Abstract

The ability of toxicogenomics to predict drug toxicity is limited in part because animal and cell- based models are not able to capture the range of responses often seen in genetically diverse patient populations. To address this problem, a genetically diverse and reproducible Rat Test Population was created using a combinatorial breeding strategy. Drug studies have shown that the Rat Test Population can detect an adverse drug response that is sometimes missed by the inbred F344 and outbred Sprague-Dawley strains. In one case study, the Rat Test Population was employed to identify strains that reflect the hyperlipidemia associated with HIV therapeutic agents, including the protease inhibitor, Ritonavir, in humans. Ritonavir (10-100 mg/kg/day) was administered to the Rat Test Population and Sprague-Dawley rats for five days. Following drug treatment, blood samples were evaluated for changes in serum lipids. Significant increases in triglyceride and cholesterol levels were observed only in the BNxLEW rats from the Rat Test Population that were treated with 75-100 mg/kg/day of Ritonavir. Livers from the sensitive BNxLEW rats and resistant Sprague-Dawley rats were then evaluated for differences in gene expression profiles using Affymetrix rat genome RAE230A arrays. 304 differentially expressed genes correlated with either cholesterol or triglyceride changes in the BNxLEW rats, 164 of which were also differentially expressed in the Sprague-Dawley rats. Lipid-metabolism genes that were regulated differently between the sensitive BNxLEW and resistant Sprague-Dawley rats include CITED2, G6PC and ME1. CYP7A1, a key enzyme for cholesterol metabolism, was found down regulated in the BNxLEW strain, but it was up regulated in the Sprague-Dawley rats. In a second case study, serum and tissue samples were collected from F344 rats and the Rat Test Population following treatment with the hepatotoxin acetaminophen (1,000 mg/kg for 24 hours). Liver damage was assessed by ALT and found to be 10 to 20-fold greater in F344 and two sensitive strains (F344xBN, F344xLEW) than in a resistant strain (F344xWKY). Histopathology revealed extensive hemorrhage and necrosis in the sensitive strains. Microarray analysis of liver tissues detected 300 differentially expressed genes for the F344 rat and 502 for the three strains from the Rat Test Population (all share the F344 genome). Finding both sensitive and resistant strains from the Rat Test Population that share a common genome background enabled genes/ESTs to be categorized for drug exposure (38), toxicity (47) and resistance (97). Collectively, the results demonstrate the utility of using the Rat Test Population to identify rat strains that may have more clinical relevance for a particular adverse drug effect. In both case studies, the genome-wide expression profiling on sensitive and resistant strains provided information regarding the pathways determining adverse drug responses.

Key concepts

• For preclinical drug studies, the use of the Rat Test Population offers researchers the chance to find a sensitive strain. A sensitive strain can then be reproduced for follow-up experiments (Ritonavir Case Study).
• Integrating microarray data with phenotypic data from the Rat Test Population enables complex gene expression signatures to be simplified. Use of sensitive and resistant strains helps to distinguish the different causes of gene expression responses following administration of a drug (Ritonavir and Acetaminophen Case Study).

Combinatorial breeding of the Rat Test Population

Four of the most genetically diverse inbred rat strains were selected using a Strain Calculator to create a panel of six F1 offspring. These six strains capture 57% of the allelic diversity in strains of rats.

• Advantages of the Rat Test Population:

1. Six different genome background combinations maximize heterogeneity (like outbred rats, but unlike inbred rats).
2. Rats from each F1 hybrid strain are genetically equivalent (like inbred rats).
3. The genotype of the each strain can be reproduced for further studies (unlike outbred rats).

Ritonavir Case Study

Male rats were dosed orally for 5 days with vehicle (5% ethanol: 95% propylene glycol) or Ritonavir (10-100 mg/kg/day) and sacrificed on the sixth day. Blood samples were collected for measuring clinical chemistry parameters and drug levels. Livers were collected for histological evaluation and microarray analysis. All rats were fed pelleted diet AIN-76A 0.4% NaCl + 20% corn oil for two weeks prior to and during drug treatment.

Ritonavir-induced serum lipid elevation is strain-specific

Serum lipid levels were determined for the Rat Test Population and Sprague-Dawley rats. The BNxLEW strain was found to have elevated serum lipid levels. Changes in cholesterol and triglyceride levels are expressed as mean± SEM. *Statistically different from vehicle at p≤0.05.

A. BNxLEW Rats

B. Sprague-Dawley Rats

Genes correlated with serum lipid levels upon Ritonavir treatment.

For microarray analysis, RNA from livers was extracted, amplified, and analyzed using Affymetrix RAE230A Rat microarray chips representing approximately 15,000 genes. Results were analyzed using Rosetta Resolver software.

A. Genes correlated with serum lipid levels in BNxLEW rats. B. Genes differentially regulated between BNxLEW and Sprague-Dawley rats. Genes are selected using ANOVA analysis to identify the sequences that are differentially regulated between BNxLEW and Sprague-Dawley rats treated with 75 and 100 mg/kg RTV (false discovery rate ≤ 0.01).

Gene expression changes involved in lipid metabolism and biosynthesis upon Ritonavir treatment.

• Gene expression changes are shown relative to the vehicle-treated controls. Cyp7a1: Cytochrome P450 7A1; G6pc: Glucose-6-Phosphatase; Cited2: CBP/P300 Interacting Transactivator; Me1: Malic enzyme 1.

Acetaminophen Case Study

Rats were administered a single intraperitoneal dose of acetaminophen (1,000 mg/kg) and the study terminated after 24 hours. The susceptibility of animals was determined from clinical chemistry measurements (ALT), histopathology (hemorrhage) and having a common genome background (F344). RNA from 4 drug treated animals was hybridized with a pool of vehicle RNA (4 individuals) to 18,000 element cDNA microarrays. Duplicate arrays with a flip-dye design resulted in 16 measurements for each gene/EST per strain. Genes/ESTs were filtered based on criteria that they are differentially expressed (P<0.05) in 75% of the individuals for each strain. Drug toxicity signatures were then sub-classified into genes/ESTs for toxicity, resistance, exposure or due to strain-specific reasons by overlapping the responses from multiple strains that are either sensitive (F344, F344xBN, F344xLEW) or less affected (F344xWKY) to the drug. “S” refers to sensitive; “R”refers to less affected/resistant. For ALT: *P<0.05 for drug vs. vehicle. †P<0.05 vs. Sprague-Dawley (CD-IGS).

A. Clinical chemistry (ALT)

B. Gene Expression sub-classification

Sub-classes for the genes most responsive to acetaminophen

Conclusion

1. For the Ritonavir Case Study, a sensitive strain of rat was identified in the Rat Test Population that responded to the hyperlipidemic effects of the drug seen in humans
   patients. There were 304 differentially expressed genes that correlated with either cholesterol and triglyceride changes and four of these shown here are known to be involved in lipid metabolism and biosynthesis
2. For the Acetaminophen Case Study, we integrated microarray data with phenotypic data from Rat Test Population and F344 rats to prioritize 47 differentially expressed genes/ESTs for toxicity from over 800 that we originally detected. This enabled a complex gene expression signature to be greatly simplified.
3. Both studies yielded information regarding the pathways determining an adverse drug response.

This work was supported by NIH grants 1R43 CA106153-01, 5R44 ES011432-03

Steven H. Nye, Yi Yang, Annette J. Dahly-Vernon, Darin L. Evan1, Eric A. Blomme, Dale J. Kempf, Larry Klein,
Kennan C. Marsh, Jeffrey F. Waring, Howard J. Jacob, and Richard J. Roman.