60/432, 811, filed December 11,2002.

BACKGROUND OF THE INVENTION 1. Field of the Invention This invention relates to methods for predicting the response to cancer therapy in an individual.

2. Background of the Invention Cellular growth and differentiation processes involve growth factors that exert their actions through specific receptors expressed in the surfaces of responsive cells.

Ligands binding to surface receptors, such as those that carry an intrinsic tyrosine kinase activity, trigger a cascade of events that eventually lead to cellular proliferation and differentiation (Carpenter et al., Biochem. , 48: 193-216,1979 ; Sachs et al., Cancer Res., 47 : 1981-1986,1987). Receptor tyrosine kinases can be classified into several groups on the basis of sequence similarity and distinct features. One of these groups includes the epidermal growth factor receptor family, which included erbB-1 (EGFR or HER-1) (Carpenter et al., Biochem. , 48: 193-216,1979) ; erbB-2 (HER-2/neu) (Semba et al., Proc.

Natl. Acad. Sci. , 82: 6497-6501,1985 ; Coussens et al., Science, 230: 1130-1139,1985, Bargmann et al., Cell, Vol. 45,649-657, 1986); erbB-3 (HER-3) (Kraus et al., Proc. Natl.

Acad. Sci. , 86: 9193-9197,1989 ; Carraway et al., R. A. J. Biol. Chem. , 269: 14303- 14306, 1994), and erbB-4 (HER-4) (Plowman et al., Nature, 366: 473-475,1993 ; Tzahar et al., Biol. Chem. , 269: 25226-25233,1994).

NDF (neu differentiation factor)/Heregulin is a receptor tyrosine kinase ligand that can stimulate the tyrosine phosphorylation of erB-2 through heterodimerization with its receptors erbB-3 or erbB-4. (Peles, et al., Cell, 69: 205-216,1992, 1993; Peles, et al., EMBO J. Mar; 12 (3): 961-71.1993 ; Holmes et al, Science, 256: 1205-1210,1992. Tzahar et al., Biol. Chem. , 269: 25226-25233,1994 ; Plowman et al., Nature, 366: 473-475, 1993; Pinkas-Kramarski et al., Proc. Natl. Acad. Sci. , 91: 9387-9391,1994 ; Pinkas- Kramarski et al., The Journal of Biological Chemistry, Vol. 271, No. 32: 19029-19032,

1996; Pinkas-Kramarski et al., Oncogene, 16,1249-1258, 1998. ) Depending on the cell line studied, NDF/Heregulin can either elicit a growth arrest and differentiation phenotype, resulting in morphological changes, induction of lipids, and expression of intracellular adhesion molecule-1, or induce a mitogenic response (Holmes et al., Science, 256: 1205-1210,1992 ; Peles et al., Cell, 69: 205-216,1992 ; Bacus et al., Cancer Res. 53: 5251-5261,1993).

Activation of erbB receptor heterodimers is coupled to and stimulates downstream MAPK-Erkl/2 and PI3K-AKT growth and survival pathways whose deregulation in cancer has been linked to disease progression and refractoriness to therapy (Olayioye, M. A., et al., Mol. Cell. Biol. 18,5042-5051 (1998), Fukazawa, T., et al., J. Biol. Chem.

271,14554-14559 (1996), Hackel, P. O., et al., Curr. Opin. Cell Biol. 11,184-189 (1999); Tzahar, E., et al., Mol. Cell. Biol. 16,5276-5287 (1996); Lange, C. A., et al., J. Biol.

Chem. 273,31308-31316 (1998). HER-3 is a major docking site for phoshoinositide-3- kinase (PI3K). In addition, NDF/Heregulin stimulation causes activation of the PI3K pathway and phosphorylation of AKT (Altiok et al., J. Biol. Chem. , 274,32274-32278, 1999; Liu et al., Res. Comm. , 261,897-903, 1999; Xing et al., Nature Med. , 6,189-195, 2000). These observations implicate PI3K/AKT in the signaling cascade that results from HER-3 heterodimerization with overexpressed HER-2/neu receptors in breast cancer cells; activation of PI3K/AKT promote cell survival and enhanced tumor aggressiveness (Shak, Semin. Oncol., Suppl 12: 71-77,1999 ; Huang et al., Clinical Cancer Res. , Vol. 7: 2166-2174,2000). In addition, AKT2 was reported to be activated and overexpressed in HER-2/neu-overexpressing breast cancers (Bacus et al., Oncogene, 21: 3532-3540,2002).

Most tumors of epithelial origin express multiple erbB (HER) receptors and co- express one or more EGF-related ligands suggesting that autocrine receptor activation plays a role in tumor cell proliferation. Because these ligands activate different erbB/HER receptors, it is possible that multiple erbB receptor combinations might be active in a tumor, a characteristic that could influence its response to an erbB-targeted therapeutic. Fore example, erbB-2/HER-2 is overexpressed in 20 to 30% of all breast cancers, and its overexpression is associated with poor prognosis, suggesting that it could be used as a target for anti-tumor agents (Slamon et al., Science, 235: 177-182,1987 ; Tagliabue et al. , Int. J. Cancer, 47: 933-937,1991 ; Hudziak et al. , Mol. Cell. Biol. , 9: 1165-1172,1989). Studies have shown that in erbB-2 overexpressing breast cancer cells, treatment with antibodies specific to HER-2/erbB-2 in combination with chemotherapeutic agents (e. g. , cisplatin, doxoubicin, taxol) elicits a higher cytotoxic

response than treatment with chemotherapy alone (Hancock et al., Cancer Res. , 51: 4575- 4580,1991 ; Arteaga et al., Cancer, 54: 3758-3765,1994 ; Pietras et al. , Oncogene, 9: 1829-1838,1994). One possible mechanism by which HER-2/erbB-2 antibodies might enhance cytotoxicity to chemotherapeutic agents is through the modulation of the HER- 2/erbB-2 protein expression, (Bacus et al. , Cell Growth & Diff. , 3: 401-411,1992, Bacus et al., Cancer Res. 53: 5251-5261,1993 ; Stancovski et al. , Proc Natl Acad Sci USA 88: 8691-8695,1991 ; Klapper et al. , Oncogene 14,2099-2109, 1997, and Klapper et al., Cancer Res. , 60: 3384-3388,2000), or by interfering with DNA repair (Arteaga et al., Cancer, 54: 3758-3765,1994, and Arteaga et al., J Clinical Oncology, Vol. 19, No 18s, 32s-40s, 2001; Pietras et al. , Oncogene, 9: 1829-1838,1994).

Because of the effect of anti-HER-2/erbB-2 antibodies on cellular growth, a number of approaches have been used to therapeutically target HER-2/erbB-2 or EGFR overexpressing cancers. For clinical use, one approach is to interfere with the kinase activity of the receptor by using inhibitors that block the nucleotide binding site of HER- 2/neu or EGFR (Bruns, et al., Cancer Research, 60,2926-2935, (2000); Christensen, et al, Clinical Cancer Research, Vol. 7,4230-4238, 2001, Erlichman, et al., Cancer Research 61,739-748, 2001, Fujimura, et al. , Clinical Cancer Research, Vol. 8,2448-2454, 2002; Herbst, et al. , Journal of Clincal Oncology, Vol. 20, No. 18, 3815-3825, 2002; Hidalgo, et al, J. Clinical Oncology, Vol 19, No 13: pp 3267-3279,2001 ; Moasser, et al, Cancer Res., 61: 7184-7188,2001 ; Normanno, et al, Ann. of Oncol., 13: 65-72,2002). A second approach is using ansamycins to influence the stability of HER2/neu receptors (Munster, et al., Cancer Research 62,3132-3137, 2002; Basso et al, Oncogene, 21: 1159-1166, 2002). Another approach is the use of antibodies directed to various erbB receptors specifically EGFR or HER-2/neu (Alaoui-Jamali, et al Biochem. Cell. Biol. , 75: 315-325, 1997; Albanell, et al. , J. National Cancer Institute, Vol 93, No. 24,1830-31, 2001; Baselga, et al., Pharmacol Ther 64: 127-154,1994 and Baselga, et al., Annuals of Oncology 13: 8-9,2002 ; Mendelsohn, Seminars in Cancer Biology, Vol. 1, pp. 339-344, 1990). A number of monoclonal antibodies and small molecule, tyrosine kinase inhibitors targeting EGFR or erbB-2 have been developed. For example, HERCEPTIN is approved for treating the 25% of women whose breast cancers overexpress erbB-2 protein or demonstrate erbB-2 gene amplification (Cobleigh, M. A., et al., J. Clin. Oncol. 17, 2639-2648 (1999) ). Analysis of various antibodies to HER-2/neu led to the identification of the murine monoclonal, 4D5. This antibody recognizes an extracellular epitope (amino acids 529 to 627) in the cysteine-rich II domain that resides very close to the

transmembrane region. Treatment of breast cancer cells with 4D5 partially blocks NDF/heregulin activation of HER-2-HER-3 complexes, as measured by receptor phosphorylation assays. To allow for chronic human administration, murine 4D5 was fully humanized to generate HERCEPTINO (trastuzumab) (Sliwkowski et al, Sem. in Oncol. , 26: 60-70,1999 ; Ye et al. , Oncogene, 18: 731-738,1999 ; Carter et al, Proc. Natl Acad Sci USA 89: 4285-4289,1992 ; Fujimoto-Ouchi et al, Cancer Chemother Pharmacol, 49: 211-216,2002 ; Vogel, et al., Oncology, 61 (suppl 2): 37-42,2001 ; Vogel, et al. , Journal of Clinical Oncology, Vol 20, No. 3: 719-726,2002). In addition, several EGFR-targeted therapies are currently under clinical investigation (Mendelsohn, J. , & Baselga, J., Oncogene 19,6550-6565 (2000); Xia, W., et al. Oncogene 21,6255-6263 (2002)).

Historically, cytotoxic cancer therapies have been developed based on maximum tolerated doses (MTD), treating patients without understanding the tumor profile for likely responders. Hence, patients were often subjected to toxic therapies with limited therapeutic benefit. Recently, elucidating tumor growth and survival pathways has led to the development of tumor-targeted therapies. An example of this approach is Gleevec, an inhibitor of the c-abl family of tyrosine kinases approved for treating chronic myeloid leukemia and gastrointestinal stromal tumors (Druker, B. J. et al., N. Engl. J. Med. 344, 1031-1037 (2001); Demitri, G. D., et al. ; N. Engl. J. Med. 347,472-480 (2002) ). In contrast, most erbB-receptor targeted therapies primarily exert cytostatic anti-tumor effects, necessitating their chronic administration. Identification of biologically effective doses (BED), the dose or dose range that maximally inhibits the intended target, beyond which dose escalation is likely to add toxicity without benefit, is therefore essential.

Moreover, many of these agents will be used in combination with cytotoxic therapies, where added toxicity may not be tolerable, further supporting BED-based dosing.

Targeted-therapy implies that populations of likely responders exists, and can be identified. In this context, the inability of the mono-EGFR inhibitor IressaTM to demonstrate a survival advantage when added to first-line chemotherapy in metastatic non-small cell lung cancer highlights the importance of identifying how to optimally utilize these agents in the clinic.

In view of the severe and deleterious consequences of administering an inappropriate or ineffective therapy to a human cancer patient, there exists a need in the art for predicting the response to cancer therapy in an individual.

L

SUMMARY OF THE INVENTION This invention provides methods for predicting a response of an individual to a particular cancer treatment regimen.

In a first aspect, the invention provides methods for predicting a response to a HERl/HER2-directed cancer therapy in a human subject, the method comprising the step of assaying a tumor sample from the human subject before therapy with one or a plurality of reagents that detect expression and/or activation of predictive biomarkers for cancer; and determining a pattern of expression and/or activation of at least two of said predictive biomarkers, wherein the pattern predicts the human subject's response to the cancer therapy. In certain embodiments, the predictive biomarkers consist of growth factor receptors, growth factor receptor ligands, and growth factor receptor-related downstream signaling molecules. The growth factor receptors can be HER1 (EGFR), HER2/neu, HER3, and IGFR. The growth factor receptor ligands can be NDF (Heregulin) and TGF- a. The growth factor receptor-related downstream signaling molecules can be pAKT, pERK, and cyclin D. In other embodiments of the invention, the cancer therapy comprises a dual HER1/HER2 kinase inhibitor. In other embodiments, the cancer therapy comprises GW572016, obtained from Glaxo Smith-Kline.

In a second aspect, the invention provides methods for predicting a response to a HERl/HER2-directed cancer therapy in a human subject, the method comprising the step of assaying a tumor sample from the human subject before therapy with one or a plurality of reagents that detect expression and/or activation of predictive biomarkers for cancer; and determining a pattern of expression and/or activation of at least one of said predictive biomarkers, wherein the pattern predicts the human subject's response to the cancer therapy. In certain embodiments, the predictive biomarkers consist of growth factor receptors, growth factor receptor ligands, and growth factor receptor-related downstream signaling molecules. The growth factor receptors can be HER1 (EGFR), HER2/neu, HER3, and IGFR; the growth factor receptor ligands can be NDF (Heregulin) and TGF-a, and the growth factor receptor-related downstream signaling molecules can be pAKT, pERK, and cyclin D. In other embodiments, the predictive biomarker is pERK. In other embodiments of the invention, the pattern of expression and/or activation is the cellular distribution of total ERK within the cell. In further embodiments, the presence of ERK in

the nucleus is predictive of a response to a HERl/HER2-directed cancer therapy in the human subject.

Specific embodiments of the present invention will become evident from the following more detailed description of certain preferred embodiments and the claims.

DESCRIPTION OF THE DRAWINGS Figures 1A and 1B illustrate inhibition of growth and survival pathways by GW572016 in tumor cell lines and xenografts. Figure 1A represents a Western blot analysis of GW572016-treated HN5, an EGFR-overexpressing tumor cell line, and BT474, an erbB-2 overexpressing tumor cell line. Figure 1B represents quantitative immunohistochemical analysis of p-erbB-2, p-EGFR, p-Erkl/2, p-AKT, and cyclin D1 from BT474 xenografts obtained from tumor-bearing C. B-17 SCID mice treated with either GW572016 (100 mg/kg) or vehicle alone for 5 doses. The optical density (OD) values assigned to each specimen are shown. OD values of <10, 10-15, >15 roughly correlate with 1,2, 3+ in the HercepTest standards, respectively.

Figures 2A through 2D illustrates that inhibition of the pro-survival factor, p- AKT, in response to GW572016 therapy correlates with tumor cell apoptosis. Figure 2A represents an immunohistochemical analysis of EGFR and erbB-2 in a metastatic breast cancer nodule from patient 361. Figure 2B shows reduced expression of intra-tumor p- AKT by day 21 (d 21) of therapy in a woman (patient 372) with metastatic breast cancer overexpressing erbB-2 and erbB-3, thereby demonstrating that GW572016 leads to the inhibition of p-AKT. Figure 2C demonstrates inhibition of intra-tumor p-AKT in Patients 372 and 369. Both had favorable clinical responses to therapy, with 372 achieving a partial remission and 369 experiencing disease stabilization. Day 21 tumor biopsies from both patients revealed a 9-to 10-fold increase in tumor cell apoptosis as quantified by TUNEL assay. Figure 2D represents the results of TUNEL assay using 10X and 40X magnification for Patient 361. The percentage of TUNEL positive cells observed prior to day 0 (d 0), and at day 21 of therapy are indicated. TUNEL staining reveals increased tumor cell apoptosis. Patient 361 also achieved a partial remission in response to GW572016 therapy.

Figures 3A and 3B illustrates that GW572016 therapy inhibits p-Erkl/2 expression in tumors that overexpress both EGFR and erbB-2. Figure 3A shows that

patients 361 and 364 had markedly elevated p-Erk indices prior to therapy. Figure 3B shows sequential tumor biopsies from a metastatic nodule (patient 361) analyzed for total Erkl/2 protein expression prior to (=d 0) and after 21 days of therapy (d 21). Whereas Erkl/2 was located in the nucleus prior to therapy, it was exclusively cytoplasmic at d 21.

Two fields representative of the entire biopsy are shown. The shift in the intra-cellular localization of total Erkl/2 was consistent with the inhibition of p-Erkl/2.

Figure 4 shows that Cyclin D1 immunoreactivity is reduced in response to GW572016 therapy. Inhibition of cyclin Dl immunoreactivity is shown in sequential tumor biopsies obtained prior to (d 0) and after 21 days of GW572016 therapy from patients 361 and 364.

DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS This invention provides methods for predicting response in cancer subjects to cancer therapy, including human cancer patients.

In contrast to traditional anticancer methods, where chemotherapeutic drug treatment is undertaken as an adjunct to and after surgical intervention, neoadjuvant (or primary) chemotherapy consists of administering drugs as an initial treatment in cancer patients. One advantage of such an approach is that, for primary tumors of more than 3 cm, it permits the use of conservative surgical procedures (as opposed to, e. g., radical mastectomy in breast cancer patients) for the majority of patients, due to the tumor- shrinking effect of the chemotherapy. Another advantage is that for many cancers, a partial and/or complete response is achieved in about two-thirds of all cases. Finally, because the majority of patients are responsive after two to three cycles of chemotherapeutic treatment, it is possible to monitor the in vivo efficacy of the chemotherapeutic regimen employed, which is important for a timely identification of those cancers which are non-responsive to chemotherapeutic treatment. Timely identification of non-responsive tumors, in turn, allows the clinician to limit the cancer patient's exposure to unnecessary side-effects of treatment and to institute alternative treatments. However, the methods present in the art, including histological examination, are insufficient for such a timely and accurate identification. The present invention provides methods by which a more informed and effective region of therapy can be administered.

A cancer diagnosis, both an initial diagnosis of disease and subsequent monitoring of the disease course (before, during, or after treatment) is conventionally confirmed through histological examination of cell or tissue samples removed from a patient.

Clinical pathologists need to be able to accurately determine whether such samples are benign or malignant and to classify the aggressiveness of tumor samples deemed to be malignant, because these determinations often form the basis for selecting a suitable course of patient treatment. Similarly, the pathologist needs to be able to detect the extent to which a cancer has grown or gone into remission, particularly as a result of or consequent to treatment, most particularly treatment with chemotherapeutic or biological agents.

Histological examination traditionally entails tissue-staining procedures that permit morphological features of a sample to be readily observed under a light microscope. A pathologist, after examining the stained sample, typically makes a qualitative determination of whether the tumor sample is malignant. It is difficult, however, to ascertain a tumor's aggressiveness merely through histological examination of the sample, because a tumor's aggressiveness is often a result of the biochemistry of the cells within the tumor, such as protein expression or suppression and protein activation, which may or may not be reflected by the morphology of the sample.

Therefore, it is important to be able to assess the biochemistry of the cells within a tumor sample. Further, it is desirable to observe and quantitate both gene expression and protein activation of tumor related genes or proteins, or more specifically cellular components of a tumor-related signally pathway.

Cancer therapy can be based on molecular profiling of tumors rather than histology or site of disease. Elucidating the biological effects of targeted-therapies in tumor tissue and correlating these effects with clinical response helps identify the predominant growth and survival pathways operative in tumors, thereby establishing a profile of likely responders and conversely providing a rational for designing strategies to overcoming resistance.

It is necessary to consider additional biomarkers beyond the presence of the target, such as HER-2/neu, for subjects who are considered for treatment with, for example, HERCEPTIN. Not all tumor cells respond to inhibition of ErbB receptors, despite exhibiting aberrant ErbB-1 and/or ErbB-2 expression. Examples include MKN7 and BT474 ErbB receptor-overexpressing carcinoma cell lines, wherein BT474 cells respond to HERCEPTIN but MKN7 cells do not. In addition, the proliferate block induced as a

consequence of decreased ErbB-1 or ErbB-2 receptor activity can be overcome by the presence of EGF-related ligands such as EGF or NDF/Heregulin (Lane et al). This phenomenon can be attenuated using a bispecific ErbB-1/ErbB-2 inhibitor, thus supporting increased efficacy through simultaneous inhibition of multiple ErbB receptors.

These observations have clear implications for ErbB-directed therapeutics, considering the expression of multiple erbB receptors and their ligands in tumors.

A combination of the ErbB-1-directed mAb 225 and mAb 4D5 (the mAb from which HERCEPTIN was derived) inhibited proliferation of an ovarian tumor cell line more strongly than either mAb alone. In addition to ErbB-targeted mAbs, a number of different ErbB-1/ErbB-2-bispecific inhibitors, also referred to as dual EGFR/erbB-2 kinase inhibitors, have been described recently. GW572016 is a non-limiting example of a dual EGFR/erbB-2 kinase inhibitor obtained from Glaxo Smith-Kline.

HER-2/neu overexpression alone was not the only predictor of response to molecules such as HERCEPTIN. From a biological point of view, HER-2/neu is a ligandless orphan receptor in need of its partner HER-1 and HER-3 in order to exert its activity. HER-1 and HER-3 heterodimerization with HER-2 is enhanced by the presence of EGF or NDF, and correspondingly HER-2 activity is dependant on HER-1 or HER-3.

In addition, in many cancers NDF/Heregulin or TGF a are expressed to provide an autocrine loop of HER-2/HER-1 heterodimerization. Downregulation of HER-2/neu is an important pathway for inhibition of signals generated by these heterodimers. This occurs by treatment with HERCEPTU-40, which affects the HER-2/neu receptor downregulation by receptor endocytosis. Furthermore, high levels of phosphorylated ERK (or pAKT) are a negative predictor, in conjunction with the expression of HER-1 and the presence of NDF, pointing to other pathways that might promote proliferation the tumor cellular growth. High pERK was associated with resistance to HERCEPTIN through downregulation of p27, and may indicate that other signals such as the estrogen receptor's cross activation of the MAPK and AKT pathways may contribute to high levels of pERK and to tumor cell growth or survival in a manner independent of the tumor growth for the erbB receptor. Interestingly, data from clinical trials have shown that using a dual inhibitor to HER-1/neu and HER-2/neu has clinical efficacy in patients when treatment induced downregulation of pERK and pAKT, but not in those patients where pERK and pAKT levels didn't diminish after treatment. Thus those patients having tumors that expressed HER-1, HER-2, pERK and pAKT showed a clinical response to dual EGFR/erbB-2 kinase inhibitors such as GW572016, whereas little or no clinical response

was observed in patients where pERK and pAKT activity remained high after treatment.

In addition, patients bearing tumors that expressed elevated levels of phosphorylated AKT showed a poor response to HERCEPTS@ ; this effect is accompanied by and may be related to high level expression of other growth factor receptors in such cells, including insulin-like growth factor receptor (IGFR-1) and platelet-derived growth factor receptor (PDGFR). Another important predictor of response to HERCEPT1Ne is upregulation of the AKT/mTOR pathway by Heregulin/NDF. Patients having this phenotype showed tumor cell growth inhibition after HERCEPTINX treatment associated with upregulation of p27. As provided herein, the invention provides methods for using analysis of diagnostic tumor markers, specifically related to the erbB receptor family, to predict response to erbB-directed therapies. In particular, the results disclosed herein indicate that HERCEPTIN treatment is effective when a patient's tumor cell growth is dependent on and regulated by a cellular pathway involving AKT/mTOR that is driven by erbB receptor and not by other tyrosine kinases, like IGFR-1 and PDGFR. Thus, when elevated levels of downstream effector molecules such as AKT/mTOR occur independently of erbB receptor activation, HERCEPTIN treatment was found to be ineffective. The inventive methods provide a diagnostic basis for determining prior to treatment whether erbB-directed therapies, such as HERCEPTIN are likely to be effective, thus minimizing the risk of administering ultimately ineffective treatments to cancer patients.

Automated (computer-aided) image analysis systems known in the art can augment visual examination of samples. In a representative system, the cell or tissue sample is exposed to detectably labeled reagents specific for a particular biological marker, and the magnified image of the cell is then processed by a computer that receives the image from a charge-coupled device (CCD) or camera such as a television camera.

Such a system can be used, for example, to detect and measure expression and activation levels of Herl, HER2, HER3, HER4, ERK, pERK, NDF, TGF a, IGFR, c-kit, SCF, pc- kit, and pAKT in a sample. Additional biomarkers are also contemplated by this invention. This methodology provides more accurate diagnosis of cancer and a better characterization of gene expression in histologically identified cancer cells, most particularly with regard to expression of tumor marker genes or genes known to be expressed in particular cancer types and subtypes (i. e. , different degrees of malignancy).

This information permits a more informed and effective regimen of therapy to be

administered, because drugs with clinical efficacy for certain tumor types or subtypes can be administered to patients whose cells are so identified.

Another drawback of conventional anticancer therapies is that the efficacy of specific chemotherapeutic agents in treating a particular cancer in an individual human patient is unpredictable. In view of this unpredictability, the art is unable to determine, prior to starting therapy, whether one or more selected agents would be active as anti- tumor agents or to render an accurate prognosis of course of treatment in an individual patient. This is especially important because the same clinical cancer may present the clinician with a choice of treatment regimens, without any current way of assessing which regimen will be most efficacious for a particular individual. It is an advantage of the methods of this invention that they are able to better assess the expected efficacy of a proposed therapeutic agent (or combination of agents) in an individual patient.

Additional advantageous features of the claimed methods for assessing the efficacy of chemotherapeutic regimens are that they are both time-and cost-effective and minimally traumatic to cancer patients.

For example, expression and activation of proteins expressed from tumor-related genes can be detected and quantitated using methods of the present invention. Further, expression and activation of proteins that are cellular components of a tumor-related signaling pathway can be detected and quantitated using methods of the present invention. Further, proteins associated with breast cancer can be quantified by image analysis using a suitable primary antibody against biomarkers, such as, but not limited to, Her-1, Her-2, p-Her-1, p-Her-2, p-ERK, or p-AKT, and a secondary antibody (such as rabbit anti-mouse IgG when using mouse primary antibodies) and/or a tertiary avidin (or Strepavidin) biotin complex ("ABC"). Examples of such reagents include a rabbit polyclonal antibody specific for pAKT and obtained from Cell Signaling Technology (Beverly, MA, Cat. No. 9277); anti-SCF antibody, obtained from Santa Cruz Biotechnology (Santa Cruz, CA, Cat. No. SC-9132); and a polyclonal anti-c-kit antibody obtained from Neomarkers, Inc. (Fremont, CA, Cat. No. RB-1518).

In practicing the method of the present invention, staining procedures can be carried out by a technician in the laboratory. Alternatively, the staining procedures can be carried out using automated systems. In either case, staining procedures for use according to the methods of this invention are performed according to standard techniques and protocols well-established in the art.

The amount of target protein can then be quantitated by the average optical density of the stained antigens. Also, the proportion or percentage of total tissue area stained may be readily calculated, as the area stained above an antibody threshold level in the second image. Following visualization of nuclei containing biomarkers, the percentage or amount of such cells in tissue derived from patients after treatment may be compared to the percentage or amount of such cells in untreated tissue or said tissue prior to treatment. For purposes of the invention herein, "determining"a pattern of expression and/or activation of a biomarker is understood broadly to mean merely obtaining gene expression and/or gene product activation information on such biomarkers.

Particularly useful embodiments of the present invention and the advantages thereof can be understood by referring to Examples 1-7. These Examples are illustrative of specific embodiments of the invention, and various uses thereof. They are set forth for explanatory purposes only, and are not to be taken as limiting the invention.

Example 1 Staining procedure for biomarkers Human tumor tissue samples were stained as follows. Tumor tissue in 10% Neutral Buffered Formalin Paraffin blocks were sectioned at 4 microns and the sections placed onto coated slides. Her-1 and Her-2 immunostaining was performed by using the pre-diluted Her-1 and Her-2 antibodies from Ventana Medical Instruments, Inc. (VMSI, Tucson, AZ.). Her-3 (1: 10) and Heregulin (1: 25) antibodies were obtained from NeoMarkers (Fremont, CA.). Her-1, Her-2, Her-3, and Heregulin were immunostained using the"BenchMark" (VMSI) with I-VIEW (VMSI) detection chemistry. p-ERK (1: 100) and p-AKT (1: 75) were obtained from Cell Signaling Technology (Beverly, MA), and immunostained using a labeled streptavidin peroxidase technique. Slides for p-ERK (1: 100) and p-AKT (1: 100) were antigen retrieved using 0.1 M citrate buffer, pH 6.0 in the"decloaker" (Biocare Corp. ) and the sections incubated overnight with the primaries at 4 °C. The next day, the slides for pERK and pAKT, were placed onto the Autostainer (Dako Corp. ) and the"LSAB2"kit (Dako) was employed as the detection chemistry.

DAB (Dako) was used as the chromogen. After immunostaining, all immunomarkers, Her-1, Her-2, Her-3, Heregulin, p-AKT, and p-ERK were counterstained manually with 4% ethyl green (Sigma).

Example 2 Procedure for Western Blot Analysis Western blot analyses to determine the relative amounts of ERB pathway-related gene products were performed as follows. Cells were lysed in ice-cold buffer (50 mM Tris-HCI (pH 7.5), 137 mM NaCI, 1mM EDTA, 1% Nonidet P-40,0. 2% Triton X-100, 10% glycerol, 0.1 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 20 mM 3- glycerophosphate, 50 mM NaF, 1 mM phenylmethylsulfonyl fluoride, 2, uM leupeptin, and 2, ug/ml aprotinin). Total protein concentration was determined with a BioRad Protein Assay Kit (BioRad Laboratories, Hercules, CA). Equal amounts of protein, such as 15 jU. g protein per lane, were separated by gel electrophoresis, such as pre-cast 4-12% Bis-Tris NuPage gradient gels (Invitrogen) or 7.5% or 4-15% gradient SDS-PAGE under reducing conditions, and transferred to membranes, such as HyBond-C nitrocellulose (Amersham Life Science) or Immobilon-P membranes. Membranes were blocked and then incubated with primary antibodies, such as antibodies against p-AKT and p-ERK (Cell Signaling Technology) overnight at 4°C in Tris-buffered saline containing 3% bovine serum albumin/0. 1% Tween 20. Detection of signal was by means of chemiluminescence (PerkinElmer Life Sciences), or by SuperSignal West Femto Maximum sensitivity substrate kit from Pierce (Rockford, IL) as described (Xia, W., et al., Oncogene 21,6255-6263 (2002) ).

Antibodies utilized for Western blotting include anti-phosphotyrosine purchased from Sigma Chemical (St. Louis, MO); anti-EGFR (Ab-12) and anti-erbB-2 (Ab-11) from NeoMarkers; antibodies to p-AKT (Ser 437), p-Erkl/2, Erkl/2 purchased from Santa Cruz Biotechnology; and anti-cyclin D 1 and 2 from Upstate (Lake Placid, NY).

Example 3 Procedure for Immunohistochemistry Quantitative immunohistochemistry (IHC) was performed as previously described (Bacus, S., et al., Analyt. Quant. Cytol. Histol. 19,316-328 (1997) ). EGFR, erbB-2, and cyclin D1 immunostaining was performed using the pre-diluted EGFR, erbB-2, and cyclin D1 antibodies from Ventana on the VMSI automated"BenchMark"staining module as described. The VMSI"I-View"detection kit was used for all three of the VMSI pre-diluted primary antibodies. Erkl/2 (1: 1200), erbB-3 (1: 10), Heregulin (1: 25), and TGFa (1: 20), were also immunostained using the"BenchMark"with I-VIEW

detection chemistry. Phospho-Erkl/2 (1: 100) and p-AKT (1: 75) were immunostained using a labeled streptavidin peroxidase technique. Phospho-Erkl/2 and p-AKT slides were antigen retrieved as described (Bacus, S., et al., Analyt. Quant. Cytol. Histol. 19, 316-328 (1997) ). Slides were placed onto the Autostainer (Dako Corp. ) and the "LSAB2"kit (Dako) employed as the detection chemistry. p-EGFR (1: 500) and p-erbB2 (1: 40) were immunostained in a similar labeled streptavidin peroxidase technique. p- EGFR slides were antigen retrieved with 1 mM EDTA and slides for p-erbB-2 with 0. 1M citrate buffer, pH 6 : 0, in the"decloaker". After staining, EGFR, erbB-2, erbB-3, heregulin, Erkl/2, p-AKT, p-Erkl/2, p-EGFR, p-erbB-2, cyclin D1, IGFR-1, and TGFa were counterstained manually with 4% ethyl green (Sigma). TUNEL assay (Roche Diagnostics, Indianapolis) was performed according to the manufacturer's instructions. Investigators preparing and analyzing tissue sections were blinded to both patient tumor type and response to therapy.

For IHC, antibodies to EGFR, erbB-2, and cyclin D antibodies were from Ventana Medical Scientific Instruments (VMSI) (Tucson, AZ); anti-p-AKT (Ser 437) and p- Erkl/2 were from Cell Signaling Technology Inc. (Beverly, MA); anti p-EGFR and p- erbB2 were from Chemicon (Temecula, CA) and NeoMarkers (Fremont, CA), respectively; antibodies to TGFa, erbB3, heregulin, IGFR-1, and phospho-erbB2 were from NeoMarkers; anti-Erkl/2 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA).

Example 4 Response to GW572016 in tumor cell lines and xenografts The HER1/HER2 combination inhibitor GW572016 (obtained from Glaxo Smith- Kline) reversibly inhibits both EGFR and erbB-2 and also inhibits steady state protein levels of pErkl/2, pAKT, and cyclin D in erbB-2- (BT474) and EGFR- (HN5) overexpressing tumor cell lines. The erbB-2 overexpressing human breast adenocarcinoma line, BT474, was obtained from the American Type Culture Collection (Rockville, MD) and the EGFR-expressing LICR-LON-HN5 head and neck carcinoma cell line (HN5) was kindly provided by Helmout Modjtahedi at the Institute of Cancer Research, Surrey, UK. Cells were cultured as previously described (Xia, W., et al., Oncogene 21,6255-6263 (2002) ). The results of Western blot analysis of GW572016- treated HN5 and BT474 tumor cells lines is shown in Figure 1A. HN5 and BT474 cells were co-cultured for 24 hr with 5 and 1 M GW572016, respectively. Equal amounts of

protein were then separated by SDS-PAGE and steady-state protein levels of both total and phosphorylated forms of EGFR, erbB-2, Erkl/2, and AKT assessed by Western-blot. Cyclin D1 and 2 steady state protein was also analyzed. Actin served as a control for equal loading of protein. Cells treated with vehicle (DMSO) alone served as a control.

These results demonstrated that inhibition of pERKI/2 and pAKT correlated with inhibition of pEGFR and p-erb2, respectively.

Given the limited amount of tissue obtained from sequential biopsies and the heterogeneous nature of those biopsies, quantitative immunohistochemistry (qIHC) was selected for application in clinical samples. Quantitative IHC offers advantages over Western-blot by enabling direct visualization of the biological effects of therapy on tumor cells interspersed amongst surrounding fibrotic tissue, normal cells, and stroma using limited amounts of tissue. Inhibitory effects similar to that observed by Western-blot were demonstrated by qIHC in tumor biopsies from mice bearing established BT474 tumor xenografts that had been treated with either GW572016 or vehicle alone. Figure 1B shows the results of quantitative IHC analysis of p-erbB2, p-EGFR, p-Erkl/2, p-AKT, and cyclin D1 from BT474 xenografts obtained from tumor-bearing C. B-17 SCID mice treated with either the duel EGFR/erbB-2 kinase inhibitor (100 mg/kg) or vehicle alone for 5 doses. The optical density (OD) values assigned to the specimen are shown. OD values of <10, 10-15, >15 roughly correlate with 1,2, 3+ in the HercepTest standards, respectively. These results demonstrated that inhibition of MAPK-Erk and/or PI3K-AKT pathways in erbB-2-or EGFR-overexpressing tumor cell lines or xenografts correlated with GW572016's inhibition of p-EGFR and erbB2, and anti-tumor activity (Xia, W., et al., Oncogene 21,6255-6263 (2002); Rusnak, D. W., et al., Mol. Cancer Therap. 1,85-94 (2001) ).

Example 5 Response to GW572016 in cancer patients After obtaining informed consent, patients were enrolled in an open-labeled, randomized trial. GW572016 (in dosages of 500,650, 900,1200, or 1600 mg) was administered orally once a day. Patients whose tumors (i) overexpressed EGFR or erbB-2 (2-3+ IHC staining in >10% of tumor cells), (ii) demonstrated erbB-2 gene amplification by FISH, or (iii) expressed activated, p-EGFR or erbB-2 by IHC (2-3+ staining in >10% of tumor cells) were enrolled if they satisfied other eligibility criteria. Tumor biopsies were obtained immediately prior to initiating the duel EGFR/erbB-2 kinase inhibitor therapy and again on day 21, within 12 hr of receiving drug. Biopsies were fixed in 10%

Neutral Buffered Formalin (NBF) and paraffin-embedded sections prepared. Patients were monitored by physical exams, clinical chemistry and hematology blood tests, and formally re-staging with imaging modalities after 8 weeks of therapy. RECIST criteria were used to assess clinical response (Therasse, P., et al., J. Natl. Cancer Inst. 92,205-216 (2000)).

The pEGFR and perbB-2 antibodies used for IHC lacked sufficient specificity to detect changes in phospho-EGFR/erbB-2 status. Therefore, analysis focused on downstream molecules mediating the growth and survival effects of EGFR and/or erbB-2, i. e. , pErkl/2, pAKT, and cyclin D, which reproducibly correlate with GW572016 inhibition of EGFR and erbB-2 and anti-tumor activity (Xia, W., et al., Oncogene 21,6255- 6263 (2002); Rusnak, D. W., et al., Mol. Cancer Therap. 1,85-94 (2001) ).

Two patients (361,372) achieved partial remissions following therapy (Table 1).

Both have metastatic breast cancer that was resistant to HERCEPTIN and multiple prior chemotherapy regimens. Figure 2A represents a sample from patient 361 stained for EGFR and erbB-2. This analysis of EGFR and erbB-2 immunoreactivity in a metastatic breast cancer nodule from patient 361 showed overexpression of both receptors. Pre- (d 0) and post-treatment (d 21) biopsies show that expression of total EGFR/erbB-2 protein is unchanged in response to GW572016. Patient 372 overexpressed erbB-2 and erbB-3 (Table 1).

Preclinically, inhibition of pAKT was associated with apoptosis of erbB-2 overexpressing tumor cell lines (Xia, W., et al., Oncogene 21,6255-6263 (2002); Rusnak, D. W., et al., Mol. Cancer Therap. 1,85-94 (2001) ). GW572016 led to the inhibition of pAKT. Figure 2B represents a sample from patient 372 stained for pAKT. Reduced expression of intra-tumor pAKT was observed by d 21 of therapy in this woman (patient 372) with metastatic breast cancer overexpressing erbB-2 and erbB-3. Activated pAKT expression was also inhibited in d 21 tumor biopsies from patients 361 (Table 1). In addition, patient 369 with head and neck carcinoma overexpressing EGFR and erbB-2 demonstrated 70% inhibition of pAKT following therapy, and experienced disease stabilization (Table 1).

The effect of pAKT inhibition on tumor cell survival using TUNEL assay as an indicator of apoptosis was studied as a pro-survival factor (Vivanco, I. , & Sawyers, C. L., Nature Reviews/Cancer 2,489-501 (2002); Tzahar, E., et al., Mol. Cell. Biol. 16,5276-5287 (1996); Thakkar, H., et al., Oncogene 20,6073-6083 (2001); Brognard, et al., Cancer Res. 61,3986-3997 (2001); Cheng, J., et al., Proc. Natl. Acad. Sci. USA 89,9267-9271

(1992) ). Increased tumor cell apoptosis (TUNEL +) was observed in d 21 biopsies from the three patients who also responded to therapy with decreased intra-tumor pAKT.

Figure 2C and 2D represent the results of inhibition of intra-tumor pAKT. Patients 372 and 369 had favorable clinical responses to therapy, with 372 achieving a partial remission and 369 experiencing disease stabilization. Day 21 tumor biopsies from both patients revealed a 9-to 10-fold increase in tumor cell apoptosis as quantified by TUNEL assay. Figure 2D shows that patient 361 also achieved a partial remission in response to GW572016 therapy. TUNEL staining reveals increased tumor cell apoptosis. Results of TUNEL assay are shown using 10X and 40X magnification. The percentage of TUNEL positive cells observed prior to (d 0), and at d 21 of therapy are indicated.

Example 6 Effect of GW572016 on MAPK/Erk signaling and cyclin D in tumors In preclinical studies, inhibition of pErkl/2 consistently correlated with GW572016-mediated inhibition of pEGFR/erbB-2 and tumor cytostasis (Rusnak, D. W., et al., Mol. Cancer Therap. 1,85-94 (2001) ). To quantify Erkl/2 activation in clinic tumor specimens, a p-Erk index was calculated for each biopsy. For example, Figure 3A represents staining tumor specimens for pERKl/2. The pErk index in patient 361 was markedly elevated (4015) at baseline (Table 1 and Figure 3A). By d 21, it was reduced to 0, with intra-tumor p-Erkl/2 immunoreactivity completely inhibited (Fig. 3A). In addition, patient 364 had the markedly elevated pERK index of 1634 prior to therapy (Table 1 and Figure 3A). pErkl/2 was inhibited 92% in patient 364 after treatment with GW572016. In both cases, p-Erkl/2 inhibition was associated with a favorable clinical response to therapy.

Activated Erkl/2 resides in the nucleus, where it regulates gene transcription promoting tumor growth and survival (Albanell, J. , et al., Cancer Res. 61,6500-6510 (2001); Kharitonenkov, A., et al. ; Nature 386,181-186 (1997) ). Consistent with inhibition of pErkl/2, total Erkl/2 cellular distribution shifted following therapy. Figure 3B represents staining tumor specimens for pERKl/2. Sequential tumor biopsies from a metastatic nodule (patient 361) were analyzed for total Erkl/2 protein expression prior to (d 0) and after 21 days of therapy (d 21). Whereas Erkl/2 was intra-nuclear prior to therapy, it was exclusively cytoplasmic at d 21. Two fields representative of the entire biopsy are shown.

The shift in the intra-cellular localization of total Erkl/2 was consistent with the inhibition of p-Erkl/2, as seen in Figure 3A.

GW572016-treated tumor cell lines undergo G1 cell cycle arrest (Rusnak, D. W., et al. , Mol. Cancer Therap. 1,85-94 (2001) ). As shown in Figures 1A and 1B, growth arrest was associated with inhibition of cyclin D protein, a regulator of the G1/S phase transition (Resnitzky, D. , & Reed, S. I. , Mol. Cell. Biol. 15,3463-3469 (1995); Lukas, J., et al., Mol. Cell. Biol. 16,6917-1925 (1996) ). Deregulation of cyclin D has been implicated in the pathogenesis of breast cancer, particularly tumors overexpressing erbB-2 (Lee, R. I., et al., Mol. Cell. Biol. 20,672-683 (2000); Zwijsen, R. M. L., et al. Mol. Cell. Biol. 16,2554-2560 (1996); Weinstat-Saslow, D., et al. ; Nat. Med. 1, 1257-1260 (1995); Sicinski, P., et al. ; Cell 82, 621-630 (1995); Musgrove, E. A., et al. ; Proc. Natl. Acad. Sci. USA 91,8022-8026 (1994) ).

Cyclin D protein was inhibited in several patients treated with GW572016 (Table 1).

Figure 4 represents staining of sequential tumor biopsies for cyclin D. Inhibition of cyclin Dl is shown in sequential tumor biopsies obtained prior to (d 0) and after 21 days of therapy with GW572016 from patients 361 and 364.

In contrast, lack of inhibition of p-AKT, p-Erkl/2, or cyclin D was generally associated with progressive disease (Table 1).

Example 7 Biomarkers that influence response to GW572016 Expression of TGF-a and heregulin, ligands for EGFR and erbB-3/erbB-4, respectively, (Reese, D. J. I., & Stern, D. F.; Bioassays 20,41-48 (1998) ) was examined in biopsies to determine if the presence of these ligands influenced biological and clinical response to therapy. Baseline elevated TGF-a and heregulin immunoreactivity was increased in two patients achieving partial response (PR). In some cases, therapy appeared to reduce tumor immunoreactivity of TGF-a and/or heregulin, which occurred more frequently in responding patients, otherwise there was no apparent correlation with response (Table 1).

The presence of insulin-like growth factor receptor (IGFR-1) has been implicated in mediating resistance to EGFR inhibitors and Herceptin. (Chakravarti, A. , et al., Cancer Res. 62,200-207 (2002); Lu, Y., et al. ; J. Natl. Cancer Inst. 93: 1852-1857 (2001) ) Patient 371 is a woman with metastatic breast cancer overexpressing EGFR and erbB2 who had a "beneficial"biological response to therapy, yet still had disease progression (Table 1).

Phospho-Erkl/2 was completely inhibited following therapy, and cyclin D protein inhibited by 75% (Table 1). However, intra-tumor p-AKT was unaffected, nor was there an indication of tumor cell apoptosis in response to therapy (data not shown). The tumor

expressed high levels of IGFR-1, suggesting that AKT activation and therefore tumor survival, may have been dependent upon IGFR-1 signaling rather than EGFR/erbB-2 (data not shown).

It should be understood that the foregoing disclosure emphasizes certain specific embodiments of the invention and that all modifications or alternatives equivalent thereto are within the spirit and scope of the invention as set forth in the appended claims.

5 Table 1. Summary of intra-tumor biological effects of GW572016 therapy according to clinical response Patient and Tumor Type . o. o. a ." s,"r w L 1 WI I W W u dl E- S -o H <'B u a o. 1 & ? M Patients achieving a partial remission at 8 weeks 361 Breast 1200 d 0 20 43 4015 36 31 0 35 20 d ll 19 41 0 24 3 25 16 372 Breast 1200 _ 7 sr 378 48 20 6Y 38 10 d 21 2 41 10 30 4 25 6 Patients with stable disease at 8 weeks 364 *H & N 1600 d 0 23 11 1634 51 66 4 59 15 d 21 23 5 100 43 33 20 8 369 H&N 1200 d 0 26 25 0 24 42 0 49 7 d 2l 35 23 0 7 10 23 6 367 **A. U. P 650 d 0 17 3 230 35 46 0 16 0 d 21 1D _ 0 25 39 10 366 Ovarian 900 d 0 8 2 110 22 0 2 16 0 d 2/10 5 25 47 31 Patients with progressive disease at 8 weeks 362 A. U. P 900 d 0 42 10 576 61 26 0 17 0 d21 32 12 1260 81 37 18 371 Breast 900 d 0 14 44 1081 36 42 57 49 7 d 21 19 32 0 33 10 23 6 363 Sarcoma 500 d 0 10 0 20 19 11 0 13 0 d 21 6 336 32 12 44

*H & N: head and neck ; **A. U. P. adenocarcinoma of unknown primary # TUNEL positive indicates the percentage of Positive staining cells # OD values of <10. 10-15, and >15 roughly correspond to HercepTest standards of 1+, 2+, 3+, respectively.