_022361.4 GL215277003)

FAM72D AB096683 (AB096683.1 GL47826474)

KLHL23 NM 144711 (NM 144711.5 GL313151205)

NM 001099670 (NM 001099670.1

C8orf59 GI: 153792326)

KCNJ16 NM_170742 (NM_170742.1 GL25777629)

WRN NM_000553 (NM_000553.4 GI: 182507163)

FAM72D AB096683 (AB096683.1 GL47826474)

NOC3L NM 022451 (NM 022451.9 GI:31377626)

NM 001017975 (NM 001017975.3

HFM1 GI: 130484566)

RPRD1A NM_018170 (NM_018170.3 GL 142385371)

FAM72D AB096683 (AB096683.1 GL47826474)

PRTFDC1 NM_020200 (NM_020200.5 GL40255014)

ATL2 NM_022374 9 NM_022374.2 GL208609999)

IGF2BP3 NM_006547 (NM_006547.2 GL30795211)

SLC30A6 NM_017964 (NM_017964.3 GL301898370)

STARD3NL NM_032016 (NM_032016.3 GL347446704)

CXCL2 NM_002089 (NM_002089.3 GI: 148298657)

NQOl NM_000903 (NM_000903.2 GL70995356)

NHLRC2 NM_198514 (NMJ98514.3 GL304307770)

SNORD82 NR_004398 (NR_004398.1 GL 161087018)

PSMA2 NM_002787 (NM_002787.4 GI: 156071494)

BXDC2 NM_018321 (NM_018321.3 GL55770899)

ASAH2 NM 019893 (NM 019893.2 GL221218980)

NM 001083926 (NM 001083926.1

ASRGL1 GI: 145275201)

SNORD54 NR 002437 9 NR 002437.1 GL71533177)

NM 001135674 (NM 001135674.1

C8orf40 GL208610012)

THNSL1 NM_024838 (NM_024838.4 GL 153792147)

PRKAA2 NM_006252 (NM_006252.3 GI: 157909838)

ZICl NM_003412 (NM_003412.3 GL70778758)

The mRNA expression signatures described in Tables 1 and 2 are also suitable miRNA targets. For example, miRNA 193a-5p targets include SPTBNl (NM_003128.4;

GI: 112382249); CTA-216E10.6 (NM 001142964.1; GI:219282704); ZC3H6

(NMJ98581.1; GI: 118766346); (NCALD (NM 001040624.1; GL98985782); SF3B3

(NM_012426.4; GL225579068); HPS4 (NM_022081.4; GT.23110965); IGSF9B

(NM_014987.1; GL 148886751); C3orf75 (NM_001031703.2; GL89145414); (NHEJl

(NM_024782.2; GI: 187607429); KIAA1715 (NM_030650.1; GL38176150); ARPP19

(NM_006628.4; GL56549669); ATMIN (NM_015251.2; GL54792091) TXLNB

(NMJ53235.3; GL222537713); DAPK1 (NM_004938.2; GL89363046); ZNF704 (ΝΜ_001033723.2; GL300796181); Clorf21 (NM_030806.3; GL58761542); FAM168B (NM 001009993.2; GI: 1 12734856); (NOL9 (NM 024654.4; GL282165748); BHLHB9 (NM_001142524.1; GL216547638); FAM171A1 (NM_001010924.1; GL63025205);

ATG9A (NM OO 1077198.1; GI: 116089281); EEF1A1 (NM 001402.5; GL83367078); C0X19 (NM_001031617.2; GI: 110349770); ZNF33B (NM_006955.1; GL24307874);

FAM131B (NM OO 1031690.2; GI: 164565451); MRPS27 (NM_015084.2; GI: 186928849); CSNK1G1 (NM_022048.3; GL98986449); AD ATI (NM_012091.3; GL302058268);

TSPYL5 (NM_033512.2; GL49410494 ); MBD6 (NM_052897.3; GL46852160); ZNF81 (NM 007137.3; GL296010943); PTPRM (NM OO 1105244.1; GL 157419151); XYLB (NM 005108.3; GL255760023); DAP (NM 004394.2; GI: 149999367); PFKFB2

(NM_006212.2; GL64762405); PTER (NM_001001484.1; GL47933340); UNK

(NM OO 1080419.2; GI:331028524); PIP4K2A (NM 005028.4; GL 156416001); FXN (NM 000144.4; GL239787167); PSD3 (NM 015310.3; GI: 117606359); PTCD3

(NM_017952.5; GL203097228); C20orfl l (NM_017896.2; GL40804466). miRNA-21 targets include DNAJC18 (NMJ52686.3; GL301500703); (NARG1L (NM_024561.4; GI: 160707983); GPRASP2 (NM OO 1004051.3; GI:315360624); RALGPS2 (NM 152663.3; GL209977064); CENPQ (NM_018132.3; GL46255020); SFRS2IP (NM_004719.2;

GI: 117676383); RTF1 (NM_015138.4; GL 195976781); U2AF1 (NM_001025204.1;

GL68800137); FLJ23834 (NMJ52750.4 GL 194018461); FMNL3 (NM_175736.4;

GI: 119120873); EXTL3 (NM_001440.2; GL41281366); ZAK (NM_133646.2;

GL82880649); C0X18 (NMJ73827.2; GL76253916); ZNF609 (NM_015042.1;

GL71725359); ARMCX1 (NM_016608.1; GL7706142); BCL2 (NM_000633.2;

GL72198188 ); SNX29 (NM OO 1080530); KBTBD7 (NM_032138.4; GI: 142388608); C6orfl82 (NM_001083535.1; GL 134133245); ATXN7L1 (NM_152749.3; GI: 163937852); FBNl (NM_000138.4; GL281485549); TBC1D4 (NM_014832.2; GI: 114688045); KLHL15 (NM_030624.2; GL226442728); MCC (NM_001085377.1; GI: 146229339); ITPR2

(NM_002223.2; GI:95147334); GSTM3 (NM_000849.4; GL215277001); MRPL49

(NM_004927.3; GL312176379); SF3B3 (NM_012426.4; GL225579068); THAP5

(NM_001130475.1; GI: 194440667); TAF8 (NMJ38572.2; GI: 104485445); AGAP6 (NM_001077665.2; GI: 166851845); ZRANB1 (NM_017580.2; GI: 110815808); PNRC2 (NM_017761.3; GL209969688); PNRC2 (NM_017761.3; GL209969688); AN06

(NM_001025356.2; GL218156298); RANBP3L (NM_001161429.1; GL239582744); SYT13 (NM_020826.2; GI: 187936944); PAPD5 (NM_001040284.2; GL256818779); P0M121 (NM 172020.3; GL380692336); SGTB (NM 019072.2; GI:95147358); RG9MTD3 (NM 144964.2; GI: 117606329); YTHDC1 (NM 001031732.2; GL94536801); PRTFDC1 (NM_020200.5; GL40255014); FM02 (NM_001460.2; GL59938777); RASA1

(NM 002890.2; GL347360924); ZDHHC2 (NM 016353.4; GL209180414); CEP68 (NM_015147.2; GI: 154800456); PJA2 (NM_014819.4; GI: 157412254); ADNP

(NM_015339.2; GI:31563504); ATRNL1 (NM_207303.2; GI: 109715850); VPS52

(NM_022553.4; GL73747798); PTPN9 (NM_002833.2; GL 18375663); POM121

(NM_172020.3; GL380692336); SAV1 (NM_021818.2; GL 18860913); PELI1

(NM_020651.3; GL 193211454); UTRN (NM_007124.2; GI: 110611227); AKAP12 (NM 005100.3; GI: 197927427); CTDSPL (NM OO 1008392.1; GL56549682); KCNT2 (NM_198503.2; GL41349442); MRAP2 (NMJ38409.2; GI: 156523250); VPS52

(NM_022553.4; GL73747798); VPS52 (NM_022553.4; GL73747798); QSER1

(NM OO 1076786.1; GI: 1 15647980); C6orfl68 (NM 032511.2; GI: 169404007); LYRM7 (NM 181705.2; GI:91176332); ARPP19 (NM 006628.4; GL56549669); HECTD1 (NM 015382.2; GI: 118498336); RPL31 (NM OO 1099693.1; GL 153252131); FAM160B1 (NM_020940.3; GL205277426); METTL3 (NM_019852.3; GL99077115); PPP1CC (NM_002710.3; GL332688253); BTBD3 (NM_014962.2; GI:31317210); MEGF9

(NM_001080497.2; GL304434525); EPM2AIP1 (NM_014805.3; GI: 197116362); ZNF493 (NM_001076678.2; GL222418651); HGSNAT (NM_152419.2; GL 150378451); L0NRF2 (NMJ98461.3; GI: 148528974); SPTY2D1 (NM_194285.2; GI:51702221); WNT5A (NM_003392.4; GL371506361); BBS7 (NMJ76824.2; GL324073523); PM20D2

(NM 001010853.1; GL58082084); APOLDl (NM OOl 130415.1; GI: 194353994); ALGIOB (NM 001013620.3; GI: 145580624); HSD17B4 (NM 000414.3; GL313151208); ATF7 (NM OO 1130059.1; GI: 194239639); DDXl (NM 004939.2; GL379317168); TMEM184C (NM 018241.2; GL 190358511); END0D1 (NM 015036.2; GL296080748); PRICKLE2 (NM_198859.3; GI: 197333826); UBR3 (NMJ72070.3; GI: 160948609); ZNF680

(NM_178558.4; GI: 194097406); MBTD1 (NM_017643.2; GI: 158508475); SLC39A9 (NM_018375.4; GL356460991); TSHZ3 (NM_020856.2; GL 127138956); ZNF189 (NM_003452.2; GL37693524); RAB36 (NM_004914.2; GL31795534); RPS14

(NM_001025071.1; GL68160921); RBM22 (NM_018047.2; GL224967063); Clorfl28 (NM_020362.4; GL209954793 ); C13orf23 (NM_025138.3; GL 125656154); ING3 (NM 019071.2; GL38201654); ANKRD46 (NM 198401.2; GL50053931); FAM160A1 (NM OO 1109977.1; GL 158341645); FAM164A (NM_016010.2; GL223555988); CCDC3 (NM_031455.3; GI: 157676326); XP04 (NM_022459.4; GI: 148886660); PTPRB

(NM_001109754.2; GL332800996); EIF1AX (NM_001412.3; GL77404356); SLC26A2 (NM 000112.3; GI: 100913029); CNBP (NM 001127192.1; GI: 187608725); COX19 (NM_001031617.2; GI: 1 10349770); TNS1 (NM_022648.4; GL 156142195); VTA1 (NM_016485.3; GL21361740); CTR9 (NM_014633.3; GL 141803441); LIMA1

(NM 001113546.1; GL 165905588); PBRM1 (NM 018165); SFRS3 (NM 003017.4;

GI: 194239695); (NR2C1 (NM_003297.2; GL 189491738); PLXNC1 (NM_005761.2;

GI:317008579); SLC1A1 (NM_004170.5; GL292658766); ZNF449 (NM_152695.5;

GI: 194239637); TNRC6B (NM_001024843.1; GL67782329); SEC63 (NM_007214.4; GI: 189491764); C0PS4 (NM_016129.2; GL38373689); ZNF471 (NM_020813.2;

GI: 150170666); ITGB1BP1 (NM_004763.3; GI: 115527101); SYNP02 (NM_133477.2; GL 193083182); ZNF295 (NM OO 1098402.1; GL 148491087); PAFAHIBI (NM 000430.3; GI: 164419737); ZNF10 (NM_015394.4; GL38045950); DNAJB14 (NM_001031723.2; GI: 197927249); M0N2 (NM_015026.2; GI: 114326551); ZYG11B (NM_024646.2;

GL210147509); CEP97 (NM_024548.2; GI:31377704); BNIP2 (NM_004330.2;

GI: 168480079); TRAPPC2 (NM 001011658.3; GI:291045292); ZBTB2 (NM 020861.1; GL24308240); RNF6 (NM 005977.3; GI: 156071500); ALDH9A1 (NM 000696.3;

GI: 115387103); PHF14 (NM_014660.3; GL292781135); OGT (NM_181672.2;

GL262231792); FGD4 (NM_139241.2; GI: 198041927); ZNF626 (NM_001076675.2;

GI: 194248054); SMG1 (NM_015092.4; GL219277632); BDH2 (NM_020139.3;

GL66933013); ZNF737 (NM_001159293.1; GL226530354); HBP1 (NM_012257.3;

GL47834345); C6orfl62 (NM OO 1042493.1; GI: 109948309); ARPP19 (NM 006628.4; GL56549669); SMG1 (NM 015092.4; GL219277632); ATP6V1C1 (NM 001695.4;

GL 148536833); MAPK1 (NM_002745.4; GL75709178); MTMR12 (NM OO 1040446.1; GL94721262); PDZD2 (NM_178140.2; GL87196342); VAPA (NM_003574.5;

GL94721249); EIF1AX (NM_001412.3; GL77404356); SMG1 (NM_015092.4;

GL219277632); ETNK1 (NM_018638.4; GL87298841); N0P14 (NM_003703.1;

GL55769586); PELI2 (NM_021255.2; GL87080798); STAG3L2 (NM_001025202);

HMGB3 (NM_005342.2; GI:71143136); MAR6 (NM_005885.2; GL33589845). These targets (from Targetscan) are inversely correlated with the miRNA expression in the primary vs. metastatic tumors. The targets are also down-regulated on average as these miRNAs are up-regulated.

Example 1. miRNAs, such as miR-21, drive an aggressive phenotype.

Samples are collected to allow for sampling more than one location to provide sufficient material for iterative analyses including immunohistochemistry (IHC) and in situ hybridization (ISH) to probe protein and miRNA levels, respectively, to validate and further test the genomic data. Multiple measurements are detected from the tumor samples to improve confidence of the validity of the observations. This approach is pursued by national projects such as TCGA (McLendon, et al. Nature. 2008;455(7216): 1061-8). One omental metastases from each case was collected and a second metastases was collected in most cases (Table 3).

Table 3. Patient characteristics of collected specimens.

As described below, up-regulated mRNAs identify more aggressive primary tumors, and the differentially expressed miRNAs are key regulators of this more aggressive state. As described in the results herein, metastatic tumors are clonal expansions from primary tumors, but adapt to the new microenvironment and are more proliferative. These data are most consistent with a metastasis evolution model as opposed to a "primary tumor predisposition" model as significant expression differences are observed from genetically matched tumor samples.

Primary vs. metastatic tumors

To evaluate similarities between primary tumors and metastases, DNA copy number was evaluated from 12 matched primary and omental metastases using Agilent 18 OK CGH microarrays (Figure 1). These data show that 8/12 primary-met pairs are more similar to each other genetically than the other tumors, consistent with a clonal expansion model, consistent with past observations (Khalique L, et al. Int J Cancer. 2009;124(7): 1579-86). Because the DNA data suggest that tumor cells are typically genetically similar, expression differences were investigated.

Metastases are proliferating more and are more anti-apoptotic compared to primary tumors

A number of models have been proposed to explain the relationship between primary and metastatic tumors. These models mainly focus on clonal expansion vs. the need for the accumulation of additional genetic changes to establish metastases (Stoecklein NH and Klein CA Int J Cancer. 2010;126(3):589-98). Next-Generation Sequencing (NGS) suggests that new mutations evolve during metastasis, at least in pancreatic cancer (Yachida S, et al.

Nature. 2010;467(7319): 1114-7). Most data examining dissemination of tumor cells focus on DNA to ask questions about clonal expansion. Early gene expression studies in ovarian cancer suggest that mRNA expression profiles are very similar between primary and metastatic tumors (Lancaster JM, et al. Int J Gynecol Cancer. 2006;16(5): 1733-45; Lengyel E. Am J Pathol. 2010;177(3): 1053-64. PMCID: 2928939). The results presented herein differ from these early studies, because more modern microarray technologies were used, and because the experimental methods focused on pathway analysis instead of gene-centric approach.

RNA was purified from regions of tumor with >60% tumor cells and probed on Affymetrix Human Gene St v 1.0 microarrays. 478 up-regulated and 496 down-regulated genes in the metastases were selected by a paired t-test (p<0.05) and >50% fold changes. Multiple hypothesis correction was not applied because it seemed to overcorrect based on qPCR validation of the microarray data, where 10/11 genes were consistent with the microarray data.

To determine what pathways may be differentially regulated between the primary and metastatic tumors, Gene Set Enrichment Analysis (GSEA) was used (Subramanian A, et al. Proc Natl Acad Sci U S A. 2005;102(43): 15545-50). In GSEA, Normalized Enrichment Scores (NES) >1.5, p-values < 0.05, and False Discovery Rate (FDR) <0.25 are generally considered significant. G2/M cell cycle checkpoints including BUB1 and CHEK2 are expressed higher in primary tumors (Figure 2A). Higher expression of pro-survival genes including MYC, IRS1, and TNFSF10 define increased expression of apoptosis gene ontology (NES = 1.56, p<0.001) pathways in metastases. Consistent with these ideas, higher Ki-67 staining in metastases was observed (Figure 2B) and less apoptosis by terminal

deoxynucleotidyl transferase dUTP nick end labeling (TUNEL). Genes positively correlating (Pearson >0.75) with the Ki-67 proliferative index are enriched for cell cycle progress and DNA replication further indicating that a significant fraction of the expression signal is originating from tumor cells.

As a second test that the bulk tumor expression data derives from tumor cells, the cells were stained for FAPB4. FABP4 mRNA expression is higher in the metastases, as reported previously (Lancaster JM, et al. Int J Gynecol Cancer. 2006;16(5): 1733-45).

Immunohistochemistry (IHC) of FABP4 shows significant staining in tumor cells, with more staining in metastases (p< 0.03, paired t-test, n=9), consistent with the mR A microarray data. FABP4 is a fatty acid transporter with increased expression in the omental metastases suggesting that the tumor cells may be increasing their use of readily available fat as a carbon and energy source to support enhanced proliferation.

These mRNA expression data suggest that the patient cohort and measurements are consistent with past studies and thus provide confidence that the novel miRNA observations described herein are not specific to this cohort of patients. The mRNA expression was measured in three secondary site metastases. GSEA analysis of these metastases compared to the matched omental metastases indicates that these secondary metastases are resistant to platinum therapy, exemplified by increased expression of glutathione metabolism (NES=1.7, p=0.002), including higher expression of GSTA1, in the secondary site metastases (Siddik ZH. Oncogene. 2003;22(47):7265-79).

These data also indicate that at least some of the features important for metastasis reside within the primary tumor (See also, Ramaswamy S, et al. Nat Genet. 2003;33(1):49- 54). However, not all genes differentially regulated are associated with survival, suggesting that features important for metastasis may be missing from the primary tumor in somecases. Alternatively, this could be because only upon selection of specific cancer cells and subsequent adaptation to the new microenvironment, does increased expression become evident.

Metastatic genes identify more aggressive disease

To test if the differentially regulated genes may indicate particularly aggressive tumors, Oncomine was utilized to compare up-regulated genes to large expression data sets from ovarian tumors (Rhodes DR, et al. Neoplasia. 2007;9(2): 166-80). Genes up-regulated in metastases are enriched in the top 10% expressed genes in ovarian serous adenocarcinoma primary tumors (Q-value=le-15, Odds Ratio 4.1) from patients with less than one year survival (Tothill RW, et al. Clin Cancer Res. 2008;14(16):5198-208). These same genes have relatively low expression in tumors from patients that survive 5 years (Q-value=6e-10, Odds Ratio=5.5 (Tothill RW, et al. Clin Cancer Res. 2008;14(16):5198-208)). These observations are consistent with the concept that expression in primary tumors is indicative of future metastasis (Ramaswamy S, et al. Nat Genet. 2003;33(l):49-54). These data may be consistent with the "primary tumor predisposition" model or alternatively may indicate that the primary tumors still need to increase expression of key factors, already highly expressed, to form metastases. Expression of the following genes (e.g., as measured by mRNA levels) are useful for survival analysis: BGN, COL11A1, COL1A1, COL8A1, COMP, CRISPLD2, FNDC1, GADD45B, IGHM, INHBA, INTS2, INTS4, ITGA11, LTBP2, MMP14, MYH9, NARS2, NNMT, OLFML2B, PLAU, RASSF2, THY1, and TRIPlO. In Figs.21-25, the results of Kaplan-Meier analysis for overall survival and progression free survival is shown. These data indicate that the RNAs robustly classify the patients.

Patients with no macro disease have better outcomes as all the readily observable tumor is removed. However, tumors can return and the mRNA + miRNA combined expression signature clearly separates the high and low risk groups. The survival curves indicate near 100% specificity for 5 year survival. Thus, the biomarkers improve the prognostic prediction in conjunction with residual disease.

Specific miRNAs are differentially expressed between primary and metastatic tumors

377 miRNAs were measured using ABI Taqman qPCR arrays, specific for mature miRNAs (Chen C, et al. Nucleic Acids Res. 2005;33(20):el79), in 9 patients with matched primary and metastatic tumors. Approximately 180 miRNAs are expressed no global up- or down-regulation of these miRNAs between the primary and metastatic tumors was observed. Figure 3A summarizes the top ten miRNAs as identified by a paired t-test and filtered for significant expression levels. Each identified miRNA has significant expression differences on a loglO scale. miR-21 and miR-31 are well-known miRNAs highly expressed in many tumors and often linked to aggressive disease (Krichevsky AM and Gabriely G J Cell Mol Med. 2009;13(l):39-53). In some primary tumors, miR-21 is expressed at low levels, barely detectable by in situ hybridization (ISH), consistent with previous reports of relatively low expression in ovarian tumor cells (Dahiya N, et al. PLoS One. 2008;3(6):e2436. PMCID: 2410296). These results clearly suggest that miR-21 increases during disease progression. ISH was very challenging as only 1/4 of tested miRNAs provided reliable data (Figure 3C). Because a reliable signal for miR-31 was not detected, tumor cells were isolated by Laser Capture Microdissection (LCM). Subsequently, Taqman was performed, which resulted in a finding a general agreement with the original screen (Figure 3B). Increased expression of miR-21 is consistent with the increased aggressiveness (Jazbutyte V and Thum T. Curr Drug Targets. 2010;11(8):926-35; Bonci D Recent Pat Cardiovasc Drug Discov. 2010;5(3): 156- 61).

Functional analysis of miRNAs in vitro

miRNA expression was measured using ABI Taqman qPCR cards in a panel of ovarian cancer cell lines (OVCAR8, SKOV3, IGROV1) and miR-21 and miR-31 in ES-2, HEY, and SKOV3-IP. miR-21 was highly expressed in all cell lines and miR-124 was at least moderately expressed. miR-31 is expressed in all cell lines, except HEY, presenting an opportunity to explore miR-31 function using both ectopic expression and inhibition.

Analogous to Weinberg's recent breast cancer studies (Valastyan S, et al. Cell.

2009;137(6): 1032-46; Valastyan S, et al. Genes Dev. 2009), these data are used as a resource to develop the cell lines for the miRNA experiments described herein.

There are many options to use antisense nucleotides with modified backbones to improve stability and increase binding to miRNAs. Peptide nucleic acids (PNA) are particularly potent inhibitors (Fabani MM and Gait MJ. RNA. 2008;14(2):336-46. PMCID: 2212241). PNAs linked to a tat peptide have been utilized to trans feet independent of transfection agents (Fabani MM and Gait MJ. RNA. 2008;14(2):336-46. PMCID: 2212241; Oh SY, et al. Mol Cells. 2009;28(4):341-5). Decreased levels of miR-21 and miR-31 were observed, as the inhibitor sequesters miRNAs (Figure 4E). miR-21 inhibition in ES-2 cells reproducibly slows spheroid growth and reduced colony formation in soft agar assays (Figure 4). Surprisingly, no significant effects were observed in adherent 2D cell culture proliferation with ES-2 or OVCAR-8 when inhibiting either miR-21 or miR-31. miR-31 reduced OVCAR-8 spheroid formation, and both miR-21 and miR-31 reduce the number of colonies in a soft agar assay. Together, these data indicate that miR-21 and miR-31 are critical for anchorage independent growth in at least some ovarian cancer cells. Anchorage independent growth is linked to more aggressive cancer cells. The data presented herein indicate that miR-21 and miR-31 promote more aggressive, anchorage independent growth, consistent with increased expression observed in metastatic human tumors.

As described in detail below, the clinical, in vitro and in vivo experimental models are used to identify miRNAs that show significant expression differences in metastases. The possibility that miRNAs are important for metastasis is also examined. It was hypothesized that miRNAs are critical regulators to support tumor cell survival and dissemination and developing a more aggressive, and possibly more drug resistant, state in metastases. As described below, by determining the mRNA targets and phenotypes using in vitro and in vivo models, insight is gained into the mechanisms important for peritoneal dissemination of ovarian cancer and the role of miRNAs in regulating these processes.

mRNA and miRNA metastasis signatures identify aggressive disease

To test if the differentially regulated genes are important for disease progression, studies were carried out to determine if expression levels in primary tumors of these up- or down-regulated genes are associated with survival. Expression in primary tumor has been associated with metastatic potential. However, finding the key genes remains a challenge due to statistical biases and assay concerns (Shi L, et al. Nat Biotechnol. 2010;28(8):827-38; Majewski IJ and Bernards R. Nat Med. 2011;17(3):304-12). To classify patients, a Support Vector Machine (SVM) classification model was utilized for Kaplan-Meier analysis using The Cancer Genome Atlas (TCGA) data. The differentially regulated metastasis genes were mapped using the gene symbol, not the probeset, to Affymetrix U133 Plus 2.0 arrays, providing further confidence that these associations are not platform dependent. Using the TCGA data including 436 primary tumors, remarkably discriminatory Kaplan-Meier survival curves (Figure 5B) of the metastasis signature were observed. Discriminatory Kaplan-Meier curves for the miRNAs were also observed (Figure 5C). These data might be consistent with the "primary tumor predisposition" model. Alternatively, the data may indicate that the primary tumors still need to increase expression of key factors, already highly expressed, to form metastases.

These Kaplan-Meier curves indicate that mRNAs and miRNAs differentially expressed between primary and metastatic tumors are important features of disease progression and patient outcomes. As described in detail below, it is investigated by Kaplan- Meier analysis whether miRNA regulated genes in combination with their miRNAs have predictive power. Together, these observations not only provide clinical justification of the importance of these factors, but also indicate the utility of a mechanism based biomarker expression signature.

miR-31 as ovarian cancer promoter vs. breast cancer suppressor of metastasis

miR-31 inhibits metastasis in breast cancer mouse models, which is consistent with decreased expression in human metastases (Valastyan S, et al. Cell. 2009;137(6): 1032-46). However, the data presented herein indicate that miR-31 promotes ovarian cancer progression. miR-31 is highly expressed in many aggressive ovarian cancer cell lines including ES-2, SKOV3 and OVCAR-8, with higher expression in some spheroids (Figure 4D), which is consistent with its possible function to drive metastasis. Because the opposite effect is observed in ovarian metastases (Figure 2) vs. breast cancer, the role of miR-31 was examined in ovarian cancer. miRNAs, like proteins, can function as oncogenes or tumor suppressors in different contexts (Garzon R, et al. Nat Rev Drug Discov. 2010;9(10):775-89; Garzon R, et al. Annu Rev Med. 2009;60: 167-79; Iorio MV and Croce CM. M J Clin Oncol. 2009;27(34):5848-56. PMCID: 2793003). Described herein is how ovarian cancer cells depend on miR-31 for dissemination in vitro and in vivo.

Example 2. Identification of differentially expressed miRNAs in ovarian primary tumors and metastases. miRNA and mRNA expression profiles are measured from four quadrants of primary and metastatic tumors (Figure 6). Specimens are collected with careful recording of the tumor block's original location within the whole tumor. Approximately 100 ovarian cancer patients are recruited to achieve the goal of at least 50 patients with serous adenocarcinoma. Given the the data described above suggesting that many miRNAs show significant expression changes in the majority of the 9 patients measured, it is expected that two groups of patients are defined that likely are indicative of the major metastatic features.

To extend and refine the characterization of primary and metastatic tumors, four quadrants are sampled from both the primary and metastatic tumors to identify the origin of potential metastatic cancer cells (Figure 6). Tumor cells closer to the omentum are the progenitors of metastases. This may be true genetically, as recently suggested by Vogelstein and colleagues (Yachida S, et al. Nature. 2010;467(7319): 1114-7). However, as metastases may be interacting with different environments on each side, i.e., omentum and the peritoneal cavity, there may be different expression patterns indicative of how the clonally expanding tumor cells are adapting to these environments. By sampling each tumor in multiple locations, it is examined how heterogeneous primary and metastatic tumors are able to increase the probability of defining the miRNAs regulating metastasis. There is evidence that expression analysis of some primary tumors predicts aggressive disease (Ramaswamy S, Ross KN, Lander ES, Golub TR. A molecular signature of metastasis in primary solid tumors. Nat Genet. 2003;33(l):49-54; Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature. 2008;455(7216): 1061-8), but these data are often noisy and some aggressive tumors could be missed because of tumor heterogeneity. Here, expression changes are identified, and not just the level of expression that may be correlated with aggressive disease. This refinement increases the probability of defining important miRNAs analogous to improved linking of factors when the perturbation is profiled compared to correlating the innate expression levels before the perturbation (Lamb J, et al. Science. 2006;313(5795): 1929-35).

Intra-tumor heterogeneity to identify differentially expressed miRNAs

LCM is used to ensure that isolated material is from tumor cells without

complications from stroma and infiltrating immune cells, which are particularly prevalent in some metastases. Tumor cells are isolated by LCM using ethanol fixation which avoids complications from formaldehyde cross-linking and yields higher quality RNA by Agilent Bioanalyzer analysis than from formaldehyde fixed cells, consistent with published reports (Espina V, et al. Nat Protoc. 2006;l(2):586-603; Wang S, et al. BMC Genomics. 2010;11 : 163. PMCID: 2853520). A tumor sample is collected from each quadrant (Figure 6). If sampling of the first 10 patients suggests that each quadrant is very similar, then testing just one location is sufficient, and the remaining tumors are sampled only once.

miR A expression

miRNAs are measured by Taqman qPCR after linear pre-amplification following the manufacturer's instructions. To determine the similarity of the miRNA expression profiles among the quadrants, the unsupervised hierarchical clustering is used. Specific differences are identified using paired t-tests. Batch effects are minimized as each matched primary and metastatic sample are prepared together.

mRNA expression is measured from the same isolated RNA used to measure miRNAs to help gain insight into possible miRNA targets. Affymetrix Gene St microarrays are utilized to measure the mRNAs. Also, to gain an improved measurement with better dynamic range, RNA-seq is utilized to measure the gene expression differences between the primary and metastases (Wang Z, et al. Nat Rev Genet. 2009;10(l):57-63). Moreover, because the data is digital and counts each transcript, the analysis to compare conditions is simplified and more robust than microarray analog signals (Mortazavi A, et al . Nat Methods.

2008;5(7):621-8). Because the RNA from LCM is moderately degraded, techniques, such as Ribo-Zero (Epicentre Biotechnologies) are utilized to reduce rRNA and to prepare

sequencing libraries. RNA-seq is run on an Illumina HiSeq 2000, and one lane should provide more than sufficient coverage for gene expression analysis (Marioni JC, et al.

Genome Res. 2008;18(9): 1509-17).

To analyze the mRNA data, similar approaches to the miRNA data analysis are used. A combination of paired t-tests and fold changes identifies the most significantly changing transcripts and hierarchical clustering defines which blocks and quadrants are the most similar. Significantly changing genes are evaluated using Ingenuity Pathway Analysis (IP A) and GSEA. The quadrant sampling approach provides new insights into the relationship between the primary and metastatic tumors. The quadrants in primary tumors nearest omental metastases, quadrant E in Figure 6, are most similar to the metastases. Three major classes of interest are expected: 1) miRNAs with similar intra-tumor expression in all quadrants, but different expression between tumors, suggesting homogeneous expression in each tumor, consistent with clonal expansion. 2) Differential expression seen only in quadrants A and E, suggesting that clonal expansion from one region of the primary tumor establishes into a heterogeneous metastatic tumor. Similar expression within the metastases, but differential expression to primary tumor quadrant E, is consistent with this model. 3) Differential expression between all quadrants in each tumor suggests extensive heterogeneity. These miRNAs are indicative of tumor cells adapting to the unique surrounding environments. All classes are observed and other combinations are possible, but the miRNAs following class 1 and 2 are of the most interest as possible metastasis regulators. mRNA expression following similar classes that enhance correlation to miRNA and mRNA target predictions is examined. With mRNA data, pathway analysis highlights specific features of the tumors and quadrants. For example, fatty acid metabolism is enriched in quadrants nearest the omentum, possibly to take advantage of this resource for energy and carbon as proposed (Lengyel E. Am J Pathol. 2010;177(3): 1053-64. PMCID: 2928939;

Kenny HA, et al. Int J Cancer. 2007;121(7): 1463-72). The data suggest that cell cycle checkpoints are down-regulated in metastases throughout the tumor. Quadrant sampling confirms these hypotheses.

In the event that tumors are homogeneous and no significant differences within the region or across quadrants are observed in the first 10 patients, each tumor is sampled only once. Alternatively, the tumors are very heterogeneous such that each quadrant appears very different from the other quadrants {i.e., >10 fold miRNA expression differences), suggesting quadrant sampling is not sufficient to capture the range of tumor cells. This is examined by making more than one block and measuring more than one slice from each block. This deeper sampling approach is performed on three patients. Early results of miRNA expression from one quadrant suggest that there are modest differences in some miRNAs, miR-31, but not others, miR-21. These differences are not as large as the overall expression changes. Measuring more patients clarifies the degree of heterogeneity.

Tissue samples are processed using known methods so as to minimize RNA degradation.. This is not as problematic for miRNAs which are more stable than mRNA (Hoefig KP, et al. Anticancer Res. 2008;28(1A): 119-23; Porkka KP, et al. Cancer Res.

2007;67(13):6130-5; Kye MJ, et al. RNA. 2007;13(8): 1224-34. PMCID: 1924896).

However, if follow-up validation IHC, ISH or LNA qPCR data suggests the data may be unreliable, RNA is isolated from each block, after selection for >70% tumor cells from each quadrant. Many ovarian tumors are sufficiently large that this approach is feasible. This approach yields miRNA candidates that are validated by additional experiments, including ISH and examination of LCM isolated tumor cells where measuring one specific miRNA or mRNA is more feasible and controllable than genome-wide studies. Additional

measurements in the tumor along with functional testing provide confidence of the validity of the observed material even if the RNA is moderately degraded. Tumor progression from primary tumors to omental and secondary site metastases

Secondary sites from the peritoneum (peritoneal implants) and lymph nodes (LN) are collected whenever possible. In the patient population (stage IIIC serous EOC),

approximately 80% of patients have separate abdominopelvic peritoneal metastases (in addition to omental disease) and 50% have LN metastases. These secondary site metastases, from myriad locations, are more proliferative and anti-apoptotic than matched omental metastases based on mRNA expression data from three patients. mRNA and miRNA expression from these non-omental metastases is measured using the same quadrant sampling approach.

To gain insight into the role of expression changes in primary vs. metastatic tumors, a metastatic signature is defined consisting of the top changing miRNAs and mRNAs (for example, the top 100 mRNAs and the top 10 miRNAs by p-value and fold change criteria) and compare these data to large datasets such as those from Tothill, et al. (Tothill RW, et al. Clin Cancer Res. 2008;14(16):5198-208) and TCGA where key clinical metrics such as progression free survival (PFS) are available for hundreds of tumors. Kaplan-Meir survival curve analysis was used to determine if metastatic gene expression identifies more aggressive disease. This analysis examines how common the "tumor pre-disposition" model is.

The important factors for the extensive dissemination of ovarian tumor cells within the peritoneal cavity are identified herein. The order of progression is not clear, as tumors may originate from the primary or omental metastases or other metastases. The similarity of these non-omental metastases are examined using the strategies described above. There are common features important for metastasis of ovarian cancer cells that are suitable for general drug targets. The alternative hypothesis is that some miRNAs may be uniquely differentially expressed in omental or non-omental tumors. The three types of tumors are compared to each other.

miR-31 is significantly up regulated in 3/4 non-omental metastases compared to primary tumors, suggesting that miR-31 is important in the establishment of metastases through the peritoneal cavity. Measuring additional samples from the current collection (including 10 non-omental serous epithelial metastases), along with future collected samples addresses how common miR-31 or other miRNAs are. miR-21 is not up-regulated compared to the primary tumor in most of the secondary sites, which suggests that miR-21 is more specific for the omentum. Two major classes of miRNAs are expected: 1) Up/down-regulated in omental metastases, but not non-omental metastases compared to primary tumors, such as miR-21, which suggests their importance in interacting with the unique omental environment.

2) Up/down-regulated in both omental and non-omental metastases compared to primary, such as miR-31 , which suggests their importance for establishing metastases in multiple environments. Differential expression in metastases and differential expression within the primary tumor suggests extensive heterogeneity and possible importance for adoption to the local environment.

Similar classes are defined for mRNAs. mRNA expression data complements the miRNA data to support analysis of putative miRNA targets. Analogous to miRNAs, specific pathways and genes are differentially expressed. Certain features are expected to be similar in all metastases, such as increased expression of integrin signaling and extracellular matrix (ECM) interactions relative to primary tumors. mRNA expression data from three non- omental metastases suggest that non-omental metastases up-regulate cell cycle pathways compared to omental metastases (NES=2.1, p<le-3 ), including CCNA1, and increase expression of ECM receptor interactions compared to primary tumors (NES=1.6, p=0.03). Ki-67 staining suggests similar high levels of proliferation in non-omental metastases, and higher than primary tumors. Measuring more non-omental metastases is necessary to make firmer conclusions and these early observations are reproduced in a larger patient cohort.

Smaller tumors, such as LN metastases, are not sufficiently large for quadrant sampling. In this case, RNA is isolated from one region to represent the profile for the tumor to gain insights into the disease progression. It is possible that no dominant trend is observed in the majority of patients as suggested by the data described above. However, the data also suggest that it is likely that a common pattern is observed for a minority fraction (>10%). Example 3. The function of miRNAs in vitro and in vivo

miRNAs with differential expression are metastatic regulators. Their function is examined using in vitro and in vivo models. miR-21 is differentially expressed from multiple blocks in primary and metastatic tumors. miR-21 has previously been associated with tumor progression (Jazbutyte V and Thum T Curr Drug Targets. 2010;11(8):926-35) and there is limited data on its function in ovarian cancer (Gao W, et al. J Cancer Res Clin Oncol. 2010). These experiments are performed for miR-21, and similar experiments are pursued for at least one miRNAs from each class identified in the clinical data described above.

The function of miRNAs mediating metastatic properties in vitro miRNAs are tested for transformation, proliferation, and drug sensitivity in vitro in a representative panel of ovarian cancer cell lines. These include the wound healing assay, growth curves with and without chemotherapeutics such as cisplatin, anoikis assays, soft agar colony formation assays, trans-well assays, collagen invasion assays, and spheroid formation. Together, these assays probe how these cell properties, necessary to survive when not attached to a solid surface (anoikis, soft agar, spheroid), move around the omentum (wound healing, trans-well), invade the collagen rich omentum (invasion), respond to

chemotherapeutics, and establish a new colony of tumor cells.

Because PNAs form a stable complex with the miRNA, knock-down of effective miRNA levels is monitored to determine the effectiveness of the PNA (Fabani MM and Gait MJ RNA. 2008;14(2):336-46. PMCID: 2212241; Oh SY, Mol Cells. 2009;28(4):341-5). To ensure that the PNA is effectively reducing miRNA function, artificial constructs are generated with six miRNA sites that should respond and inform the efficacy of miRNA manipulations. More native 3' UTR reporters are also utilized.

miRNAs are examined in a panel of 14 cell lines to model the range of genetic backgrounds in ovarian tumors. This is a proven approach in other cancers where any single cell line may not recapitulate tumors, but the average response and genetics across a panel of cell lines effectively models tumors (Neve RM, et al. Cancer Cell. 2006;10(6):515-27). The most aggressive cell lines are examined, such as ES-2, HEY, SKOV3-IP, and A2780, as well as those that were derived from serous epithelial cancers, such as DOV13, OVCA433 and OVCA429. Most of these cell lines form soft agar colonies in a reasonable time frame and are amenable to xenografts (Lange TS, et al. Invest New Drugs. 2009;28(5):543-53; Robison K, Kim K, Singh R, Lange T, Granai C, Brard L, editors. The use of a vitamin D derivative in a mouse xenograft ovarian cancer model. GOG; 2008).

These in vitro experiments provide insights into which steps a specific miRNA are affected during metastasis. In one aspect, miRNAs are pleiotropic, affecting more than one step of the metastasis cascade (Nguyen DX, et al. Nat Rev Cancer. 2009;9(4):274-84;

Valastyan S, et al. Cell. 2009;137(6): 1032-46; Valastyan S, et al. Genes Dev. 2009;

Valastyan S, et al. Cancer Res. 2010;70(12):5147-54. PMCID: 2891350). miR-21 promotes in vitro metastatic properties extending the preliminary observations in multiple cell lines described above (Figure 4). In particular, miR-21 's effects on soft-agar colony are the best predictor of the xenograft experiments. miR-21 affecting multiple cell lines would further suggest miR-21 's general role in metastasis. Another miRNA, miR-31, appears to behave differently in ovarian tumors than breast tumors. Based on the clinical data, miR-31 promotes tumor survival and dissemination. These experiments test the how ovarian cancer cells depend on miR-31 for metastatic properties in vitro. miR-31 is highly expressed in SKOV3 and OVCAR8, consistent with its possible importance for these cells metastatic behaviors. The p53 wild-type, IGROV-1, has relatively low expression of miR-31 , and introducing miR-31 with a pre-miR may induce stronger spheroid formation or soft-agar colony formation of this relatively less aggressive cell line.

The in vitro experiments mostly focus on the tumors cells in relative isolation of fat, stromal, immune and other cells, typically found in the omentum and other metastatic sites that may contribute to tumor progression within a real tumor. Thus, part of miRNA control of tumor cells are in response to the unique micro environment at the metastatic site.

Determining the function of miRNAs in vivo using xenograft models

A panel of cell lines (SKOV3-IP, ES-2, HEY) stably expressing luciferase is created by G418 selection for live in vivo imaging of the tumors and dissemination of the tumor cells (Edinger M, et al. Neoplasia. 1999;1(4):303-10. PMCID: 1508101). Selected cell lines are tested by soft agar assays as described iabove, in addition to pilot xenograft experiments to ensure that the luciferase expressing cell lines behave similarly to the parental cells.

To test the role of miR-21 , miRNA function is inhibited by expressing an miRNA sponge to compete with endogenous miR-21 target mRNAs reducing miR-21 activity (Ebert MS, Nat Methods. 2007;4(9):721-6). Because miR-21 does not significantly affect proliferation, it is possible to select a stable line inhibiting miR-21. Stably expressing sponge lines are tested to recapitulate the in vitro observations using PNAs and transient sponge inhibition of miR-21. The tumor cells expressing miR-21 or scrambled sequence sponge constructs are injected into the bursa sac (IB) of nude mice to generate a localized primary tumor, which spreads within the abdominal cavity, including the peritoneum. Intraperitoneal (IP) injection of these cell lines is also be performed as a separate dissemination model (vide infra). Using live animal luciferase imaging (IVIS) metastases are examined, both for size and number, primary tumor size and mouse survival similar to published approaches (Shaw TJ, et al. Mol Ther.

2004;10(6): 1032-42). Ten animals in 3 groups are tested for the parental cell line, mutant sponge, and the stably expressed sponge to gain statistical significance. The chosen sample size is commonly used for these types of studies and allows generation of statistically meaningful data (Lange TS, et al. Chem Biol Drug Des. 2010;76(2): 164-73; Saulnier Sholler GL, et al. J Pediatr Hematol Oncol. 2009;31(3): 187-93). The function of miRNAs is examined in xenograft models. To test whether patient- identified miRNAs are critical for metastasis and tumor progression, a panel of cell lines are used in xenograft tumor models. In the past, SKOV-3 xenograft animal models were used. In addition to this cell line, more aggressive metastatic lines, such as SKOV3-IP, ES-2 and HEY are also used. These cell lines readily form soft agar colonies in 2-3 weeks, and are reported to rapidly disseminate within the abdominal cavity in xenograft models (Shaw TJ, et al. Mol Ther. 2004; 10(6): 1032-42). miRNAs are tested in a panel of cell lines to mimic the range of phenotypes and genetic backgrounds of tumors (Sharma SV, Nat Rev Cancer.

2010;10(4):241-53).

Inhibition of miR-21 reduces dissemination of ovarian cancer cells. In one aspect, a reduction in the size of the primary tumor in the IB groups is seen, which indicates effects on growth in addition to effects on dissemination. This suggests that miR-21 affects both proliferation and metastasis. In one aspect, because miR-31 expression increases in omental and non-omental metastases, miR-31 affect dissemination throughout the peritoneal cavity. It is also possible that miR-21 affects the amount of tumor cells in ascites, when tumor cells are not attached, consistent with its possible role in suppressing apoptosis found in other tumors (Asangani IA, et al. Oncogene. 2008;27(15):2128-36; Frankel LB, et al. J Biol Chem.

2008;283(2): 1026-33; Krichevsky AM and Gabriely G J Cell Mol Med. 2009;13(l):39-53).

It is possible that the miRNA affects primary tumor growth, reducing the metastatic potential. It is possible that this is the case with miR-21, which affects 3D growth. In this case, it is determined if the miRNA specifically affects metastasis by allowing the primary tumor to form and disseminate before adding a miRNA inhibitor IP, such as LNA or PNA. This model better represents the clinical scenario as it is examined if inhibiting a miRNA reverses, or least prevents, further metastasis. To avoid complications of inhibitor delivery, the sponge is modified with a drug-inducible promoter such as tamoxifen (Brown BD, Naldini L. Exploiting and antagonizing microRNA regulation for therapeutic and

experimental applications. Nat Rev Genet. 2009;10(8):578-85). IB has the advantage of more clearly defining the primary tumor compared to IP. In the event that problems are encountered with the degree of dissemination with the IB model, strong metastasis are induced with IP, using a similar experimental design of sponge or LNA's to inhibit miRNA activity.

Example 4. The role of miR-21 and miR-31

miR-21 and miR-31 control disease progression by promoting survival and proliferation in anchorage independent conditions. Because miRNAs may be pleiotropic, the ability of miR-21 and miR-31 to mediate different properties of cancer cells important in the metastatic cascade is examined (Nguyen DX, et al. Nat Rev Cancer. 2009;9(4):274-84;

Valastyan S, et al. Cell. 2009;137(6): 1032-46; Valastyan S, et al. Genes Dev. 2009;

Valastyan S, et al. Cancer Res. 2010;70(12):5147-54. PMCID: 2891350). Together, these assays probe how these cell properties, necessary to survive when not attached to a solid surface (anoikis, soft agar, spheroid), migrate in the omentum (wound healing, trans-well), invade the collagen rich omentum (invasion), respond to chemotherapeutics, and establish a new colony of tumor cells.

Table 4. Cell lines and their properties .

Determining the function of miR-21 and miR-31 in mediating migration and invasion

Migration and invasion allow ovarian tumor cells to penetrate the omentum, break through the existing collagen-rich tissue, and establish a lesion. Correlations between migration, invasion, and expression levels in patient tumors have previously been identified (i.e., integrins and matrix metalloproteinases (MMPs); Lengyel E. Am J Pathol.

2010;177(3): 1053-64. PMCID: 2928939; Sawada K, et al. Cancer Res. 2008;68(7):2329-39. PMCID: 2665934; Cowden Dahl KD, et al. Cancer Res. 2008;68(12):4606-13). miR-21 and miR-31 are reported to mediate these processes in a variety of cancers (Medina PP, et al. Nature. 2010;467(7311):86-90; Connolly EC, et al. Mol Cancer Res. 2010;8(5):691-700; Bonci D. Recent Pat Cardiovasc Drug Discov. 2010;5(3): 156-61 ; Krichevsky AM and Gabriely G. J Cell Mol Med. 2009;13(l):39-53; Jazbutyte V and Thum T. Curr Drug Targets. 2010;11(8):926-35). Predicted negatively correlated targets include factors likely important for cancer cell mobility (Figure 7). Therefore, miR-21 and miR-31 mediate cell migration and invasion in addition to their likely role promoting spheroid and colony formation

(Figure 4).

miR-21 and miR-31 are examined for migration and invasion in a representative panel of ovarian cancer cell lines including ES-2, SKOV-3 and OVCAR8. These include the wound healing assay, trans-well assays, and collagen invasion assays. To test migration, cell lines are screened with the wound healing assay. Subsequently, the more quantitative trans- well assay is utilized. Invasion tests the cells' ability to move through a membrane which is dependent on the cells' ability to move through and break down collagen, or other tested membranes. Invasion often depends on MMPs (Cowden Dahl KD, et al. Cancer Res.

2008;68(12):4606-13) and an miR-21 down-regulated target in the metastases, RECK, is an inhibitor of MMPs (Clark JC, et al. Cancer Metastasis Rev. 2007;26(3-4):675-83), suggesting a role for miR-21 in promoting invasion.

Manipulating miRNA levels

Overexpressed miRNAs can induce non-specific effects even at relatively modest overexpression levels (Baek D, et al. Nature. 2008;455(7209):64-71; Selbach M, Nature. 2008;455(7209):58-63) and therefore may not be physiological relevant. Moreover, because the miRNA-mRNA interaction is concentration dependent (Ebert MS and Sharp PA. RNA. 2010;16(11):2043-50. PMCID: 2957044), miRNA overexpression can drive interactions that follow the "seed" rules of miRNA-mRNA predicted interactions, but may not be

physiologically relevant. There are a number of commercial options to inhibit miRNA function. Because PNAs form a stable complex with the miRNA, knock-down of effective miRNA levels is monitored to determine the effectiveness of the PNA (Fabani MM and Gait MJ. RNA. 2008;14(2):336-46. PMCID: 2212241; Oh SY, et al. Mol Cells. 2009;28(4):341-5) and Figure 4E. Introducing the PNA with a tat peptide or a transfection agent leads to equivalent levels of miRNA knock-down. The tat peptide allows for the introduction of the inhibitor without transfection agents, which may have their own effects on cell behavior. Specificity is determined by comparing to a random sequence inhibitor. To ensure that the PNA is effectively reducing miRNA function, artificial constructs are generated with six miRNA sites downstream of luciferase that indicate the efficacy of the miRNA

manipulations.

Migration and invasion in multiple cell lines

miR-21 and miR-31 are examined in a panel of cell lines to model the range of genetic backgrounds in ovarian tumors (Table 4). This is a proven approach in other cancers where any one single cell line may not recapitulate tumor features, but the average behavior and genetics across a panel of cell lines effectively models tumors (Neve RM, et al. Cancer Cell. 2006; 10(6):515-27). These cell lines form soft agar colonies and are amenable to xenografts (Lange TS, et al. Invest New Drugs. 2009;28(5):543-53; Robison K, et al. editors. The use of a vitamin D derivative in a mouse xenograft ovarian cancer model. GOG; 2008). To gain confidence that the observation is not cell line specific, the desired phenotype is observed in at least two cell lines. For ovarian cancer cells to establish metastases, they need to invade the omentum and need to be mobile to find a niche in which to grow in the new tissue (Lengyel E. Am J Pathol. 2010;177(3): 1053-64. PMCID: 2928939). Factors that affect invasion and migration (e.g., integrins and metalloproteinases (MMPs)) have higher expression in established metastases compared to primary tumors (Sawada K, et al. Cancer Res. 2008;68(7):2329-39. PMCID: 2665934). For the wound assay, after inhibiting the miRNA, the cells' ability to migrate across a region removed of cells is monitored. More quantitative trans-well assays are used to further examine any promising observations from the wound assay.

Invasion assays through collagen membranes best model the omentum as opposed to matrigel (Sodek KL, et al. BMC Cancer. 2008;8:223). A second complimentary assay is to mix cells with collagen and monitor how the ensuing aggregates spread into the collagen as developed by Stack and colleagues (Barbolina MV, et al. Mol Cancer Res. 2010;8(5):653-64. PMCID: 2883461). Together, these assays investigate whether the miR-21 or miR-31 mediate cell properties thought to be required to establish new lesions.

Table 5. In vitro assays to probe miRNA role in ovarian cancer metastasis.

miR-21 and miR-31 in some cell lines are vital for migration and invasion, based on the enriched pathways of the predicted targets (Figure 7C). It is also likely that miR-21 and miR-31 promote migration and invasion in some cell lines, consistent with enhancing "aggressiveness" and metastasis.

These in vitro experiments examine tumors cells in the absence of fat, stromal, immune, and other cells, typically found in tumors that may contribute to tumor progression (i.e., microenvironment). Thus, part of miRNA control of tumor cells may be in response to the unique microenvironment at the metastatic site. Therefore, these miRNAs are also examined in vivo. The observations of miR-31 action might conflict with the report of miR-31 over-expression inhibiting proliferation of p53 mutant ovarian cancer cell lines (Creighton CJ, et al. Cancer Res. 2010;70(5): 1906-15. PMCID: 2831102). IfmiR-31 inhibits proliferation of these cell lines in 2D culture, it may not have the same effect in 3D models such as spheroid or soft agar. On the other hand, in breast cancer over-expressing miR-31 enhanced growth, but inhibited metastasis, which suggests that opposing effects between these processes are feasible for miRNAs (Valastyan S, et al. Cell. 2009; 137(6): 1032-46). It is not uncommon for miRNAs {e.g., miR-31) to have apparently conflicting functions in different contexts (Valastyan S and Weinberg RA. Cell Cycle. 2010;9(11)). The experiments presented herein determine if miR-31 mediates in vitro metastatic properties and correlates with the increased expression observed in human metastases.

If only modest effects of miR-21 and miR-31 are observed, it is likely that a stronger phenotype may be possible when miRNAs are inhibited together. miR-21 and miR-31 expression is strongly correlated in primary and metastatic tumors (Pearson =0.72), where similar lower expression was observed in the same two cases, while up-regulated expression was observed in the same seven cases. Moreover, there are 9 overlapping Targetscan predicted mRNA targets negatively correlated in the human tumor data, including bromo adjacent homology domain containing 1 (BAHD1), a chromatin regulator, and nuclear factor of activated T-cells 5 (NFAT5), whose repression has been previously linked to metastasis (Levy C, et al. Mol Cell. 2010;40(5):841-9. PMCID: 3004467), which may require both miRNAs to be sufficiently repressed to induce a phenotype. Together, these observations indicate that inhibiting both miR-21 and miR-31 in ovarian cancer cells leads to stronger phenotypes.

The role of miR-21 and miR-31 in 2D and 3D cell culture systems

There is increasing interest in 3D culture systems such as spheroids. Spheroids, or multicellular aggregates (MCAs), are amenable to screening multiple perturbations and may model multicellular aggregates in the ascites of ovarian cancer patients that likely establish metastases (Grun B, et al. Cell Prolif. 2009;42(2):219-28). Moreover, MCAs may better model both the histology of tumors and the steric hindrance of drug access (Grun B, et al. Cell Prolif. 2009;42(2):219-28). The data presented herein indicate that both miR-21 and miR-31 promote 3D, but not 2D culture proliferation (Figure 4). The ability to not only resist anoikis, but to thrive and grow in anchorage independent conditions is often an indicator of tumorigenesis and aggressiveness. Here, the observations in spheroid and soft agar assays are extended to other cell lines. miRNA expression profiles in 2D and 3D determine which miR As are differentially expressed, perhaps mimicking disease progression. These experiments test the hypothesis that miR-21 and miR-31 drive anchorage independent growth.

miRNA levels are manipulated to examine the effect of miR-21 and miR-31 on anchorage independent growth. Spheroids are formed using micromolds to reliably form and measure >9 spheroids in each experiment (Figure 4). Soft agar experiments are examined by introducing anti-mir before mixing with agar. Additionally, more anti-mir is introduced during the weeks-long experiment, as the anti-mir may be diluted or degraded over time. For anoikis assays, transfected cells are plated on low adherent plates for 24 hours to induce anoikis before being moved to adherent plates for 24 hours. The percentage of viable cells compared to a negative control scrambled sequence anti-mir is determined by the modified MTT assays, Wst-1 (Promega) and cell counting. To test if these miRNAs reduce apoptosis, a TUNEL assay is utilized to directly observe the number of apoptotic cells in spheroids. The anti-mir may inhibit the cell cycle and reduce the number of cycling cells, which is determined by Ki-67 immunostaining. In the case of miR-21, which may inhibit apoptosis through target genes such as PDCD4 (Figure 7), it is determined if inhibition of miR-21 and miR-31 increases sensitivity to the common ovarian therapeutic, cisplatin, in 2D and 3D culture. These data would examine whether natural progression of the disease not only leads to more aggressive cells (Ki-67 and TUNEL tumor staining, Figure 2), but also leads to more chemoresistant cells.

miRNA and mRNA expression in 2D and 3D culture

The data presented herein indicate that changes in miR-21 and miR-31 expression in some cell lines mimic the changes in metastasis. Previous expression studies suggest that adhesion and integrin remodeling occur when cells are moved into 3D culture models. To test the hypothesis that expression changes in 3D culture compared to 2D culture mimic some of the adhesion and integrin remodeling observed in ovarian metastasis, miRNA expression profiles are examined in 2D and 3D cultures. A global expression profile using the Taqman qPCR to measure 377 miRNAs would determine the specificity of the increased expression compared to a larger background (as opposed to picking one or two miRNAs). It is determined if miRNA expression changes resemble disease progression in the two clinical models (primary vs. metastasis expression changes and Kaplan-Meier analysis of the TCGA primary tumors) to test which miRNAs are indicative of patient prognosis.

These experiments extend the observations suggesting that miR-21 and miR-31 mediate 3D growth. These data determine the general applicability of early observations presented in Figure 4 and Table 4). To determine how miR-21 and miR-31 perform these functions, it is determined if the number of apoptotic cells increased and/or the number of cycling cells is reduced when these miRNAs are inhibited. These are two possible mechanisms that could explain why spheroids are smaller when the miRNAs are inhibited suggesting their importance for survival and progression. This would indicate that these miRNAs and at least some of their mRNA targets are viable therapeutic agents.

Spheroids are powerful, but relatively crude 3D culture models that do not mimic tumor microenvironments, as the cancer cells are examined in the absence of other components that constitute a tumor such as fibroblasts, immune cells, and fat bodies. Co- culture models with fibroblasts or fatty tissue are examined to determine if these better model the progression from primary to metastatic tumors. Similar experiments are performed to examine the dependency on miR-21 and miR-31 in proliferation and apoptosis using Ki-67 staining and TUNEL assays, as well as measuring miRNA expression changes compared to cancer cells in standard 2D and 3D culture. These data begin to link expression changes to particular microenvironments.

Each cell line may respond differently to miR-21 and miR-31. For example, miR-31 may promote anchorage independent growth in OVCAR-8, but not affect migration and invasion. Alternatively, in another cell line (e.g., ES-2 cell line), miR-31 may not affect 3D growth, but may mediate invasion and migration. In this case, some of the differences between OVCAR-8 and ES-2 or other cell lines are examined to determine why miR-31 has these different functions in different backgrounds, such as P53 status and expression of other miRNAs. Even if miR-21 and miR-31 effects are not consistent across multiple cell lines, these experiments characterize MCAs and generate insight into these in vitro models and how they compare to human tumors as well as xenografts.

Example 5. Identification of miRNA-mRNA targets and mechanisms of miRNA control of metastasis

A combination of computational and experimental approaches is utilized to identify mRNA targets to determine which factors and pathways are responsible for the miRNA phenotypes. It is possible that the activity of specific signaling pathways and expression of mRNA targets dictate miRNA function. Therefore, to understand how miR-21 and miR-31 regulate the phenotypes, the mRNA targets regulated by miR-21 and miR-31 are identified to induce their effects. mRNA targets are computationally predicted from the tumor data, and are also consistent with the model systems. Most miR-31 validated targets have been characterized in situations where miR-31 inhibits aggressive disease and therefore may not be relevant to ovarian cancer.

Determining mRNA targets of miRNAs using computational approaches

In order to understand how miR-21 and miR-31 control metastatic tumors, putative targets that are likely being repressed by miRNAs are identified. Comparing mRNA and miRNA expression levels is a proven approach to identify mRNA targets as miRNAs often destabilize mRNAs, causing lower total RNA expression levels (Lim LP, et al. Nature.

2005;433(7027):769-73; Guo H, et al. Nature. 2010;466(7308):835-40). It is important to identify potential targets in the context of ovarian primary and metastatic tumors because of the large number of target genes that may or may not be expressed and regulated by miRNAs. In the case of miR-21, some previously proposed target genes, such as PDCD4, are viable candidates. In the case of miR-31 , which appears to behave differently in ovarian compared to breast cancer, identifying targets in the context of ovarian tumors is crucial.

Computational approaches are utilized to identify candidate mRNA targets from the patient data using correlation analysis. To determine significance, permutations of similar sized sets of mRNAs are examined. A variety of prediction databases are used to explore the full range of potential miRNA targets. Target mRNAs that are predicted from multiple algorithms suggest higher confidence predictions.

Correlation analysis to identify mRNA targets

Determining which transcripts are controlled by miRNAs provides insights into mechanisms of the observed phenotypes (Valastyan S, et al. Cancer Res. 2010;70(12):5147- 54. PMCID: 2891350; Cohen EE, et al. Cancer Res. 2009;69(l):65-74. PMCID: 2746005; Guo H, I et al. Nature. 2010;466(7308):835-40). Figure 7 demonstrates that the global distribution of Spearman correlation coefficients is notably different for miR31 and miR-21 predicted targets compared to randomly selected sets of mRNAs. Spearman correlation tests for a linear relationship between miRNA and targets across the patients and provides stronger support for a concentration dependent relationship than a Pearson correlation. These observations link the change in miRNA level to a concomitant change in mRNA expression of predicted targets, suggesting that the observed miRNA change is likely functional in the tumor setting.

Determining mRNA targets is complicated by the observation that miRNAs often subtly control the expression of hundreds of genes by relatively small amounts, i.e., tuning expression (Flynt AS and Lai EC. Nat Rev Genet. 2008;9(11):831-42. PMCID: 2729318). Moreover, prediction algorithms suffer from very high false positive rates, as recently reviewed (Thomas M, et al. Nat Struct Mol Biol. 2010; 17( 10): 1169-74). Nonetheless, the prediction programs define the range of interactions possible and comparing to the mRNA expression data leads to insights into miRNA action. Because many of the prediction algorithms use a variety of parameters including the seed region, conservation,

thermodynamic stability, and local sequence context, a common approach is to use multiple prediction programs (TargetScan, Pictar, PITA, and miRBase) and focus on the overlap to identify the most likely mRNA targets (Thomas M, et al. Nat Struct Mol Biol.

2010;17(10): 1169-74). Literature -validated databases, such as TarBase and MiRDisease, are becoming more extensive and represent refined sets of miRNA targets. miRNA regulated pathways are assessed by examining enrichment for the negatively correlated miRNA targets using tools such as Ingenuity Pathway Analysis (IP A). To gain confidence that the predicted mRNA targets are functionally relevant as a group, their ability to classify patients by survival is examined by Kaplan-Meier analysis, similar to Figure 6.

The analysis described herein has identified likely mRNA targets involved in processes associated with metastasis such as cellular movement (STK38L), apoptosis (PDCD4, BCL2) and pathways such as ERK5 signaling (Figure 7). This approach indicates that other miR-21 target genes in the literature such as RhoB (Connolly EC, et al. Mol Cancer Res. 2010;8(5):691-700; Sabatel C, et al. PLoS One. 2011;6(2):el6979. PMCID: 3037403) may not be active in ovarian tumors (Spearman = -0.06) and thus reduces the number of candidate targets to consider. This commonly-used subjective approach is utilized to identify a list of 12-16 genes to examine because of the high false positive rates from the predictions. Candidate genes of interest from the data presented herein using TargetScan and PITA miRNA target predictions are shown in Figure 7C. To gain more confidence in these associations from the tumors, the putative interaction is validated and the functional relationship is defined.

For example, if miR-21 mediates apoptosis, a literature validated miR-21 target such as the tumor suppressor PDCD4, which promotes apoptosis is a good candidate to examine (Figure 7B, 7C). Candidate genes such as the MMP-9 inhibitor RECK are strong candidates to mediate migration and/or invasion based on strong negative correlation with miR-21 expression. For miR-31 , predicted targets such as NFAT5, previously linked to metastasis (Levy C, e al. Mol Cell. 2010;40(5):841-9. PMCID: 3004467) negatively correlated in the tumor data are viable candidates. Antibodies are utilized when available to allow protein level studies. A number of refinements to improve miRNA prediction are used in this analysis. A recent model indicates that higher expression of putative miRNA targets reduces miRNA efficacy on each individual target (Arvey A, et al. Mol Syst Biol. 2010;6:363. PMCID:

2872614). This additional metric is incorporated to refine the predictions for the most likely mRNA targets. The requirements for miR-21 and miR-31 activity is also examined. Because identifying miRNA mRNA targets can be challenging, a second approach is to determine which common growth pathways {e.g., A T or WNT) are critical for the mediation of miRNA activity.

Examining predicted targets in luciferase 3' UTR assays in 2D and 3D cell culture systems

The prediction programs have high false positive rates compared to experimental data (Thomas M, Nat Struct Mol Biol. 2010; 17( 10): 1169-74). This has recently been exemplified by Weinberg's miR-31 study where only 6 of 16 tested mRNA targets responded in 3' UTR luciferase assays (Valastyan S, et al. Cell. 2009;137(6): 1032-4). Therefore, 12-16 predicted miRNA-mRNA pairs are examined using luciferase reporters to gain confidence that a proposed interaction is real in an effort to identify five validated interactions for each miRNA. Because phenotypes are observed in MCAs, but not 2D culture for miR-21 and miR-31 , it is hypothesized that the targets of most interest are more repressed, or even specifically repressed in MCAs compared to 2D culture.

To test the predicted target mRNA-miRNA relationships in ovarian cancer cell lines, 3 ' UTR luciferase reporters are used in at least two cell lines to model the genetic diversity of ovarian cancer. Because miR-21 and miR-31 effects in spheroids, but not standard 2D culture conditions are observed, miRNA repression is tested with luciferase assays in 2D and 3D culture. To validate computationally identified mRNA targets, luciferase 3 'UTR reporters are utilized. Reporters for most genes are commercially available. Co-transfection of anti-mir with luciferase reporters and Renilla luciferase to control for transfection and general translation tests for two-fold increases in luciferase activity upon miRNA inhibition. Mutant binding sites test if the predicted miRNA binding is responsible for any observed miRNA dependent changes in luciferase expression.

To identify targets likely important for the observed miRNA phenotypes, miRNA dependent luciferase expression is compared in 2D and 3D culture. True miRNA targets may respond stronger, or even specifically, in 3D culture based on the observations described herein that miR-21 and miR-31 affect multiple cell lines in 3D, but not 2D culture conditions. MCAs are collected into 96 well plates for measuring luciferase luminescence.

Complementing the reporter assays, it is also determined if the endogenously expressed predicted mRNA targets increase expression when the miRNA is inhibited in 2D and 3D cell culture. This would provide evidence of a relationship at physiological expression levels.

Identifying targets helps establish the mechanism of miRNA function and increase confidence in the interpretation of the global analysis. Inhibition of miRNAs with PNAs increases reporter activity while not affecting mutant reporters. Despite concern of non- physiological expression levels, 3' UTR luciferase reporters remain the standard to test miRNA-mRNA targets. miRNAs often lead to strong phenotypes by affecting multiple transcripts and the effects may be relatively modest on individual mRNAs (Karres JS, Cell. 2007;131(l): 136-45; Biryukova I, et al. Dev Biol. 2009;327(2):487-96; Iovino N, et al. Dev Cell. 2009;17(l): 123-33; Wu CI, et al. Genome Res. 2009;19(5):734-43). In this case, genomic approaches are utilized to directly measure mRNA targets in vitro from 2D and 3D cultures. RNA is collected after inhibition and interrogated by RNA-seq to determine if mRNAs with seed regions are significantly enriched. RNA-seq is preferred to microarrays because of the better dynamic range. To link miRNA-mRNA targets, prediction algorithms are utilized. Transcripts coding for genes likely to be involved in adhesion and integrin remodeling are of particular interest, as are genes regulating the cell cycle and growth factor pathways. Functional enrichment of the mRNA targets provides insight into possible functions mediated by the miRNA.

Determining mRNA targets responsible for miRNA function in vitro

Although miRNAs appear to regulate multiple target mRNAs, paradoxically, specific mRNA targets phenocopy the miRNA, suggesting that much of the miRNA function depends on regulating a single gene. Various hypotheses have been proposed to explain these observations including that many targets may be non-functional and serve to control active miRNA concentrations (Seitz H. Curr Biol. 2009;19(10):870-3). Here, the conventional model often seen in cancer cells is examined, where one or two genes are able to phenocopy the miRNA. To determine if the putative miRNA-mRNA targets confer the observed miRNA phenotypes, a combination of ectopic expression and knock-down experiments is used. About five targets are examined for each miRNA. The strongest miRNA phenotypes suitable for transient transfection experiments, i.e., spheroid formation or migration, are examined first.

Upon miRNA inhibition, mRNA targets have higher expression. Two general approaches are undertaken to test the function of mRNA targets. 1) mRNA targets are ectopically expressed to determine if the gene phenocopies miRNA inhibition. Using miRNA resistant mutant 3'UTRs, the targets are tested to complement the miRNA inhibition. Careful selection of promoters aims to minimize the over-expression to levels comparable to those seen in endogenous transcripts. If necessary, stable expression using lentiviral constructs are used to reduce the copy number. miRNA targets are co-transfected with mutated 3'UTRs resistant to miRNA regulation, with anti-mirs, to test for rescue of miRNA inhibition. Particular attention is focused on where inhibition of miRNAs has their strongest effects, such as in spheroid formation. 2) mRNA targets are knocked-down by siRNA to test if this inhibits miRNA inhibition. Reducing the target expression level prevents the expected increase in target expression from miRNA inhibition and inhibit the effects of miRNA inhibition. mRNA expression levels are determined by real-time qPCR. Western blot for factors with available antibodies determines protein levels to check knock-downs.

A reporter validated mRNA target that pheno copies miR inhibition suggests that a single target likely is responsible for the miRNA effects. For consistency, and to help show that ectopic expression is not non-specific, the expression of the predicted target is knocked down to test for inhibition of the miRNA phenotype. If targets do not phenocopy the miRNA phenotypes, it is possible that multiple targets need to be repressed to induce the miRNA phenotype. One model suggests that miRNAs induce their phenotypes by modestly repressing expression of many genes as opposed to strong repression of a gene or two.

Multiple (2, 3, 4, or perhaps all 5) putative targets are expressed at the same time to test if multiple genes need to be expressed to phenocopy the miRNA effects. To pursue a more genomic approach if the candidate approach fails, RNA is collected from miRNA inhibited cells and all significantly expressed mRNAs are measured using multiplexed RNA-seq. These data identify transcripts with predicted miR-21 or miR-31 binding sites that have increased expression. Any transcripts that are also negatively correlated in the tumor data are strong candidates for validation and functional analysis.

The experiments described herein evaluate the dissemination of ovarian cancer.

Collecting tumor tissue in a defined manner uncovers specific tumor cells that are most likely to metastasize within each tumor and improve identification of miRNAs important for metastasis. Also described herein are specific miRNAs involved in mediating metastasis and tumor progression of this lethal disease.

Example 6. Determining whether miR-21 and miR-31 inhibit dissemination in xenograft models.

By definition, metastasis is a process that occurs in vivo; therefore, the function of miR-21 and miR-31 and their predicted mRNA targets is examined in xenograft models. It is determined if miR-21 or miR-31 are important for dissemination of ovarian cancer cells in order to establish these factors and their regulated networks as therapeutic targets. To test if the xenograft models mimic progression observed in human disease, the miR A expression levels between primary tumors and metastases are compared.

Mouse model

miR-21 and miR-31 expression levels may go up as the cancer cells metastasize, consistent with the clinical observations. The intrabursal injections (IB) model is used, where a primary tumor is isolated and omental metastases are collected for expression analysis. The expression changes in four commonly used cell lines (SKOV-3, OVCAR-8, ES-2 and A2780) are characterized. These data are used, in combination with the in vitro data, to determine which cell line to use to test the role of miR-21 and miR-31 directly.

Tumors are collected from four mice to control for variation similar to previously published experiments (Levy C, et al. Mol Cell. 2010;40(5):841-9. PMCID: 3004467). To test how similar the animal model may be with clinical data, primary and omental metastases are collected from each animal and miRNA expression is measured by Taqman qPCR. The whole panel is measured using Taqman qPCR, with particular focus on the candidate miRNAs, miR-21 and miR-31. By measuring hundreds of miRNA, the background is determined and it is possible to discriminate between global and specific effects on miRNA expression. The expression levels of each miRNA may not be similar to human tumors, but the change up or down may be similar.

Similar changes in expression of some of the same miRNAs as observed in human tumors support the use of these xenograft models. The four cell lines (SKOV-3, OVCAR-8, ES-2 and A2780) are expected to induce a range of animal survival (Table 4) allowing a correlation between miRNA expression and aggressiveness. Other metrics such as the number of lesions, and time needed to disseminate using luciferase longitudinal

measurements provide additional insight in the evolving xenograft and miRNA expression. miRNA expression is monitored over time to improve the ability to capture the key miRNA features of the tumors during disease progression. Other strongly associated miRNAs identified in these studies are also examined.

Inhibition of miRNAs in xenograft models

It is examined if miR-21 or miR-31 affects metastasis by allowing the primary tumor to form and disseminate before adding a miRNA inhibitor by Intraperitoneal injection (IP), such as LNA or PNA used in mice previously (Huynh C, et al. Oncogene. 2011 ;30(12): 1481- 8). This model may best represent the clinical scenario because it would be determined if inhibiting a miRNA can reverse, or least prevent, further dissemination in the abdominopelvic cavity.

Cancer cells are delivered by injecting the human cancer cells orthotopically into the bursa surrounding the ovaries of nude mice (IB). In this model, a localized tumor develops before spreading within the abdominal cavity, presenting the opportunity to define a primary site. The alternative IP injection leads to disseminated tumor cells throughout the peritoneal cavity. Either model is suitable to test the effects of miRNA inhibition, but characterizing expression changes to compare the mouse model with metastasis in human disease would be easier in the IB model. Tumor growth and dissemination is monitored longitudinally using luciferase expressing ovarian cancer cell lines. This would allow more precise timing to deliver the inhibitor as well as determine its effects. SKOV-3 luciferase has been extensively used (Huynh C, et al. Oncogene. 2011;30(12): 1481-8). If other cell lines prove more miR-21 or miR-31 sensitive, then stably expressing luciferase lines are created using G418 selection or lentiviral integration. Luciferase expressing cell lines are tested to ensure that they behave similarly to the parental cells in xenograft assays in terms of survival and metastasis as evaluated by visual inspection of the peritoneal cavity.

Metastases size and number, primary tumor size, and mouse survival are examined. The experiments are carried out with fifteen animals, with ten animals in each group treated with specific and non-specific inhibitors. A miRNA inhibitor such as Locked Nucleic Acids from Exiqon (LNA) delivered IP is utilized. These inhibitors are synthesized with modifications, such as phosphorothioate backbones, to allow delivery to cancer cells.

Twenty- five mg/kg of an anti-mir has been shown to inhibit liver metastasis (Huynh C, et al. Oncogene. 2011;30(12): 1481-8). The anti-mir is compared to PBS treated tumors. Delivery efficacy and specificity is determined by measuring miRNA and mRNA target levels from the xenografts. Ki-67 staining and TUNEL assays help determine the robustness of the tumor and possible anti-miR effects.

To determine if miRNA inhibition affects the same mRNA targets studies in vitro, the expression of identified mRNA targets is examined in anti-mir and control xenograft tumors. This correlation would link the in vitro mechanisms to the mouse models.

Based on the spheroid, soft agar, and clinical observations, it is anticipated that miR-21 and miR-31 inhibit dissemination and tumor growth in the xenograft model, as determined by fewer lesions and perhaps smaller tumor size. This is examined by

longitudinal luciferase imaging and confirmed by visual inspection at the end of the experiment (and necropsy). In the case of miR-21, which inhibits colony formation of ES-2 cells, it is likely to affect tumor growth. This would establish these two miRNAs as regulators of metastasis or metastamirs (Hurst DR, et al. Cancer Res. 2009;69(19):7495-8).

Because the dissemination xenograft models in combination with delivery of an anti- mir is hallenging, the role of miR-21 and miR-31 is examined by using genetic manipulations such as miRNA sponges (Ebert MS, Nat Methods. 2007;4(9):721-6). Sponges are transcripts with multiple miRNA binding sites designed to sequester the miRNA from its endogenous targets and thus reduce its efficacy. Because miR-21 and miR-31 do not significantly affect proliferation in at least some cell lines (OVCAR-8 and ES-2), a stably expressing sponge ovarian cell line is generated without significant concerns of inducing a variant subtype. Stably expressing sponge cells include eGFP for in vivo imaging of the dissemination of tumor cells. Sponges targeting miR-31 (Valastyan S, et al. Cell. 2009; 137(6): 1032-46) and miR-21 (Ebert MS, et al. Nat Methods. 2007;4(9):721-6) have been successfully constructed and used. Since a sponge-expressing cell line has reduced miRNA activity before injection into nude mice, the miRNA may also affect tumorigenesis allowing for the monitoring of the role of these miRNAs in both tumor formation and dissemination. If the tumor formation effect is very strong, dissemination effects are isolated by expressing the sponge under the control of a drug-inducible promoter (Ebert MS and Sharp PA. RNA. 2010;16(11):2043-50. PMCID: 2957044; Brown BD and Naldini L. Nat Rev Genet. 2009;10(8):578-85) to allow the tumors to develop before inhibiting the miRNA.

Example 7: Expression Profiling of Primary and Metastatic Ovarian Tumors Reveals a Metastatic Predictive Gene Signature

Serous epithelial ovarian cancer (EOC) patients often succumb to aggressive metastatic disease, yet little is known about the behavior and genetics of ovarian cancer metastasis. As described herein, the DNA copy number and mRNA expression differences between matched primary human tumors and omental metastases were characterized. mRNA expression data reveal hundreds of genes with recurring significant expression differences across the patient cohort. Pathway analysis reveals differential expression of pathways including increased expression of anti-apoptotic factors in metastases. Metastatic cancer cells are more proliferative and less apoptotic than in primary tumors, explaining the large size and aggressive nature of these lesions. Copy number aberrations were detected that differ between primary and metastatic tumors, but are not recurrent and do not explain the mRNA expression data. Genes differentially expressed between metastatic and primary tumors robustly identify a poor prognosis subtype in primary tumors, suggesting that primary EOC tumors encode significant metastatic potential and the molecular features that manifest in omental lesions are clinically important for patient survival. A refined 16 gene prognostic expression signature was identified, validated in multiple datasets, expression platforms, and independent of clinical variables by multivariate analysis. Together, these data highlight how metastatic tumors evolve into a state distinct from primary tumors, and that many of the differentially expressed genes identify a promising independent prognostic expression signature of survival of EOC patients.

Serous Epithelial Ovarian Cancer (EOC) is an aggressive disease for which there are few effective biomarkers and therapies. EOC is often diagnosed after tumor cells have disseminated within the peritoneal cavity (Lengyel E (2010) Am J Pathol 177 (3): 1053-1064) and metastases account for the majority of disease-related deaths. Despite its vital role in disease progression, however, the features required for ovarian cancer metastasis remain poorly understood (Lengyel E (2010) Am J Pathol 177 (3): 1053-1064). Ovarian tumors do not typically spread through a hematogenous route, but rather shed from the primary tumor and enter the peritoneal fluid. Primary ovarian tumors typically spread within the peritoneal cavity, most often to the omentum.

In other cancers, genomic remodeling has been observed when comparing primary and metastatic tumors (Paris PL, et al. (2006) Neoplasia 8 (12): 1083-1089; Shah SPet al. (2009) Nature 461 (7265):809-813; Yachida Set al. (2010) Nature 467 (7319): 1114-1117). Copy number changes, large structural variants, and more recently, point mutations identified by next-generation sequencing suggest that specific genetic differences are found in metastases compared to primary tumors from the same patient (Shah SPet al. (2009) Nature 461 (7265):809-813; Yachida Set al. (2010) Nature 467 (7319): 1114-1117). In ovarian cancer, studies suggest that metastases form by clonal expansion based on 22 microsatellite markers (Khalique Let al. (2009) Int J Cancer 124 (7): 1579-1586).

Separate studies, comparing matched ovarian primary and metastatic tumors from the same patient by copy number and mRNA expression, support a "primary tumor

predisposition" model (Ramaswamy S, et al. (2003) Nat Genet 33 (l):49-54; Colella S, et al. (2008) Head Neck 30 (10): 1273-1283; Liu CJ, et al. (2008) J Pathol 214 (4):489-497; Paris PL, et al. (2004) Hum Mol Genet 13 (13): 1303-1313). mRNA expression data using early generation microarrays suggest there are few significant expression differences between omental lesions and primary tumors (Adib TR, et al. (2004) Br J Cancer 90 (3):686-692; Hibbs K, et al. (2004) Am J Pathol 165 (2):397-414; Lancaster JM, et al. (2006) Int J Gynecol Cancer 16 (5): 1733-1745); however, numerous studies testing specific functions observe differential expression of factors between primary tumors and metastases including MMPs (Cowden Dahl KD, et al. (2008) Cancer Res 68 (12):4606-4613; Moss NM, et al. (2009) Cancer Res 69 (17):7121-7129) and integrins (Sawada K, et al. (2008) Cancer Res 68 (7):2329-2339).

The purpose of the experiments described herein was to identify features that are important in establishing metastases. As the disease progresses, specific cancer cells with distinct genomes and phenotypes are selected. Comparing primary and metastatic tumors has generated important insights into disease progression in both animal models (Bos PD, et al. (2009) Nature 459 (7249): 1005-1009) and in patients (Paris PL, et al. (2006) Neoplasia 8 (12): 1083-1089). To improve treatment of metastatic disease, it is vital to understand the genes and pathways expressed in metastases as many differences are possible from primary tumors. As described herein, significant differences were observed between the primary and omental metastatic tumors by gene expression microarray analysis. The copy number alterations that differ between matched primary and metastatic tumors do not explain the recurring expression differences that define common features of metastasis. Up-regulated signaling pathways, including TGFP signaling, suggest that tumor cells are adapting to the new omental environment. qPCR and immunohistochemistry support the microarray findings that metastases appear to be more proliferative and have less apoptotic cells than primary tumors. Described herein is a "metastatic expression signature" of the most significantly differentially expressed genes between primary and metastatic tumors. This signature, enriched for adhesion and extracellular matrix genes, identifies poor prognosis patients by Kaplan-Meier analysis in two large independent primary tumor datasets. In sum, the data described herein suggest that metastatic tumors progress into a more aggressive state with features distinct from the primary tumors that prove to be more indicative of the disease that needs to be treated.

Materials and Methods

Patient material

Ovarian tumors were collected from de-identified cases using standard methods All patients were over age 55, stage III or later, and all tumors were chemo-nai^'ve. A pathologist specializing in gynecological cancers examined all specimens (MS). Samples were snap frozen with no fixation.

RNA isolation and Affymetrix microarrays

Tumor tissue with >70% cancer cells was homogenized with a Tekmar Tissumizer (Cincinnati, OH). RNA was purified using miRNeasy kit (Qiagen, Valencia, CA) following the manufacturer's instructions. Nugen WT-Ovation Pico kit with the WT-Ovation Exon Module was used to prepare the RNA for Affymetrix Human Gene St vl .0 microarrays following manufacturer instructions in the Brown University Center for Genomics and Proteomics core facility. Data was quantile normalized and signals were estimated using Robust Multi-array Average (RMA). Samples were prepared in two batches and a modest difference in array performance between the two batches was observed for some genes. Genes with consistent signal below the lowest quartile were removed. P-values were calculated by paired t-tests. To identify enriched pathways, Ingenuity Pathway Analysis (IP A) and Gene Set Enrichment (GSEA) was used. Data are deposited to the Gene

Expression Omnibus in series GSE30587.

DNA purification and copy number analysis

DNA was isolated using Qiagen DNeasy Blood and Tissue Kit following

manufacturer's instructions. DNA quality was determined on an agarose gel. DNA was prepared for Agilent 180K CGH microarrays using the Roche NimbleGen (Madison, WI) enzymatic labeling protocol using random nonamers and hybridized following

manufacturer's protocols at the Microarray Centre at the Prostate Centre in Vancouver, British Columbia, Canada. Each sample was hybridized with Promega (Madison, WI) female reference DNA. The log 10 copy number ratios were smoothed using a standard deviation- based outlier detection method and segmented using Circular Binary Segmentation (CBS) (Venkatraman ES, Olshen AB (2007) Bioinformatics 23 (6):657-663) as implemented in the R package DNAcopy ('smooth.region=10' for the smoothing method, 'alpha=0.05' for the segmentation, and default parameters for all other arguments). A log 10 copy number for each gene was computed by averaging the smoothed and segmented log 10 ratios for each probe located within the gene region. Only genes that contained three or more probes were considered.

Quantitative Real-Time PCR

Equal amounts of total RNA were reverse transcribed using Superscript III and random hexamers (Invitrogen, Carlsbad, CA). Resulting cDNA was renormalized using Quant-iT PicoGreen (Invitrogen) before mixing with Power SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA). Reactions were performed in an Applied Biosystems 7900HT Fast Real-Time PCR System. The fold change was calculated as described previously using the internal control as stated for each patient (Bronson MW, et al. (2010) Mol Endocrinol 24 (6): 1120-1135).

Survival analysis Univariate Cox proportional hazard analysis was used to select the most significant predictive genes in the ovarian cancer TCGA datasets (The Cancer Genome Atlas Research Network (2011) Integrated genomic analyses of ovarian carcinoma. Nature 474 (7353):609- 615). Support vector machine (SVM) classified patients by the expression of these differentially expressed genes for Kaplan-Meier analysis. 1/3 of patients with most extreme survivals (1/6 shortest lived without censoring and 1/6 longest survivors) were used to train the classifier which was then applied to the remaining 2/3 of patients. They were labeled together with the training patients. The classifier was tested using a log-rank significance test.

Immunohistochemistry

4 μιη slices of formalin fixed paraffin embedded (FFPE) tissue were prepared from each tumor. Proliferation index was determined by staining with anti-Ki-67 antibody (Dako, Cat # IS626). Ki-67 percentage was evaluated manually by comparison with a schematic representation of various percentages of positive-staining nuclei. Quantification of apoptosis in primary and metastatic serous epithelial carcinoma was performed with a Terminal deoxynucleotidyl transferase dUTP Nick End Labeling assay (TUNEL). Each slice was stained with ApopTag Plus Peroxidase In Situ Apoptosis Kit (Millipore, Billerica, MA) according to the manufacturer's protocol. The whole tumor was imaged in a series of images with a Photometries CoolSNAP camera (Photometries, Tucson, AZ) under lOx magnification and assessed for positive apoptosis staining. An apoptotic index was calculated as the total number of apoptotic cells divided by the total number of tumor cells from 12 images across each slice. The distribution of percent apoptotic cells was not normal, but is log-normal as determined by Shapiro- Wilk tests and evaluation of quantile-quantile plots. Paired t-tests were calculated after log transformation of the percent apoptotic cells. The distribution of Ki-67 positive cells is normal and log transformation yields the same paired t-test p-values. Results

Patient and Sample Collection

Nineteen matched primary and omental metastatic tumor specimens were identified from patients with serous adenocarcinomas including 14 serous epithelial ovarian, 4 serous epithelial fallopian tube and one serous epithelial primary peritoneal (Figure 11). All patients are post-menopausal and had metastatic disease. mRNA gene expression was measured using Affymetrix Gene St microarrays and copy number using Agilent 180K CGH microarrays in each tumor.

Identification of Recurring Differentially Expressed Pathways To determine if there are recurring phenotypic changes between primary and metastatic tumors, the gene expression profiles were measured in nine of the twelve pairs of primary and metastatic tumors used for copy number analysis with sufficient RNA quality for analysis. The gene expression data reveal significant expression differences between the primary tumor and omental lesions. To identify transcripts with significant changes across the patient cohort, transcripts were selected with average fold changes >40% and p<0.05 in a paired t-test, identifying 467 up-regulated and 168 down-regulated genes between matched primary and omental tumors. To validate the microarray observations, 11 transcripts were sampled by real-time qPCR in two cases revealing good correlation with the microarray data (Figure 12). A number of genes previously reported to be up-regulated in metastases including ITGA5 (Sawada K, et al. (2008) Cancer Res 68 (7):2329-2339; Mitra AK, et al. (2010) Ligand-independent activation of c-Met by fibronectin and alpha(5)beta(l)-integrin regulates ovarian cancer invasion and metastasis. Oncogene) FABP4 (Lancaster JM, et al. (2006) Int J Gynecol Cancer 16 (5): 1733-1745; Nieman KM, et al. (2011) Nat Med 17 (11): 1498-1503), and FAK were observed (Mitra AK, et al. (2010) Ligand-independent activation of c-Met by fibronectin and alpha(5)beta(l)-integrin regulates ovarian cancer invasion and metastasis. Oncogene). A number of MMPs are up-regulated including MMP2, MMP13 and MMP14, consistent with their likely role in supporting invasion and establishing new lesions (Moss NM, et al. (2009) Cancer Res 69 (17):7121-7129).

Tumors evolve to be more proliferative and anti-apoptotic during metastasis

To identify cellular pathways differentially expressed between primary and metastatic tumors without using an arbitrary threshold, Gene Set Enrichment Analysis (GSEA) was used (Subramanian A, et al. (2005) Proc Natl Acad Sci U S A 102 (43): 15545-15550). Higher expression in metastases of cell communication, cell adhesion, and extracellular matrix receptor interactions are consistent with expected changes necessary for colonizing and establishing metastases. In primary tumors, higher expression of cell cycle and DNA repair factors were observed, suggesting possible changes in proliferation rates between the tumors (Figure 13). In particular, GSEA suggests that G2/M checkpoint factors are more highly expressed in primary tumors compared to metastases (Figure 8A). The expression levels of a subset of these genes were validated by qRT-PCR (Figure 8 A). Higher expression of key checkpoint regulators such as BUB IB in primary tumors suggests that metastases have a more significantly dysregulated cell cycle. This was examined by staining for Ki-67 in each pair of matched tumors from 19 cases (Figure 8A). Ki-67 showed higher staining in metastases in 12/19 cases, including 6/9 of the cases used for expression microarrays, suggesting that omental metastases are often proliferating more rapidly than their matched primary tumors (p=0.01, paired t-test).

To determine which expression changes may originate from cancer cells, the genes with expression changes that correlate with the change in Ki-67 immunostaining were identified. The top 250 genes were exammined, all with Pearson correlations >0.72, in DAVID(Huang da W, et al. (2007) Genome Biol 8 (9):R183), and observe significant enrichment of cell cycle, M phase, cell cycle progress, and DNA replication genes. This correlation analysis suggests that many of the cell cycle expression changes observed by microarray originate from cancer cells and not tumor infiltrating cells.

Higher expression of pro-survival genes is observed in the omental lesions including MYC, IRFl, BCL2L2, and TNFSFIO leading to biased expression of the gene ontology anti- apoptosis gene set as determined by GSEA (Figure 8B). Consistent with these gene expression changes, decreased TUNEL staining is observed in 12/15 tested metastases compared to primary tumors, including 7/9 of the cases measured on the microarrays (p=0.02, paired t-test). Together with the Ki-67 observations, these data suggest that cancer cells in metastases are often more proliferative and may be more inherently resistant to cell death induced by chemotherapeutics. Thus, ovarian metastases often evolve and adapt to the new omental environment in a more anti-apoptotic and proliferative state compared to primary tumors.

Changes in expression levels of a number of growth pathways were observed in metastases compared to primary tumors. Many of these changes likely contribute to the increased proliferation and reduced apoptosis. Higher expression of Axin2, DKK2, NKD1, and NKD2 contribute to enrichment of WNT/p-Catenin signaling in metastases (Figure 13) consistent with recent studies (Bitler BG, et al. (2011) Wnt5a Suppresses Epithelial Ovarian Cancer by Promoting Cellular Senescence. Cancer Res.; Burkhalter RJ, et al. (2011) J Biol Chem 286 (26):23467-23475). Other pathways linked to aggressive cancer include increased calcium signaling (NES=2.03, FDR=0.003), and increased Jak-Stat , Stat3, and NOTCH signaling in metastases.

Metastatic signature identifies more aggressive tumors

To evaluate whether the expression changes observed in metastases are important for disease progression, the prognostic capability of these genes was evaluated. Patients were classified using a Support Vector Machine (SVM) and evaluated the partition between high and low risk groups by Kaplan-Meier analysis. The top 100 genes with the lowest p-values, up-regulated in metastases, identify more aggressive disease using 432 patients from Affymetrix U133 Plus 2.0 arrays from TCGA data (Figure 9A) (The Cancer Genome Atlas Research Network (2011) Integrated genomic analyses of ovarian carcinoma. Nature 474 (7353):609-615). To reduce the number of genes in the signature, univariate proportional Cox hazard analysis (HR>1, p< 0.05) identifies 32 genes with strong predictive power starting from the list of 467 up-regulated genes (Figure 9B). To test if these genes were platform dependent, the same approach applied to the Agilent TCGA microarray data was used. A 28 gene signature was identified (Figure 9C). The 16 genes that are in both the Affymetrix 32 and the Agilent 28 gene signature were chosen for further evaluation. The reduced 16 gene signature maintains a large fraction of the discriminatory power of the larger 100 gene metastasis signature (Figure 9D-E and Table 7). This refined 16 gene signature maintains strong predictive power in an independent dataset from the Australian Ovarian Cancer Study Group (Tothill RW, et al. (2008) Clin Cancer Res 14 (16):5198-5208) (Figure 9F, Table 7). The 16 genes have higher expression in the high risk group, consistent with their up-regulation in metastases (Figure 14). Together, these data suggest that genes up- regulated in metastases identify more aggressive primary tumors that are more pre-disposed for aggressive disease and poorer patient survival.

Across multiple platforms, the high risk group has less than 20% survival (Figure 9A- C) and sometimes only 10%> survival (Figure 9E), suggesting strong stratification. Debulking status remains the strongest clinical predictor of survival in advanced EOC (Bristow RE, et al. (2002) J Clin Oncol 20 (5): 1248-1259). Combining the 16 gene signature with residual disease yields more distinct separation (Figure 9G-H and Figure 15). The low risk group has an approximately 70% five year survival for optimal debulking (residual disease <1 mm), depending on the platform, while the high risk group has approximately 20% survival in the TCGA datasets (Figure 9E and Figure 15). In the Tothill dataset (Tothill RW, et al. (2008) Clin Cancer Res 14 (16):5198-5208), the same 16 gene signature separates the high risk and low risk groups into almost no surviving patients, after censoring, and 90%> survival for patients with optimal debulking, <1 mm residual disease.

Multivariate Cox analysis indicates that the 16 gene signature is an independent predictor of outcome relative to residual disease and age (Table 6). Together, these observations suggest that the gene expression signature identifies aggressive tumors, independent of interventions. This 16 gene expression signature is enriched for cell-to-cell signaling and cellular movement, and 11 of the genes define a network enriched for tumor morphology and lipid metabolism (Figure 16). This signature includes a number of genes involved in intercellular signaling such as LOX and PLAUR, highlighting how expression originating from both cancer and infiltrating cells contribute to aggressive disease (Table 7). The expression of the plasminigoen activator and inhibitor PLAUR and SERPINE1 has been associated with cancer progression in a variety of cancers including ovarian (Kenny HA, et al. (2010) Targeting the Urokinase Plasminogen Activator Receptor Inhibits Ovarian Cancer Metastasis. Clin Cancer Res; Schmitt M, et al. (2011) Expert Rev Mol Diagn 11 (6):617- 634). Prior to the invention described herein, this combination of genes has not been identified to predict patient outcomes across expression platforms and patient datasets.

Genetically diverse cancer cells establish omental lesions with similar phenotvpes

To determine if copy number aberrations contribute to the recurring gene expression changes, DNA copy number from 12 matched primary and omental metastases was evaluated using Agilent 180K CGH microarrays (Figure 10A). Hierarchical clustering was performed to investigate relationships among 24 tumors from 12 patients based on their genomic alterations. These data suggest that 8/12 primary-met pairs are more genetically similar to each other than the other tumors, consistent with a clonal expansion model and past observations (Khalique L, et al. (2009) Int J Cancer 124 (7): 1579-1586). Notably, almost every region of the genome is represented in the cumulative spectrum of CNAs observed in the collection of tumors, consistent with the known extensive genomic instability of EOC (Gorringe KL, et al. (2010) PLoS One 5 (9)) (Figure 17).

Evaluation of each patient revealed a different CNA spectrum in matched primary and metastatic tumors (Figure 10B). Many CNAs are similar between the primary and metastatic tumor, as suggested by hierarchical clustering (Figure 10A); however, notable differences are observed in each patient suggesting the metastases are genetically distinct from the primary tumor.

Metastatic specific CNAs may indicate DNA copy number changes that are necessary to establish metastases. To determine if any metastatic specific CNAs are recurring, which may indicate functional importance, segmented regions were mapped to genes and a log2 ratio score for each gene was calculated. Genes in amplified or copy loss regions with high or low expression were selected. One amplified gene, ERBB2, unique to metastases and four genes uniquely amplified in primary tumors, and not in omental lesions, were observed in at least two patients where the expression and copy number are correlated (Pearson > 0.7) (PDCD10, ZNF507, DTNA, AMN1). Each of these genes has lower expression in all metastases, consistent with the role of copy number in modulating expression. However, a recent study examining ERBB2 expression and amplification, did not detect noticeable differences between primary and omental metastases, perhaps because of significant intratumoral heterogeneity (Tuefferd M, et al. (2007) PLoS One 2 (1 l):el 138). These data suggest that few, if any, specific CNAs, unique to metastases, are providing a selective advantage to establish metastases in the major clone in the omental lesions. Alternatively, the primary and metastatic tumors may be so heterogeneous that they are not dominated by a single clone detectable in bulk tumor samples.

The key findings described herein suggest that metastatic and primary tumors have different expression profiles reflecting their proliferation and apoptosis status. Significant recurrence is observed at the mRNA expression level, as hundreds of the same genes are differentially expressed between primary and metastatic tumors in most patients. On the other hand, copy number analysis did not reveal recurrent genetic changes in this cohort, suggesting that the dominant clone in metastases may differ from the primary tumor, or that these tumors have high intratumoral heterogeneity. The expression changes indicative of aggressive disease between primary tumors and omental lesions are common and important for disease progression. Importantly, genes differentially expressed between primary tumors and omental lesions identify a predictive expression signature of patient outcome, suggesting the functional importance of these genes in metastasis and disease progression. Together, these observations support the idea that ovarian primary tumors encode significant metastatic potential.

Past studies have concluded that ovarian primary and metastatic tumors have very similar expression profiles (Adib TR, et al. (2004) Br J Cancer 90 (3):686-692; Lancaster JM, et al. (2006) Int J Gynecol Cancer 16 (5): 1733-1745), differing from the observations. In part, this apparent discrepancy may be because of improvements in microarray technology, which are now more sensitive and robust than earlier generations. Second, these past studies applied multiple hypothesis corrections leading to the observation of very few differentially expressed genes. An uncorrected paired t-test, justified by qPCR sampling, was used to find hundreds of genes differentially expressed that are enriched for specific networks and gene ontologies. Third, differentially expressed pathways were observed recurring across all the patients including and support these observations with Ki-67 immunohistochemistry and TUNEL staining. Genes up-regulated in omental lesions identify patients with poor outcomes. Together, these observations support the conclusion that omental metastases have significantly different expression profiles than primary ovarian tumors and these expression differences identify key features of aggressive disease. Importantly, these data may help explain why omental metastases are particularly aggressive and large tumors as they are more proliferative and are more resistant to cell death than primary tumors. These data also may suggest that the omental metastases are more inherently resistant to chemotherapy than primary tumor cells in many patients.

These data suggest that ovarian cancer metastases often evolve to be more anti- apoptotic and more proliferative. These observations highlight the changing state of ovarian tumors that should be considered as various treatment options are developed. The data support the premise that the disease evolves into a state distinct from the primary tumor, e.g., a state that is influenced by the omental microenvironment. Thus, deeper examination of metastases is necessary to treat the disease, as distinct pathways may be activated in metastases. For example, WNT signaling is up-regulated in metastases (Figure 13) and may be more active in metastases and significant role in many metastases compared to primary tumors.

A number of differentially expressed pathways were identified using Ingenuity Pathway Analysis (IP A) for the 635 genes with p<0.05 and the more unbiased enrichment approach of GSEA. Most identified pathways have been linked to ovarian cancer. Top enriched functions in genes up-regulated in metastases include cellular movement (p<le-4), cell-to-cell signaling (p<le-4), cell proliferation and cell death (p<1.5e-4) and lipid metabolism (p<l .4e-4). These data suggest that the most significantly up-regulated genes are enriched for functions that promote more aggressive, invasive tumors, and enable cancer cells to adapt to the fat-rich omentum. After Benjamini-Hochberg multiple hypothesis correction, no pathways are significantly enriched in genes down-regulated in omental lesions. The top scoring IPA identified network enriched in genes up-regulated in metastases includes genes involved in intercellular signaling between the stroma and tumor cells, centered on TGFB1 (Figure 18). TGFB1 has recently been shown to stimulate ovarian tumor growth and metastasis in a xenograft mouse model (Padua D, Massague J (2009) Cell Res 19 (1):89-102; Yamamura S, et al. (201 1) The activated transforming growth factor-beta signaling pathway in peritoneal metastases is a potential therapeutic target in ovarian cancer. Int J Cancer.) and has been associated with tumor progression and metastasis in a number of cancers. These data provide confidence that differential expression of genes and pathways important for ovarian cancer progression were observed, and many of these pathways in combination may be critical to define the more aggressive state observed in the omental lesions.

For many of the identified networks, it is unclear if the expression changes are due to adaptation of the cancer cells or changes in the type and number of infiltrating cells. The evaluation by immunohistochemistry and correlation of expression with Ki-67 signal does suggest that many of the expression changes originate from cancer cells, as expected, given that the large majority, >70%, of cells are cancer cells. Nonetheless, some signal from infiltrating cells is certainly present as evidenced by increased expression of CD3 and other T-cell related genes. As described herein, because most large tumor characterization studies also use the same criteria as applied here, a predictive expression signature that appears to include genes that originate from both cancer and infiltrating cells was derived.

Genes differentially expressed between matched omental metastases and primary tumors identify more aggressive disease by Kaplan-Meier analysis, supporting their functional importance. A number of expression signatures have been proposed for ovarian cancer based on analysis of primary tumors (The Cancer Genome Atlas Research Network (2011) Integrated genomic analyses of ovarian carcinoma. Nature 474 (7353):609-615;

TothiU RW, et al. (2008) Clin Cancer Res 14 (16):5198-5208; Jochumsen KM, et al. (2007) Int J Gynecol Cancer 17 (5):979-985; Bonome T, et al. (2008) Cancer Res 68 (13):5478- 5486; Mok SC, et al. (2009) Cancer Cell 16 (6):521-532; Berchuck A, et al. (2009) Clin Cancer Res 15 (7):2448-2455; Konstantinopoulos PA, et al. (2011) PLoS One 6 (3):el8202).

As described herein, an alternative approach developed a candidate expression signature based on a biological foundation, namely the differential expression between matched primary and metastatic tumors. This gene expression signature classifies patients into high and low risk groups in two large independent datasets, which appears more significant than those found by classification of patients within each dataset based on clustering (Tothill RW, Tinker AV, George J, Brown R, Fox SB, Lade S, Johnson DS, Trivett MK, Etemadmoghadam D, Locandro B, Traficante N, Fereday S, Hung JA, Chiew YE, Haviv I, Gertig D, DeFazio A, Bowtell DD (2008) Novel molecular subtypes of serous and endometrioid ovarian cancer linked to clinical outcome. Clin Cancer Res 14 (16):5198-5208; The Cancer Genome Atlas Research Network (2011) Integrated genomic analyses of ovarian carcinoma. Nature 474 (7353):609-615).

The identification of a strong predictive expression signature for EOC is an important finding. Expression measurements were taken advantage of using different platforms performed by TCGA to assess the expression signature across different microarray platforms. In this manner, 16 genes were identified that behave similarly in different assays, providing more confidence of their true predictive power. The expression signature was further validated in a second large independent dataset where it was found that the 16 gene expression signature actually performs better than in TCGA data with higher hazard ratios (Table 6) and lower p-values by Kaplan-Meier analysis (compare Figure 9D, E and F).

Importantly, multivariate analysis suggests that the expression signature is independent of residual disease consistent with the identification of the genes as important in metastasis and thus these genes are not related to the amount of tumor not removable by surgery. Together, these observations strongly suggest that the 16 gene metastatic gene expression signature may be useful as a predictive biomarker, similar to Oncotype in breast cancer (Conlin AK, Seidman AD (2007) Mol Diagn Ther 11 (6):355-360) to reduce overtreatment with cytotoxic chemotherapy for patients who may not significantly benefit. Furthermore, the predictive expression signature, together with the Ki-67 and TUNEL data, suggests that the features differentially regulated between primary tumors and omental lesions identify a particularly aggressive form of the disease.

A common expression phenotype emerging in omental metastases that differs from primary tumors as observed. On the other hand, recurrent genetic changes uniquely emerging in omental metastases are less common. Together, these data suggest that various cancer genomes lead to the same metastatic phenotype, defined by common expression patterns. The diversity of genomes between primary and metastatic tumors in this cohort may be due to intratumoral or interpatient genetic heterogeneity or both. The data presented here suggest that expression changes are much more common than genetic changes between primary and metastatic tumors. Nonetheless, it appears that common expression changes are readily detectable and are critical for disease progression, as indicated by the classification of patient outcomes by the metastatic expression signature. Understanding the molecular features unique to omental lesions is critical to understanding and treating advanced ovarian cancer.

Table 6. Multivariate Analysis of the 16 gene expression signature.

Note: The difference in the residual and age parameters with the two TCGA datasets originates from slightly different patient lists with the data currently available.

Table 7. Summary of the 16 Gene Expression Signature.

Affymetrix Gene Previous reference

Gene Ketseq_lD Known Gene Name Function

St Probeset ID in EOC

ALDH1A3 NM. .000693 7986446 aldehyde dehydrogenase 1 family, member A3 Metabolism, glycolysis

ASNAl NM _004317 8026024 arsA arsenite transporter, ATP -binding, homolog 1 ATPase, anion-transporting, ion 1

(bacterial) homeostasis

CALB2 NM_ .001740 7997139 calbindin 2, transcript variant CALB2 Cell junction, calcium binding

CCDC49 NM. .017748 8014738 coiled-coil domain containing 49 (CWC25) Splicing factor

COPZ2 NM _016429 8016390 coatomer protein complex, subunit zeta 2 Vesicle transport, protein

transport

FXYD5 NM. .144779 8027778 FXYD domain containing ion transport regulator 5 ion transport, regulation cell

adhesion

GFPT2 NM. .005110 8116418 glutamme-fructose-6-phosphate transaminase 2 Metabolism, energy, glutamine

metabolism

KIF26B NM _018012 7911114 kinesin family member 26B Microtubule, kinesin motor

LOX NM. .002317 8113709 lysyl oxidase Blood vessel development, ECM

organization

LPL NM. .000237 8144917 lipoprotein lipase Fatty acid metabolism, PPAR

signaling

NNMT NM. .006169 7943998 nicotinamide N-methyltransferase Nicotinate and nicotinamide

metabolism

NUCB1 NM. 006184 8030133 nucleobindin 1 Calcium binding, Golgi calcium

homeostasis

PLAUR NM. .002659 8037374 plasminogen activator, urokinase receptor Plasminogen activation, 2

RGS1 NM. .002922 7908388 regulator of G-protein signaling 1 GTPase activating protein

S100A10 NM_ _002966 7920123 S 100 calcium binding protein Al 0 Calcium ion binding

SERPINEl NM. .000602 8135069 serpin peptidase inhibitor, clade E (nexin, plasminogen Fibrinolysis inhibitor 3,4

activator inhibitor type 1)

1. Hemmingsson O, Nojd M, Kao G, Naredi P (2009) Increased sensitivity to platinating agents and arsenite in human ovarian cancer by downregulation of ASNAl . Oncol Rep 22 (4):869-875.

2. Henic E, Borgfeldt C, Christensen U, Casslen B, Hoyer-Hansen G (2008) Cleaved forms of the urokinase plasminogen activator receptor in plasma have diagnostic potential and predict postoperative survival in patients with ovarian cancer. Clin Cancer Res 14 (18):5785-5793. doi: 14/18/5785 [pii] 10.1158/1078-0432. CCR-08-0096.

3. Schmitt M, Harbeck N, Brunner N, Janicke F, Meisner C, Muhlenweg B, Jansen H, Dorn J, Nitz U, Kantelhardt EJ, Thomssen C (2011) Cancer therapy trials employing level-of-evidence-1 disease forecast cancer biomarkers uPA and its inhibitor PAI-1. Expert Rev Mol Diagn 11 (6): 617-634. doi:10.1586/erm.l l.47.

4. Koensgen D, Mustea A, Klaman I, Sun P, Zafrakas M, Lichtenegger W, Denkert C, Dahl E, Sehouli J (2007) Expression analysis and RNA localization of PAI-RBP1 (SERBP1) in epithelial ovarian cancer: association with tumor progression. Gynecol Oncol 107 (2):266-273. doi:S0090-8258(07)00427-l [pii] 10.1016/j.ygyno.2007.06.023.

Example 8 : Metastatic miRNA Expression Signature Identifies High Risk Patients

As shown in Table 8, combinations of metastatic miRNAs identify high risk patients better than randomly selected miRNA combinations from the same data set. Figure 19 describes the proportion survival of patients from 0 to 120 months after diagnosis.

Table 8:

Table 9:

Exemplary metastatic miRNA combinations are provided in Tables 10-12 below.

Table 10: Combinations of 2 microRNAs (and their corresponding MiRBase Accession No. as of filing date of application).

MiRBase Accession Number.

microRNA microRNA Log rank p-value hsa-miR-193a-5p hsa-miR-370

(MI0000487) (MI0000778) 4.47E-03 hsa-miR-150 hsa-miR-370

(MI0000479) (MI0000778) 4.56E-03 hsa-let-7d hsa-miR-431

(MI0000065) (MI0001721) 5.00E-03 hsa-miR-150 hsa-let-7d

(MI0000479) (MI0000065) 5.50E-03 hsa-miR-31 hsa-miR-370

(MI0000089) (MI0000778) 5.75E-03 hsa-miR-150 hsa-miR-508-3p

(MI0000479) (AMI0003195) 6.03E-03 hsa-miR-31 hsa-miR-150

(MI0000089) (MI0000479) 6.21E-03 hsa-let-7d hsa-miR-370

(MI0000065) (MI0000778) 6.37E-03 hsa-miR-370

hsa-miR-124 (MI0000778) 7.39E-03 hsa-miR-214 hsa-miR-370

(MI0000290) (MI0000778) 7.91E-03 hsa-miR-21 hsa-miR-193a-5p

(MI0000077) (MI0000487) 9.68E-03 hsa-miR-193a-5p hsa-let-7d

(MI0000487) (MI0000065) 1.02E-02 hsa-miR-21 hsa-miR-370

(MI0000077) (MI0000778) 1.14E-02 hsa-miR-193a-5p hsa-miR-431

(MI0000487) (MI0001721) 1.20E-02 hsa-miR-152 hsa-miR-431

(MI0000462) (MI0001721) 1.20E-02 hsa-miR-150 hsa-miR-509-3-5p

(MI0000479) (MI0005717) 1.37E-02 hsa-let-7d hsa-miR-152

(MI0000065) (MI0000462) 1.51E-02 hsa-miR-150 hsa-miR-185

(MI0000479) (MI0000482) 1.60E-02 hsa-miR-708 hsa-miR-146a

(MI0005543) (MI0000477) 2.04E-02 hsa-miR-185 hsa-miR-431

(MI0000482) (MI0001721) 2.08E-02 hsa-miR-185 hsa-miR-193a-5p

(MI0000482) (MI0000487) 2.31E-02 hsa-miR-31 hsa-miR-431

(MI0000089) (MI0001721 ) 2.49E-02 hsa-miR-150 hsa-miR-214

(MI0000479) (MI0000290) 2.65E-02 hsa-miR-150

(MI0000479) hsa-miR-886-3p 2.76E-02 hsa-miR-150 hsa-miR-431

(MI0000479) (MI0001721 ) 2.84E-02 microRNA microRNA Log rank p-value

hsa-mi -31 hsa-miR-21

(MI0000089) (MI0000077) 3.09E-02

hsa-miR-708 hsa-miR-370

(MI0005543) (MI0000778) 3.13E-02

hsa-miR-150 hsa-miR-21

(MI0000479) (MI0000077) 3.14E-02

hsa-miR-150 hsa-miR-29a

(MI0000479) (MI0000087) 3.22E-02

hsa-miR-431

(MI0001721) hsa-miR-886-3p 3.40E-02

hsa-miR-21 hsa-miR-152

(MI0000077) (MI0000462) 4.14E-02

hsa-let-7d hsa-miR-214

(M 10000065) (Ml 0000290) 4.77E-02

Table 11: Combinations of 3 microRNAs (and their corresponding MiRBase Accession No as of filing date of patent application).

² MiRBase Accession Number. Log rank p- microRNA microRNA microRNA value hsa-miR-31 hsa-miR-146a

(M 10000089) (MI0000477) hsa-miR-886-3p 4.83E-06 hsa-miR-193a-5p hsa-miR-146a hsa-miR-508-5p

(M 10000487) (MI0000477) (M 10003195) 4.93E-06 hsa-miR-150 hsa-miR-193a-5p hsa-miR-152

(M 10000479) (MI0000487) (M 10000462) 5.02E-06 hsa-miR-508-3p hsa-miR-146a

(AMI0003195) (MI0000477) hsa-miR-886-3p 5.27E-06 hsa-miR-150 hsa-miR-508-3p hsa-miR-146a

(M 10000479) (AMI0003195) (M 10000477) 5.38E-06 hsa-miR-508-3p hsa-miR-193a-5p hsa-miR-146a

(AMI0003195) (MI0000487) (M 10000477) 5.56E-06 hsa-miR-150 hsa-miR-146a hsa-miR-509-3-5p

(M 10000479) (MI0000477) (MI0005717) 6.07E-06 hsa-miR-150 hsa-miR-146a hsa-miR-370

(M 10000479) (MI0000477) (MI0000778) 7.16E-06 hsa-miR-31 hsa-miR-21 hsa-miR-146a

(MI0000089) (MI0000077) (M 10000477) 7.72E-06 hsa-miR-150 hsa-miR-146a hsa-miR-508-5p

(M 10000479) (MI0000477) (MI0003195) 7.72E-06 hsa-miR-185 hsa-miR-370 hsa-miR-431

(M 10000482) (MI0000778) (MI0001721) 8.25E-06 hsa-miR-29a hsa-miR-146a hsa-miR-509-3-5p

(M 10000087) (MI0000477) (MI0005717) 8.62E-06 hsa-miR-508-3p hsa-miR-185 hsa-miR-146a

(AMI0003195) (MI0000482) (M 10000477) 8.63E-06 hsa-let-7d hsa-miR-146a hsa-miR-509-3-5p

(M 10000065) (MI0000477) (MI0005717) 8.72E-06 hsa-miR-29a hsa-miR-146a hsa-miR-152

(M 10000087) (MI0000477) (M 10000462) 8.77E-06 hsa-miR-508-3p hsa-miR-146a

hsa-miR-124 (AMI0003195) (M 10000477) 8.89E-06 hsa-miR-29a hsa-miR-146a hsa-miR-370

(M 10000087) (MI0000477) (MI0000778) 9.57E-06 hsa-miR-185 hsa-miR-146a hsa-miR-370

(M 10000482) (MI0000477) (MI0000778) 9.98E-06 hsa-miR-31 hsa-miR-150 hsa-miR-146a

(M 10000089) (MI0000479) (M 10000477) 1.10E-05 hsa-miR-31 hsa-miR-146a hsa-miR-431

(M 10000089) (MI0000477) (M 10001721) 1.16E-05 hsa-miR-150 hsa-miR-508-3p hsa-miR-370

(M 10000479) (AMI0003195) (MI0000778) 1.29E-05 hsa-miR-508-3p hsa-let-7d hsa-miR-146a

(AMI0003195) (MI0000065) (M 10000477) 1.34E-05 hsa-miR-146a hsa-miR-370

hsa-miR-124 (MI0000477) (MI0000778) 1.47E-05 hsa-let-7d hsa-miR-146a hsa-miR-370

(M 10000065) (MI0000477) (MI0000778) 1.48E-05 hsa-miR-146a hsa-miR-509-3-5p

hsa-miR-124 (MI0000477) (MI0005717) 1.52E-05 Log rank p- microRNA microRNA microRNA value hsa-miR-146a hsa-miR-431 hsa-miR-508-5p

(M 10000477) (MI0001721) (M 10003195) 1.54E-05 hsa-let-7d hsa-miR-146a

hsa-miR-124 (MI0000065) (M 10000477) 1.78E-05 hsa-miR-21 hsa-miR-146a hsa-miR-370

(M 10000077) (MI0000477) (MI0000778) 1.78E-05 hsa-miR-508-3p hsa-miR-146a hsa-miR-431

(AMI0003195) (MI0000477) (M 10001721) 1.94E-05 hsa-miR-31 hsa-let-7d hsa-miR-146a

(M 10000089) (MI0000065) (M 10000477) 2.11E-05 hsa-miR-508-3p hsa-miR-29a hsa-miR-146a

(AMI0003195) (MI0000087) (M 10000477) 2.21E-05 hsa-miR-193a-5p hsa-miR-146a hsa-miR-214

(M 10000487) (MI0000477) (MI0000290) 2.22E-05 hsa-miR-146a hsa-miR-509-3-5p

(M 10000477) (MI0005717) hsa-miR-886-3p 2.36E-05 hsa-miR-146a hsa-miR-370

(M 10000477) (MI0000778) hsa-miR-886-3p 2.42E-05 hsa-miR-185 hsa-let-7d hsa-miR-146a

(M 10000482) (MI0000065) (M 10000477) 2.43E-05 hsa-miR-31 hsa-miR-146a hsa-miR-508-5p

(M 10000089) (MI0000477) (MI0003195) 2.60E-05 hsa-miR-21 hsa-miR-185 hsa-miR-146a

(M 10000077) (MI0000482) (M 10000477) 2.86E-05 hsa-miR-31 hsa-miR-193a-5p hsa-miR-146a

(M 10000089) (MI0000487) (M 10000477) 2.95E-05 hsa-miR-193a-5p hsa-miR-146a hsa-miR-152

(M 10000487) (MI0000477) (M 10000462) 3.28E-05 hsa-miR-150 hsa-let-7d hsa-miR-370

(M 10000479) (MI0000065) (MI0000778) 3.42E-05 hsa-miR-185 hsa-miR-146a hsa-miR-509-3-5p

(M 10000482) (MI0000477) (MI0005717) 3.51E-05 hsa-miR-146a hsa-miR-152 hsa-miR-370

(M 10000477) (MI0000462) (MI0000778) 3.75E-05 hsa-miR-150 hsa-miR-185 hsa-miR-370

(M 10000479) (MI0000482) (MI0000778) 3.79E-05 hsa-miR-185 hsa-miR-146a

(M 10000482) (MI0000477) hsa-miR-886-3p 3.85E-05 hsa-miR-508-3p hsa-miR-146a hsa-miR-214

(AMI0003195) (MI0000477) (MI0000290) 4.12E-05 hsa-miR-185 hsa-miR-146a

hsa-miR-124 (MI0000482) (M 10000477) 4.85E-05 hsa-miR-508-3p hsa-miR-146a hsa-miR-509-3-5p

(AMI0003195) (MI0000477) (MI0005717) 4.87E-05 hsa-miR-150 hsa-miR-146a

(M 10000479) (MI0000477) hsa-miR-886-3p 5.12E-05 hsa-miR-150 hsa-miR-152 hsa-miR-431

(M 10000479) (MI0000462) (MI0001721) 5.55E-05 hsa-miR-150 hsa-miR-146a hsa-miR-152

(M 10000479) (MI0000477) (M 10000462) 5.79E-05 Log rank p- microRNA microRNA microRNA value hsa-miR-146a hsa-miR-370 hsa-miR-431

(M 10000477) (MI0000778) (MI0001721) 6.28E-05 hsa-miR-150 hsa-miR-508-3p hsa-miR-152

(M 10000479) (AMI0003195) (M 10000462) 6.30E-05 hsa-miR-150 hsa-miR-370 hsa-miR-509-3-5p

(M 10000479) (MI0000778) (MI0005717) 6.38E-05 hsa-miR-150 hsa-miR-152 hsa-miR-508-5p

(M 10000479) (MI0000462) (MI0003195) 6.43E-05 hsa-miR-21 hsa-miR-29a hsa-miR-146a

(M 10000077) (MI0000087) (M 10000477) 6.60E-05 hsa-miR-150 hsa-miR-152

(M 10000479) (MI0000462) hsa-miR-886-3p 6.65E-05 hsa-miR-146a hsa-miR-152 hsa-miR-508-5p

(M 10000477) (MI0000462) (MI0003195) 6.77E-05 hsa-miR-21 hsa-miR-146a hsa-miR-509-3-5p

(M 10000077) (MI0000477) (MI0005717) 7.03E-05 hsa-miR-150 hsa-miR-370

(M 10000479) (MI0000778) hsa-miR-886-3p 7.28E-05 hsa-let-7d hsa-miR-146a

(M 10000065) (MI0000477) hsa-miR-886-3p 7.86E-05 hsa-miR-193a-5p hsa-miR-146a hsa-miR-431

(M 10000487) (MI0000477) (MI0001721) 8.19E-05 hsa-miR-31 hsa-miR-185 hsa-miR-146a

(M 10000089) (MI0000482) (M 10000477) 9.22E-05 hsa-miR-150 hsa-miR-152 hsa-miR-370

(M 10000479) (MI0000462) (MI0000778) 9.52E-05 hsa-miR-185 hsa-miR-193a-5p hsa-miR-370

(M 10000482) (MI0000487) (MI0000778) 9.76E-05 hsa-miR-193a-5p hsa-miR-29a hsa-miR-370

(M 10000487) (MI0000087) (MI0000778) 1.02E-04 hsa-miR-150 hsa-let-7d hsa-miR-146a

(M 10000479) (MI0000065) (M 10000477) 1.02E-04 hsa-miR-146a hsa-miR-214 hsa-miR-370

(M 10000477) (MI0000290) (MI0000778) 1.03E-04 hsa-miR-31 hsa-miR-146a hsa-miR-509-3-5p

(M 10000089) (MI0000477) (MI0005717) 1.09E-04 hsa-let-7d hsa-miR-370 hsa-miR-509-3-5p

(M 10000065) (MI0000778) (M 10005717) 1.14E-04 hsa-miR-193a-5p hsa-miR-29a hsa-miR-146a

(M 10000487) (MI0000087) (M 10000477) 1.19E-04 hsa-miR-150 hsa-miR-193a-5p hsa-miR-146a

(M 10000479) (MI0000487) (M 10000477) 1.22E-04 hsa-miR-31 hsa-miR-150 hsa-miR-193a-5p

(M 10000089) (MI0000479) (M 10000487) 1.26E-04 hsa-miR-31 hsa-miR-508-3p hsa-miR-146a

(M 10000089) (AMI0003195) (M 10000477) 1.26E-04 hsa-miR-31 hsa-miR-29a hsa-miR-146a

(M 10000089) (MI0000087) (M 10000477) 1.33E-04 hsa-miR-193a-5p hsa-let-7d hsa-miR-146a

(M 10000487) (MI0000065) (M 10000477) 1.37E-04 Log rank p- microRNA microRNA microRNA value hsa-miR-185 hsa-miR-193a-5p hsa-miR-431

(M 10000482) (MI0000487) (MI0001721) 1.43E-04 hsa-miR-708 hsa-miR-146a hsa-miR-370

(M 10005543) (MI0000477) (MI0000778) 1.51E-04 hsa-miR-31 hsa-miR-146a hsa-miR-214

(M 10000089) (MI0000477) (MI0000290) 1.54E-04 hsa-miR-185 hsa-miR-193a-5p hsa-miR-146a

(M 10000482) (MI0000487) (M 10000477) 1.59E-04 hsa-miR-21 hsa-miR-146a

hsa-miR-124 (MI0000077) (M 10000477) 1.68E-04 hsa-miR-21 hsa-miR-193a-5p hsa-miR-370

(M 10000077) (MI0000487) (MI0000778) 1.71E-04 hsa-let-7d hsa-miR-146a hsa-miR-152

(M 10000065) (MI0000477) (M 10000462) 1.71E-04 hsa-miR-193a-5p hsa-miR-146a

(M 10000487) (MI0000477) hsa-miR-886-3p 1.72E-04 hsa-miR-29a hsa-miR-370 hsa-miR-431

(M 10000087) (MI0000778) (MI0001721) 1.79E-04 hsa-miR-146a hsa-miR-214 hsa-miR-509-3-5p

(M 10000477) (MI0000290) (MI0005717) 1.80E-04 hsa-miR-193a-5p hsa-miR-146a

hsa-miR-124 (MI0000487) (M 10000477) 1.86E-04 hsa-miR-508-3p hsa-miR-21 hsa-miR-146a

(AMI0003195) (MI0000077) (M 10000477) 1.88E-04

Table 12: Combinations of 4 microRNAs (and their corresponding MiRBase Accession No as of filing date of patent applicaiton).

MiRBase Accession Number. Log rank p- microRNA microRNA microRNA microRNA value

hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

hsa-miR-124 (MI0000477) (MI0000778) (M 10003195) 2.24E-08 hsa-miR-508-3p hsa-miR-146a hsa-miR-370 hsa-miR-431

(AMI0003195) (MI0000477) (MI0000778) (M 10001721 ) 2.53E-08 hsa-miR-508-3p hsa-miR-708 hsa-miR-146a hsa-miR-370

(AMI0003195) (MI0005543) (M 10000477) (M 10000778) 2.83E-08 hsa-miR-21 hsa-miR-146a hsa-miR-370 hsa-miR-509-3-5p

(M 10000077) (MI0000477) (MI0000778) (M 10005717) 3.33E-08 hsa-miR-185 hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10000482) (MI0000477) (MI0000778) (M I0003195) 3.97E-08 hsa-miR-146a hsa-miR-370 hsa-miR-431 hsa-miR-508-5p

(M 10000477) (MI0000778) (MI0001721 ) (M 10003195) 3.97E-08 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10000487) (MI0000477) (MI0000778) (M I0003195) 5.59E-08 hsa-miR-29a hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10000087) (MI0000477) (MI0000778) (M I0003195) 5.76E-08 hsa-let-7d hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10000065) (MI0000477) (MI0000778) (M I0003195) 7.63E-08 hsa-miR-31 hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10000089) (MI0000477) (MI0000778) (MI0003195) ¹ 7.68E-08 hsa-miR-508-3p hsa-miR-21 hsa-miR-146a hsa-miR-370

(AMI0003195) (MI0000077) (M 10000477) (M 10000778) 8.05E-08 hsa-miR-150 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370

(M 10000479) (M 10000487) (M 10000477) (M 10000778) 8.33E-08 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370 hsa-miR-431

(M 10000487) (MI0000477) (MI0000778) (M 10001721 ) 9.09E-08 hsa-miR-29a hsa-miR-146a hsa-miR-370 hsa-miR-509-3-5p

(M 10000087) (MI0000477) (MI0000778) (M 10005717) 1.04E-07 hsa-miR-150 hsa-miR-508-3p hsa-miR-146a hsa-miR-508-5p

(M 10000479) (AMI0003195) (M 10000477) (M I0003195) 1.44E-07 hsa-miR-150 hsa-miR-193a-5p hsa-miR-370

(M 10000479) (M 10000487) (MI0000778) hsa-miR-886-3p 1.47E-07 hsa-miR-21 hsa-miR-29a hsa-miR-146a hsa-miR-509-3-5p

(M 10000077) (MI0000087) (M 10000477) (M I0005717) 1.48E-07 hsa-miR-150 hsa-miR-193a-5p hsa-miR-146a hsa-miR-509-3-5p

(M 10000479) (M 10000487) (M 10000477) (M I0005717) 1.71E-07 hsa-miR-31 hsa-miR-146a hsa-miR-370

(M 10000089) (MI0000477) (MI0000778) hsa-miR-886-3p 1.95E-07 hsa-miR-193a-5p hsa-miR-146a hsa-miR-152 hsa-miR-370

(M 10000487) (MI0000477) (M 10000462) (M 10000778) 2.00E-07 hsa-miR-150 hsa-let-7d hsa-miR-146a hsa-miR-370

(M 10000479) (MI0000065) (M 10000477) (M 10000778) 2.13E-07 hsa-miR-31 hsa-miR-150 hsa-miR-146a hsa-miR-370

(M 10000089) (MI0000479) (M 10000477) (M 10000778) 2.22E-07 hsa-miR-31 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370

(M 10000089) (M 10000487) (M 10000477) (M 10000778) 2.38E-07 hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10000477) (MI0000778) (MI0003195) hsa-miR-886-3p 2.40E-07 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370 hsa-miR-886-3p 2.42E-07 Log rank p- microRNA microRNA microRNA microRNA value

(M 10000487) (MI0000477) (MI0000778)

hsa-miR-150 hsa-miR-193a-5p hsa-miR-146a hsa-miR-152

(M 10000479) (M 10000487) (M 10000477) (M 10000462) 2.62E-07 hsa-miR-508-3p hsa-miR-29a hsa-miR-146a hsa-miR-370

(AMI0003195) (MI0000087) (M 10000477) (M 10000778) 2.64E-07 hsa-miR-508-3p hsa-miR-21 hsa-miR-29a hsa-miR-146a

(AMI0003195) (MI0000077) (M 10000087) (M 10000477) 2.71E-07 hsa-miR-193a-5p hsa-miR-146a hsa-miR-152 hsa-miR-214

(M 10000487) (MI0000477) (M 10000462) (M 10000290) 2.73E-07 hsa-miR-708 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370

(M 10005543) (M 10000487) (M 10000477) (M 10000778) 3.02E-07 hsa-miR-150 hsa-miR-146a hsa-miR-214 hsa-miR-509-3-5p

(M 10000479) (MI0000477) (MI0000290) (M I0005717) 4.25E-07 hsa-miR-150 hsa-miR-193a-5p hsa-miR-370 hsa-miR-431

(M 10000479) (M 10000487) (MI0000778) (M 10001721 ) 4.38E-07 hsa-miR-508-3p hsa-miR-146a hsa-miR-152

(AMI0003195) (MI0000477) (M 10000462) hsa-miR-886-3p 4.76E-07 hsa-let-7d hsa-miR-146a hsa-miR-152 hsa-miR-509-3-5p

(M 10000065) (MI0000477) (M 10000462) (M I0005717) 4.78E-07 hsa-miR-150 hsa-miR-193a-5p hsa-miR-146a hsa-miR-214

(M 10000479) (M 10000487) (M 10000477) (M 10000290) 5.52E-07 hsa-miR-150 hsa-miR-193a-5p hsa-miR-152 hsa-miR-370

(M 10000479) (M 10000487) (M 10000462) (M 10000778) 5.63E-07 hsa-imiR-21 hsa-miR-193a-5p hsa-miR-146a hsa-miR-508-5p

(M 10000077) (M 10000487) (M 10000477) (M I0003195) 5.90E-07 hsa-miR-146a hsa-miR-152 hsa-miR-370 hsa-miR-508-5p

(M 10000477) (M 10000462) (MI0000778) (M I0003195) 6.18E-07 hsa-imiR-21 hsa-miR-193a-5p hsa-miR-146a hsa-miR-509-3-5p

(M 10000077) (M 10000487) (M 10000477) (M I0005717) 6.44E-07 hsa-miR-150 hsa-miR-146a hsa-miR-508-5p hsa-miR-509-3-5p

(M 10000479) (MI0000477) (MI0003195) (M 10005717) 6.63E-07 hsa-miR-508-3p hsa-miR-146a hsa-miR-152 hsa-miR-509-3-5p

(AMI0003195) (MI0000477) (M 10000462) (M I0005717) 6.64E-07 hsa-miR-185 hsa-miR-193a-5p hsa-miR-146a hsa-miR-214

(M 10000482) (M 10000487) (M 10000477) (M 10000290) 6.90E-07 hsa-miR-508-3p hsa-miR-21 hsa-miR-146a hsa-miR-152

(AMI0003195) (MI0000077) (M 10000477) (M 10000462) 7.30E-07 hsa-miR-508-3p hsa-miR-21 hsa-miR-193a-5p hsa-miR-146a

(AMI0003195) (MI0000077) (M 10000487) (M 10000477) 7.36E-07 hsa-miR-508-3p hsa-let-7d hsa-miR-146a hsa-miR-152

(AMI0003195) (MI0000065) (M 10000477) (M 10000462) 8.01E-07 hsa-miR-29a hsa-let-7d hsa-miR-146a hsa-miR-509-3-5p

(M 10000087) (MI0000065) (M 10000477) (M I0005717) 8.16E-07 hsa-imiR-31 hsa-miR-508-3p hsa-miR-146a hsa-miR-152

(M 10000089) (AMI0003195) (M 10000477) (M 10000462) 8.25E-07 hsa-miR-21 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370

(M 10000077) (M 10000487) (M 10000477) (M 10000778) 8.29E-07 hsa-miR-508-3p hsa-miR-185 hsa-miR-146a hsa-miR-152

(AMI0003195) (MI0000482) (M 10000477) (M 10000462) 8.33E-07 hsa-miR-508-3p hsa-miR-146a hsa-miR-431 hsa-miR-508-5p 8.93E-07 Log rank p- microRNA microRNA microRNA microRNA value

(AMI0003195) (MI0000477) (MI0001721 ) (M I0003195)

hsa-miR-193a-5p hsa-miR-146a hsa-miR-370 hsa-miR-509-3-5p

(M 10000487) (MI0000477) (MI0000778) (M 10005717) 9.05E-07 hsa-miR-150 hsa-miR-508-3p hsa-miR-193a-5p hsa-miR-146a

(M 10000479) (AMI0003195) (M 10000487) (M 10000477) 9.40E-07 hsa-miR-150 hsa-miR-146a hsa-miR-152 hsa-miR-509-3-5p

(M 10000479) (MI0000477) (M 10000462) (M 10005717) 9.48E-07 hsa-miR-146a hsa-miR-214 hsa-miR-370 hsa-miR-508-5p

(M 10000477) (MI0000290) (MI0000778) (M I0003195) 9.68E-07 hsa-miR-150 hsa-miR-193a-5p hsa-miR-214 hsa-miR-370

(M 10000479) (M 10000487) (MI0000290) (M 10000778) 1.01E-06 hsa-miR-146a hsa-miR-431 hsa-miR-509-3-5p

(M 10000477) (M 10001721 ) (M 10005717) hsa-miR-886-3p 1.02E-06 hsa-miR-708 hsa-miR-146a hsa-miR-370 hsa-miR-508-5p

(M 10005543) (MI0000477) (MI0000778) (M I0003195) 1.04E-06 hsa-miR-21 hsa-miR-146a hsa-miR-508-5p hsa-miR-509-3-5p

(M 10000077) (MI0000477) (MI0003195) (M I0005717) 1.05E-06 hsa-miR-150 hsa-miR-185 hsa-miR-193a-5p hsa-miR-370

(M 10000479) (MI0000482) (M 10000487) (M 10000778) 1.06E-06 hsa-miR-146a hsa-miR-152 hsa-miR-431 hsa-miR-509-3-5p

(M 10000477) (M 10000462) (MI0001721 ) (M I0005717) 1.08E-06 hsa-miR-150 hsa-miR-193a-5p hsa-miR-152 hsa-miR-431

(M 10000479) (M 10000487) (M 10000462) (M 10001721 ) 1.09E-06 hsa-miR-193a-5p hsa-miR-146a hsa-miR-152 hsa-miR-509-3-5p

(M 10000487) (MI0000477) (M 10000462) (M I0005717) 1.10E-06 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370 hsa-miR-124 (M 10000487) (M 10000477) (M 10000778) 1.13E-06 hsa-let-7d hsa-miR-146a hsa-miR-431 hsa-miR-509-3-5p

(M 10000065) (MI0000477) (MI0001721 ) (M I0005717) 1.21E-06 hsa-miR-508-3p hsa-miR-185 hsa-miR-146a hsa-miR-508-5p

(AMI0003195) (MI0000482) (M 10000477) (M I0003195) 1.23E-06 hsa-miR-31 hsa-miR-146a hsa-miR-152 hsa-miR-509-3-5p

(M 10000089) (MI0000477) (M 10000462) (M I0005717) 1.27E-06 hsa-miR-508-3p hsa-miR-146a hsa-miR-152 hsa-miR-508-5p

(AMI0003195) (MI0000477) (M 10000462) (M I0003195) 1.28E-06 hsa-miR-150 hsa-miR-193a-5p hsa-miR-29a hsa-miR-370

(M 10000479) (M 10000487) (M 10000087) (M 10000778) 1.32E-06 hsa-miR-29a hsa-miR-146a hsa-miR-152 hsa-miR-431

(M 10000087) (MI0000477) (M 10000462) (MI0001721 ) 1.36E-06 hsa-miR-193a-5p hsa-let-7d hsa-miR-146a hsa-miR-509-3-5p

(M 10000487) (MI0000065) (M 10000477) (M I0005717) 1.46E-06 hsa-miR-31 hsa-miR-193a-5p hsa-miR-146a hsa-miR-509-3-5p

(M 10000089) (M 10000487) (M 10000477) (M I0005717) 1.48E-06 hsa-miR-193a-5p hsa-miR-146a hsa-miR-509-3-5p

(M 10000487) (MI0000477) (MI0005717) hsa-miR-886-3p 1.50E-06 hsa-miR-31 hsa-miR-150 hsa-miR-193a-5p hsa-miR-370

(M 10000089) (MI0000479) (M 10000487) (M 10000778) 1.54E-06 hsa-miR-150 hsa-miR-152 hsa-miR-370 hsa-miR-509-3-5p

(M 10000479) (M 10000462) (MI0000778) (M I0005717) 1.55E-06 hsa-miR-124 hsa-miR-150 hsa-miR-146a hsa-miR-370 1.55E-06 Log rank p- microRNA microRNA microRNA microRNA value

(Ml 0000479) (M 10000477) (M 10000778)

hsa-mi -185 hsa-miR-146a hsa-miR-152 hsa-miR-509-3-5p

(M 10000482) (Ml 0000477) (M 10000462) (M I0005717) 1.56E-06

hsa-miR-150 hsa-miR-508-3p hsa-let-7d hsa-miR-146a

(M 10000479) (AMI0003195) (M 10000065) (M 10000477) 1.62E-06

hsa-miR-31 hsa-miR-146a hsa-miR-431 hsa-miR-508-5p

(M 10000089) (MI0000477) (MI0001721 ) (M I0003195) 1.64E-06

hsa-miR-21 hsa-let-7d hsa-miR-146a hsa-miR-370

(M 10000077) (MI0000065) (M 10000477) (M 10000778) 1.65E-06

hsa-miR-185 hsa-miR-193a-5p hsa-miR-146a hsa-miR-370

(M 10000482) (M 10000487) (M 10000477) (M 10000778) 1.68E-06

OTHER EMBODIMENTS

While the invention has been described in conjunction with the detailed description thereof, the foregoing description is intended to illustrate and not limit the scope of the invention, which is defined by the scope of the appended claims. Other aspects, advantages, and modifications are within the scope of the following claims.

The patent and scientific literature referred to herein establishes the knowledge that is available to those with skill in the art. All United States patents and published or unpublished United States patent applications cited herein are incorporated by reference. All published foreign patents and patent applications cited herein are hereby incorporated by reference. Genbank and NCBI submissions indicated by accession number cited herein are hereby incorporated by reference. All other published references, documents, manuscripts and scientific literature cited herein are hereby incorporated by reference.

While this invention has been particularly shown and described with references to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention encompassed by the appended claims.

Claims

What is claimed is:

1. A composition for predicting presence of secondary site metastases or a predisposition thereto in a subject diagnosed as having a primary tumor comprising a detection reagent specific for at least one microRNA sequence selected from the group consisting of hsa-miR- 146a, hsa-miR-150, hsa-miR- 193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa-miR-29a, hsa-miR-508-5p, hsa-miR- 152, hsa-miR- 509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR- 886-3p.

2. The composition of claim 1, wherein said composition further comprises a second detection reagent specific for at least one mRNA selected from the group consisting of INTS4, NARS2, SNORD31, INTS2, TRIP10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1, ITGA11, IGHM, COL8A1, NNMT, COL1A1, BGN, INHBA, and COL11A1.

3. The composition of claim 1, wherein said composition further comprises a second detection reagent specific for at least one mRNA selected from the group consisting of COPZ2, NUCB1, LPL, CCDC49, GFPT2, LOX, NNMT, RGSl, ASNAl, FXYD5, SERPINEl, KIF26B, SIOOAIO, ALDH1A3, CALB2, and PLAUR.

4. The composition of claim 1, wherein said miRNA is selected from the group consisting of hsa-let-7d, hsa-miR-146a, hsa-miR-29a, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-708, hsa-miR-152, hsa- miR-214, and hsa-miR-150.

5. The composition of claim 1, wherein said miRNA is selected from the group consisting of hsa-let-7d, hsa-miR-146a, hsa-miR-29a, hsa-miR-193a-5p, hsa-miR-31, and hsa-miR-150.

6. The composition of claim 1, wherein said miRNA is selected from the group consisting of hsa-miR- 146a, hsa-miR- 193a-5p, hsa-miR-31, and hss-miR-150.

7. The composition of claim 1, wherein said microRNA comprises a combination of two microRNAS selected from the combinations listed in Table 10.

8. The composition of claim 1, wherein said microRNA comprises a combination of three microRNAS selected from the combinations listed in Table 11.

9. The composition of claim 1, wherein said microRNA comprises a combination of four microRNAS selected from the combinations listed in Table 12.

10. The composition of claim 2, wherein said mRNA is selected from the group consisting of mRNAs listed in Table 1 A or IB.

11. The method of claim 1 , wherein said detection reagent comprises a nucleic acid comprising a sequence complementary to at least 10 nucleotides of said microRNA.

12. The method of claim 1, wherein said detection reagent comprises a nucleic acid comprising a sequence complementary to 17-25 nucleotides of said microRNA.

13. The method of claim 1, wherein said detection reagent comprises a nucleic acid comprising a sequence complementary to at least 10 nucleotides of said mRNA.

14. The method of claim 1, wherein said detection reagent comprises a nucleic acid comprising a sequence complementary to 17-25 nucleotides of said mRNA.

15. A method for predicting the presence of secondary site metastases or a predisposition thereto in a subject diagnosed as having a primary tumor comprising:

providing a tissue sample obtained from said primary tumor;

detecting in said tissue sample at least one biomarker selected from the group consisting of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa- miR-29a, hsa-miR-508-5p, hsa-miR-152, hsa-miR-509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR-886-3p;

comparing the levels of said biomarker in said tissue sample to a control level of said biomarkers, wherein a higher level of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa- miR-370, hsa-let-7d, hsa-miR-29a, hsa-miR-152, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, or hsa-miR-886-3p compared to said control level of said biomarkers indicates that said subject is suffering from or at an increased risk of developing secondary site metastases and wherein a lower level of has- miR-508-5p, hsa-miR-509-3-5p, or hsa-miR-508-3p indicates that said subject is suffering from or at an increased risk of developing secondary site metastases.

I l l

16. A method for predicting the presence of secondary site metastases or a predisposition thereto in a subject diagnosed as having a primary tumor comprising:

providing a tissue sample obtained from said primary tumor;

detecting in said tissue sample at least two biomarker selected from the group consisting of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa- miR-29a, hsa-miR-508-5p, hsa-miR-152, hsa-miR-509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR-886-3p;

comparing the levels of said at least two biomarkers in said tissue sample to a control level of said biomarkers, wherein a higher level of said at least two biomarkers compared to said control level of said biomarkers indicates that said subject is suffering from or at an increased risk of developing secondary site metastases.

17. The method of claim 16, further comprising:

detecting in said tissue sample a messenger ribonucleic acid (mRNA) transcript encoding at least one of INTS4, NARS2, SNORD31, INTS2, TRIP 10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1 , ITGA11, IGHM, COL8A1, NNMT, COLlAl, BGN, INHBA, and COLl 1 Al or a mRNA transcript in Table 1 or Table 2;

comparing the levels of said mRNA transcript in said tissue sample to a control level of said mRNA, wherein a lower level of said INTS4, NARS2, SNORD31, INTS2 mRNA transcript compared to said control level of said mRNA indicates that said subject is suffering from or at an increased risk of developing secondary site metastases and wherein a higher level of said TRIP 10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDC1, ITGA11, IGHM, COL8A1, NNMT, COLlAl, BGN, INHBA, and COLl 1A1 level indicates that said subject is suffering from or at an increased risk of developing secondary site metastases.

18. The method of claim 16, further comprising:

detecting in said tissue sample a messenger ribonucleic acid (mRNA) transcript encoding at least one of COPZ2, NUCBl, LPL, CCDC49, GFPT2, LOX, NNMT, RGSl, ASNAl, FXYD5, SERPINEl, KIF26B, S100A10, ALDH1A3, CALB2, and PLAUR;

comparing the levels of said mRNA transcript in said tissue sample to a control level of said mRNA, wherein a higher level of said mRNA indicates that said subject is suffering from or at an increased risk of developing secondary site metastases.

19. The method of claim 16, wherein said primary tumor tissue sample comprises serous, clear cell, endometrioid, or mucinous carcinoma cells.

20. The method of claim 19, wherein said primary tumor is within the peritoneal cavity.

21. The method of claim 20, wherein said primary tumor is ovarian cancer, pancreatic cancer, gastric, kidney, colorectal cancer, hepatic cancer, bladder cancer, or breast cancer.

22. The method of claim 16, wherein said at least two biomarkers are detected via quantitative real time polymerase chain reaction (qPCR).

23. A method for predicting the presence of secondary site metastases or a predisposition thereto in a subject diagnosed as having a primary tumor comprising:

providing a tissue sample obtained from said primary tumor;

detecting in said tissue sample an mRNA transcript encoding programmed cell death protein 4 (PDCD4), serine/threonine-protein kinase 38-like (STK38L), reversion-inducing-cysteine-rich protein with kazal motifs (RECK), ELAV-like protein 1 (HuR), protein numb homolog (NUMB), 14-3-3 protein epsilon (YWHAE), AT-rich interactive domain-containing protein 1A (ARID 1 A) or a mRNA transcript in Table 1 or Table 2;

comparing the levels of said mRNA transcript in said tissue sample to a control level of said mRNA, wherein a lower level of said mRNA transcript compared to said control level of said mRNA indicates that said subject is suffering from or at an increased risk of developing secondary site metastases.

24. The method of claim 23, wherein said primary tumor tissue sample comprises serous, clear cell, endometrioid, or mucinous carcinoma cells

25. The method of claim 24, wherein said primary tumor is within the peritoneal cavity.

26. The method of claim 25, wherein said primary tumor is ovarian cancer, pancreatic cancer, colo-rectal cancer, liver cancer, or bladder cancer.

27. The method of claim 23, wherein said mRNA transcript is detected via quantitative real time reverse transcription polymerase chain reaction (qRT-PCR).

28. The method of claim 16, wherein said control level is obtained from a non-cancerous tissue sample.

29. A method for predicting the presence of secondary site metastases or a predisposition thereto in a subject diagnosed as comprising a primary tumor comprising:

providing a tissue sample obtained from said primary tumor;

detecting in said tissue sample at least one biomarker selected from the group consisting of hsa-miR- 193a-5p, hsa-miR-708, hsa-miR-31, hsa-miR-508-3p, hsa-miR-124, and hsa-miR-185;

comparing the levels of said at least one biomarker in said tissue sample to a control level of said biomarker, wherein a higher level of said at least one biomarker compared to said control level of said biomarker indicates that said subject is suffering from or at an increased risk of developing secondary site metastases.

30. The method of claim 29, wherein said primary tumor is a peritoneal cavity tumor selected from the group consisting of ovarian cancer, pancreatic cancer, colo-rectal cancer, liver cancer, and bladder cancer.

31. A method of treating a peritoneal cavity tumor in a subject in need thereof comprising administering to said subject a therapeutically effective amount of a composition comprising an inhibitor of a miRNA selected from the group consisting of miR-31, miR-21, miR-124, miR-150, miR-185, miR-708, miR-193-5p, miR-29a, let-7d, miR-886-3p, miR-270, miR-152, miR-146a, miR-431, and miR-214.

32. The method of claim 31, wherein said inhibitor is a peptide nucleic acid (PNA) or locked nucleic acid (LNA).

33. A method of treating a peritoneal cavity tumor in a subject in need thereof comprising administering to said subject a therapeutically effective amount of a composition comprising miR-508-3p, miR-509-3-5p, or miR-508-5p.

34. The method of claim 33, further comprising administering to said subject an miRNA mimic, wherein said miRNA mimic is a miR-509-3-5p mimic, a miR-508-3p mimic, or a miR-508-5p mimic.

35. A method of inhibiting cancer metastases in the peritoneal cavity of a subject comprising modulating the level of miR-31, miR-21, miR-124, miR-150, miR-185, miR-708, miR-193-5p, miR-29a, let-7d, miR- 886-3p, miR-270, miR-152, miR-146a, miR-431, miR-214, miR-508-3p, miR-509-3-5p, or miR-508-5p to change mRNA transcript levels, thereby inhibiting cancer metastases in said subject.

36. The method of claim 35, wherein said mRNA encodes PDCD4, STK38L, RECK, HuR, NUMB, YWHAE, or ARID 1 A.

37. A method of determining whether a tumor is resistant to a chemotherapeutic agent, comprising

detecting in a tissue sample from a subject at least one biomarker selected from the group consisting of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa-miR-370, hsa-let-7d, hsa- miR-29a, hsa-miR-508-5p, hsa-miR-152, hsa-miR-509-3-5p, hsa-miR-508-3p, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, hsa-miR-886-3p;

comparing the levels of said biomarker in said tissue sample to a control level of said biomarkers, wherein a higher level of hsa-miR-146a, hsa-miR-150, hsa-miR-193a-5p, hsa-miR-31, hsa-miR-21, hsa- miR-370, hsa-let-7d, hsa-miR-29a, hsa-miR-152, hsa-miR-708, hsa-miR-214, hsa-miR-431, hsa-miR-185, hsa-miR-124, or hsa-miR-886-3p compared to said control level of said biomarkers indicates that said subject comprises a drug resistant tumor and wherein a lower level of hsa-miR-508-5p, hsa-miR-509-3-5p, or hsa-miR-508-3p indicates that said subject comprises a drug-resistant tumor.

38. A method of determining whether a tumor is resistant to a chemotherapeutic agent, comprising

detecting in a tissue sample from a subject a messenger ribonucleic acid (mRNA) transcript encoding at least one of INTS4, NARS2, SNORD31, INTS2, TRIP 10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1, CRISPLD2, COMP, FNDCl, ITGAl l, IGHM, COL8A1, NNMT, COLlAl, BGN, INHBA, and COL11A1 or a mRNA transcript in Table 1 or Table 2;

comparing the levels of said mRNA transcript in said tissue sample to a control level of said mRNA, wherein a lower level of said INTS4, NARS2, SNORD31, INTS2 mRNA transcript compared to said control level of said mRNA indicates that said subject comprises a drug resistant tumor and wherein a higher level of said TRIP10, RASSF2, GADD45B, LTBP2, MYH9, MMP14, PLAU, OLFML2B, THY1,

CRISPLD2, COMP, FNDCl, ITGAl l, IGHM, COL8A1, NNMT, COLlAl, BGN, INHBA, and COLl lAl level indicates that said subject comprises a drug resistant tumor.

39. A method of determining whether a tumor is resistant to a chemotherapeutic agent, comprising

detecting in a tissue sample from a subject a messenger ribonucleic acid (mRNA) transcript encoding encoding at least one of COPZ2, NUCBl, LPL, CCDC49, GFPT2, LOX, NNMT, RGSl, ASNAl, FXYD5, SERPINE1, KIF26B, S100A10, ALDH1A3, CALB2, and PLAUR ;

comparing the levels of said mRNA transcript in said tissue sample to a control level of said mRNA, wherein a higher level of said mRNA indicates that said subject comprises a drug resistant tumor.

40. The method of claim 37, 38 or 39, wherein said chemotherapeutic agent comprises a platinum compound or a taxane compound.

41. The method of claim 37, 38 or 39, wherein said chemotherapeutic agent comprises both a platinum compound and a taxane compound.

42. The method of claim 15, wherein the levels of said biomarker comprises a prognostic indicator of overall survival or progression- free survival of said subject.

43. The method of claim 17 or 18, wherein the levels of said mRNA comprises a prognostic indicator of overall survival or progression-free survival of said subject.