Biological assessment of robust noise models in microarray data analysis.

MOTIVATION: Although several recently proposed analysis packages for microarray data can cope with heavy-tailed noise, many applications rely on Gaussian assumptions. Gaussian noise models foster computational efficiency. This comes, however, at the expense of increased sensitivity to outlying observations. Assessing potential insufficiencies of Gaussian noise in microarray data analysis is thus important and of general interest. RESULTS: We propose to this end assessing different noise models on a large number of microarray experiments. The goodness of fit of noise models is quantified by a hierarchical Bayesian analysis of variance model, which predicts normalized expression values as a mixture of a Gaussian density and t-distributions with adjustable degrees of freedom. Inference of differentially expressed genes is taken into consideration at a second mixing level. For attaining far reaching validity, our investigations cover a wide range of analysis platforms and experimental settings. As the most striking result, we find irrespective of the chosen preprocessing and normalization method in all experiments that a heavy-tailed noise model is a better fit than a simple Gaussian. Further investigations revealed that an appropriate choice of noise model has a considerable influence on biological interpretations drawn at the level of inferred genes and gene ontology terms. We conclude from our investigation that neglecting the over dispersed noise in microarray data can mislead scientific discovery and suggest that the convenience of Gaussian-based modelling should be replaced by non-parametric approaches or other methods that account for heavy-tailed noise.

Studies on the pharmacokinetics and metabolism of a gamma-secretase inhibitor BMS-299897, and exploratory investigation of CYP enzyme induction.

BMS-299897 is a gamma-secretase inhibitor that was effective in reducing amyloid beta-peptide (A beta) in transgenic mice and guinea pigs. Therefore, pharmacokinetic and drug metabolism studies were conducted in animals to support its clinical development. The compound appeared to have low to intermediate total body clearance and was orally bioavailable (24-100%). The oral absorption of BMS-299897 from solid dosage forms appeared to be dissolution rate-limited. BMS-299897 was distributed into extravascular space (V(ss) >or= 1.3 l kg(-1)), including brain (brain-to-plasma ratio = 0.13-0.50). BMS-299897 appeared to be a P-glycoprotein (P-gp) substrate as the brain-to-plasma ratio was two-fold higher in the mdr1a knockout mouse as compared with the wild-type. Apparent autoinduction by BMS-299897 was observed in murine and rat efficacy and toxicity studies. In vitro, BMS-299897 was a weaker inducer of cytochrome P450 3A4 (CYP3A4) and a weaker transactivator of human pregnane X receptor (hPXR) as compared with rifampicin. Induction of human UGT1A and UGT2B was evaluated in primary human hepatocytes, but the results were inconclusive. A low potential for autoinduction in humans was predicted at a clinical dose of 250 mg and the prediction was consistent with the findings from a clinical multiple-dose study with BMS-299897 in probable Alzheimer's patients.

Influence of pop hole dimensions on the number of laying hens outside on the range.

The aim was to evaluate whether pop hole width is a factor influencing the number of laying hens on the range. Eight groups of 256 birds each were kept in 8 compartments in a deep litter system. Hens could leave each compartment through two equally-sized pop holes arranged evenly along the side of each compartment. Pop hole dimensions were varied every second week in each compartment in a random order from 30, 60, 90, 120 up to 150 x 30 cm (width x height). Range per hen (10 m(2)) were provided. The number of laying hens on range was counted hourly from 07:00 to 20:00 h. Pop hole width did not significantly influence the number of laying hens on the range. Our findings show that, within the limits of the dimensions investigated, other factors are more important than pop hole dimensions.

Dynamics of {beta}-amyloid reductions in brain, cerebrospinal fluid, and plasma of {beta}-amyloid precursor protein transgenic mice treated with a {gamma}-secretase inhibitor.

gamma-Secretase inhibitors are one promising approach to the development of a therapeutic for Alzheimer's disease (AD). gamma-Secretase inhibitors reduce brain beta-amyloid peptide (Abeta), which is believed to be a major contributor in the etiology of AD. Transgenic mice overexpressing the human beta-amyloid precursor protein (APP) are valuable models to examine the dynamics of Abeta changes with gamma-secretase inhibitors in plaque-free and plaque-bearing animals. BMS-299897 2-[(1R)-1-[[(4-chlorophenyl)sulfony](2,5-difluorophenyl)amino]ethyl]-5-fluorobenzenepropanoic acid, a gamma-secretase inhibitor, showed dose- and time dependent reductions of Abeta in brain, cerebrospinal fluid (CSF), and plasma in young transgenic mice, with a significant correlation between brain and CSF Abeta levels. Because CSF and brain interstitial fluid are distinct compartments in composition and location, this correlation could not be assumed. In contrast, aged transgenic mice with large accumulations of Abeta in plaques showed reductions in CSF Abeta in the absence of measurable changes in plaque Abeta in the brain after up to 2 weeks of treatment. Hence, CSF Abeta levels were a valuable measure of gamma-secretase activity in the central nervous system in either the presence or absence of plaques. Transgenic mice were also used to examine potential side effects due to Notch inhibition. BMS-299897 was 15-fold more effective at preventing the cleavage of APP than of Notch in vitro. No changes in the maturation of CD8(+) thymocytes or of intestinal goblet cells were observed in mice treated with BMS-299897, showing that it is possible for gamma-secretase inhibitors to reduce brain Abeta without causing Notch-mediated toxicity.

Medical care for individual patients: concepts beyond evidence-based medicine.

Todays evidence-based medicine has brought the practicing physician a vast amount of statistical evidence from which various stakeholders in the healthcare system obtain their arguments for and against the use of new therapies. Physicians assume an obligation to prescribe these new treatment options for their patients, firstly because of their eagerness to provide the best medicine, and secondly because of their fear of litigations. On looking at the published data, however, we have observed that the arguments for saving lives with a new treatment are not always supported by the underlying data. Sometimes the data show that the effect of treatment, in real terms, is only a relatively small gain in life-time prolongation. It is concluded that EBM-based concepts such as NNT (number needed to treat), absolute risk and relative risk fall short in ensuring real benefit for the patient. We have, therefore, put forward a mathematic model which takes into account the benefit of a treatment for the individual patient in terms of expected gain in lifetime duration. This model is readily applicable to published results on the clinical effects of a medical therapy and gives the practicing physician a useful tool for deciding whether to administer a medical therapy or not. By looking at the duration of treatment and the individual gain in lifetime expected, we have derived an effectiveness coefficient which can be used to categorize medical treatments into highly effective (close to 100%) and not effective (< 5%), and at the same time arrive at a cost-benefit analysis of the treatment in question. These simple concepts will help physicians and individual patients to make informed decisions based only on those medical therapies which are proven and appropriate. The model we have developed provides a new perspective in therapy for the individual patient using medicines that constitute a rational therapy i.e. a therapy that makes "sense" (sense-orientated medicine = SOM).

Expression and analysis of presenilin 1 in a human neuronal system: localization in cell bodies and dendrites.

Mutations in the recently identified presenilin 1 gene on chromosome 14 cause early onset familial Alzheimer disease (FAD). Herein we describe the expression and analysis of the protein coded by presenilin 1 (PS1) in NT2N neurons, a human neuronal model system. PS1 was expressed using recombinant Semliki Forest virions and detected by introduced antigenic tags or antisera to PS1-derived peptides. Immunoprecipitation revealed two major PS1 bands of approximately 43 and 50 kDa, neither of which were N-glycosylated or O-glycosylated. Immunoreactive PS1 was detected in cell bodies and dendrites of NT2N neurons but not in axons or on the cell surface. PS1 was also detected in BHK cells, where it was also intracellular and colocalized with calnexin, a marker for the rough endoplasmic reticulum. A mutant form of PS1 linked to FAD did not differ from the wild-type protein at the light microscopic level. The model system described here will enable studies of the function of PS1 in human neurons and the role of mutant PS1 in FAD.

Alzheimer-associated presenilins 1 and 2: neuronal expression in brain and localization to intracellular membranes in mammalian cells.

Mutations in two recently identified genes appear to cause the majority of early-onset familial Alzheimer's disease (FAD). These two novel genes, presenilin 1 (PS1) and presenilin 2 (PS2) are members of an evolutionarily conserved gene family. The normal biological role(s) of the presenilins and the mechanism(s) by which the FAD-associated mutations exert their effect remain unknown. Employing in situ hybridization, we demonstrate that the expression patterns of PS1 and PS2 in the brain are extremely similar to each other and that messages for both are primarily detectable in neuronal populations. Immunochemical analyses indicate that PS1 and PS2 are similar in size and localized to similar intracellular compartments (endoplasmic reticulum and Golgi complex). FAD-associated mutations in PS1 and PS2 do not significantly modify either their migration patterns on SDS-polyacrylamide gel electrophoresis or their overall subcellular localization, although subtle differences in perinuclear staining were noted for mutant PS1.

The upstream stimulatory factor functionally interacts with the Alzheimer amyloid beta-protein precursor gene.

The amyloid beta-protein precursor (APP) gives rise to the A beta peptide, which is deposited in the brains of patients with Alzheimer's disease and Down's syndrome. Overexpression of APP due to a third copy of the gene appears to correlate with very early onset of Alzheimer's disease neuropathology in the brains of Down's syndrome patients. Thus, the identification of the factors involved with transcriptional regulation of the APP gene could provide critical clues regarding the events leading to the formation of amyloid deposits. An overlapping AP-1/AP-4 site in the proximal promoter region (-39 to -49) of the human APP gene has previously been shown to increase transcription 4-fold. Here we identify the factor binding specifically to this element as the upstream stimulatory factor USF, unrelated to the c-fos/c-jun complex or the AP-4 factor. In vitro transcription and co-transfection studies show that USF activates transcription from the APP promoter and that the AP-1/AP-4 element participates in this activation. Modulation of APP expression via regulation of USF could potentially ameliorate the production of Alzheimer-augmented beta-amyloid.

A novel protein, amyloid precursor-like protein 2, is present in human brain, cerebrospinal fluid and conditioned media.

A monoclonal antibody, 3B11, was raised to a novel protein, amyloid precursor-like protein 2, which did not recognize amyloid precursor protein. Multiple bands were detected in human brain fractions and cell lysate by Western blotting, indicating the presence of isoforms, 3B11 immunoreactivity was also detected in cerebrospinal fluid and conditioned medium, indicating that the protein is secreted. Immunocytochemistry revealed 3B11 immunoreactivity in sections of human brain.

Reversal of the Swedish familial Alzheimer's disease mutant phenotype in cultured cells treated with phorbol 12,13-dibutyrate.

Protein phosphorylation mediated by phorbol ester stimulates secretion of the beta-amyloid precursor protein (beta-APP) in the cell culture. This increase in secretion is produced by a transient increase in cleavage to produce non-amyloidogenic protease nexin II products mediated by the alpha-secretase activity, and a concomitant decrease in beta-protein production. Cells expressing the Swedish familial Alzheimer's disease (FAD) variant of beta-APP produce more beta-protein and potentially amyloidogenic fragments than cells expressing wild-type protein; furthermore, cleavage shifts from the alpha- to the beta-secretase cleavage site of the precursor. We show that treatment with phorbol 12,13-dibutyrate (PDBu) of cells expressing the Swedish FAD reverses the mutant phenotype to wild-type. The alpha-secretase cleavage increases with a concomitant loss of beta-protein and other beta-secretase cleaved products. These results show that modulating beta-secretase cleavage directly affects beta-protein production. It suggests that activating protein kinase C through, for example, muscarinic receptor agonists could reduce amyloidosis by modulating the level of beta-protein produced.

Study of the phorbol ester effect on Alzheimer amyloid precursor processing: sequence requirements and involvement of a cholera toxin sensitive protein.

Phorbol esters (PDBu) stimulate alpha-secretase cleavage and secretion of the Alzheimer amyloid precursor protein (APP). To determine whether any cytoplasmic residues or sequence motifs mediate the PDBu effect on APP processing, this region of APP was altered by point mutations or deletions. To differentiate the mutated APP from the endogenous APP, the APP751 ectodomain between amino acids 1 and 647 was replaced by a human secreted alkaline phosphatase derivative (SEAP). The resultant fusion protein (SEAP-APP751) was cleaved by alpha-secretase at the same site as full-length APP, and its secretion was stimulated by PDBu at a level similar to APP751. However, PDBu-stimulated secretion of the SEAP-APP751 fusion protein reached its maximum level after 30 min of treatment, while secretion of APP751 reached its maximum after 60 min, suggesting that the APP ectodomain affects the kinetics of APP secretion. Mutation of the cytoplasmic serines to alanines had no effect on the PDBu-stimulated secretion of the SEAP-APP, indicating that protein kinase C (PKC) phosphorylation of the cytoplasmic domain of APP is not important for stimulation of APP secretion. Similarly, deletion of the cytoplasmic domain between amino acids 719 and 751 had no effect on the PDBu-stimulated secretion. However, deletion of amino acids 707-751 resulted in a significant increase in the secretory cleavage of the SEAP-APP707 delta C construct, suggesting that the sequence 707-719 is important for the regulated secretion of APP. Cholera toxin, but not pertussis toxin, reduced the PDBu-induced secretion of APP by more than two-fold, suggesting that the PDBu response may be modulated by a cholera toxin sensitive heterotrimeric G-protein.

Altered cleavage and secretion of a recombinant beta-APP bearing the Swedish familial Alzheimer's disease mutation.

Mutations within the beta-amyloid precursor protein gene cosegregate with the early-onset form of familial Alzheimer's Disease (FAD). It is not known how these mutations result in disease; however, one early-onset AD mutation in a Swedish kindred increases potentially amyloidogenic fragments and beta-protein production in cells expressing the mutant beta-APP. Using a novel recombinant reporter system we found a qualitative change in the secreted product, from cleavage within the beta-protein sequence to cleavage near the N-terminal region of the beta-protein, even though the total amount of secreted mutant product is similar to wild-type. The results suggest that the increased formation of potentially amyloidogenic fragments in cells expressing the Swedish FAD occurs by enzymatic cleavage in the secretory pathway. Alterations in the secretory process may predispose an individual to AD.

Non-amyloidogenic cleavage of the beta-amyloid precursor protein by an integral membrane metalloendopeptidase.

The beta-amyloid precursor protein (beta-APP) is a membrane spanning glycoprotein. The small beta-protein domain within the precursor is presumed to be the source of amyloid found in plaques characteristic of Alzheimer's disease. The amino terminus of beta-APP is released from cells by cleavages that produce both potentially amyloidogenic and nonamyloidogenic fragments of the carboxyl terminus. We developed a cell free system that imposes specificity and co-localization to characterize the proteolytic activity that cleaves the precursor within the beta-protein domain. A reporter protein containing the carboxyl-terminal 105 amino acids of beta-APP provided a specific substrate for cleavage at Lys16 of the beta-protein. The protease inhibitor profile and solubility characteristics of the activity demonstrate the cleavage is produced by an integral membrane metalloendopeptidase.

Processing of the beta-amyloid precursor protein carrying the familial, Dutch-type, and a novel recombinant C-terminal mutation.

Mutations within the beta-amyloid precursor protein (beta-APP) gene that cosegregate with early onset familial Alzheimer's disease (FAD) and hereditary cerebral hemorrhage with amyloidosis of the Dutch-type (HCHWA-D) have been reported. The effects of these mutations on the products of both the non-amyloidogenic and potentially amyloidogenic processing pathways of the beta-APP protein were examined in stably transfected cells. Processing of these mutants appeared to be the same as wild-type. These results contrasted sharply to those observed with a mutation near the amino terminus of the beta-protein domain of beta-APP. This mutation resulted in a two-fold decrease of a potentially amyloidogenic 11 kDa peptide fragment. The data suggest that the FAD and HCHWA-D mutations have no effect on the formation of potentially amyloidogenic fragments in this cell system, possibly implicating an alternative mechanism for their effects.

Mutational analysis of the gag-pol junction of Moloney murine leukemia virus: requirements for expression of the gag-pol fusion protein.

The gag-pol polyprotein of the murine and feline leukemia viruses is expressed by translational readthrough of a UAG terminator codon at the 3' end of the gag gene. To explore the cis-acting sequence requirements for the readthrough event in vivo, we generated a library of mutants of the Moloney murine leukemia virus with point mutations near the terminator codon and tested the mutant viral DNAs for the ability to direct synthesis of the gag-pol fusion protein and formation of infectious virus. The analysis showed that sequences 3' to the terminator are necessary and sufficient for the process. The results do not support a role for one proposed stem-loop structure that includes the terminator but are consistent with the involvement of another stem-loop 3' to the terminator. One mutant, containing two compensatory changes in this stem structure, was temperature sensitive for replication and for formation of the gag-pol protein. The results suggest that RNA sequence and structure are critical determinants of translational readthrough in vivo.

Expression of the gag-pol fusion protein of Moloney murine leukemia virus without gag protein does not induce virion formation or proteolytic processing.

A mutant of Moloney murine leukemia virus was generated in which the UAG termination codon at the 3' end of the gag gene was changed to a CAG codon encoding glutamine. Cells carrying the mutant provirus constitutively express the gag-pol fusion protein and no detectable gag protein. The precursor is stable, is not processed by the protease domain within the precursor, and does not induce assembly and release of virion particles.

Nematode repetitive DNA with ARS and segregation function in Saccharomyces cerevisiae.

Several members of a repetitive DNA family in the nematode Caenorhabditis elegans have been shown to express ARS and centromeric function in Saccharomyces cerevisiae. The repetitive family, denoted CeRep3, consists of dispersed repeated elements about 1 kilobase in length, present 50 to 100 times in the nematode genome. Three elements were sequenced and found to contain DNA sequences homologous to yeast ARS and CEN consensus sequences. Nematode DNA segments containing these repeats were tested for ARS and CEN (or SEG) function after ligation to shuttle vectors and introduction into yeast cells. Such nematode segments conferred ARS function to the plasmid, as judged by an increased frequency of transformation compared with control plasmids without ARS function. Some, but not all, also conferred to the plasmid increased mitotic stability, increased frequency of 2+:2- segregation in meiosis, and decreased plasmid copy number. These effects are similar to those of yeast centromeric DNA. In view of these results, we suggest that the CeRep3 repetitive family may have replication and centromeric functions in C. elegans.

Structure and evolution of a family of interspersed repetitive DNA sequences in Caenorhabditis elegans.

The structure of three members of a repetitive DNA family from the genome of the nematode Caenorhabditis elegans has been studied. The three repetitive elements have a similar unitary structure consisting of two 451-bp sequences in inverted orientation separated by 491 bp, 1.5 kb, and 2.5 kb, respectively. The 491-bp sequence separating the inverted 451-bp sequences of the shortest element is found adjacent to one of the repeats in the other two elements as well. The combination of the three sequences we define as the basic repetitive unit. Comparison of the nucleotide sequences of the three elements has allowed the identification of the one most closely resembling the primordial repetitive element. Additionally, a process of co-evolution is evident that results in the introduction of identical sequence changes into both copies of the inverted sequence within a single unit. Possible mechanisms are discussed for the homogenization of these sequences. A direct test of one possible homogenization mechanism, namely homologous recombination between the inverted sequences accompanied by gene conversion, shows that recombination between the inverted repeats does not occur at high frequency.