A Bayesian network approach to feature selection in mass spectrometry data.

BACKGROUND: Time-of-flight mass spectrometry (TOF-MS) has the potential to provide non-invasive, high-throughput screening for cancers and other serious diseases via detection of protein biomarkers in blood or other accessible biologic samples. Unfortunately, this potential has largely been unrealized to date due to the high variability of measurements, uncertainties in the distribution of proteins in a given population, and the difficulty of extracting repeatable diagnostic markers using current statistical tools. With studies consisting of perhaps only dozens of samples, and possibly hundreds of variables, overfitting is a serious complication. To overcome these difficulties, we have developed a Bayesian inductive method which uses model-independent methods of discovering relationships between spectral features. This method appears to efficiently discover network models which not only identify connections between the disease and key features, but also organizes relationships between features--and furthermore creates a stable classifier that categorizes new data at predicted error rates. RESULTS: The method was applied to artificial data with known feature relationships and typical TOF-MS variability introduced, and was able to recover those relationships nearly perfectly. It was also applied to blood sera data from a 2004 leukemia study, and showed high stability of selected features under cross-validation. Verification of results using withheld data showed excellent predictive power. The method showed improvement over traditional techniques, and naturally incorporated measurement uncertainties. The relationships discovered between features allowed preliminary identification of a protein biomarker which was consistent with other cancer studies and later verified experimentally. CONCLUSIONS: This method appears to avoid overfitting in biologic data and produce stable feature sets in a network model. The network structure provides additional information about the relationships among features that is useful to guide further biochemical analysis. In addition, when used to classify new data, these feature sets are far more consistent than those produced by many traditional techniques.

HTLV-I trans-activator protein, tax, is a trans-repressor of the human beta-polymerase gene.

Human T cell leukemia virus type I (HTLV-I) is the etiological agent for adult T cell leukemia (ATL). The HTLV-I trans-activator protein Tax can activate the expression of its own long terminal repeat (LTR) and many cellular and viral genes. Tax down-regulated the expression of human beta-polymerase (hu beta-pol), a cellular enzyme involved in host cell DNA repair. This finding suggests a possible correlation between HTLV-I infection and host chromosomal damage, which is often seen in ATL cells.

Discrete serum protein signatures discriminate between human retrovirus-associated hematologic and neurologic disease.

The human T-cell leukemia virus type I (HTLV-I) is the causative agent for adult T-cell leukemia (ATL) and HTLV-I-associated myelopathy/tropical spastic paraparesis (HAM/TSP). Approximately 5% of infected individuals will develop either disease and currently there are no diagnostic tools for early detection or accurate assessment of disease state. We have employed high-throughput expression profiling of serum proteins using mass spectrometry to identify protein expression patterns that can discern between disease states of HTLV-I-infected individuals. Our study group consisted of 42 ATL, 50 HAM/TSP, and 38 normal controls. Spectral peaks corresponding to peptide ions were generated from MS-TOF data. We applied Classification and Regression Tree analysis to build a decision algorithm, which achieved 77% correct classification rate across the three groups. A second cohort of 10 ATL, 10 HAM and 10 control samples was used to validate this result. Linear discriminate analysis was performed to verify and visualize class separation. Affinity and sizing chromatography coupled with tandem mass spectrometry was used to identify three peaks specifically overexpressed in ATL: an 11.7 kDa fragment of alpha trypsin inhibitor, and two contiguous fragments (19.9 and 11.9 kDa) of haproglobin-2. To the best of our knowledge, this is the first application of protein profiling to distinguish between two disease states resulting from a single infectious agent.

HTLV-I Tax and Cytokeratin: Tax-Expressing Cells Show Morphological Changes in Keratin-Containing Cytoskeletal Networks.

Human T cell leukemia virus type I (HTLV-I) has been linked to the development of an aggressive lymphoproliferative disorder (adult T cell leukemia), a chronic neurodegenerative presentation (HTLV-I-associated myelopathy/tropical spastic paraparesis) and numerous less well-defined inflammatory conditions. The viral regulatory protein Tax has been implicated in cellular transformation events leading to the onset of adult T cell leukemia. Details on the stepwise processes through which Tax induces morphological changes in cells are poorly understood. We show here that Tax can bind to a class of intermediate filaments, the cytokeratins (Ker). Tax interacts with the 1B helical coil of keratin 8, a domain critical for higher-order intermediate filament matrix formation. Expression of Tax in epithelial cells visibly altered the structural pattern of the Ker network. In a T lymphocyte cell line, induction of Tax expression resulted in increased cellular adherence/invasion of Matrigel filters. We propose that one aspect of Tax function is the induction of morphological changes in cellular cytoskeletal structures. This finding for Tax-expressing cells might be one factor contributing directly to the pathogenesis of HTLV-I disease(s). Copyright 1997 S. Karger AG, Basel

Divergent Subcellular Locations of HTLV-I Tax and Int-6: A Contrast between in vitro Protein-Protein Binding and Intracellular Protein Colocalization.

Protein-protein interactions define many important molecular and cellular processes in prokaryotic and eukaryotic biology. In trying to delineate the contact between two proteins, the yeast two-hybrid assay has emerged as a powerful technique. Complementing the yeast two-hybrid assay are in vitro techniques (e.g. GST-fusion-protein chromatography) that can also yield information on protein-protein associations. However, unambiguous functional significance to these interactions is best supported through a finding of colocalization of two proteins inside cells. In instances where two proteins interact in vitro but have divergent localizations within cells one needs to reconsider the biological importance of the former finding. Here, we present evidence for different subcellular locations of HTLV-I Tax and the Int-6 protein. We suggest a reexploration of the functional significance between Tax and Int-6 in cellular transformation.

Substitution of norleucine for methionine residues in a crustacean pigment-dispersing hormone.

In order to evaluate the structural/functional roles of Met residues in an octadecapeptide pigment-dispersing hormone (PDH: Asn-Ser-Gly-Met-Ile-Asn-Ser-Ile-Leu-Gly-Ile-Pro-Arg-Val-Met-Thr-Glu-Ala- NH2), first described as light-adapting distal retinal pigment hormone (DRPH) from Pandalus, three analogs were synthesized: Nle4-PDH, Nle15-PDH, and Nle4,15-PDH. When tested for melanophore pigment-dispersing activity in destalked Uca, all three Nle-analogs were more potent than unsubstituted PDH. Performic acid oxidation caused a marked loss of potency of PDH, Nle4-PDH, and Nle15-PDH. The analog Nle4,15-PDH was resistant to oxidation and displayed 6-fold higher potency than PDH. Thus Met4 and Met15 are not essential for the PDH activity. The oxidation-induced loss of activity of unsubstituted PDH may result from introduction of oxygen (in methionine sulfone) and a consequent conformational change in the octadecapeptide.

C-terminal deletion analogs of a crustacean pigment-dispersing hormone.

This study deals with the effect of deamidation and C-terminal truncation on the potency of an octadecapeptide pigment-dispersing hormone (PDH: Asn-Ser-Gly-Met-Ile-Asn-Ser-Ile-Leu-Gly-Ile-Pro-Arg-Val-Met-Thr-Glu-Ala- NH2), first described as light-adapting distal retinal pigment hormone (DRPH) from Pandalus borealis. Bioassay of synthetic analogs for melanophore pigment dispersion in destalked fiddler crabs (Uca pugilator) showed that deamidation causes a 300-fold decrease in potency. The analogs 1-17 NH2 and 1-16 NH2 were about 3 times more potent than 1-18-OH. Further truncation led to decreases in potency, with the peptide 1-9-NH2 being the smallest C-terminal deletion analog to display activity (0.001% potency). Smaller analogs (1-8-NH2, 1-6-NH2 and 1-4-NH2) were inactive when tested in doses as high as 500 nmoles/crab. On the basis of our earlier work on N-terminal deletion analogs and the present findings the residues 6 to 9 seem to be important for PDH action.

Molecular pathology of prostate cancer.

This chapter includes discussion of the molecular pathology of tissue, blood, urine, and expressed prostatic secretions. Because we are unable to reliably image the disease in vivo, a 12 core method that oversamples the peripheral zone is widely used. This generates large numbers of cores that need to be carefully processed and sampled. In spite of the large number of tissue cores, the amount of tumor available for study is often quite limited. This is a particular challenge for research, as new biomarker assays will need to preserve tissue architecture intact for histopathology. Methods of processing and reporting pathology are discussed. With the exception of ductal variants, recognized subtypes of prostate cancer are largely confined to research applications, and most prostate cancers are acinar. Biomarker discovery in urine and expressed prostatic secretions would be useful since these are readily obtained and are proximate fluids. The well-known challenges of biomarker discovery in blood and urine are referenced and discussed. Mediators of carcinogenesis can serve as biomarkers as exemplified by mutations in PTEN and TMPRSS2:ERG fusion. The use of proteomics in biomarker discovery with an emphasis on imaging mass spectroscopy of tissues is discussed. Small RNAs are of great interest, however, their usefulness as biomarkers in clinical decision making remains the subject of ongoing research. The chapter concludes with an overview of blood biomarkers such as circulating nucleic acids and tumor cells and bound/free isoforms of prostate specific antigen (PSA).

Definition of a minimal activation domain in human T-cell leukemia virus type I Tax.

Fourteen mutants were used to delineate a minimal activation domain in the Tax protein of human T-cell leukemia virus type I. In an assay using a Gal4-Tax (GalTx) fusion protein and a responsive promoter containing Gal4 consensus binding sites, we found that activation was "squelched" by coexpression of wild-type Tax protein in trans. When Tax mutants were tested for squelching, many competed effectively against GalTx. However, those containing changes in amino acids 289 to 322 failed to inhibit activity. In particular, three mutants that were expressed stably, with changes at amino acids 289, 296, and 320 respectively, did not squelch GalTx activity. On the other hand, mutants with individual changes at amino acid 3, 9, 29, 41, 273, and 337 efficiently inhibited GalTx function. Three other mutants failed to be stably expressed. In separate experiments, when fused alone to the DNA-binding domain of Gal4, amino acids 289 to 322 of Tax conferred trans activation ability. This fusion protein was able to activate a core promoter. These findings suggest that amino acids 289 to 322 define a region that contacts an essential transcription factor and that this region is a modular activation domain.

Mutational analysis of human T-cell leukemia virus type I Tax: regions necessary for function determined with 47 mutant proteins.

We have made 47 mutations that span the length of the human T-cell leukemia virus type I (HTLV-I) Tax open reading frame. Of the 47 mutations, 38 were substitutions of single amino acids, 5 were missense changes in two or more amino acids, and 4 were deletions. A subset of these mutations includes individual changes of all 26 naturally occurring serines to alanines. By assaying each mutant protein separately on the HTLV-I long terminal repeat (LTR) and the human immunodeficiency virus type 1 (HIV-1) LTR in parallel, we were able to identify regions of Tax selectively necessary for each promoter. A small region in the carboxyl terminus, amino acids 315 to 325, was found to be selectively important for activation of the HTLV-I LTR. Three changes at serine 113, serine 160, and serine 258 were found to specifically affect function on the HIV-1 LTR. Surprisingly, we found that the great preponderance of missense changes (32 of 42) in Tax did not affect function.

HTLV-I Tax is a zinc-binding protein: role of zinc in Tax structure and function.

We have examined the functional significance of the cysteine- and histidine-rich region (amino acids 22-53) of HTLV-I Tax. A modeling of this region suggests two possible overlapping zinc-finger-like motifs. Using a zinc blotting technique, we show that Tax binds zinc. An N-terminal deletion in Tax that removed this zinc-finger region abolished the ability to bind zinc. Site-directed mutagenesis was used to generate 10 separate mutations so as to discriminate between the two alternative zinc-binding structures. Each Tax mutant was studied for its ability to trans-activate the HTLV-I LTR. Five of the ten mutations inactivated trans-activation. Our results support that one of the two putative zinc fingers is an integral element of Tax structure.

Localization of human T-cell leukemia virus type 1 tax to subnuclear compartments that overlap with interchromatin speckles.

Tax, the virally encoded activator of the human T-cell leukemia virus type 1 long terminal repeats, regulates the expression of many cellular genes. This protein has been implicated in transformation events leading to the development of adult T-cell leukemia. Because subcellular localization contributes importantly to protein function, we determined the compartment(s) within the cell in which Tax is found. Using confocal microscopy, we found that Tax localizes to subnuclear domains which overlap with structures previously identified as interchromatin granules or spliceosomal speckles. These Tax speckled structures are coincident with a subset of nuclear transcriptional hot spots. Disruption of the Tax speckled structures by heat shock revealed the existence of different populations of Tax. One population of Tax is tightly associated with nuclear speckles. A second population exists outside of the speckles and is transcriptionally active for some promoters.

Induction of micronuclei by HTLV-I Tax: a cellular assay for function.

Cellular chromosomal damage is ubiquitously seen in HTLV-I-transformed lymphocytes. It is also characteristic of cells that have been exposed to mutagens. A sensitive measurement for mutagen-induced DNA damage is the formation of micronuclei in treated cells. Because current evidence suggests that HTLV-I Tax is etiologically linked to transformation, we tested for its activity in inducing micronuclei. We show here that transfection into cells of a Tax-producing plasmid rapidly induced the formation of micronuclei. This effect cooperated with that of a mutagen (mitomycin C) and was correlated with the inherent trans-activation capacity of Tax. These findings suggest that a commonly used mutagen assay could be a quick biological test for putatively oncogenic proteins.

Human T-cell leukemia virus type I tax masks c-Myc function through a cAMP-dependent pathway.

Human T-cell leukemia virus type I Tax is a pleiotropic gene regulator that functions through CREB/ATF- and NF-kappaB-mediated pathways. In most contexts, Tax is a potent gene activator. Here, we describe an unexpected finding of Myc repression by Tax. In cells that overexpress human T-cell leukemia virus type I Tax, the detection of c-Myc protein in the nucleus by a monoclonal antibody was masked. Tax prevented immunological visualization of a Myc epitope contained within amino acids 45-104, resulting in interference with Myc function in transcription and in anchorage-independent cell growth. Tax did not affect steady-state protein levels since detection of c-Myc with other antibodies was unperturbed. Four observations suggest that this Tax-Myc interaction is mediated through CREB/ATF signal transduction. 1) Tax point mutants, selectively defective for activation of CREB/ATF but not NF-kappaB, failed to mask c-Myc; 2) masking of Myc was abolished when Tax-expressing cells were treated with protein kinase inhibitor H-9; 3) Tax-specific shielding of Myc is absent in cells (B1R) that are genetically defective for cAMP signaling; and 4) forskolin treatment of cells mimicked Tax in masking the Myc epitope. Considered collectively, these findings suggest a regulation of Myc function at the level of localized protein conformation.

Tunicamycin inhibits function and expression of the high-affinity IL-2 receptor in a murine IL-2-dependent cell line.

Murine interleukin-2-dependent T-lymphocytes (CT6) were treated with tunicamycin, an inhibitor of both glycoprotein and ganglioside synthesis, to study the involvement of glycosylation in the IL-2 proliferative response. Tunicamycin inhibited proliferation in a dose-dependent manner at concentrations which did not inhibit protein synthesis (10-50 ng/ml). Swainsonine, a glycoprotein processing inhibitor, had no effect on proliferation. Inhibition of proliferation by tunicamycin was accompanied by an inhibition of binding of 125I-IL-2 to its high-affinity receptor. Scatchard analysis showed that receptor number was decreased by tunicamycin treatment. On the other hand, tunicamycin did not affect either the binding of the monoclonal antibody 7D4, specific for the 55 kDa low-affinity protein subunit of the IL-2 receptor, or the recycling of the IL-2 receptor. To determine the specific effects of tunicamycin on the biosynthesis of particular CT6 glycoconjugates, cells were radiolabeled with 3H-glucosamine and incorporation into ganglioside, neutral glycolipid and glycoprotein fractions was measured. Low doses of tunicamycin inhibited ganglioside synthesis and glycoprotein glycosylation to the same extent, whereas no effect on neutral glycolipid synthesis was observed. These results suggest that glycosylation of glycoprotein and/or gangliosides might play an important role in the formation of a functional high-affinity IL-2 receptor complex in CT6 cells.

The amino terminus of Tax is required for interaction with the cyclic AMP response element binding protein.

Tax of human T-cell leukemia virus type 1 was analyzed for interaction with the cyclic AMP response element binding protein (CREB) in vitro with and without Tax response element DNA. Mutations in the carboxy terminus of Tax (L296G and L320G) did not affect binding to CREB and led to supershifts. In contrast, mutants with changes in the amino-terminal cysteine-rich region lost the ability to bind to CREB. The S10A mutant protein bound moderately. Thus, the amino terminus of Tax is essential for Tax-CREB interaction.

Human T-cell leukemia virus type 1 Tax shuttles between functionally discrete subcellular targets.

Human T-cell leukemia virus type 1 (HTLV-1) Tax is a nuclear protein with striking pleiotropic functionality. We recently demonstrated that Tax localizes to a multicomponent nuclear speckled structure (Tax speckled structure [TSS]). Here, we examine these structures further and identify a partial overlap of TSS with transcription hot spots. We used a strategy of directed expression via fusion proteins to determine if these transcription sites are the subtargets within TSS required for Tax function. When fused to human immunodeficiency virus type 1 (HIV-1) Tat, the resulting Tat-Tax fusion protein displayed neither a Tat-like nor a Tax-like pattern but rather was targeted specifically to the transcription subsites. The Tat-Tax fusion was able to activate both the HIV-1 long terminal repeat (LTR) and the HTVL-1 LTR at the same level as the individual component; thus, targeting proteins to transcription hot spots was compatible with both Tax and Tat transcription function. In contrast, the fusion with HIV-1 Rev, Rev-Tax, resulted in a pattern of expression that was largely Rev-like (nucleolar and cytoplasmic). The reduced localization of Rev-Tax to transcription sites was reflected in a 10-fold drop in activation of the HTLV-1 LTR. However, there was no loss in the ability of Tax to activate via NF-kappaB. Thus, NF-kappaB-dependent Tax function does not require targeting of Tax to these transcription sites and suggests that activation via NF-kappaB is a cytoplasmic function. Selective mutation of the nuclear localization signal site in the Rev portion resulted in retargeting of Rev-Tax to TSS and subsequent restoration of transcription function, demonstrating that inappropriate localization preceded loss of function. Mutation of the nuclear export signal site in the Rev portion had no effect on transcription, although the relative amount of Rev-Tax in the cytoplasm was reduced. Finally, in explaining how Tax can occupy multiple subcellular sites, we show that Tax shuttles from the nucleus to the cytoplasm in a heterokaryon fusion assay. Thus, pleiotropic functionality by Tax is regulatable via shuttling between discrete subcellular compartments.

A human T-cell leukemia virus Tax variant incapable of activating NF-kappaB retains its immortalizing potential for primary T-lymphocytes.

The human T-cell leukemia virus type 1 (HTLV-1) transactivator (Tax) has been shown to interfere with regulated cellular proliferation. Many studies have focused on the ability of Tax to transform rodent fibroblasts; however, none has defined the molecular requirements for Tax transformation of human lymphoid cells. We show here that tax induces permanent growth of human primary T-lymphocytes by using a transformation/immortalization defective rhadinovirus vector. The cells phenotypically resemble HTLV-immortalized lymphocytes and contain episomally persisting recombinant rhadinoviral sequences, which stably express functional Tax protein. As Tax can activate major cellular signal transducing pathways including NF-kappaB and cAMP-responsive element binding protein (CREB), we asked for the relevance of these routes in the immortalization of human lymphocytes. By using Tax mutants that either activate exclusively CREB/activating transcription factor or are defective in activating this signaling pathway, we delineated that Tax can induce immortalization of primary human T-lymphocytes through a mechanism independent of NF-kappaB activation.

Modulation of Sp1 phosphorylation by human immunodeficiency virus type 1 Tat.

We previously reported (K. T. Jeang, R. Chun, N. H. Lin, A. Gatignol, C. G. Glabe, and H. Fan, J. Virol. 67: 6224-6233, 1993) that human immunodeficiency virus type 1 (HIV-1) Tat and Sp1 form a protein-protein complex. Here, we have characterized the physical interaction and a functional consequence of Tat-Sp1 contact. Using in vitro protein chromatography, we mapped the region in Tat that contacts Sp1 to amino acids 30 to 55. We found that in cell-free reactions, Tat augmented double-stranded DNA-dependent protein kinase (DNA-PK)-mediated Sp1 phosphorylation in a contact-dependent manner. Tat mutants that do not bind Sp1 failed to influence phosphorylation of the latter. In complementary experiments, we also found that Tat forms protein-protein contacts with DNA-PK. We confirmed that in HeLa and Jurkat cells, Tat expression indeed increased the intracellular amount of phosphorylated Sp1 in a manner consistent with the results of cell-free assays. Furthermore, using two phosphatase inhibitors and a kinase inhibitor, we demonstrated a modulation of reporter gene expression as a consequence of changes in Sp1 phosphorylation. Taken together, these findings suggest that activity at the HIV-1 promoter is influenced by phosphorylation of Sp1 which is affected by Tat and DNA-PK.

HTLV-I and HTLV-II Tax: differences in induction of micronuclei in cells and transcriptional activation of viral LTRs.

Human T-cell leukemia virus (HTLV) types I and II are highly related viruses that differ in disease manifestations. HTLV-I has been linked unmistakably to adult T-cell leukemia-lymphoma. On the other hand, there is little evidence that prior infection with HTLV-II increases risk for lymphoproliferative disorders. Both viruses encode homologous transcriptional-activating proteins (respectively designated as Tax1 and Tax2) which have been suggested to be important mediators of viral pathogenesis. Previously, we reported that Tax1 is a potent inducer of micronuclei formation in cells. Here, we present evidence that Tax2 lacks micronuclei inductive ability. We contrast this phenotypic difference between Tax1 and Tax2 at the cellular level with their similarities at the molecular level in transcriptional activation.