TREATMENT COURSE OF ACTION

CROSS REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of U.S. provisional application Serial No.

62/443,246, filed January 6, 2017, which is incorporated herein by reference in its entirety.

FIELD OF THE INVENTION

The present invention relates to compositions, systems, and methods for determining a treatment course of action. In particular, the present invention relates to compositions, systems, and methods for utilizing gene expression profiles to determine drug sensitivity in colorectal cancer.

BACKGROUND OF THE INVENTION

Colorectal cancer (CRC) is a major, worldwide health burden with a high incidence and mortality rate, representing the fourth most common cause of cancer deaths (1). There is great heterogeneity in the clinical outcome, both with respect to prognosis and response to treatment (2, 3). Treatment decisions are primarily based on cancer stage and tumor location, and the repertoire of targeted treatments, as well as the extent of stratified treatment based on prognostic and predictive factors is limited (4, 5).

CRC is a heterogeneous disease also at the molecular level (3) and this is a clinical challenge, both with respect to precise interpretation of prognostic and predictive markers, and as a result of the potential to confer primary or secondary resistance to targeted treatment (6, 7). Traditionally, molecular subtyping of CRC has primarily been based on microsatellite instability (MSI) status, describing two main types of tumors that are associated with distinct DNA mutation patterns, tumor microenvironments, patient prognoses and treatment response (8-11). Recently, four consensus molecular subtypes (CMS) of primary CRC were described, based on intrinsic gene expression patterns (12). This classification encompasses both MSI status and the role of the tumor microenvironment, and has distinct clinicopathological and biological properties, with prognostic value independently of cancer stage. CMS1 includes most of the MSI+ tumors and tumors with BRAF mutations, and is a de-differentiated, immunogenic and inflammatory subtype. CMS2 is the largest, and a canonical CRC subtype with epithelial characteristics and WNT activation. CMS3 is also epithelial, but is specifically characterized by metabolic de-regulation. CMS4 is a mesenchymal subtype with activation of epithelial-to-mesenchymal transition (EMT), has a high level of stromal infiltration and is associated with a poor patient survival. CMS represents a consensus of several gene expression based classifications of CRC, put forth by the international CRC Subtyping Consortium (12); however, independent validation is pending. Furthermore, there is great potential clinical benefit in exploring CMS classification as a basis for molecularly stratified treatment of CRC, with identification of subtype-specific prognostic associations and drug vulnerabilities.

Immortalized cancer cell lines are valuable in vitro models for efficient, large-scale analyses of drug sensitivities. In CRC in particular, cell lines have repeatedly been shown to recapitulate the molecular properties of primary tumors (13-16). Although the tumor microenvironment has a strong potential influence on drug response, three-dimensional cultures (organoids) of CRC cells derived from patients' tumors have demonstrated good correlations between specific drug sensitivities and genomic alterations of the targeted pathways (17). Infiltration of immune and stromal cells in tumors also has an important influence on gene expression measured in bulk tumor tissue samples (18). Indeed, the tumor microenvironment is an important constituent of CMS classification (12), and in vitro explorations of the subtypes therefore require a cancer cell-specific classification approach.

Customized therapies for CRC are needed.

SUMMARY OF THE INVENTION

The present invention relates to compositions, systems, and methods for determining a treatment course of action. In particular, the present invention relates to compositions, systems, and methods for utilizing gene expression profiles for stratified patient treatment in colorectal cancer, based on subtype-specific drug sensitivities.

In experiments described herein, distinct molecular and clinicopathological characteristics of the four gene expression-based consensus molecular subtypes (CMS) of colorectal cancer (CRC) were validated in a consecutive, single-hospital series of 409 patients with stage I-IV CRC. For analyses of CRC cell lines, we developed a cancer cell- specific and stroma-independent CMS classifier, and showed in a panel of 150 cell lines that these CRC models recapitulate the biology of CMS classification. For in vitro exploration of subtype-specific drug vulnerabilities, high-throughput drug screening of 35 cell lines representing all the four CMS classes was performed. This revealed CMS-specific response patterns, including strong sensitivity to Hsp90-inhibitors in CMS1 and CMS4 relative to CMS2 and CMS3. This was validated by independent drug screens of additional cell lines.

Accordingly, in some embodiments, provided herein is a method for determining a treatment course of action in a subject diagnosed with colorectal cancer (CRC), comprising: a) identifying a consensus molecular subtype (CMS) of a colorectal cancer sample; and b) determining a treatment course of action based on the CMS classification. In some preferred embodiments, the methods comprise the step of stratifying a colorectal cancer sample from the subject by determining a consensus molecular subtype (CMS) classification selected from the group consisting of CMS1, CMS2, CMS3, and CMS4 for the sample. In some preferred embodiments, the stratification may performed by the clinician, hospital, or health maintenance organization treating the subject, or may be performed by an entity remote or distinct from the treating clinician, hospital, or health maintenance organization treating the subject. In some embodiments, the CMS is CMS1, CMS2, CMS3, or CMS4. The present disclosure is not limited to a particular treatment course of action. In some embodiments, the treatment course of action is administration of an HSP90 inhibitor, a topoisomerase II inhibitor, 2-methoxyestradiol, disulfiram, atorvastatin, PF-03758309, rigosertib, tipifamib, YM155, indibulin, cytarabine, 8-chloro-adenosine, serdemetan, TH588, RAF265, KU-60019, BGB324, auranofin or SGC-CBP30 when the CMS is CMS1 or CMS4. In some

embodiments, the HSP90 inhibitor is, for example, luminespib, ganetespib, onalespib, SGX- 301, radicicol or a derivative thereof. In some embodiments, the radicicol derivative is OS- 47720. In some embodiments, the topoisomerase II inhibitor is valrubicin, daunorubicin, doxorubicin or idarubicin. In preferred embodiments, the treatment course of action is administration of luminespib, ganetespib, 2-methoxyestradiol, atorvastatin or disulfiram. In some embodiments, the biological sample is, for example, a tissue sample, a biopsy sample, a blood sample or a stool sample.

In some embodiments, the colorectal cancer is stage I, II, III or IV. In some embodiments, identifying the CMS comprises determining the level of expression or presence of a mutation in one or more genes or proteins. In some preferred embodiments, the CMS is determined by application of a CMS classifier algorithm. Suitable algorithms are described for example, in Guinney et al, Nature Medicine, 21: 1350; Eide et al., CMScaller: an R package for consensus molecular subtyping of colorectal cancer pre-clinical models, Scientific Reports 7: 16618 (2017) and Sveen et al. Colorectal cancer Consensus Molecular Subtypes translated to preclinical models uncover potentially targetable cancer-cell dependencies, (2017) Clin. Can. Res. DOI: 10.1158/1078-0432.CCR-17-1234, the entire contents of each of which are incorporated by reference herein in their entirety). In some preferred embodiments, CMS subtypes are determined by analysis of a combination of genomic, mRNA, miRNA, and proteomic distinctions (e.g., expression, copy number, mutations, microsatellite stability, etc.) that distinguish the subtypes from each other.

In some embodiments, the identification comprises the use of colorectal cancer informative reagent selected from, for example, a nucleic acid probe or probes that hybridizes to a respective gene product of the one or more genes, nucleic acid primers for the amplification and detection or sequencing of a respective gene product of the one or more genes, and an antigen binding protein that binds to a respective gene product of the one or more genes. In some embodiments, the gene product is an RNA transcript from the gene and the colorectal informative reagent is a nucleic acid probe or probes that hybridizes to the respective gene product of the one or more genes or nucleic acid primers for the amplification and detection or sequencing of the respective gene product of the one or more genes. In some embodiments, the gene product is a protein and the colorectal cancer informative reagent is an antibody recognizing the respective product of the one or more genes. In some

embodiments, the method further comprises the step of administering the treatment course of action to the subject.

Additional embodiments provide a method for treating a subject diagnosed with colorectal cancer (CRC), comprising: a) identifying a consensus molecular subtype (CMS) of a colorectal cancer sample; b) determining a treatment course of action based on the CMS classification; and c) administering the treatment course of action to the subject.

Additional embodiments will be apparent to persons skilled in the relevant art based on the teachings contained herein. DESCRIPTION OF THE DRAWINGS

FIG. 1. Validation of molecular and clinicopathological characteristics of the CMS subtypes a) From a consecutive series of 409 patients with stage I-IV CRC, totally 323 (79%) tumors were confidently assigned to a CMS subtype (posterior probability larger than 0.5 from the random forest CMS classifier), while 46 tumors (11%) displayed mixed characteristics between two of the subtypes (posterior probability larger than 0.3 for both subtypes) and 40 tumors (10%) were indeterminate. Among the confidently classified tumors, known associations with the CMS subtypes (12) were validated for b) MSI status, BRAF mutations, KRAS mutations and TP53 mutations, c) patient gender, tumor localization, tumor differentiation grade and cancer stage, and d) patient survival, showing a poor outcome for patients with CMS4 tumors, e) Among patients with MSS tumors (n = 257), CMS1 and CMS4 were both associated with a poor survival compared with CMS2 and CMS3. In parts b-e, the color code is the same as indicated in part a. OS, overall survival; RFS, relapse-free survival; HR, hazard ratio from Cox's proportional hazards analyses; 95% CI, 95% confidence interval.

FIG. 2. CMS4 is associated with a poor patient survival Excluding from the in- house patient series those that had emergency surgery, incomplete resection, bowel perforation, or received pre-operative radiation therapy (n = 67), CMS4 tumors (n = 45) were consistently associated with a significantly poorer patient survival than CMS 1-3 (n = 210). HR, hazard ratio from Cox's proportional hazards analyses

FIG. 3. CMS subtyping of CRC cell lines a) Confident CMS classification was obtained for 131 (87%) of 150 unique CRC cell lines using the cancer cell-specific CMS classifier, with similar distribution among the subtypes as for the consecutive patient series. In the cell line classification, the molecular and biological characteristics of the CMS subtypes were also recapitulated, as shown b) for MSI status, BRAF mutations, KRAS mutations and TP53 mutations, as well as c) by gene set expression enrichment analyses.

FIG. 4. Differential drug sensitivities between individual CMS subtypes In each plot, cell lines of two CMS subtypes were compared, as indicated, with respect to sensitivity to each of 460 drugs included in a high-throughput drug screen. Each dot represents one drug, and selected drugs are colored according to molecular targets, as indicated. In comparison with CMS2, CMS1 cell lines were more sensitive to inhibitors of topoisomerases (orange), Hsp90-inhibitors (red; in particular luminespib, ganetespib and radicicol) and 2ME (green; combined angiogenesis- and tubulin-inhibitor). Also in comparison with CMS3, CMS 1 cell lines were more sensitive to Hsp90-inhibitors and 2ME. There was no differential drug sensitivity between CMS1 and CMS4 cell lines. CMS2 cell lines were, in comparison with CMS3, more sensitive to EGFR-inhibitors, while the opposite was found for mitotic inhibitors. In comparison with CMS2 and CMS3, CMS4 cell lines showed strong sensitivity to Hsp90-inhibitors, atorvastatin (blue; HMG-CoA reductase-inhibitor), 2ME and disulfiram (pink; inhibitor of acetaldehyde dehydrogenase). 2ME, 2-methoxyestradiol; DSS, drug sensitivity score

FIG. 5. Strong relative drug sensitivities in CMS1 and CMS4 a) High-throughput drug screening (n = 460 drugs) of CMS classified CRC cell lines (n = 29) showed high relative sensitivity to three Hsp90-inhibitors (red; luminespib, ganetespib and radicicol), 2ME (green; combined angiogenesis- and tubulin-inhibitor), atorvastatin (dark blue; HMG-CoA reductase-inhibitor), indibulin (pale blue; tubulin-inhibitor) and disulfiram (pink; inhibitor of acetaldehyde dehydrogenase) in CMS 1 and CMS4 compared with CMS2 and CMS3. b) A validation drug screen of five additional CMS1 and CMS4 cell lines confirmed strong sensitivity (red) to Hsp90-inhibitors, 2ME, atorvastatin and disulfiram in comparison both with two CMS3 cell lines included in the validation screen, and with the average sensitivity in CMS2 and CMS3 cell lines in the initial screen. 2ME, 2-methoxy estradiol; DSS, drug sensitivity score

FIG. 6. Validation of strong relative ganetespib-sensitivity in CMS1 and CMS4 cell lines in published data Strong relative sensitivity to Hsp90-inhibition in CMS1 and CMS4 cell lines compared with CMS2 and CMS3 cell lines was validated in two published datasets; a) using the Hsp90-inhibitor ganetespib in He, S. et al. Invest. New Drugs 32, 577- 586 (2014); and b) using the Hsp90-inhibitor CCT018159 in the Genomics of Drug

Sensitivity in Cancer Project, Yang, W. et al. Nucleic Acids Res. 41, D955-961 (2013). In both datasets, drug sensitivities were reported as ICso-values, representing the drug concentration that gives half of maximal inhibition. Accordingly, low values indicate high drug sensitivity. The ICso-values are plotted in inverse order on the vertical axes. In panel b, the ICio-values are plotted as the natural logarithm.

FIG. 7: CMS and drug-sensitivity associations in extended panel of CRC cell lines. /^-values are for Kruskal-Wallis rank sum tests. Higher scores indicate higher drug sensitivity. Afatinib is included as an example of a drug with larger effect within CMS2.

FIG. 8. CMS-selective activity of luminespib in PDX models. In CMS4 PDX models (n = 34) of a liver metastasis from a chemotherapy-naive CRC patient (top panel), combined administration of 5-FU and luminespib showed stronger anti -tumor activity than single agent treatment with 5-FU or luminespib, or in vehicle-treated controls. Tumor growth is plotted as the mean ± standard error of tumor volume fold changes of all mice per treatment arm at the indicated time points. No significant changes in Ki67 protein expression in post-treatment samples (relative to vehicle-treated controls) confirmed that the CMS4 model was chemoresistant, while increased expression of HSP70 after luminespib treatment showed a targeted effect of HSP90 inhibition (P-values were calculated by Welch's t-test; sample numbers vary due to availability of high-quality samples or data). CMS2 PDX models (n = 34) were highly chemosensitive, as shown by a strong anti-tumor activity of 5-FU monotherapy and reduced Ki67 expression in post-treatment samples, and there was no synergistic effect of combining 5-FU with luminespib (bottom panel). In contrast with the in vitro data, luminespib monotherapy had a moderately stronger anti-tumor activity in CMS2 (relative to vehicle-treated controls) than in CMS4, but this was not associated with changes in HSP70 expression in CMS2 post-treatment samples. DEFINITIONS

To facilitate an understanding of the present invention, a number of terms and phrases are defined below:

As used herein, the term "sensitivity" is defined as a statistical measure of performance of an assay (e.g., method, test), calculated by dividing the number of true positives by the sum of the true positives and the false negatives.

As used herein, the term "specificity" is defined as a statistical measure of performance of an assay (e.g., method, test), calculated by dividing the number of true negatives by the sum of true negatives and false positives.

As used herein, the term "informative" or "informativeness" refers to a quality of a marker or panel of markers, and specifically to the likelihood of finding a marker (or panel of markers) in a positive sample.

As used herein, the terms "colorectal cancer informative reagent" refers to a reagent or reagents that are informative for identification of cancer gene markers described herein. In some embodiments, reagents are primers, probes or antibodies for detection of gene expression products (e.g., RNA transcripts or proteins) associated with CRC types CMSl-4.

As used herein, the "consensus molecular subtype" or CMS (e.g., CMS1, CMS2, CMS3, and CMS4) refer to molecular subtypes of colorectal cancer (CRC) as defined in Guinney et al. (Nature Medicine, 21 : 1350; herein incorporated by reference in its entirety), and derivatives or adaptations of said subtypes comprising essentially the same molecular features, including the adapted CMS classifiers described herein (see, e.g., the CMS classifiers described in in Eide et al, CMScaller: an R package for consensus molecular subtyping of colorectal cancer pre-clinical models, Scientific Reports 7: 16618 (2017) and Sveen et al. Colorectal cancer Consensus Molecular Subtypes translated to preclinical models uncover potentially targetable cancer-cell dependencies, (2017) Clin. Can. Res. DOI:

10.1158/1078-0432.CCR-17-1234, the entire contents of each of which are incorporated by reference herein in their entirety). In some embodiments, the CMS is determined by application of the CMScaller algorithm, available on the world wide web at

github.com/Lothelab/CMScaller and which is described in Eide et al. In some embodiments, CMS subtypes are a combination of genomic, mRNA, miRNA, and proteomic distinctions (e.g., expression, copy number, mutations, microsatellite stability, etc.) that distinguish the subtypes from each other.

As used herein, the term "metastasis" is meant to refer to the process in which cancer cells originating in one organ or part of the body relocate to another part of the body and continue to replicate. Metastasized cells subsequently form tumors which may further metastasize. Metastasis thus refers to the spread of cancer from the part of the body where it originally occurs to other parts of the body. As used herein, the term "metastasized colorectal cancer cells" is meant to refer to colorectal cancer cells which have metastasized; colorectal cancer cells localized in a part of the body other than the colorectal.

As used herein, "an individual is suspected of being susceptible to metastasized colorectal cancer" is meant to refer to an individual who is at an above-average risk of developing metastasized colorectal cancer. Examples of individuals at a particular risk of developing colorectal cancer are those whose family medical history indicates above average incidence of colorectal cancer among family members and/or those who have already developed colorectal cancer and have been effectively treated who therefore face a risk of relapse and recurrence. Other factors which may contribute to an above-average risk of developing metastasized colorectal cancer which would thereby lead to the classification of an individual as being suspected of being susceptible to metastasized colorectal cancer may be based upon an individual's specific genetic, medical and/or behavioral background and characteristics.

The term "neoplasm" as used herein refers to any new and abnormal growth of tissue. Thus, a neoplasm can be a premalignant neoplasm or a malignant neoplasm. The term "neoplasm-specific marker" refers to any biological material that can be used to indicate the presence of a neoplasm. Examples of biological materials include, without limitation, nucleic acids, polypeptides, carbohydrates, fatty acids, cellular components (e.g., cell membranes and mitochondria), and whole cells. The term "colorectal neoplasm-specific marker" refers to any biological material that can be used to indicate the presence of a colorectal neoplasm (e.g., a premalignant colorectal neoplasm, a malignant colorectal neoplasm, a metastatic colorectal neoplasm). Examples of colorectal neoplasm-specific markers include, but are not limited to, the 13 gene signature described herein.

As used herein, the term "amplicon" refers to a nucleic acid generated using primer pairs. The amplicon is typically single-stranded DNA (e.g., the result of asymmetric amplification), however, it may be RNA or dsDNA. The term "amplifying" or "amplification" in the context of nucleic acids refers to the production of multiple copies of a polynucleotide, or a portion of the polynucleotide, typically starting from a small amount of the polynucleotide (e.g., a single polynucleotide molecule), where the amplification products or amplicons are generally detectable.

Amplification of polynucleotides encompasses a variety of chemical and enzymatic processes. The generation of multiple DNA copies from one or a few copies of a target or template DNA molecule during a polymerase chain reaction (PCR) or a ligase chain reaction (LCR; see, e.g., U.S. Patent No. 5,494,810; herein incorporated by reference in its entirety) are forms of amplification. Additional types of amplification include, but are not limited to, allele-specific PCR (see, e.g., U.S. Patent No. 5,639,611; herein incorporated by reference in its entirety), assembly PCR (see, e.g., U.S. Patent No. 5,965,408; herein incorporated by reference in its entirety), helicase-dependent amplification (see, e.g., U.S. Patent No.

7,662,594; herein incorporated by reference in its entirety), hot-start PCR (see, e.g., U.S. Patent Nos. 5,773,258 and 5,338,671; each herein incorporated by reference in their entireties), intersequence-specfic PCR, inverse PCR (see, e.g., Triglia, et al. (1988) Nucleic Acids Res., 16:8186; herein incorporated by reference in its entirety), ligation-mediated PCR (see, e.g., Guilfoyle, R. et al, Nucleic Acids Research, 25: 1854-1858 (1997); U.S. Patent No. 5,508,169; each of which are herein incorporated by reference in their entireties), methylation-specific PCR (see, e.g., Herman, et al, (1996) PNAS 93(13) 9821-9826; herein incorporated by reference in its entirety), miniprimer PCR, multiplex ligation-dependent probe amplification (see, e.g., Schouten, et al, (2002) Nucleic Acids Research 30(12): e57; herein incorporated by reference in its entirety), multiplex PCR (see, e.g., Chamberlain, et al, (1988) Nucleic Acids Research 16(23) 11141-11156; Ballabio, et al, (1990) Human Genetics 84(6) 571-573; Hayden, et al, (2008) BMC Genetics 9:80; each of which are herein incorporated by reference in their entireties), nested PCR, overlap-extension PCR (see, e.g., Higuchi, et al, (1988) Nucleic Acids Research 16(15) 7351-7367; herein incorporated by reference in its entirety), real time PCR (see, e.g., Higuchi, etl al., (1992) Biotechnology 10:413-417; Higuchi, et al, (1993) Biotechnology 11: 1026-1030; each of which are herein incorporated by reference in their entireties), reverse transcription PCR (see, e.g., Bustin, S.A. (2000) J. Molecular Endocrinology 25: 169-193; herein incorporated by reference in its entirety), solid phase PCR, thermal asymmetric interlaced PCR, and Touchdown PCR (see, e.g., Don, et al, Nucleic Acids Research (1991) 19(14) 4008; Roux, K. (1994) Biotechniques 16(5) 812-814; Hecker, et al, (1996) Biotechniques 20(3) 478-485; each of which are herein incorporated by reference in their entireties). Polynucleotide amplification also can be accomplished using digital PCR (see, e.g., Kalinina, et al, Nucleic Acids Research. 25; 1999- 2004, (1997); Vogelstein and Kinzler, Proc Natl Acad Sci USA. 96; 9236-41, (1999);

International Patent Publication No. WO05023091 A2; US Patent Application Publication No. 20070202525; each of which are incorporated herein by reference in their entireties).

As used herein, the terms "complementary" or "complementarity" are used in reference to polynucleotides (i.e., a sequence of nucleotides) related by the base-pairing rules. For example, the sequence "5'-A-G-T-3'," is complementary to the sequence "3'-T-C-A-5'." Complementarity may be "partial," in which only some of the nucleic acids' bases are matched according to the base pairing rules. Or, there may be "complete" or "total" complementarity between the nucleic acids. The degree of complementarity between nucleic acid strands has significant effects on the efficiency and strength of hybridization between nucleic acid strands. This is of particular importance in amplification reactions, as well as detection methods that depend upon binding between nucleic acids.

As used herein, the term "primer" refers to an oligonucleotide, whether occurring naturally as in a purified restriction digest or produced synthetically, that is capable of acting as a point of initiation of synthesis when placed under conditions in which synthesis of a primer extension product that is complementary to a nucleic acid strand is induced (e.g., in the presence of nucleotides and an inducing agent such as a biocatalyst (e.g. , a DNA polymerase or the like) and at a suitable temperature and pH). The primer is typically single stranded for maximum efficiency in amplification, but may alternatively be double stranded. If double stranded, the primer is generally first treated to separate its strands before being used to prepare extension products. In some embodiments, the primer is an

oligodeoxyribonucleotide. The primer is sufficiently long to prime the synthesis of extension products in the presence of the inducing agent. The exact lengths of the primers will depend on many factors, including temperature, source of primer and the use of the method. In certain embodiments, the primer is a capture primer.

As used herein, the term "nucleic acid molecule" refers to any nucleic acid containing molecule, including but not limited to, DNA or RNA. The term encompasses sequences that include any of the known base analogs of DNA and RNA including, but not limited to, 4 acetylcytosine, 8-hydroxy-N6-methyladenosine, aziridinylcytosine, pseudoisocytosine, 5- (carboxyhydroxyl-methyl) uracil, 5-fluorouracil, 5-bromouracil, 5- carboxymethylaminomethyl-2-thiouracil, 5-carboxymethyl-aminomethyluracil,

dihydrouracil, inosine, N6-isopentenyladenine, 1-methyladenine, 1 -methylpseudo-uracil, 1- methylguanine, 1-methylinosine, 2,2-dimethyl-guanine, 2-methyladenine, 2-methylguanine, 3-methyl-cytosine, 5-methylcytosine, N6-methyladenine, 7-methylguanine, 5- methylaminomethyluracil, 5-methoxy-amino-methyl-2-thiouracil, beta-D-mannosylqueosine, 5'-methoxycarbonylmethyluracil, 5-methoxyuracil, 2-methylthio-N- isopentenyladenine, uracil-5-oxyacetic acid methylester, uracil-5-oxyacetic acid, oxybutoxosine, pseudouracil, queosine, 2-thiocytosine, 5-methyl-2-thiouracil, 2-thiouracil, 4-thiouracil, 5-methyluracil, N- uracil-5-oxyacetic acid methylester, uracil-5-oxyacetic acid, pseudouracil, queosine, 2- thiocytosine, and 2,6-diaminopurine.

As used herein, the term "nucleobase" is synonymous with other terms in use in the art including "nucleotide," "deoxynucleotide," "nucleotide residue," "deoxynucleotide residue," "nucleotide triphosphate (NTP)," or deoxynucleotide triphosphate (dNTP).

An "oligonucleotide" refers to a nucleic acid that includes at least two nucleic acid monomer units (e.g., nucleotides), typically more than three monomer units, and more typically greater than ten monomer units. The exact size of an oligonucleotide generally depends on various factors, including the ultimate function or use of the oligonucleotide. To further illustrate, oligonucleotides are typically less than 200 residues long (e.g., between 15 and 100), however, as used herein, the term is also intended to encompass longer polynucleotide chains. Oligonucleotides are often referred to by their length. For example a 24 residue oligonucleotide is referred to as a "24-mer". Typically, the nucleoside monomers are linked by phosphodiester bonds or analogs thereof, including phosphorothioate, phosphorodithioate, phosphoroselenoate, phosphorodiselenoate, phosphoroanilothioate, phosphoranilidate, phosphoramidate, and the like, including associated counterions, e.g., H ⁺ , NH ₄ ⁺ , Na ⁺ , and the like, if such counterions are present. Further, oligonucleotides are typically single-stranded. Oligonucleotides are optionally prepared by any suitable method, including, but not limited to, isolation of an existing or natural sequence, DNA replication or amplification, reverse transcription, cloning and restriction digestion of appropriate sequences, or direct chemical synthesis by a method such as the phosphotriester method of Narang et al. (1979) Meth Enzymol. 68: 90-99; the phosphodiester method of Brown et al. (1979) Meth Enzymol. 68: 109-151; the diethylphosphoramidite method of Beaucage et al. (1981) Tetrahedron Lett. 22: 1859-1862; the triester method of Matteucci et al. (1981) J Am Chem Soc. 103:3185-3191; automated synthesis methods; or the solid support method of U.S. Pat. No. 4,458,066, entitled "PROCESS FOR PREPARING POLYNUCLEOTIDES," issued Jul. 3, 1984 to Caruthers et al, or other methods known to those skilled in the art. All of these references are incorporated by reference. A "sequence" of a biopolymer refers to the order and identity of monomer units (e.g., nucleotides, etc.) in the biopolymer. The sequence (e.g., base sequence) of a nucleic acid is typically read in the 5' to 3' direction.

As used herein, the term "subject" refers to any animal (e.g., a mammal), including, but not limited to, humans, non-human primates, rodents, and the like, which is to be the recipient of a particular treatment. Typically, the terms "subject" and "patient" are used interchangeably herein in reference to a human subject.

As used herein, the term "non-human animals" refers to all non-human animals including, but are not limited to, vertebrates such as rodents, non-human primates, ovines, bovines, ruminants, lagomorphs, porcines, caprines, equines, canines, felines, aves, etc.

The term "gene" refers to a nucleic acid (e.g. , DNA) sequence that comprises coding sequences necessary for the production of a polypeptide, RNA (e.g. , including but not limited to, mRNA, tRNA and rRNA) or precursor. The polypeptide, RNA, or precursor can be encoded by a full length coding sequence or by any portion of the coding sequence so long as the desired activity or functional properties (e.g. , enzymatic activity, ligand binding, signal transduction, etc.) of the full-length or fragment are retained. The term also encompasses the coding region of a structural gene and the including sequences located adjacent to the coding region on both the 5' and 3' ends for a distance of about 1 kb on either end such that the gene corresponds to the length of the full-length mRNA. The sequences that are located 5' of the coding region and which are present on the mRNA are referred to as 5' untranslated sequences. The sequences that are located 3' or downstream of the coding region and that are present on the mRNA are referred to as 3' untranslated sequences. The term "gene" encompasses both cDNA and genomic forms of a gene. A genomic form or clone of a gene contains the coding region interrupted with non-coding sequences termed "introns" or "intervening regions" or "intervening sequences". Introns are segments of a gene that are transcribed into nuclear RNA (hnRNA); introns may contain regulatory elements such as enhancers. Introns are removed or "spliced out" from the nuclear or primary transcript;

introns therefore are absent in the messenger RNA (mRNA) processed transcript. The mRNA functions during translation to specify the sequence or order of amino acids in a nascent polypeptide.

The term "locus" as used herein refers to a nucleic acid sequence on a chromosome or on a linkage map and includes the coding sequence as well as 5' and 3' sequences involved in regulation of the gene. DETAILED DESCRIPTION OF THE INVENTION

The present invention relates to compositions, systems, and methods for determining a treatment course of action. In particular, the present invention relates to compositions, systems, and methods for utilizing gene expression profiles to determine drug sensitivity in colorectal cancer.

For example, in some embodiments, the present invention provides a for determining a treatment course of action in a subject diagnosed with colorectal cancer (CRC), comprising: a) identifying a consensus molecular subtype classification (CMS) for a colorectal cancer sample; and b) determining a treatment course of action based on the CMS classification.

I. Identification of CMS classification

In some embodiments, the CMS classification is determined by assaying a sample for the level of expression and/or presence or absence of a mutation in one or more genes.

Any patient sample suspected of containing the genes may be tested according to methods of embodiments of the present invention. By way of non-limiting examples, the sample may be tissue (e.g. , a colorectal biopsy sample or other tissue sample), blood, stool or a fraction thereof (e.g. , plasma, serum, etc.).

In some embodiments, the patient sample is subjected to preliminary processing designed to isolate or enrich the sample for the pseudogenes or cells that contain the pseudogenes. A variety of techniques known to those of ordinary skill in the art may be used for this purpose, including but not limited to: centrifugation; immunocapture; cell lysis; and, nucleic acid target capture (See, e.g. , EP Pat. No. 1 409 727, herein incorporated by reference in its entirety).

While the present invention exemplifies several markers specific CMS classification, any marker that is correlated with the CMS classification or drug sensitivity may be used, alone or in combination with the markers described herein. A marker, as used herein, includes, for example, nucleic acid(s) whose production or mutation or lack of production is characteristic of a colorectal neoplasm or a prognosis or treatment thereof. Depending on the particular set of markers employed in a given analysis, the statistical analysis will vary. For example, where a particular combination of markers is highly specific for sensitivity of colorectal cancer to a particular treatment, the statistical significance of a positive result will be high. It may be, however, that such specificity is achieved at the cost of sensitivity (e.g., a negative result may occur even in the presence of colorectal cancer). By the same token, a different combination may be very sensitive (e.g., few false negatives), but has a lower specificity.

Particular combinations of markers may be used that show optimal function with different ethnic groups or sex, different geographic distributions, different stages of disease, different degrees of specificity or different degrees of sensitivity. Particular combinations may also be developed which are particularly sensitive to the effect of therapeutic regimens on disease progression. Subjects may be monitored after a therapy and/or course of action to determine the effectiveness of that specific therapy and/or course of action. Markers for other cancers, diseases, infections, and metabolic conditions are also contemplated for inclusion in a multiplex or panel format.

The methods are not limited to a particular type of mammal. In some embodiments, the mammal is a human. In some embodiments, the colorectal neoplasm is premalignant. In some embodiments, the colorectal neoplasm is malignant. In some embodiments, the colorectal neoplasm is colorectal cancer without regard to stage of the cancer (e.g., stage I, II, III, or IV). In some embodiments, the colorectal cancer is stage II.

A. DNA and RNA Detection - Colorectal Cancer Informative Reagents

Expression of the cancer marker genes of the present invention are detected using a variety of nucleic acid techniques known to those of ordinary skill in the art, including but not limited to: nucleic acid sequencing; nucleic acid hybridization; and nucleic acid amplification. These techniques utilize colorectal informative reagents such as nucleic acid probes and primers that hybridize to or can be used to amplify gene products of the cancer marker genes so that the level of expression of the respective cancer marker gene can be determined.

1. Sequencing

Illustrative non-limiting examples of nucleic acid sequencing techniques include, but are not limited to, chain terminator (Sanger) sequencing and dye terminator sequencing. Those of ordinary skill in the art will recognize that because RNA is less stable in the cell and more prone to nuclease attack experimentally RNA is usually reverse transcribed to DNA before sequencing.

Chain terminator sequencing uses sequence-specific termination of a DNA synthesis reaction using modified nucleotide substrates. Extension is initiated at a specific site on the template DNA by using a short radioactive, or other labeled, oligonucleotide primer complementary to the template at that region. The oligonucleotide primer is extended using a DNA polymerase, standard four deoxynucleotide bases, and a low concentration of one chain terminating nucleotide, most commonly a di-deoxynucleotide. This reaction is repeated in four separate tubes with each of the bases taking turns as the di-deoxynucleotide. Limited incorporation of the chain terminating nucleotide by the DNA polymerase results in a series of related DNA fragments that are terminated only at positions where that particular di- deoxynucleotide is used. For each reaction tube, the fragments are size-separated by electrophoresis in a slab polyacrylamide gel or a capillary tube filled with a viscous polymer. The sequence is determined by reading which lane produces a visualized mark from the labeled primer as you scan from the top of the gel to the bottom.

Dye terminator sequencing alternatively labels the terminators. Complete sequencing can be performed in a single reaction by labeling each of the di-deoxynucleotide chain- terminators with a separate fluorescent dye, which fluoresces at a different wavelength.

A variety of nucleic acid sequencing methods are contemplated for use in the methods of the present disclosure including, for example, chain terminator (Sanger) sequencing, dye terminator sequencing, and high-throughput sequencing methods. Many of these sequencing methods are well known in the art, See, e.g., Sanger et al, Proc. Natl. Acad. Sci. USA 74:5463-5467 (1997); Maxam et al, Proc. Natl. Acad. Sci. USA 74:560-564 (1977);

Drmanac, et al, Nat. Biotechnol. 16:54-58 (1998); Kato, Int. J. Clin. Exp. Med. 2: 193-202 (2009); Ronaghi et al, Anal. Biochem. 242:84-89 (1996); Margulies et al., Nature 437:376- 380 (2005); Ruparel et al, Proc. Natl. Acad. Sci. USA 102:5932-5937 (2005), and Harris et al., Science 320: 106-109 (2008); Levene et al, Science 299:682-686 (2003); Korlach et al., Proc. Natl. Acad. Sci. USA 105: 1176-1181 (2008); Branton et al, Nat. Biotechnol.

26(10): 1146-53 (2008); Eid et al, Science 323: 133-138 (2009); each of which is herein incorporated by reference in its entirety.

In some embodiments, deep sequencing is utilized to provide an analysis of the sequence and frequency of RNA molecules in the samples. Suitable deep sequencing techniques include, but are not limited to, next generation sequencing techniques such as single molecule real time sequencing (Pacific Biosciences), sequencing by synthesis

(Illumina, Inc.), 454 pyrosequencing (Roche Diagnostics, Inc.), SOLiD sequencing (Life Technologies, Inc.), and ion semiconductor sequencing (Life Technologies, Inc.).

2. Hybridization Illustrative non-limiting examples of nucleic acid hybridization techniques include, but are not limited to, in situ hybridization (ISH), microarray, nuclease protection assay, and Southern or Northern blot.

In situ hybridization (ISH) is a type of hybridization that uses a labeled

complementary DNA or RNA strand as a probe to localize a specific DNA or RNA sequence in a portion or section of tissue (in situ), or, if the tissue is small enough, the entire tissue (whole mount ISH). DNA ISH can be used to determine the structure of chromosomes. RNA ISH is used to measure and localize mRNAs and other transcripts (e.g., pseudogenes) within tissue sections or whole mounts. Sample cells and tissues are usually treated to fix the target transcripts in place and to increase access of the probe. The probe hybridizes to the target sequence at elevated temperature, and then the excess probe is washed away. The probe that was labeled with either radio-, fluorescent- or antigen-labeled bases is localized and quantitated in the tissue using either autoradiography, fluorescence microscopy or immunohistochemistry, respectively. ISH can also use two or more probes, labeled with radioactivity or the other non-radioactive labels, to simultaneously detect two or more transcripts.

In some embodiments, gene expression is detected using fluorescence in situ hybridization (FISH). In some embodiments, FISH assays utilize bacterial artificial chromosomes (BACs). These have been used extensively in the human genome sequencing project (see Nature 409: 953-958 (2001)) and clones containing specific BACs are available through distributors that can be located through many sources, e.g., NCBI. Each BAC clone from the human genome has been given a reference name that unambiguously identifies it. These names can be used to find a corresponding GenBank sequence and to order copies of the clone from a distributor.

The present invention further provides a method of performing a FISH assay on human colorectal cells, human colorectal tissue or on the fluid surrounding the human colorectal cells or tissue. Specific protocols are well known in the art and can be readily adapted for the present invention. Guidance regarding methodology may be obtained from many references including: In situ Hybridization: Medical Applications (eds. G. R. Coulton and J. de Belleroche), Kluwer Academic Publishers, Boston (1992); In situ Hybridization: In Neurobiology; Advances in Methodology (eds. J. H. Eberwine, K. L. Valentino, and J. D. Barchas), Oxford University Press Inc., England (1994); In situ Hybridization: A Practical Approach (ed. D. G. Wilkinson), Oxford University Press Inc., England (1992)); Kuo, et al, Am. J. Hum. Genet. 49: 112-119 (1991); Klinger, et al , Am. J. Hum. Genet. 57:55-65 (1992); and Ward, et al. , Am. J. Hum. Genet. 52:854-865 (1993)). There are also kits that are commercially available and that provide protocols for performing FISH assays (available from e.g. , Oncor, Inc., Gaithersburg, MD). Patents providing guidance on methodology include U.S. 5,225,326; 5,545,524; 6,121,489 and 6,573,043. All of these references are hereby incorporated by reference in their entirety and may be used along with similar references in the art and with the information provided in the Examples section herein to establish procedural steps convenient for a particular laboratory.

In some embodiments, the present invention utilizes nuclease protection assays. Nuclease protection assays are useful for identification of one or more RNA molecules of known sequence even at low total concentration. The extracted RNA is first mixed with antisense RNA or DNA probes that are complementary to the sequence or sequences of interest and the complementary strands are hybridized to form double-stranded RNA (or a DNA-RNA hybrid). The mixture is then exposed to ribonucleases that specifically cleave only single-stranded RNA but have no activity against double-stranded RNA. When the reaction runs to completion, susceptible RNA regions are degraded to very short oligomers or to individual nucleotides; the surviving RNA fragments are those that were complementary to the added antisense strand and thus contained the sequence of interest. Suitable nuclease protection assays, include, but are not limited to those described in US 5,770,370; EP 2290101A3; US 20080076121; US 20110104693; each of which is incorporated herein by reference in its entirety. In some embodiments, the present invention utilizes the quantitative nuclease protection assay provided by HTG Molecular Diagnostics, Inc. (Tuscon, AZ).

3. Microarrays

Different kinds of biological assays are called microarrays including, but not limited to: DNA microarrays (e.g. , cDNA microarrays and oligonucleotide microarrays); protein microarrays; tissue microarrays; transfection or cell microarrays; chemical compound microarrays; and, antibody microarrays. A DNA microarray, commonly known as gene chip, DNA chip, or biochip, is a collection of microscopic DNA spots attached to a solid surface (e.g. , glass, plastic or silicon chip) forming an array for the purpose of expression profiling or monitoring expression levels for thousands of genes simultaneously. The affixed DNA segments are known as probes, thousands of which can be used in a single DNA microarray. Microarrays can be used to identify disease genes or transcripts (e.g., genes described herein) by comparing gene expression in disease and normal cells. Microarrays can be fabricated using a variety of technologies, including but not limiting: printing with fine-pointed pins onto glass slides; photolithography using pre-made masks; photolithography using dynamic micromirror devices; ink-jet printing; or, electrochemistry on microelectrode arrays.

Southern and Northern blotting is used to detect specific DNA or RNA sequences, respectively. DNA or RNA extracted from a sample is fragmented, electrophoretically separated on a matrix gel, and transferred to a membrane filter. The filter bound DNA or RNA is subject to hybridization with a labeled probe complementary to the sequence of interest. Hybridized probe bound to the filter is detected. A variant of the procedure is the reverse Northern blot, in which the substrate nucleic acid that is affixed to the membrane is a collection of isolated DNA fragments and the probe is RNA extracted from a tissue and labeled.

4. Amplification

Nucleic acids (e.g., cancer marker genes) may be amplified prior to or simultaneous with detection. Illustrative non-limiting examples of nucleic acid amplification techniques include, but are not limited to, polymerase chain reaction (PCR), reverse transcription polymerase chain reaction (RT-PCR), transcription-mediated amplification (TMA), ligase chain reaction (LCR), strand displacement amplification (SDA), and nucleic acid sequence based amplification (NASBA). Those of ordinary skill in the art will recognize that certain amplification techniques (e.g., PCR) require that RNA be reversed transcribed to DNA prior to amplification (e.g., RT-PCR), whereas other amplification techniques directly amplify RNA (e. g. , TMA and NASBA).

The polymerase chain reaction (U.S. Pat. Nos. 4,683,195, 4,683,202, 4,800,159 and 4,965,188, each of which is herein incorporated by reference in its entirety), commonly referred to as PCR, uses multiple cycles of denaturation, annealing of primer pairs to opposite strands, and primer extension to exponentially increase copy numbers of a target nucleic acid sequence. In a variation called RT-PCR, reverse transcriptase (RT) is used to make a complementary DNA (cDNA) from mRNA, and the cDNA is then amplified by PCR to produce multiple copies of DNA. For other various permutations of PCR see, e.g., U.S. Pat. Nos. 4,683,195, 4,683,202 and 4,800,159; Mullis et al, Meth. Enzymol. 155: 335 (1987); and, Murakawa et al, DNA 7: 287 (1988), each of which is herein incorporated by reference in its entirety.

Transcription mediated amplification (U.S. Pat. Nos. 5,480,784 and 5,399,491, each of which is herein incorporated by reference in its entirety), commonly referred to as TMA, synthesizes multiple copies of a target nucleic acid sequence autocatalytically under conditions of substantially constant temperature, ionic strength, and pH in which multiple RNA copies of the target sequence autocatalytically generate additional copies. See, e.g., U.S. Pat. Nos. 5,399,491 and 5,824,518, each of which is herein incorporated by reference in its entirety. In a variation described in U.S. Publ. No. 20060046265 (herein incorporated by reference in its entirety), TMA optionally incorporates the use of blocking moieties, terminating moieties, and other modifying moieties to improve TMA process sensitivity and accuracy.

The ligase chain reaction (Weiss, R., Science 254: 1292 (1991), herein incorporated by reference in its entirety), commonly referred to as LCR, uses two sets of complementary DNA oligonucleotides that hybridize to adjacent regions of the target nucleic acid. The DNA oligonucleotides are covalently linked by a DNA ligase in repeated cycles of thermal denaturation, hybridization and ligation to produce a detectable double-stranded ligated oligonucleotide product.

Strand displacement amplification (Walker, G. et al, Proc. Natl. Acad. Sci. USA 89: 392-396 (1992); U.S. Pat. Nos. 5,270,184 and 5,455,166, each of which is herein

incorporated by reference in its entirety), commonly referred to as SDA, uses cycles of annealing pairs of primer sequences to opposite strands of a target sequence, primer extension in the presence of a dNTPaS to produce a duplex hemiphosphorothioated primer extension product, endonuclease-mediated nicking of a hemimodified restriction endonuclease recognition site, and polymerase-mediated primer extension from the 3' end of the nick to displace an existing strand and produce a strand for the next round of primer annealing, nicking and strand displacement, resulting in geometric amplification of product.

Thermophilic SDA (tSDA) uses thermophilic endonucleases and polymerases at higher temperatures in essentially the same method (EP Pat. No. 0 684 315).

Other amplification methods include, for example: nucleic acid sequence based amplification (U.S. Pat. No. 5,130,238, herein incorporated by reference in its entirety), commonly referred to as NASBA; one that uses an RNA replicase to amplify the probe molecule itself (Lizardi et al., BioTechnol. 6: 1197 (1988), herein incorporated by reference in its entirety), commonly referred to as replicase; a transcription based amplification method (Kwoh et al, Proc. Natl. Acad. Sci. USA 86:1173 (1989)); and, self-sustained sequence replication (Guatelli et al, Proc. Natl. Acad. Sci. USA 87: 1874 (1990), each of which is herein incorporated by reference in its entirety). For further discussion of known amplification methods see Persing, David H., "In Vitro Nucleic Acid Amplification

Techniques" in Diagnostic Medical Microbiology: Principles and Applications (Persing et al, Eds.), pp. 51-87 (American Society for Microbiology, Washington, DC (1993)). 5 Detection Methods

Non-amplified or amplified nucleic acids can be detected by any conventional means. For example, the cancer marker genes described herein can be detected by hybridization with a detectably labeled probe and measurement of the resulting hybrids. Illustrative non- limiting examples of detection methods are described below.

One illustrative detection method provides for quantitative evaluation of the amplification process in real-time. Evaluation of an amplification process in "real-time" involves determining the amount of amplicon in the reaction mixture either continuously or periodically during the amplification reaction, and using the determined values to calculate the amount of target sequence initially present in the sample. A variety of methods for determining the amount of initial target sequence present in a sample based on real-time amplification are well known in the art. These include methods disclosed in U.S. Pat. Nos. 6,303,305 and 6,541,205, each of which is herein incorporated by reference in its entirety. Another method for determining the quantity of target sequence initially present in a sample, but which is not based on a real-time amplification, is disclosed in U.S. Pat. No. 5,710,029, herein incorporated by reference in its entirety.

Amplification products may be detected in real-time through the use of various self- hybridizing probes, most of which have a stem-loop structure. Such self-hybridizing probes are labeled so that they emit differently detectable signals, depending on whether the probes are in a self-hybridized state or an altered state through hybridization to a target sequence. By way of non-limiting example, "molecular torches" are a type of self-hybridizing probe that includes distinct regions of self-complementarity (referred to as "the target binding domain" and "the target closing domain") which are connected by a joining region (e.g. , non- nucleotide linker) and which hybridize to each other under predetermined hybridization assay conditions. In a preferred embodiment, molecular torches contain single-stranded base regions in the target binding domain that are from 1 to about 20 bases in length and are accessible for hybridization to a target sequence present in an amplification reaction under strand displacement conditions. Under strand displacement conditions, hybridization of the two complementary regions, which may be fully or partially complementary, of the molecular torch is favored, except in the presence of the target sequence, which will bind to the single- stranded region present in the target binding domain and displace all or a portion of the target closing domain. The target binding domain and the target closing domain of a molecular torch include a detectable label or a pair of interacting labels (e.g. , luminescent/quencher) positioned so that a different signal is produced when the molecular torch is self-hybridized than when the molecular torch is hybridized to the target sequence, thereby permitting detection of probe:target duplexes in a test sample in the presence of unhybridized molecular torches. Molecular torches and a variety of types of interacting label pairs are disclosed in U.S. Pat. No. 6,534,274, herein incorporated by reference in its entirety.

In some embodiments, a TaqMan™ detection system is utilized to detect and quantify expression of the cancer marker genes. The TaqMan probe system relies on the 5 '-3' exonuclease activity of Taq polymerase to cleave a dual-labeled probe during hybridization to the complementary target sequence and fluorophore-based detection. As in other real-time PCR methods, the resulting fluorescence signal permits quantitative measurements of the accumulation of the product during the exponential stages of the PCR; however, the TaqMan probe significantly increases the specificity of the detection. TaqMan probes consist of a fluorophore covalently attached to the 5 '-end of the oligonucleotide probe and a quencher at the 3'-end. Several different fluorophores (e.g. 6-carboxyfluorescein, acronym: FAM, or tetrachlorofluorescein, acronym: TET) and quenchers (e.g. tetramethylrhodamine, acronym: TAMRA, or dihydrocyclopyrroloindole tripeptide minor groove binder, acronym: MGB) are available. The quencher molecule quenches the fluorescence emitted by the fluorophore when excited by the cycler's light source via FRET (Fluorescence Resonance Energy Transfer). As long as the fluorophore and the quencher are in proximity, quenching inhibits any

fluorescence signals. TaqMan probes are designed such that they anneal within a DNA region amplified by a specific set of primers. As the Taq polymerase extends the primer and synthesizes the nascent strand (again, on a single-strand template, but in the direction opposite to that shown in the diagram, i.e. from 3' to 5' of the complementary strand), the 5' to 3' exonuclease activity of the polymerase degrades the probe that has annealed to the template. Degradation of the probe releases the fluorophore from it and breaks the close proximity to the quencher, thus relieving the quenching effect and allowing fluorescence of the fluorophore. Hence, fluorescence detected in the real-time PCR thermal cycler is directly proportional to the fluorophore released and the amount of DNA template present in the PCR.

Another example of a detection probe having self-complementarity is a "molecular beacon." Molecular beacons include nucleic acid molecules having a target complementary sequence, an affinity pair (or nucleic acid arms) holding the probe in a closed conformation in the absence of a target sequence present in an amplification reaction, and a label pair that interacts when the probe is in a closed conformation. Hybridization of the target sequence and the target complementary sequence separates the members of the affinity pair, thereby shifting the probe to an open conformation. The shift to the open conformation is detectable due to reduced interaction of the label pair, which may be, for example, a fluorophore and a quencher (e.g., DABCYL and EDANS). Molecular beacons are disclosed in U.S. Pat. Nos. 5,925,517 and 6,150,097, herein incorporated by reference in its entirety.

Other self-hybridizing probes are well known to those of ordinary skill in the art. By way of non-limiting example, probe binding pairs having interacting labels, such as those disclosed in U.S. Pat. No. 5,928,862 (herein incorporated by reference in its entirety) might be adapted for use in the present invention. Probe systems used to detect single nucleotide polymorphisms (SNPs) might also be utilized in the present invention. Additional detection systems include "molecular switches," as disclosed in U.S. Publ. No. 20050042638, herein incorporated by reference in its entirety. Other probes, such as those comprising intercalating dyes and/or fluorochromes, are also useful for detection of amplification products in the present invention. See, e.g. , U.S. Pat. No. 5,814,447 (herein incorporated by reference in its entirety).

Another illustrative detection method, the Hybridization Protection Assay (HP A) involves hybridizing a chemiluminescent oligonucleotide probe (e.g. , an acridinium ester- labeled (AE) probe) to the target sequence, selectively hydrolyzing the chemiluminescent label present on unhybridized probe, and measuring the chemiluminescence produced from the remaining probe in a luminometer. See, e.g., U.S. Pat. No. 5,283,174 and Norman C. Nelson et al., Nonisotopic Probing, Blotting, and Sequencing, ch. 17 (Larry J. Kricka ed., 2d ed. 1995, each of which is herein incorporated by reference in its entirety).

B. Protein Detection - Colorectal Cancer Informative Reagents

The cancer marker genes described herein may be detected as proteins using a variety of protein techniques known to those of ordinary skill in the art, including but not limited to: protein sequencing; and, immunoassays.

1. Sequencing

Illustrative non-limiting examples of protein sequencing techniques include, but are not limited to, mass spectrometry and Edman degradation.

Mass spectrometry can, in principle, sequence any size protein but becomes computationally more difficult as size increases. A protein is digested by an endoprotease, and the resulting solution is passed through a high pressure liquid chromatography column. At the end of this column, the solution is sprayed out of a narrow nozzle charged to a high positive potential into the mass spectrometer. The charge on the droplets causes them to fragment until only single ions remain. The peptides are then fragmented and the mass- charge ratios of the fragments measured. The mass spectrum is analyzed by computer and often compared against a database of previously sequenced proteins in order to determine the sequences of the fragments. The process is then repeated with a different digestion enzyme, and the overlaps in sequences are used to construct a sequence for the protein.

In the Edman degradation reaction, the peptide to be sequenced is adsorbed onto a solid surface (e.g. , a glass fiber coated with polybrene). The Edman reagent,

phenylisothiocyanate (PTC), is added to the adsorbed peptide, together with a mildly basic buffer solution of 12% trimethylamine, and reacts with the amine group of the N-terminal amino acid. The terminal amino acid derivative can then be selectively detached by the addition of anhydrous acid. The derivative isomerizes to give a substituted

phenylthiohydantoin, which can be washed off and identified by chromatography, and the cycle can be repeated. The efficiency of each step is about 98%, which allows about 50 amino acids to be reliably determined.

2. Immunoassays

Illustrative non-limiting examples of immunoassays include, but are not limited to: immunoprecipitation; Western blot; ELISA; immunohistochemistry; immunocytochemistry; flow cytometry; and, immuno-PCR. Polyclonal or monoclonal antibodies detectably labeled using various techniques known to those of ordinary skill in the art (e.g. , colorimetric, fluorescent, chemiluminescent or radioactive) are suitable for use in the immunoassays.

Immunoprecipitation is the technique of precipitating an antigen out of solution using an antibody specific to that antigen. The process can be used to identify protein complexes present in cell extracts by targeting a protein believed to be in the complex. The complexes are brought out of solution by insoluble antibody -binding proteins isolated initially from bacteria, such as Protein A and Protein G. The antibodies can also be coupled to sepharose beads that can easily be isolated out of solution. After washing, the precipitate can be analyzed using mass spectrometry, Western blotting, or any number of other methods for identifying constituents in the complex.

A Western blot, or immunoblot, is a method to detect protein in a given sample of tissue homogenate or extract. It uses gel electrophoresis to separate denatured proteins by mass. The proteins are then transferred out of the gel and onto a membrane, typically polyvinyldiflroride or nitrocellulose, where they are probed using antibodies specific to the protein of interest. As a result, researchers can examine the amount of protein in a given sample and compare levels between several groups. An ELISA, short for Enzyme-Linked Immunosorbent Assay, is a biochemical technique to detect the presence of an antibody or an antigen in a sample. It utilizes a minimum of two antibodies, one of which is specific to the antigen and the other of which is coupled to an enzyme. The second antibody will cause a chromogenic or fluorogenic substrate to produce a signal. Variations of ELISA include sandwich ELISA, competitive ELISA, and ELISPOT. Because the ELISA can be performed to evaluate either the presence of antigen or the presence of antibody in a sample, it is a useful tool both for determining serum antibody concentrations and also for detecting the presence of antigen.

Immunohistochemistry and immunocytochemistry refer to the process of localizing proteins in a tissue section or cell, respectively, via the principle of antigens in tissue or cells binding to their respective antibodies. Visualization is enabled by tagging the antibody with color producing or fluorescent tags. Typical examples of color tags include, but are not limited to, horseradish peroxidase and alkaline phosphatase. Typical examples of fluorophore tags include, but are not limited to, fluorescein isothiocyanate (FITC) or phycoerythrin (PE).

Flow cytometry is a technique for counting, examining and sorting microscopic particles suspended in a stream of fluid. It allows simultaneous multiparametric analysis of the physical and/or chemical characteristics of single cells flowing through an

optical/electronic detection apparatus. A beam of light (e.g. , a laser) of a single frequency or color is directed onto a hydrodynamically focused stream of fluid. A number of detectors are aimed at the point where the stream passes through the light beam; one in line with the light beam (Forward Scatter or FSC) and several perpendicular to it (Side Scatter (SSC) and one or more fluorescent detectors). Each suspended particle passing through the beam scatters the light in some way, and fluorescent chemicals in the particle may be excited into emitting light at a lower frequency than the light source. The combination of scattered and fluorescent light is picked up by the detectors, and by analyzing fluctuations in brightness at each detector, one for each fluorescent emission peak, it is possible to deduce various facts about the physical and chemical structure of each individual particle. FSC correlates with the cell volume and SSC correlates with the density or inner complexity of the particle (e.g. , shape of the nucleus, the amount and type of cytoplasmic granules or the membrane roughness).

Immuno-polymerase chain reaction (IPCR) utilizes nucleic acid amplification techniques to increase signal generation in antibody-based immunoassays. Because no protein equivalence of PCR exists, that is, proteins cannot be replicated in the same manner that nucleic acid is replicated during PCR, the only way to increase detection sensitivity is by signal amplification. The target proteins are bound to antibodies which are directly or indirectly conjugated to oligonucleotides. Unbound antibodies are washed away and the remaining bound antibodies have their oligonucleotides amplified. Protein detection occurs via detection of amplified oligonucleotides using standard nucleic acid detection methods, including real-time methods.

II. Data Analysis

In some embodiments, a computer-based analysis program is used to translate the raw data generated by the detection assay (e.g. , the expression level a given marker or markers) into data of predictive value for a clinician. The clinician can access the predictive data using any suitable means. Thus, in some preferred embodiments, the present invention provides the further benefit that the clinician, who is not likely to be trained in genetics or molecular biology, need not understand the raw data. The data is presented directly to the clinician in its most useful form. The clinician is then able to immediately utilize the information in order to optimize the care of the subject.

The present invention contemplates any method capable of receiving, processing, and transmitting the information to and from laboratories conducting the assays, information provides, medical personal, and subjects. For example, in some embodiments of the present invention, a sample (e.g. , a biopsy or a serum or stool sample) is obtained from a subject and submitted to a profiling service (e.g., clinical lab at a medical facility, genomic profiling business, etc.), located in any part of the world (e.g. , in a country different than the country where the subject resides or where the information is ultimately used) to generate raw data. Where the sample comprises a tissue or other biological sample, the subject may visit a medical center to have the sample obtained and sent to the profiling center, or subjects may collect the sample themselves (e.g. , a urine sample) and directly send it to a profiling center. Where the sample comprises previously determined biological information, the information may be directly sent to the profiling service by the subject (e.g. , an information card containing the information may be scanned by a computer and the data transmitted to a computer of the profiling center using an electronic communication systems). Once received by the profiling service, the sample is processed and a profile is produced (i.e. , expression data), specific for the diagnostic or prognostic information desired for the subject.

The profile data is then prepared in a format suitable for interpretation by a treating clinician. For example, rather than providing raw expression data, the prepared format may represent a diagnosis or risk assessment for the subject, along with recommendations for particular treatment options. The data may be displayed to the clinician by any suitable method. For example, in some embodiments, the profiling service generates a report that can be printed for the clinician (e.g. , at the point of care) or displayed to the clinician on a computer monitor.

In some embodiments, the information is first analyzed at the point of care or at a regional facility. The raw data is then sent to a central processing facility for further analysis and/or to convert the raw data to information useful for a clinician or patient. The central processing facility provides the advantage of privacy (all data is stored in a central facility with uniform security protocols), speed, and uniformity of data analysis. The central processing facility can then control the fate of the data following treatment of the subject.

For example, using an electronic communication system, the central facility can provide data to the clinician, the subject, or researchers.

In some embodiments, the subject is able to directly access the data using the electronic communication system. The subject may chose further intervention or counseling based on the results. In some embodiments, the data is used for research use. For example, the data may be used to further optimize the inclusion or elimination of markers as useful indicators of a particular condition or stage of disease or as a companion diagnostic to determine a treatment course of action. III. Compositions & Kits

Compositions for use in the diagnostic methods described herein include, but are not limited to, kits comprising one or more colorectal cancer informative reagents as described above. In some embodiments, the kits comprise one or more colorectal cancer informative reagents for detecting altered gene expression in a sample from a subject having or suspected of having colorectal cancer of one or more two or more, five or more, 10 or more, 11 or more, 12 or more or 13. In some embodiments, the kits contain colorectal cancer informative reagents specific for a cancer gene marker, in addition to detection reagents and buffers. In preferred embodiments, the colorectal informative reagent is a probe(s) that specifically hybridizes to a respective gene product(s) of the one or more genes, a set(s) of primers that amplify a respective gene product(s) of the one or more genes, an antigen binding protein(s) that binds to a respective gene product(s) of the one or more genes, or a sequencing primer(s) that hybridizes to and allows sequencing of a respective gene product(s) of the one or more genes. The probe and antibody compositions of the present invention may also be provided in the form of an array. In preferred embodiments, the kits contain all of the components necessary to perform a detection assay, including all controls, directions for performing assays, and any necessary software for analysis and presentation of results.

In some embodiments, the kits include instructions for using the reagents contained in the kit for the detection and characterization of cancer in a sample from a subject. In some embodiments, the instructions further comprise the statement of intended use required by the U.S. Food and Drug Administration (FDA) in labeling in vitro diagnostic products. The FDA classifies in vitro diagnostics as medical devices and requires that they be approved through the 510(k) procedure. Information required in an application under 510(k) includes: 1) The in vitro diagnostic product name, including the trade or proprietary name, the common or usual name, and the classification name of the device; 2) The intended use of the product; 3) The establishment registration number, if applicable, of the owner or operator submitting the 510(k) submission; the class in which the in vitro diagnostic product was placed under section 513 of the FD&C Act, if known, its appropriate panel, or, if the owner or operator determines that the device has not been classified under such section, a statement of that determination and the basis for the determination that the in vitro diagnostic product is not so classified; 4) Proposed labels, labeling and advertisements sufficient to describe the in vitro diagnostic product, its intended use, and directions for use. Where applicable, photographs or engineering drawings should be supplied; 5) A statement indicating that the device is similar to and/or different from other in vitro diagnostic products of comparable type in commercial distribution in the U.S., accompanied by data to support the statement; 6) A 510(k) summary of the safety and effectiveness data upon which the substantial equivalence determination is based; or a statement that the 510(k) safety and effectiveness information supporting the FDA finding of substantial equivalence will be made available to any person within 30 days of a written request; 7) A statement that the submitter believes, to the best of their knowledge, that all data and information submitted in the premarket notification are truthful and accurate and that no material fact has been omitted; 8) Any additional information regarding the in vitro diagnostic product requested that is necessary for the FDA to make a substantial equivalency determination. Additional information is available at the Internet web page of the U.S. FDA. III. Methods of Use

As disclosed herein, the present invention provides colorectal cancer informative reagents and methods for determining a prognosis and/or treatment course of action for colorectal cancer in a subject. The colorectal cancer can be stage I, II, III, or IV colorectal cancer. In some embodiments, the treatment course of action is administration of an HSP90 inhibitor, a topoisomerase II inhibitor, 2-methoxy estradiol, disulfiram, atorvastatin, PF- 03758309, rigosertib, tipifarnib, YM155, indibulin, cytarabine, 8-chloro-adenosine, serdemetan, TH588, RAF265, KU-60019, BGB324, auranofin or SGC-CBP30 when the CMS is CMS1 or CMS4. In some embodiments, the HSP90 inhibitor is, for example, luminespib, ganetespib, onalespib, SGX-301, radicicol or a derivative thereof. In some embodiments, the radicicol derivative is OS-47720. In some embodiments, the topoisomerase II inhibitor is valrubicin, daunorubicin, doxorubicin, or idarubicin. In some embodiments, the treatment course of action is administration of an alcohol dehydrogenase inhibitor or an Axl- inhibitor when the CMS is CMS4. In some embodiments, the alcohol dehydrogenase inhibitor is disulfiram and the Axl-inhibitor is BGB324.

In preferred embodiments, the treatment course of action is administration of luminespib, ganetespib, 2-methoxyestradiol, atorvastatin or disulfiram.

In some embodiments, treatments described herein are administered with one or more conventional treatments for CRC or in combination with surgical or radiation therapies. In some embodiments, treatments described herein are administered together with 5'- fluorouracil (5-FU).

In some embodiments, the CMS classification is determined at one or more points during treatment (e.g., before, during, or after treatment with a particular agent).

EXPERIMENTAL

The following examples are provided in order to demonstrate and further illustrate certain preferred embodiments and aspects of the present invention and are not to be construed as limiting the scope thereof.

Example 1

MATERIALS AND METHODS

Patient material

A consecutive series of 409 patients treated surgically for stage I-IV CRC at Oslo

University Hospital, Oslo, Norway, between 2005 and 2013 was included (Table 1). The study was approved by the Regional Committee for Medical and Health Research Ethics, South Eastern Norway (REC number 1.2005.1629). All patients provided written informed consent, and the study was conducted in accordance with the Declaration of Helsinki. Details of DNA/RNA extraction, as well as MSI status and mutation analyses are included as Supplementary Text.

CRC cell lines

Totally 169 CRC cell lines were analyzed (Supplementary Table S2), including 38 cell lines in-house (details of growth conditions in Supplementary Text) and publicly available gene expression data from 136 cell lines (five overlapping with the in-house dataset; obtained from Gene Expression Omnibus [GEO] accession numbers GSE36133 (Barretina J, et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 2012;483:603-7), GSE57083 and GSE59857 (Medico E, et al. The molecular landscape of colorectal cancer cell lines unveils clinically actionable kinase targets. Nat Commun 2015;6:7002). The number of cell lines derived from unique patients was 148. Cell line identities were verified by fingerprinting according to the

AmpFLSTR Identifiler PCR Amplification Kit (Life Technologies by Thermo Fisher Scientific), and matched to the profiles reported by the American Type Culture Collection. Cell lines were regularly tested for mycoplasma contamination according to the My co Alert Mycoplasma Detection Assay (Lonza Walkersville Inc., Walkersville, MD, USA).

Gene expression analyses

The primary CRCs were analyzed for gene expression using Affymetrix GeneChip Human Exon 1.0 ST Arrays (HuEx; n = 201 CRCs) or Human Transcriptome 2.0 Arrays (HTA; n = 208 CRCs) according to the manufacturer's instructions (Affymetrix Inc., Santa Clara, CA, USA).

The in-house cell lines were analyzed on HTA arrays. The data have partly been published previously (GEO accession numbers GSE24550, GSE29638, GSE69182,

GSE79959, and GSE97023) and the remaining samples (n = 174 CRCs) have been deposited to GEO with accession number GSE96528. The tumors were classified according to the CMS subtypes of CRC using the RF predictor implemented in the R package CMSclassifier (12) for tumors analyzed on HuEx and HTA arrays separately. Gene set expression enrichment analyses were performed using the R package GSA (Efron et al. AnnAppl Stat 2007;1 : 107- 29) and a customized collection of 51 CRC -related gene sets. Sample-wise gene set expression enrichment scores were calculated using the R package GSVA (Hanzelmann et al. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 2013;14:7). Genes were matched with Entrez IDs, identifying x (x%) and x (x%) of the x genes in the RF predictor for tumors analyzed with the Human Exon 1.0 ST and Human Trans criptome 2.0 Arrays, respectively. A default posterior probability of 0.5 was used as threshold for confident sample classification. Tumors that were indeterminate according to this criterion, but had a larger posterior probability than 0.3 for more than one subtype, were denoted as mixed tumors. The remaining tumors were indeterminate.

Development of the cancer cell-specific CMS classifier

A CMS classifier enriched for cancer cell-intrinsic gene expression signals was developed based on RNA sequencing data from primary CRCs in The Cancer Genome Atlas (TCGA; n = 560) and CRC cell lines (n = 37 unique) (Klijn et al. A comprehensive transcriptional portrait of human cancer cell lines. Nat Biotechnol 2015;33:306-12), as well as a public microarray dataset of PDX tumors and primary CRCs (n = 40 and 30, respectively) ( ien et al. Characterization of a large panel of patient-derived tumor xenografts representing the clinical heterogeneity of human colorectal cancer. Clin Cancer Res

2012;18:5314-28). For the TCGA data, preprocessed gene-level RSEM expression values were downloaded from the Broad GDAC Firehouse (level 3; doi: 10.7908/CHG0KM9) and CMS assignments from the Colorectal Cancer Subtyping Consortium web site at SAGE Synapse (Guinney et al. The consensus molecular subtypes of colorectal cancer. Nat Med 2015;21 : 1350-6). The samples were randomly assigned to a training (75%, n = 417) and a test (25%, n = 143) dataset.

Genes with subtype-specific expression were identified as genes with high relative expression in each CMS group in the TCGA training set. Differential expression analysis was done by comparing each subtype with the rest using the voom approach with quantile normalization in the R package limma, and genes with a log ₂ fold-change > 1 and adjusted P- value < 0.1 in each subtype were retained. To enrich for genes likely to be informative in cell lines and PDX models, and to exclude genes with high expression in the tumor

microenvironment, two additional filters were applied. First, only genes with high expression in CRC cell lines (top 25% expressed genes in at least three samples) and high expression variation (top 25% inter-percentile range [10 ^th to 90 ^th ] among the samples) in the RNA sequencing cell line dataset were retained. Second, genes with high expression in primary

CRCs compared to PDX tumors were filtered out, retaining only genes with a mean log2 fold- change below 2 in the primary CRC versus PDX dataset.

Based on this filtered template gene set representing cancer cell-adapted expression signatures of each CMS group, a collection of 148 CRC cell lines derived from unique patients (totally 169 cell lines) was classified using the Nearest Template Prediction (NTP) algorithm (Hoshida, Y. Nearest template prediction: a single-sample-based flexible class prediction with confidence assessment. PLoS One 2010;5:el5543) with cosine correlation distances to predict the proximity of each sample to the four template signatures. P-values and false discovery rates (FDRs) were calculated based on random resampling (n = 1,000) of the template genes.

CMS-specific genes (n = 343) were identified as differentially expressed genes among the different CMS subtypes in The Cancer Genome Atlas (TCGA) RNA sequencing dataset (n = 319 samples). Prior to classification of cell lines, the CMS-specific gene list was filtered to exclude genes reported to have expression associated with the tumor stroma (33-34); genes higher expressed in the stromal than epithelial compartment of tissue samples from laser micro-dissected CRCs (n = 13) and normal colorectal mucosa (n = 4)(35); genes reported to have expression associated with immune cells (34, 36); genes not highly expressed in CRC cell lines (identified as genes with lower expression than the upper 25 ^th percentile in at least 3 of 38 cell lines analyzed by RNA sequencing (37)); and genes with low expression variation in CRC cell lines (identified as genes with expression variance in the lower 25 ^th percentile among the 38 cell lines- ^' ). Based on this filtered gene set consisting of 343 template genes, nearest template prediction (NTP)-' was used to classify a collection of 150 unique CRC cell lines (gene expression data was obtained from publicly available datasets ¹² ' ^™ ' ^™ and in-house expression analysis of 35 cell lines) using CMS assignments and RNA sequencing data from TCGA as templates. The performance of the cancer cell-specific classifier was assessed in patient samples from three independent series by comparisons with class assignments obtained with the RF CMSclassifier. The correspondence in assigned CMS subtypes was 86% to 93% across the TCGA RNA sequencing dataset, GSE39582 (n = 566) and the in- house dataset (HTA arrays; n = 207).

P-value adjusted for multiple comparisons by the Benjamini-Hochberg method below 0.01 and fold change larger than 1.5.

Drug screening in CRC cell lines

An in-house collection of 35 cell lines, which represents all the four CMS subtypes was analyzed for drug sensitivities in an in vitro screen using an established high-throughput platform. (27) The screen included a library of 460 drugs, with 143 clinically approved and 317 investigational drugs representing different molecular target classes. A drug sensitivity score (DSS) (28) was calculated per drug and cell line relative to a negative and a positive control, based on cell viability after drug treatment at five different concentrations. Drugs (n = 217) with low variance in DSS scores across all cell lines (DSSmax-min < 7 and DSSmax <7 in less than two cell lines) were excluded from further analyses. Quality control showed strong reproducibility of DSS scores between replicate drug screens of the same cell line (RKO; Pearson correlation 0.99, standard deviation of difference between replicates (SD) 1.36) and between different cell lines from the same patient (HT29 and WIDR; Pearson correlation 0.94, SD = 2.77). For analysis of differential drug responses among CMS subtypes, non- canonical CRC cell lines (CCD841CON, COLO320 and COL0741) and paired cell lines from the same patients were excluded (HCT15, WIDR, SW480, IS1). For the 29 remaining cell lines (n = 6, 10, 6 and 7 in CMS1, 2, 3 and 4, respectively), differential drug sensitivities among individual CMS subtypes were detected, using a threshold for the difference in DSS scores between the sample groups of 2.0 and P-values from independent samples t-tests of 0.01 (Figure 4 and Table 5).

Transcriptional profiling and western blotting in cells treated with luminespib

Three CMS4 cell lines with varying levels of sensitivity to HSP90 inhibition (CAC02, LIM2099 and SW480) were seeded in 60 mm dishes 24 hours prior to exposure to DMSO (control) or 50 nM luminespib. RNA was isolated after treatment for 6 hours (Qiagen Allprep DNA/RNA/miRNA Universal kit) and analyzed on Affymetrix HTA microarrays. Differential gene expression analysis was performed by paired samples t-tests comparing treated and control cells using limma. Protein expression of HSP70 and HSP40 was analyzed by western blotting.

CMS classification of PDX models

PDX models of primary CRCs or liver metastases (n = 32) were established as previously described (Puig I, et al. A personalized preclinical model to evaluate the metastatic potential of patient-derived colon cancer initiating cells. Clin Cancer Res

2013;19:6787-801). One tumor from each mouse and samples from four matching primary CRCs were analyzed for gene expression on Affymetrix Human Gene 2.0 ST arrays. Sample classification was performed using the adapted CMS classifier.

Animals, xenotransplantation and treatments

Among the 32 PDX models classified according to CMS, one model characteristic of

CMS4 (patient ID 43) and one of CMS2 (patient ID 1) were selected for drug treatment. Experiments were conducted following the European Union's animal care directive (2010/63/EU) and were approved by the Ethical Committee of Animal Experimentation of VHIR (Vail d'Hebron Institute of Research)/VHIO (Vail d'Hebron Institute of Oncology; ID: 18/15 CEEA). NOD-SCID (NOD.CB17-Prfefc^/NcrCrl) mice were purchased from Charles River Laboratories (Wilmington, MA, USA). One hundred thousand patient-derived cells suspended in PBS were mixed with Matrigel (1 : 1 v/v-ratio; BD Biosciences, San Jose, CA, USA) and injected subcutaneously into both flanks of NOD-SCID mice. When the tumor reached 0.5 cm ³ in volume, mice (n = 34 for both models) were randomized to each of four different treatment arms, including a control arm (empty vehicle), 5-fluorouracil (5-FU) monotherapy, luminespib (HSP90 inihibitor) monotherapy, and 5-FU + luminespib combination therapy. Luminespib (25 mg/kg in PBS, MCE, Monmouth Junction, NJ, USA) was administered by intraperitoneal injection three times per week. 5-FU (40mg/kg in PBS; Sigma- Aldrich, St. Louis, MO, USA) was administered by intraperitoneal injection twice per week. When matching end-point criteria, mice were euthanized and complete necropsies were performed. Protein expression of HSP70 and Ki67 was analyzed in post-treatment tissue samples by immunohistochemistry.

Statistical analyses

Statistical tests were conducted in R (v.3.3.3), including Fisher's exact test of contingency tables with the function fisher. test, t-tests with equal or unequal variances (Welch's t-test) using the function t.test, prediction accuracy using the confusionMatrix- function in the package caret, and two one-sided test for equivalence using the tost-function in the package equivalence, with the magnitude of similarity determined by the parameter epsilon. Unsupervised principal components analysis (PCA) was done using the prComp function. Univariable and multivariable survival analyses were conducted with Cox's proportional hazards regression, with calculation of P-values from Wald's tests for predictive potential using the SPSS software version 21 (IBM Corporation, Armonk, NY, USA).

Kaplan-Meier survival curves were compared with the log-rank test. Five-year relapse-free survival (RFS, considering relapse after complete resection or death from any cause as events) and overall survival (OS, considering death from any cause as events) were used as endpoints. Anti-tumor activity in PDX models was analyzed using a generalized linear mixed model of tumor volume fold changes, with random effects and treatment arm and time as covariates.

RESULTS Validation of CMS in an independent series of patients with CRC

A consecutive series of primary CRCs (n = 409; Table 1) were classified according to CMS based on their transcriptomic profiles using the random forest (RF) predictor implemented in the R package CMSclassifier (12). Confident classification (posterior probability above 0.5) was obtained for 323 (79%) of the tumors, with a similar distribution among the subtypes as in the original publication (12); 64 (16%) tumors in CMSl, 139 (34%) in CMS2, 56 (14%) in CMS3 and 64 (16%) in CMS4 (Figure la). Another 46 tumors (11%) displayed mixed characteristics between two of the subtypes, while 40 tumors (10%) were indeterminate.

The previously described molecular (MSI status, BRAF mutations, KRAS mutations and TP 53 mutations) and clinicopathological (patient gender, tumor localization, tumor differentiation grade and cancer stage) associations of each subtype were confirmed. Tumors in CMSl were enriched for MSI and BRAF mutations (P = 2xl0 ^"16 from Fisher's exact test of both), while tumors in CMS3 and CMS2 were enriched for KRAS and TP 53 mutations, respectively (P = 0.0002 and 2xl0 ^"6 ; Figure lb). With respect to clinicopathological characteristics, tumors with CMSl were more frequently found in female patients (P = 0.001), in the right side of the colon (P = 9xl0 ^"15 ) and had a low differentiation grade (P = 2xl0 ^"8 ), compared with tumors with CMS2-4 (Figure lc). Furthermore, CMS2 tumors were most frequently found in the rectum (P = 3xl0 ^"8 ), whereas CMS4 was enriched for cancers with advanced stage (stages III and IV; P = 0.01) and was most frequently found in male patients (P = 0.0008).

Prognostic associations of CMS

The poor prognosis associated with the CMS4 subtype was validated in our patient series. Patients with CMS4 tumors had a 5-year overall survival (OS) and relapse-free survival (RFS) rate of 49% and 47%, respectively, significantly lower than the corresponding survival rates of 71% and 67% for patients with CMSl-3 tumors (HR = 2.0 [1.3-3.1] and 1.8 [1.2-2.7]; P = 0.001 and 0.005 for OS and RFS, respectively; Figure Id). There was no significant difference in the 5-year OS or RFS rates among patients with CMSl, CMS2 and CMS3 tumors (P > 0.2). In multivariable analyses, the prognostic value of CMS4 for 5-year OS was independent of patient age and gender, cancer stage, the localization and MSI status of the tumor, as well as adjuvant chemotherapy (multivariable HR = 1.7 [1.1-2.7], P = 0.02; Table 2). For 5-year RFS, the prognostic value of CMS4 remained borderline significant in the same multivariable model (multivariable HR = 1.5 [0.9-2.3], P = 0.09). Also when excluding patients that had emergency surgery, incomplete resection, bowel perforation, or received pre-operative radiotherapy (n = 67), patients with CMS4 tumors had worse OS and RFS rates than patients with CMS 1-3 (Figure 2).

CMS classification of CRC models

To explore the properties of individual CMS subtypes in vitro, a CMS classifier specific to cancer cells, independent of the tumor microenvironment and optimized for CRC cell lines, was generated. In short, (i) candidate genes with high relative expression in each subtype were identified in The Cancer Genome Atlas (TCGA) CRC gene expression data; (ii) genes highly expressed in immune and stromal compartments were filtered out; and (iii) remaining genes with high expression levels and high expression variation in a public gene expression dataset of CRC cell lines were used as features for Nearest Template Prediction (NTP (23); n = 343 genes). To confirm that the classifier was able to correctly identify tumors of each subtype, prediction accuracy was assessed in three sets of primary CRCs. The prediction accuracy in independent tumors from TCGA (n = 113 tumors analyzed by RNA sequencing), GSE14333 (n = 290 tumors analyzed on Affymetrix HG U133 Plus 2.0 arrays) and the in-house patient series (n = 207 tumors analyzed on Affymetrix HTA arrays) was 86% [75%-93%], 92% [86%-96%] and 93% [88%-96%], respectively, demonstrating robust performance independent of gene expression platform (Table 3).

Next, the classifier was applied on a collection of 150 unique CRC cell lines (gene expression data was obtained from multiple publicly available datasets (13, 16, 24)and in- house expression analysis of 35 cell lines). Confident CMS classification was obtained for 131 (87%) of the cell lines, using a threshold for the false discovery rate (FDR) from NTP of 0.2. The distribution of cell lines among the subtypes was similar to the in-house patient series; 28 cell lines (19%) in CMS1; 51 (34%) in CMS2; 24 (16%) in CMS3; 28 (19%) in CMS4 (Figure 3 a).

To determine whether key characteristics of CMS could be recapitulated in CRC cell lines, we explored associations between CMS and other molecular data. Similarly to primary CRCs, cell lines of the CMS1 subtype showed strong enrichment for MSI (P = 8xl0 ^"5 ) and BRAF mutations (P = 3xl0 ^"4 ; Figure 3b and Table 4). Furthermore, CMS3 cell lines were also frequently MSI+. TP 53 mutations were enriched in CMS2, although not statistically significant, while KRAS mutations were more evenly distributed among CMS2-4 than in the patient material. Gene set enrichment analyses confirmed that the properties of the CMS subtypes were recapitulated in the cell line classification (Figure 3c). CMS1 and CMS4 were dedifferentiated, while CMS2 and CMS3 showed clear epithelial characteristics. CMS2 additionally had up-regulation of HNF4A targets, while CMS3 was enriched for metabolic pathways. CMS4 was specifically characterized by EMT activation, extracellular matrix organization and TGFp responses. Taken together, these results demonstrate that CRC cell lines retain key hallmarks of CMS, and are suitable tools for in vitro exploration of the biology of the expression subtypes.

Differential drug sensitivities among the CMS subtypes

Representative CRC cell lines of all the four CMS subtypes (n = 6, 10, 6 and 7 in

CMS1, 2, 3 and 4, respectively), were analyzed for drug sensitivities in an in vitro screen using an established high-throughput platform (n = 460 drugs). (27)Quality control showed strong reproducibility of drug sensitivity scores (DSS) (28) between independent drug screens of the same cell line (RKO; Pearson correlation 0.99, standard deviation of difference between repeated screens (SD) 1.36) and between different cell lines from the same patient (HT29 and WIDR; Pearson correlation 0.94, SD = 2.77).

Differential drug sensitivities among individual CMS subtypes were detected (Figure 4 and Table 5). Cell lines of the CMS 1 subtype were, in comparison with CMS2, more sensitive to chemotherapeutic drugs, particularly inhibitors of topoisomerases. However, CMS1 cell lines were also more sensitive to Hsp90-inhibitors and to 2-methoxy estradiol (2ME; combined angiogenesis- and tubulin-inhibitor). CMS 1 -specific sensitivity to these drugs was also found in comparison with CMS3 cell lines, which frequently also were MSI+ (four MSI+ cell lines of totally six in both CMS1 and CMS3). There were few drugs with differential sensitivities between CMS1 and CMS4 cell lines. However, cell lines in CMS4 showed strong sensitivity to Hsp90-inhibitors and atorvastatin (HMG-CoA reductase- inhibitor) compared with both CMS2 and CMS3, and also to 2ME and disulfiram (inhibitor of acetaldehyde dehydrogenase) in comparison with CMS2. Considering the similarity between CMS1 and CMS4 relative to CMS2 and CMS3, both in drug responses and gene expression patterns, a two-way comparison between these two groups of cell lines was performed. Strong sensitivity to Hsp90-inhibitors (luminespib, ganetespib and radicicol) and 2ME, as well as atorvastatin and indibulin, another tubulin-inhibitor, was found specifically in CMS1 and CSM4, compared with CMS2 and CMS3 (Figure 5a and Tables 6-7). Other drugs with notable selectivity for CMS 1 and CMS4 over CMS2 and CMS3 include the topoisomerase II inhibitors valrubicin, daunorubicin and doxorubicin, the survivin inhibitor YM155, the PAK inhibitor PF-03758309, the PLK inhibitor rigosertib, the farsenyltransferase inhibitor tipifarnib, and the anti-metabolite cytarabine. Additional drugs with differential sensitivity supporting CMS subtype-specific treatment of CRC are listed in tables 5-7.

Technical and biological validation of sensitivity to Hsp90-inhibitors in CMSl and CMS4

For independent technical validation, published data on sensitivity of 15 CRC cell lines (9 overlapping with our drug screen) to the Hsp90-inhibitor ganetespib (29) was analyzed in relation to CMS classification. Lower ICso-values in CMSl and CMS4 (average 22.6 nM) compared with CMS2 and CMS3 (average 42 nM) confirmed higher sensitivity in the first group of cell lines (Figure 6a). Similarly, among 34 CRC cell lines from the

Genomics of Drug Sensitivity in Cancer Project (30) (17 overlapping with our drug screen), higher sensitivity to the Hsp90-inhibitor CCT018159 was confirmed in CMSl and CMS4 (average IC ₅ o 64 μΜ) compared with CMS2 and CMS3 (average IC ₅ o 519μΜ; Figure 6b). The association between Hsp90-inhibition and CMS subtypes was not found in data from the Cancer Cell Line Encyclopedia. (13)

For independent biological validation of the CMS-associated drug sensitivities, five additional cell lines were identified as belonging to either the CMS l (LIM2405) or CMS4 (CAR1, HCA7, LIM2099 and OUMS23) subtypes based on publicly available gene expression data, and subsequently screened for drug sensitivities with the same experimental setup as in the initial discovery screen. In addition, two of the CMS3 cell lines from the initial screen were included as controls. Clear differential sensitivity in CMS l and CMS4 compared with CMS2 and CMS3 was validated for the three Hsp90-inhibitors (luminespib, ganetespib and radicicol), 2-ME, atorvastatin and disulfiram (Figure 5b). Notably, disulfiram had a biphasic dose-response curve.

Extended analyses on additional CRC cell lines (for a total of 67 lines) confirmed the CMS1+CMS4 selective activity of luminespib, ganetespib, radicicol, indibulin, 2- methoxy estradiol, atorvastatin, disulfiram, YM155 (also known as sepantronium bromide) and the Axl inhibitor BGB324 (Figure 7).

HSP90 inhibition alleviates chemoresistance in CMS4 in vivo

In our drug screen panel, CMS4 had a particularly poor response to fluoropyrimidines (P < 0.05 among MSS cell lines). Previous studies have suggested that HSP90 inhibition may sensitize CRC cell lines to chemotherapy, and although monotherapy with HSP90 inhibitors has shown low efficacy in metastatic CRC (Cercek A, et al. Ganetespib, a novel Hsp90 inhibitor in patients with KRAS mutated and wild type, refractory metastatic colorectal cancer. Clin Colorectal Cancer 2014;13:207-12), response has been obtained by combination therapy with HSP90 inhibitors and capecitabine (5-FU pro-drug) in patients who have progressed on fluoropyrimidines (Bendell JC, et al. A phase I study of the Hsp90 inhibitor AUY922 plus capecitabine for the treatment of patients with advanced solid tumors. Cancer Invest 2015;33:477-82). Accordingly, to analyze a potential effect of HSP90 inhibition in vivo, we selected a CMS4 PDX model (MSS, KRAS/NRAS wild type, BRAF ^Y600E mutated) for treatment in a randomized and controlled set-up. Immunodeficient NOD-SCID mice (n = 34) were injected with cells derived from a liver metastasis of a chemotherapy -naive CRC patient and randomized to four treatment arms: (i) control arm with vehicle; (ii) single agent 5-FU; (iii) single agent luminespib; and (iv) combination therapy with 5-FU + luminespib.

Consistent with the cell line data, this CMS4 model showed poor response to chemotherapy (Figure 8, top panel). Chemoresistance was confirmed by staining for the proliferation marker Ki67 in post-treatment samples, and there were no significant changes in Ki67 expression in mice receiving 5-FU compared to vehicle-treated controls. Furthermore, monotherapy with luminespib did not impact on tumor growth, but combined administration of 5-FU + luminespib resulted in significantly greater anti-tumor activity compared to vehicle-treated control (50% reduction in tumor growth, P < 0.001 in generalized linear model) and 5-FU single agent (33% reduction in tumor growth, P < 0.001). Significant up-regulation of HSP70 after treatment with luminespib (both as monotherapy and combined with 5-FU) indicated a specific pharmacodynamic effect of HSP90 inhibition and therefore target dependency. The combination of fluoropyrimidines with HSP90 inhibition was well tolerated, on the basis of minimal changes in mouse body weight. For control, a CMS2 PDX model (MSS,

KRAS/NRAS/BRAF wild type, TP 53 mutated) was treated with the same experimental setup. Inconsistent with the cell line data, single-agent luminespib had a stronger effect on tumor growth in this model, however, HSP90 inhibition (monotherapy or in combination with 5- FU) was not associated with increased expression of HSP70 in post-treatment samples, suggesting that the inhibitory activity was likely a result of off-target effects (Figure 8, bottom panel). Furthermore, this model was highly chemosensitive, as shown by a strong reduction in tumor growth and reduced proliferation in post-treatment samples (Ki67 expression) after treatment with 5-FU compared to vehicle-treated controls, and no synergistic effect of combination treatment with luminespib was detected at the end of the experiment. Discussion

Response to standard oncological treatment is limited in CRC and there is great potential to improve treatment efficacy by molecularly-guided repurposing of targeted drugs. We identify strong relative activity of HSP90 inhibitors in in vitro models of the

transcriptomic CMS 1 and CMS4 groups of CRC by high-throughput drug screening, using a new and cancer cell-adapted CMS classifier. HSP90 inhibition has previously been extensively investigated in cancer and has demonstrated anti-tumor activity in several solid tumor types, mainly as combination therapies (Wang H, et al. Effects of treatment with an Hsp90 inhibitor in tumors based on 15 phase II clinical trials. Mol Clin Oncol 2016;5:326- 34.). However, low response rates are observed in unstratified patient populations. In the only phase II trial reported in CRC, single-agent treatment with ganetespib demonstrated good tolerance but low efficacy in chemotherapy -refractory metastatic disease, independent of KRAS mutation status (Cercek A, et al. Ganetespib, a novel Hsp90 inhibitor in patients with KRAS mutated and wild type, refractory metastatic colorectal cancer. Clin Colorectal Cancer 2014;13:207-12). Higher anti-tumor activity was seen in early clinical trials exploring combinations of HSP90 inhibitors with chemotherapies, including fluoropyrimidines (5-FU and capecitabine) (Bendell JC, et al. A phase I study of the Hsp90 inhibitor AUY922 plus capecitabine for the treatment of patients with advanced solid tumors. Cancer Invest

2015;33:477-82). Our study confirms stronger in vivo anti-tumor activity of combination therapy with HSP90 inhibitors and 5-FU in a chemoresistant CMS4 PDX model. This is concordant with published in vitro data showing that HSP90 inhibition sensitizes CRC cell lines to the effect of 5-FU, oxaliplatin and topoisomerase inhibitors (He S, et al. The HSP90 inhibitor ganetespib has chemosensitizer and radiosensitizer activity in colorectal cancer. Invest New Drugs 2014;32:577-86; Nagaraju GP, et al. HSP90 inhibition downregulates thymidylate synthase and sensitizes colorectal cancer cell lines to the effect of 5FU-based chemotherapy. Oncotarget 2014;5:9980-91; McNamara AV, et al. Hsp90 inhibitors sensitise human colon cancer cells to topoisomerase I poisons by depletion of key anti-apoptotic and cell cycle checkpoint proteins. Biochem Pharmacol 2012;83:355-67). Specifically, our PDX results are in line with a CMS4 cell line-derived xenograft (HCT116) experiment, where ganetespib significantly potentiated the anti -tumor efficacy of capecitabine, causing tumor regression in a model that is intrinsically resistant to fluoropyrimidine therapy. The encouraging preclinical data presented here suggest that targeted inhibitors can overcome chemoresistance in selected CRC populations, opening the door for future targeted therapies based on a patient's CMS classification.

Table 1. Clinicopathological and molecular characteristics of the consecutive CRC patient series (n = 409)

Characteristic Value

Patient age at diagnosis, mean (range) 71 (27 - 97)

Patient gender (female, male) 207, 202

Tumor localization (right, left, rectum, synchronous) 173, 122, 111, 3

Cancer stage (I, II, III, IV) 84, 153, 115, 57

Tumor differentiation grade (high, medium, low, unknown) 9, 336, 56, 8

5-year survival rate (overall, relapse-free) 67%, 63%

Post-operative chemotherapy (stage I, II, III, IV) 0, 4%, 48%, 55%

MSI-status (MSI+, MSS, not scored) 72, 328, 9

KRAS (mutation frequency, number of samples scored) 32.5%, 409

BRAF (mutation frequency, number of samples scored) 17.1%, 409

TP53 (mutation frequency, number of samples scored) 56.7%, 337

PIK3CA (mutation frequency, number of samples scored) 8.9%, 337

PTEN (mutation frequency, number of samples scored) 8.5%, 328

MSI, microsatellite instability; MSS, microsatellite stable

Table 2. Multivariable survival analy

Five-year progression-free survival Five-year overall survival

Univariate Multivariate Univariate Multivariate

HR P- HR P- HR P- HR P-

[95% erf value ^b [95% erf value ^b [95% Clf value ^b [95% Clf value ^b

1.8 1.5 2.0 1.7

CMS4 vi. CMS 1-3 0.005 0.09 0.001 0.02

[1.2-2.7] [0.9-2.3] [1.3-3.1] [1.1-2.7]

Patient age (above vs. 1.1 1.4 1.3 1.6

0.7 0.1 0.1 0.03 below median) [0.8-1.5] [0.9-2.1] [0.9-1.8] [1.0-2.5]

Patient gender (female vs. 0.9 0.9 0.9 0.9

0.3 0.2 0.2 0.3 male) [0.8-1.1] [0.7-1.1] [0.8-1.1] [0.7-1.1]

Cancer stage (III or IV vs. 2.8 2.5 3.3 3.0

2E-09 3E-05 lE-10 4E-06 I or II) [2.0-3.9] [1.6-3.9] [2.3-4.7] [1.9-4.8]

Tumor localization (right 0.9 1.2 1.0 1.2

0.6 0.4 0.8 0.3 vs. left or rectum) [0.7-1.3] [0.8-1.8] [0.7-1.4] [0.8-1.9]

MSI status (MSI+ vs. 0.6 0.8 0.7 0.9

0.08 0.5 0.2 0.8 MSS) [0.4-1.0] [0.5-1.5] [0.4-1.2] [0.5-1.7]

Post-operative 2.1 1.0 2.2 0.9

2E-05 0.9 2E-05 0.8 chemotherapy (yes vs. no) [1.5-3.0] [0.6-1.7] [1.5-3.2] [0.5-1.6] 'Hazard ratios from Cox's regression; ^b P -values from Wald's test of predictive potential

Table 3. Prediction accuracy of cancer cell-specific CMS subtype in patient samples

aCMS classes obtained from Guinney et al., Nat. Med. 21, 1350-1356 (2015); ^b CMS classes obtained using the RF algorithm in the CMSclassifier

Table 4. Enrichment of molecular characteristics in CMS subtypes of CRC cell lines

, , , Number of cell Number of cell Fisher's exact test Molecular ,. .

, _^ . lines in enriched lines in the other

characteristic „,„ , _^ Odds 95% confidence „ ,

CMS subtype subtypes . _^ , P-value ratio interval

MSI-status CMSl CMS2-4

MSI+ vs. 19 26 6.5 2.3 - 20.5 8E-05

MSS 7 63

MSI-status CMS3 CMS2, 4

MSI+ vs. 14 12 6.8 2.2 - 22.7 3E-04

MSS 9 54

BRAF CMSl CMS2-4

mutation vs. 11 8 6.9 2.1 - 23.3 3E-04 wild-type 16 82 Table 5. Differential drug sensitivities between cell lines of individual CMS subtypes

Average

Comparison Drag" difference P-value Molecular targets/mechanisms in DSS ^b

Rigosertib 11.6 0.009 PLK1 inhibitor, non-ATP-comp inhibitor

YM155 11.3 0.005 Survivin inhibitor

Gemcitabine 10.2 0.005 Antimetabolite; Nucleoside analog

2-methoxyestradiol 9.8 0.0003 Angiogenesis inhibitor; tubulin inhibitor

Idarubicin 9.4 0.007 Topoisomerase II inhibitor

Indibulin 9.3 0.006 Mitoric inhibitor. Microtubule depolymenzer

Valrubicin 9.0 0.005 Topoisomerase II inhibitor

Luminespib 8.2 0.006 Hsp90 inhibitor

Daunorabicin 7.5 0.007 Topoisomerase II inhibitor

Carfilzomib 7.4 0.009 Proteasome inhibitor (20S subunit)

Radicicol 7.3 0.009 Hsp90 inhibitor

CMS1 vs. Omacetaxine 7.1 0.007 Protein synthesis inhib (80S ribosome)

CMS2 Doxorubicin 6.7 0.009 Topoisomerase II inhibitor

Serdemetan 6.0 0.003 HDM2-P53 antagonist

Aldoxorubicin 5.8 0.007 Topo II, albumin

OTX015 5.6 0.005 BRD2, 3, 4

TH588 5.2 0.007 MTH1 inhibitor

I-BET151 4.9 0.002 BET family inhibitor

KU-60019 4.2 0.005 ATM inhibitor

Atorvastatin 3.9 0.009 HMG CoA reductase inhibitor

Auranofin 3.7 0.004 Antirheumatic agent

UNC0642 3.5 0.005 G9a/GLP inhibitor

AT 101 3.4 0.005 Bel family inhibitor

SGC-CBP30 3.0 0.005 CREBBP/EP300 bromodomain inhibitor

2-methoxyestradiol 10.3 0.004 Angiogenesis inhibitor; tubulin inhibitor

CMS1 vs.

CMS3 Mepacrine aq 5.1 0.002 PLA2 inhibitor. NF-kB inhibitor, p53 activator

SGC-CBP30 4.0 0.002 CREBBP/EP300 bromodomain inhibitor

CMS2 vs. Carfilzomib -6.6 0.010 Proteasome inhibitor (20S subunit)

CMS3 Oprozomib -5.1 0.003 Irreversible proteasome (20 S) inhibitor

Disulfiram -12.7 0.006 Alcohol dehydrogenase inhibitor

Luminespib -8.1 0.0002 Hsp90 inhibitor

Radicicol -7.1 0.003 Hsp90 inhibitor

Ganetespib -7.1 0.0004 Hsp90 inhibitor

2-methoxyestradiol -6.1 0.002 Angiogenesis inhibitor; tubulin inhibitor

Talazoparib -4.9 0.010 PARPl/2 inhibitor

Daunorabicin -4.9 0.010 Topoisomerase II inhibitor

CMS2 vs. Atorvastatin -4.6 0.0001 HMG-CoA-reductase inhibitor

CMS4 LY-2874455 -4.6 0.002 FGFR inhibitor

RAF265 -4.6 0.008 C-Raf inhibitor

Cisplatin aq -4.5 0.009 Platinum-based antineoplastic agent

Ponatinib -4.2 0.001 Broad TK inhibitor

Nutlin-3 -3.8 0.007 MDM2 inhibitor

Rocilinostat -3.3 0.007 HDAC-6 selective inhibitor

Nintedanib -3.2 0.006 VEGFR, PDGFR, FGFR inhibitor

UNC0642 -3.0 0.007 G9a/GLP inhibitor CMS3 v _i Luminespib -9.7 0.001 Hsp90 inhibitor

CMS4 Ganetespib -8.3 0.009 Hsp90 inhibitor

Atorvastatin -4.5 0.0002 HMG-CoA-reductase inhibitor ^a All drugs with average difference in DSS scores between the sample groups above 2.0 and P-values from independent samples t-tests below 0.01. Sorted by average difference in DSS scores; ^b Positive values indicate drugs with strongest effect in the first subtype, and vice versa for negative values. DSS, drug sensitivity score

Table 6. Differential drug sensitivities between cell lines of the CMSl and CMS4 versus CMS2 and CMS3 subtypes

Average

Drug ³ difference P-value Molecular targets/mechanisms

in DSS ^b

YM155 8.8 0.003 Survivin inhibitor

Luminespib 8.7 0.00001 Hsp90 inhibitor

Disulfiram 8.3 0.007 Alcohol dehydrogenase inhibitor

2-methoxyestradiol 8.0 0.00001 Angiogenesis inhibitor; tubulin inhibitor

Ganetespib 7.4 0.0001 Hsp90 inhibitor

Radicicol 6.6 0.0005 Hsp90 inhibitor

Indibulin 6.3 0.006 Mitoric inhibitor; microtubule depolymerizer

Cytarabine 6.1 0.009 Anti-metabolite, interferes with DNA synthesis

Valrubicin 5.7 0.007 Topoisomerase II inhibitor

8-chloro-adenosine 5.7 0.001 Nucleoside analog, RNA synthesis inhibitor

Daunorubicin 4.9 0.004 Topoisomerase II inhibitor

Doxorubicin 4.7 0.004 Topoisomerase II inhibitor

Atorvastatin 4.3 0.00001 HMG-CoA-reductase inhibitor

Serdemetan 3.8 0.005 HDM2-P53 antagonist

TH588 3.6 0.005 MTH1 inhibitor

RAF265 3.5 0.004 C-Raf inhibitor

KU-60019 3.0 0.008 ATM inhibitor

BGB324 2.8 0.007 Axl inhibitor

Auranofin 2.5 0.009 Antirheumatic agent

SGC-CBP30 2.4 0.002 CREBBP/EP300 bromodomain-inhibitor ^a All drugs with average difference in DSS scores between CMS1/CMS4 (n = 13) and CMS2/CMS3 (n = 16) cell lines above 2.0 and P-values from independent samples t-tests below 0.01; sorted by average difference in DSS scores. ^b Positive values indicate drugs with strongest effect in

CMS1/CMS4 cell lines. DSS, drug sensitivity score Table 7. Differential drug sensitivity between CMSl/4 and CMS2/3 cell lines - extended analysis

Average

Drug ³ difference in P-value FDR Molecular targets/mechanisms

DSS ^b

PF-03758309 10.8 8xl0 ^"4 8xl0 ^"3 PAK inhibitor

Rigosertib 10.5 4xl0 ^"4 6xl0 ^"3 PLK1 inhibitor

Disulfiram 9.0 3xl0 ^"3 l xlO ^"2 Alcohol dehydrogenase inhibitor

YM155 9.0 3xl0 ^"3 l xlO ^"2 Survivin inhibitor

Tipifarnib 8.8 lxlO ^"3 9xl0 ^"3 Farnesyltransferase inhibitor

Luminespib 8.3 lxlO ^"4 4xl0 ^"3 HSP90 inhibitor

Ganetespib 8.1 3xl0 ^"5 2xl0 ^"3 HSP90 inhibitor

Idarubicin 7.9 4xl0 ^"4 6xl0 ^"3 Topoisomerase II inhibitor

Teniposide 7.8 2xl0 ^"3 l xlO ^"2 Topoisomerase II inhibitor

Mitotic inhibitor; microtubule

Indibulin 7.4 8xl0 ^"4 8xl0 ^"3

depolymerizer

Dactinomycin 7.2 4xl0 ^"3 2xl0 ^"2 RNA and DNA synthesis inhibitor

Clofarabine 7.2 3xl0 ^"3 l xlO ^"2 Anti-metabolite; Purine analog

Aurora, Ret, TrkA, FGFR-1

Danusertib 7.1 6xl0 ^"3 2xl0 ^"2

inhibitor

2-methoxy estradiol 7.0 lxlO ^"4 4xl0 ^"3 Angiogenesis inhibitor

Radicicol 6.8 7xl0 ^"4 8xl0 ^"3 HSP90 inhibitor

Anti-metabolite, interferes with

Cytarabine 6.7 4xl0 ^"3 2xl0 ^"2

DNA synthesis

Gemcitabine 6.6 lxlO ^"2 4xl0 ^"2 Antimetabolite; Nucleoside analog

PHA-793887 6.4 2xl0 ^"3 l xlO ^"2 CDK inhibitor

Valrubicin 6.4 2xl0 ^"3 l xlO ^"2 Topoisomerase II inhibitor

Nucleoside analog; RNA synthesis

8-chloro-adenosine 5.9 9xl0 ^"4 8xl0 ^"3

inhibitor

aTop 20 drugs (FDR from independent samples t-tests below 0.05) sorted by average

difference in DSS values between CMS1/CMS4 (n = 15) and CMS2/CMS3 (n = 14) cell lines. ^b Positive values indicate drugs with strongest effect in CMS1/CMS4 cell lines.

References

I Ferlay, J. et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int. J. Cancer 136, E359-386 (2015).

2 Linnekamp, J. F., Wang, X., Medema, J. P. & Vermeulen, L. Colorectal cancer heterogeneity and targeted therapy: a case for molecular disease subtypes. Cancer Res. 75, 245-249 (2015).

3 Sottoriva, A. et al. A Big Bang model of human colorectal tumor growth. Nat. Genet. 47, 209-216 (2015).

4 Dienstmann, R., Salazar, R. & Tabernero, J. Personalizing colon cancer adjuvant therapy: selecting optimal treatments for individual patients. J. Clin. Oncol. 33, 1787-1796 (2015).

5 Walther, A. et al. Genetic prognostic and predictive markers in colorectal cancer. Nat.Rev.Cancer 9, 489-499 (2009).

6 Azuara, D. et al. Nanoflui die Digital PCR and Extended Genotyping of RAS and BRAF for Improved Selection of Metastatic Colorectal Cancer Patients for Anti-EGFR Therapies. Mol. Cancer Ther. 15, 1106-1112 (2016).

7 Misale, S., Di, N. F., Sartore-Bianchi, A., Siena, S. & Bardelli, A. Resistance to anti- EGFR therapy in colorectal cancer: from heterogeneity to convergent evolution. Cancer Discov. 4, 1269-1280 (2014).

8 TCGA Network. Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330-337 (2012).

9 Kloor, M., Michel, S. & von Knebel Doeberitz, M. Immune evasion of microsatellite unstable colorectal cancers. Int. J. Cancer 127, 1001-1010 (2010).

10 Popat, S., Hubner, R. & Houlston, R. S. Systematic review of microsatellite instability and colorectal cancer prognosis. J.Clin.Oncol. 23, 609-618 (2005).

I I Le, D. T. et al. PD-1 Blockade in Tumors with Mismatch-Repair Deficiency. N. Engl. J. Med. 372, 2509-2520 (2015).

12 Guinney, J. et al. The consensus molecular subtypes of colorectal cancer. Nat. Med. 21, 1350-1356 (2015).

13 Barretina, J. et al. The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483, 603-607 (2012).

14 Ahmed, D. et al. Epigenetic and genetic features of 24 colon cancer cell lines.

Oncogenesis 2, e71 (2013). 15 Mouradov, D. et al. Colorectal cancer cell lines are representative models of the main molecular subtypes of primary cancer. Cancer Res. 74, 3238-3247 (2014).

16 Medico, E. et al. The molecular landscape of colorectal cancer cell lines unveils clinically actionable kinase targets. Nature communications 6, 7002 (2015).

17 van de Wetering, M. et al. Prospective derivation of a living organoid biobank of colorectal cancer patients. Cell 161, 933-945 (2015).

18 Calon, A. et al. Stromal gene expression defines poor-prognosis subtypes in colorectal cancer. Nat. Genet. 47, 320-329 (2015).

19 Fessler, E. et al. A multidimensional network approach reveals microRNAs as determinants of the mesenchymal colorectal cancer subtype. Oncogene (2016).

20 Efiron, B. & Tibshirani, R. On testing the significance of sets of genes. Ann. Appl. Stat. 1, 107-129 (2007).

21 Becht, E. et al. Immune and stromal classification of colorectal cancer is associated with molecular subtypes and relevant for precision immunotherapy. Clin. Cancer Res.16, 4057-4066 (2016).

22 Rooney, M. S., Shukla, S. A, Wu, C. J., Getz, G. & Hacohen, N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell 160, 48-61 (2015).

23 Hoshida, Y. Nearest template prediction: a single-sample-based flexible class prediction with confidence assessment. PLoS One 5, el5543 (2010).

24 Garnett, M. J. et al. Systematic identification of genomic markers of drug sensitivity in cancer cells. Nature 483, 570-575 (2012).

25 Brabletz, T. EMT and MET in metastasis: where are the cancer stem cells? Cancer Cell 22, 699-701 (2012).

26 Lamouille, S., Xu, J. & Derynck, R. Molecular mechanisms of epithelial - mesenchymal transition. Nat. Rev. Mol. Cell Biol. 15, 178-196 (2014).

27 Pemovska, T. et al. Individualized systems medicine strategy to tailor treatments for patients with chemorefractory acute myeloid leukemia. Cancer Discov. 3, 1416-1429 (2013).

28 Yadav, B. et al. Quantitative scoring of differential drug sensitivity for individually optimized anticancer therapies. Sci. Rep. 4, 5193 (2014).

29 He, S. et al. The HSP90 inhibitor ganetespib has chemosensitizer and radiosensitizer activity in colorectal cancer. Invest. New Drugs 32, 577-586 (2014).

30 Yang, W. et al. Genomics of Drug Sensitivity in Cancer (GDSC): a resource for therapeutic biomarker discovery in cancer cells. Nucleic Acids Res. 41, D955-961 (2013). 31 Hieronymus, H. et al. Gene expression signature-based chemical genomic prediction identifies a novel class of HSP90 pathway modulators. Cancer Cell 10, 321-330 (2006).

32 Maloney, A. et al. Gene and protein expression profiling of human ovarian cancer cells treated with the heat shock protein 90 inhibitor 17-allylamino-17- demethoxygeldanamycin. Cancer Res. 67, 3239-3253 (2007).

33 Isella, C. et al. Stromal contribution to the colorectal cancer transcriptome. Nat. Genet. 47, 312-319 (2015).

34 Yoshihara, K. et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat. Commun. 4, 2612 (2013).

35 Nishida, N. et al. Microarray analysis of colorectal cancer stromal tissue reveals upregulation of two oncogenic miRNA clusters. Clin. Cancer Res. 18, 3054-3070 (2012).

36 Newman, A. M. et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods 12, 453-457 (2015).

37 lijn, C. et al. A comprehensive transcriptional portrait of human cancer cell lines. Nat. Biotechnol. 33, 306-312 (2015).

All publications and patents mentioned in the above specification are herein incorporated by reference. Various modifications and variations of the described method and system of the invention will be apparent to those skilled in the art without departing from the scope and spirit of the invention. Although the invention has been described in connection with specific preferred embodiments, it should be understood that the invention as claimed should not be unduly limited to such specific embodiments. Indeed, various modifications of the described modes for carrying out the invention that are obvious to those skilled in the medical sciences are intended to be within the scope of the following claims.