WO2002077176A2

WO2002077176A2 - Aberrantly expressed proteins in laser capture microdissected tumors

Info

Publication number: WO2002077176A2
Application number: PCT/US2002/008847
Authority: WO
Inventors: Emanuel Frank Petricoin, Iii; Lance A. Liotta; Michael R. Emmert-Buck; Yingming Zhao
Original assignee: The Government Of The United States Of America As Represented By The Secretary, Department Of Health And Human Services; Mt. Sinai School Of Medicine
Priority date: 2001-03-22
Filing date: 2002-03-22
Publication date: 2002-10-03
Also published as: WO2002077176A3; AU2002255880A1

Abstract

Provided herein are specific examples of tumor or tumor-stage linked protein fingerprints and specific proteins identified from these fingerprints as being aberrantly expressed in specific diseased cell types. Methods of using these fingerprints, sub-sets thereof, and individual proteins in the diagnosis, prognosis, treatment, treatment selection, and drug development for disease are also provided.

Description

ABERRANTLY EXPRESSED PROTEINS IN LASER CAPTURE MICRODISSECTED TUMORS

FIELD The present invention relates to proteins that are differentially expressed in diseased cells, in comparison to cells from the same tissue that is not suffering from that disease.

BACKGROUND

Proteomics is the study of variations in cellular protein levels between normal and disease states. As the field of molecular biology moves beyond genomics to proteomics, there is a growing need for a direct method to monitor the levels of expressed proteins in developing, diseased or genetically altered tissues. Direct monitoring of tissues is difficult because of their heterologous, three-dimensional structure. This structure is the result of the strong adhesive interaction of the various cell types with adjacent cells, connective stroma, blood vessels, glandular and muscle components, adipose cells, and inflammatory or immune cells. The amount and type of protein expressed by cells in such a native tissue environment may be quite different from that of more easily studied cultured or transplanted cells. This consideration requires a direct means of measuring protein levels to obtain results reflecting in vivo conditions.

Prior methods for extracting and analyzing biomolecules from tissue subpopulations were complicated, labor intensive, and did not utilize protein stabilizers. Because of these limitations, a number of direct comparisons between tumor cells and normal cells have not been achievable. There has been no way to directly compare, without the danger of cross-contamination, the spectrum of proteins contained in normal cells with the proteins in tumor cells in a single tissue. Many hypotheses about altered protein levels in tumor cells have been based on work on cells lines, and the ability for continued growth in culture by the cell line adds yet another variable into the results. There also has been no way to directly quantify differences in protein amount between tumor and normal tissue. Much of the work in this area has been done using immunohistochemistry, imparting limitations as to quantification and specificity. Moreover, there has been no way to compare the protein content of various stages of tumor development, or to compare the characteristics of proteins found in tumor cells to those in normal cells. Nor could analysis of a protein's amino acid sequence or binding characteristics be performed. Because such comparative protein studies could not be done, it was difficult to reliably determine the source of a tumor metastasis, unless the probable tumor source was already known.

Proteomic research has been complicated by the fact that cells undergoing the progression to a diseased state may constitute less than 5% of the volume of a tissue biopsy sample. In a mixture of diseased and normal cells, differential protein expression in the diseased cells may be masked by the "background" of proteins expressed in normal cells. Immunohistochemistry cannot provide a solution to this problem, because it can provide only the location of the targeted antigen. This technique is ineffective at providing quantitative data because it is not possible to calibrate the intensity of immunohistochemical staining with the actual number of antigen molecules in the stained tissue cells. Additionally, many antibodies cannot differentiate between pro-enzyme and active enzyme species. Immunologically based studies cannot easily provide a complete picture ofa cell's protein contents.

SUMMARY OF THE DISCLOSURE

The present disclosure describes devices and methods for extracting proteins from samples of microdissected cells, and applying various analytic processes to the extracted proteins, such as immunoassays, ID and 2D gel electrophoresis characterization, Western blotting, Matrix Assisted Laser Desorption Ionization/Time of Flight (MALDI/TOF), liquid chromatography quadrapole ion trap electrospray (LCQ-MS), and Surface Enhanced Laser Desorption Ionization Spectroscopy (SELDI). These methods allow for direct comparison of qualitative and quantitative protein content of tumor cells and normal cells from the same tissue sample. The methods also allow for investigation of protein characteristics of tumor cells, such as binding ability and amino acid sequence. The present methods also provide, through the use of protein fingerprinting, a rapid and reliable way to identify the source tissue of a tumor metastasis.

In this method, the protein content of a selected population of cells from a tissue sample is analyzed, by extracting the population of cells from the tissue sample using laser capture microdissection (LCM). The population of cells that is extracted can be, for example, cells ofa particular cellular substructure (such as cells from epithelium on the lumen of an organ, or pockets of cells that have undergone malignant transformation against a background ofa larger population of more normal cells). Proteins are isolated from the isolated population of cells, and characteristics of the proteins may be analyzed to provide information about the protein characteristics of the selected isolated population of cells.

In particularly disclosed examples, differential expression of proteins in isolated malignant cells can be used to study changing patterns of protein expression during malignant transformation. In other examples, cells at different stages of biological transformation (such as neoplastic progression from normal cells to metaplastic cells to invasive carcinoma) can be studied to analyze differential protein expression at different stages of neoplastic progression. Similar methods can be used to analyze cells for therapeutic purposes (such as selecting drugs targeted against expression of particular proteins), or in drug response assays (to assess changes in protein expression as an indication of drug response).

Also provided are proteins identified as being differentially (e.g., aberrantly) expressed in diseased versus normal tissues, and methods of using the identification of these proteins in disease diagnosis, prognosis, disease treatment, treatment selection, and drug development. BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a relative size comparison of the laser beam diameter and the epithelial tissue sample being microdissected. Laser Capture Microdissection (LCM) was performed to selectively transfer only the epithelial lining of the prostate gland to the polymer film. The 30 micron laser beam spot size is shown in relation to the thickness of the gland. EP: epithelial, LU: lumen, stroma are as marked.

Figure 2 shows how cells of interest are selected using LCM and transferred onto a cap, where the proteins are solubilized directly into a tube.

Figures 3 A and B are graphic representations of the number of molecules of PSA as compared to the number of laser shots used to harvest the cells. Figure 3 A shows the results of a 500 shot sample of tumor cells serially diluted to verify sensitivity. RLU values were converted to PSA molecule numbers. Linear regression indicated an r value of 0.998. Figure 3B reports the results of a linearity analysis conducted on six different tissue sections from the same patients. The number of laser shots are indicated, r = 0.95. Figure 4 is a table showing the number of PSA molecules, the coefficient of variation of the process, and the mean number of cells used in the immunoassays. These sensitivity and precision analyses were done on 10 separated microdissection replicates. The mean readout in RLUs is reported as well as the corresponding number of PSA molecules. The sensitivity, defined as two standard deviations above background, was one laser shot. Imprecision (%CV) was inversely correlated with number of shots per specimen.

Figure 5 is a table comparing the number of PSA molecules with the immunohistochemistry score. Independent immunohistochemistry scoring of coded specimens was done by a pathologist and scored as high (+++), medium (++), and low (+). PIN: prostate intraepithelial neoplasia, Tumor: invasive carcinoma, Normal: histologically normal prostate glands. The relative concentrations of PSA per cell among different progression stages varies greatly and this corresponds directly to the semiquantitative immunohistochemistry scoring.

Figures 6A, B, and C are 2-D or 1-D gel comparisons of the proteins in microdissected normal and tumor epithelium. Fifty thousand cells were procured by LCM, directly lysed in IEF buffer and run on a 3-10 NL Pharmacia IPG IEF strip for 100 kVhr. The second dimension runs were performed on 8-18% linear gradient SDS-PAGE gels and the gels were stained with silver. Figures 6A and 6B show match tumor and normal fingerprints for each patient. A representative pi and molecular weight ruler for direct comparison and alignment is shown in panel A. Figure 6C shows the α-tubulin immunoblot that was used to normalize for relative protein load.

Figure 7 is a 2-D gel comparison of microdissected normal and tumor epithelium compared to a Western blot using an anti-pan-cytokeratin type II antibody. The gels were run as described for Figure 6. The completed gel was transferred to PVDF membrane and western blotted with a 1 : 1000 dilution of anti-pan type II cytokeratin. Figure 8 is a 2-D gel comparison of microdissected normal, tumor epithelium, and stroma tissue indicating the differences by circling the altered proteins. The gels were prepared and run as described in connection with Figure 6. Figure 9 is a 2-D gel comparison of microdissected normal and tumor epithelium as compared to whole tissue cryostat where circles indicate the locations of the "altered" proteins. The gels were prepared and run as described in Figure 6.

Figure 10 is a ID-gel anti-PSA Western blot of cell lysates from microdissected normal and tumor prostatectomy specimens. Lanes 1 & 3 are benign epithelium and lanes 2 & 4 are from malignant epithelium.

Figures 11 A and B are anti-PSA Western blots, where lanes 1-4 are benign epithelium and lanes 5-8 are malignant. Lanes 1 & 5 are untreated, lanes 2 & 6 are with ACT added, and lanes 3 & 7 are incubation controls, while lanes 4 & 8 are ACT added with 120 minutes of incubation at 37° C.

Figures 12A and B are anti-PSA Western blots showing complexed and non-complexed species.

Figures 13 A, B, and C show data that indicate SELDI protein profiles of LCM-derived cellular lysates are reproducible and sensitive. Figure 13A shows two separate microdissections of prostate tumor epithelium form the same tissue section from the same patient (1200) cells each) which were analyzed by SELDI protein fingerprinting. The raw mass spectroscopic mass map is shown for each microdissection along with the Gel- View^® display from the same data set. Figure

13B shows two separate microdissections of prostate tumor epithelium from a tissue section from two different patients (1200 cells each) which were analyzed by SELDI protein fingerprinting. The raw mass spectroscopic mass map is shown for each microdissection along with the Gel- View^® display from the same data set. Figure 13C shows four separate microdissections of decreasing number of cells that were analyzed by SELDI protein fingerprinting. The Gel- View^® display is shown as a representation of the direct alignment of each of these four mass spectra to each other.

Figures 14A, B, and C show that SELDI protein profiles of LCM-derived cellular lysates are discriminatory between different tumor epithelial cell types from different patients and between tumor and normal epithelial cells from the same patient. In Figures 14A and 14B LCM derived tumor epithelial cells (1200 cells) from prostate, breast, and colon frozen tissue sections were separately acquired and analyzed via SELDI. The Gel-View^® representation is shown in Figure 14A and the spectrographic mass profile is shown in panel B. Figure 14C shows the SELDI analysis of four separate patient-matched microdissections of 1200 cells of colon normal epithelium, colon tumor epithelium, colon tumor epithelium from the colon tumor that has metastasized to the liver, and normal liver cells next to the metastasis. The Gel- View^® display is shown as a representation of the direct alignment of each of these four mass spectra to each other.

Figure 15 is a SELDI fingerprint comparison between a variety of tumor types. Figure 16 shows a device developed for solubilizing proteins in cells taken from a biological specimen by LCM.

Figures 17A and 17B are SELDI analyses of microdissected esophageal epithelium showing proteins disregulated in a disease-specific manner. Figures 18A and 18B are SELDI analyses of 8 different esophageal cancer cases, where three separate microdissections of eight different patients' matched tumor and normal cells were subjected to SELDI analysis via the use of a hydrophobic interaction C18 binding surface. Each replicate was run in triplicate, giving a total of 72 data points for each protein peak analyzed. The analysis of the protein fingerprint in the low mass region is shown in Figure 17A, the higher mass region in Figure 17B. A representative mass map from one case (case #1) is shown on the left side of each panel with the normal and tumor fingerprint shown (top and bottom, respectively) for each mass region. A gel-like representation is displayed for that particular case as well as the fingerprint for two other cases. Proteins 1, 2, 6, and 7 are labeled for orientation. All cases analyzed in the study set were then subjected to analysis as a ratio of relative intensity of the selected proteins to one another and the statistical results shown on the right side of each figure. Figure 17A: average C.V. = 12.7%, 29.4% for the normal and tumor microdissections respectively. Figure 17B: average C.V. = 10.5% and 18.9% for the normal and tumor microdissections, respectively.

Figure 18 shows the SELDI analysis of prostate carcinogenesis. Figure 18A shows a mass map that represents the profile from 1500 normal, pre-invasive neoplasia (PIN) and invasive carcinoma cells acquired by LCM from one case (case #2). Additionally, the corresponding patient- matched stromal cells (1500 cells) were also microdissected for analysis. Figure 18B shows a gellike image of the raw mass data shown in Panael A. All samples from this patient were run in triplicate, with the representation of one experiment shown. Two proteins, A and B, having molecular weights of 28,000 and 32,000 respectively, were found to be reproducibly differentially expressed in this patient and are indicated in both Figures 18A and 18B. Figure 18C shows the ratio analysis of A vs. B from an additional study set of two other (cases #1 and #3) patient matched tumor and normal prostatic epithelium (average C.V> = 17.2 % and 10.1% for the tumor and normal microdissections, respectively).

Figure 19 is a diagram of the multiplexed tissue array method for a high throughput target validation and drug tissue interaction assay, using microdissected human breast cells in various stages of malignancy as an example biological sample.

Figures 20A, 20B and 20C are representative results using colorimetric and chemiluminescent detection methods for prostate soluble antigen (PSA). The rows of the upper arrays are in all three cases (1) a protein standard, (2) prostate stroma, and (3) normal prostate tissue, while the lower arrays include (4) prostate intraepithelial neoplasia (PIN), (5) tumor tissue, and (6) invasive tumor tissue. The amount of total protein loaded on the gel is reduced across the rows. Figure 20A shows the colorimetric results, Figure 20B is the positive and Figure 20C is the negative image of the fluorescence results. Figures 21 A, 2 IB, and 21C diagram the reproducibility of protein analysis for samples microdissected from a mixed sample of epithelial cells of the esophagus. All the following data was obtained from normal cells present in the sample. Figure 21 A shows the reproducibility of the annexin I protein data in normal cells over a variation of "shot" size within one slide and between multiple slide sets. Figure 2 IB shows a histogram of the coefficient of variance for these data sets. Figure 21C graphically shows reproducibility of the total protein obtained from the cells as shown by fluorescence detection.

Figure 22 shows the results of interaction between immobilized binding agents and biotin- labeled lysates from microdissected normal and tumor cells.

DETAILED DESCRIPTION

Laser Capture Microdissection (LCM) is a recently developed technology that enables the user to obtain pure cell populations from stained heterogeneous tissue under direct, high power microscopic visualization. See Emmert-Buck et al, 274 Science 998 (1996); Bonner et al, 278 Science 1481 (1997). A laser beam focally activates a special transfer film that specifically bonds to cells identified and targeted by microscopy within the tissue section. The transfer film with the bonded cells is then lifted off the thin tissue section, leaving all unwanted cells behind. This process allows the collected sample to include only cells exhibiting the desired morphology for study. This application of LCM to the analysis of the DNA of isolated cells is described in U.S. Patent No. 5,843,657, issued to Liotta et al. , which is incorporated herein by reference.

Briefly, the LCM process works as follows. First, a transparent transfer film is applied to the surface of the tissue section. Under the microscope, the operator views the thin tissue section through the glass slide on which it is mounted and chooses microscopic clusters of cells to study. When the cells of choice are in the center of the field of view, the operator pushes a button that activates a near-IF laser diode integrated with the microscope optics. The pulsed laser beam activates a precise spot on the transfer film immediately above the cells of interest. At this precise location the film melts and fuses with the underlying cells of choice. When the film is removed, the chosen cell(s) are tightly held within the focally expanded polymer, while the rest of the tissue is left behind. The exact morphology of the procured cells is retained and held on the transfer film, ensuring preservation of the cells' intracellular components such as DNA, RNA, and proteins for future analysis. The removed transfer film and cells are transferred onto a plastic cap (referred to as the LCM cap) for subsequent analysis.

Isolation Isolation of these preserved intracellular components from the cells on the LCM cap has necessitated the development of new procedures, buffers, and devices. In particular, the present methods utilize a new procedure for extracting proteins from very small sample sizes, on the order of about 1500 to about 5 cells, the number obtained in a typical LCM laser shot. This procedure generally involves the extraction of proteins in one solubilizing step, using a very small volume ofa unique buffer. To facilitate the handling of such small volumes and small sample sizes, a device in which the solubihzation step can be conveniently performed was developed. The results of this new procedure are intact proteins, substantially free of cross-contamination from other nontumor or normal cell types. The isolated proteins maintain activity, allowing analysis through any number of immunological and biochemical assays.

The buffers for the protein isolation step can include one or more of buffer components, salt(s), detergents, protease inhibitors, and phosphatase inhibitors. In particular, one effective buffer for extracting proteins to be analyzed by immunohistochemistry includes the buffer Tris-HCl, NaCl, the detergents Nonidet® P-40, EDTA, and sodium pyrophosphate, the protease inhibitors aprotinin and leupeptin, and the phosphatase inhibitors sodium deoxycholate, sodium orthovanadate, and 4-2- aminoethyl benzenesulfonylfluroride (AEBSF). Another salt that could be used is LiCl, while glycerol is a suitable emulsifying agent that can be added to the fraction buffer. Additional optional protease inhibitors include soybean trypsin inhibitor and pepstatin. Other suitable phosphatase inhibitors include phenylmethylsufonyl fluoride, sodium molybdate, sodium fluoride, and beta- glycerol phosphate.

For 2-D gel analysis, simple lysis with a 1% SDS solution was effective, while ultimate analysis using the SELDI process required Triton-X-100, a detergent (Sigma, St. Louis, MO), MEGA 10® (ICN, Aurora, OH), and octyl B-glucopyranoside (ESA, Chelmsford, MA) in a standard PBS base. Another buffer which was used prior to 2-D gel analysis was 7M urea, 2M thiourea, CHAPS, MEGA 10, octyl B-glucopyranoside, Tris, DTT, tributyl phosphine, and Pharmalytes. A preferred 1:100 concentration buffer is as follows:

The buffer is made and diluted 1 : 100 in distilled water for use. The buffer must be kept frozen at - 20° C. It can only be used unfrozen for a few hours. In all cases, the buffer is used in very low volumes, from about 1 μl to about 15 μl, and is applied directly to the laser capture dissected cells while still on the LCM cap.

To simplify the process of solubilizing the proteins directly on the LCM cap, a device was developed to handle the small volumes. This device (shown in Fig. 16) includes a small chamber 1 where the solubilization process can occur. The chamber has an upper opening 2 structured such that it can directly accept the LCM cap with its adhered cells. The LCM cap is placed in the chamber 1 with the surface having cells upon it facing the interior of the chamber. Once the LCM cap is attached to the device, the cells are therefore strategically positioned within the chamber and solubilization of the proteins will occur upon introduction of the solubilization buffer. The chamber is supplied with at least one inlet port 3, which can be equipped with a syringe 4 as a means of introducing the small volume of the solubilizing buffer into the chamber 1. The inlet port 3 is connected to the chamber by a narrow inlet canal 5, where liquids placed into the canal will move toward the solubilization chamber by capillary action. The device can include more than one inlet chamber, if desired, each equipped with a means of introducing liquid, such as a syringe. The device also includes an outlet port 6, connected to the chamber 1 with a narrow outlet canal 7, which can be supplied with a means of collecting and removing the solubilized proteins. One embodiment of this device has the chamber 1, the inlet ports 3, inlet canals 5, outlet port 6, and outlet canal 7, all machined in a hard plastic block, such as lucite.

Analysis Once the proteins have been solubilized, a number of different immunological or biochemical analyses can be used to characterize the isolated proteins. Because the cells are usually initially identified for LCM capture based on particular morphological characteristics, the results of these assays are assured to be representative of the cell type collected.

One type of assay that can be performed is a soluble immunoassay, where an antibody specific for a protein of interest is used. The antibody can be labeled with a variety of markers, such as chemiluminescent, fluorescent, or radioactive markers. For best results, a high sensitivity assay can be used, such as a microparticle enzyme immunoassay (MEIA). By applying a calibration curve used to estimate immunodetected molecules in serum, the number of molecules per cell can be estimated. Thus, the presently described methods provide a quantitative immunoassay, which can measure the actual number of the protein molecules of interest in vivo.

A second type of assay that can be used to analyze the extracted proteins is two-dimensional polyacrylamide gel electrophoresis (2D-PAGE). By running both proteins extracted from normal cells of the sample and proteins extracted from tumor cells ofa sample, and comparing the blots, differential protein expression can be seen. In particular, by scanning the stained gels into a computer, and using image comparison software, the location of proteins that are present in one cell type and absent (or vice versa) in the other can be determined. Furthermore, these altered proteins can be isolated from the gel where they are present, and mass spectroscopy MS-MS sequencing can be used to identify the protein, if the sequence exists in a database. In this way, the protein differences between normal and tumor cells can be more fully understood. Additionally, proteins of interest isolated from a 2-D gel may be used in binding studies, where the protein is functionally tested for an alteration in its ability to bind with a putative or known ligand. The disclosed comparative analysis need not be between normal and tumor cells, but can be between isolated stages of a tumor, where the different stages exhibit sufficient morphological differences to allow separate isolation of populations using the LCM technique.

A further analysis that may be performed involves the use of the surface enhanced laser desorption ionization spectroscopy technique, or SELDI (Ciphergen Biosystems Inc., Palo Alto, CA). This process can separate proteins that would not be separately focused by 2-D gel analysis, in particular those proteins which are very basic, very small (<7000 Daltons) or are expressed at low or moderate levels in the cells. The lower level of expression becomes critical in experiments analyzing LCM derived spectrum because of the extremely small sample size of cells used. SELDI also separates proteins more rapidly than gel analysis. SELDI utilizes a "protein chip" that allows for desorption and detection of intact proteins at the femtomole levels from crude samples. Proteins of interest are directly applied to a defined small surface area of the protein chip formatted in 8 to 24 predetermined regions on an aluminum support. These surfaces are coated with defined chemical "bait" matrices comprised of standard chromatographic supports, such as hydrophobic, cationic, or anionic or biochemical bait molecules such as purified protein ligands, receptors, antibodies, or DNA oligonucleotides (see Strauss, Science 282:1406, 1998). In the case of LCM collected samples, the solubilized proteins are applied to the surface of the SELDI chip. Binding of the proteins to the surface is dependent on the nature of the bait surface and the wash conditions employed. The mixture of bound proteins is then characterized by laser desorption and ionization and subsequent time-of- flight (TOF) mass analysis generated from a sensitive molecular weight detector. These data produce a protein fingerprint for the sample, with SELDI having a practical resolution and detection working range of 1000 to 300,000 Daltons, depending on the energy-absorbing molecule utilized and the bait surface/wash conditions employed.

The produced protein "fingerprints" have proven to be both disease-specific and organ- specific. That is to say, the protein fingerprint of a tumor from a particular tissue type remains characteristic of that tissue type, whether it is normal, tumor, or a metastasis. This allows determination of the origin of an unknown metastasis to be made, through comparison to protein fingerprints of normal or tumor tissue of likely organ sources.

The principles of several embodiments are further illustrated by the following Examples.

EXAMPLE 1

Quantification of Intracellular PSA from Benign and Malignant Prostate Epithelium by Immunoassay

Case Materials

Tissue was obtained following an IRB approved protocol from both the Urologic Oncology Branch in the National Cancer Institute, Bethesda, MD and the Mayo Clinic in Rochester, MN. After surgery, the tissue samples were snap frozen in liquid nitrogen. The tissue supplies were then embedded in Optical Coherence Tomography (O.C.T.) compound (Tissue Tek, Miles, Elkhart, IN) and stored at -80°C. Cases were selected based on the histology present in the tissue sections, so that normal glands, Prostate Intraepithelial Neoplasia (PIN), and adjacent carcinoma, could be compared within the same patient. Prostate tissue cases were selected to include ample stroma to serve as a negative control. Lung tissue was used as a second negative control. Sectioning and Staining

The O.C.T. embedded tissue blocks were cut into 8 μm sections with a cryostat. After cutting, the sections were immediately placed on dry ice and then stored at -80°C. Only one section was thawed and dissected at a time, to minimize degradation of proteins. After fixation in 70% ethanol for 10 seconds, the section was stained with hematoxylin and eoxin, and dehydrated in xylene. Laser Capture Microdissection

The PixCell system incorporates an Olympus IX-50 Microscope containing a microscope slide stage that is moved by a joystick. The operator uses the joystick to position the tissue under a fixed laser beam that can be focused from 5 to 60 μm in diameter. The LCM transfer film is fixed to the undersurface of a vial cap (CapSure^® TF- 100 transfer film carrier, 5 mm dia. optical grade transparent plastic; matching vial is Brinkmann #22 36 430-8). LCM cell procurement time was always less than 15 minutes. A droplet of the extraction buffer was applied to the surface of the film containing the selected cells. The cap with the droplet on its undersurface was inserted into the mouth of the matching vial containing 50 μl of the immunoassay dilution buffer. The sealed vial sample receptacle was frozen at -20°C and stored for less than 48 hours prior to assay. The thawed sample was held at 4°C for no longer than 2.5 hours prior to introduction into the immunoassay module.

Figure 1 compares the size of a 30 μm laser shot with the size of an example prostate gland that was microdissected by movement of the joystick. The yield of procured pure epithelial or carcinoma cells, and the precision of cellular procurement, were tested by visually counting the number of cells removed from the tissue and transferred to the film. Table 1 compares the number of laser shots with the average total number of microdissected cells.

Table 1

A standard laser spot diameter of 30 μm encompasses 5-7 cells. Individual tissue cells can vary in their packing density, their shape, and their volume. Consequently, the imprecision of the cell yield is greater with fewer laser shots. Solubilization of proteins from LCM procured cells

Extraction of the proteins from the exposed, transferred cells was done by pipetting a small volume of buffer directly on the inverted capture surface of the polymer film (Figure 2). Following placement of 5 μl of extraction buffer onto the transfer surface containing the tissue cells, visual confirmation was obtained establishing the complete solubilization of the stained cells. The transfer cap with the extracted cellular protein was inserted into the mouth of a vial containing the immunoassay buffer. After analysis, the transfer caps were counterstained to ensure complete solubilization of the cells. The diluted extraction buffer did not interfere with the immunoassay background level or linear calibration. Extraction Buffer

The final buffer composition was a 1/1000 dilution of the following stock: 50 mM Tris HCl, 1% NP-40, 0. 1%> Na Deoxycholate , 150 mM NaCl, 4mM EDTA, Aprotinin (10 mg/ml), Leupeptin (lOmg/ml), Na Pyrophosphate (10 mM) Na orthovanate (2mM), and AEBSF (100 mM). Assay Principle The Immulite Third Generation PSA assay (Diagnostic Products Corp., Los Angeles, CA) was adapted to measure PSA solubilized from tissue cells procured by LCM. Photon production is measured by a luminometer (output: relative light units (RLUs)). Negative tissues or zero controls produced a value of 80-140,000 RLUs, compared to PSA positive samples that ranged from 600,000 to 10,000,000 RLUs. PSA immunoassay: Microparticle Enzyme Immunoassay (MEIA)

PSA was measured by an automated two site chemiluminescent assay, using the ultrasensitive PSA reagent kit on the Immulite immunoassay analyzer (Diagnostic Products Corp., Los Angeles, CA). RLUs reflect the photons detected by the photomultiplier tube, which is proportional to the concentration of PSA. The sensitivity of the assay is .004 ng/mL of PSA, or approximately 4 x 10⁶ molecules of PSA per assay. Immunohistochemistry

Immunohistochemistry staining for PSA was conducted as follows. The frozen sections were desiccated and then fixed in acetone. Following a wash with 3% goat serum, primary antibody incubation was 1 hour at room temperature. The secondary antibody was labeled by Avidin/Biotin and the substrate system was Peroxidase/DAB. The secondary reaction was performed essentially as suggested by the commercial product insert. Calibration Curve

The calibration curve used for measurement of PSA in serum according to the Immulite package insert was applied to the measurement of cellular tissue proteins procured by LCM. The calibration curve designed for serum was reproduced with the LCM extracted cellular proteins in the buffer solution. Serum or buffer samples with a known concentration of PSA were mixed with equal volumes of 100 laser shots of PSA negative tissue cells. There was approximately 100% recovery in all cases, with no detectable elevation of the zero dose or diminution of the signal. The number of laser shots of actual tissue cells was compared to the calculated number of PSA molecules for individual patient cases as shown in Figure 3. Linear regression analysis yielded an r value of greater than 0.95 over a dynamic range of 0.004 ng/mL to 1 ng/mL. Sensitivity The detection limit of the assay for microdissected cellular proteins was 0.004 ng/ml PSA as defined as the concentration two standard deviations above the signal response ofa sample free of PSA or a PSA negative tissue. The sensitivity achieved with these criteria was one 30 μm laser shot (5-7 cells). Imprecision The imprecision of cell capture for 100 shots is 3.3% C.V. compared to a C.V. of 15.8% for the PSA measured in the proteins extracted from 15 laser shots. In keeping with the assumed biologic heterogeneity, one laser shot exhibits a PSA measurement C.V. of 60.2%, even though the mean magnitude of the signal (804,416 RLUs) is substantially greater than the mean value for combined 100 shots of background tissue (132,759) (Figure 4). Thus, capturing 15 (analytical sensitivity threshold) to 100 laser shots of a tissue sample (this takes less than five minutes) provides an acceptable level of imprecision for a routine assay. Correlation with Tissue Immunohistochemistry

Analysis using LCM coupled with PSA immunoassays for PSA was conducted on prostate tissue sections that contained normal epithelium, carcinoma, and prostate intraepithelial neoplasia (PIN). Based on the imprecision analysis described above, the number of laser shots was 100. For each specimen studied, a standard curve was prepared for a series of different numbers of laser shots, to verify that conclusions based on 100 shots remained in the linear portion of the curve (Figure 3B). Immunohistochemistry for PSA was conducted on adjacent sections. The PSA immune-reactivity of cellular populations identical to those sampled by LCM on matched coded sections was scored independently and ranked on a scale of zero to five plus (Figure 5). The average number of molecules per cell ranged from IO⁴ to 10^δ. The immunohistochemistry scoring values paralleled the quantitation and mirrored the heterogeneity in PSA production by normal and neoplastic cell populations. For example, in case C the normal epithelium contained 6.3 x IO⁶ molecules and scored three plus by immunohistochemistry. In contrast, the PIN cells contained 3.7 x 10⁵ molecules and scored two plus, and the tumor cells contained 1.99 x IO⁴, and stained one plus.

Application of this new technology, as reported in Figure 5, provides quantitative confirmation of the heterogeneity in PSA expression, which was previously detected only by qualitative staining. The average numbers of PSA molecules harvested per cell ranged over several logs. An immunohistochemical staining difference could be discriminated only if there was greater than a ten-fold difference in PSA molecule number per cell.

The present technology provides one of the first direct estimates of the actual number of protein molecules per tissue cell in vivo for a single specific known protein of moderate to low abundance. The number of total PSA molecules in normal prostate epithelium cell ranged from IO⁴ to 10^s. PSA is an important serum analyte used to clinically monitor prostate cancer, but it is not a specific maker of prostate cancer. Previous investigators have reported great heterogeneity in the intensity of PSA immunohistochemical staining among various neoplastic and non-neoplastic populations in the prostate. Populations of microdissected cells of a pure histologic class would be expected to contain some level of heterogeneity in PSA expression among the population members, as presently seen.

EXAMPLE 2 Comparative Protein Analysis of Normal and Cancerous Esophageal Epithelium Patients and Tissue Samples

The two specimens studied were from patients who presented to the Shanxi cancer Hospital in Taiyuan, Shanxi Province, People's Republic of China and were diagnosed with esophageal cancer. Both patients were considered candidates for curative surgical resection. Both cases were stage two squamous cell carcinomas of the esophagus. Microdissection

Frozen section slides were prepared from each case and microdissected by LCM (Pixcell 100, Arcturus Engineering, Mountain View, CA) by selectively aiming for and capturing normal epithelium or tumor cells. LCM was performed as previously described except AEBSF (Boehringer Mannheim) was added to the staining baths at a final concentration of 2 mM to inhibit proteases. In each case 50,000 cells were procured. Based on careful review of the histologic sections, each microdissection is estimated to contain >98% of the desired cell type (normal versus tumor). Sample Preparation

One hundred microliters of IEF lysing solution (containing 7M Urea, 2M Thiourea, 4% CHAPS, 1% MEGA- 10, 1% Octyl-b-Glucopyranoside, 40 mM Tris, 50 mM DTT, and 2 mM tri-butyl phosphine (TBP) and 0.5% (v/v) Pharmalytes) was applied directly to the microdissected cells adhered on the LCM cap, placed into an Eppendorf tube and vortexed vigorously for one minute until all cells were completely lysed. The IEF lysing solution was then re-applied to another cap containing cells from the same microdissected material and the procedure repeated until each 100 μl contained lysate from 50,000 cells (approximately 7000 LCM transfer pulses). 2D PAGE and Image Analysis

First-dimension isoelectric focusing was carried out on a Pharmacia Immobiline IPG Dry-strip system essentially as described by the manufacturer. Pre-cast immobilized pH gradient strips (18 cm, 3-10 non-linear) were employed for the first dimensional separation for a total focusing time of 120kVh. The strips were re-equilibrated with a solution containing SDS and Tris pH 6.9, reduced with TBP (2 mM), alkylated with iodoactemide (2.5% w/v), and directly applied to a 8-18% linear gradient SDS-PAGE gel for electrophoresis overnight at 40 volts constant voltage. The gels were stained with silver, and direct scanning and image analysis was performed using an Umax scanner with Adobe Photoshop software and a Tektronix IISDX photographic-quality printer. Scanned images were analyzed and compared using the MELANIE II software package (BioRad). Comparison of protein fingerprints were performed using images representing protein spots readily apparent by direct visualization. Only spots that were present/completely absent between normal-tumor cells were defined as altered. Each experiment was performed in duplicate and produced similar results. Normalization of sample load was by anti-α-tubulin immunoblot analysis prior to the first-dimension run. Scoring of the blots included comparison of multiple exposure times. Analysis of α-tubulin

Ten μl of the IEF lysate was diluted 1:1 in 2X SDS sample buffer, boiled for 5 minutes, and applied to a 4-20% NO VEX Tris-glycine SDS gel and electrophoresed for 1 hour. Immunoblotting was performed for 1.5 hour using a BIO-RAD Semi-dry blotting apparatus with Immobilon PVDF membrane as the capture surface. Blots were blocked with IX TBS containing 1% ovalbumin and incubated with a monoclonal anti-α-tubulin antibody for 3 hours. Anti-α-tubulin antibodies were purchased from SIGMA and used at final dilution of 1 : 1000. Blots were washed with IX TBS three times for 5 minutes and each secondary antibody was added. HRP-coupled rabbit anti-mouse secondary antibodies were purchased from SIGMA and used at a final dilution of 1 :10,000. Blots were washed and ECL substrate (Amersham) was added for chemiluminescent detection on Kodak Bio-Max film. Mass Spectrometry

A separate 2D PAGE gel was run with microdissected tumor from case #1, as above, except the gel was stained and analyzed with zinc-imidazole. Two proteins that were aberrantly regulated in both tumors were chosen for mass spectrometry MS-MS sequencing. In-gel proteolytic digestion

The acrylamide plug was taken from the gel, washed once with HPLC-grade water for 5 minutes, and subjected to in-gel tryptic digestion as previously described. Gel pieces were excised and washed by end-over-end mixing in 12 ml 30% methanol for 30 minutes (room temperature), washed twice for 30 minutes with 150 μl of 1:1 acetonitrile/100 mM ammonium bicarbonate pH 8. Each spot was sliced into fourths and rehydrated 10 μl of 100 mM ammonium bicarbonate pH 8 containing trypsin (2 pmol/μl) (Promega modified trypsin). Additional 10 μl of digestion buffer without trypsin was then added. After incubation at 37° C for 20 hours, the condensate was collected by spinning the tubes briefly and excess liquid was removed into a new tube. Peptides remaining in the gel matrix were extracted twice with 150 μl 60% aqueous acetonitrile, 0.1% TFA at 30° C for 30 minutes. The extracted volumes were reduced to about 5 μl using an Eppendorf speed-vacuum concentrator. Electrospray mass spectrometry Mass spectrometric data regarding tryptic peptides from in-gel digestions were obtained from collision induced dissociation (CID) spectra with a Finnigan-MAT LCQ ion trap mass spectrometer after introduction via a polyamide-coated fused silica microcapillary HPLC. Immunoblot Analysis of Cytokeratin 1 and Annexin I

Two 2D gels were run simultaneously, both containing identical amounts of lysates of microdissected tumor from case #1. One gel was silver stained and the other was immunoblotted to PVDF membrane as described above. 2D western blot analysis was performed as described above, except antibody to type II cytokeratin or annexin I was used as probe. Anti-pan type II cytokeratin antibody was purchased from SIGMA and used at a final dilution of 1:1000. Anti-annexin I antibody was purchased from Transduction Labs and used at a final dilution of 1 :5000. Tissue Fixation/Embedding

A male BALB C mouse was sacrificed with immediate surgical excision of the liver. One piece was embedded in O.C.T. compound (Tissue Tek, Miles, Elkhart, IN) and immediately frozen on dry ice. Two pieces were fixed in 70% ethanol containing proteinase inhibitor (Complete, Mini Boehringer Mannheim Corp., Indianapolis, IN) for 1 hour at room temperature. One of these pieces was then embedded and blocked in paraffin wax and the other in polyester wax (Gallard-Schlesinger Industries, Inc., Carle Place, NY). For polyester wax embedding, the tissue was dehydrated in 90% ethanol at 4° C for 1 hour, followed by 99% ethanol at 4° C for 1 hour, then by 100% ethanol at room temperature for 1 hour. Then the tissue was infiltrated with 50% polyester wax:ethanol at 42° C for 2 hours followed by 90%> polyester wax: ethanol for one hour at 42° C. Finally, the tissue was blocked with 90% polyester wax:ethanol in a cryomold on ice. A separate piece of liver was fixed in 10%> neutral-buffered formalin and embedded and blocked in paraffin wax. Eight-micron thick sections were cut from each of the blocks onto glass slides, stained with hematoxylin and eosin, and approximately 5000 cells (1200 laser pulses) were dissected by LCM. The transfer cap with attached cells was applied to an Eppendorf tube containing 30 microliters of SDS buffer to lyse the cells, and the recovered protein were analyzed by SDS-PAGE. Protein Profiles of Matched Normal and Tumor Esophageal Cells Protein profiles from microdissected normal epithelium and tumor cells from two patients were analyzed by 2D PAGE (Figure 6). For the purposes of this study proteins were scored as "altered" only if there was a clear-cut "on-off ' difference between the comparison groups. No quantitative evaluation was made of the level of those proteins that were present in both sample groups and showed subtle differences in levels. Immunoblot analysis of α-tubulin using a small aliquot of each sample was used to verify that equal amounts of total protein were analyzed from each dissection (Figure 6C).

The overall protein profiles between the normal-tumor pairs were highly similar. For example, the microdissected normal epithelium from case #1 shared 98% of the observed proteins with the corresponding tumor (Table 2). Table 2

However, seventeen distinct differences were observed between the paired normal-tumor samples, 12 of which occurred identically in both cases. The circled spots in panels A and B of Figure 8 show proteins which are either up-or down-regulated in each tumor, as compared to the matching normal epithelium. As a control to assess inherent 2D PAGE variability, each experiment was performed twice and the results compared. One hundred percent of proteins were scored as identical in each duplicate run. Protein Identification Two proteins, one overexpressed and the other underexpressed in both tumors, were selected for further analysis to determine the feasibility of obtaining identification of proteins derived from microdissected cells. Spots on the 2D gel marked were chosen for mass spectrometry MS-MS sequencing. A second 2D PAGE gel was run, except the gel was stained and analyzed with zinc-imidazole followed by protein elusion and in-gel digestion with trypsin. The volume ratio of trypsin to gel plug was strictly controlled to maximize digestion. In addition, the volatile salt ammonium bicarbonate in the digestion buffer was mostly removed before the samples were used without purification to obtain molecular weights by ESI-MS. For each sample, the peptide mass peaks in the ESI-MS spectra were easily distinguished by the impurities and the masses were corrected by using internal calibration standard. With the protocols used in sample handling and MALDI or ESI loading, a large number of authentic peptide mass peaks were present and this made identification by peptide matching computer programs straight forward and very reliable. The results are reported in Table 3. By matching the experimental molecular weights of trypsin digestion products with theoretical predictions, multiple peptides were identified for each of the two proteins, spanning broad regions of the primary amino acid sequence. The protein over-expressed in tumors was identified as cytokeratin 1, and the protein under-expressed in tumors was identified as annexin I.

Table 3

Protein Confirmation - Immunoblot

To confirm the mass spectrometry results, microdissected normal epithelium and tumor samples from case #2 were subjected to 2D PAGE separation and immunoblot analysis using an antibody directed against a human pan-type II cytokeratin, and an antibody against annexin I. Figure 7 shows the results confirming the identity of the protein up-regulated in tumor as cytokeratin 1. The identity of the protein down-regulated in tumor as annexin I was similarly confirmed. Epithelium vs. Stroma

To ensure the present approach is efficient at detecting protein alterations between microdissected cell populations, a region of stroma was microdissected and compared to dissected normal epithelium and tumor (Figure 8). Stroma cells represent a lineage and phenotype markedly distinct from epithelium. Less than 25% (approximately 150 of 675) of the observed proteins were scored as identical between the two populations (Table 2). None of the 17 proteins that were altered in the normal epithelium-tumor comparisons were present in the stroma. Whole Tissue Sections vs. Microdissected Cells Samples from whole tissue sections prepared without dissection were analyzed by 2D PAGE and compared to microdissected normal epithelial and tumor cell populations from the same section. The experiment was performed to assess the impact of potential laser capture microdissection induced modifications, such as protein oxidation and proteolysis on 2D PAGE fingerprints. It also determined the percentage of proteins present in a microdissected cell population that could be reliably identified in the whole tissue section from which they were derived. Figure 9 shows the 2D PAGE comparison of the undissected whole tissue section and corresponding microdissected samples. All of the proteins in the 2D gel from the microdissected cells, including those found to be differentially regulated in the tumor (circled), were visualized at the same Mr and pi in the undissected whole tissue section indicating LCM had no apparent effect on proteins recovered from microdissected cells.

Validation and Extension to Other Cancers

These same methods have been used on eight additional normal/squamous cell cancer of the esophagus cells sets. The results indicated that annexin I is absent in all carcinoma cells, but strongly present in the nonnal epithelium. Immunohistochemistry experiments, as described above, indicate that annexin I abruptly disappears in this cancer at the pre-malignant state, and continues to be absent as the cells become cancerous. The presently described techniques have also been extended to examine a possible role of annexin I in cancers other than those of the esophagus. In particular, experiments were preformed to determine if the level of annexin I was altered in prostate cancer. Using Western blot analysis and immunohistochemical assays, as described above, 2 of the 3 prostate cancers were found to have completely lost the expression of annexin I, while the third cancer had reduced expression. Thus, these results indicate the applicability of the present techniques to the elucidation of protein alterations in a variety of cancer types and other biological conditions (such as normal organ development). -li

Previous studies of microdissected cell populations have focused primarily on DNA and more recently on analyses of messenger RNA, and have been aimed at identifying gross genomic alterations, specific genes with mutations, or mRNA levels. While these efforts have been successful and hold great promise for identifying molecular profiles in normal and diseased cells, they provide only partial information regarding the protein products of mutated or dysregulated genes. Efficiency of translation, post-translational modifications, protein stability, phosphorylation state, protein-protein interactions, and protein DNA-binding affinities are examples of parameters which cannot be studied by DNA and mRNA efforts alone. Protein-based analyses are required to address these questions. Moreover, protein profiling may more easily detect certain types of alterations than genomic or expression-based approaches, for example, a tumor suppresser gene mutation that results in protein truncation. Therefore, high-throughput protein studies will be an important component of future efforts to determine the molecular anatomy of normal and diseased cells.

Precise recovery of specific cell populations from human tissue is technically challenging, particularly when downstream molecular applications require on the order of tens of thousands of cells. In this study, procurement of normal epithelium was the most difficult challenge. LCM-based microdissection was critical and allowed recovery of 50,000 normal epithelial cells (and tumor cells) with a high degree of purity in a reasonably short period of time and did not appear to damage or degrade proteins. Fifty thousand microdissected cells revealed approximately 675 distinct protems as visualized by 2D-PAGE stained with silver. Assuming the lower limit of detection to be one nanogram, the analysis identified proteins in the abundance range of 50,000 to 1,000,000 copies per cell.

Dysregulated proteins of high or moderate abundance will have sustained utility for basic research and clinical applications, as they will be the easiest to detect, study, and monitor. However, proteomic studies of microdissected nonnal and tumor cells will also benefit from increased sensitivity enabling a larger percentage of the cellular proteins to be analyzed. Additionally, there are many microscopic, biologically interesting (and potentially clinically important) lesions that consist of less than 50,000 cells. In these circumstances, protein labeling with (for example) 125-1 labeling or biotinylation dramatically increases the number of proteins visualized from microdissected cells. Similarly, scanning immunoblotting with class-specific antibodies allows for more sensitive detection of specific subsets of proteins, for example, all known proteins involved with cell cycle regulation. In some circumstances, the small amount of material analyzed is not ideal for obtaining highly sensitive protein fingerprints, and identifying proteins of interest. In such circumstances, a useful strategy is to produce parallel "diagnostic" and "sequencing" 2D gels from each case. The diagnostic fingerprints are derived from microdissected cells and provide maximal sensitivity for detection of normal-tumor differences. Sequencing fingerprints then allow for determination of protein identity. The sequencing 2D gels are generated from serial, whole tissue section cryostat recuts that contain abundant amounts of protein representing all cell types present in the tissue, including the dissected cell population(s). Alignment of the diagnostic and sequencing 2D gels permits determination of proteins of interest for subsequent mass spectrometry or N-terminal sequence analysis. This technique successfully visualized and aligned all 675 proteins present in the microdissected tumor cells from case #1 in the corresponding whole tissue section 2D gel. Two of the tumor-specific alterations were selected for subsequent mass spectrometry and identified as cytokeratin I (over-expressed) and annexin I (under-expressed). The identity of both proteins was confirmed by immunoblot. Consistent with this result cytokeratins and annexin I have been previously reported to be dysregulated in epithelial tumors.

It is therefore advantageous, in some embodiments, to compare matched normal and tumor cells from the same patient. Of the seven proteins found to be down-regulated in the tumor, two were observed uniquely in the normal epithelium of case #1, and one was present uniquely in case #2.

These proteins are "patient-specific" and would not have been identified if a single reference normal sample had been used as comparator for the tumors. Therefore, direct intrapatient comparison of normal-tumor pairs is advantageous to ensure detection of all protein alterations that occur during tumorigenesis. Moreover, since little is known of the similarities and differences in protein profiles from person to person, the disclosed methods and devices can be used to provide interpatient analyses of both normal and diseased cells to help reveal patient-unique protein profiles related to disease susceptibility or progression.

The esophageal specimens utilized in this study were frozen in liquid nitrogen shortly after surgical resection and histologic tissue sections were prepared on a cryostat. Frozen sections are often used for molecular analyses of tissue due to the relatively high quality DNA, RNA, and proteins that can be recovered. Tissue processed through standard fonnalin fixation and paraffin embedding is less advantageous due to the molecular cross-linking which occurs during fixation and the prolonged exposure to elevated temperatures during embedding.

In summary, 2D-PAGE protein profiles and subsequent determination of the identity of selected proteins is possible from microdissected cell populations, and is uniquely capable of detecting disease-specific alterations. These techniques and methodologies can determine the abundance and status of proteins in vivo for studying normal cells and their associated pathologies.

EXAMPLE 3 Comparative Protein Analysis of Normal and Cancerous Prostate Epithelium and In Vitro

Prostate Cell Lines Patients and Tissue Samples

Tissue from prostate cancer patients undergoing radical prostatectomy at the Clinical Center of the National Cancer Institute (Bethesda, MD) and at the Mayo Clinic (Rochester, MN) was used. Tissue from the peripheral zone of the prostate was procured, immediately snap frozen, and stored at -80° C. Matching normal and tumor cell lines were prepared from the prostatectomy specimen obtain at the NCI and immortalized as described by Bright etal., Cancer Res. 57: 995-1002 (1997). LnCaP and PC3 cells were purchased from the American Type Culture Collection (Manassas, VA). Microdissection was performed and the cells were prepared for 2D-PAGE analysis as described above in Example 2. 2D-PAGE gels were done using α-tubulin to normalize the sample load, as described in Example 2. The anti-α-tubulin antibodies, used at 1:1000, and HRP-coupled rabbit anti- mouse secondary antibodies, used at 1:10,000) were purchased from Sigma (St. Louis, MO). Blots were washed using conditions described above and ECL substrate (Amersham, Piscataway, NJ) was added for chemiluminescent detection on Kodak Bio-Max film. For PSA, anti-PSA antibodies were purchased from Scripps Laboratories (San Diego, CA). Protein Profiles of Matched Normal and Tumor Prostate Epithelial Cells

Analysis of the microdissected cells from two different cases revealed approximately 750 distinct proteins, as shown by silver staining. Altered proteins were scored as discussed in Example 2 and the results are reported in Table 4. Twelve total proteins were altered between the normal and tumor prostate tissue. Six tumor-specific alterations were present in both cases. Two proteins were exclusively expressed in the normal cells of both cases. Three other proteins were uniquely expressed by normal cells from one case, while an additional protein was uniquely expressed by normal cells in the other case. These results confirm the ability of the present methods to make meaningful comparisons between the protein content of microdissected cell samples. Other proteins were overexpressed in the cancerous cells, one of which was confirmed by immunoblot to be PSA. Comparison was also made between epithelial and stromal cells, where less than 45% of the proteins were shared. This result confirms that protein expression differs between epithelium and stroma cell types in various tissue types. The six altered proteins in the epithelial cells were all epithelial- specific, indicating that the alterations seen were not due to contamination of the normal cells with stromal cell populations.

Table 4

Comparison to Prostate Cell Line Protein Expression

To determine if the protein profiles from prostate cancer cell lines are representative of prostate cancer in vivo, lysates of cells cultured in vitro were subjected to 2D-PAGE and the resulting patterns compared to those of LCM-derived normal malignant prostatic epithelium. The 2D-PAGE pattern of two common cell lines, LnCaP and PC3, were first compared to the pattern of in vivo cancerous cells. Protein expression between the two cell lines was similar, but markedly different from the protein profiles of epithelium in vivo, exhibiting less than 20% identity. To assess whether differences in 2D-PAGE profiles were the result of qualitative or quantitative alterations in expression, immunoblots were done using PSA as a test molecule. PSA was not detected in normal cells, PC3, or a tumor cell line developed from the in vivo dissected tumor cells. In contrast, LnCaP expressed PSA, but there was an alteration in the migration, reflective of a qualitative change in the protein. These results indicate that there are both quantitative and qualitative changes in cell expression when comparing immortal and in vivo isolated cells. Finally, a direct comparison between the in vivo tumor cells pre-immortalization and a cell line developed from those tumor cells shows substantial alterations in protein expression. This observation means at least some of the differences in protein expression seen between the LnCaP and PC3 cell lines and the tumor cells reflect changes in protein expression due to the immortalization process.

EXAMPLE 4 Characterization of PSA Proteins in Tumor and Nontumor Cell Samples

LCM of benign and malignant epithelium

Frozen tissue was obtained from radical prostatectomy specimens and embedded in O.C.T. compound (Tissue-Tek, Miles, Elkhart, IN). Eight-micron sections were made with a standard cryostat and stained with hemotoxylin and eosin using standard protocols. Benign and malignant histology was identified by a pathologist, and LCM was performed to obtain cells from each population by directing the laser at those populations of cells. LCM was performed as previously described, except AEBSF (Boehringer Mannheim) was added to the staining baths at a final concentration of 2 M, to inhibit proteases. For 1 -dimensional and 2-dimensional electrophoresis analysis, 2,000 (approximately 8 - 10,000 cells) and 5,000 (approximately 20 - 25,000 cells) 30- micron laser shots of each cell population were used, respectively. Based on careful review of histologic sections each dissection is estimated to contain > 95% of desired cells. Western blot analysis of intracellular PSA

For 1-dimensional electrophoresis (ID- PAGE) analysis, 2X SDS buffer was used to lyse cells directly from the EVA film (from the LCM cap) and the lysates were run on 20% Tris-glycine non-denaturing gels and transferred to a nylon membrane using the Novex system. A murine monoclonal antibody purchased from Scripps Laboratories was used as the primary antibody, at a concentration of 1 : 1 ,000, and a murine horseradish peroxidase tagged anti-IgG antibody from Sigma was used as the secondary antibody. The ECL reaction was generated by the ultra-ECL from Pierce. Purified free PSA and PSA bound to ACT purchased from Scripps were used as controls.

For 2-dimensional electrophoresis (2D-PAGE), IEF buffer (7M Urea, 2M Thiourea, 4%> CHAPS, 1%) MEGA-10, 1% Octyl-b-Glucopyranoside, 40 mM Tris, 50 mM DTT, and 2 mM tri-butyl phosphine (TBP) and 0.5% (v/v) Pharmalytes) was used to lyse cells off the EVA film and the lysates were used to swell pre-cast immobilized pH gradient strips (18 cm, 3-10 non-linear from Pharmacia) overnight. First-dimensional separation was performed for a total focusing time of 120 kVh. The strips were re-equilibrated with a solution containmg SDS and Tris pH 6.9, reduced with DTT (50 mM), alkylated with iodoactemide (2.5% w/v), and directly applied to a 9% isocratic SDS-PAGE gel for electrophoresis overnight at 40 volts constant voltage. The protein was transferred to a nylon membrane and western blot performed using previously described reagents. Twenty 8 micron cryostat sections containing both malignant and benign epithelium were suspended in EF buffer and analyzed by 2D PAGE and western blotting as just described. PSA/ACT binding studies.

Malignant and normal prostatic epithelium were dissected using LCM and lysis buffer (containing protease inhibitors and 100 mM NaCl) was used to lyse the cells off of the EVA film and solubilize the protein. One microgram of purified ACT (purchased from Scripps) was added to 10 μl of lysis buffer containing 2,000 laser shots (approximately 10,000 cells) of normal or malignant epithelium and incubated for 2 hours at 37°C. Results PSA is a serine protease that is produced as an inactive zymogen and then activated by release of a 17 amino acid signal peptide, followed by liberation of a 7 amino acid propeptide. The catalytically active form of PSA is highly glycosylated with a molecular mass of approximately 30 kDa. Although it has been shown that PSA is constitutively produced in LnCaP cells, the relative proportion of pre- and processed forms of PSA within uncultured prostatic epithelium has not been reported. Furthermore, the site at which PSA binds to ACT is also not known.

In order to answer these questions, anti-PSA western blot analysis was performed on cellular lysates from LCM-derived normal and malignant epithelial cells, which revealed a band at 30 kDa (Figure 10). These results were highly reproducible in several cases, and demonstrate that a majority of intracellular PSA exists as a 30 kDa form. This finding suggests that PSA binds to ACT within either the extracellular space or the serum and not intracellularly. The total abundance of cellular

PSA varied such that in some cases the malignant cells contained more PSA while in other cases the benign cells contained more PSA. This variability of intracellular PSA levels within both normal and malignant prostatic epithelium is consistent with previous reports. Binding Experiments To ascertain whether the 30 kDa protein represents the active form of PSA experiments were performed to determine if it would bind to ACT. As shown in Figure 11, the majority of cellular PSA complexed with ACT after 2 hours of incubation and there was no difference in the binding capacity of PSA derived form normal and malignant epithelium. This finding suggests that the 30 kDa form represents active PSA but does not explain why serum percent free PSA is lower in men with cancer. Although the PSA found intracellularly in normal and malignant epithelial cells appeared identical by molecular weight and ACT binding capacity, other protein modifications that do not alter patterns on ID-PAGE gels may have been missed by this investigative methodology. Therefore, 2D PAGE was performed on cellular lysates from LCM derived tissue. As can be seen in Figure 12A, three distinct 30 kd PSA isoforms similarly exist in benign and malignant epithelial cells, indicating that the PSA found in malignant prostatic epithelium is not mutated or differentially glycosylated, because this would be reflected in a change in isoelectric focusing point on a 2D Western. Furthermore, none of these three isoforms is a proteolytically cleaved form of PSA since all remained intact under reducing conditions. The fact that the purified PSA/ACT complex remains intact under the reducing conditions (Figure 12B) validates that the findings demonstrated in Figure 11 are not an artifact of 2- D-PAGE. In addition, whole tissue frozen sections were analyzed (that had not been exposed to LCM) containing both benign and malignant epithelium to demonstrate that firing the laser on the cells did not alter PSA structure (Figure 12A). These experiments show that PSA made in normal and malignant prostatic epithelium does not differ in regards to molecular weight, iso-electric focusing point or ability to bind ACT, and that differences in serum percent free PSA levels are not the result of altered ACT production by malignant prostatic epithelium.

EXAMPLE S

Production of SELDI Protein Fingerprints for Tumor and Nontu or Cell Samples

To assess the reproducibility of the protein fingerprint generated by SELDI, two separate cell populations of prostatic tumor epithelial cells (1200 cells per dissection) from the same patient (Case 1) were separately microdissected (Microdissections 1 and 2) from frozen tissue sections.

AEBSF (Boehringer Mannheim) was added to the staining baths at a final concentration of 2 mM to inhibit proteases. Based on careful review of the histologic sections by a pathologist, each microdissection was estimated to have greater than 95% purity. Microdissected cells were lysed directly on an LCM cap with 2 μl of an extraction buffer containing 1% (w/v) Triton-X-100 (Sigma, St. Louis, MO), 1% (w/v) MEGA 10 (ICN, Aurora, OH), and 1% (w/v) octyl-B-glucopyranoside (ESA, Chelmsford, MA) in a standard IX PBS base. SELDI analysis was performed using an aliphatic reverse phase chip (Ciphergen, Palo Alto, CA). The bait surfaces on the chip were pretreated with 2 μl of acetonitrile (Sigma, St. Louis, MO). Shortly before the acetonitrile completely evaporated, 2 μl of the lysate was applied to the bait surface. The analyte was allowed to concentrate by air drying, followed by the application of 0.3 ml of 3,5-dimethoxy-4-hydroxycinnamic acid (sinapinic acid, 98%, Sigma, St. Louis, MO) as the energy-absorbing molecule of choice for all experiments in this study.

The results of this experiment (Figure 13 A) are displayed in two of several possible formats offered by the SELDI software analysis package. The first representation is a standard chromatographic mass map with the respective molecular weight range displayed on the x-axis. Each "peak" represents a protein isoform with a different molecular weight. The second representation is a Gel- View^® display which takes the peak data from the mass-map chromatogram and presents the data as if one is looking at a standard ID PAGE gel "stained" for proteins, with the molecular weight ranges displayed at the same scale as that seen in the chromatogram. The protein profiles of the lysates of separate microdissected regions of the same tumor are nearly identical. These results indicate that reproducible protein profiles are generated from LCM-derived cells. Similar experiments were repeated several times with identical results. To assess the reproducibility and relative homogeneity of captured cell populations of the same cell type between patients, prostatic tumor epithelium from two different patients (Case 1 and Case 2) was dissected (1200 cells per dissection) and analyzed as above. These results are shown by both mass-mapping and Gel- View™ display in Figure 13B, and show that the protein fingerprint generated from case 1 and case 2 is highly reproducible. Similar results were obtained with multiple samples from the two cases. The minor differences in protein expression patterns observed between case 1 and case 2 were no more variable than that seen between microdissections from the same case. To assess the relative sensitivity of the SELDI instrument in the detection of protein profiles from LCM-acquired cells, four separate microdissections of decreasing numbers of prostatic tumor epithelium were isolated and the resulting lysate subjected to SELDI analysis using the reverse-phase aliphatic chip as above. The results, shown in Figure 13C in Gel- View™ display, indicate that a protein profile from as little as 450 cells can be visualized, and that the spectra is comparable to that seen from 1500 cells. The mass maps from fewer numbers of cells (300 and 150) gave less reproducible results, reflecting microheterogeniety in protein expression from region to region in the same tissue section, or indicating that such lower numbers of cells may not provide the most optimal results when assessing complex protein mixtures. As a control for each experiment, lysing buffer alone, or LCM with a blank glass slide, were analyzed. Each produced a flat baseline SELDI spectra. The Gel- View^® display normalizes the relative intensities of each separate spectra and does not reflect what was seen in the mass map; that is, as the total number of cells decreases, so does the relative intensity of the peaks seen. The results using a range of cell numbers indicate that the relative intensity of the peaks may be proportional to the number of cells, the relative fidelity of the spectra is not altered when studying the protein fingerprints of 450 cells to 1500 cells.

For the purposes of these studies only the molecular weight range of 15 kDa to 70 kDa was analyzed, although the SELDI has a practical resolution and detection working range of 1000 to 300,000 Daltons depending on the energy-absorbing molecule utilized and the bait surface/wash conditions employed. Western blot analysis of LCM derived cells in our laboratory using anti α- tubulin as a housekeeping marker for a protein expression reference standard indicated that procurement of cells by an equal number of laser shots generates a nearly equivalent (+/- 5%) final protein yield. Discriminatory SELDI analysis of different tumor types

A practical application for both clinical and research studies utilizing SELDI fingerprinting of LCM-acquired cell populations would be to develop diagnostic fingerprints which are disease-specific and organ-specific. To assess the feasibility of this concept, we tested the ability of SELDI to discriminate between tumor epithelium from prostate, breast, and colon tumors from different patients. Twelve hundred cells were analyzed as described above, and the resultant spectra are shown in Figure 14A and B. Each tumor type shows a unique fingerprint, illustrating the wide range of protein differences that occur in epithelial cells from different solid tumor types. Novel insights into the protein expression patterns between different types of tumors could aid in the development of new treatment strategies, or the choice of the best strategy. Diagnosis and prognosis from the limiting amounts of cells generated by fine needle aspirants or sentinel node analysis could be achieved very rapidly if a tumor-specific or grade-specific profile could be generated and used as a template. SELDI protein fingerprinting of colon cancer with liver metastasis

The ability to assess changes in protein expression occurring during tumor progression will aid in the elucidation of the fundamental mechanisms underlying carcinogenesis in patients. To investigate the potential of SELDI analysis of LCM derived cells to study this process, the colonic normal epithelium, primary cancer, and hepatic metastasis were analyzed from one patient. Normal liver cells were also analyzed as a comparison; the tissue was microdissected from the same case. Results are shown in Figure 14C, and demonstrate that a normal epithelial cell and tumor epithelial cell-specific fingerprint can be identified.

Additionally, the protein profile from the colon tumor that had metastasized resembles the colonic epithelium, regardless of tumor or normal cell state, and not the liver cell protein profile. Importantly, the cells from the metastasis have their own distinct protein fingerprint. This uniqueness may arise from the fact that the metastatic tumor has changed its expression profile as a result of its new environment, or it may reflect protein expression changes that enabled the cells to metastasize, or a combination of both. A direct application of this type of analysis would be to identify protein profiles of primary tumors that predict that a tumor has likely metastasized indicating a need for more aggressive patient treatment and follow-up. If the protein fingerprint from the colon to liver metastasis is compared to the fingerprints of the prostate or breast tumor epithelium (Figure 14A), one can see that the fingerprint most closely resembles that of the colon, regardless of normal or tumor status. A practical application of this type of analysis would be the generation of protein fingerprints that are organ-specific so that metastasis of unknown origin could be characterized and matched to a particular organ type. See Figure 15 for an example of a battery of fingerprints that could be used.

These results indicate that SELDI analysis of LCM-derived cells can be achieved with a high degree of reproducibility, fidelity, sensitivity, and discrimination. These operating conditions allow for the exciting possibility of proteomic analysis of small populations of diseased cells directly from patient tissue. This enables, for the first time, visualization of disease-specific protein changes occurring in defined patient-matched cells directly from tissue sections. Evaluation of Cancer Disease Progression

Normal, pre-malignant prostatic intraepithelial neoplasia (PIN) and frankly invasive cancer cells and normal cells were microdissected from one stained human tissue section and a protein biomarker fingerprint obtained, as discussed above. The comparisons were made between proteins isolated from approximately 1500 cells and all samples were run in triplicate. Two proteins, having molecular weights of approximately 28,000 and 32,000 Daltons, were reproducibly differentially expressed during the progression of the cells from normal to pre-cancerous to frankly invasive. The results are disclosed in Figure 17A, 17B, and 17C. In particular, the relative ratio of expression of these two proteins can be correlated to the state of the cell and is reproducible in the three cases analyzed. Although this is an extremely small sample set, this experiment demonstrates the applicability of the present techniques to analysis of the progression ofa cell sample to cancer. Extension to Other Cancers

These same methods have been extended to determine the presence of protein "biomarkers" in eight normal/squamous cell cancer of the esophagus cell sets. Three separate independent microdissections of eight different patients' tumor and normal esophageal epithelium were performed and the whole cell lysate subjected to SELDI analysis via the use of a hydrophobic interaction Cl 8 binding surface. Each replicate was run in triplicate, giving a total of 72 data points for each protein peak analyzed. The results are reported in Figure 18A and 18B. Two proteins were found to be present in a disease specific manner, one of which was up regulated in a tumor-specific manner in 7 out of 8 cases analyzed, and one was found to be specifically present in nonnal epithelium from all 8 cases examined. Coefficients of variation and standard deviations for these analyses indicate that the relative abundance of these two protems may serve as valid "biomarkers" that correlate with disease. Thus, these results indicate the applicability of the present techniques to the elucidation of protein alterations in a variety of cancer types.

EXAMPLE 6 Multiplexed Tissue Arrays for High-Throughput Target Validation and Drug-Tissue

Interaction Assay

The ability to compare protein content of finite cell samples as described above can be applied to high throughput assays utilizing tissue arrays. Examples of tissue array techniques that can be used with the protein extraction and analysis methods are disclosed in PCT publications

W099/44063 and W099/44062, whose disclosures are hereby incorporated by reference. The present methods, in a tissue array format, are useful for screening potential therapeutic agents and analyzing their impact on very small subsets of cells. In particular, the methods are used to analyze the impact of therapeutic agents on specific cellular pathways, such as signaling pathways. These results can, for example, indicate the efficacy or toxicity of a particular agent on the cell.

To apply the present methods to this type of analysis, the tissue from multiple patients is exposed to the agent to be tested. The exposure can be done in vivo, prior to collection of the sample, or in vitro, after collection of the sample and/or after laser microdissection. In vivo exposure would involve administration of the agent to the subject, while in vitro administration could be to cultured cells. Some agents that could be tested include pharmacological agents, imaging agents, labeled proteins, such as ligands, or other agents known to have particular effects on cells, such as cytokines. After exposure, microdissection techniques are used to isolate the cells of interest from the sample, the cells are lysed to allow isolation of the proteins or other cellular components, such as nucleic acids or other subcellular structures, from the sample, and the lysate contents are transferred to a confined zone ofa substrate. The lysate contents, or cellular components, are placed in identifiable positions on a substrate, where such positions are confined zones. One example ofa confined zone is the coordinates of an array. The array is constructed by applying microspots of the isolated proteins on any suitable matrix, such as nitrocellulose, nylon, or silica. Referring to Figure 19, the microspots are arranged on the matrix in any manner that produces meaningful data. The micropsots can be placed on the matrix using, for example, a micropipette, and examples of the substrate include a glass or plastic slide, a section of embedding medium, or a nitrocellulose matrix. In one embodiment, the microspots can be arranged in the y- dimension by patient number and in the x-dimension by a criterion for categorizing the various samples obtained, such as by stage of malignancy. Other alternative criteria include before and after treatment samples, various cells types, such as epithelial and stromal, or stages of development for embryonic samples. The microspots are then subjected to some type of analysis, for example, to determine if an amount ofa particular protein is altered. Such analysis can include probing with an antibody as the binding agent. Alternative analyses include probing with other binding agents such as nucleic acids, labeled or unlabeled DNA or RNA, or aptameric or phage display screening. A consistent alteration in the cellular content of a protein is then correlated to exposure of the tissue or cells to the agent of interest.

The matrix can be a general capture matrix, such that all isolated proteins from the sample are present, or embedded with a specific binding agent that would result in the maintenance of only proteins of interest. In this embodiment, the microdissected proteins in a soluble form can be labeled, for example with a radioactive or fluorescent tag, and capture by the binding agent is detected by the presence of the label. Alternatively, the protein spots themselves can act as a binding agent or attractant for an interaction, such as detection ofa labeled target or ligand within the protein spot after treatment with a labeled test solution. Because of the method of isolation of the proteins, the test is highly sensitive, as each spot may be the protein extracted from a pure population of cells from an individual tissue, or a combination of particular populations of interest.

By using samples from multiple patients, the results have greater statistical significance. Alternatively, a large number of samples can be gathered from one patient's tissue or tissues to achieve the desired statistical results. Other sample sources besides patient tissue include formalin and ethanol fixed paraffin embedded stained tissue sections. With such a sample, it is possible to view the stained tissue using microscopy, then microdissect cells from the precise area of the slide of interest for further analysis. The results can then be referenced back to photographs or other records of the slide image.

Additionally, it is anticipated that the tissue arrays can include a body tissue set, which allows comparison of the results obtained from the experimental tissue set to samples from other tissues of the patient's body. Thus, any variation in individual patient protein composition, or any global effect a therapeutic may have, can be taken into account during interpretation of the results. Other possible controls include known protein mixtures applied to the matrix in spots or "spiking" each spot with a positive control.

A schematic drawing illustrating a sample tissue array for examining protein content of breast tissue of multiple patients is illustrated in Figure 19. In this illustration, the tissue from multiple patients is used to examine the effect of a particular therapeutic agent administered to a subject on protein content of breast tissue in various stages of malignancy, from normal tissue to stromal invasion. These tissues are microdissected from a tissue specimen using LCM. Because of the sensitivity of the test, detailed questions about effects on the level of cellular biochemical pathways in each stage subset can be obtained. An additional application of tissue array technology to the described methods is the use of the protein content assay to prescreen various tissue types for the target of diagnostic or therapeutic moieties. In this embodiment, a therapeutic or diagnostic agent is being evaluated for its appropriateness in a patient's treatment. If .the target of this agent is known, the presence or absence of this target in various cell types of the patient can be tested. Thus, using the breast tissue of Figure 19 as an example, if it was questionable whether a target enzyme for a chemotherapeutic agent was present in a particular patient's cancerous cells, the samples are generated and placed in an array as illustrated. Then an antibody or other means of identification of the target is used to determine if the target agent is present. This allows diagnostic, imaging, or therapeutic agents to be specifically selected for patients or disease types based on the presence of the target molecule in the diseased tissue.

To prepare a comparison assay between a tissue sample and the results of an array analysis, tissue may, for example, be preserved by fixation in 70% ethanol and embedded in a paraffin block. The paraffin block is cut in 5 to 10 μm sections and mounted on microscope slides. The tissue on these slides is deparaffmized by placing the slide in room temperature xylene for 5 minutes, followed by xylene heated to 70°C for 5 minutes. The slide is next placed in acetone for 10 seconds, Diff ■ Quick Solution II (Dade Behringer) for 5 seconds, and acetone for 10 seconds. The stained slide is allowed to rinse in xylene for a minimum of 30 seconds to dehydrate.

Cells of interest are dissected from the slide by LCM using the PixCell II (Arcturus Engineering). Using the 30 μm spot size, 3000 laser pulses are fired to collect cells from a subceliular structure of interest, and the cells are embedded in a thermoplastic polymer film supported on a plastic cap. Extraction buffer, consisting of one part SDS sample buffer (Novex) with 4% BME and one part T-PER (Pierce), is heated to boiling. A 40 μl aliquot of the hot extraction buffer is added to a standard 500 μl microcentrifuge tube. The cap containing the tissue embedded on the thermoplastic film is inserted into the tube and inverted to allow the extraction buffer to cover the tissue. The inverted tube is vortexed for 20-30 seconds to aid in the disruption of the cells and release of the cellular components. Still inverted, with the buffer flooding the cap surface, the tube and cap assembly is placed in a 70°C incubator for overnight incubation. After the incubation, the tube is once more vortexed in an inverted position, then it is turned upright and centrifuged for 10 seconds. Dilutions are made from the resulting lysate using T-PER. A minimum of 10 μl of lysate or dilution is placed in the bottom of a well of a 96-well round-bottom microtiter plate.

This plate is positioned on the stage of Genetic Micro Systems 417 Arrayer. Oncyte Slides (Schleicher & Schuell) covered in pure nitrocellulose are placed in the appropriate positions on the arrayer. After the desired number of slides and replicates has been programmed into the accompanying software, the instrument is activated. A metal micro tube lowers into the appropriate well and lifts 1 μl of the lysate, trapping it inside the bore of the tube. After positioning the tube over the nitrocellulose slide, a pin pierces the trapped lysate, spotting a minute aliquot onto the nitrocellulose surface measuring 500 μm in diameter. Two different detection systems have been used successfully to visualize and quantitate the concentration of the proteins applied to the nitrocellulose and are representative of the detection means that can be used with the described methods. System One

The Western-Star Chemiluminescent Detection System (TROPIX, Inc., Bedford, MA) is an immunoblot detection system designed for membrane-bound proteins. The protocol outlined for nitrocellulose membranes is followed for the Oncyte slides. After the nitrocellulose slide is prepared on the arrayer and dried, it is submerged in I-Block blocking buffer (provided in the kit) prepared as directed, and agitated on an orbital mixer for at least 1 hour. The primary antibody of choice is diluted to the recommended concentration using I-Block buffer. The slides are drained after the blocking step and the diluted primary antibody is poured over the slides and for incubation, with agitation, overnight at 4°C. After the overnight incubation, the slides are washed with agitation in a IX PBS/ 0.1 %> TWEEN washing buffer, one time for 15 minutes and twice for 5 minutes. The secondary antibody, which is conjugated to alkaline phosphatase, is diluted in I-Block blocking buffer and poured over the slide. It is allowed to incubate for 4-5 hours at room temperature on an orbital mixer. At the end of this incubation, the slide is washed with PBS as described above. The 10X Assay Buffer (provided in the kit) is diluted with deionized water and the slide is incubated twice for two minutes submerged in the IX Assay Buffer. The slides are blotted dry on a piece of tissue paper. For the nitrocellulose slides, a 1:20 dilution of Nitro-Block-II (provided in the kit) is prepared using the CDP-Star Ready-To-Use substrate solution. This substrate solution is pooled on the slide and allowed to sit undisturbed for 5 minutes. At the end of this incubation, the slide is blotted and placed in a plastic development folder, after which it is exposed to x-ray film. Representative results using this system for analysis of the content of PSA in prostate tissue is shown in Figures 20B and 20C. System Two The second illustrative example uses the DAKO Autostainer Universal Staining System

(DAKO Coφoration, Carpinteria, CA). This automated slide processing system uses a primary antibody followed by the addition of a patented horseradish peroxidase labeled polymer (DAKO En Vision™ System). This is an example ofa colorimetric-detectable labeled antibody. The prepared Oncyte slide is blocked in I-Block (TROPIX, Inc.) for a minimum of one hour. The DAKO Autostainer is programmed with the incubation times and volumes of the reagents to be dispensed on the slides. The nitrocellulose slide is washed with a Wash Buffer wash (prepackaged by DAKO). The primary antibody, diluted as specified for the antibody being used, is applied in the predetermined amount of 600 μl, and allowed to incubate for 30 minutes. At the end of the incubation, the slide is washed with 600 μl of the Wash Buffer. This step is followed by the addition of the HRP-labeled polymer for 10 minutes, followed by washing with the Wash Buffer. The substrate is added to the slide and allowed to sit for 5 minutes, followed by a final wash, and is then allowed to dry for optimal visualization. The positive results are visible staining of the spotted proteins on the slide. Representative results using this system for analysis of the content of PSA in prostate tissue are shown in Figure 20A.

The tissue array method for analyzing protein content of small sample sizes is very reproducible, even when applied to very small numbers of cells. Experiments to examine the reproducibility of the technique were done using five separate slides of normal esophageal tissue, and detection of annexin I was performed. The size of the samples varied in the number of microdissection "shots," from a high of 500 shots to a low of 16 shots. Samples of varying cell numbers were collected, the proteins were isolated as described above, and the isolated proteins were placed on a matrix of nitrocellulose paper. Using standard Western techniques, the microspotted proteins were probed using the antibody systems described above. The results of these tests are diagrammed in Figures 21A, 21B, and 21C. Figure 21 A indicates that the results from each data set (average of triplicate results for each slide) remains consistent within the set. As is expected, the amount of variance increases with the smaller number of "shots" included in the sample. However, as indicated graphically in Figure 2 IB, the results for multiple samples containing as few as 125 shots varies only by 18%. These results indicate that this method of analyzing protein content of small tissue samples is reliable. The results shown in Figure 21C also support this conclusion and indicate that the relative value between the various series remains generally consistent between samples of varying numbers of shots. In summary, the results indicate that decreasing the number of "shots" in a sample affects only the quantitative, not qualitative, character of the results.

Example results of an embodiment where the binding agent has been immobilized, and the cell lysates placed in contact with the immobilized agent, are shown in Figure 22. In this experiment, lysates from microdissected normal and tumor esophageal cells were labeled with biotin. An array of binding agents, including anti-annexin I, anti-α-tubulin, anti-phospho ERK, and mouse IgGl, and normal rabbit serum were placed on nitrocellulose paper. Biotin labeled lysates were placed in contact with the binding agent arrays, and the interaction was detected by avidin-peroxidase. As shown in Figure 22, the tumor lysates contained a reduced concentration of annexin I, and an increased concentration of phospho ERK, when compared to the normal cell lysates. EXAMPLE 7

Identification of Aberrantly Expressed Proteins Using Laser Capture Microdissected Tumors

Prostate and Esophageal Cancers:

Using techniques substantially similar to those disclosed above, proteins have been identified that are up-regulated or down-regulated in prostate or esophageal cancers.

Direct comparisons between lysates separated by two-dimensional gel electrophoresis of patient-matched laser capture microdissected normal and tumor epithelium were analyzed by silver staining and/or fluorescent staining. Protein spots that were found to substantially (more than twofold) differ in intensity between the two populations (microdissected normal epithelium versus microdissected tumor epithelium) were identified. Specifically, spots were identified that were up- regulated in the cancer tissue compared to the normal tissue, or down-regulated in cancer tissue compared to normal tissue (thus appearing as "elevated" in normal tissue when it was compared to the cancer tissue).

Two-dimensional gel electrophoretic separations of undissected, whole tissue specimens from the same patients were analyzed concomitantly by the same methodology, and those spots identified on the laser capture microdissected "road-map" gels were identified on the whole tissue gels by gel alignment. The spots selected in this manner were excised, subjected to in-gel trypsin digestion and routine protein microsequencing via Finnegan LCQ-MS-MS mass spectroscopy. The identified proteins are listed in Table 5 (human prostate adenocarcinoma and normal epithelium) and Table 6 (human esophageal squamous cancer and normal epithelium). This embodiment particularly includes the following features:

(1) patient-matched laser capture microdissected material is used as the comparator starting material for protein discovery and identification; and

(2) the laser capture microdissected 2D-PAGE gels were used as a road map to select which protein were chosen for further analysis, but the spots collected for microsequencing came off of the whole tissue 2D-PAGE.

Table 5

Table 6

Invasive Ovarian Cancer:

To detect aberrantly expressed invasive ovarian cancer-related proteins, a comparison was made between microdissected invasive ovarian cancer epithelium and microdissected ovarian borderline or low metastatic potential (LMP) cancer epithelial cells, rather than between patient matched normal and tumor cells (as described above for prostate and esophageal cancers). Otherwise, methods used to identify invasive ovarian cancer-related proteins were essentially similar to those described above. The proteins identified as being elevated in human ovarian invasive cancer epithelium are listed in Table 7.

These proteins are believed to have particular clinical significance. Each protein that has been identified as being invasive-specific (over-expressed in invasive versus borderline or LMP cancer epithelium cells) also is associated with the metastatic phenotype itself. This phenotype is of critical importance in ovarian cancer, since the non-metastatic cancer form (borderline or LMP) and early stage disease is highly curable. The invasive ovarian cancer related proteins identified have gone through a particularly rigorous comparison in that not only are they aberrantly expressed, but are over expressed when compared to cancer epithelium microdissected from tumors that do not metastasize.

It is believed that ovarian cancer metastasis may be blocked by targeting one (or a combination) of the identified proteins. This may provide a treatment for ovarian cancer. Also, the identified proteins can be used as early markers for diagnosis of ovarian cancer, particularly with metastatic potential.

Table 7

Breast Cancer, including DCIS:

Using techniques substantially similar to those disclosed above, proteins have been identified that are differential regulated in human breast cancer or in ductal carcinoma in situ (DCIS; a prelmalignant early stage form of breast cancer) when compared to each other or to normal tissue.

Direct comparisons between lysates separated by two-dimensional gel electrophoresis of patient-matched laser capture microdissected normal and tumor epithelium, or normal and DCIS, or DCIS and tumor, were analyzed by silver staining and/or fluorescent staining. Protein spots that were found to substantially (more than two-fold) differ in intensity between the two populations were identified. Specifically, spots were identified that were up-regulated in the cancer or DCIS tissue compared to the normal tissue, or down-regulated in cancer or DCIS tissue compared to normal tissue (thus appearing as "elevated" in normal tissue when it was compared to the cancer tissue), or up- or down-regulated in DCIS tissue compared to tumor tissue. Where available, invasive tumor samples were also analyzed.

Two-dimensional gel electrophoretic separations of undissected, whole tissue specimens from the same patients were analyzed concomitantly by the same methodology, and those spots identified on the laser capture microdissected "road-map" gels were identified on the whole tissue gels by gel alignment. The spots selected in this manner were excised, subjected to in-gel trypsin digestion and routine protein microsequencing via Finnegan LCQ-MS-MS mass spectroscopy. Proteins identified in each type of comparison are indicated in Table 8 (listed alphabetically) and Table 9 (reorganized based on relative protein abundance between the indicated tissue samples).

It is believed that breast cancer metastasis and/or progression may be blocked by targeting one (or a combination) of the identified proteins. This may provide a treatment for breast cancer. Also, the identified proteins can be used as early markers for diagnosis of breast cancer, particularly with metastatic potential, as well as for monitoring the effectiveness of treatment.

It is useful to subdivide and reorganize proteins listed Table 8 into those that are more highly expressed in tumor than in normal samples (see Table 9, Part A), more highly expressed in normal than in tumor samples (see Table 9, Part B), more highly expressed in DCIS than in normal samples (see Table 9, Part C), more highly expressed in normal than in DCIS samples (see Table 9, Part D), more highly expressed in tumor than in DCIS samples (see Table 9, Part E), or more highly expressed in DCIS than in tumor samples (see Table 9, Part F).

Because laser capture microdissection was used to generate the starting cellular material for discovery of these proteins, direct comparisons could be (and were) performed within the same patient for different cell types. The proteins themselves, not the genes, are the drug targets, imaging targets, and biomarkers for early detection.

Identification of proteins specifically upregulated early in the disease process (i.e., premalignant or DCIS for breast cancer) provides new biomarkers for early disease (or disease progression) detection, new imaging targets for patient monitoring, and new drug targets for early intervention. Because these proteins have specifically been discovered in patient-matched material, their dysregulation is assured.

Proteins the expression of which is lost in the disease process, and particularly those which are expressed only in the normal counterpart, can be used as biomarkers for early detection. For instance, their loss in tissue biopsy samples would signal important early changes for the clinician.

Proteins whose expression resides solely with the invasive counterpart provide new biomarkers for detection, for imaging, and new drug targets. The proteins that are increased early in the breast cancer development process and remain highly expressed would fall into the same category.

EXAMPLE 8 Expression of Cancer-Related Polypeptides

The disclosed cancer-related proteins (and fragments thereof) can be expressed by standard laboratory technique. After expression, the purified protein or polypeptide may be used for functional analyses, antibody production, diagnostics, prognostics, and patient therapy, e.g., for prevention or treatment of prostate cancer (including hormone-refractory or metastatic prostate cancer), esophageal cancer, ovarian cancer (particularly invasive ovarian cancer), or breast cancer (or DCIS). Furthermore, the DNA sequences encoding the disclosed cancer-related proteins can be manipulated in studies to understand the expression of these genes and the function of their products, in particular how these proteins function in the control of, or response, to cancer or cancer development.

Mutant forms of human cancer-related proteins (and corresponding encoding sequences) may be isolated based upon information contained herein, and may be studied in order to detect or further study alteration in expression patterns in terms of relative quantities, tissue specificity and functional properties of the encoded mutant cancer-related protein.

Partial or full-length cDNA sequences, which encode the subject protein, may be ligated into bacterial expression vectors. Methods for expressing large amounts of protein from a cloned gene introduced into Escherichia coli (E. coli) or other prokaryotes may be utilized for the purification, localization, and functional analysis of proteins. For example, fusion proteins consisting of amino terminal peptides encoded by a portion of the E. coli lacZ or trpE gene linked to a cancer-related protein may be used to prepare polyclonal and monoclonal antibodies against these proteins. Thereafter, these antibodies may be used to purify proteins by immunoaffinity chromatography, in diagnostic assays to quantitate the levels of protein and to localize proteins in tissues and individual cells by immunofluorescence. Intact native protein may also be produced in E. coli in large amounts for functional studies.

Methods and plasmid vectors for producing fusion proteins and intact native proteins in bacteria are described in Sambrook et al. (In Molecular Cloning: A Laboratory Manual, Ch. 17, CSHL, New York, 1989). Such fusion proteins may be made in large amounts, are easy to purify, and can be used to elicit antibody response. Native proteins can be produced in bacteria by placing a strong, regulated promoter and an efficient ribosome-binding site upstream of the cloned gene. If low levels of protein are produced, additional steps may be taken to increase protein production; if high levels of protein are produced, purification is relatively easy. Suitable methods are presented in Sambrook et al (In Molecular Cloning: A Laboratory Manual, CSHL, New York, 1989) and are well known in the art. Often, proteins expressed at high levels are found in insoluble inclusion bodies. Methods for extracting proteins from these aggregates are described by Sambrook et al. (In Molecular Cloning: A Laboratory Manual, Ch. 17, CSHL, New York, 1989). Vector systems suitable for the expression of lacZ fusion genes include the pUR series of vectors (Ruther and Muller-Hill, EMBO J. 2:1791, 1983), pEXl-3 (Stanley and Luzio, EMBO J. 3:1429, 1984) and pMRlOO (Gray et at, Proc. Natl. Acad. Sci. USA 79:6598, 1982). Vectors suitable for the production of intact native proteins include pKC30 (Shimatake and Rosenberg, Nature 292: 128, 1981), pKK177-3 (Amann and Brosius, Gene 40: 183, 1985) and pET-3 (Studiar and Moffatt J. Mol. Biol. 189:113, 1986).

Fusion proteins, for instance fusions that incorporate a portion ofa cancer-related protein, may be isolated from protein gels, lyophilized, ground into a powder and used as an antigen. The DNA sequence can also be transferred from its existing context to other cloning vehicles, such as other plasmids, bacteriophages, cosmids, animal viruses and yeast artificial chromosomes (YACs) (Burke et al, Science 236:806-812, 1987). These vectors may then be introduced into a variety of hosts including somatic cells, and simple or complex organisms, such as bacteria, fungi (Timberlake and Marshall, Science 244:1313-1317, 1989), invertebrates, plants (Gasser and Fraley, Science

244:1293, 1989), and animals (Pursel et al, Science 244:1281-1288, 1989), which cell or organisms are rendered transgenic by the introduction of the heterologous cancer-related cDNA (such as a cDNA encoding one of the proteins listed in Table 5, 6, 7, or 8).

For expression in mammalian cells, the cDNA sequence may be ligated to heterologous promoters, such as the simian virus (SV) 40 promoter in the pSV2 vector (Mulligan and Berg, Proc. Natl. Acad. Sci. USA 78:2072-2076, 1981), and introduced into cells, such as monkey COS-1 cells (Gluzman, Cell 23:175-182, 1981), to achieve transient or long-term expression. The stable integration of the chimeric gene construct may be maintained in mammalian cells by biochemical selection, for example with neomycin (Southern and Berg, J. Mol. Appl. Genet. 1:327-341, 1982) or mycophenolic acid (Mulligan and Berg, Proc. Natl. Acad. Sci. USA 78:2072-2076, 1981).

DNA sequences can be manipulated with standard procedures such as restriction enzyme digestion, fill-in with DNA polymerase, deletion by exonuclease, extension by terminal deoxynucleotide transferase, ligation of synthetic or cloned DNA sequences, site-directed sequence- alteration via single-stranded bacteriophage intermediate or with the use of specific oligonucleotides in combination with PCR.

The cDNA sequence (or portions derived from it) or a mini gene (a cDNA with an intron and its own promoter) may be introduced into eukaryotic expression vectors by conventional techniques. These vectors are designed to permit the transcription of the cDNA in eukaryotic cells by providing regulatory sequences that initiate and enhance the transcription of the cDNA and ensure its proper splicing and polyadenylation. Vectors containing the promoter and enhancer regions of the SV40 or long terminal repeat (LTR) of the Rous Sarcoma virus and polyadenylation and splicing signal from SV40 are readily available (Mulligan et al, Proc. Natl. Acad. Sci. USA 78:1078-2076, 1981; Gorman et al, Proc. Natl. Acad. Sci USA 78:6777-6781, 1982). The level of expression of the cDNA can be manipulated with this type of vector, either by using promoters that have different activities (for example, the baculovirus pAC373 can express cDNAs at high levels in S. frugiperda cells (Summers and Smith, In Genetically Altered Viruses and the Environment, Fields et al. (Eds.) 22:319-328, CSHL Press, Cold Spring Harbor, New York, 1985) or by using vectors that contain promoters amenable to modulation, for example, the glucocorticoid-responsive promoter from the mouse mammary tumor virus (Lee et al., Nature 294:228, 1982). The expression of the cDNA can be monitored in the recipient cells 24 to 72 hours after introduction (transient expression).

In addition, some vectors contain selectable markers such as the gpt (Mulligan and Berg, Proc. Natl. Acad. Sci. USA 78:2072-2076, 1981) or neo (Southern and Berg, J. Mol. Appl. Genet. 1:327-341, 1982) bacterial genes. These selectable markers permit selection of transfected cells that exhibit stable, long-term expression of the vectors (and therefore the cDNA). The vectors can be maintained in the cells as episomal, freely replicating entities by using regulatory elements of viruses such as papilloma (Sarver et al, Mol. Cell Biol. 1:486, 1981) or Epstein-Barr (Sugden et al., Mol. Cell Biol. 5:410, 1985). Alternatively, one can also produce cell lines that have integrated the vector into genomic DNA. Both of these types of cell lines produce the gene product on a continuous basis. One can also produce cell lines that have amplified the number of copies of the vector (and therefore of the cDNA as well) to create cell lines that can produce high levels of the gene product (Alt et al, J. Biol. Chem. 253:1357, 1978).

The transfer of DNA into eukaryotic, in particular human or other mammalian cells, is now a conventional technique. The vectors are introduced into the recipient cells as pure DNA

(transfection) by, for example, precipitation with calcium phosphate (Graham and vander Eb, Virology 52:466, 1973) or strontium phosphate (Brash et al, Mol. Cell Biol. 7:2013, 1987), electroporation (Neumann et al, EMBO J 1:841, 1982), lipofection (Feigner et al, Proc. Natl. Acad. Sci USA 84:7413, 1987), DEAE dextran (McCuthan et al, J. Natl. Cancer Inst. 41:351, 1968), microinjection (Mueller et al, Cell 15:579, 1978), protoplast fusion (Schafner, Proc. Natl. Acad. Sci. USA 77:2163-2167, 1980), or pellet guns (Klein et al., Nature 327:70, 1987). Alternatively, the cDNA, or fragments thereof, can be introduced by infection with virus vectors. Systems are developed that use, for example, retroviruses (Bernstein et al, Gen. Engr'g 7:235, 1985), adenoviruses (Ahmad et al, J. Virol. 51:261, 1986), or Herpes virus (Spaete et al, Cell 30:295, 1982). Encoding sequences can also be delivered to target cells in vitro via non-infectious systems, for instance liposomes.

These eukaryotic expression systems can be used for studies of cancer-related nucleic acids (such as those encoding the proteins listed in Tables 5, 6, 7, and 8) and mutant forms of these molecules, as well as cancer-related proteins and mutant forms of these protein. Such uses include, for example, the identification of regulatory elements located in the 5' region of cancer-related genes on genomic clones that can be isolated from human genomic DNA libraries. The eukaryotic expression systems may also be used to study the function of the normal cancer-related proteins, specific portions of these proteins, or of naturally occurring or artificially produced mutant versions of cancer-related proteins. Using the above techniques, the expression vectors containing cancer-related gene sequence or cDNA, or fragments or variants or mutants thereof, can be introduced into human cells, mammalian cells from other species or non-mammalian cells as desired. The choice of cell is determined by the purpose of the treatment. For example, monkey COS cells (Gluzman, Cell 23 : 175- 182, 1981) that produce high levels of the SV40 T antigen and permit the replication of vectors containing the SV40 origin of replication may be used. Similarly, Chinese hamster ovary (CHO), mouse NIH 3T3 fibroblasts or human fibroblasts or lymphoblasts may be used.

Recombinant vectors that comprise all or part of a cancer-related gene or cDNA sequence for expression in a suitable host are also provided. The nucleic acid sequence is operatively linked in the vector to an expression control sequence to form a recombinant DNA molecule, so that the encoded polypeptide can be expressed. The expression control sequence may be selected from the group consisting of sequences that control the expression of genes of prokaryotic or eukaryotic cells and their viruses and combinations thereof. The expression control sequence may be specifically selected from the group consisting of the lac system, the trp system, the tac system, the trc system, major operator and promoter regions of phage lambda, the control region of fd coat protein, the early and late promoters of SV40, promoters derived from polyoma, adenovirus, retrovirus, baculovirus and simian virus, the promoter for 3-phosphoglycerate kinase, the promoters of yeast acid phosphatase, the promoter of the yeast alpha-mating factors and combinations thereof. The host cell, which may be transfected with the vector of this invention, may be selected from the group consisting of E. coli, Pseudomonas, Bacillus subtilis, B. stear other mophilus or other bacilli; other bacteria; yeast; fungi; insect; mouse or other animal, such as human tissue cells; or plant hosts.

It is appreciated that for mutant or variant cancer-related DNA sequences, similar systems are employed to express and produce the mutant product. In addition, fragments ofa cancer-related protein can be expressed essentially as detailed above, as can fusion proteins that include at least a fragment of an identified cancer-related protein (such as those listed in Tables 5, 6, 7, and 8). Such fragments include individual cancer-related protein domains or sub-domains, as well as shorter fragments such as peptides. Cancer-related protem fragments having therapeutic properties may be expressed in this manner also.

EXAMPLE 9 Suppression of Cancer-Linked Protein Expression

A reduction of target protein expression in a transgenic cell may be obtained by introducing into cells an antisense construct based on a sequence encoding the target protein, such as a cDNA or gene sequence,, or flanking regions thereof, of any one of the proteins listed in Tables 5, 6, 7, or 8. For antisense suppression, a nucleotide sequence encoding a specific cancer-linked protein, e.g. all or a portion ofa sequence (such as a cDNA or gene) encoding a protein that is aberrantly expressed in prostate cancer (such as those listed in Table 5), esophageal cancer (such as those listed in Table 6), invasive ovarian cancer (such as those listed in Table 7), or breast cancer (such as those listed in Table 8) is arranged in reverse orientation relative to the promoter sequence in the transformation vector. Other aspects of the vector may be chosen as for any other expression vector (see, e.g., Example 5). The introduced sequence need not be a full-length cDNA or gene, and need not be exactly homologous to the equivalent sequence found in the cell type to be transformed. Generally, however, where the introduced sequence is of shorter length, a higher degree of homology to the native sequence likely will be needed for effective antisense suppression. The introduced antisense sequence in the vector may be at least 30 nucleotides in length, and improved antisense suppression will typically be observed as the length of the antisense sequence increases. The length of the antisense sequence in the vector advantageously may be greater than 100 nucleotides.

Although the exact mechanism by which antisense RNA molecules interfere with gene expression has not been elucidated, it is believed that antisense RNA molecules bind to the endogenous mRNA molecules and thereby inhibit translation of the endogenous mRNA.

Suppression of endogenous gene expression can also be achieved using ribozymes. Ribozymes are synthetic RNA molecules that possess highly specific endoribonuclease activity. The production and use of ribozymes are disclosed in U.S. Patent No. 4,987,071 to Cech and U.S. Patent No. 5,543,508 to Haselhoff. The inclusion of ribozyme sequences within antisense RNAs may be used to confer RNA cleaving activity on the antisense RNA, such that endogenous mRNA molecules that bind to the antisense RNA are cleaved, which in turn leads to an enhanced antisense inhibition of endogenous gene expression.

In addition, dominant negative mutant forms of the disclosed cancer-related sequences may be used to block endogenous activity of the corresponding gene products. Suppression of expression of one or more cancer-related gene(s) or the encoded protein(s) can be used, for instance, to treat, reduce, or prevent cell proliferative and other disorders caused by over- expression or unregulated expression of the corresponding cancer-related gene. In particular, suppression of expression of some or all of the sequences disclosed herein as being up-regulated in prostate, esophageal, ovarian, or breast cancer can be used to treat, reduce, or prevent progression of prostate, esophageal, ovarian, or breast cancer (respectively).

EXAMPLE 10 Identification of Therapeutic Compounds

The cancer-related molecules disclosed herein, and more particularly the linkage of these molecules to cancer and cancer progression, can be used to identify compounds that are useful in treating, reducing, or preventing prostate, esophageal, ovarian, or breast cancer, or development or progression of these cancers.

By way of example, a test compound is applied to a cell, for instance a test cell, and at least one cancer-related molecule level and/or activity in the cell is measured and compared to the equivalent measurement from a test cell (or from the same cell prior to application of the test compound). If application of the compound alters the level and/or activity of a cancer-related molecule (for instance by increasing or decreasing that level, such as the level of a protein listed in Table 5, 6, 7, or 8), then that compound is selected as a likely candidate for further characterization. In particular examples, a test agent that opposes or inhibits a cancer-related change is selected for further study, for example by exposing the agent to a prostate, esophageal, ovarian, or breast cancer cell in vitro, to determine whether in vitro growth is inhibited. Such identified compounds may be useful in treating, reducing, or preventing cancer or development or progression of cancer, particularly the cancer it was identified as being associated with. In particular embodiments, the compound isolated will inhibit, inactivate, enhance, or activate at least one cancer-related molecule, for instance one or more of the proteins listed in Table 5, 6, 7, or 8. For example, in those embodiments where the cancer-related molecule level is increased in the associated cancer, compounds can be identified that decrease the level (or activity) of that molecule. Similarly, where the level of a molecule is reduced in a subject suffering from the associated cancer, compounds can be identified that increase the level (or activity) of that molecule.

With the provision herein of profiles that contain a plurality of molecules each linked to a specific cancer, identification of therapeutic compounds using the responses of multiple molecules in the profile (whether they increase or decrease) is now enabled. It is particularly contemplated that such changes in such profiles with the treatment of a potential therapeutic compound may reflect increases in some of the molecules in the profile and decreases in others. Beneficially, compounds are isolated that tend to reverse the apparent effect(s) of the cancer on the profile, and thus that tend to yield a protein profile from a treated sample that looks "more normal" (or less cancerous) than a protein profile from an untreated sample. Methods for identifying such compounds optionally can include the generation ofa specific cancer-related protein expression profile, as described herein. Control protein expression profiles useful for comparison in such methods may be constructed from normal tissue (particularly normal tissue corresponding to the target cancer), or a cancer tissue that is a different stage (or possibly at a different stage) than the experimental tissue.

EXAMPLE 11 Protein Expression Profiles (Fingerprints)

With the provision herein of methods for determining proteins that are linked (through aberrant expression) to a specific cancer, and the provision ofa collection of such cancer-linked molecules (as represented for instance by the prostate-cancer linked proteins listed in Table 5, the esophageal cancer-linked proteins listed in Table 6, the invasive ovarian cancer-linked proteins listed in Table 7, and the breast cancer-linked proteins listed in Table 8), protem expression profiles that provide information on the cancer-state ofa subject are now enabled.

Specific cancer-related expression profiles (e.g., prostate cancer-related, esophageal cancer- linked, invasive ovarian cancer-linked profiles, or breast cancer-linked or DCIS-linked profiles) comprise the distinct and identifiable pattern of expression (or level) of sets of cancer-related proteins, for instance a pattern of high and low expression ofa defined set of proteins, or molecules that can be correlated to such proteins, such as mRNA levels or protein activities.

The set of molecules in a particular profile will usually include at least one that is represented by (or correlated to) a protein listed in Table 5 (if it is a profile for prostate cancer), listed in Table 6 (if it is a profile for esophageal cancer), in Table 7 (if it is a profile of invasive ovarian cancer), or in Table 8 (if it is a profile of breast cancer). Particular breast cancer profiles will include at least one molecule that is represented by (or correlated to) a protein listed in one of the Parts of Table 9. In other examples of cancer-related gene expression profiles, more than one molecule corresponding to the proteins listed in these tables is included in the profile. By way of example, any subset of the proteins listed in Table 5 (or 6 or 7 or 8 or 9) (or corresponding to the molecules in this list) may be included in a single expression profile. Specific examples of such subsets include those molecules that show increased expression in the specified cancer, those that show decreased expression as compared to non-malignant tissue, and so forth, such as those subdivisions indicated in the relevant tables.

By comparison to a cancer-linked profile, a "normal" profile for a certain tissue type is the protein profile of a histologically "normal" (or more nearly normal) cell or cell sample from the same tissue. A change in a profile over time, or after an intervention such as a treatment or the application of a compound, can be said to change the protein profile so that it is more like a normal profile or more like a cancer-linked profile. In such instances, it is understood that the level of at least one protein in the profile is changed to be more nearly like the level that is seen in a reference profile (e.g., a reference normal profile for that tissue, or a reference cancer-linked profile for that tissue). The change will be either an increase or a decrease, depending on which brings it closer to the amount seen in the reference profile. It is understood that a change in a profile may be reflected as a change in only one or a few proteins in a whole profile that comprises a plurality of proteins. Thus, a protein profile can change so that it more closely resembles a normal protein profile for a tissue, without it being necessary that every protein in the profile reflect a normal (or more nearly normal) expression level.

Specific examples of profiles can be as follows:

(1) For prostate cancer, certain example profiles will include at least DJ403A15.3 protein (index # 6706675), phosphoglucomutase-related protein (index # 2135916), glial factor 1 (index # 183250), serotransferrin (index # 136191), aldehyde dehydrogenase (index # 118495), or GRP78 (index # 121567). In other embodiments, prostate linked-profiles will include any two (or more) proteins in Table 5, up to and including all of the proteins listed in that table. In other embodiments, at least two (or three, or four, etc.) proteins that are up-regulated in the cancer tissue, or at least two (or three, or four, etc.) proteins thβt are down-regulated in the cancer tissue, or both, are included in the profile. In other embodiments, prostate cancer profiles will include at least glyoxyalse I (index #

2135301), cofilin, annexin V (index # 113690), a 14-3-3 protein (e.g., index # 112690), annexin I (index # 113944), or tropomyosin (index # 136090), or a combination of two or more of these, three or more, four or more, five or more, or all six of these proteins. Examples of such profiles may additionally include at least one other protein listed in Table 5, or other proteins. (2) For esophageal cancer, certain example profiles will include at least glutathione fransferase (index # 87564), nm23 protein (index # 35068), enolase 1, or fascin. In other embodiments, esophageal cancer-linked profiles will include any two (or more) proteins in Table 6, up to and including all of the proteins listed in that table. In other embodiments, at least two (or three, or four, etc.) proteins that are up-regulated in the cancer tissue, or at least two (or three, or four, etc.) proteins that are down-regulated in the cancer tissue, or both, are included in the profile.

In other embodiments, esophageal cancer profiles will include at least NM23 protein (index # 35068), fascin (index # 2498357, or annexin 1 (index # 113944), or a combination of two or more of these, three or more, or all four of these proteins. Examples of such profiles may additionally include at least one other protein listed in Table 6, or other proteins.

(3) For invasive ovarian cancer, certain example profiles will include at least pyruvate kinase (index # 266427), FK506-binding protein (index #3916257), rho-GDP dissociation inhibitor protein (index # 11433318), lactoglutathione lyase (index # 219664), or a protein with a tryptic fragment with a sequence of SEAFTQQLNSR. In other embodiments, invasive ovarian cancer-linked profiles will include any two (or more) proteins of these proteins (which are also listed in Table 7), up to and including all five of these proteins. In other embodiments, at least two (or three, or four, etc.) proteins that are up-regulated in the cancer tissue, or at least two (or three, or four, etc.) proteins that are down- regulated in the cancer tissue, or both, are included in the profile. Further embodiments of invasive ovarian cancer profiles will include any two, any three, or any four of the proteins listed in Table 7.

(4) For breast cancer, certain example profiles will include at least one of the proteins listed in Table 8. It is particularly contemplated that specific profiles will include at least one of the following: 14-3-3 zeta/delta; actin-related protein 3 beta; alpha enolase; annexin A2; annexin II; annexin V; annexin VII; antithrombin III precursor; cathepsin B; cathepsin B; cellular retinoic acid-binding protein 2; chloride intracellular channel protein 1 ; cofilin 1 ; collagen Type VI, alpha 1 precursor; collagen VI αl chain precursor; collagen VI alpha 2 chain precursor; collagen VI alpha 3 chain precursor; crystallin, alpha B; cyclophilin; dimethylarginine dimethylaminohydrolase 1; elongation factor 1 -alpha; endoplasmin Precursor; eukaryotic translation initiation factor 3; F-actin capping protein beta subunit, CapZ; fatty acid binding protein; fatty acid binding protein 5; FK506 BP; GDP-mannose pyrophosphorylase B isoform 2; glucose-regulated protein precursor (grp 78); glycyl-tRNA synthetase; HSP 27; HSP 90β; heterogeneous nuclear protein similar to Rat Helix destabilizing protein; heterogeneous nuclear ribonucleoprotein L; heterogeneous nuclear ribonucleoprotein X; hnRNP A2; hnRNP a2/bl; hnRNP C; lactoferrin (diferric); macrophage capping protein; Mn-superoxide dismutase*; Neuropolypeptide H3; Nm 23 (acidic isoform); Nm 23 (basic isoform); Nm 23 ; Nm 23 B; nucleoside diphosphate kinase; peptidylprolyl isomerase B; peroxiredoxin 1; peroxiredoxin 5; phosphatidylethanolamine-binding protein (Neuropolypeptide H3); polymeric immunoglobulin receptor precursor; profϊlin 1; prohibitin; prostatic binding protein; protein kinase C substrate 80K-H; protein phosphatase; Ras-related protein Rab 11a; Rho GDI; SI 00 A8; S100A11 (S100C, calgizzarin); S100A7; selenium binding protein 1; stathmin; superoxide dismutase 2; superoxide dismutase, mitochondrial precursor; thioredoxin peroxidase 2; transgelin; tropomyosin; tropomyosin 4; tropomyosin 4-anaplastic lymphoma kinase fusion protein (TPM4-ALK); ubiquitin conjugating enzyme E21; ubiquitin conjugating enzyme Ubc9; or voltage-dependent anion channel protein 1. In other embodiments, breast cancer-linked profiles will include any two (or more) proteins of these proteins (which are also listed in Table 8), up to and including all of these proteins. In other embodiments, at least two (or three, or four, etc.) proteins that are up-regulated in cancer tissue (or DCIS tissue), or at least two (or three, or four, etc.) proteins that are down-regulated in the cancer tissue(or DCIS tissue), or both, are included in the profile. Further embodiments of breast cancer profiles include any two, any three, or any four, any five, any six, any eight, any ten, any 15, any 20, any 30, and so forth, of the proteins listed in Table 8.

Particular contemplated profiles are those including one or more proteins that are present at a higher level in tumor than in normal samples, such as: 19343 (gil4769716); aldolase A protein (P04075); Annexin V (P08758); Annexin VII (P20073); bA139H14.1 (gi8217500); Cathepsin B (P07858); Cellular retinoic acid-binding protein 2 (gi4503029); Coagulation Factor (gi4503635); Coflin 1, Non-muscle (gi6680924); Endoplasmic reticulum lumenal protein (gi5803013); Ferritin (gil3655121); GDP-mannose pyrophosphorylase B, isoform 2 (gil 1761621); Globin, alpha-2 (HBA 2) (Q9Y5P6); Glucose-regulated protein precursor (grp 78) (PI 1021); Glycyl-tRNA Synthetase (gi493066); Hypothetical protein DKFZp761K0511.1 (gil l277141); Hypothetical protein FLJ 10652 (gi8922573); Hypothetical protein XP-033918 (gil4775389); isoAsp protein carboxyl methylotransferase (P22061); L-lactate dehydrogenase mu chain (P00338); Lamin A (P25391); Mn- superoxide dismutase (P04179); Nm23 (Q96NZ2); Non-metastatic cells 2, protein (NM23B) exp'd in (gi4505409); Omithine aminotransferase precursor (gi4557809); Peptidylprolyl isomerase B (gil4755229); Peroxiredoxin 1 (Q06830); Phosphatidylethanolamine-binding protein

(Neuropolypeptide H3) (P30086); Prohibiting (P35232); Proteasome subunit gamma (Q12920); Protein phosphatase (gi7657475); Ras-related protein Rabl la (gi5453555); Rho GDI (P52565); S100A11 (S100C, calgizzarin) (P31949); S100A7 (P31151); Similar to hypothetical protein (gil3528771); Stathmin (gi5031851); Survival/ evasion peptide gene (Ay044239); Thioredoxin peroxidase 2 (Q06830); Thioredoxin peroxidase B (P48822); Transaldolase 1 (P37837); Triose- phophate isomerase (P00938); Tropomyosin 4 (P07226); Tropomyosin 4-anaplastic lymphoma kinase fusion protein (TPM4-ALK) (Q96PF4); tRNA pseudouridine synthase A (Q9Y606); Tyrosyl tRNA ligase (P54577); or Ubiquitin (P02248). In some embodiments, such profiles include at least two, at least three, at least 5, at least 10, at least 15, at least 25, at least 50, at least 100, or more of these proteins. In some examples, other proteins from Table 8 are included in the profile.

Other particular contemplated profiles are those including one or more proteins that are present at a higher level in normal than in tumor samples, such as: 7150 (gi4759461); Acri- Coenzyme A Dehydrogenase (gi4557235); Apolipoprotein Al (P02647); Collagen VI αl chain precursor (P12109); Collagen VI alpha 3 chain precursor (P12111); Cytovillin 2 (gi340217, P15311); Dimethylarginine dimethylaminohydrolase 1 (gi6912328); Elastin B Precursor (gil63026); Epidermal autoantigen 450k (clone pE450-B) (gil082355); Nucleoside diphosphate kinase (gil421609); Similar to P29266 (gi416873); or Transferrin (P02786). In some embodiments, such profiles include at least two, at least three, at least 5, at least 10, or more of these proteins. In some examples, other proteins from Table 8 are included in the profile.

Other particular contemplated profiles are those including one or more proteins that are present at a higher level in DCIS than in normal samples, such as: 14-3-3 zeta/delta (P29312); Alpha enolase (P06733); bA139H14.1 (Q9NTI9); Biotin Carboxylase (P5074J); Carbonic anhydrase I

(P00915); Cellular retinoic acid-binding protein 2 (P29373); Chloride intracellular channel protein 1 (P000299); Cofilin 1 (P23528); Crystallin, alpha B (P02511); Cyclophilin (Q08752); D-3- phosphoglycerate dehydrogenase (P043175); Elastin B Precursor (P15502); Elongation factor 1-alpha (P04720); Endoplasmin precursor (P14625); F-actin capping protein beta subunit, CapZ (P47756); Fatty acid binding protein 5 (P15090); GDP-mannose pyrophosphorylase B isofonn 2 (Q9Y5P6);

Glucose-regulated protein precursor (grp 78) (PI 1021); Heat shock protein 27 (P04792); Heat shock protein 90 beta (P08238); Heterogeneous nuclear Rib nucleoprotein L (P14866); Heterogeneous nuclear Rib nucleoprotein X (G 971674); hn RNP A2 (P22626); hnRNP C (P07910); Human erythrocyte catalase 3-amino- 1,2,4-triazole complex, chain B (G 9257062); Hypothetical protein DKFZp761K0511.1 (Q9NTK6); Hypothetical protein FLJ10652 (Q9NVL6); KIAA0120 (P37802); Lactoferrin (diferric) (P02788); Lamin A (P25391); L-lactate dehydrogenase mu chain (P00338); Macrophage capping protein (P40121); Omithine aminotransferase precursor (P04181); Phosphatase 2C motif (P075688); Phosphoglycerate mutase B (P18669); Profilin 1 (P07737); Proteasome subunit, beta type, 9 (P28065); Rho GDI (P52565); Protein Kinase C substrate 80K-H (P14314); S100 A8 (P05109); S100A11 (S100C, calgizzarin) (P31949); Similar to mRNA for KIAA0120 gene GenBank Ace # D21261.1 (P37802); Stathmin (P16949); Superoxide dismutase, mitochondrial precursor (P04179); Thioredoxin peroxidase B (P48822); Tyrosyl tRNA ligase (P54577); or Ubiquitin conjugating enzyme E21 (P50550). In some embodiments, such profiles include at least two, at least three, at least 5, at least 10, at least 15, at least 25, or more of these proteins. In some examples, other proteins from Table 8 are included in the profile.

Still other particular contemplated profiles are those including one or more proteins that are present at a higher level in normal than in DCIS samples, such as: 19343 Fibrinogen Beta Chain (P19343); 3-hydroxyisobutyryl-CoA hydralase (Q9LKJ1); Aconitase (Q99798); Adenylate Kinase 1 (P00568); Annexin A2 (P07355); Antithrombin III precursor (P01008); Apolipoprotein Al (P02647); Collagen VI αl chain precursor (P12109); Cytovϊllin 2 (gi340217, P15311); Dimethylarginine dimethylaminohydrolase 1 (P094760); DNF 15S2 (P13798); Epidermal autoantigen 450k (clone pE450-B) (GI1082355); Eukaryotic translation initiation factor 3 (Q13347); Fatty acid binding protein (P15090); FK506 BP (P20071); High mobility group nonhistone chromosomal protein 1 (Q9NQJ4); KIAA0106 mRNA (P30041); Multicatalytic endopeptidase complex, proteasome precursor, beta subunit (023716); NADH dehydrogenase (ubiquinone) Fe-S protein (Q969P3); NADH - ubiquinone oxidoreductase 75 kD subunit (P28331); Neuropolypeptide H3 (P30086); Peptidylprolyl isomerase B (Q9BVK5); Peroxiredoxin 5 (P30044); polymeric immunoglobulin receptor precursor (P01833); Prostatic Binding protein (P08723); Selenium binding protein 1 (Q13228); Short chain 3-hydroxyacyl-CoA-dehydrogenase (Q16836); Similar to P29266 (gi416873); Superoxide dismutase 2 (P04179); Transferrin (P02786); Transgelin (Q01995); Transketolase (P29401); Tropomyosin (P09493); Tyrosine 3-monooxygenase/ tryptophan 5- monooxygenase activatioprotein (Q9CSZ6); or Voltage-dependent anion channel protein 1 (P21796). In some embodiments, such profiles include at least two, at least three, at least 5, at least 10, at least 15, at least 25, or more of these proteins. In some examples, other proteins from Table 8 are included in the profile.

Also contemplated are profiles that include one or more proteins that are present at a higher level in tumor than in DIGS samples, such as: L-lactate dehydrogenase mu chain (P00338); Lamin A (P25391); Rho GDI (P52565); S100A11 (S100C, calgizzarin) (P31949); or Tyrosyl tRNA ligase (P54577). In some embodiments, such profiles include at least two, at least three, at least four, or even all five of these proteins. In some examples, other proteins from Table 8 are included in the profile.

Further specific contemplated profiles are those that include one or more proteins that are present at a higher level in DCIS than in tumor samples, such as: Apolipoprotein Al (P02647);

Calreticulin precursor (P27797); Collagen VI αl chain precursor (P12109); Thioredoxin peroxidase B (P48822); or Transferrin (P02786). In some embodiments, such profiles include at least two, at least three, at least four, or even all five of these proteins. In some examples, other proteins from Table 8 are included in the profile. Also contemplated are profiles that include breast cancer-linked proteins from two or move of these categories, for instance both proteins that are more highly expressed in tumor than normal and proteins that are expressed at lower levels in tumor than in normal. Specific examples of cancer- linked protein profiles will include at least one protein from each of the categories, thus at least one protein from each of the individual Parts of Table 9.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The singular terms "a," "an," and "the" include the plural equivalents unless context clearly indicates otherwise. "Comprises" means "includes." Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, suitable methods and materials are described herein. All publications, patent applications, patents, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including explanations of terms, will control. In addition, the materials, methods, and examples are illustrative only and not intended to be limiting. It is understood that the invention is not limited to the specific embodiments shown and described, but changes or modifications can be made in the embodiments without departing from the contemplated scope of the present invention.

Claims

We claim:

1. A method of diagnosing or prognosing development or progression of prostate cancer in a subject, comprising detecting a change in expression level of one or more proteins of the ■ subject, wherein at least one such protein is DJ403A15.3 protein (index # 6706675), phosphoglucomutase-related protein (index # 2135916), glial factor 1 (index # 183250), serotransferrin (index # 136191), aldehyde dehydrogenase (index # 118495), or GRP78 (index # 121567).

2. The method of claim 1, further comprising detecting a change in expression level of at least one additional protein selected from the proteins listed in Table 5. 3. A method of screening for a compound useful in treating, reducing, or preventing prostate cancer or development or progression of prostate cancer, comprising determining if application of a test compound alters a protein expression profile ("fingerprint") so that the profile more closely resembles a normal prostate protein profile than it did prior to such treatment, and selecting a compound that so alters the protein expression profile, wherein the protein expression profile includes at least DJ403A15.

3 protein (index # 6706675), phosphoglucomutase-related protein (index # 2135916), glial factor 1 (index # 183250), serotransferrin (index # 136191), aldehyde dehydrogenase (index # 118495), or GRP78 (index # 121567.

4. The method of claim 3, wherein the protein expression profile further comprises at least one additional protein selected from the proteins listed in Table 5.

5. A method for treating or inhibiting prostate cancer in a subject, comprising administering a therapeutically effective amount of a pharmaceutical compound directed to restore normal expression of a prostate-cancer linked protein listed in Table 5, thereby treating the subject or inhibiting prostate cancer in the subject.

6. The method of claim 5, wherein the method comprises: identifying a protein listed in Table 5 that is over- or under-expressed in a subject having prostate cancer; administering a pharmaceutical compound directed to restore normal expression of that protein; monitoring expression of the protein until expression approaches normal levels, thereby treating prostate cancer.

7. A method of diagnosing or prognosing development or progression of prostate cancer in a subject, comprising detecting an increase or decrease in protein level of at least one protein listed in Table 5 in a prostate sample from the subject, when compared to a second prostate sample.

8. A method of diagnosing or prognosing development or progression of esophageal cancer in a subject, comprising detecting a change in expression level of one or more proteins of the subject, wherein at least one such protein is glutathione fransferase (index # 87564), nm23 protein (index # 35068), or enolase 1.

9. The method of claim 8, further comprising detecting a change in expression level of at least one additional protein selected from the proteins listed in Table 6.

10. A method for treating or inhibiting esophageal cancer in a subject, comprising administering a therapeutically effective amount of a pharmaceutical compound directed to restore normal expression of an esophageal-cancer linked protein listed in Table 6, thereby treating the subject or inhibiting esophageal cancer in the subject.

11. The method of claim 10, wherein the method comprises: identifying a protein listed in Table 6 that is over- or under-expressed in a subject having esophageal cancer; administering a pharmaceutical compound directed to restore normal expression of that protein; monitoring expression of the protein until expression approaches normal levels, thereby treating esophageal cancer.

12. A method of screening for a compound useful in treating, reducing, or preventing esophageal cancer or development or progression of esophageal cancer, comprising determining if application of a test compound alters a protein expression profile so that the profile more closely resembles a normal esophageal protein profile than it did prior to such treatment, and selecting a compound that so alters the protein expression profile, wherein the protein expression profile includes at least glutathione fransferase (index # 87564), nm23 protein (index # 35068), or enolase 1.

13. The method of claim 12, wherein the protein expression profile further comprises at least one additional protein selected from the proteins listed in Table 6.

14. A method of diagnosing or prognosing development or progression of esophageal cancer in a subject, comprising detecting an increase or decrease in protein level of at least one protein listed in Table 6 in a esophageal sample from the subject, when compared to a second esophageal sample.

15. A method of diagnosing or prognosing development or progression of invasive ovarian cancer in a subject, comprising detecting a change in expression level of one or more proteins of the subject, wherein at least one such protein comprises a tryptic fragment with a sequence of SEAFTQQLNSR, or is pyruvate kinase (index # 266427) or FK506-binding protein (index #3916257).

16. The method of claim 15, further comprising detecting a change in expression level of at least one additional protein selected from the proteins listed in Table 7.

17. A method for treating or inhibiting invasive ovarian cancer in a subject, comprising administering a therapeutically effective amount of a phannaceutical compound directed to restore normal expression of an invasive ovarian-cancer linked protein listed in Table 7, thereby treating the subject or inhibiting ovarian cancer in the subject.

18. The method of claim 17, wherein the method comprises: identifying a protein listed in Table 7 that is over- or under-expressed in a subject having invasive ovarian cancer; administering a pharmaceutical compound directed to restore normal expression of that protein; monitoring expression of the protein until expression approaches normal levels, thereby treating invasive ovarian cancer.

19. A method of screening for a compound useful in treating, reducing, or preventing invasive ovarian cancer or development or progression of invasive ovarian cancer, comprising determining if application of a test compound alters a protein expression profile so that the profile more closely resembles a normal ovarian protein profile than it did prior to such treatment, and selecting a compound that so alters the protein expression profile, wherein the protein expression profile includes at least pyruvate kinase (index # 266427), FK506-binding protein (index #3916257), or a protein comprising a tryptic fragment with a sequence of SEAFTQQLNSR.

20. The method of claim 19, wherein the protein expression profile further comprises at least one additional protein selected from the proteins listed in Table 7.

21. A method of diagnosing or prognosing development or progression of ovarian cancer in a subject, comprising detecting an increase or decrease in protein level of at least one protein listed in Table 7 in an ovarian sample from the subject, when compared to a second ovarian sample.

22. A method for treating or inhibiting breast cancer in a subject, comprising administering a therapeutically effective amount of a pharmaceutical compound directed to restore normal expression ofa breast-cancer linked protein listed in Table 8, thereby treating the subject or inhibiting ovarian cancer in the subject.

23. The method of claim 22, wherein the method comprises: identifying a protein listed in Table 8 that is over- or under-expressed in a subject having breast cancer; administering a pharmaceutical compound directed to restore normal expression of that protein; monitoring expression of the protein until expression approaches normal levels, thereby treating breast cancer.

24. A method of diagnosing or prognosing development or progression of breast cancer in a subject, comprising detecting an increase or decrease in protein level of at least one protein listed in Table 8 in a breast sample from the subject, when compared to a second breast sample.

25. The method of claim 24, comprising an increase or decrease in protein level in at least 5, at least 10, at least 15, at least 25, at least 50, at least 100, or at least 150 proteins listed in

Table 8.

26. The method of claim 24, which is a method of diagnosis of breast cancer, comprising detecting an increase in 19343 (gil 4769716); aldolase A protein (P04075); Annexin V (P08758); Annexin VII (P20073); bA139H14.1 (gi8217500); Cathepsin B (P07858); Cellular retinoic acid-binding protein 2 (gi4503029); Coagulation Factor (gi4503635); Coflin 1, Non-muscle (gi6680924); Endoplasmic reticulum lumenal protein (gi5803013); Ferritin (gil3655121); GDP- mannose pyrophosphorylase B, isoform 2 (gil 1761621); Globin, alpha-2 (HBA 2) (Q9Y5P6); Glucose-regulated protein precursor (grp 78) (PI 1021); Glycyl-tRNA Synthetase (gi493066); Hypothetical protein DKFZp761K0511.1 (gil 1277141); Hypothetical protein FLJ10652 (gi8922573); Hypothetical protein XP-033918 (gil4775389); isoAsp protein carboxyl methylotransferase (P22061); L-lactate dehydrogenase mu chain (P00338); Lamin A (P25391); Mn- superoxide dismutase (P04179); Nm23 (Q96NZ2); Non-metastatic cells 2, protein (NM23B) exp'd in (gi4505409); Omithine aminotransferase precursor (gi4557809); Peptidylprolyl isomerase B (gil4755229); Peroxiredoxin 1 (Q06830); Phosphatidylethanolamine-binding protein (Neuropolypeptide H3) (P30086); Prohibiting (P35232); Proteasome subunit gamma (Q12920); Protein phosphatase (gi7657475); Ras-related protein Rabl la (gi5453555); Rho GDI (P52565); S100Al l (S100C, calgizzarin) (P31949); S100A7 (P31151); Similar to hypothetical protein (gil3528771); Stathmin (gi5031851); Survival/ evasion peptide gene (Ay044239); Thioredoxin peroxidase 2 (Q06830); Thioredoxin peroxidase B (P48822); Transaldolase 1 (P37837); Triose- phophate isomerase (P00938); Tropomyosin 4 (P07226); Tropomyosin 4-anaplastic lymphoma kinase fusion protein (TPM4-ALK) (Q96PF4); tRNA pseudouridine synthase A (Q9Y606); Tyrosyl tRNA ligase (P54577); Ubiquitin (P02248); or two or more thereof.

27. The method of claim 24, which is a method of early diagnosis of breast cancer, comprising detecting a decrease in 7150 (gi4759461); Acri-Coenzyme A Dehydrogenase (gi4557235); Apolipoprotein Al (P02647); Collagen VI al chain precursor (P12109); Collagen VI alpha 3 chain precursor (P12111); Cytovillin 2 (gi340217, PI 5311); Dimethylarginine dimethylaminohydrolase 1 (gi6912328); Elastin B Precursor (gil 63026); Epidermal autoantigen 450k (clone pE450-B) (gi,1082355); Nucleoside diphosphate kinase (gil421609); Similar to P29266 (gi416873); Transferrin (P02786); or two or more thereof.

28. The method of claim 24, which is a method of early diagnosis of breast cancer, comprising detecting an increase in 14-3-3 zeta/delta (P29312); Alpha enolase (P06733); bA139H14.1 (Q9NTI9); Biotin Carboxylase (P50747); Carbonic anhydrase I (P00915); Cellular retinoic acid-binding protein 2 (P29373); Chloride intracellular channel protein 1 (P000299); Cofilin 1 (P23528); Crystallin, alpha B (P02511); Cyclophilin (Q08752); D-3-phosphoglycerate dehydrogenase (P043175); Elastin B Precursor (P15502); Elongation factor 1-alpha (P04720); Endoplasmin precursor (P14625); F-actin capping protein beta subunit, CapZ (P47756); Fatty acid binding protein 5 (P15090); GDP-mannose pyrophosphorylase B isoform 2 (Q9Y5P6); Glucose- regulated protein precursor (grp 78) (P 11021 ); Heat shock protein 27 (P04792); Heat shock protein 90 beta (P08238); Heterogeneous nuclear Rib nucleoprotein L (P14866); Heterogeneous nuclear Rib nucleoprotein X (G 971674); hn RNP A2 (P22626); hnRNP C (P07910); Human erythrocyte catalase 3-amino- 1,2,4-triazole complex, chain B (G 9257062); Hypothetical protein DKFZp761K0511.1 (Q9NTK6); Hypothetical protein FLJ10652 (Q9NVL6); KIAA0120 (P37802); Lactoferrin (diferric) (P02788); Lamin A (P25391); L-lactate dehydrogenase mu chain (P00338); Macrophage capping protein (P40121); Omithine aminotransferase precursor (P04181); Phosphatase 2C motif (P075688); Phosphoglycerate mutase B (P18669); Profilin 1 (P07737); Proteasome subunit, beta type, 9 (P28065); Rho GDI (P52565); Protein Kinase C substrate 80K-H (P14314);

S100 A8 (P05109); S100A11 (S100C, calgizzarin) (P31949); Similar to mRNA for KIAAO 120 gene GenBank Ace # D21261.1 (P37802); Stathmin (P16949); Superoxide dismutase, mitochondrial precursor (P04179); Thioredoxin peroxidase B (P48822); Tyrosyl tRNA ligase (P54577); Ubiquitin conjugating enzyme E21 (P50550); or two or more thereof.

29. The method of claim 24, which is a method of early diagnosis of breast cancer, comprising detecting a decrease in 19343 Fibrinogen Beta Chain (P19343); 3 -hydroxy isobutyryl- CoA hydralase (Q9LKJ1); Aconitase (Q99798); Adenylate Kinase 1 (P00568); Annexin A2 (P07355); Antithrombin III precursor (P01008); Apolipoprotein Al (P02647); Collagen VI al chain precursor (P12109); Cytovillin 2 (gi340217, P15311); Dimethylarginine dimethylaminohydrolase 1 (P094760); DNF 15S2 (P13798); Epidermal autoantigen 450k (clone pE450-B) (GI1082355);

Eukaryotic translation initiation factor 3 (*075822); Fatty acid binding protein (P15090); FK506 BP (P20071); High mobility group nonhistone chromosomal protein 1 (Q9NQJ4); KIAA0106 mRNA (P30041); Multicatalytic endopeptidase complex, proteasome precursor, beta subunit (023716); NADH dehydrogenase (ubiquinone) Fe-S protein (Q969P3); NADH - ubiquinone oxidoreductase 75 kD subunit (P28331); Neuropolypeptide H3 (P30086); Peptidylprolyl isomerase B (Q9BVK5); Peroxiredoxin 5 (P30044); polymeric immunoglobulin receptor precursor (P01833); Prostatic Binding protein (P08723); Selenium binding protein 1 (Q13228); Short chain 3-hydroxyacyl-CoA- dehydrogenase (Q16836); Similar to P29266 (gi416873); Superoxide dismutase 2 (P04179); Transferrin (P02786); Transgelin (Q01995); Transketolase (P29401); Tropomyosin (P09493); Tyrosine 3-monooxygenase/ tryptophan 5- monooxygenase activatioprotein (Q9CSZ6); Voltage- dependent anion channel protein 1 (P21796); or two or more thereof.

30. The method of claim 24, which is a method of detecting progression of breast cancer, comprising detecting an increase in L-lactate dehydrogenase mu chain (P00338); Lamin A (P25391); Rho GDI (P52565); S100A11 (S100C, calgizzarin) (P31949); Tyrosyl tRNA ligase (P54577); or two or more thereof.

31. The method of claim 24, which is a method of detecting progression of breast cancer, comprising detecting a decrease in Apolipoprotein A 1 (P02647); Caireticulin precursor (P27797); Collagen VI al chain precursor (P12109); Thioredoxin peroxidase B (P48822); Transferrin (P02786); or two or more thereof.