Estimating dataset size requirements for classifying DNA microarray data.

A statistical methodology for estimating dataset size requirements for classifying microarray data using learning curves is introduced. The goal is to use existing classification results to estimate dataset size requirements for future classification experiments and to evaluate the gain in accuracy and significance of classifiers built with additional data. The method is based on fitting inverse power-law models to construct empirical learning curves. It also includes a permutation test procedure to assess the statistical significance of classification performance for a given dataset size. This procedure is applied to several molecular classification problems representing a broad spectrum of levels of complexity.     

Methylation of high molecular weight fibroblast growth factor-2 determines post-translational increases in molecular weight and affects its intracellular distribution.

The high molecular weight (HMW) forms (24, 22.5, and 22 kDa) of basic fibroblast growth factor-2 (FGF-2) contain an N-terminal extension responsible for their predominantly nuclear localization. These forms of FGF-2 are post-translationally modified, resulting in a 1- to 2-kDa increase in apparent molecular mass. Here we show that this post-translational modification is inhibited by methionine starvation and by the methyltransferase inhibitors 5'-deoxy-5'-methylthioadenosine (MTA) and 3-deaza-adenosine. Inhibition of the methylation-dependent modification results in a significant decrease in HMW FGF-2 nuclear accumulation, suggesting that methylation is relevant to the intracellular distribution of these forms of FGF-2. Treatment with MTA does not affect either the synthesis or the intracellular fate of another nuclear protein, the SV40 large T antigen, demonstrating that this drug does not have a generalized effect on nuclear protein accumulation. These results link HMW FGF-2 post-translational modification to its intracellular distribution.     

Lipopolysaccharide inhibits activation of latent transforming growth factor-β in bovine endothelial cells

The activation of latent transforming growth factor-β (TGF-β) by vascular endothelial cells (ECs) is regulated by cellular plasminogen activator (PA)/plasmin, transglutaminase (TGase), and latent TGF-β levels. Because lipopolysaccharide (LPS) has been reported to reduce EC surface plasmin levels by increasing the production of the inhibitor of PA, PA inhibitor-1 (PAI-1), we have tested whether LPS might suppress latent TGF-β activation in ECs using two different systems, namely, bovine aortic ECs (BAECs) cocultured with smooth muscle cells (SMCs) and BAECs treated with retinol. BAECs were either cocultured with SMCs after treatment with 15 ng/ml LPS or were treated with 2 μM retinol and/or 10 ng/ml LPS, and the expression of PA, surface plasmin, TGase, and the amounts of active and latent TGF-β secreted into the culture modium were measured. The downregulation of surface PA/plasmin levels with LPS was accompanied by a profound decline of both TGase and latent TGF-β expression as well as the suppression of surface activation of latent TGF-β. The effect was dependent on the concentration of LPS and on treatment time. The formation of TGF-β did not occur in cells maintained in LPS-contaminated culture medium. © 1995 Wiley-Liss, Inc.     

Latent Transforming Growth Factor-β Binding Protein Domains Involved in Activation and Transglutaminase-dependent Cross-Linking of Latent Transforming Growth Factor-β

Transforming growth factor-β (TGF-β) is secreted by many cell types as part of a large latent complex composed of three subunits: TGF-β, the TGF-β propeptide, and the latent TGF-β binding protein (LTBP). To interact with its cell surface receptors, TGF-β must be released from the latent complex by disrupting noncovalent interactions between mature TGF-β and its propeptide. Previously, we identified LTBP-1 and transglutaminase, a cross-linking enzyme, as reactants involved in the formation of TGF-β. In this study, we demonstrate that LTBP-1 and large latent complex are substrates for transglutaminase. Furthermore, we show that the covalent association between LTBP-1 and the extracellular matrix is transglutaminase dependent, as little LTBP-1 is recovered from matrix digests prepared from cultures treated with transglutaminase inhibitors. Three polyclonal antisera to glutathione S–transferase fusion proteins containing amino, middle, or carboxyl regions of LTBP-1S were used to identify domains of LTBP-1 involved in crosslinking and formation of TGF-β by transglutaminase. Antibodies to the amino and carboxyl regions of LTBP-1S abrogate TGF-β generation by vascular cell cocultures or macrophages. However, only antibodies to the amino-terminal region of LTBP-1 block transglutaminase-dependent cross-linking of large latent complex or LTBP-1. To further identify transglutaminase-reactive domains within the amino-terminal region of LTBP-1S, mutants of LTBP-1S with deletions of either the amino-terminal 293 (ΔN293) or 441 (ΔN441) amino acids were expressed transiently in CHO cells. Analysis of the LTBP-1S content in matrices of transfected CHO cultures revealed that ΔN293 LTBP-1S was matrix associated via a transglutaminasedependent reaction, whereas ΔN441 LTBP-1S was not. This suggests that residues 294–441 are critical to the transglutaminase reactivity of LTBP-1S.     

Isolation of a protease inhibitor from tissues resistant to tumor invasion

We have purified to homogeneity the major trypsin inhibitors from both bovine cartilage and aorta, two tissues reported to be highly resistant to invasion. The two inhibitors appear to be identical and they resemble the Kunitz inhibitor with respect to molecular weight, amino acid composition, range of susceptible proteases, and antigenicity. Each of these inhibitors accounts for 100% of the antitrypsin activity found in extracts of bovine cartilage and aorta.     

Tetradecanoyl phorbol acetate stimulates latent collagenase production by cultured human endothelial cells

Angiogenesis is associated with the fragmentation of blood vessel basement membranes. Since collagen is a major constituent of basement membranes, cultured human endothelial cells derived from umbilical cord veins were assayed for their ability to produce collagenase. Unstimulated cultured human endothelial cells did not secrete detectable levels of active collagenase into the culture medium. However, if the post-culture medium was treated with trypsin or plasmin, low levels of collagenolytic activity were detected, indicating that endothelial cells secrete small amounts of latent collagenase. Addition of the tumor promoter 12-O-tetradecanoyl phorbol-13-acetate (TPA) to the culture medium stimulated the secretion of collagenase by endothelial cells 5–30 fold. More than 90% of the collagenase was secreted in the latent form. Stimulation of collagenase production was detected at 10^?9 M TPA and was maximal at 10^?8 M TPA. An increase in the rate of collagenase production could be detected within 3 hr after the addition of TPA, and full induction occurred by 12 hr. Cycloheximide (3 μg/ml) or actinomycin D (0.1 μg/ml) inhibited both basal levels of collagenase production and the stimulation of collagenase production by TPA. Phorbol-12,13-didecanoate (PDD), a tumor-promoting analog of TPA, also stimulated collagenase production when administered at the same concentrations that were effective for TPA. However, 4-O-methyl TPA and 4-αPDD, two analogs of TPA which are not tumor promoters, did not stimulate collagenase production at concentrations up to 10^?7 M. The collagenase produced by endothelial cells was a typical vertebrate collagenase as judged by the following criteria: it cleaved collagen into only two fragments which were three quarters and one quarter of the length of the intact molecule; it was inhibited by EDTA and human serum; it was not inhibited by inhibitors of serine, thiol or aspartate proteases. Thus TPA causes an increase in the production of latent collagenase by cultured human endothelial cells.     

Lipoprotein (a) inhibits the generation of transforming growth factor beta: an endogenous inhibitor of smooth muscle cell migration

Conditioned medium (CM) derived from co-cultures of bovine aortic endothelial cells (BAECs) and bovine smooth muscle cells (BSMCs) contains transforming growth factor-beta (TGF-beta) formed via a plasmin-dependent activation of latent TGF-beta (LTGF beta), which occurs in heterotypic but not in homotypic cultures (Sato, Y., and D. B. Rifkin. 1989. J. Cell Biol. 107: 1199-1205). The TGF-beta formed is able to block the migration of BSMCs or BAECs. We have found that the simultaneous addition to heterotypic culture medium of plasminogen and the atherogenic lipoprotein, lipoprotein (a) (Lp(a)), which contains plasminogen-like kringles, inhibits the activation of LTGF-beta in a dose-dependent manner. The inclusion of LDL in the culture medium did not show such an effect. Control experiments indicated that Lp(a) does not interfere with the basal level of cell migration, the activity of exogenous added TGF-beta, the release of LTGF-beta from cells, the activation of LTGF-beta either by plasmin or by transient acidification, or the activity of plasminogen activator. The addition of Lp(a) to the culture medium decreased the amount of plasmin found in BAECs/BSMCs cultures. Similar results were obtained using CM derived from cocultures of human umbilical vein endothelial cells and human foreskin fibroblasts. These results suggest that Lp(a) can inhibit the activation of LTGF-beta by competing with the binding of plasminogen to cell or matrix surfaces. Therefore, high plasma levels of Lp(a) might enhance smooth muscle cell migration by decreasing the levels of the migration inhibitor TGF-beta thus contributing to generation of the atheromatous lesions.     

Role of phospholipids in the cytotoxic action of high density lipoprotein on trypanosomes.

Host range among the African trypanosomes, protozoa that cause fatal diseases both in humans and livestock, may be, in part, regulated by toxic properties associated with host high density lipoproteins (HDL). High density lipoproteins from hosts resistant (baboon, human) or susceptible (rabbit, rat) to Trypanosoma brucei infection were isolated and their trypanocidal activity was determined in in vitro cell lysis assays. Rabbit and rat HDL were not cytotoxic while baboon and human HDL rapidly lysed trypanosomes within 2 h at 37 degrees C. Analysis of the phospholipid composition of HDL preparations from these species suggested a correlation between trypanocidal activity and low phosphatidylinositol content. Phospholipase digestion of HDL resulted in a loss of trypanocidal activity, indicating the importance of native phospholipids in maintaining this biological activity of HDL. Cell lysis and loss of trypanosome infectivity induced by baboon HDL could be inhibited either by addition of rabbit or rat HDL to the incubation medium or by addition of purified phospholipids, phosphatidylinositol being the most effective inhibitor. Although the mechanism by which HDL lyses trypanosomes remains to be elucidated, these results suggest an important role for phospholipids in determining the specificity of this cytotoxic property of HDL.     

Release of basic fibroblast growth factor,an angiogenic factor devoid of secretory signal sequence: A trivial phenomenon or a novel secretion mechanism?

Basic fibroblast growth factor (bFGF), a potent angiogenesis inducer, lacks a signal sequence. Therefore, it has been proposed that bFGF is primarily released from dead or damaged cells. Other proteins devoid of secretion signals, interleukin 1 beta (IL-1 beta) and the muscle lectin L-14, have been shown to be released via exocytosis, a novel secretion pathway independent of the "classic" endoplasmic reticulum-Golgi route. In the light of these findings and of our own recent results, we discuss evidence that bFGF can be released from single, uninjured cells and mediate functions in an autocrine manner. As is the case for IL-1 beta and L-14, externalization of bFGF may occur via exocytosis, a pathway utilized during development and differentiation.     

A specific and potent immunotoxin composed of antibody ZME-018 and the plant toxin gelonin

Murine monoclonal antibody ZME-018 recognizes a 240 Kda glycoprotein present on the surface of most human melanoma cells and on over 80% of human biopsy specimens tested. Gelonin is a ribosome-inactivating plant toxin similar in nature and rivaling the activity of ricin A chain. ZME-018 was coupled to purified gelonin using the reagents SPDP and 2-iminothiolane. The ZME-gelonin conjugate was purified by S-300 Sephacryl and Blue Sepharose chromatography, removing unreacted gelonin and antibody, respectively. PAGE analysis showed that ZME was coupled to 1, 2, or 3 gelonin molecules. The ZME-gelonin conjugate was 10(6)-fold more active than gelonin itself in inhibiting the growth of log-phase human melanoma cells in culture. The immunoconjugate was not cytotoxic to antigen negative T-24 (human bladder carcinoma) cells. Treatment of melanoma cells with recombinant IFN-alpha or TNF substantially augmented the cytotoxicity of the immunoconjugate while treatment with IFN-gamma had a minor effect. Using the human tumor colony assay of melanoma cells obtained from fresh biopsy specimens, greater than 90% growth suppression was observed in 2 of 4 samples tested at a concentration of 250 ng/ml. In addition, 25% growth suppression was observed with a third sample tested, and no growth suppression was observed in 1 sample. Thus, clonogenic melanoma cells are sensitive in vitro to the cytotoxic activity of this immunotoxin at concentrations which we presume are pharmacologically relevant.