Michael P. Lawton

Pathway analysis using random forests classification and regression

MOTIVATION Although numerous methods have been developed to better capture biological information from microarray data, commonly used single gene-based methods neglect interactions among genes and leave room for other novel approaches. For example, most classification and regression methods for microarray data are based on the whole set of genes and have not made use of pathway information. Pathway-based analysis in microarray studies may lead to more informative and relevant knowledge for biological researchers. RESULTS In this paper, we describe a pathway-based classification and regression method using Random Forests to analyze gene expression data. The proposed methods allow researchers to rank important pathways from externally available databases, discover important genes, find pathway-based outlying cases and make full use of a continuous outcome variable in the regression setting. We also compared Random Forests with other machine learning methods using several datasets and found that Random Forests classification error rates were either the lowest or the second-lowest. By combining pathway information and novel statistical methods, this procedure represents a promising computational strategy in dissecting pathways and can provide biological insight into the study of microarray data. AVAILABILITY Source code written in R is available from http://bioinformatics.med.yale.edu/pathway-analysis/rf.htm.

Increased Serum Enzyme Levels Associated with Kupffer Cell Reduction with No Signs of Hepatic or Skeletal Muscle Injury

Macrophage colony-stimulating factor (M-CSF) is a hematopoietic growth factor that is responsible for the survival and proliferation of monocytes and the differentiation of monocytes into macrophages, including Kupffer cells (KCs) in the liver. KCs play an important role in the clearance of several serum enzymes, including aspartate aminotransferase and creatine kinase, that are typically elevated as a result of liver or skeletal muscle injury. We used three distinct animal models to investigate the hypothesis that increases in the levels of serum enzymes can be the result of decreases in KCs in the apparent absence of hepatic or skeletal muscle injury. Specifically, neutralizing M-CSF activity via a novel human monoclonal antibody reduced the CD14(+)CD16(+) monocyte population, depleted KCs, and increased aspartate aminotransferase and creatine kinase serum enzyme levels in cynomolgus macaques. In addition, the treatment of rats with clodronate liposomes depleted KCs and led to increased serum enzyme levels, again without evidence of tissue injury. Finally, in the osteopetrotic (Csf1(op)/Csf1(op)) mice lacking functional M-CSF and having reduced levels of KCs, the levels of serum enzymes are higher than in wild-type littermates. Together, these findings support a mechanism for increases in serum enzyme levels through M-CSF regulation of tissue macrophage homeostasis without concomitant histopathological changes in either the hepatic or skeletal system.

The Role of Hypoxia in 2-Butoxyethanol–Induced Hemangiosarcoma

To understand the molecular mechanisms underlying compound-induced hemangiosarcomas in mice, and therefore, their human relevance, a systems biology approach was undertaken using transcriptomics and Causal Network Modeling from mice treated with 2-butoxyethanol (2-BE). 2-BE is a hemolytic agent that induces hemangiosarcomas in mice. We hypothesized that the hemolysis induced by 2-BE would result in local tissue hypoxia, a well-documented trigger for endothelial cell proliferation leading to hemangiosarcoma. Gene expression data from bone marrow (BM), liver, and spleen of mice exposed to a single dose (4 h) or seven daily doses of 2-BE were used to develop a mechanistic model of hemangiosarcoma. The resulting mechanistic model confirms previous work proposing that 2-BE induces macrophage activation and inflammation in the liver. In addition, the model supports local tissue hypoxia in the liver and spleen, coupled with increased erythropoeitin signaling and erythropoiesis in the spleen and BM, and suppression of mechanisms that contribute to genomic stability, events that could be contributing factors to hemangiosarcoma formation. Finally, an immunohistochemistry method (Hypoxyprobe) demonstrated that tissue hypoxia was present in the spleen and BM. Together, the results of this study identify molecular mechanisms that initiate hemangiosarcoma, a key step in understanding safety concerns that can impact drug decision processes, and identified hypoxia as a possible contributing factor for 2-BE–induced hemangiosarcoma in mice.

Acute Drug-Induced Vascular Injury in Beagle Dogs: Pathology and Correlating Genomic Expression

Acute vascular injury that leads to vascular inflammation is a common finding in the preclinical toxicity testing of drugs in rats and dogs. However, the relevance of this finding for risk to humans is unclear. Concern about the safety of these drugs is heightened by the current lack of noninvasive clinical methods to predict the onset of vascular damage in animals or humans. Determining the relevance of this poorly understood preclinical outcome for humans requires a better understanding of the molecular mechanisms of injury in addition to the development of sensitive and specific leading biomarkers for the clinical diagnosis of acute vascular damage. Most molecular research on this toxicity has been performed in rats, but recent development of canine gene expression microarrays makes transcriptomic studies now possible in the dog. In this study, we investigated the molecular mechanisms of drug-induced vascular injury in dogs using gene arrays. After treating Beagles with toxic doses of CI-947, an adenosine receptor agonist, we profiled gene expression in the coronary arteries and correlated those changes with histopathology at 16 and 24 hours after dosing. The results demonstrated that pathobiological processes such as stimulation of the innate immune response, increased extracellular matrix turnover and oxidative stress were active at times of very early injury.

Nonclinical Safety Biomarkers of Drug-induced Vascular Injury Current Status and Blueprint for the Future

Better biomarkers are needed to identify, characterize, and/or monitor drug-induced vascular injury (DIVI) in nonclinical species and patients. The Predictive Safety Testing Consortium (PSTC), a precompetitive collaboration of pharmaceutical companies and the U.S. Food and Drug Administration (FDA), formed the Vascular Injury Working Group (VIWG) to develop and qualify translatable biomarkers of DIVI. The VIWG focused its research on acute DIVI because early detection for clinical and nonclinical safety monitoring is desirable. The VIWG developed a strategy based on the premise that biomarkers of DIVI in rat would be translatable to humans due to the morphologic similarity of vascular injury between species regardless of mechanism. The histomorphologic lexicon for DIVI in rat defines degenerative and adaptive findings of the vascular endothelium and smooth muscles, and characterizes inflammatory components. We describe the mechanisms of these changes and their associations with candidate biomarkers for which advanced analytical method validation was completed. Further development is recommended for circulating microRNAs, endothelial microparticles, and imaging techniques. Recommendations for sample collection and processing, analytical methods, and confirmation of target localization using immunohistochemistry and in situ hybridization are described. The methods described are anticipated to aid in the identification and qualification of translational biomarkers for DIVI.

Three-Dimensional Quantitative Structure-Activity Relationship Analysis of Human CYP51 Inhibitors

CYP51 fulfills an essential requirement for all cells, by catalyzing three sequential mono-oxidations within the cholesterol biosynthesis cascade. Inhibition of fungal CYP51 is used as a therapy for treating fungal infections, whereas inhibition of human CYP51 has been considered as a pharmacological approach to treat dyslipidemia and some forms of cancer. To predict the interaction of inhibitors with the active site of human CYP51, a three-dimensional quantitative structure-activity relationship model was constructed. This pharmacophore model of the common structural features of CYP51 inhibitors was built using the program Catalyst from multiple inhibitors (n = 26) of recombinant human CYP51-mediated lanosterol 14α-demethylation. The pharmacophore, which consisted of one hydrophobe, one hydrogen bond acceptor, and two ring aromatic features, demonstrated a high correlation between observed and predicted IC50 values (r = 0.92). Validation of this pharmacophore was performed by predicting the IC50 of a test set of commercially available (n = 19) and CP-320626-related (n = 48) CYP51 inhibitors. Using predictions below 10 μM as a cutoff indicative of active inhibitors, 16 of 19 commercially available inhibitors (84%) and 38 of 48 CP-320626-related inhibitors (79.2%) were predicted correctly. To better understand how inhibitors fit into the enzyme, potent CYP51 inhibitors were used to build a Cerius2 receptor surface model representing the volume of the active site. This study has demonstrated the potential for ligand-based computational pharmacophore modeling of human CYP51 and enables a high-throughput screening system for drug discovery and data base mining.

Analysis of changes in hepatic gene expression in a murine model of tolerance to acetaminophen hepatotoxicity (autoprotection)

Pretreatment of mice with a low hepatotoxic dose of acetaminophen (APAP) results in resistance to a subsequent, higher dose of APAP. This mouse model, termed APAP autoprotection was used here to identify differentially expressed genes and cellular pathways that could contribute to this development of resistance to hepatotoxicity. Male C57BL/6J mice were pretreated with APAP (400mg/kg) and then challenged 48h later with 600mg APAP/kg. Livers were obtained 4 or 24h later and total hepatic RNA was isolated and hybridized to Affymetrix Mouse Genome MU430_2 GeneChip. Statistically significant genes were determined and gene expression changes were also interrogated using the Causal Reasoning Engine (CRE). Extensive literature review narrowed our focus to methionine adenosyl transferase-1 alpha (MAT1A), nuclear factor (erythroid-derived 2)-like 2 (Nrf2), flavin-containing monooxygenase 3 (Fmo3) and galectin-3 (Lgals3). Down-regulation of MAT1A could lead to decreases in S-adenosylmethionine (SAMe), which is known to protect against APAP toxicity. Nrf2 activation is expected to play a role in protective adaptation. Up-regulation of Lgals3, one of the genes supporting the Nrf2 hypothesis, can lead to suppression of apoptosis and reduced mitochondrial dysfunction. Fmo3 induction suggests the involvement of an enzyme not known to metabolize APAP in the development of tolerance to APAP toxicity. Subsequent quantitative RT-PCR and immunochemical analysis confirmed the differential expression of some of these genes in the APAP autoprotection model. In conclusion, our genomics strategy identified cellular pathways that might further explain the molecular basis for APAP autoprotection.

Identification and Elucidation of the Biology of Adverse Events: The Challenges of Safety Assessment and Translational Medicine

There has been an explosion of technology-enabled scientific insight into the basic biology of the causes of adverse events. This has been driven, in part, by the development of the various “omics” tools (e.g., genomics, proteomics, and metabolomics) and associated bioinformatics platforms. Meanwhile, for decades, changes in preclinical testing protocols and guidelines have been limited. Preclinical safety testing currently relies heavily on the use of outdated animal models. Application of systems biology methods to evaluation of toxicities in oncology treatments can accelerate the introduction of safe, effective drugs. Systems biology adds insights regarding the causes and mechanisms of adverse effects, provides important and actionable information to help understand the risks and benefits to humans, focuses testing on methods that add value to the safety testing process, and leads to modifications of chemical entities to reduce liabilities during development. Leveraging emerging technologies, such as genomics and proteomics, may make preclinical safety testing more efficient and accurate and lead to better safety decisions. The development of a U.S. Food and Drug Administration guidance document on the use of systems biology in clinical testing would greatly benefit the development of drugs for oncology by communicating the potential application of specific methodologies, providing a framework for qualification and application of systems biology outcomes, and providing insight into the challenges and limitations of systems biology in the regulatory decision-making process. Clin Cancer Res; 17(21); 6641–5. ©2011 AACR.

Effects of Modulating In Vivo Nitric Oxide Production on the Incidence and Severity of PDE4 Inhibitor–Induced Vascular Injury in Sprague-Dawley Rats

Drug-induced vascular injury (DIVI) is observed in rat mesenteric arterioles in response to treatment with phosphodiesterase-4 inhibitors (PDE4i). However, the mechanisms responsible for causing the characteristic vascular lesions are unclear. Nitrotyrosine (NT) adducts, markers of local nitric oxide (NO) production, have been observed in close proximity to the arterial lesions and in the inflammatory cells associated with DIVI. To determine if NO has a direct role in DIVI, rats were treated with the PDE4i CI-1044 at 10, 20, or 40 mg/kg alone or in combination with the nitric oxide synthase inhibitor L-NAME (60 mg/kg) or the nitric oxide donor SIN-1 (30 mg/kg). Mesenteries were collected and processed for microscopic evaluation. NT formation was evaluated in situ via immunohistochemical staining. Serum nitrite (SN), a marker of in vivo NO production, was measured. Compared with vehicle controls, treatment with CI-1044 alone resulted in dose-related increases in the frequency and severity of vascular injury, SN levels, and NT residues. SIN-1 coadministration caused vascular injury to occur at lower doses of CI-1044, compared with CI-1044 alone, with the overall incidence and severity of injury being greater across all CI-1044-dose groups. Following administration of 20 or 40 mg/kg CI-1044, there were also increases in NT immunoreactivity when SIN-1 was coadministered and significant increases in SN. Conversely, coadministration of L-NAME resulted in marked reduction of injury, NT, and SN when compared with CI-1044 alone. The present study suggests that NO production is closely linked to PDE4i-induced vascular injury.

The Role of eNOS Phosphorylation in Causing Drug-induced Vascular Injury

Previously we found that regulation of eNOS is an important part of the pathogenic process of Drug-induced vascular injury (DIVI) for PDE4i. The aims of the current study were to examine the phosphorylation of eNOS in mesentery versus aorta at known regulatory sites across DIVI-inducing drug classes and to compare changes across species. We found that phosphorylation at S615 in rats was elevated 35-fold 2 hr after the last dose of CI-1044 in mesentery versus 3-fold in aorta. Immunoprecipitation studies revealed that many of the upstream regulators of eNOS activation were associated with eNOS in 1 or more signalosome complexes. Next rats were treated with drugs from 4 other classes known to cause DIVI. Each drug was given alone and in combination with SIN-1 (NO donor) or L-NAME (eNOS inhibitor), and the level of eNOS phosphorylation in mesentery and aorta tissue was correlated with the extent of vascular injury and measured serum nitrite. Drugs or combinations produced altered serum nitrite levels as well as vascular injury score in the mesentery. The results suggested that phosphorylation of S615 may be associated with DIVI activity. Studies with the species-specific A2A adenosine agonist CI-947 in rats versus primates showed a similar pattern.