Eighth Principal Investigators Meeting

Aleksandar Milosavljevic, Oliver Hampton, Adrian Lee

Department of Molecular and Human Genetics, Human Genome Sequencing Center, Baylor College of Medicine, Houston, Texas

The general aim of our project is to employ highly multiplexed, next-generation genomic sequencing technologies to map chromosomal aberrations in cancer. The proposal aims to develop methods that map rearrangements in pools of bacterial artificial chromosomes (BACs) and then scale the methods to the whole-genome level. This technology development project uses both the commercially available BAC library obtained from the well-established MCF-7 breast cancer cell line as well as genomic DNA from an extant MCF-7 cell line. The BAC library was obtained from the original tumor tissue and thus may lack aberrations that may have accumulated in the cell line over decades.

The next-generation technologies will allow mapping of extensive collections of tags sequenced from ends of clone inserts. Genomic aberrations in pooled BACs and whole genomes will be detected by analyzing patterns of tag mapping. The method will provide an extensive view of chromosomal aberrations across multiple cell lines and primary tumors at an unprecedented level of resolution. A dense set of mapped tags will allow precise delineation of fused genes and the design of PCR-based assays for economical detection of aberrant joins. The biological function of rearrangements will be tested by cloning and expression in both immortalized and transformed breast epithelial cells. Biological assays will depend on the rearrangement or breakpoints of interest but will focus on proliferation, survival, invasion, and transformation.

The developed methods will open new opportunities for understanding recurrent chromosomal rearrangements and molecular mechanisms, identifying drug targets, and developing biomarkers for effective tumor-specific therapy.

Brian M. Balgley¹, Shan-Rong Shi², Li Yang³, Wang Weijie1, Tao Song1, Cheng Lee³, Clive Taylor<sup>2

1</sup>Calibrant Biosystems, Inc., Gaithersburg, Maryland; ²University of Southern California, Los Angeles, Los Angeles, California; ³University of Maryland, College Park, Maryland

Introduction: Histopathology plays a major role in current-day pathological diagnoses in which immunohistochemistry (IHC) is a critical tool for evaluating protein expression at a cellular level where morphological information is preserved. A drawback to IHC is that it has been impossible to predict which proteins or, more specifically, which epitopes can be unmasked for subsequent immunodetection. We present our findings cataloging indepth thousands of proteins and tens of thousands of potential epitopes unmasked under varying antigen retrieval (AR) conditions, including pH of the extraction buffer and application of heat during the AR process.

Methods: Paired fresh-frozen and formalin-fixed human liver samples were analyzed. The fresh-frozen sample was subject to extraction using RIPA buffer. The formalin-fixed samples were subject to AR using 2% SDS in Tris-HCl at varying pH, including 2, 7, 9, and 12. AR was performed by heating the samples at 100 °C 20 minutes then reducing the temperature to 60 °C for 2 hours. AR was performed on an additional pH 7 sample without heat. All samples were dialyzed to remove detergents, then denatured, reduced, alkylated, and digested with trypsin. The resulting peptide mixtures werefractionated by transient capillary isotachophoresis/capillary zone electrophoresis. Approximately 18 fractions were taken of each sample. Each fraction was then analyzed by nanoscale reverse-phase liquid chromatography coupled by electrospray ionization to a linear ion trap mass spectrometer performing tandem mass spectrometry. All samples were analyzed in duplicate. Tandem mass spectra were searched using the OMSSA algorithm and were thresholded to accept hits only below a 1% false discovery rate as determined by the decoyed sequence library search.

Results: Proteins recovered from formalin-fixed liver samples using the AR method compare favorably with those retrieved from fresh tissue using RIPA buffer. The fresh-frozen tissue yielded more than 3,500 proteins, and the pH and pH 9 heated extractions and the pH 7 unheated extraction each retrieved more than 2,500 proteins. The pH 2 and pH 12 extractions yielded fewer proteins, 2,300 and 1,400, respectively. Although clear differences were seen in the physical properties (pI and length) of the peptides extracted by each condition, very little difference was seen in the unique proteins extracted by each condition. That is, proteins from the worst performing extractions, pH 2 and pH 12, were almost wholly contained within the set of proteins extracted from the pH 7 heated and unheated extractions, which in turn were almost wholly contained within the set extracted from fresh-frozen tissue. In each condition, the predicted proteins showed no bias by molecular weight, isoelectric point, or hydrophobicity. Overall, more than 15,000 tryptic peptides, each corresponding to a potential antibody epitope, were identified for more than 3,000 proteins extracted from formalin-fixed human liver. This catalog offers the possibility of predicting which antigen epitopes are available for IHC under varying AR conditions. This is useful for minimizing the trial and error that is currently commonplace when performing IHC and also offers the possibility of raising new antibodies to specific epitopes known to be unmasked by the desired AR condition.

Jingfang Ju, Yaguang, Xi

Cancer Genetics Laboratory, Mitchell Cancer Institute, University of South Alabama, Mobile, Alabama

Translational control plays a key role during development, cell cycle control, and mechanisms related to acute drug resistance. In particular, noncoding miRNAs can potentially regulate more than 30% of gene expression at the translational level. Gene expression analysis of actively translated mRNA transcripts provides a unique approach to study posttranscriptional regulation. The current method relies on a traditional sucrose gradient ultracentrifugation procedure to isolate polysome complexes and requires a large number of cells (up to 500 million cells). As a result, this still remains a major bottleneck for the investigation of posttranscriptional regulation with limited quantities of clinical samples. Therefore, there is an urgent need to develop a novel approach to isolate actively translated polysomes from a small number of cells. The new approach will allow us to systematically study translational regulation with limited clinical samples. It has been shown that actively translated mRNAs are associated with multiple units of ribosomes, and the newly synthesized polypeptides are closely associated with molecular chaperones such as hsp73. These molecular chaperones assist in the proper folding of nascent polypeptides into higher ordered structures. These chaperones will provide the anchor to separate actively translated mRNAs associated with polysomes from free mRNAs. Affinity antibody capture beads will be developed to capture hsp73 chaperones associated with the polysome complexes so that all polysomes can be separated from monosomes and free mRNAs. The isolated actively translated mRNAs will be used for high-throughput gene expression analysis. This technology will make investigation of translational control feasible from limited clinical specimens.

Lance L. Munn

Massachusetts General Hospital/Harvard Medical School, Charlestown, Massachusetts

The isolation of rare cells from peripheral blood is a difficult and tedious step in the diagnosis and treatment of cancer. Examples include leukocytes, which are of interest in leukemia; circulating endothelial cells, which can serve as a surrogate marker for solid-tumor treatment; and circulating cancer cells, which may provide a tool for diagnosis and prognosis. In most patients, each of these cell types is outnumbered by red blood cells by a factor of at least 1,000. Therefore, isolation of these cells from a sample of whole blood is the required first step of many clinical and basic research assays. We recently described a microfluidic device that takes advantage of plasma skimming and leukocyte margination—intrinsic features of blood flow in the microcirculation—to enrich nucleated cells such as leukocytes directly from whole blood. It consists of a simple network of rectangular microchannels manufactured using standard photolithography and silicone molding techniques and requires only pressure-driven flow to operate. Its initial channel is designed to enhance lateral migration of spherical cells, which, once near the wall, are easily extracted through small extraction channels. In our preliminary design, a single pass through the device produced a 34-fold enrichment of the leukocyteto- erythrocyte ratio.

We propose to further develop the microfluidics to provide simple, efficient, and inexpensive technology for use as an initial stage in lab-on-a-chip analyses. Its integration into microanalytical devices that require rare cell enrichment will provide less expensive, more reliable clinical assays that are also convenient and portable for point-of-care testing. Specifically, we will maximize the purity and efficiency of nucleated blood cell isolation and compare the performance with traditional separation techniques. When fully developed, this technology will be a necessary and integral component of any microfluidic device analyzing nucleated blood cell populations by eliminating the need for preliminary blood processing steps.

Richard J. Cote¹, Y-C Ta², Jiang (John) Zhong¹, Henry K. Lin1, Siyang Zheng<sup>2

1</sup>Keck School of Medicine, University of Southern California, Los Angeles, California; ²Division of Engineering and Applied Sciences, Department of Electrical Engineering, California Institute of Technology, Pasadena, California

Sensitive detection of earliest metastatic spread of tumor in a minimally invasive and user-friendly manner will revolutionize the clinical management of cancer patients. The current methodology for circulating tumor cell (CTC) capture and identification has significant barriers, including multiple procedural steps, handling of relatively large volumes of blood, substantial human intervention, extremely high cost, and, importantly, lack of reliability and standardization for the detection methods. We report the development and optimization of a novel parylene membrane filter microdevice with a manual syringe injection system for capturing CTCs from undiluted human peripheral blood, which is capable of greater than 80% recovery with high enrichment factor and outperforms most current methods used in the field. Moreover, less than 10 minutes is required for each CTC capture operation compared with current multistep processing requiring more than 1 hour. We have demonstrated a superior recovery rate in comparison with the FDA-approved Cellsearch® system. We have also characterized our filter-based microdevice with a pressure-controlled system in a model system using cultured cancer cells admixed in blood. The effects of changing flow rate, fixative concentration, fixation time, blood dilution factor, and delivery pressure on this novel filter system will be reported. In addition, both off-chip DNA and RNA amplification from captured tumor cells has been demonstrated to work successfully with the potential of on-chip integration.Blood drawn fromcastration-resistant prostate cancer patients have also been studied, and we have demonstrated the feasibility of using our microdevice to work with clinical specimens. This novel filter-based CTC enrichment device will provide a cost-effective method for CTC monitoring, with a higher recovery rate, faster processing time, and more reliable results due to minimal human intervention.

This work was supported by R21-CA123027-01. Funds for the project were made available on 9/26/2006.

Rick A. Rogers, Renee Dickie, Glen Deloid, Jean Lai, Rosalinda Sepulveda

Biomedical Imaging Laboratory, Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts

We are developing sample preparation techniques that permit high-resolution analysis of tumorassociated microvasculature for the evaluation of vascular change in cancer. Source material will be tissue samples from animal models and fixed archival specimens. It is anticipated that these specimen preparation protocols will produce image data of sufficient quality, and at sufficient tissue depth, to enable automated three-dimensional (3D) computer analysis of vascular morphology within neovascularized peritumoral and viable tumor tissue. Vascular morphology abnormalities are characteristic of cancerous tumors and are associated with a switch from quiescent to aggressively invasive, metastatic behavior. Thus, angiogenic activity, as reflected by morphological microvascular change, is a critical point of assessment in cancer research. Visualization and quantification of microvascular attributes permit monitoring of disease progression and response to therapy. Yet currently available sample preparation methods to produce data at capillary-resolution and suitable for automated 3D quantitative analysis are not well established.

Our goal is to create and refine specimen preparation techniques that allow collection of highresolution 3D vascular image data of tumors and the supporting peritumoral tissue. Simultaneous immunohistochemical labeling of related molecules of interest will advance angiogenesis research in animal model systems. The desired outcome of this project is to develop specimen protocols that produce image data of sufficient quality to allow automated computer analysis of morphological attributes describing the 3D microvessel architecture associated with tumor growth and development.

Our methods will combine protocols used in vascular biology and adapt procedures developed in other fields, such as bioimaging, computer science, and medicine. To accomplish this task, we will undertake perfusion-based preparation methods, combined with optical clearing methods, to produce serial section image data used to generate 3D reconstructions of the microvasculature of whole-mount specimens, while preserving the ability to label other molecules of interest. We will first optimize perfusion-based preparation methods for vessel filling to provide robust imaging of microvascular beds. Specimens prepared in this manner will also furnish comparative data for evaluating the success of our second goal, whole-mount endothelial labeling, which will provide the flexibility to use fixed, archived tissue. To accomplish this aim, we will employ non-perfusion-based, diffusive whole-mount preparations. Methods for labeling other structures of interest will then be incorporated into these two types of protocols. In all cases, image data quality will also be evaluated with respect to its suitability for automated analysis of vascular morphology. Because of the time and expense entailed in generating tumor tissue, specimen preparation methods will be developed and evaluated in healthy tissue prior to being applied to tumor tissue.

Timothy J. O’Leary, Jeffrey T. Mason, Carol B. Fowler, David L. Evers, Robert E. Cunningham

Armed Forces Institute of Pathology, U.S. Department of Veterans Affairs, Rockville, Maryland

The identification of molecular changes such as mutations or altered expression underlying malignancy has led to specific and effective treatments for several forms of cancer. When the clinical course is typically short, it is reasonable to use fresh or frozen tissue for which molecular biologic analyses are relatively straightforward. However, many cancers (such as breast cancer) have a time course in which many years may elapse between treatment of the primary tumor and the appearance of metastasis. In these cases, the time required to obtain clinical correlations could be significantly reduced by performing molecular analyses on formalin-fixed, paraffin-embedded (FFPE) archival tissues for which the time course of the disease is already established. However, this is not currently feasible because both genomic expression and proteomic analyses using this tissue are less than robust, particularly for low-abundance transcripts and proteins.

The long-term goal of our research program is to employ high-throughput proteomic and molecular biologic screening of archival tissue specimens to identify the proteomic and genetic signatures of cancer. To achieve this goal, we are attempting to gain a detailed understanding of the chemical reactions and physical processes that occur from the time a tissue is first exposed to formaldehyde through the dehydration and embedding process and subsequently during efforts to extract molecules for proteomic and genomic analyses. A variety of chemical tools, including gel and liquid chromatography, mass spectrometry, and circular dichroism spectroscopy, have been used to follow these changes. Among these observations are the following:

1. It is possible to create tissue surrogates of one or two proteins that can be used to study in detail the chemical processes that occur during tissue fixation, dehydration, and embedding.
2. The process of high-temperature antigen retrieval, which is used to improve the immunocytochemical reactivity of proteins, results in the reversal of some, but not all, of the crosslinks and methylol adducts formed when proteins react with formaldehyde.
3. Dehydration with alcohol causes additional chemical changes beyond those that occur when formaldehyde reacts with proteins or nucleic acids in water. These changes are less readily reversible than are the changes induced by aqueous fixation.
4. Reaction of formaldehyde with proteins results in increased molecular heterogeneity, which is likely to significantly complicate proteomic analyses under the best of circumstances.
5. Solvent systems that are commonly used for molecular biologic analyses of fresh or frozen tissues are likely to be nonoptimal for recovery of DNA and RNA from formalin-fixed tissues.

To date, this work has enabled identification of some highly effective methods for isolating proteins from FFPE tissue and at the same time has highlighted some difficulties associated with use of these extracts in less than thoughtful proteomic analyses. We expect that our efforts will lead to further improvements in protein recovery for proteomics, as well as improved methods for recovering RNA for gene expression analysis.

Andrew Ellington, Austin Matt Levy, Ted Chu, Supriya Pai, Amos Yan, Dmitriy Ovcharenko, Jessica Ebright

Department of Chemistry and Biochemistry, College of Natural Sciences, The University of Texas at Austin, Austin, Texas

The Ellington lab has selected aptamers against a wide variety of tumor cells and has shown that these aptamers can be used to specifically label cells. By adapting aptamers to the proximity ligation assay, it has proven possible to identify as few as 10 tumor cells against a background of 100,000 noncognate cells. These methods may therefore prove very useful for the identification of circulating tumor cells in serum samples. In addition, aptamers that bind to cell surfaces have been shown to internalize into cells and can carry cargoes such as toxins and siRNAs. Such internalizing aptamers may prove to be extremely useful for the delivery of antitumor and other siRNAs and RNA therapeutics. We also are developing cell chips that can be used to reproducibly assay reagents, identify novel biomarkers, and help determine strategies for typing and staging tumors.

Binghe Wang

Department of Chemistry, Center for Biotechnology and Drug Design, Georgia State University, Atlanta, Georgia

Glycoprofiling is a very important way to study the correlation of glycosylation patterns and pathological states. This is currently done with lectins, which are sugar-binding proteins. However, several issues hinder the application of lectins in glycoprofiling, including limited availability of lectins and their limited structural specificity. Our lab has been working on the development of lectin mimics using both small-molecule sugar sensors and DNA-based fluorescent aptamers. This presentation will discuss the latest results from our lab in developing fluorescent lectin mimics for glycorecognition and glycoprofiling applications. The end products of our platform technologies will be very useful to glycobiologists interested in studying glycoprofiling and correlation of glycosylation pattern variations with biological and pathological processes.

David Andrew Largaespada¹, Lara S. Collier¹, Laura E. Green¹, Eric P. Rahrmann¹, Timothy K. Starr¹, Raha Allaei1, Adam J. Dupuy², Nancy A. Jenkins³, Neal C. Copeland³, Paul C. Marker<sup>1

1</sup> Department of Genetics, Cell Biology, and Development, University of Minnesota Cancer Center, Minneapolis, Minnesota; ²University of Iowa, Iowa City, Iowa; ³Institute of Molecular and Cellular Biology, Republic of Singapore

Recent large-scale sequencing studies have identified an average of 90 somatic mutations in a typical epithelial tumor. The next step must be to distinguish passenger from driver mutations in carcinogenesis. In this study, we employ a synthetic DNA transposon called Sleeping Beauty (SB) to create random somatic mutations in the prostate of mice, which we hypothesize will result in prostate cancer. This approach utilizes somatic transposition of an SB transposon vector, called T2/Onc, which can induce gain-of-function mutations or loss-of-function mutations upon insertion within or near genes in living mice. Thus, T2/Onc is capable of activating oncogenes or disrupting tumor suppressors. In this study, T2/Onc mutations are directed to the prostate using a CRE-controlled SB transposase knockin allele called Rosa26-LoxP-STOP-LoxP-SB11 (R26-LSL-SB11) in combination with a CRE recombinase transgene driven by the rat probasin promoter (PB-CRE) plus the T2/Onc transposon transgene array. This experiment is being done in otherwise wild-type mice and in mice homozygous for the Ptenflox conditional knockout allele. We hope to determine genes and genetic pathways that can induce prostate cancer in cooperation with loss of PTEN, a common event in human prostate cancer. Thus, the samples generated for this work are mouse prostate tumors. These samples will be analyzed by shotgun cloning and sequencing of T2/Onc transposon insertions. The output is a list of genes that are candidate prostate cancer genes. Proof-of-principle data have been obtained by studying early prostate lesions that develop in T2/Onc mice in which the SB11 transposase protein is expressed ubiquitously from an R26-SB11 transgene. These mice succumb to lymphoma, but lymphomatous mice also show an increase in the frequency of nests of PIN-like, Ki67+ prostate epithelial cells when compared with control prostates. These lesions harbor T2/Onc insertions, a subset of which we have shown to affect the same genes—thus, defining common insertion sites (CISs) and associated cancer gene candidates. Tissue-specific SB mutagenesis studies also are being conducted in other organs of the mouse to attempt to model hepatocellular carcinoma, lung cancer, and gastrointestinal (GI) tract cancer. Current results from the GI tract model are promising. We have combined T2/Onc, R26-LSL-SB11, and the Villin-Cre transgenes. More than 50% of the experimental animals have died between 9 and 15 months compared with 15% of control mice. The majority of experimental mice had GI tract tumors ranging from a few polyps to more than 100 polyps. Several animals also had liver tumors in addition to GI tract neoplasia. We are currently attempting to map transposon insertion sites from multiple independently generated tumors and identify CISs. These studies are conducted in mice, which may develop some cancer types along genetic pathways not relevant to human cancer. Also, it is possible that some genes cannot be mutated by T2/Onc insertion in a way that can contribute to cancer. Nevertheless, these studies provide the advantage of allowing an in vivo functional screen for potential cancer genes. The results of the in vivo SB-based screens will be useful for scientists searching for cancer-causing events in human samples by exon resequencing, expression microarray, comparative genome hybridization, and other similar strategies.

Faqing Huang, Yan-Lin Guo

Department of Chemistry and Biochemistry, University of Southern Mississippi, Hattiesburg, Mississippi

Posttranscriptional gene silencing by small interfering RNA (siRNA) has evolved into a powerful tool for downregulation of any target gene(s), with both high efficiency and sequence specificity. In principle, siRNA may become the basis for developing the next generation of anticancer agents with high potencies and low side toxicities. However, no therapeutically acceptable delivery methods of siRNA are currently available. Folate receptor (FR) has been shown to be highly expressed by some cancer cells to facilitate the uptake of folate to meet its increased need for biosynthesis, although most normal cells do not express FR. Therefore, FR can be used for targeted drug delivery to certain cancer cells. FR-mediated delivery of a number of chemical agents by endocytosis to cancer cells has been demonstrated to be both efficient and highly specific toward FR-positive cancer cells. In principle, folate can be linked to siRNA to achieve specific siRNA delivery to FR-expressing cancer cells. However, the chemistry of direct coupling between folate and siRNA has not been readily achieved until now. Capitalizing on our recent RNA bioconjugation methods, the current research will develop a novel folate receptor-based siRNA delivery strategy against specific target genes in FR-expressing cancer cells. First, we will chemically synthesize the transcriptional initiator folate-AMP conjugate. Second, a folate-conjugated siRNA against a cancer cell marker, urokinase plasminogen activator receptor (uPAR), will be prepared by our T7 transcription system. Using KB cancer cells (a human nasopharyngeal epidermoid carcinoma cell line) and RNA fluorescent labeling, we will then determine folate-mediated siRNA delivery efficiency and cell specificity (normal cells vs. cancer cells). After successful FR-mediated siRNA delivery, RNAi effects will be assessed by determining the expression of uPAR at both mRNA and protein levels by quantitative RT-PCR and Western blot analysis. Finally, we will test the hypothesis that downregulation of uPAR expression in cancer cells by folate-conjugated siRNA will effectively inhibit the cellular activity of uPAR associated with tumor growth. We expect that the delivery of folate-siRNA against uPAR in KB cells will be more efficient and specific than other currently available methods. As a consequence, tumor growth suppression through uPAR silencing is expected. Results from the proposed research will likely lead to the development of general strategies and methods for FR-mediated delivery of siRNA against specific target genes in FR-expressing cancer cells and, therefore, may lead directly to cancer therapeutic applications.

Han Cao, Paru Deshpande, Mike Austin, Ming Xiao, Mike Boyce-Jacino

BioNanomatrix, LLC, Philadelphia, Pennsylvania

We are developing a nanochip device for manipulating long genomic DNA for high-resolution (kilobase), whole-genome analysis of cancer biomarkers such as gene amplifications, deletions, and translocations. These chromosome structural aberrations are strongly implicated in the process of malignant transformation and are important diagnostic, prognostic, and therapeutic indicators for many types of cancer.

Techniques that rely on probing chromosomes, such as metaphase fluorescence in situ hybridization, although providing a pangenomic view, cannot resolve structures below the Mb range. By probing uncompressed interphase DNA, resolution can be improved, but spatial organization of the genome is lost, so multiplexed and quantitative information is difficult to obtain. By stretching out DNA using techniques such as “molecular combing” or “optical mapping,” it is possible to probe specific loci in a spatially significant way, with resolutions in the kb range. However, techniques for mechanically fixing DNA on surface are inherently variable, leading to inconsistent stretching of molecules, which often cross over and retract on themselves. This makes it difficult to standardize such techniques as highthroughput methods for the biomedical community.

We have found that an individual dsDNA molecule, essentially a long polymer chain, will elongate and linearize in a consistent manner when streamed into nanofabricated long confining nanochannels with extremely small diameters. In the past 8 months, we have set up a single-molecule imaging lab, developed novel nanofabrication methods to reliably manufacture centimeter-long nanochannels in solid-state material, and demonstrated that megabase-size human genomic DNA can be consistently streamed and analyzed in these nanochannels by direct imaging at the single-molecule level. Conventional genomic DNA sample preparation methods were used, a very small amount of sample is needed, and in vitro cell culturing and/or subsequent PCR amplification are not required steps. This not only reduces cost and time but also, most importantly, allows fast analysis of samples of a scare source, such as needle aspirates, to obtain the most accurate information without introducing artifacts during sample/clonal amplification. Since an intact single molecule is analyzed in its individual channel, this provides the opportunity to interrogate haplotyping information directly from the native patient sample.

Working with our collaborator, we have also obtained preliminary data on imaging locus-specific probes along linearized long genomic DNA molecules as well as short PCR amplicons. This technology does not limit the capability of using PCR products on our chip in case specific target areas from complex or pooled samples are analyzed.

Flexible chip design could allow many samples to be analyzed in parallel in individually registered fluidic compartments on the same chip. DNA region-specific signature probes and multicolor labeling can be incorporated to interrogate the fine details of genomic structural information.

Our ultimate goal is to develop the nanochannel array chip integrated system for the routine and standardized quantitative analysis of genome structure, which will enable archiving and crosslaboratory comparison of data in cancer diagnostics.

Ian Burbulis, Roger Brent, Kumiko Yamaguchi, Orna Resnekov, Richard Yu

The Molecular Sciences Institute, Berkeley, California

There is great optimism that different cancer disease states will be assessed by quantifying changes in the abundance of human analytes, including posttranslationally modified proteins involved in key signal transduction pathways. We previously developed a technology that measures the abundance of specific analytes in biological samples, such as serum, using protein-DNA fusions known as tadpoles [1]. This technology quantifies multiple analytes over six to eight orders of magnitude, with limits of detection in the hundreds of molecules. Outlined below is our recent progress in accomplishing our NCI-funded goals:

1. We have developed and validated three dsDNA constructs for assembling multiplex assays and two internal amplification control DNAs. We have found that internal calibration templates exceeding 1,000 copies reduce the amplification efficiency of other templates in the same reaction regardless of whether the assay and control templates are amplified by the same primers. Thus, having the number of calibration templates at or below 1,000 copies ensures accurate and precise quantification even in the presence of each other.

2. We have developed a universal antibody-counting tadpole for implementing existing antibodies that bind known cancer markers. Using this tadpole, we measure human interleukin-6 and human prostate specific antigen in human serum as a test system [2].

3. We further extend the use of this universal antibody-counting tadpole to elucidate intracellular signaling state. As a model system, we used this tadpole with specific capture IgY to quantify 1,000 MAPK proteins per cell from as few as 50 yeast cells. Using rabbit antiphospho-p42/44 IgG, we quantify the amount of phosphorylated MAPK produced in response to pheromone exposure for 15 minutes. Our platform will enable the signaling state of primary tumor cells to be detected, mapped, and quantified.

4. We have developed antibody-DNA conjugates by attaching DNA directly to heavy chain carbohydrates, which do not interfere with antigen binding. These reagents serve as alternative detectors for quantifying antigens without the need to develop new affinity proteins, as is necessary for implementing the tadpole. We are optimizing conditions necessary to achieve multiplex quantification of CA125, prolactin, osteopontin, insulin-like growth factor II, and leptin in human serum.

1. Burbulis I.E., Yamaguchi K., Gordon A., et al. Using Protein-DNA Chimeras To Detect and Count Small Numbers of Molecules. Nat Methods 2(1):31-7, 2005.
2. Burbulis I.E., Yamaguchi K., et al. "Quantifying Small Numbers of Molecules With a Near-Universal Protein-DNA Chimera." Nat Methods, submitted.

Jeffrey E. Segall

Albert Einstein College of Medicine, Bronx, New York

Current metastasis assays tend to evaluate one cell line at a time, resulting in the use of large numbers of nonhuman animals due to the variability in measurements from animal to animal. This project will develop methods for parallel analysis of multiple cell lines for metastatic properties in a single animal, with the goal of evaluating up to 50 genes at a time. The development of such an approach would allow the screening of genes for metastatic effects. Such a screen could indicate the contributions of proteins to the steps of primary tumor growth, intravasation, and lung colonization. The first aim will focus on production of pools of cells expressing or suppressing selected proteins. The second aim will evaluate detection technologies for measuring construct distributions in a pool. The third aim will determine the appropriate formation of pools to be screened. The fourth and fifth aims will perform an initial screen and validate candidates identified in the screen. Sample preparation will involve isolation of viable tumor cells from the primary tumor, blood, and lungs of animals carrying breast cancer tumors. The likely end use of the data is determination of whether specific genes contribute to tumor formation, intravasation, or lung metastasis. Cell lines will be used for production of pools of cells expressing or suppressed for expression of particular proteins. Endogenous variations in cell lines can affect our evaluation of the contributions of the targeted proteins to metastasis.

John Welsh, Gaelle Rondeau, Rosana Risques, Martin Judex

Sidney Kimmel Cancer Center, San Diego, California

We have developed two technologies for the analysis of tumors. The first is a method we refer to as “vertical arrays,” which are similar to dot blots but use a microarray format and low-complexity representations (LCRs) of total mRNA as the spotted features. These modifications result in improved signal-to-noise ratio, portability, and other advantages of glass slide microarrays. Vertical arrays can be used to assess the expression behavior of a gene in thousands of biological samples. The goal of our research is to prepare vertical arrays from tumor specimens so that the expression of a gene that is implicated in one tumor can be assessed quickly and conveniently in thousands of other tumor specimens, including tumors of the same and different types. Addressing the generality of a molecular marker is an important step in biomarker validation.

We have recently completed “spike-in” sensitivity and coverage measurements for vertical arrays. “Sensitivity” refers to the number of molecules in a sample for which a change of a certain magnitude can be measured reliably, and “coverage” refers to the number of LCRs needed to measure most of the transcripts in a cell. We find that fourfold changes can be measured very reliably in the neighborhood of less than 1 transcript per cell for 80% of the spike-in transcripts using six LCRs, such that 11 LCRs should cover about 95% of all transcripts in or above this range. The major strengths of this approach are measurement of changes in transcript steady-state levels comparable to quantitative PCR, with the associated conveniences of the microarray format. Specifically, vertical arrays prepared from a common set of tumors could be widely distributed with greater economy than achievable with, for example, microtiter plates for quantitative PCR. Also, all aspects of the preparation of a vertical array are performed robotically, including sample preparation, LCR preparation, and microarray spotting.

One of the obstacles to constructing such a pan-cancer vertical array is assembling a collection of thousands of microdissected cancer samples. For this and other purposes, we have devised a “transverse microtome,” which can tile tumor tissue sections into thousands of well-documented 50 x 50 x 10 micrometer voxels. The transverse microtome can perform about 40,000 cuts per day, and each tissue voxel is mapped to a composite photomicrograph of the entire tissue section. A graphic interface allows the user to select regions of interest from the composite image, and the selections are automatically tabulated in a form that allows a liquid-handling robot to access the corresponding samples. This will allow the user of the eventual pan-cancer vertical array to assess, for example, the ratio of tumor to stroma represented in an LCR spot. We are currently adapting molecular methods, including RNA arbitrarily primed polymerase chain reaction (used to make LCRs) as well as wholegenome amplification, which will be useful for experiments we contemplate to address intratumoral heterogeneity, to samples dissected using the transverse microtome.

Our IMAT project will establish proof of principle through the preparation of a small vertical array focusing initially on prostate cancer. The experiments that we have performed so far on LCR preparation and vertical arrays have used RNA from cells grown in culture and indicate that reproducible LCRs can be prepared from as few as 10 cells. However, we have not yet constructed a vertical array from samples taken from frozen sections of surgically resected tumors. Consequently, we do not yet have an estimate for the smallest amount of tumor tissue that will be needed per LCR.

In an eventual, large pan-cancer vertical array, tumors will be snap frozen as soon as possible after surgical resection. Every LCR will derive from a region of a tumor that is captured in a photomicrograph, and the more information that can be attached to such a sample, particularly the record of presurgical treatment and clinical outcome, the more likely it is that the array will be useful. Frozen tissues have poor morphology relative to fixed tissues, and RNA cannot be recovered from tissue stained with hematoxylin. Still, in many cases, the proportion of cancer to normal tissue can be determined with good accuracy, and molecular markers (e.g., distinguishing stroma from tumor) may partially solve this problem. Vertical arrays have an inherent self-correcting mechanism, in that mishandled samples would be expected to give rise to outliers. Thus, the high potential throughput of this approach is expected to provide a good buffer of average behavior for occasional mishaps. Nevertheless, good practice in sample handling will allow for the detection of outliers of biological significance.

Kristie M. Melnik1, Arfaan A. Rampersaud¹, Francesca Carpino2, Stephen Williams², Xiaoxia Jin², Maciej Zborowski², Ira Yudovin-Farber³, Abraham Domb<sup>3

1</sup>Columbus NanoWorks, Inc. (CNW), Columbus, Ohio; ²Cleveland Clinic, Cleveland, Ohio; ³The Hebrew University of Jerusalem, Jerusalem, Israel

Magnetic cell separation as a preenrichment step prior to downstream cellular characterization techniques is becoming more widely accepted. The aim of this project was to develop high-resolution magnetic reagents for use in the quadrupole magnetic cell sorter (QMCS). The QMCS is a flow-through immunomagnetic separation system that can provide sensitive enrichment of circulating tumor cells (CTCs) in blood as well as in other biological fluids and tissues. Defining the criteria to which magnetic reagents must adhere is essential in obtaining successful enrichments using magnetic separation technology. These criteria include high magnetic susceptibility with a well-defined shelf life, narrow particle size distribution, high-density attachment sites for antibodies, and a low level of nonspecific binding to nontargeted cells. CNW has met these criteria through the use of novel techniques, and quantitative specifications of particle manufacturing have been established. Magnetic field-flow fractionation (MgFFF) was used as a means of verifying the mass of iron, including and validating the shelf life of CNW materials. MgFFF is an analytical separation and characterization technique for nanosize and microsize magnetic particles. It is a separative elution technique similar to chromatography in which different components of a small sample elute from the channel at different times. A longer retention time in the MgFFF is identified with particles that interact strongly with the magnetic field and have high magnetic susceptibility. Furthermore, size (or mass) distribution of theinclusions can be determined from the elution profile using a data reduction method. Results from MgFFF indicate that CNW materials have a size distribution of 200 to 400 nm and an average retention time of 15 minutes, indicating high magnetic susceptibility within the magnetic field. MgFFF results also showed that there was no change in the elution profile over time, demonstrating that coating techniques used on CNW materials prevent the iron from oxidizing and losing magnetism.

Nonspecific studies were also performed with CNW particles. Jurkat cells were incubated with varying concentrations of CNW particles for 30 minutes and subsequently washed. Cells were measured by CTV, which measures the movement of cells within a well-defined magnetic energy gradient and generates a value known as the magnetophoretic mobility of a cell. Cells that moved above 2.5 x 10-5 mm3/T.A.sec were considered magnetic and therefore nonspecifically labeled with CNW nanoparticles. At higher particle concentrations, 48.8% of the cells exhibited a magnetophoretic mobility above 2.5 x 10-5 mm3/T.A.sec, indicating problematic nonspecific binding of CNW particles. When particle concentrations were reduced relative to cell concentration, the nonspecific binding measured by CTV was reduced to 5%. Chemistry modifications were pursued to reduce nonspecific binding of particles at higher concentrations while retaining a high specificity for targeted cells. To reduce nonspecific binding, modified polyethylene glycol (PEG) units were introduced by reductive amination. PEG-modified particles showed a reduction of nonspecific binding. At the higher particle concentrations, less than 1% of cells were determined to be magnetic and nonspecifically bound with the PEG-modified CNW particles.

Leslie B. Poole, Chananat Klomsiri, Kimberly Nelson, Sarah A. Knaggs, Jacquelyn Fetrow, Larry Daniel, Bruce King

Department of Biochemistry, Center for Structural Biology, School of Medicine, Wake Forest University, Winston-Salem, North Carolina

Signal transduction processes rely on a cascade of posttranslational modifications (PTMs), proteinprotein interactions, modulated catalytic activities, and translocations within a response network of interacting species to generate specific biological outputs. Based on the generation of hydrogen peroxide and other reactive oxygen species (ROS) through the “oxidative burst” that accompanies a number of receptor-mediated signaling processes, we propose that a largely overlooked PTM, cysteine oxidation to sulfenic acid (and subsequent disulfide bond formation), provides the major molecular mechanism through which redox-based modulation of phosphorylation cascades takes place.

Our new technology, a rapidly reacting chemical trap for sulfenic acids (R-SOH) on proteins [Poole et al. 2005], allows us to assess the molecular and spatial location and timing of sulfenic acid formation on proteins in a proteomics-friendly manner. The modifying agents, with fluorescent or biotinylated tags attached to an analog of dimedone, are uniquely reactive toward R-SOH and “lock in” this chemical information in cell-culture- and potentially tissue-derived proteins for later readout by gel and mass spectrometry (MS)-based methods. Proposed improvements to our technology (three-color fluorescent labeling for differential gel electrophoresis and isotope-coded samples for MS) will also allow for enhanced quantitative abilities to assess variations in levels of these modifications among different samples in multiplexed approaches. Bioinformatics tools are also being developed to elucidate “signatures” of reactive sites to better understand the basis for the specificity of given cysteines toward peroxide-mediated oxidation and to predict previously unknown reactive sites across the proteome. Although it is at an early stage, our research using these compounds indicates that an initial “burst” of R-SOH formation is observed within 1 to 2 minutes after addition of TNF-α; to HEK-293 cells. A growing list of oxidized proteins that we have identified using two-dimensional gel electrophoresis and MS methods includes a number of signaling-relevant proteins (including both phosphatases and kinases) as well as proteins with other functions; the majority of the proteins identified so far have not previously been known to be redox regulated.

As indicated above, implementation of our new technology will allow us to capture a new type of information on a redox PTM that can be used to report on the redox status or responsiveness of signaling-relevant proteins in given cell samples under controlled conditions. This novel technology is also likely to have broad applicability in molecular profiling to stratify patients with cancers that are sensitive to ROS-generating therapies and in the development of novel cancer therapies based on the inhibition of ROS-dependent proliferative signaling.

Poole L.B., Zeng B.B., Knagg S.A., et al. Synthesis of Chemical Probe To Map Sulfenic Acid Modifications on Proteins. Bioconjug Chem 16(6):1624-8 2005.

This work is supported by R21-CA112145 (LBP, Principal Investigator; JSF, LWD, SBK, Co-Investigators).

Luke Tolley

Department of Chemistry, Southern Illinois University, Carbondale, Illinois

Proteomics requires high-resolution separations due to the large number of different proteins in a typical sample. One method that has shown great promise for protein separation is capillary isoelectric focusing (IEF). Capillary IEF allows for a large number of proteins to be separated in a short period of time, but efficiently interfacing it to other separation or analysis methods is difficult and leads to a drastic reduction in resolution and sensitivity. Dynamic IEF is a new technique that overcomes many of the problems of capillary IEF by using a dynamic electric field within the capillary. Manipulation of the electric field using additional high-voltage power supplies permits adjustment of the pH gradient, enabling both the location and width of the focused protein bands to be accurately controlled. Each protein can be migrated to a designated sampling point, while remaining focused, where it can be collected for further analysis. This ability to collect and isolate the protein bands while maintaining a high peak capacity demonstrates the great potential of dynamic IEF for sample separation.

In addition to its utility for proteomics, dynamic IEF can also be used to identify active proteins in complex samples using bioassay-guided fractionation (BGF), which combines analytical separation methods with biological testing to identify compounds responsible for an observed effect. Dynamic IEF has the ability to fractionate a sample into more than 1,000 fractions with little overlap, providing unprecedented potential for BGF. To minimize the number of bioassays performed, successive fractionations will be used. The first round will fractionate a sample into 11 parts, and a Plackett- Burman design will be used to identify which fractions or combinations of fractions are active. Active fractions will be further fractionated using the same experimental design. Eventually, the active fraction will be greatly simplified compared with the original sample, and liquid chromatography/mass spectrometry (MS) will be used to identify the remaining proteins. The capabilities and benefits of dynamic IEF for both proteomics-guided fractionation and BGF will be demonstrated. Dynamic IEF is able to provide a peak capacity of greater than 1,000, as shown using prostate cancer cell lysates. Samples are then analyzed by MS, giving a total system peak capacity of more than 300,000.

BGF will be demonstrated using culture medium from Fusarium virguliforme, the fungus responsible for soybean sudden death syndrome. This fungus produces an unidentified protein toxin causing foliar necrosis. Data will be presented showing the ability of dynamic IEF to perform BGF and to isolate fractions spanning less than 0.01 pH unit.

Neal F. Gordon, James Graham, Jeffrey Radding, Tim Nadler, Cheryl Murphey

Epitome Biosystems, Inc., Waltham, Massachusetts

The genesis of cancer involves several genetic changes that culminate in uncontrolled cellular proliferation. Several key oncogenes and tumor suppressor genes are widely recognized to influence the development and progression of several cancers. This recognition has come in large part from numerous efforts that have defined key mutations in some of these genes associated with altered functions. However, some genes associated with cancer have no apparent mutations that alter their normal function. In these cases, there is evidence that splice variants of a given gene and the ratio of isoforms are significantly changed in the cancerous state. One of the major obstacles in elucidating the significance of cancer-associated splice variants is the lack of a systematic technology that can reliably detect and quantitate variants at the protein level.

Antibody-based methods to measure proteins are powerful, are extensively used in research and diagnostic applications, and fit well in a clinical setting. However, development of immunoassays to discriminate among protein isoforms is challenging due to the large overlap in sequence. Epitome Biosystems, Inc., has developed methodologies that solve this problem through a combination of antibody design and sample treatment aimed to expose differences among protein isoforms. Starting with in silico techniques to identify continuous linear sequences for any protein, unique EpiTags™ are generated. Antibodies are raised against synthetic peptides that make up these unique EpiTag™ sequences rather than against the protein itself. EpiTags™ are made accessible to the antibody by fragmenting proteins in the sample prior to analysis, yielding predictable antibody performance.

For protein isoform detection, a “sandwich” assay format is utilized for unambiguous detection and quantification of each variant form. Specifically, the antibody sandwich is formed by antibodies raised to two different EpiTags™ that exist within a single protein fragment, liberated by protease digestion of the sample. The protein fragment is selected such that it spans the unique junction region between two exons. Importantly, although individual EpiTag™ sequences are shared between the full-length protein and one or more isoforms, the combination of both EpiTags™ on a single protein fragment liberated by protease digestion is unique. Following sample denaturation and digestion, each individual isoform generates a unique peptide fragment signature that can be quantified based on novel antibody sandwich pair formation. Therefore, Epitome’s approach allows for an overall systematic methodology for the detection of virtually any splice variant within the proteome.

Proof of concept has been demonstrated using a model system developed using commercially available peptide-specific antibodies and a set of corresponding synthetic peptides. Assays are being developed to measure the cancer-related splice forms of Bcl-X and CD44. Bcl-X is a member of the Bcl- 2 family of apoptotic regulators, is highly expressed in many lymphomas, and may have a significant role in the genesis of these cancers. This gene has two distinct splice variants (the long and short forms) with opposing apoptotic functions for which EpiTag™ assays are being developed. CD44 has a very large number of potential splice forms. In many human cancers, including carcinomas of the breast, certain CD44 variants are found to predominate. Epitome’s initial focus is on these forms, specifically, isoforms Meta-1 and Meta-2.

Norman J. Dovichi

Department of Chemistry, University of Washington, Seattle, Washington

The cell is the organizing unit of life, and characterization of single cells provides a level of detail and biological understanding that is lost when cell cultures and tissues are homogenized for analysis. Chemical cytometry employs microscale separation technologies and ultrasensitive laser-induced fluorescence to characterize the composition of single cells. Chemical cytometry reveals details on heterogeneous cell types, such as those found in tumors, neurons, and certain bacterial systems. Cellular heterogeneity may prove to be a fundamental characteristic of life, and understanding that heterogeneity provides insight into disease prognosis.

Philip J. Lee, Terry A. Gaige, Paul J. Hung

CellASIC Corporation, San Leandro, California

We are developing a microfluidic system to allow control of the cell culture environment for longterm, time-lapse microscopy of adherent cells. Building on previous work in microfluidic cell culture [1], the current method integrates a customized digital flow controller to deliver defined fluid inputs to the cultured cells during microscopy.

As the trend toward “systems biology” continues, it will become increasingly important to study dynamic behavior in individual live cells. The use of fluorescence microscopy offers a promising method to investigate live cell response. However, currently there is no system available to control the delivery of fluids to cultured cells. Current techniques use either static culture plates or bulky flow chambers.

We describe a microfluidic flow chamber that overcomes the major limitations of time-lapse microscopy studies. The microfluidic chamber is designed to have a cell culture region separated from the flow path by an artificial endothelial barrier as previously described [2]. The device is formatted to a standard 96-well plate, allowing liquid and cell samples to be directly pipeted into the appropriate inlet reservoirs. A custom pneumatic flow controller is then used to load the cells into the culture regions as well as to switch between two different exposure solutions. A digital software interface allows the user to program specific inputs (pulses, ramps, etc.) over time to expose the cells to complex functions during time-lapse imaging. Anticipated applications include gene expression kinetics, protein localization, apoptosis, and siRNA silencing.

A key aspect of this work is the potential to standardize live cell experiment data. Since each condition can be well controlled in the microfluidic format (cell type, cell number, temperature, time, flow solutions, exposure profile, etc), the reproducibility of cellular responses from lab to lab should be greatly improved. Toward this goal, we are also working on developing an open-source standard that describes the experimental conditions used to collect any given data. This will make cell biology experiments much more quantitative and portable between research groups.

The format of the microfluidic plate also lends itself to future automation in a screening format. Currently, the prototype plate can run eight independent flow experiments. By adopting this to a 384- well plate and maximizing the usage of each well, it is possible to create 96 independent controllable flow units on a single plate, which can be run using the most current robotic instruments and analyzed via high-content analysis. This can benefit the cancer biology field by enabling a type of analysis not currently available at any scale, much less for medium-throughput screening.

1. Hung P.J., Lee P.J., Sabounchi P., et al. A Novel High Aspect Ratio Microfluidic Design To Provide a Stable and Uniform Microenvironment for Cell Growth in a High Throughput Mammalian Cell Culture Array. Lab Chip 5(1):44-8, 2005.
   
2. Lee P.J., Hung P.J., Lee L.P. An Artificial Liver Sinusoid With a Microfluidic Endothelial-Like Barrier for Primary Hepatocyte Culture. Biotechnol Bioeng February 7, 2007.

Roger Cubicciotti1, Martin Guthold², Lu Peng², Brian J. Stephens^2,3, Keith Bonin<sup>2

1</sup>NanoMedica, Inc., Montclair, New Jersey²Wake Forest University, Winston-Salem, North Carolina; ³Vanderbilt University, Nashville, Tennesee

We are developing a method that utilizes a combined atomic force microscope (AFM)/fluorescence microscope and small-copy-number PCR to affinity-select individual aptamer species in a single cycle from a small pool of random sequence oligonucleotides. In this method, a library of small beads, each of which is functionalized with fluorescent oligonucleotides of different sequences, is created. This library of oligonucleotide-functionalized beads is flowed over immobilized target molecules on a glass cover slip. High-affinity, target-specific aptamers bind tightly to the target for prolonged periods and resist subsequent washes, resulting in a strong fluorescence signal on the substrate surface. This signal is observed from underneath the sample via fluorescence microscopy. The AFM tip, situated above the sample, is then directed to the coordinates of the fluorescence signal and is used to capture a threedimensional, high-resolution image of the surface-bound bead and to extract the bead (plus attached oligonucleotide). The extracted oligonucleotide is PCR-amplified, sequenced, and may then be subjected to further biochemical analysis.

We describe the underlying principles of this method, the required microscopy instrumentation, and the results of proof-of-principle experiments. In these experiments, we selected aptamers in eight trials from a binary pool containing a 1:1 mixture of thrombin aptamer oligonucleotide and a nonsense oligonucleotide. In each of the eight trials, the positive control aptamer was successfully detected, imaged, extracted, and characterized by PCR amplification and sequencing. In no case was the nonsense oligonucleotide selected, indicating good selectivity at this early stage of technology development.

Ye Sun, Val Golovlev

SciTec, Inc., Knoxville, Tennessee

A new highly sensitive and cost-effective enzymatic luminescence assay for high-throughput detection and quantification of microRNA in biological samples will be developed in this project. The assay utilizes the same detection concept known from pyrosequencing, yet expands the pyrosequencing detection methodology for an accurate quantification of small RNA molecules. This technique is based on the detection of released inorganic pyrophosphate during reverse transcription of microRNA to cDNA, which is subsequently converted to adenosine 5-triphosphate (ATP) by ATPsulfurylase, and provides energy for luciferase to oxidize luciferin and generate light. Preliminary results show that the proposed method has unique dynamic range and is capable of detecting less than 5 fg of microRNA in a complex RNA mixture. After optimization of reagents and protocols, the assay is expected to outperform real-time PCR for the analysis of small RNA molecules. Compared with current RT-PCR and microarray platforms for RNA quantification, the proposed assay is simpler and faster and requires less expensive reagents. The bioluminescence signals are detected with the widely available luminometer, which has the capability for high-throughput microRNA analysis in a 96-well format but is less expensive than real-time PCR and microarray systems. The accomplishment of the proposed microRNA detection technique will provide a simple, fast, sensitive, and less expensive platform for RNA detection and quantification to be used in life science research, drug discovery, and clinical diagnosis.

Yue Wang¹, Jason Xuan¹, Guoqiang Yu¹, Li Chen1, Robert Clarke<sup>2

1</sup>Department of Electrical, Computer, and Biomedical Engineering, Virginia Polytechnic Institute and State University, Arlington, Virginia; ²Georgetown University Medical Center, Washington, D.C.

The challenge of cancer treatment has been to target specific therapies to pathogenetically distinct cancer types or subtypes to maximize efficacy and minimize toxicity. However, cancers with similar histopathological appearance can follow significantly different clinical courses and show different responses to therapy. The recent development of gene microarrays provides an opportunity to take a genome-wide approach to predict clinical heterogeneity in cancer treatment and potentially discover new diagnostic and therapeutic targets. This IMAT project aims to develop a bioinformatics tool suite for data modeling and analysis consisting of most major computational tasks in cancer research.

Diagnostic marker selection and classifier design are two important tasks for molecular classification of multicategory cancers. However, disconnection between marker selection and classifier design exists in two popular scenarios: (1) Classifier-independent marker selection is universally applicable but not optimized for specific classifier designs, and (2) classifier-dependent marker selection may be optimized for specific classifier designs but often is not directly interpretable. For the molecular classification of multicategory cancers, we have developed a biologically guided, joint marker selection and classifier design algorithm called phenotypic-upregulated gene-supported one-versusrest support vector machine (PUG-OVRSVM). To date, we have tested PUG-OVRSVM on four large published and one in-house oligonucleotide microarray data sets. We compared PUG-OVRSVM with four popular benchmark marker selection methods (SNR, t-statistics, BW, SVM-RFE) and two popular benchmark classifiers (K-Nearest Neighbor, Naive Bayes Classifier). PUG-OVRSVM outperforms all other methods on the four cancer data sets in terms of lower error rate, higher performance stability, and least number of genes required for the lowest error rate. PUG-OVRSVM achieved a comparable performance on one muscular dystrophy data set. Furthermore, a two-step gene selection method is also proposed. In the first step, individually discriminatory genes (IDGs) are identified by using onedimensional weighted signal-to-noise ratio. In the second step, jointly discriminatory genes (JDGs) are selected by sequential search methods from the IDGs, based on their joint class separability measured by multidimensional weighted Fisher criterion. By applying the proposed IDG/JDG approach to a microarray study of small round blue cell tumors (SRBCTs), we successfully identified a much smaller yet efficient set of JDGs (nine genes) for diagnosing SRBCTs, with a misclassification error rate of 3.1%.

Multichannel biomedical imaging promises powerful tools for the visualization and elucidation of important developmental or disease-causing biological processes. Recent research aims to simultaneously assess the spatial-temporal/spectral distributions of multiple biomarkers, where the signals often represent a composite of more than one distinct source independent of spatial resolution. We report a novel blind source separation method for quantitative dissection of mixed yet correlated biomarker distributions. The close-form algebraic solution is based on a linear latent variable model whose parameters are estimated using geodetically principled, nonnegative, leastcorrelated component analysis. We demonstrate the principle of the approach on the mixtures of real cancer images acquired by dual-energy x-ray and dynamic contrast-enhanced magnetic resonance imaging. We observed accurate and robust source separation into component biomarker distributions in agreement with the ground truth or biomedical expectations. With superior performance compared with existing techniques, this method has powerful features that are of considerable widespread applicability.

Zhiyong Ding¹, Jiyong Liang¹, Yiling Lu1, Qinghua Yu1, Zhou Songyang², Shiaw-Yih Lin1, Gordon B. Mills1

¹Department of Systems Biology, University of Texas M.D. Anderson Cancer Center, Houston, Texas; ²Baylor College of Medicine, Houston, Texas

Identification of novel protein-protein interactions is a fundamental step to understanding protein function and signaling networks and allowing efficient implementation of targeted cancer therapy. The majority of protein-protein interactions are currently identified using yeast two hybrid (Y2H), coimmunoprecipitation, and mass spectrometry or protein libraries. Each of these approaches has its own set of major limitations in failing to mimic native physiological conditions (Y2H and protein libraries) or efficiently identify protein interactions on the cytoskeleton or membrane, due either to the location of the interaction (Y2H) or to difficulties in coimmunoprecipitation of cytoskeletal or membrane proteins. Furthermore, conventional Y2H approaches yield false positive signals with transcription factors precluding screening. Therefore, a novel screening method that efficiently identifies biologically relevant protein interactions bypassing the limitations of current screening methods would have wide applicability.

We propose to develop and validate a readily applicable, context-dependent, subcellular localization-, cDNA library-, and cell-type-independent retrovirus-based mammalian two hybrid (ReMTH) screen method for identification of novel protein-protein interactions, including cytoskeletal and membrane proteins, in mammalian cells, allowing native protein folding and posttranslational modifications. In ReMTH, bait protein is fused to one fragment of a rationally dissected fluorescent protein such as GFP. The second, complementary fragment of GFP is fused to an endogenous protein by the retrovirusmediated exon trap vector. An interaction between bait and host protein (prey) can bring the two halves of the GFP molecule into proximity, resulting in reconstitution of fluorescence. The resultant cells will be reagents for the study of the localization and function of the novel protein-protein interaction complex as well as resources for high-content drug or siRNA screening. The fully developed technology will identify functional protein-protein interactions more efficiently than current methods and identify interactions not discoverable by current methods, particularly in context-dependent mammalian screens. Furthermore, the proposed ReMTH screen has the unique potential to stabilize or trap transient/weak interactions such as enzyme/substrate interactions, allowing identification of components of signaling pathways and networks in previously undetectable cancers. We have completed an initial proof-of-concept screen in HeLa cells for identification of interaction partners of the oncogene AKT1 (Ding et al. 2006). We identified a series of previously known AKT1 interaction partners and substrates, as well as novel interaction partners, including cytoskeleton and membrane proteins. We have confirmed that one novel interaction partner, ACTN4, interacts physically and functionally with AKT1. Thus, the technology will uncover functional proteinprotein interactions not detectable by other approaches and advance our understanding of protein functions and signaling networks in cancer.

Ding Z., Liang J., Lu Y., et al. A Retrovirus-Based Protein Complementation Assay Screen Reveals Functional AKT1-Binding Partners. Proc Natl Acad Sci U S A 103(41):15014-9, 2006.

Bevin P. Engelward, Dominika M. Wiktor-Brown, Hyuk-Sang Kwon, Yoon Sung Nam, Peter T. So

Biological Engineering Division, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts

The clonal expansion of cells with mutations in genes that provide growth and survival advantages is one of the pivotal first steps in carcinogenesis. To understand the earliest stages of cellular transformation, a method to identify and analyze these premalignant cells is needed. We have created a transgenic fluorescent yellow direct repeat (FYDR) mouse in which cells that have undergone a rare DNA sequence rearrangement (via a homologous recombination event) express a fluorescent protein, enabling the labeling of phenotypically normal cells. To measure clonal expansion in situ, we have integrated one- and two-photon microscopy to create a sensitive imaging system that spans four orders of magnitude on the length scale and provides three-dimensional (3D) analysis within intact unfixed tissue. This imaging platform rapidly identifies very rare fluorescent cells within an entire mouse organ (at the cm scale) and subsequently provides 3D images of each fluorescent cell or cluster of cells (at the micron scale). We applied these techniques to study the effect of age on clonal expansion of fluorescent cells in the pancreata of FYDR mice. Results show that as mice age, there is a significant increase in the number of cells within fluorescent cell clusters, indicating that pancreatic cells can clonally expand with age. This combination of mechanico-optical engineering technologies with genetically engineered FYDR mice can be applied to study the effects of genetic and environmental exposures on the risk of clonal expansion.

Darryl Shibata, Joanna Wu

Department of Pathology, Keck School of Medicine, University of Southern California, Los Angeles, California

The times required for cancer progression are uncertain. Knowledge of how quickly cancers develop would be useful to guide prevention and screening strategies. However, it is currently impossible to serially observe human cancers because they are removed as soon as they are found. In this study, we seek to develop a method to retrospectively count numbers of divisions during progression by counting numbers of replication errors; that is, the greater the number of divisions or genome replications, the greater the number of replication errors (a “molecular clock” hypothesis). Although somatic mutations are relatively infrequent in most cancers, colorectal cancers deficient in DNA mismatch repair have mutation rates that are 100- to 1,000-fold higher than normal. Nearly all microsatellite (MS) loci are mutated in these cancers. Therefore, it is possible to “count” somatic MS mutations by isolating tumor DNA and then genotyping the sizes of their MS alleles. To “calibrate” our MS molecular clock, “interval” cancers that arise shortly (1-3 years) after a negative clinical examination (colonoscopy) are examined, because visible cancer growth should occur after the negative examination. Preliminary data indicate that sporadic cancers and interval cancers have similar mitotic ages, which suggests that many cancers frequently grow to detectable sizes within a small period of time (years).

A limitation of this study is sample preparation, which uses archival formalin-fixed, paraffin-embedded tissues. Results are confounded by mixtures of tumor and normal DNA, which are currently “filtered” after data collection by various algorithms.

David G. Kaufman, Penny Hawkins, Paul D. Chastain

Department of Pathology and Laboratory Medicine, School of Medicine,The University of North Carolina at Chapel Hill, Chapel Hill, North Carolina

Our goal is to develop a technique that can determine the relative density of oxidative damage sites in bulk DNA and in specific regions of the genome. A single method that can accomplish all these goals is not yet available, but it is currently possible to determine the overall number of abasic (AP) sites in genomic DNA using slot blots, the presence of hotspots for AP formation in a population of the same DNA fragments using ligation-mediated PCR, and the random or clustered distribution of AP sites (using electron microscopy or atomic force microscopy). We will use DNA combing, a method that generates samples of genomic DNA that are aligned, straightened, and easily visualized. We will then detect the distribution of AP sites by immunofluorescence and refine our methods until we are able to quantitate AP sites within 10% of the measurements made by the slot blot technique, which is the current standard of detection. When this is achieved, our next goal will be to optimize the reaction conditions so that, in addition to quantitation of AP sites, we can employ fluorescence in situ hybridization to identify a specific DNA region and visualize AP sites at that locus. Once we have this ability, our last goal will be to standardize these conditions so that, in cells exposed to different doses of hydrogen peroxide (H2O2), our detection of AP sites is found to vary with the H2O2 dose at this locus. We expect that the detection of AP sites will fall within 10% of the levels measured by the current standard method of detection of AP sites, with less than 10% variability within repeats of the same sample. The proposed experiments will prove that these methods can be used to investigate complex biological problems.

We believe that a method that could detect the distribution of AP sites, both in the genome as a whole and within a specific region, would be an extremely valuable addition to existing diagnostic tools. It would enable us to determine the extent and distribution of DNA damage in specific regions of the genome, such as origins of replication or promoter regions; lesion formation during normal cell metabolic processes (e.g., replication and transcription); and lesions formed as a consequence of the cells being exposed to carcinogens or to oxidative stress. This method may allow the study of the role of epigenetic changes in DNA damage (e.g., a change in DNA methylation status and/or histone acetylation in active and inactive regions of the genome or in areas of DNA replicated at different times during the S phase). It could also be used for the characterization of similarities and differences in the DNA damage that occurs in different cell types (e.g., normal and breast cancer cells, different breast cancer lines, and cells from different patients affected by the same disease, such as lung cancer, breast cancer, and Alzheimer’s disease). Ultimately, these studies might lead to a better understanding of the health risks (with regard to the probability, degree, and distribution of DNA damage) for different genotoxic and epigenetic events. This project seeks to achieve a proof of principle that demonstrates the feasibility of a method that would make the aforementioned studies possible.

Jung-Hye Choi, Rebecca Busch, Joon Park, Ie-Ming Shih,Tian-Li Wang

Departments of Pathology, Oncology, and Gynecology, Johns Hopkins Medicine, Johns Hopkins University, Baltimore, Maryland

Interaction between Notch receptor and its ligand activates Notch signaling, which constitutes a molecular circuit in development and tumorigenesis. We have previously reported Notch3 gene amplification in ovarian serous carcinoma and Notch3-dependent growth in cancer cells. In this study, we characterize the roles of the Notch3 ligand Jagged-1 in the tumor progression of ovarian carcinomas. We demonstrated a significant correlation of Notch3 expression and Jagged-1 expression in ovarian cancer based on serial analysis of gene expression (SAGE) data. Quantitative real-time PCR analysis showed significant overexpression of Jagged-1-in ovarian cancer cells compared with ovarian surface epithelial cells and benign cystadenomas. Treatment with Jagged-1-specific shRNA resulted in a significant decrease in cell proliferation and colony formation in OVCAR3 cells and, to a lesser extent, in A2780 and TOVG21 cells. In addition, transwell migration and in vitro wound healing assays revealed that Jagged-1 knockdown reduced cellular motility in OVCAR3, A2780, and TOVG21 cells. These phenotypic changes were accompanied by a decrease in the levels of the active (intracellular) form of Notch3, which has been demonstrated to be involved in the proliferation and survival in Notch3 overexpressing OVCAR3 and A2780 cells. Furthermore, PBX-1, a potential target gene of Notch3 and a tumor suppressor in ovarian carcinoma, was also downregulated by treating OVCAR3 cells with Jagged-1 shRNA. These data suggest that interaction between Jagged-1 and Notch3 is involved in activating Notch3 signaling in ovarian carcinoma, and concomitant expression of Notch3 and Jagged-1 may contribute to tumor progression.

Lewis K. Pannell, Julio Ruiz, Mathur Rajesh, Rajeev Samant, Lalita Shevde

Proteomics and Mass Spectrometry Research Facility, Mitchell Cancer Institute, University of South Alabama, Mobile, Alabama

Objective: The objective of this proposal was to develop methods to perform glycosylation profiling of cancer-associated proteins. Since most secreted and membrane proteins are glycosylated, they were the targets. The concept was to use immunoprecipitation (IP) on the selected proteins and then analyze them using mass spectrometry and mass gap analysis to find the glycosylation profiles. The use of IP has been modified, since too many proteins are often pulled down with the antibodies or associated with the proteins of interest. There was also considerable concern that these antibodies may selectively enrich certain glycoforms of the proteins and thus alter the profiles. The large amount of antibody in the final sample also makes the glycoanalysis difficult.

Solutions: We have developed a highly efficient method to obtain enriched secreted proteomes from cell lines. This work was published and has recently been applied to the analysis of the MCF10 isogenic breast cancer cell series. Proteins have been detected that are highly associated with aggressiveness (accepted subject to revision). The proteins are fractionated into about 10 subproteomes using liquid chromatography on a reverse-phase column; these are then used as the input samples for glycoanalysis, each containing multiple proteins. Samples are digested and run on a mass spectrometer (MS) using accurate mass data acquisition. As a second part of the solution, we have developed a new analysis strategy and computerized this for the analysis of glycosylation in protein mixtures. Although this still uses the mass gap approach to match together the masses from a single glycosylation site, it uses a subdatabase development and automated glycosylation site prediction routine to identify all possible glycosylated peptides in a sample. The use of a fragmentation strategy also provides information on the likely peptide mass for lookup in the table, but this is not an essential step. To help in the analysis of these complex mixtures, we have also moved the technology to the recently acquired, new, and highly accurate Orbitrap MS instrument, which has significantly enhanced the analysis within these mixtures. As a third step, we are using a property of the glycopeptides to help their enrichment during the MS analysis stage. They consistently have higher m/z (mass/charge) ratios than most normal peptides, and by excluding the low m/z data from the scan, the interpretation of the data files and the finding of the glycopeptide masses are greatly enhanced. This combination of altered strategies is now being applied to the analysis of site-specific, glycoprotein profiles with the secreted subproteomes from a number of cell lines, and data from some of these will be presented. For membrane proteins, we are using a similar strategy, and we will enrich these following minimal biotinylation of the cell surface using a fusible linker. This work is currently under way, and the results will also be shown. All of these data can now be analyzed using repeat injections and isolations and the variations processed using both commercial software and a set of differential analysis routines separately developed for clinical studies in the facility. Although the work is behind the 2-year schedule, we are surmounting the problems in the original concept and are requesting a no-cost extension to complete the work.

Spinoffs: The techniques we have developed are separately being applied to the analysis of cancer biomarker proteins in body fluids. It has successfully allowed us to analyze prostate specific antigen (PSA) glycosylation in urine, and this is the subject of a grant application. PSA is a secreted protein and is an excellent case of where there are observed differences in glycosylation in cancer, but these cannot be analyzed in clinical samples. Our routines allow the analysis of such changes within clinical trials. In addition, the secreted proteome method has identified at least one glycoprotein biomarker that is common to multiple cancer histiotypes, which will be the subject of another funding request.

Sebastian Szpakowski¹, Xueguang Sun¹, José M. Lage1, Jill Rubinstein¹, Andrew Dyer¹, Diane Kowalski¹, Michel Krauthammer¹, David Tuck¹, Perry Miller⁴, Hongyu Zhao², Janet Brandsma¹, Clarence Sasaki³, Jose Costa1, Paul M. Lizardi¹

Departments of ¹Pathology, ²Epidemiology and Public Health, and ³Surgery and Anesthesiology and ⁴Center for Medical Informatics, Yale University School of Medicine, New Haven, Connecticut

We are using head and neck cancer as a model to establish and validate procedures for genome-wide molecular analysis of neoplastic lesions. Samples of head and neck malignancies are being collected prospectively at Yale-New Haven Hospital. DNA is extracted using the Epicenter Master Pure kit, and aliquots are stored before and after whole-genome amplification (Qiagen REPLI-g). All archival DNA samples are assayed for human papillomavirus (HPV) infection. The samples are also processed for analysis of gene losses and gains, using array-comparative genomic hybridization (CGH) in highdensity microarrays. Unamplified DNA from the same tumor DNA archives is analyzed using a custom NimbleGen microarray and an endonuclease-based protocol designed to report changes in DNA methylation status. A unique feature of the method is its ability to report methylation changes for loci associated with repetitive DNA. Such loci represent nearly 50% of the human genome and have formerly resisted analysis using microarray approaches. The microarray profiles comprise relative methylation levels at 25,000 promoter CpG islands, 46,000 non-promoter-associated CpG islands, and more than 200,000 CpG islands mapping to interspersed repeats or tandem repeats. A subset of the DNA methylation data has been validated by bisulfite PCR analysis. Analysis of methylation changes in a data set comprising 40 tumors shows striking hypermethylation of a number of candidate tumor suppressor genes and reveals clusters of linked hypermethylated genes within “chromatin neighborhoods.” Abnormal hypomethylation in tumor DNA shows a recurrent, complex component comprising CpG islands associated with interspersed repeat loci. The data underscore for the first time the complexity of the loss of DNA methylation at multiple repetitive DNA loci in tumors and reveals the existence of a genome-wide epigenetic framework that is now accessible to study using relatively small DNA samples. A significant number of DNA methylation abnormalities are observed in “morphologically normal” tissue from the same patients. Given the potential for tumor classification as well as for risk stratification of “morphologically normal” tissue, based on profiles of abnormally methylated genomic loci, this novel approach opens new avenues for a large-scale discovery effort based on non-repeat-masked DNA methylation analysis in any human cancer. The data sets showing DNA methylation abnormalities in head and neck squamous cell carcinoma tumors are being combined with the gene locus gain/loss information obtained by array-CGH to enable improved tumor class comparison, class discovery, and class prediction. Computational and statistical tools are being used to construct classification schemes based on distance-based trees, as well as different clustering algorithms, utilizing the complete data set of array-CGH, DNA methylation, and HPV infection status observations. One of our goals is to derive a conditional risk model identifying those patients who are most likely to develop additional head and neck cancers in the future. However, future application of this technology would benefit from the availability of genetic information for all patient samples, particularly single nucleotide polymorphism haplotypes, which unfortunately are not available in the present study and may limit the power of our tumor morphism classification analysis.

Ramila Philip, Sidhartha Murthy, Jonathan Krakover, Gomathinayagam Sinnathamby, Jennifer Zerfass, Lorraine Keller, Mohan Philip

Immunotope, Inc./Pennsylvania Biotechnology Center, Doylestown, Pennsylvania

Introduction: Elimination of cancer through early detection and treatment is the ultimate goal of cancer research. This is especially critical for ovarian cancer, which is typically diagnosed at very late stages with very poor response rates. Immunoproteomics, which defines the subset of proteins involved in the immune response, holds considerable promise for providing better understanding of early-stage immune response to cancer as well as important insights into antigens that may be suitable for immunotherapy. Early administration of immunotherapeutic vaccines can potentially have profound effects on prevention of metastasis and may potentially cure through efficient and complete tumor elimination.

Method: We developed a mass spectrometry (MS) method to identify novel autoantibody-based serum biomarkers for the early diagnosis of ovarian cancer. The method uses native tumor-associated proteins immunoprecipitated by autoantibodies from sera obtained from cancer patients and from cancer-free controls to identify autoantibody signatures that occur at high frequency only in cancer patient sera. Antigen-antibody complexes were immunoprecipitated from normal and ovarian cancer patient composites. Autoantibody-reactive antigens were separated from the antibodies and fractionated first by size exclusion chromatography (SEC), then further fractionated by reverse-phase high-performance liquid chromatography (HPLC). Each HPLC fraction was individually treated with trypsin. Major histocompatibility peptide complexes (MHCs) from two ovarian cancer cell lines were isolated by immunoaffinity purification, and the peptide fraction was separated from the protein fraction by boiling, then further fractionated by SEC and HPLC. Tryptic peptide fractions and MHC peptide fractions were analyzed using a liquid chromatography-MS system. The mass spectral data were analyzed using Sequest and Mascot software and the SwissProt human database.

Results: We identified a subset of more than 50 autoantigens that were also processed and presented by MHC class I molecules on the surfaces of ovarian cancer cells and thus common to the two immunological processes of humoral and cell-mediated immunity.

Discussion: These shared autoantigens were highly representative protein families with roles in key processes in carcinogenesis and metastasis, such as cell cycle regulation, cell proliferation, apoptosis, tumor suppression, and cell adhesion. Autoantibodies appearing at the early stages of cancer suggest that this detectable immune response to the developing tumor can be exploited as early-stage biomarkers for the development of ovarian cancer diagnostics. Correspondingly, because the T-cell immune response depends on MHC class I processing and presentation of peptides, the identification of proteins that go through this pathway are potential candidates for the development of immunotherapeutics designed to activate a T-cell immune response to cancer. To the best of our knowledge, this is the first comprehensive study that identifies and categorizes proteins that are involved in both humoral and cell-mediated immunity against ovarian cancer and may have broad implications for the discovery and selection of theranostic molecular targets for cancer therapeutics and diagnostics in general.

Rodney J. Rothstein¹,Samantha L. Ciccone¹, John Dittmar², Robert J.D. Reid<sup>1

1</sup>Department of Genetics and Development, College of Physicians and Surgeons, Columbia University Medical Center, New York, New York; ²Department of Biology, Columbia University, New York, New York

We have developed a novel approach to identify new drug targets by uncovering specific genetic interactions involving genes overexpressed in cancer cells. The genetic interaction called synthetic dosage lethality (SDL), which was first described in yeast, occurs when a normally nonessential gene becomes essential due to overexpression of a second gene. To identify potential therapeutic targets, we define SDL interactions in yeast using orthologs of genes that are overexpressed in different cancers. For example, hRad9 is overexpressed in 33% of cases of non-small cell lung carcinomas. hRad9 and its yeast ortholog Ddc1 are part of the sliding clamp [Rad9/Hus1/Rad1 (9-1-1) in humans and Ddc1/Mec3/Rad24 in yeast], one of the damage sensors recruited to stalled replication forks and DNA breaks. This complex subsequently activates cell-cycle checkpoints in collaboration with ATR, Chk1, and downstream effectors. To search for SDL interactions, we mimicked overexpression of hRAD9 by introducing the yeast Ddc1 gene into every strain in the viable haploid gene disruption library and inducing high-level expression of the Ddc1 protein. We identified a strong SDL interaction in the absence of Cik1, a binding partner of the kinesin motor protein Kar3 that functions to increase the velocity of Kar3 movement affecting microtubule dynamics. Cik1 is important during mitosis for assembly and/or maintenance of the mitotic spindle preventing sister chromatid separation, perhaps by microtubule crosslinking. NuMA, the human counterpart of Cik1, interacts with dynein/dynactin, kinesin motor proteins, and KIF2a, the likely human functional ortholog of yeast Kar3. This target is exciting because anticancer drugs such as taxol are known to function by stabilizing microtubules.

We have also screened for mutations in yeast that are sensitive to overexpression of activated RasG19V in budding yeast, which corresponds to G12V in mammalian cells. We chose Ras since activating mutations in N-Ras and K-Ras are found in a large percentage of human cancers. We have screened 50% of the yeast library, and thus far, no SDL interacting partners have been identified when wild-type Ras is overexpressed. Interestingly, more than 6% of the gene disruptions show slower growth or no growth when the RasG19V allele is overexpressed. Among the genes that we have identified are several in the autophagy pathway. Our experiments support an essential role of autophagy in cancer cell survival and underscore the importance of these genes as targets for therapy.

Finally, to show that the same interactions that we define in yeast are occurring in cancer cells, we will validate our yeast results in mammalian cells. We have established cell lines that overexpress hRad9 and are in the process of validating the Ddc1-Cik1 SDL interaction in these cell lines by reducing expression of the Cik1 homolog NuMA using siRNA and assaying cell survival. The ultimate goal is to show that these same interactions can be found in lung cancer cell lines that exhibit overexpression of hRad9. Similar experiments will be done targeting genes in the autophagy pathway in cells expressing RasG12V. The validation of the interactions that we find in yeast may identify novel therapeutic targets that can selectively kill cancer cells. Such therapies would be a major advance compared with current therapies that often nonselectively target all proliferating cells.

This work is supported by R33-CA125520.

Shoshana Frankenburg

Sharett Institute of Oncology, Hadassah University Hospital, Jerusalem, Israel

The major goal of this project is to establish the proof of principle that topical delivery of tumorassociated antigens can elicit effective antitumor responses and can be used as a novel means for cancer immunotherapy. It is now an accepted view that successful cancer treatment will eventually include a combination of different modalities. Thus, effective cell immunotherapy that is simple and eliminates the need for a specialized laboratory or hospitalization will represent a significant contribution to the pool of available antimelanoma treatments.

As model antigens, we cloned, expressed, and purified a 60-Kd recombinant melanoma protein derived from a shortened sequence of the native gp100 gene (HR-gp100) and a synthetic multiepitope polypeptide. As adjuvants and to enhance transcutaneous delivery, we evaluated two forms of heat-labile enterotoxin and a cell active peptide.

HR-gp100, despite its size, entered the skin without the need of adjuvant. This was demonstrated by (1) production of specific antibodies in mice after topical application of the protein and (2) dosedependent epidermal Langerhans cell (LC) activation following transcutaneous delivery to intact human skin. LC activation was measured using a novel model of human skin transplanted to the highly vascularized chorioallantoic membrane of the chicken egg. Current experiments are analyzing cellular immune responses elicited by transdermal protein delivery.

This project aims at developing tools that will facilitate cell immunotherapy. We expect the data produced to demonstrate the feasibility of transdermal vaccination and to characterize the immune responses elicited by this treatment. We expect the results to be affected by the intrinsic characteristics of the antigenic proteins evaluated.

William G. Kerr¹, Sander Diks², Maikel Petrus Peppelenbosch², Kim Paraiso1, Daniela Wood<sup>1

1</sup>Department of Immunology, H. Lee Moffitt Cancer Center and Research Institute,Tampa, Florida; ²University Medical Center Groningen, Groningen, The Netherlands

Multiple myeloma (MM) is a neoplasm of plasma cell origin that is largely confined to the bone marrow (BM). Chromosomal translocations and other genetic events are known to contribute to deregulation of signaling pathways, which leads to the transformation of plasma cells and progression to malignancy. To provide a more comprehensive molecular analysis of signaling disruptions in this disease, we compared the kinome of normal plasma cells to neoplastic MM plasma cells. The kinome of normal BM stromal cells and diseased stromal cells were also compared. To do this comparison, we prepared highly purified samples of the above cell populations from the BM of MM patients and normal controls via high-speed cell sorting. The purified cell populations were then lysed in the presence of phosphatase inhibitors and then incubated in the presence of 33P-ATP on PepChip arrays of 1,152 different pseudopeptides representing substrates for most of the kinases present in the mammalian cell. Thus far, we have compared the kinome of four different MM cell samples with that of four different normal plasma cells samples as well as stromal cells from both tumor and normal marrow. These comparisons revealed deregulation of multiple signaling pathways in MM cells but not in the supporting stroma compared with their normal counterparts. The deregulated kinases identified are potential novel molecular targets in this lethal disease.

Xiang Li, Charles J. Bieberich

Department of Biological Sciences, University of Maryland, Baltimore County, Baltimore, Maryland

Signal transduction is a central component of nearly all biological processes. Protein kinases play critical roles in signaling pathways by phosphorylating proteins involved in signal amplification and in executing the cellular response to extrinsic or intrinsic stimuli. Aberrant kinase signaling plays a role in the etiology of nearly all cancers and in hundreds of other diseases. Systematic approaches to the elucidation of the signaling pathways driven by kinases have the potential to illuminate new points of therapeutic intervention. The identification of physiologic kinase substrates is an important component of this endeavor. Currently, there is a need for technologies that can be widely deployed to approach this problem. Using protein kinase CK2 as a paradigm, we are developing a simple, straightforward technique for kinase substrate. In this approach, the roles of kinase and substrate in a classic in-gel kinase assay are reversed. In the reverse in-gel kinase assay (RIKA), a kinase is copolymerized in a polyacrylamide gel that is then used to resolve a tissue or cell protein extract. Refolding by buffer exchange to restore kinase activity and substrate structure is followed by a kinase reaction to phosphorylate substrates in situ. We demonstrate that this method can be used to profile true CK2 substrates and show that RIKA can detect as little as 50 femtomoles of a substrate. With further development and validation, this assay has the potential to identify the physiologic substrates of many cancer-relevant serine/threonine kinases.

Xiaowei Xu, Hongtao Zhang, Lynn Schuchter, Suzanne McGettigan, Mark Greene, Pat Van Belle, David Elder

Department of Pathology and Laboratory Medicine, School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania

We have developed a new antigen detection and quantification method that we term “fluorescent amplification catalyzed by T7 RNA polymerase technology” (FACTT). FACTT uses principles similar to enzyme-linked immunosorbent assay (ELISA); however, the detection antibody in FACTT is directly coupled to a double-stranded DNA template that contains a T7 promoter to accommodate the attachment of the T7 RNA polymerase enzyme. The interaction of T7 leads to the production of RNA species that can be monitored by adding a fluorescent RNA intercalating dye. FACTT assay is developed in 96- or 384-well plates. It is an innovative isothermal, quantitative, high-throughput immunoassay platform. Our preliminary data demonstrated that FACTT assays consistently have at least a 1,000-fold higher sensitivity than ELISA.

Malignant melanoma is a deadly disease. Melanoma cells express melanocyte lineage-specific markers, such as Melan-A, tyrosinase, and TRP-1. Melanoma cells secrete soluble tumor markers, such as protein melanoma-inhibitory activity (MIA) and S-100beta. Tumor cell apoptosis and necrosis are common in malignant neoplasms even at an early stage; therefore, lineage-specific markers may be released into the bloodstream.

Our goal is to develop ultrasensitive assays to monitor melanoma progression and response to therapy. FACTT is a versatile platform that also can be used to detect biomarkers for other cancers.

We have set up FACTT assays to detect tyrosinase, TRP-1, and MIA. Pairs of antibodies were purchased from commercial sources. The capture antibody is coated in carbonate-bicarbonate buffer (pH 9.6) to 384-well plates at 5 µg/mL/well and 20 µL/well overnight at 4 °C. 1:100 dilution of serum in the amount of 20 µL per well was added to the coated plate for a 1-hour incubation at room temperature. Twenty µL of diluted biotinylated detection antibody (180 ng/mL, or an optimized concentration for each antibody) was used for each well and incubated at room temperature for 1 hour. Streptavidin and the biotin-DNA template (amplification module [AM]) were added sequentially at 5 µg/mL and 250 ng/mL, respectively, with a 1-hour room temperature incubation for each step. We washed the plate six times with PBST between each binding incubation. After excess AM and proteins were removed by washing, 20 µL of reaction mixture (containing 60 units of T7 RNA polymerase plus [Ambion], 1.25 mM NTP, 1x T7 buffer [Ambion]) was added to each well. RNA amplification was performed at 37 °C for 3 hours. The RNA intercalating dye RiboGreen was added to the reaction mixture (20 µL, 1:200 diluted in the TE buffer supplied by the manufacturer), and the plates were read at Ex 485nm/Em 535 nm in a TECAN SpectraFluor reader. Our preliminary data showed that tyrosinase, TRP-1, and MIA levels were significantly increased in patients with metastatic melanoma.

We plan to optimize the assays to have a concentration curve correlation coefficiency of at least 0.95; coefficient of variation of intra-assay and interassay variance of <10% and a quantification accuracy of >90%. We will establish normal ranges for serum tyrosinase, TRP-1, and MIA in control populations using FACTT assays. We have already started to collect sera and plasma from patients with history of melanoma. We will use the assays to measure the serum protein levels in these samples to see whether these serum proteins can be used as biomar