Systematic variation in gene expression patterns in human cancer cell lines.

We used cDNA microarrays to explore the variation in expression of approximately 8,000 unique genes among the 60 cell lines used in the National Cancer Institute's screen for anti-cancer drugs. Classification of the cell lines based solely on the observed patterns of gene expression revealed a correspondence to the ostensible origins of the tumours from which the cell lines were derived. The consistent relationship between the gene expression patterns and the tissue of origin allowed us to recognize outliers whose previous classification appeared incorrect. Specific features of the gene expression patterns appeared to be related to physiological properties of the cell lines, such as their doubling time in culture, drug metabolism or the interferon response. Comparison of gene expression patterns in the cell lines to those observed in normal breast tissue or in breast tumour specimens revealed features of the expression patterns in the tumours that had recognizable counterparts in specific cell lines, reflecting the tumour, stromal and inflammatory components of the tumour tissue. These results provided a novel molecular characterization of this important group of human cell lines and their relationships to tumours in vivo.

A gene expression database for the molecular pharmacology of cancer.

We used cDNA microarrays to assess gene expression profiles in 60 human cancer cell lines used in a drug discovery screen by the National Cancer Institute. Using these data, we linked bioinformatics and chemoinformatics by correlating gene expression and drug activity patterns in the NCI60 lines. Clustering the cell lines on the basis of gene expression yielded relationships very different from those obtained by clustering the cell lines on the basis of their response to drugs. Gene-drug relationships for the clinical agents 5-fluorouracil and L-asparaginase exemplify how variations in the transcript levels of particular genes relate to mechanisms of drug sensitivity and resistance. This is the first study to integrate large databases on gene expression and molecular pharmacology.

Charge-mosaic membranes: enhanced permeability and negative osmosis with a symmetrical salt.

Charge-mosaic membranes are prepared by embedding a single layer of alternating cation and anion exchange beads in silicone resin. Membranes made in an identical manner but containing only one type of exchanger serve as controls. The mosaic membranes are 50 to 100 times more permeable to potassium chloride than the controls; and furthermore they give rise to net volume flow from concentrated to dilute solutions of potassium chloride in the absence of a pressure gradient ("negative osmosis"), whereas the controls exhibit normal osmotic behavior. The negative reflection coefficients of the mosaics suggest potential applications in desalination.

Electric power from differences in salinity: the dialytic battery.

An array of alternating anion and cation exchange membranes can be used to generate electric power from the free energy of mixing of river and sea waters. A simple mathematical model, which predicts experimental results well, is useful in exploring conditions for optimization of the process. Major, but not impossible, improvements in technology would be required to bring the cost of power from the dialytic battery into line with foreseeable energy prices.

Charge-mosaic membranes: dialytic separation of electrolytes from nonelectrolytes and amino acids.

Charge-mosaic membranes were used for dialytic separations of potassium chloride from low-molecular-weight nonelectrolytes and neutral amino acids. The permeability ratio (potassium chloride to uncharged species) ranged from about 6 in the case of methanol to about 86 in that of mannitol. A theoretical model predicts that optimum rates of dialysis should be achieved by dialyzing against salt concentrations other than zero; this prediction was confirmed by experiment. These observations suggest potential applications of mosaics in laboratory separations, industrial processing, and hemodialysis.

Liposomes and local hyperthermia: selective delivery of methotrexate to heated tumors.

Liposomes with phase transitions a few degrees above physiological temperature delivered more than four times as much methotrexate to murine tumors heated to 42 degrees C as to unheated control tumors. Most of the accumulated drug appeared to be intracellular and bound to dihydrofolate reductase, the enzyme blocked by methotrexate in its role as an antineoplastic agent.

Design of liposomes for enhanced local release of drugs by hyperthermia.

Liposomes can be designed to release an entrapped drug preferentially at temperatures attainable by mild local hyperthermia. In a test system in vitro, protein synthesis by Escherichia coli is inhibited and killing of the cells is enhanced by heating neomycin-containing liposomes to their phase transition temperature to maximize drug release. In the presence of serum the ratio of release at 44 degrees C to that at 37 degrees C can be made greater than 100:1, suggesting possible applications in the treatment of tumors or local infection.

Liposome-cell interaction: transfer and intracellular release of a trapped fluorescent marker.

When small, unilamellar lipid vesicles containing a high concentration of the fluorescent dye 6-carboxyfluorescein are incubated with either frog retinas or human lymphocytes, fluroescence distributes widely throughout each cell. Since "self-quenching" largely prevents the dye from fluorescing as long as it remains sequestered in vesicles, it is clear that a considerable amount of dye is released from the vesicles and diluted into the much larger volume of the cell.

Monoclonal anitbodies in the lymphatics: toward the diagnosis and therapy of tumor metastases.

Monoclonal antibodies subcutaneously injected into mice track to regional lymph nodes and specifically label target cells there. The lymphatic route of administration can be expected to provide much higher sensitivity, higher target-to-background ratio, faster localization, and lower toxicity than the intravenous route when the aim is to diagnose or treat tumor metastases or lymphoma in the lymph nodes.

Hydroxyurea as an inhibitor of human immunodeficiency virus-type 1 replication.

Hydroxyurea, a drug widely used in therapy of several human diseases, inhibits deoxynucleotide synthesis--and, consequently, DNA synthesis--by blocking the cellular enzyme ribonucleotide reductase. Hydroxyurea inhibits human immunodeficiency virus-type 1 (HIV-1) DNA synthesis in activated peripheral blood lymphocytes by decreasing the amount of intracellular deoxynucleotides, thus suggesting that this drug has an antiviral effect. Hydroxyurea has now been shown to block HIV-1 replication in acutely infected primary human lymphocytes (quiescent and activated) and macrophages, as well as in blood cells infected in vivo obtained from individuals with acquired immunodeficiency syndrome (AIDS). The antiviral effect was achieved at nontoxic doses of hydroxyurea, lower than those currently used in human therapy. Combination of hydroxyurea with the nucleoside analog didanosine (2',3'-dideoxyinosine, or ddl) generated a synergistic inhibitory effect without increasing toxicity. In some instances, inhibition of HIV-1 by hydroxyurea was irreversible, even several weeks after suspension of drug treatment. The indirect inhibition of HIV-1 by hydroxyurea is not expected to generate high rates of escape mutants. Hydroxyurea therefore appears to be a possible candidate for AIDS therapy.

Monoclonal antibodies in the lymphatics: selective delivery to lymph node metastases of a solid tumor.

After subcutaneous injection, monoclonal antibodies directed against a tumor can enter local lymphatic vessels, pass to the draining lymph nodes, and bind to metastases there. Lymphatic delivery of antibody to early metastases is more efficient than intravenous administration, and the lymphatic route can be used to image smaller metastatic deposits. Perhaps more important, the lymphatic route minimizes binding of antibodies to circulating tumor antigens and to cross-reactive antigens present on normal tissues. Antibodies inappropriate for intravenous use because of binding to normal tissues may therefore be useful against lymph node metastases when injected subcutaneously or directly into lymphatic vessels.

Neural computing in cancer drug development: predicting mechanism of action.

Described here are neural networks capable of predicting a drug's mechanism of action from its pattern of activity against a panel of 60 malignant cell lines in the National Cancer Institute's drug screening program. Given six possible classes of mechanism, the network misses the correct category for only 12 out of 141 agents (8.5 percent), whereas linear discriminant analysis, a standard statistical technique, misses 20 out of 141 (14.2 percent). The success of the neural net indicates several things. (i) The cell line response patterns are rich in information about mechanism. (ii) Appropriately designed neural networks can make effective use of that information. (iii) Trained networks can be used to classify prospectively the more than 10,000 agents per year tested by the screening program. Related networks, in combination with classical statistical tools, will help in a variety of ways to move new anticancer agents through the pipeline from in vitro studies to clinical application.

An information-intensive approach to the molecular pharmacology of cancer.

Since 1990, the National Cancer Institute (NCI) has screened more than 60,000 compounds against a panel of 60 human cancer cell lines. The 50-percent growth-inhibitory concentration (GI50) for any single cell line is simply an index of cytotoxicity or cytostasis, but the patterns of 60 such GI50 values encode unexpectedly rich, detailed information on mechanisms of drug action and drug resistance. Each compound's pattern is like a fingerprint, essentially unique among the many billions of distinguishable possibilities. These activity patterns are being used in conjunction with molecular structural features of the tested agents to explore the NCI's database of more than 460,000 compounds, and they are providing insight into potential target molecules and modulators of activity in the 60 cell lines. For example, the information is being used to search for candidate anticancer drugs that are not dependent on intact p53 suppressor gene function for their activity. It remains to be seen how effective this information-intensive strategy will be at generating new clinically active agents.

Gadd45, a p53-responsive stress protein, modifies DNA accessibility on damaged chromatin.

This report demonstrates that Gadd45, a p53-responsive stress protein, can facilitate topoisomerase relaxing and cleavage activity in the presence of core histones. A correlation between reduced expression of Gadd45 and increased resistance to topoisomerase I and topoisomerase II inhibitors in a variety of human cell lines was also found. Gadd45 could potentially mediate this effect by destabilizing histone-DNA interactions since it was found to interact directly with the four core histones. To evaluate this possibility, we investigated the effect of Gadd45 on preassembled mononucleosomes. Our data indicate that Gadd45 directly associates with mononucleosomes that have been altered by histone acetylation or UV radiation. This interaction resulted in increased DNase I accessibility on hyperacetylated mononucleosomes and substantial reduction of T4 endonuclease V accessibility to cyclobutane pyrimidine dimers on UV-irradiated mononucleosomes but not on naked DNA. Both histone acetylation and UV radiation are thought to destabilize the nucleosomal structure. Hence, these results imply that Gadd45 can recognize an altered chromatin state and modulate DNA accessibility to cellular proteins.

Use of the Kohonen self-organizing map to study the mechanisms of action of chemotherapeutic agents.

BACKGROUND: Many natural and synthetic compounds might prove to be effective in cancer chemotherapy. To identify potentially useful agents, the National Cancer Institute screens over 10,000 compounds annually against a panel of 60 distinct human tumor cell lines in vitro. This screening program generates large amounts of data that are organized into relational databases. Important questions concern the information content of the data and ways to extract that information. Previously, statistical techniques have revealed that compounds with similar patterns of activity against the 60 cell lines are often similar in structure and mechanism of action. Feed-forward, back-propagation neural networks have been trained on this type of data to predict broadly defined mechanisms of action of chemotherapeutic agents. PURPOSE AND METHOD: In this report, we examine the information that can be extracted from the screening data by means of another type of neural network paradigm, the Kohonen self-organizing map. This is a topology-preserving function, obtained by unsupervised learning, that nonlinearly projects the high-dimensional activity patterns into two dimensions. Our dataset is almost identical to that used in the earlier neural network study. RESULTS: The self-organizing maps we constructed have several important characteristics. 1) They partition the two-dimensional array into distinct regions, each of which is principally occupied by agents having the same broadly defined mechanism of action. 2) These regions can be resolved into distinct subregions that conform to plausible submechanisms and chemically defined subgroups of submechanism. 3) These results (and exceptions to them) are consistent with those obtained with the use of such deterministic measures of similarity among activity patterns as the Euclidean distance or Pearson correlation coefficient. CONCLUSIONS: Our results indicate that the activity patterns obtained from the screen contain detailed information about mechanism of action and its basis in chemical structure. The self-organizing map can be used to suggest the mechanism of action of compounds identified by the screen as potentially useful chemotherapeutic agents and to probe the biology of the cell lines in the cancer screen. Kohonen self-organizing maps, unlike the previously applied neural networks, preserve and reveal the relationships among compounds acting by similar mechanisms and therefore have the potential to identify compounds that act by novel cytotoxic mechanisms.

Identification of epidermal growth factor receptor and c-erbB2 pathway inhibitors by correlation with gene expression patterns.

BACKGROUND: Growth factor receptor-signaling pathways are potentially important targets for anticancer therapy. The interaction of anticancer agents with specific molecular targets can be identified by correlating target expression patterns with cytotoxicity patterns. We sought to identify new agents that target and inhibit the activity of the epidermal growth factor (EGF) receptor and of c-erbB2 (also called HER2 or neu), by correlating EGF receptor, transforming growth factor (TGF)-alpha (a ligand for EGF receptor), and c-erbB2 messenger RNA (mRNA) expression levels with the results of cytotoxicity assays of the 49000 compounds in the National Cancer Institute (NCI) drug screen database. METHODS: The levels of mRNAs were measured and used to generate a molecular target database for the 60 cell lines of the NCI anticancer drug screen. The computer analysis program, COMPARE, was used to search for cytotoxicity patterns in the NCI drug screen database that were highly correlated with EGF receptor, TGF-alpha, or c-erbB2 mRNA expression patterns. The putative EGF receptor-inhibiting compounds were tested for effects on basal tyrosine phosphorylation, in vitro EGF receptor tyrosine kinase activity, and EGF-dependent growth. Putative ErbB2-inhibiting compounds were tested for effects on antibody-induced ErbB2 tyrosine kinase activity. RESULTS: EGF receptor mRNA and TGF-alpha mRNA levels were highest in cell lines derived from renal cancers, and c-erbB2 mRNA levels were highest in cells derived from breast, ovarian, and colon cancers. Twenty-five compounds with high correlation coefficients (for cytotoxicity and levels of the measured mRNAs) were tested as inhibitors of the EGF receptor or c-erbB2 signaling pathways; 14 compounds were identified as inhibitors of these pathways. The most potent compound, B4, inhibited autophosphorylation (which occurs following activation) of ErbB2 by 50% in whole cells at 7.7 microM. CONCLUSIONS: Novel EGF receptor or c-erbB2 pathway inhibitors can be identified in the NCI drug screen by correlation of cytotoxicity patterns with EGF receptor or c-erbB2 mRNA expression levels.

Growth and chemotherapeutic response of cells in a hollow-fiber in vitro solid tumor model.

BACKGROUND: Cancer treatments that appear promising in tissue culture are often less effective in solid tumors, in part because of the proliferative and microenvironmental heterogeneity that develops in these tumors as they grow. Heterogeneous tumor models are thus needed for drug screening. PURPOSE: Our goal was to develop and test for drug evaluation a solid tumor model based on cell growth inside biocompatible hollow fibers. METHODS: Building on the experience of Hollingshead and co-workers with a sparse-cell, hollow-fiber tumor model, we tested six human tumor cell lines for in vitro growth inside 450-microns internal-diameter polyvinylidine fluoride fibers and examined them histologically. Human SW620 colon carcinoma cells grown in hollow fibers were also examined using electron microscopy, and their doxorubicin sensitivity was assessed. A colorimetric assay based on sulforhodamine B was adopted to replace the more cumbersome clonogenic cell survival assay. RESULTS: Five of the human tumor cell lines tested grew to confluence, forming heterogeneous in vitro tumors with subpopulations of viable and necrotic cells. For SW620 hollow-fiber tumors, maximum viable cell populations in excess of 10(8) cells/mL were obtained after 8 days of growth. This viable cell density remained roughly constant for 3-4 days, permitting dose-response experiments over this time interval. Tumor cells in hollow fibers were much more resistant to a 4-hour doxorubicin exposure than were tumor cells in monolayers: LC50 values (i.e., the drug concentrations at which the plating efficiency equals one-half the plating efficiency of untreated cells) of 3.5 microM and 0.16 microM were obtained for hollow-fiber tumors and monolayers, respectively. LC50 values decreased when drug exposure time was increased. Results from the colorimetric assay were in agreement with those from the clonogenic assay. CONCLUSION: The successful growth of tumor cells to confluence in hollow fibers and the feasibility of performing in vitro drug dose-response experiments with a relatively easy colorimetric assay demonstrate the potential of the hollow-fiber solid tumor model as a tool for experimental therapeutic research. IMPLICATION: Hollow-fiber solid tumors may prove useful for experimental drug evaluation.

Early intervention in cancer using monoclonal antibodies and other biological ligands: micropharmacology and the "binding site barrier".

Monoclonal antibodies and other biological ligands tend to distribute nonuniformly in bulky tumors after systemic administration. In part, that observation reflects intrinsic heterogeneity of the tumor; in part, it represents poor percolation through tumor substance. Theoretical considerations led us several years ago to formulate the "binding site barrier" hypothesis, the idea that macromolecular ligands could be prevented from penetrating tumors by the very fact of their successful binding to the target receptor. All else being equal, the higher the density of target moieties (e.g., antigens) and the higher the affinity, the greater the barrier. Experimental evidence for this hypothesis remained circumstantial until we recently obtained direct experimental verification in an animal tumor system. As shown by calculations in the present study, metabolism of ligand in free form or once it has bound to the target can also limit dramatically the extent of penetration. The PERC program package, developed to examine these issues in the case of monoclonal antibodies, has now been applied to other types of ligands as well. We speculate that the same microscopic factors have influenced the evolution of biological ligands, such as the autocrine-paracrine and chemotactic factors. Micropharmacological issues (binding sites, molecular size, and charge) should be taken into account as we design the next generation of biological ligands for systemic administration. The same issues are, perhaps, even more important with respect to molecular design of biological factors to be secreted by genetically modified cells in the treatment of cancer and in cancer vaccines. Since the PERC calculations and experiments relate to aggregates of tumor cells no more than a few hundred micron across, the ideas appear relevant to the problems of early detection and intervention. However, barriers associated with organized epithelial cell layers and basement membrane in the case of early carcinomas and carcinomatous change remain to be understood.

An analysis of monoclonal antibody distribution in microscopic tumor nodules: consequences of a "binding site barrier".

Rational in vivo application of monoclonal antibodies for diagnosis and therapy of cancer requires an understanding of both the global and microscopic pharmacology of macromolecular ligands. Here, we introduce a new mathematical model for antibody distribution into small, prevascular, densely packed nodules (representing either primary or metastatic tumor). For the analysis, we link together several aspects of antibody pharmacology: the global (whole body) pharmacokinetics; transcapillary transport into normal tissue interstitium surrounding the nodule; diffusion into the nodule; nonspecific binding and/or partitioning; specific binding to tumor antigen; metabolism; and lymphatic outflow from the tissue space. Input parameter values are estimated from experimental studies in vitro, in animals, and in clinical trials. Our aim is to explore the sensitivity of antibody localization to variation in three of the important parameters of this model: the rate of transcapillary transport; the rate of lymphatic outflow; and the antigen density. Predictions based on this analysis include the following: (a) High rates of transcapillary transport influx or low rates of lymphatic efflux will enhance antibody percolation into the tumor nodule at early times after injection and increase the average antibody concentration in the tumor at all times; (b) Changes in antibody influx rate will affect the antibody distribution in the tumor at earlier times than do changes in the efflux rate; (c) Reducing the antigen concentration will increase the uniformity of antibody penetration but lower the average concentration in the tumor at all times after injection; and (d) Counter to intuition, lowering the antigen concentration can increase the peak concentrations achieved toward the center of the nodule. If, in addition, there is any metabolism of bound antibody, the concentration-time integral (i.e., the "area under the curve") for the center of the nodule will also be increased by decreasing the antigen concentration. These predictions directly reflect the "binding site barrier" hypothesis of Weinstein et al. (Ann. NY Acad. Sci., 507: 199-210, 1987) and Fujimori et al. (Cancer Res., 49:5656-5663, 1989; J. Nucl. Med., 31:1191-1198, 1990). In general, and perhaps surprisingly until one considers the problem carefully, the parameters governing antibody percolation can have opposite effects on the uniformity of antibody distribution at early and late times. These calculations, using the PERC program set, were done for antibodies, but we believe that the "binding site barrier" will also prove important for other injected macromolecules, for at least some highly bindable injected small molecules, for lymphokines and cytokines released from transfected cells injected in vivo, and, indeed, for endogenous species such as the autocrine-paracrine factors.

Integrated microscopic-macroscopic pharmacology of monoclonal antibody radioconjugates: the radiation dose distribution.

Accurate dosimetry is essential for the assessment of radioimmunotherapy. Most often studied to date has been the macroscopic dosimetry related to organ and tumor distribution of the radiolabeled antibody, but the question of microscopic dose heterogeneity is also important. To address the latter issue, we have taken an integrated approach to the pharmacology, taking into account whole-body distribution, transcapillary transport, percolation through the tumor interstitial space, antigen-antibody interaction, and antibody metabolism. The first step is to simulate the spatial antibody concentration profile in a tumor as a function of time after i.v. (e.g., bolus) injection, using reasonable values for the parameters involved. The second step is to calculate, also as a function of time, the absorbed radiation dose distribution resulting from each concentration profile. Parameter values for IgG pharmacology and a radiation point source function for 131I are used to explore the effect of antibody distribution profiles on absorbed dose in the tumor. The geometry simulated corresponds to a spherical nodule of densely packed tumor cells. Absorbed doses are calculated for radiation from a single nodule (e.g., a micrometastasis or prevascular primary tumor) and for a cubic lattice of such nodules (e.g., corresponding to nodular lymphoma). As noted in our previous studies, there is a "binding site barrier." Binding to antigen retards antibody percolation into the nodules; high antibody affinity tends to decrease percolation and give a higher absorbed dose near the surface of each nodule. Heterogeneous antibody distribution results in a heterogeneous absorbed dose. This is more apparent in the case of radiation from a single nodule than it is for radiation from within an array of nodules. Dehalogenation results in a lower absorbed dose over time, and the effect is more apparent at later times after injection. PERC-RAD, the computer program package developed for these analyses, provides a convenient and flexible way to assess the impact of macroscopic and microscopic parameters on the distribution of radioimmunoconjugates and on the consequent profile of absorbed radiation dose in tumors. This mathematical model and the general principles developed here can be applied as well to other radiolabeled biological ligands.