Gene Expression Analysis of the Acute Phase Response Using a Canine Microarray

M. A. Higgins*, B. R. Berridge{dagger}, B. J. Mills{ddagger}, A. E. Schultze{dagger}, H. Gao*, G. H. Searfoss*, T. K. Baker* and T. P. Ryan*,1

* Department of Lead Optimization Toxicology and {dagger} Department of Pathology, Lilly Research Laboratories and {ddagger} Elanco Animal Health, Divisions of Eli Lilly and Company, Greenfield, IN 46140

Received February 17, 2003; accepted April 22, 2003


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
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.

Key Words: acute phase response; canine genomics; dog; gene expression; inflammation; lipopolysaccharide (LPS); liver; microarray; toxicogenomics.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
In drug development, safety assessment is typically based on toxicology data from one rodent and one nonrodent species. A recent study by Olson et al. (2000)Go demonstrated that nonrodent studies were more predictive of human toxicities than rodent studies. Although preclinical toxicology findings were frequently observed in both dog and rat, the dog demonstrated better concordance with humans when toxicities were observed in only one species. In general, toxicology data obtained from dog studies have demonstrated good potential to predict human toxicities from many therapeutic classes; however, a number of toxicities observed in dog are not relevant to human risk assessment (Olson et al., 1998Go). Without additional information, interpretation of these data could result in erroneous extrapolation of preclinical data to safety assessment (reviewed in Louden et al., 2001). As a result, a better understanding of species-specific responses to xenobiotic exposure is needed to determine the extent to which animal models reflect the human situation. Investigations at the molecular level are needed to provide critical insight into cases in which metabolic and toxic outcomes differ between sites of action and target organs across species. With this improved understanding, we will be able to more fully utilize toxicology data obtained from preclinical animal studies.

Microarray technology is a powerful tool that can be used to gain insight into species differences in toxicological responses, and is currently being employed to elucidate molecular mechanisms and identify markers of toxicity in rodents and humans (Baker et al., 2003Go; Pennie et al., 2000Go; Ulrich and Friend, 2002Go). In addition, gene expression patterns have recently been shown to successfully predict molecular sequelae and clinical outcomes, with notable success in the field of cancer research (Bertucci et al., 2000Go; Devilard et al., 2002Go; Mitsiades et al., 2002Go; Niedergethmann et al., 2002Go). The abundance of sequence information for human (~100,000), rat (~32,000), and mouse (~40,000) in public databases such as GenBank (http://www.ncbi.nlm.nih.gov) has resulted in the construction of species-specific microarrays by several academic and commercial laboratories. In comparison, there are only ~5500 canine sequence entries in the public domain, many of which are not suitable for array construction. This paucity of publicly available canine sequence information, coupled with the high cost associated with sequencing efforts and array design, has precluded the development of a dog-specific oligonucleotide array. Consequently, the majority of preclinical gene expression studies designed to address toxicity issues and gain mechanistic insight have been limited to rodents.

The development of a canine array is of critical importance to the pharmaceutical industry, as it provides opportunity for canine biomarker identification and mechanism of action elucidation. Furthermore, availability of dog-specific genomic tools will increase the ability to monitor toxicities in real time and potentially improve human risk assessment. More sensitive preclinical and clinical toxicology biomarkers are needed, as a number of adverse events are missed or poorly predicted with current markers. For example, bacterial-mediated sepsis is associated with a 10–40% mortality rate despite the availability of several inflammatory and acute phase response biomarkers (cited in Hewett and Roth, 1993Go). The high fatality rate associated with this condition highlights the need for new, highly specific markers that can be monitored in real time.

Herein we describe the design, construction, and application of a custom high-density oligonucleotide microarray specifically designed to monitor transcriptional responses in dog. Over 10,000 canine nonredundant expressed sequence tags (ESTs) were obtained through the DNA sequencing efforts of LION bioscience and 3650 canine sequences were obtained from the public domain (GenBank). The addition of 8945 probe sets designed to human reference sequences (RefSeqs) and 100 human normalization probe sets increased the transcript monitoring capability of the canine array and facilitated direct comparison with corresponding probe sets on the Affymetrix human U133A array (Santa Clara, CA). To test the ability to detect differential gene expression, the canine array was employed to identify mechanisms and biomarkers of the acute phase response in the liver following administration of lipopolysaccharide (LPS) to male beagle dogs. The LPS-induced model of endotoxemia was selected because it has well-characterized clinical chemistry and gene expression endpoints in many species (Hewett and Roth, 1993Go; Panciera et al., 2003Go). Furthermore, changes in hepatic gene expression during the acute phase response are robust (Bulera et al., 2001Go; Tygstrup et al., 2002Go) and appear to mirror serum changes in acute phase proteins. Since traditional markers of the acute phase response change in a similar manner across species, comparison of transcript profiles in dog with hepatic gene expression patterns, blood chemistry, and hematological parameters in other species was used to validate the canine array. In the present study, profound early (4 h) and late (24 h) responses to bacterial endotoxin were observed at the transcriptional level. A marked induction of transcripts encoding inflammatory mediators was observed at 4 h, whereas transcripts involved with glucose metabolism, biotransformation, and extracellular matrix remodeling were downregulated 24 h after LPS administration. In addition, application of the canine array identified a number of potential novel biomarkers of acute inflammation, such as stromal cell-derived factor 2-like 1 (SDF2L1) and protein transport protein sec61 alpha subunit (sec61a).


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Canine array design: Features and sequence content.
An oligonucleotide array (12.8 mm) was constructed based on the current design format of the Affymetrix human U133 array set—18 micron feature size, 11 probe pairs per sequence, and 25-mer oligonucleotide probes, as described in technical note "Array Design for the GeneChip® Human Genome U133 Set." The canine array contains 22,774 probe sets, utilizing 11 Perfect Match/Mismatch pairs per transcript to fill approximately 250,000 array features. A total of 13,729 dog-specific probe sets were tiled on the array, including 10,079 proprietary sequences purchased from LION bioscience and 3650 public canine sequences obtained from GenBank. Dog sequences were subjected to rigorous assessment to remove low quality sequences, vector contamination, and sequence tagged site (STS) primers prior to submission for probe design. Directionality (5' to 3' orientation) of a subset of dog sequences was determined by manual inspection. An additional 8945 probe sets for fully annotated human RefSeqs were included on the array (GenBank). Microarrays were synthesized by Affymetrix using photolithographic methods as described in the literature (Jacobs and Fodor, 1994Go; Mazzola and Fodor, 1995Go).

Probe design and selection.
Probes were designed using standard design guidelines and algorithms as outlined in Affymetrix literature "GeneChip® Custom Array Design Guide" and "New Statistical Algorithms for Monitoring Gene Expression on GeneChip® Probe Arrays." To maximize cross-species hybridization efficiency, probe selection regions differed for dog and human sequences. As shown in Figure 1Go, dog-specific probes were designed with a 3'-bias using a standard Affymetrix algorithm targeting the terminal 600 bases of the represented transcript. To capitalize on higher homology in the coding region compared to the 3' UTR, human-derived probes were designed to the terminal 600 bases of the coding region. Each probe was assigned a quality score based on predicted hybridization efficiency, and then ranked with respect to the following parameters: probe sort score, probe count, average raw probe score, raw standard deviation, and length of the actual probe selection region (sif length) for the probe set. Scores were penalized for cross-hybridizing and overlapping probes. Prioritization was also based on type of probe set (unique, identical, or mixed), with the highest probe set score chosen as the exemplar for identical probe types. All probe sets with a sort score greater than 2.0 for dog or 4.0 for human, based on slope response and independence of each probe in the set, were included on the array. Multiple probe sets exist on the array for some genes due to redundancy within the public dog database and overlap between proprietary and public sequences.



View larger version (14K):
[in this window]
[in a new window]
 
FIG. 1. Schematic representation of design strategy for canine and human-derived probes on the canine microarray.

 
Array control parameters.
To assist in evaluation of chip performance and scaling of intensity data, several types of control probe sets were included on the array. Noneukaryotic Affymetrix hybridization (bioB, bioC, bioD, and cre) and polyA (dap, lys, phe, and thr) probe sets were included on the array to assess inter-array hybridization using spiked-in labeled controls at known concentrations. Edge controls, corner checkerboards, and a center cross were included on each array for grid alignment purposes. Canine-specific probe sets from the Affymetrix GeneChip® Test3 array representing beta-actin, beta-3 adrenergic receptor, glucose-6-phosphatase, glyceraldehyde-3-phosphate dehydrogenase, and cytochrome P450 (CYP) 2E1 were also tiled on the canine array. Probes were designed to the 5', middle, and 3' regions of these transcripts to control for RNA quality and labeling efficiency. An additional 100 normalization control probe sets identical to those contained on the human U133A array were tiled on the canine array for cross-array comparisons. To compare hybridization efficiency between probes targeting the 3' and coding region of human sequences, liver samples from four naive dogs were sequentially hybridized to canine and human U133A arrays.

Animal studies.
Nine male beagle dogs (Marshall Farms, North Rose, NY) were acclimated for a minimum of 3 weeks to caging and feeding conditions. Dogs (8–25 months, 5–11 kg) were fed approximately 300 g of Harlan Teklad 2021 every morning and had free access to drinking water. All animal procedures were performed according to protocols approved by the Eli Lilly Animal Care and Use Committee. Dogs were randomly assigned to one of three experimental groups: Control (n = 3, Group 01), LPS 4 h (n = 3, Group 02), or LPS 24 h (n = 3, Group 03). Treated dogs received a single intravenous dose of 0.2 mg/kg of LPS (Escherichia coli Serotype 055:B5) suspended in saline (Sigma Chemical Co., St. Louis, MO). Control animals received vehicle only. Dose selection was based on a study by Cullen et al. (1998)Go that demonstrated endotoxin-mediated effects on the gastrointestinal tract without producing mortality in dogs at the selected dose. Dogs were fasted overnight prior to dosing and 4 h prior to necropsy. Animals were euthanized 4 and 24 h post-dosing, and various tissues were collected in RNAlaterTM (Ambion, Austin, TX) for gene expression analysis.

Clinical chemistry and hematological evaluation.
Blood samples were collected pre-study, 2, 4, 8, and 24 h post-dose and processed for clinical chemistry and hematological evaluation. Whole blood for complete blood counts was collected into Monoject blood collection vacuum tubes containing liquid EDTA(K3) (Sherwood Medical, St. Louis, MO). Reticulocyte, erythrocyte, platelet, total leukocyte, neutrophil, lymphocyte, monocyte, eosinophil, and basophil counts, hemoglobin concentration, and mean corpuscular volume (MCV) were determined using an Advia 120 Hematology System (Bayer Corporation, Tarrytown, NY). Hematocrit, mean corpuscular hemoglobin (MCH), and mean corpuscular hemoglobin concentration (MCHC) were calculated values. Smears of fresh EDTA-anticoagulated blood were stained with Wright-Giemsa stain using the HMS Series Programmable Slide Stainer (Carl Zeiss Inc., Thornwood, NY) and evaluated for blood cell morphology.

Whole blood for coagulation studies was collected in vacuum tubes containing 3.8% sodium citrate anticoagulant (Becton Dickinson, Franklin Lakes, NJ). Citrated plasma was collected following centrifugation of samples at 2–8°C in an AllegraTM 6R centrifuge (Beckman Instruments, Fullerton, CA) for 10 min at 3000 rpm. Prothrombin time and activated partial thromboplastin time were determined using a STA analyzer (Diagnostica Stago, Parsippany, NJ). Whole blood for clinical chemistry determinations was collected in vacuum tubes containing no anticoagulant, and samples were allowed to clot at room temperature. Serum was collected following centrifugation of samples in a Beckman GS-6 centrifuge for 10 min at 3000 rpm. Concentrations of urea nitrogen, creatinine, total bilirubin, sodium, potassium, chloride, calcium, inorganic phosphorus, glucose, cholesterol, triglycerides, total protein, albumin, and the activities of alkaline phosphatase (ALP), gamma glutamyltransferase (GGT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and creatine kinase in serum were determined using standard methods on a Hitachi 917 Automatic Analyzer (Roche Diagnostics Corporation, Indianapolis, IN). Concentrations of c-reactive protein (CRP) and SAA were determined with commercially available immunoassay kits (Tri-Delta Diagnostics, Morris Plains, NJ) using the SPECTRAmax® 340 micro-plate spectrophotometer (Molecular Devices Corporation, Sunnyvale, CA). Concentration of cardiac troponin I was determined using the Immulite Immunoanalyzer (Diagnostic Products Corporation, Los Angeles, CA). Histopathological evaluations were performed on the liver, kidney, heart, and gastrointestinal tissues.

Gene expression analysis: Sample preparation.
The canine array was used to examine hepatic gene expression 4 and 24 h following administration of a single 0.2 mg/kg intravenous dose of LPS to beagle dogs. Approximately 150 mg of liver from naive or LPS-treated dogs was homogenized in 1 ml RNA Stat60 (Tel-Test, Friendswood, TX) and total RNA was extracted according to the manufacturer’s protocol. Further purification of RNA was carried out with RNeasy columns (Qiagen, Valencia, CA). Samples were processed both individually and in pools containing an equivalent amount of total RNA from dogs within each treatment group (n = 3). Total RNA (15 µg) was reverse transcribed into double-stranded cDNA using Superscript II (Invitrogen, Carlsbad, CA) and an oligo T-7-(dT)24 primer (Operon, Alameda, CA). Biotinylated cRNA was synthesized using the BioArray T-7 polymerase labeling kit (Enzo, Farmingdale, NY) and fragmented prior to overnight hybridization with either the canine array or human U133A array. As per the "Affymetrix GeneChip® Expression Analysis Technical Manual," arrays were washed and stained using the EukGE-WS2 antibody amplification protocol. Data analysis was performed using MicroArray Suite (MAS) 5.0 and Data Mining Tool (DMT) Affymetrix-based software packages. Total chip fluorescence intensities were scaled to 1500 prior to comparison analysis. Data were filtered in DMT for signal log ratio (-1 < SLR > 1) and gene detection call ("present" or "absent"). For the purpose of this manuscript, all transcripts called "marginal" by the Affymetrix algorithm were considered "present." Since the response to LPS within each of the treatment groups was remarkably consistent, gene expression data are presented from pooled liver RNA samples rather than individual animal samples. Criteria for differential gene expression required that at least two out of three LPS-treated dogs responded with like gene changes relative to control dogs, and that individual animal findings were confirmed in pooled samples.

Quantitative polymerase chain reaction (PCR).
Quantitative real-time PCR was performed to confirm a subset of transcriptional changes detected by the canine array. Gene-specific primers were designed using Primer Express 1.5 software (Applied Biosystems, Foster City, CA) for SAA, interleukin-8 (IL-8), tissue inhibitor of metalloproteinase 1 (TIMP1), NADH ubiquinone oxidoreductase chain 5 (ND5), SDF2L1, and 18s ribosomal RNA. Gene-specific primer sequences used in this study are presented in Table 1Go. Briefly, total RNA from pooled samples was DNase treated (DNA-freeTM, Ambion) and reverse transcribed using the SuperScriptTM First-Strand Synthesis System for RT-PCR kit (Invitrogen). PCR products were electrophoresed on a 4% agarose gel to confirm primer specificity and amplicon size prior to quantitative fluorescence-based detection using the Applied Biosystems 7700 Sequence Detector System. PCR reactions containing template, 100 nM of each primer (Qiagen), and SybrGreenTM PCR Master Mix (Applied Biosystems) were incubated at 50°C for 2 min, then denatured for 10 min at 95°C. Following this initial incubation, 50 cycles of 95°C for 15 s and 60°C for 1 min were performed. Each assay was performed in triplicate. Data were analyzed by the comparative CT method of relative quantitation, according to the calculations described in the Applied Biosystems User Bulletin #2. All values are expressed as fold change relative to the vehicle-only control group (Fig. 3Go).


View this table:
[in this window]
[in a new window]
 
TABLE 1 Gene-Specific Primer Sequences Used in Quantitative PCR Analysis
 


View larger version (15K):
[in this window]
[in a new window]
 
FIG. 3. Comparison of quantitative real-time PCR and canine array analyses from pooled liver samples. Primers were designed to canine sequences and PCR reactions were carried out as described in the Materials and Methods section. Data (array = black, PCR = gray) are presented as fold change relative to control values.

 

    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Canine Array Performance
Array performance parameters, such as overall signal intensity and gene "present" or "absent" determinations as calculated by standard Affymetrix algorithms, were summarized from transcript profiling of naive dog liver samples (n = 4). As shown in Table 2Go, probes derived from dog sequences exhibited higher average expression levels than probes derived from human sequences. Since overall signal intensity is a major determinant in assigning gene detection calls, increased intensity values resulted in more canine-derived genes being called "present." Thus, 28% of canine-specific probe sets were determined to be "present" compared to only 8% of human-derived probe sets. Although overall signal intensity was greater with canine-derived probes, this finding was dependent on the probe set being analyzed. Maximal signal intensities obtained for genes derived from dog (455,880) and human (410,923) sources were similar, demonstrating that significant cross-species hybridization is possible with oligonucleotide arrays on an individual gene basis.


View this table:
[in this window]
[in a new window]
 
TABLE 2 Canine Microarray Performance: Detection of Basal Gene Expression
 
Additional experiments were performed to examine whether the design strategy devised for human-derived probes on the canine array improved cross-species hybridization (Fig. 1Go). To determine if the source of lower signal intensity of human-derived probes resulted from poor species cross-reactivity or decreased labeling efficiency, identical dog naive liver samples were sequentially hybridized to canine and human U133A arrays (Table 3Go). Of the 8945 human RefSeq-derived probe sets on the canine array, 965 transcripts were determined to be "present" on at least one array type when hybridized with naive dog liver samples. Of these 965 probe sets, only 10% were determined to be "present" on both array types. On average, 32% of the 965 transcripts were uniquely "present" on the canine array, with the remaining 58% of transcripts determined to be "present" only on the human U133A array. The high percentage of genes detected exclusively by 3' biased probe sets on the U133A array suggests that probe location is an important factor in detection with oligonucleotide arrays.


View this table:
[in this window]
[in a new window]
 
TABLE 3 Assessment of Human-Derived Probe Sets to Detect Canine Gene Expression
 
Differential Gene Expression in Response to LPS
To test the ability of the canine array to detect differential gene expression, male beagle dogs were treated with a sub-lethal dose of bacterial endotoxin, and hepatic gene expression was measured 4 and 24 h following LPS administration. As summarized in Table 4Go, the number of detected transcripts did not vary greatly between treatment Groups 02 (LPS 4 h) and 03 (LPS 24 h). Although the total number of differentially regulated transcripts was similar at both time points, a greater number of genes were downregulated at 4 h (n = 1140) than at 24 h (n = 853).


View this table:
[in this window]
[in a new window]
 
TABLE 4 Hepatic Basal and LPS-Mediated Differential Gene Expression Using the Canine Microarray
 
Clinical observations, serum chemistry, hematology parameters, and histopathology were monitored to gauge the magnitude of the inflammatory response in LPS-treated dogs and served as correlates for gene expression data obtained from the canine array. Due to a remarkably consistent response among animals within each of the treatment groups, pathology data are presented as treatment group average, and gene expression data are from pooled liver RNA samples rather than individual animals. Clinical observations following LPS administration to dogs were consistent with the generation of endotoxemia, with decreased activity and vomiting observed in the majority of LPS-treated dogs and decreased food consumption observed in Group 03. In addition, several LPS-mediated hematological effects were noted in this study (Table 5Go). Total leukocyte count was markedly reduced at 2 and 4 h, but increased at 24 h in endotoxin-treated dogs. Moderate neutropenia occurred 2–4 h following LPS challenge, followed by moderate neutrophilia at 24 h. A slight prolongation of activated partial thromboplastin time and prothrombin time occurred in the majority of LPS-treated dogs. A minimal increase in reticulocyte count, consistent with sudden onset of hypoxia, occurred 2 h after LPS administration and returned to baseline by 24 h. In addition, morphologic changes such as congestion, single cell necrosis, and hepatocellular vacuolation were observed in liver from LPS-treated dogs (data not shown).


View this table:
[in this window]
[in a new window]
 
TABLE 5 Hematology Evaluations following Exposure of Male Beagle Dogs to a Single Intravenous Dose of 0.2 mg/kg LPS
 
As previously described in the literature, administration of LPS caused robust changes in clinical chemistry markers of the acute inflammatory response (Table 6Go). Serum concentration of CRP, a classical marker of inflammation, progressively increased from 4 to 24 h post-LPS exposure. Activity of serum markers of hepatocellular injury, ALT and AST, were increased relative to control values at all time points following LPS challenge. Availability of DNA sequence for markers SAA, ALP, albumin, and AST permitted the comparison of clinical chemistry changes with expression data generated from the canine array. Serum determinations of these four proteins are presented along with hepatic gene expression data in Figure 2Go. Serum SAA concentrations increased within 2 h of endotoxin administration and continued to rise throughout the study period (Fig. 2AGo). Similarly, SAA mRNA expression was markedly upregulated in the liver at both 4 and 24 h compared to vehicle-treated dogs. As determined by the average of multiple SAA probe sets present on the canine array, transcript levels were induced greater than 100-fold at both 4 and 24 h. Additionally, multiple probe sets indicated that ALP mRNA expression was increased at 4 and 24 h, correlating with increased ALP activity throughout the course of study (Fig. 2BGo). Transcript levels of albumin, a negative acute phase reactant, decreased 4-fold in dogs treated with LPS for 24 h. This finding is consistent with a 15% decrease in serum albumin observed at the 24 h time point only (Fig. 2CGo). Serum AST activity was markedly increased above controls 4 h post-dose, whereas hepatic AST transcript levels were decreased 2-fold at 4 and 24 h in LPS-treated dogs (Fig. 2DGo).


View this table:
[in this window]
[in a new window]
 
TABLE 6 Clinical Chemistry Evaluations following Exposure of Male Beagle Dogs to a Single Intravenous Dose of 0.2 mg/kg LPS
 


View larger version (30K):
[in this window]
[in a new window]
 
FIG. 2. Correlation of clinical chemistry markers with hepatic gene expression following exposure of male beagle dogs to a single intravenous dose of 0.2 mg/kg LPS. Hepatic gene expression (bar) was measured at 4 and 24 h following exposure of male beagle dogs to LPS. Clinical chemistry correlates (line) for SAA, ALP, albumin, and AST were measured pre-study, 2, 4, 8, and 24 h post-LPS administration.

 
Several differentially expressed genes were chosen for confirmation by real-time quantitative PCR. Selection of genes was based on observed changes in expression level, signal intensity, or probe design region. Quantitative RT-PCR for SAA, IL-8, TIMP1, ND5, and SDF2L1 are shown with expression data from the canine array in Figures 3AGo–3E. All values were normalized to 18s ribosomal RNA as levels of this transcript did not vary throughout the course of study. RT-PCR results for the selected genes confirmed microarray data with respect to direction and pattern of change. For example, SAA induction was confirmed at both 4 and 24 h following LPS treatment, with a trend toward increased expression at the late time point detected by both technologies (Fig. 3AGo). Similarly, the rapid and robust induction of IL-8 (Fig. 3BGo) and TIMP1 (Fig. 3CGo) was corroborated by PCR analysis. In addition to these marked changes, the slight upregulation of ND5 detected by array and PCR technologies at both time points is shown in Figure 3DGo. PCR reproducibly detected differential gene expression of transcripts with lower signal intensity and magnitude of change, as observed with SDF2L1 (Fig. 3EGo). Consistent with array findings, traditional and quantitative PCR confirmed that primers designed to human SDF2L1 were not as efficient at amplifying the target as canine-specific primers (data not shown).

Biological Correlations and Implications of Measured Transcriptional Responses
As monitored by the canine array, Table 7Go includes expression data from several functional classes of genes that demonstrated significant responses to LPS challenge. Gene expression patterns exhibiting clearly discernable early and late transcriptional responses to bacterial endotoxin were observed in this study. A profound induction in the expression of transcripts encoding cytokines and other inflammatory mediators was observed 4 h following LPS administration. For example, interleukin-6 (IL-6) was highly induced at 4 h, with transcript levels returning to baseline by 24 h. Similarly, the average of multiple probe sets (n = 6) for IL-8 demonstrated transcript levels greater than 20-fold above control values at 4 h. Rapid and marked increases in transcripts involved with extracellular matrix remodeling, such as tissue inhibitor of metalloproteinase I (TIMP1), were observed as early as 4 h following LPS challenge. A time dependent upregulation of additional extracellular matrix proteins, such as elafin and inter-alpha-trypsin inhibitor family heavy chain-related protein (IHRP), was also noted in LPS-treated dogs.


View this table:
[in this window]
[in a new window]
 
TABLE 7 Hepatic Gene Expression 4 and 24 Hours following Exposure of Male Beagle Dogs to a Single Intravenous Dose of 0.2 mg/kg LPS  Human aldehyde dehydrogenase 9 (gamma-aminobutyraldehyde dehydrogenase, E3 isozyme)AFFX-Cf_P450–2E1_M4 h FC24 h FC
 
Decreased plasma levels of glucose were observed at all time points (Table 6Go). Transcripts involved with glucose regulation were downregulated in endotoxin-treated dogs primarily at the 24 h time point (Table 7Go). Examples of genes repressed in response to LPS challenge included UDP-glucose pyrophosphorylase 2, glucose transporter type 5, and phosphoglucomutase. In addition, good agreement in expression level was demonstrated between probe sets designed to different regions of target mRNA transcripts. For example, transcript levels were decreased 2.3, 2.9, and 2.8-fold for probe sets targeting the 5', middle, and 3' untranslated region of glucose-6-phosphatase, respectively (Table 7Go).

In general, expression of Phase I and II conjugating enzymes was significantly repressed at the late time point (24 h). All members of the CYP450 superfamily present on the array were dramatically downregulated at 24 h (Table 7Go). Examples include CYP2D15, 39A1, 2C21, and 2E1, all of which were decreased up to 10-fold at the late time point. Similar transcript changes were detected for Phase II biotransformation enzymes, including several glutathione s-transferase (GST) isozymes and UDP-glucuronosyltransferase. As shown in Table 7Go, GST isozymes pi and mu5-5 were downregulated 2.5- and 3.6-fold at 24 h, respectively.

A delayed increase in serum cholesterol and triglycerides was observed at 8 and 24 h in endotoxin-treated dogs (Table 6Go). Interestingly, profound decreases in gene expression were detected at 24 h for enzymes responsible for cholesterol metabolism via the bile acid pathways. For example, acyl-coA cholesterol acyltransferase (ACAT), lecithin-cholesterol acyltransferase (LCAT), and CYP7A transcripts were decreased as much as 7-fold 24 h post-dosing, with no changes observed at the 4 h time point (Table 7Go). Taken collectively, these changes in gene expression, and clinical chemistry parameters suggest a sparing of cholesterol in response to LPS challenge.


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Canine Array Performance
Preclinical drug candidates are often terminated relatively late in development as a result of toxicity observed in dog studies. Since canine toxicity is deemed particularly relevant to human risk assessment (Olson et al., 1998Go, 2000Go), insight into the underlying mechanisms of damage would help circumvent toxicity issues in future development. Furthermore, identification of biomarkers with improved diagnostic utility would aid in the development of clinical plans to better measure margins of safety, thereby increasing the likelihood of clinical success. We developed a dog-specific tool to address toxicity at the molecular level by designing a high-density oligonucleotide microarray to profile thousands of canine transcripts simultaneously. This is the first reported use of microarray technology to assess biological responses in dog, an approach that represents a significant addition to the set of tools necessary to exploit the promises of toxicogenomics.

Recent technological advances in array design, such as decreased feature size and reduction in the number of oligonucleotide probe pairs required for mRNA detection, allowed an increased number of genes to be represented on the canine array compared to previous commercial arrays. Comparable in size to the human U133A array recently developed by Affymetrix, the canine array contains over 22,000 probe sets. At the time of construction, approximately 13,729 canine sequences were available through public databases and a sequencing partnership with LION bioscience. The remaining space was filled by tiling probes targeting the coding region of fully annotated human reference sequences (RefSeqs), resulting in the creation of a hybrid array designed to detect canine gene expression. The impetus to change human probe location was based on pilot cross-species studies demonstrating fewer than 10% of genes to be "present" when naive dog liver samples were hybridized to commercially available human U95Av2 (data not shown) and U133A arrays (Table 3Go). In an attempt to improve cross-species hybridization efficiency on the canine array, human-derived probes were designed to the terminal 600 bases of the coding region since homology is higher than in the 3' UTR (Lewin, 2000Go). This probe design strategy devised to improve cross-species hybridization is illustrated in Figure 1Go. The drawback of moving the probe region farther away from the 3' UTR is that sample labeling utilizes reverse transcription that begins at the 3' end of mRNA targets, resulting in probe-target interactions biased for the terminal region. This bias for the poorly conserved 3' UTR of the targeted mRNAs, coupled with the sensitivity of oligonucleotide-based hybridization to single base mismatches, would be expected to decrease the number of "present" transcripts when performing cross-species hybridizations. As shown in Table 2Go, probes derived from dog sequences resulted in more intense overall hybridization than human-derived probe sets. Demonstrating that canine-specific probes on the custom array performed well, more than 3000 of the 13,000 canine-specific sequences were determined to be "present" when naive dog liver RNA was hybridized to the array. Despite the lower average signal intensity of human-derived probe sets on the canine array, we gained the ability to monitor an additional 715 liver transcripts (8% of 8945 RefSeqs) that served to increase the overall expression profile and utility of the array (Table 2Go).

Several observations were made regarding human sequence-derived probe sets on the hybrid canine array with regard to cross-species hybridization efficiency. The alternative probe design strategy outlined in Figure 1Go was evaluated by comparing samples hybridized to the canine array (coding region design) and the human U133A array (3' design). All transcripts detectable by at least one array type were grouped, and the performance of this subset of genes was analyzed in greater detail. Of the human RefSeq probe sets common to both array types, 965 were determined to be "present" on at least one array type. When probes from both sources were compared for each of these transcripts, it was found that approximately 10% were determined to be "present" on both array types (Table 3Go). The finding that 58% of the 965 transcripts were called uniquely "present" on the U133A array compared to 32% on the canine array confirms that probe location is an important determinant in gene detection with oligonucleotide arrays. The difference in number of "present" genes likely reflects the positioning of the probes relative to the 3' end of the targeted mRNA, as the probe design algorithm and sort score acceptance criteria were similar for both probe types. The tradeoff between labeling efficiency and probe specificity demonstrates that both are important factors, with the 3' bias being more critical with respect to current labeling practices. As shown in Table 7Go, 20 of the 150 transcripts were detected by human sequence-derived probe sets. These probe sets were particularly useful in monitoring transcriptional changes in glucose metabolism, where key gene changes would have been missed if only dog sequences were used to construct the array. Additionally, the number of detectable transcripts will likely increase with future application of the arrays in new treatment paradigms.

Differential Gene Expression in Response to LPS
The LPS-induced model of systemic inflammation was chosen to study differential gene expression with the canine array since the biological effects of this gram-negative bacterial endotoxin have been well characterized (Hewett and Roth, 1993Go; Panciera et al., 2003Go). Clinical chemistry and hematology parameters presented in Tables 5 and 6GoGo were used to gauge the inflammatory response and toxicity in dogs at various time points following exposure to LPS. As expected, clinical chemistry and hematology markers demonstrated changes indicative of an acute phase response within 2 h of LPS administration. In addition to these changes in inflammatory markers, LPS caused alterations in markers of injury to multiple organ systems including liver, muscle, kidney, and the gastrointestinal tract (Table 6Go). Increased activities of ALT and AST were consistent with hepatotoxicity and coincided with morphological changes such as hepatocellular vacuolation and single cell necrosis in individual animals. Increased serum bilirubin concentration and activities of gamma glutamyltransferase and ALP were consistent with cholestasis.

Hepatic gene expression results obtained from LPS-treated dogs confirm previously described transcriptional responses in rodent models of inflammation (Bulera et al., 2001Go; Liu et al., 2002Go; Olivier et al., 1999Go; Saban et al., 2002Go) and LPS-treated cell lines (Cohen et al., 2000Go; Joyce et al., 2001Go; Zhao et al., 2001Go). In the present study, approximately 1200 genes were differentially expressed in the liver in response to LPS treatment. The biological significance of these transcriptional changes was tested by comparing microarray data with serum levels of acute phase proteins. A strong positive correlation was apparent between hepatic gene expression and measured clinical chemistry data, demonstrating that transcriptional events translated into changes at the protein level that could be monitored non-invasively. The most robust induction from both a gene expression and serum protein standpoint was with SAA, a well-known and sensitive marker of the acute phase response. Figure 2Go illustrates transcriptional changes and corresponding serum changes for acute phase and stress markers SAA, ALP, albumin, and AST. Expression data support the early and sustained induction of SAA and ALP in serum (Figs. 2A and 2BGo). Similarly, decreased transcript levels of the negative acute phase reactant pre-albumin were only detected at 24 h, consistent with a 15% decrease in serum observed 24 h postdosing (Fig. 2CGo). Hepatic AST mRNA expression was decreased 2-fold at both 4 and 24 h following LPS administration, whereas serum activity was increased at all time points (Fig. 2DGo). This finding is indicative of LPS-mediated hepatocellular injury, which is known to result in leakage of AST from the hepatocyte into the bloodstream rather than activation at the transcriptional level.

A subset of genes differentially expressed in response to LPS was chosen for confirmation by real-time quantitative PCR. Temporal PCR data verified expression data obtained from the canine array for each of the selected genes. However, absolute fold change levels differed between the two analyses, a common phenomenon largely due to differences in dynamic range and threshold detection between array and PCR technologies. Higher sensitivity of fluorescence-based PCR technology allows more accurate measurement of basal gene expression, thereby circumventing this issue.

Biological Correlations and Implications of Measured Transcriptional Responses
Robust LPS-induced transcriptional changes were observed with extracellular matrix proteins involved in proteinase inhibition (Table 7Go). Multiple probe sets representing TIMP1 (n = 3) demonstrated a marked upregulation at both 4 and 24 h post-LPS treatment, with expression levels varying less than 10% between alternate probe sets derived from public and proprietary sequencing efforts. Similarly, matrix metalloproteinase 9 (MMP9, n = 2) was upregulated, but only at the 24 h time point. TIMPs form irreversible complexes with activated metalloproteinases, and the balance between these proteins controls extracellular matrix remodeling. It is not currently known whether TIMP1 is synthesized in response to metalloproteinase production or whether these enzymes are independently regulated (Zucker et al., 1999Go). The induction of TIMP1 prior to MMP9 in our study suggests that it is independently controlled, and as a result, TIMP1 may serve as an early indicator of the acute phase response. Elafin, a recently described serine proteinase inhibitor of the troppin gene family, was markedly increased in a time-dependent manner in this study. Similar to TIMP1, elafin expression is believed to control the balance of extracellular matrix proteolysis via inhibition of elastase. Inflammation related activity of elafin has been studied primarily in the lung (Bingle et al., 2001Go; Reid et al., 1999Go), with induction of this gene proposed as a potential therapeutic strategy in pulmonary inflammation (Simpson et al., 2001Go). Another protease inhibitor, inter-alpha-trypsin inhibitor family heavy chain protein (IHRP), was also markedly increased in response to LPS treatment. Proposed as an acute phase protein (Choi-Miura, 2001Go), IHRP was identified as one of five inflammation-related proteins upregulated in a differential display-based screening model of dog vasospasm (Onda et al., 1999Go). Although IHRP has been shown to respond at the transcriptional level in other inflammatory models, this is the first report that expression of this gene is inducible by LPS. Collectively, these data indicate that the balance of extracellular protease activity in LPS-treated dogs is shifted dramatically toward proteinase inhibition. This shift may protect against destruction of the extracellular matrix, thereby preserving tissue architectural integrity. An alternate hypothesis is that LPS-mediated fibrotic damage may be attributed to protease inhibition (Zucker et al., 1999Go). In the later scenario, any one or a battery of protease inhibitors could serve as markers of matrix damage that occurs during sepsis. TIMP1, elafin, and IHRP are excellent candidates for noninvasive markers of injury associated with the acute phase response since they are found in blood, and mRNA expression is strongly induced and sustained over time.

It has previously been reported that the human response to endotoxemia consists of a symptomatic phase that peaks at 4 h, followed by an asymptomatic haemostatic response at 24 h (Dhainaut et al., 2001Go). In this study, administration of a single dose of LPS produced a very similar response profile in dog, characterized by early inflammation and secondary changes in metabolic events. As supported by extensive literature, the early response to LPS was characterized by increased expression of inflammatory mediators (Table 7Go). The canine microarray detected a marked induction of pro-inflammatory cytokines, including IL-6, IL-8, and TNF-{alpha}. This presumably results from LPS stimulation of immune cells, such as macrophages, which is known to activate intracellular signaling pathways and transcription factors involved with the induction of genes encoding inflammatory molecules.

A notable exception to the early LPS-mediated response was the delayed increase in serum cholesterol and triglycerides, demonstrating altered lipid metabolism secondary to liver injury (Table 6Go). This observation is supported by transcriptional responses measured in the liver (Table 7Go). The effect appears to be one of cholesterol sparing in the dog, as enzymes responsible for cholesterol catabolism are decreased following LPS challenge. Serum cholesterol concentrations parallel the transcriptional repression of ACAT, which is markedly downregulated at 24 h, but not 4 h, following LPS administration. This delayed response suggests that the observed cholesterol sparing effect is a secondary response controlled by early mediators of inflammation such as IL-6 and IL-8. These findings support the inhibition of cholesterol metabolism previously described in rodents (Memon et al., 2001Go) and demonstrate that altered cholesterol metabolism also occurs in nonrodent species treated with LPS.

Decreased hepatic expression and activity of various CYP isoenzymes has been described in rodents during an acute inflammatory response (Morgan, 1993Go; Roe et al., 2001Go; Sewer et al., 1997Go). Consistent with these rodent studies, expression of biotransformation enzymes was profoundly decreased in LPS-treated dogs (Table 7Go). Transcript levels of CYP2D15, 2E1, and 2B were markedly reduced 24 h following LPS administration. To our knowledge, this is the first report that transcription of CYPs is downregulated in dog during an acute inflammatory response. A similar repression of Phase II detoxification enzymes, such as GSTs and UDP glucuronosyltransferase, was observed in endotoxin-treated dogs at 24 h. These changes in drug-metabolizing enzymes may have important clinical implications, since several drugs used to treat sepsis undergo metabolism via these pathways.

Identification of Novel Biomarkers of the Acute Phase Response
An interesting and novel finding was the induction of the Sec61a transcript, a protein translocase involved with protein secretion (Gabriel et al., 2001Go; Gorlich et al., 1992Go; Hartmann et al., 1994Go). Sec61a has not previously been shown to respond to inflammatory stimuli, yet it was induced greater than 2-fold at both 4 and 24 h in this study. One possible explanation is that Sec61a is involved with the secretion of acute phase proteins synthesized in hepatocytes and transported into the blood. Protein translocation machinery may be sensitive to cholesterol concentrations in the endoplasmic reticulum membrane (Nilsson et al., 2001Go). This could account for the transcriptional effects on Sec61a, given the cholesterol changes seen in the present study. Another transcript demonstrating robust and sustained induction is the recently cloned SDF2L1. Both dog and human probe types demonstrated similar levels of transcript induction in response to LPS treatment. SDF2L1 is a secretory protein most closely related to yeast dolichyl phosphate-D-mannose:protein mannosyltransferases (Fukuda et al., 2001Go). These enzymes function as O-glycosylation enzymes in yeast, potentially connecting the link between the acute phase response and glycosylation state of circulating proteins.

In summary, the newly designed canine array identified a large number of expected and novel transcriptional changes in the liver during acute inflammation. The ability of the array to accurately detect changes in gene expression was confirmed by comparison with several parameters including clinical chemistry, hematology, histopathology, and quantitative RT-PCR analysis. Transcriptional data from this study demonstrate that the canine array is a powerful and sensitive tool that can be employed to gain insight into underlying inflammatory mechanisms. In addition, this new genomic array identified novel markers of toxicity and several genes not previously reported to be regulated during an acute inflammatory response.


    ACKNOWLEDGMENTS
 
Clinical chemistry and hematology evaluations were performed by Kerry Rodocker, Michele Langsford, Bruce Beechler, Krista Pratt, and Lee Ann Sturgill. Statistical analyses for the human and dog probe set comparisons were performed by Hui-Rong Qian and Han Weng. The authors thank Gregg Lundeen for supporting this research endeavor. Special thanks to Thomas Jones, Craig Thomas, James Stevens, and Myrtle Davis for critical review of the manuscript.


    NOTES
 
1 To whom correspondence should be addressed. Fax: (317) 277-6770. E-mail: timryan{at}lilly.com. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Baker, T. K., Higgins, M. A., Carfagna, M. A., and Ryan, T. P. (2003). Characterization of hepatocytes and their use as a model system in toxicogenomics. In An Introduction to Toxicogenomics (M. E. Burczynski, Ed.), pp. 117–143. CRC Press LLC, Boca Raton, FL.

Bertucci, F., Houlgatte, R., Benziane A., Granjeaud, S., Adelaide, J., Tagett, R., Loriod, B., Jacuemier, J., Viens, P., Jordan, B., et al. (2000). Gene expression profiling of primary breast carcinomas using arrays of candidate genes. Hum. Mol. Genet. 9, 2981–2991.[Abstract/Free Full Text]

Bingle, L., Tetley, T. D., and Bingle, C. D. (2001). Cytokine-mediated induction of the human elafin gene in pulmonary epithelial cells I regulated by nuclear factor-kappaB. Am. J. Resp. Cell. Mol. Biol. 25, 84–91.[Abstract/Free Full Text]

Bulera, S. J., Eddy, S. M., Ferguson, E., Jatkoe, T. A., Reindel, J. F., Bleavins, M. R., and De La Iglesia, F. A. (2001). RNA expression in the early characterization of hepatotoxicants in Wistar rats by high-density DNA microarrays. Hepatology 33, 1239–1258.[CrossRef][ISI][Medline]

Choi-Miura, N. H. (2001). Quantitative measurement of the novel human plasma protein, IHRP, by sandwich ELISA. Biol. Pharm. Bull. 24, 214–217.[CrossRef][ISI][Medline]

Cohen, P., Bouaboula, M., Bellis, M., Baron, V., Jbilo, O., Poinot-Chazel, C., Galiegue, S., Hadibi, E., and Casellas, P. (2000). Monitoring cellular responses to Listeria monocytogenes with oligonucleotide arrays. J. Biol. Chem. 275, 11181–11190.[Abstract/Free Full Text]

Cullen, J. J., Spates, S. T., Ephgrave, K. S., and Hinkhouse, M. M. (1998). Endotoxin temporarily impairs canine colonic absorption of water and sodium. J. Surg. Res. 74, 34–38.[CrossRef][ISI][Medline]

Devilard, E., Bertucci, F., Trempat, P., Bouabdallah, R., Loriod, B., Giaconia, A., Brousset, P., Granjeaud, S., Nguyen, C., Birnbaum, D., et al. (2002). Gene expression profiling defines molecular subtypes of classical Hodgkin’s disease. Oncogene 21, 3095–3102.[CrossRef][ISI][Medline]

Dhainaut, J. F., Marin, N., Mignon, A., and Vinsonneau, C. (2001). Hepatic response to sepsis: Interaction between coagulation and inflammatory processes. Crit. Care. Med. 29, S42–S47.[ISI][Medline]

Fukuda, S., Sumii, M., Masuda, Y., Takahashi, M., Koike, N., Teishima, J., Yasumoto, H., Itamoto, T., Asahara, T., Dohi, K., and Kamiya, K. (2001). Murine and human SDF2L1 is an endoplasmic reticulum stress-inducible gene and encodes a new member of the Pmt/rt protein family. Biochem. Biophys. Res. Commun. 280, 407–414.[CrossRef][ISI][Medline]

Gabriel, K., Buchanan, S. K., and Lithgow, T. (2001). The alpha and the beta: Protein translocation across mitochondrial and plastid outer membranes. Trends Biochem. Sci. 26, 36–40.[CrossRef][ISI][Medline]

Gorlich, D., Prehn, S., Hartmann, E., Kalies, K. U., and Rapoport, T. A. (1992). A mammalian homolog of SEC61p and SECYp is associated with ribosomes and nascent polypeptides during translocation. Cell 71, 489–503.[ISI][Medline]

Hartmann, E., Sommer, T., Prehn, S., Gorlich, D., Jentsch, S., and Rapoport, T. A. (1994). Evolutionary conservation of components of the protein translocation complex. Nature 367, 654–657.[CrossRef][ISI][Medline]

Hewett, J. A., and Roth, R. A. (1993). Hepatic and extrahepatic pathobiology of bacterial lipopolysaccharides. Pharmacol. Rev. 45, 381–411.[ISI]

Jacobs, J. W., and Fodor, S. P. (1994). Combinatorial chemistry—applications of light-directed chemical synthesis. Trends Biotechnol. 12, 19–26.[CrossRef][ISI][Medline]

Joyce, D. E., Gelbert, L., Ciaccia, A., DeHoff, B., and Grinnell, B. W. (2001). Gene expression profile of antithrombotic protein C defines new mechanisms modulating inflammation and apoptosis. J. Biol. Chem. 276, 11199–11203.[Abstract/Free Full Text]

Lewin, B. (2000). Genes VII. Oxford University Press, New York.

Liu, J., Saavedra, J. E., Lu, T., Song, J. G., Clark, J., Waalkes, M. P., and Keefer, L. K. (2002). O(2)-Vinyl 1-(pyrroliding-1-yl)diazen-1-ium-1,2-diolate protection against D-galactosamine/endotoxin-induced hepatotoxicity in mice: Genomic analysis using microarrays. J. Pharmacol. Exp. Ther. 300, 18–25.[Abstract/Free Full Text]

Louden, C., and Morgan, D. G. (2001). Pathology and pathophysiology of drug-induced arterial injury in laboratory animals and its implications on the evaluation of novel chemical entities for human clinical trials. Pharmacol. Toxicol. 89, 158–170.[CrossRef][ISI][Medline]

Mazzola, L. T., and Fodor, S. P. (1995). Imaging biomolecule arrays by atomic force microscopy. Biophys. J. 68, 1653–1660.[Abstract]

Memon, R. A., Moser, A. H., Shigenaga, J. K., Grunfeld, C., and Feingold, K. R. (2001). In vivo and in vitro regulation of sterol 27-hydroxylase in the liver during the acute phase response. J. Biol. Chem. 276, 30118–30126.[Abstract/Free Full Text]

Mitsiades, N., Mitsiades, C. S., Poulaki, V., Chauhan, D., Fanourakis, G., Gu, X., Bailey, C., Joseph, M., Libermann, T. A., Treon, S. P., et al. (2002). Molecular sequelae of proteasome inhibition in human multiple myeloma cells. Proc. Natl. Acad. Sci. U.S.A. 99, 14374–14379.[Abstract/Free Full Text]

Morgan, E. T. (1993). Down-regulation of multiple cytochrome P450 gene products by inflammatory mediators in vivo. Independence from the hypothalamopituitary axis. Biochem. Pharmacol. 45, 415–419.[CrossRef][ISI][Medline]

Niedergethmann, M., Hildenbrand, R., Wostbrock, B., Hartel, M., Sturm, J. W., Richter, A., and Post, S. (2002). High expression of vascular endothelial growth factor predicts early recurrence and poor prognosis after curative resection for ductal adenocarcinoma of the pancreas. Pancreas 25, 122–129.[ISI][Medline]

Nilsson, I., Ohvo-Rekila, H., Slotte, J. P., Johnson, A. E., and von Heijne, G. (2001). Inhibition of protein translocation across the endoplasmic reticulum membrane by sterols. J. Biol. Chem. 276, 41748–41754.[Abstract/Free Full Text]

Olivier, E., Soury, E., Risler, J., Smih, F., Schneider, K., Lochner, K., Jouzeau, J., Fey, G. H., and Salier, J. (1999). A novel set of hepatic mRNAs preferentially expressed during an acute inflammation in rat represents mostly intracellular proteins. Genomics 57, 352–364.[CrossRef][ISI][Medline]

Olson, H., Betton, G., Robinson, D., Thomas, K., Monro, A., Kolaja, G., Lilly, P., Sanders, J., Sipes, G., Bracken, W., et al. (2000). Concordance of the toxicity of pharmaceuticals in humans and in animals. Regul. Toxicol. Pharmacol. 32, 56–67.[CrossRef][ISI][Medline]

Olson, H., Betton, G., Stritar, J., and Robinson, D. (1998). The predicitivity of the toxicity of pharmaceuticals in humans from animal data—an interim assessment. Toxicol. Lett. 102103, 535–538.[CrossRef]

Onda, H., Kasuya, H., Takakura K., Hori, T., Imaizumi, T., Takeuchi, T., Inoue, I., and Takeda, J. (1999). Identification of genes differentially expressed in canine vasospastic cerebral arteries after subarachnoid hemorrhage. J. Cereb. Blood Flow Metab. 19, 1279–1288.[CrossRef][ISI][Medline]

Panciera, D. L., Ritchey, J. W., and Ward, D. L. (2003). Endotoxin-induced nonthyroidal illness in dogs. AJVR 64, 229–234.

Pennie, W. D., Tugwood, J. D., Oliver, G. J. A., and Kimber, I. (2000). The principles and practice of toxicogenomics: Applications and opportunities. Toxicol. Sci. 54, 277–283.[Free Full Text]

Reid, P. T., Marsden, M. E., Cunningham, G. A., Haslett, C., and Sallenave, J. M. (1999). Human neutrophil elastase regulates the expression and secretion of elafin (elastase-specific inhibitor) in type II alveolar epithelial cells. FEBS Lett. 457, 33–37.[CrossRef][ISI][Medline]

Roe, A. L., Poloyac, M., Howard, G., Shedlofsky, S. I., and Blouin, R. A. (2001). The effect of endotoxin on hepatocyte nuclear factor 1 nuclear protein binding: Potential implications on CYP2E1 expression in the rat. J. Pharm. Pharmacol. 53, 1365–1371.[CrossRef][ISI][Medline]

Saban, M. R., Nguyen, N., Hammond, T. G., and Saban, R. (2002). Gene expression profiling of mouse bladder inflammatory responses to LPS, substance P, and antigen-stimulation. Am. J. Physiol. 160, 2095–2110.

Sewer, M. B., Koop, D. R., and Morgan, E. T. (1997). Differential inductive and suppressive effects of endotoxin and particulate irritants on hepatic and renal cytochrome P-450 expression. J. Pharmacol. Exp. Ther. 280, 1445–1454.[Abstract/Free Full Text]

Simpson, A. J., Cunningham, G. A., Porteous, D. J., Haslett, C., and Sallenave, J. M. (2001). Regulation of adenovirus-mediated elafin transgene expression by bacterial lipopolysaccharide. Hum. Gene Ther. 12, 1395–1406.[CrossRef][ISI][Medline]

Tygstrup, N., Bangert, K., Ott, P., and Bisgaard, H. C. (2002). Messenger RNA profiles in liver injury and stress: A comparison of lethal and nonlethal rat models. Biochem. Biophys. Res. Commun. 290, 518–525.[CrossRef][ISI][Medline]

Ulrich, R., and Friend, S. H. (2002). Toxicogenomics and drug discovery: Will new technologies help us produce better drugs? Nature Reviews 1, 84–88.[ISI][Medline]

Zhao, B., Bowden, R. A., Stavchansky, S. A., and Bowman, P. D. (2001). Human endothelial cell response to gram-negative lipopolysaccharide assessed with cDNA microarrays. Am. J. Physiol. Cell Physiol. 281, 1587–1595.

Zucker, S., Hymowitz, M., Conner, C., Zarrabi, H. M., Hurewitz, A. N., Matrisian, L., Boyd, D., Nicolson, G., and Montana, S. (1999). Measurement of matrix metalloproteinases and tissue inhibitors of metalloproteinases in blood and tissues. Clinical and experimental applications. Ann. N.Y. Acad. Sci. 30, 212–227.