Toxicogenomic module associations with pathogenesis: a network-based approach to understanding drug toxicity.

Despite investment in toxicogenomics, nonclinical safety studies are still used to predict clinical liabilities for new drug candidates. Network-based approaches for genomic analysis help overcome challenges with whole-genome transcriptional profiling using limited numbers of treatments for phenotypes of interest. Herein, we apply co-expression network analysis to safety assessment using rat liver gene expression data to define 415 modules, exhibiting unique transcriptional control, organized in a visual representation of the transcriptome (the 'TXG-MAP'). Accounting for the overall transcriptional activity resulting from treatment, we explain mechanisms of toxicity and predict distinct toxicity phenotypes using module associations. We demonstrate that early network responses complement traditional histology-based assessment in predicting outcomes for longer studies and identify a novel mechanism of hepatotoxicity involving endoplasmic reticulum stress and Nrf2 activation. Module-based molecular subtypes of cholestatic injury derived using rat translate to human. Moreover, compared to gene-level analysis alone, combining module and gene-level analysis performed in sequence identifies significantly more phenotype-gene associations, including established and novel biomarkers of liver injury.

Gene expression analysis of the acute phase response using a canine microarray.

The safety of pharmaceuticals is typically assessed in the dog and rat prior to investigation in humans. As a result, a greater understanding of adverse effects in these preclinical testing species would improve safety assessment. Despite this need, there is a lack of tools to examine mechanisms and identify biomarkers in the dog. To address this issue, we developed an Affymetrix-based oligonucleotide microarray capable of monitoring the expression of thousands of canine genes in parallel. The custom canine array contains 22,774 probe sets, consisting of 13,729 canine and 9045 human-derived probe sets. To improve cross-species hybridization with human-derived probes, the detection region was moved from the variable 3' UTR to the more homologous coding region. Testing of this strategy was accomplished by comparing hybridization of naive dog liver RNA to the canine array (coding region design) and human U133A array (standard 3' design). Although raw signal intensity was greater with canine-specific probe sets, human-derived probes detected the expression of additional liver transcripts. To assess the ability of this tool to detect differential gene expression, the acute phase response was examined in beagle dogs given lipopolysaccharide (LPS). Hepatic gene expression 4 and 24 h post-LPS administration was compared to gene expression profiles of vehicle-treated dogs (n=3/group). Array data was consistent with an acute inflammatory response, with transcripts for multiple cytokines and acute phase proteins markedly induced 4 h after LPS challenge. Robust changes in the expression of transcripts involved with glucose homeostasis, biotransformation, and extracellular matrix remodeling were observed 24 h post-dose. In addition, the canine array identified several potential biomarkers of hepatic inflammation. Strong correlations were found between gene expression data and alterations in clinical chemistry parameters such as serum amyloid A (SAA), albumin, and alkaline phosphatase (ALP). In summary, this new genomic tool successfully detected basal canine gene expression and identified novel aspects of the acute phase response in dog that shed new light on mechanisms underlying inflammatory processes.

Temporal gene expression analysis of monolayer cultured rat hepatocytes.

The use of cultured primary hepatocytes within toxicology has proven to be a valuable tool for researchers, however, questions remain with regard to functional differences observed in these hepatocytes relative to the intact liver. Cultured hepatocytes have typically been described as dedifferentiated, a classification based upon the investigation of a few key cellular processes or hepatocellular markers. In the present study, parallel expression monitoring of approximately 8700 rat genes was used to characterize mRNA changes over time in hepatocyte cultures using Affymetrix microarrays. We isolated and labeled mRNA from whole rat livers, hepatocyte-enriched cell pellets, and primary cultured hepatocytes (4, 12, 24, 48, and 72 h postplating), and hybridized these samples to microarrays. From these data, several pairwise and temporal gene expression comparisons were made. Gene expression changes were confirmed by RT/PCR and by performing replicate experiments and repeated hybridizations using a rat toxicology sub-array that contained a 900-gene subset of the 8700-gene rat genomic microarray. PCR data qualitatively reproduced the temporal patterns of gene expression observed with microarrays. Cluster analysis of time course data using self-organizing maps (SOM) revealed a classic hepatocyte dedifferentiation response. Functional grouping of genes with similar transcriptional patterns showed time-dependent regulation of phase I and phase II metabolizing enzymes. In general, cytochrome P450 mRNA expression was repressed, but expression of phase II metabolizing enzymes varied by class (upregulation of glucuronidation, downregulation of sulfation). Potential metabolic targets for toxic insult, such as glutathione metabolism, gluconeogenesis, and glycolysis, were also affected at the transcriptional level. Progressive induction of several genes associated with the cellular cytoskeleton and extracellular matrix was observed in accord with physical changes in cell shape and connectivity associated with cellular adhesion. Finally, many transcriptional changes of genes involved in critical checkpoints throughout the hepatocyte cell cycle and differentiation process were observed. In total, these data establish a more comprehensive understanding of hepatocellular dedifferentiation and reveal many novel aspects of physiological and morphological hepatocyte adaptation. An assembly of all transcripts that demonstrated differential expression in this study can be found in the Supporting Information.

Identification of novel peptide antagonists for von Willebrand factor binding to the platelet glycoprotein Ib receptor from a phage epitope library.

We have constructed a fusion phage epitope library in the filamentous bacteriophage fuse5. The library was made by inserting a degenerate oligonucleotide which encodes 15 variable amino acids into the NH2-terminal region of the phage gene III protein. This library, containing over 10(7) different epitope bearing phage, has been used in an attempt to identify inhibitors of the von Willebrand factor (vWF)-platelet Glycoprotein Ib interaction. The library was screened with a monoclonal antibody (RG46) that recognizes the GPIb binding domain of vWF (amino acids 445-733). A total of 30 clones falling into 8 classes have been identified that react with the RG46 antibody. Isolates from all 8 classes are positive by immunoblot analysis. The amino acid sequence of the gene III fusion protein from positive clones showed a strong homology to the known RG46 epitope. Peptides identified from the screen were synthesized and used to demonstrate that some of the synthetic peptides exhibited inhibitory activity towards ristocetin induced binding of vWF to the GPIb receptor. Thus, we have demonstrated that screening a fusion phage epitope library with a monoclonal antibody that inhibits vWF binding to the GPIb receptor can be a useful tool not only for mapping antibody recognizing determinants, but also can serve as a source for identifying novel peptides that are antagonists for vWF binding to the platelet GPIb receptor.

Identification of a Src SH3 domain binding motif by screening a random phage display library.

A phage display library was constructed in the filamentous bacteriophage fuse5. The library was made by inserting a degenerate oligonucleotide which encodes 15 variable amino acids into the NH2-terminal region of the phage gene III protein. This library, containing over 10(7) different phage, was screened with a glutathione S-transferase (GST) fusion protein containing the Src homology 3 (Src SH3) domain and a protein kinase A phosphorylation site (GST/PKA/Src SH3). A family of proline-rich sequences was isolated following four cycles of enrichment and amplification. Phage containing these sequences were shown to specifically bind to the GST/PKA/Src SH3 protein but not to GST/PKA only. A comparison of the inferred amino acid sequence of the different phage clones revealed a consensus sequence, RPLPXXP, which conforms to a Src SH3 domain binding motif identified independently during an affinity screen of a lambda-lox mouse embryo cDNA library using a 32P-labeled Src SH3 protein fragment as the probe (Y. Ivashchenko, manuscript in preparation). Peptides based upon the 7-amino acid SH3 binding domain core motif displayed strong binding to both the Src and to the Fyn SH3 domains, but failed to bind to the SH3 domain of p21 Ras-GTPase-activating protein (Ras-GAP) and other proteins. We anticipate that further screening of the phage display library will be a useful tool for the rapid identification of additional SH3 domain binding sequences and will also help to establish the essential core motifs that define the specificity of interactions among the diverse proteins containing SH3 domains and those containing SH3 binding motifs.

Expression of human serum albumin in the milk of transgenic mice.

We have tested the feasibility of producing large quantities of human serum albumin (HSA) in the milk of transgenic livestock by generating transgenic mice as a model system. The sheep beta-lactoglobulin (BLG) 5'-regulatory promoter sequences were used to support expression of BLG or HSA in transgenic mice. Transgenic animals generated from the entire BLG gene including 3, 5.5 or 10.8 kb of 5'-sequences demonstrated that 3 kb of 5'-sequences were sufficient to support high levels of expression of BLG, and that the longer 5'-sequences did not improve upon the levels of expression. As such, the 3 kb 5'-sequences were used to drive expression of HSA in BLG-HSA constructs. HSA was not detectably expressed in eight transgenic lines generated from a BLG-HSA construct containing the HSA cDNA. Two transgenic lines of 26 generated, using five different constructs, with an HSA minigene possessing the first intron expressed HSA in their milk. One of these expressed HSA at high levels (2.5 mg ml-1) and has stably transmitted this ability to its progeny. A high percentage of transgenic mouse lines (four of six) generated from a vector containing an HSA minigene possessing introns 1 and 2 expressed HSA in their milk at levels which ranged from 1 to 35 micrograms ml-1. In a similar trend, levels of expression of HSA by transfected tissue culture cells from BLG-HSA vectors containing an introduced SV40 enhancer were low with the HSA cDNA, increased with the HSA minigene with intron 1 and increased further with the minigene containing introns 1 and 2. This study demonstrates that high levels of HSA can be expressed in the milk of transgenic animals, that introns of the HSA gene play a role in its expression and that transfected cell lines may be used to quickly evaluate the relative expression efficiencies of various vector constructs intended for future transgenic evaluation.

Functional modulation of the isolated glycoprotein Ib binding domain of von Willebrand factor expressed in Escherichia coli.

We have expressed in Escherichia coli the domain of von Willebrand factor (vWF) containing the binding site for platelet glycoprotein (GP) Ib and used it to study the regulation of vWF-platelet interaction. The recombinant fragment, comprising residues 445-733 of the mature vWF subunit and designated rvWF445-733, did not have the native conformation of the corresponding domain in the intact molecule because, in order to prevent formation of random aggregates, the seven cysteine residues in the sequence were reduced and alkylated. Unlike native vWF, rvWF445-733 bound to GP Ib in the absence of any modulator, suggesting that the lack of disulfide bonds and/or carbohydrate side chains within this domain may expose platelet interaction sites. In the presence of two modulators, the glycopeptide ristocetin and the snake protein botrocetin, rvWF445-733 inhibited native vWF binding to GP Ib as well as platelet aggregation mediated by vWF, suggesting that both the fragment and the native molecule interact with the same site on platelets. This conclusion was also supported by the observation that the recombinant fragment competed with the binding to platelets of an anti-GP Ib monoclonal antibody known to inhibit vWF binding. Botrocetin formed a complex with rvWF445-733, but the affinity of this interaction was approximately 25-fold lower than with native vWF. However, the complexes of botrocetin with either rvWF445-733 or multimeric native vWF bound to GP Ib with similar dissociation constant. Therefore, conformational attributes of vWF regulate its affinity for botrocetin, but once the complex is formed, interaction with GP Ib is independent of native vWF conformation. These findings provide insights into the regulation of vWF-platelet interaction.

Molecular cloning of a family of retroviral sequences found in chimpanzee but not human DNA.

A number of retrovirus-like sequences have been cloned from chimpanzee DNA which constitute the chimpanzee homologs of the endogenous colobus type C virus CPC-1. One of the clones contains a nearly complete viral genome, but others have sustained deletions of 1 to 2 kilobases in the polymerase gene. The pattern of related sequences detected in other primate species is consistent with the genetic transmission of these sequences for millions of years. However, the appropriately related sequences have not been detected in human, gibbon, or orangutan DNAs. These results suggest either that this family of sequences has been deleted from humans, gibbons, and orangutans, or that the genes were recently acquired in the chimpanzee and gorilla lineages.

Colobus type C virus: molecular cloning of unintegrated viral DNA and characterization of the endogenous viral genomes of Colobus.

The unintegrated viral DNA intermediates of colobus type C virus (CPC-1) were isolated from infected human cells that were permissive for viral growth. There were two major species of DNA, linear molecules with two copies of the long terminal repeat and relaxed circles containing only a single long terminal repeat. In addition, there was a minor species (approximately 10%) composed of relaxed circles with two copies of the long terminal repeat. A restriction endonuclease map of the unintegrated DNA was constructed. The three EcoRI fragments of circular CPC-1 DNA were cloned in the EcoRI site of lambda gtWES . lambda B and then subcloned in the EcoRI site of pBR322. Using these subgenomic fragments as probes, we have characterized the endogenous viral sequences found in colobus cellular DNA. They are not organized in tandem arrays, as is the case in some other gene families. The majority of sequences detected in cellular DNA have the same map as the CPC-1 unintegrated DNA at 17 of 18 restriction endonuclease sites. There are, however, other sequences that are present in multiple copies and do not correspond to the CPC-1 map. They do not contain CPC-1 sequences either in an altered form or fused to common nonviral sequences. Instead, they appear to be derived from a distinct family of sequences that is substantially diverged from the CPC-1 family. This second family of sequences, CPC-2, is also different from the sequences related to baboon endogenous type C virus that forms a third family of virus-related sequences in the colobus genome.