Mouse Strain–Specific Differences in Vascular Wall Gene Expression and Their Relationship to Vascular Disease
http://www.100md.com
《动脉硬化血栓血管生物学》
From the Donald W. Reynolds Cardiovascular Clinical Research Center, Division of Cardiovascular Medicine (R. Tabibiazar, R.A.W., J.M.S., E.A.A., P.S.T., T.Q.), Department of Health Research and Policy (B.N., B.E., R. Tibshirani), and the Department of Statistics (B.E., R. Tibshirani), Stanford University School of Medicine, Stanford, Calif; and the Genome Sciences Department (E.M.R.), Lawrence Berkeley National Laboratory, Berkeley, Calif.
Correspondence to Raymond Tabibiazar, Stanford Medical School, Division of Cardiovascular Medicine, 300 Pasteur Drive, Falk CVRC, Stanford, CA 94305 (E-mail rtabibiazar@cvmed.stanford.edu) or Thomas Quertermous, Stanford Medical School, Division of Cardiovascular Medicine, 300 Pasteur Drive, Falk CVRC, Stanford, CA 94305 (E-mail tomq1@stanford.edu).
Abstract
Objective— Different strains of inbred mice exhibit different susceptibility to the development of atherosclerosis. The C3H/HeJ and C57Bl/6 mice have been used in several studies aimed at understanding the genetic basis of atherosclerosis. Under controlled environmental conditions, variations in susceptibility to atherosclerosis reflect differences in genetic makeup, and these differences must be reflected in gene expression patterns that are temporally related to the development of disease. In this study, we sought to identify the genetic pathways that are differentially activated in the aortas of these mice.
Methods and Results— We performed genome-wide transcriptional profiling of aortas from C3H/HeJ and C57Bl/6 mice. Differences in gene expression were identified at baseline as well as during normal aging and longitudinal exposure to high-fat diet. The significance of these genes to the development of atherosclerosis was evaluated by observing their temporal pattern of expression in the well-studied apolipoprotein E model of atherosclerosis.
Conclusion— Gene expression differences between the 2 strains suggest that aortas of C57Bl/6 mice have a higher genetic propensity to develop inflammation in response to appropriate atherogenic stimuli. This study expands the repertoire of factors in known disease-related signaling pathways and identifies novel candidate genes for future study.
To gain insights into the molecular pathways that are differentially activated in strains of mice with varied susceptibility to atherosclerosis, we performed comprehensive transcriptional profiling of their vascular wall. Genes identified through these studies expand the repertoire of factors in disease-related signaling pathways and identify novel candidate genes in atherosclerosis.
Key Words: atherosclerosis ? vascular disease ? gene expression ? microarray ? inflammation ? oxidative stress
Introduction
Mouse strains with different genetic backgrounds have helped to advance the study of atherosclerosis and are an attractive model for investigating the intricate interactions between environmental and genetic factors that contribute to the development of this complex disease.1,2 Although there is great variation in disease susceptibility in a variety of mouse strains, the atherosclerosis-resistant C3H/HeJ (C3H) and the atherosclerosis-susceptible C57Bl/6 (C57) strains have emerged as popular models for cardiovascular research. Manipulation of expression of individual genes through transgenesis and gene targeting, such as the apolipoprotein E (apoE) gene knockout in C57Bl/6 mice, has added significantly to the repertoire of genetic models available for the study of atherosclerosis.3,4 Despite the availability of strains with differing susceptibilities to atherosclerosis, there remains little information regarding the underlying disease processes in the vessel wall that mediate atherosclerosis.
With the availability of genomic tools that allow genome-wide transcriptional profiling, individual environmental risk factors can be evaluated in a controlled setting to define gene expression differences that are associated with the risk, without a priori knowledge of specific candidate genes. Recent studies have shown that transcriptional profiling can be used to characterize gene expression patterns associated with inherited genetic differences.5 The integration of gene expression data with genetic analyses has been used to identify susceptibility loci for complex diseases, such as diabetes, obesity, and allergic asthma, and to prioritize candidate genes.6,7 Therefore, gene expression analysis of genetically distinct mice with different disease susceptibilities under controlled environmental conditions can potentially identify genes whose expression underlies differences in disease susceptibility.
Prior studies investigating genetic influences on atherosclerosis susceptibility in various mouse strains have provided evidence for involvement of the vascular wall and not peripheral blood cells or systemic cholesterol metabolism in the development of atherosclerosis.8 We thus hypothesized that genes differentially expressed in the vascular wall in genetically resistant C3H versus genetically susceptible C57 mice may play critical roles in the initiation and modulation of atherosclerosis. To identify these genes, we have performed genome-wide transcriptional profiling of mouse aortic tissue from these 2 mouse strains and compared their vascular wall gene expression. Using a comprehensive study design and novel analytical tools, we show that activity of specific biological processes and molecular functions distinguish disease-resistant versus disease-prone mouse strains. Further, the importance of these genes in development of atherosclerosis is validated with studies in the apoE model.
Methods
RNA Preparation and Hybridization to the Microarray
Three-week-old female C3H/HeJ, C57Bl/6J, and apoE knockout mice (C57BL/6J-Apoetm1Unc) were purchased from Jackson Laboratories (Bar Harbor, Me). At 4 weeks of age, the mice were either continued on normal chow or switched to non–cholate-containing high-fat diet, which included 21% anhydrous milkfat and 0.15% cholesterol (Dyets #101511, Dyets Inc) for a maximum period of 40 weeks. At each of the time points, including 0 (baseline), 4, 10, 24, and 40 weeks, for each of the conditions (strain-diet combination), 15 mice were harvested for RNA isolation, for a total of 450 mice. Following Stanford University animal care guidelines,9 the mice were anesthetized with Avertin and perfused with normal saline.10 The aortas from the root to the common iliacs were carefully dissected, flash frozen in liquid nitrogen (LN2), and divided into 3 pools of 5 aortas for further RNA isolation. Total RNA was isolated as described previously.10 First strand cDNA was synthesized from 10 μg of total RNA from each pool and from whole e17.5-day embryo for reference RNA in the presence of Cy5 or Cy3 dCTP, respectively, and hybridized to a mouse 60-mer oligo microarray (G4120A, Agilent Technologies), generating 3 biological replicates for each time point. For details, please see online Methods,11 available at http://atvb.ahajournals.org.
Data Processing
Array image acquisition and feature extraction was performed using the Agilent G2565AA Microarray Scanner and feature extraction software version A.6.1.1. Normalization was performed using a LOWESS algorithm, and Dye-normalized signals were used in calculating log ratios. Features with reference values of <2.5 SDs above background for the negative control features were regarded as missing values. Those features with values in at least two thirds of the experiments and present in at least 1 of the replicates were retained for further analysis. For significance analysis of microarrays (SAM), a K-nearest-neighbor algorithm was applied to impute for missing values.10
Data Analysis
Experimental design and analysis flow chart is depicted in Figure 1. SAM was used to identify genes with statistically different expression between the C3H and C57 mice at baseline.10,12–14 For partitioning clustering of the genes with K-means and self-organizing maps (SOMs), we used positive correlation for distance determination and required complete linkage, which uses the greatest distance between genes to ascribe similarity. SOMs and K-means analyses were performed using Expressionist software (GeneData Inc).15 Heatmaps were generated using HeatMap Builder.14,16 For enrichment analysis we used the EASE analysis software, which uses gene ontology (GO) annotation and the Fisher exact test to derive biological themes within particular gene sets.17 For time-course study a new statistical algorithm, the area under curve (AUC) analysis, was devised. For each sequence of 4 triplicate gene expression measurements over time, we first subtracted the measurement at time 0 from all values. We then computed the signed AUC. The area is a natural measure of change over time. These areas were then used to compute an F statistic for comparing C57 and C3H mice across the different diets. A permutation analysis, similar to that used in SAM, was performed to estimate the false discovery rate (Q value or "FDR") for different levels of the F statistic. For ease of presentation, genes which meet our FDR cutoffs will be referred to as "significant" throughout the remainder of this article. All microarray data were submitted to the National Center for Biotechnology Information Gene Expression Omnibus (GEO GSE1560; http://www.ncbi.nlm.nih.gov/geo/).
Figure 1. Experimental design and analysis flowcharts. A, Four-week-old female C3H/HeJ (C3H) and C57Bl/6 (C57) mice were fed normal chow vs high-fat diet for the maximum period of 40 weeks. Triplicate microarray experiments were performed for each time point using 3 pools of 5 aortas at 0, 4, 10, 24, and 40 weeks on either diet (total of 15 mice per time point). B, Data analysis overview. Of the 20 283 genes present on the array, 311 genes were found to be significantly differentially expressed between C3H and C57 mice at baseline (SAM FDR 10% and >1.5-fold change). Differential gene expression during aging was determined by comparing C57 vs C3H time-course differences on normal and atherogenic high-fat diets using the AUC analysis.
Aortic Lesion Analysis
For select time points within various experimental groups, 5 to 7 female mice were used for histological lesion analysis. Atherosclerosis lesion area was determined as described previously.18
Quantitative Real-Time Reverse Transcriptase–Polymerase Chain Reaction
Primers and probes for 10 representative differentially expressed genes were obtained from Applied Biosystems Assays-on-Demand. A Total of 90 reactions were performed from representative RNA samples used for microarray experiments. These included triplicate assay on 3 pools of 5 aortas. cDNA was synthesized and Taqman was performed as described previously.10
Results
Baseline Differences in Gene Expression Patterns Between the Mouse Strains
Differences in gene expression levels between the 2 strains at baseline, before effects of aging or diet become apparent, may identify genes that play a role in determining vascular wall disease susceptibility. To identify such genes we used SAM to compare the vascular wall gene expression of C3H versus C57 mice at 4 weeks of age, with all animals on normal chow diet. SAM identified 311 genes as being significantly differentially expressed (FDR<0.1 with >1.5-fold difference), and expression patterns of these genes provided a clear partition between C3H and C57 mice (Figure 2). A separate 2-class comparison (SAM, FDR<0.1) between C57 and apoE-deficient mice with a C57Bl/6 genetic background revealed only a few genes, including apoE, which were differentially expressed in the 2 groups of mice (data not shown).
Figure 2. Differential gene expression between C3H and C57 mice at baseline. The SAM analysis shown was associated with an FDR of 10%, and a total of 311 probes were identified as differentially regulated at this level of confidence. Lists represent a select group of genes (expressed sequence tags excluded) with higher expression in C3H (top 20 ranking genes) and C57 (top 45 ranking genes). The heatmap reflects normalized gene expression ratios and is organized with individual hybridizations for each of the 3 replicates for each mouse strain arranged along the x axis, with normalized expression ratios depicted by color intensity such that highest expression corresponds to bright red, and the lowest expression corresponds to bright green. See supplemental material for full list of genes (available online at http://atvb.ahajournals.org).
Comparison of C3H and C57 vascular wall gene expression at baseline provided a list of compelling candidate genes, which reflected differences in biological processes such as growth, differentiation, and inflammation, as well as molecular functions, such as cathecholamine synthesis, phosphatase activity, peroxisome function, insulin-like growth factor activity, and antigen presentation (Figure 2). These processes were exemplified by higher expression of genes such as Cdkn1a, Pparbp, protein tyrosine phosphatase-4a2, and Socs5 in C3H mice, compared with genes such as ABCC1, H2-D1, Bat5, IGFBP1, SCD1, and Serpine6b, which demonstrated higher expression in C57 mice (Tables I and II, available online at http://atvb.ahajournals.org). These fundamental baseline gene expression differences may determine disease susceptibility as the mice are exposed to age-related stimuli or dietary challenges.
Age-Related Differences in Gene Expression Patterns Between the Mouse Strains
To further examine the vascular wall gene expression differences between C57 and C3H mice, we performed analyses to identify genes differentially expressed in response to aging (Figure 3). Data were collected at 5 time points over a 40-week period. To identify such genes, we developed the AUC analysis. The AUC analysis relies on a permutation procedure to reduce the number of potential false-positives generated because of multiple testing but still uses the increase in statistical power of time-course experimental design. Comparing C57 versus C3H time-course differences on normal diet with a rigid cutoff (FDR<0.05) did not identify any genes. However, relaxing the AUC stringency (F statistic>10, FDR<0.45) allowed a large number of genes (413) to be included for pathway overrepresentation analysis using GO annotation (Table III, available online at http://atvb.ahajournals.org). Functional annotation and group overrepresentation analysis (Fisher test probability value <0.02) of the resultant differentially expressed genes revealed differences in a number of biological processes, including growth and development as well as a number of molecular functions, such as cell cycle control, regulation of mitosis, and metabolism (Figure 3B). Some of these processes are exemplified by genes with higher expression in C57 mice, such as Aoc1 (prooxidative stress), Bub1 (cell cycle check point), Cyclin B2, and genes with higher expression in C3H, including INHBA and INHBB (Table IV, available online at http://atvb.ahajournals.org).
Figure 3. Differential gene expression between C3H and C57 mice in response to normal aging. A, Response to aging was determined by comparing C57 vs C3H time-course differences on normal diet (AUC analysis F statistic>10). B, Functional annotation of the 413 differentially expressed genes reveal differences in various biological processes, including growth and differentiation. The probability rates provided are based on Fisher exact test (P<0.02).17 C, K-means clustering of the 413 genes reveals several profiles of gene expression. Clusters 1, 4, and 9 reveal increased gene expression in C3H vs C57 mice, whereas clusters 2, 6, and 14 reveal the opposite pattern. See Table VIII for lists of genes in the clusters (available online at http://atvb.ahajournals.org).
Temporally variable genes identified by AUC analysis were further characterized with K-means clustering to identify dynamic patterns of expression during the aging process (Figure 3C). Clusters 1, 4, and 9 revealed either higher overall expression or temporally increasing levels of expression in C3H mice compared with C57 mice. In contrast, clusters 2, 6, and 14 revealed the opposite pattern. Of the genes which were noted to be differentially expressed in the 2 strains during aging, 51 genes were also differentially expressed at baseline, suggesting that baseline differences of certain genes can further be affected with aging.
Diet-Related Differences in Gene Expression Patterns Between the Mouse Strains
Differential vascular wall response to atherogenic stimuli was determined by comparing temporal gene expression patterns in C57 versus C3H mice on high-fat diet (Figure 4A). Comparing C57 versus C3H time-course differences on high-fat diet with a rigid cutoff (FDR<0.05) identified 35 genes, including Hgfl and Tgfb4, which were downregulated in C57 on high-fat diet. Additional known genes as well as a number of expressed sequence tags were also identified (Table V, available online at http://atvb.ahajournals.org). Using a less stringent AUC cutoff allowed identification of a larger number of genes that could be evaluated with pathway overrepresentation analysis using GO annotation. At this level of stringency (F statistic>10, FDR<0.35), a total of 650 genes with temporally variable expression were identified (Table VI, available online at http://atvb.ahajournals.org). Genes that were also differentially regulated by the aging process (141 of 650 genes) were excluded from further analysis of this group. Thirty-eight of the remaining 509 genes were among those differentially expressed at baseline. Functional annotation and group overrepresentation analysis (Fisher test P<0.02) of these differentially expressed genes revealed differences in biological processes, such as catabolism, oxygen reactive species and superoxide metabolism, and proteo- and peptidolysis as well as molecular functions, such as fatty acid metabolism, oxidoreductase, and methyltransferase activities (Figure 4B). Interestingly, this analysis suggested important differences between the 2 mouse strains with respect to the activity of the peroxisome, microbody, and lysosome. Some of these processes were exemplified by genes with higher expression in C3H mice, such as Ccs, Ephx2, Gpx4, Prdx6 (antioxidants), Sirt3 (transcriptional repressor), PPAR, and Mcd as well as genes with higher expression in C57 mice, such as Lysyl oxidase and Cdkn1a (Table VII, available online at http://atvb.ahajournals.org). K-means clustering of these genes identified a small number of distinct expression patterns (Figure 4C), with clusters 3 and 9 revealing increased gene expression in C3H mice and clusters 8 and 10 showing the opposite pattern.
Figure 4. Differential gene expression between C3H and C57 mice in response to high-fat diet. A, Response to atherogenic stimulus was determined by comparing C57 vs C3H time-course differences on high-fat diet (AUC analysis F statistic>10). B, Functional annotation of the 509 differentially expressed genes reveals differences in various biological processes and cellular components. The probability rates provided are based on Fisher exact test (P<0.02).17 C, K-means clustering of the 509 differentially expressed genes revealed several patterns of gene expression with clusters 3 and 9 exhibiting increased gene expression in C3H vs C57 mice and clusters 8 and 10 with the opposite pattern. See Table IX for lists of genes in the clusters (available online at http://atvb.ahajournals.org).
Evaluation of Strain-Specific Differentially Regulated Genes in the ApoE Model
Using these techniques, we have identified a significant number of genes that are differentially expressed in the atherosclerosis-resistant C3H and susceptible C57 mice, some of which are likely involved in atherogenesis and some of which are likely irrelevant to the process. To further select genes most likely to be involved in atherogenesis, we examined their expression in apoE-deficient mice fed normal or high-fat diet over a period of 40 weeks (Figure 5). We used SOM analysis to visualize the expression profiles of these subsets of genes throughout the development and progression of atherosclerosis in the apoE-deficient mice. The analysis revealed several patterns of gene expression. For example, SOM cluster 8 demonstrated a consistently increasing pattern of expression which correlated with disease progression in the apoE-deficient mice (Figure 5). As evidenced by the pie chart, this cluster is enriched with genes that were identified as more highly expressed in C57 versus C3H mice at baseline (ie, potentially atherogenic). In contrast, clusters 4, 5, and 6 showed decreasing expression with disease progression. The decreased expression of genes in cluster 4 was somewhat attenuated with high-fat challenge of the apoE-deficient mice. This cluster was particularly enriched with genes that had revealed a higher expression in C3H mice (ie, potentially atheroprotective) with atherogenic stimuli and with aging.
Figure 5. Evaluation in the apoE knockout model of genes identified as differentially expressed between C3H and C57 strains. A, ApoE knockout mice (C57BL/6J-Apoetm1Unc) were fed normal chow versus high-fat diet for the maximum period of 40 weeks. Triplicate microarray experiments were performed for each time point using 3 pools of 5 aortas at 0, 4, 10, 24, and 40 weeks for regular and high-fat diet groups (total of 15 mice per time point). SOMs were used to visualize patterns of expression of genes of interest. Genes which were differentially regulated by aging (Figure 3, K-means clusters 1, 4, and 9 with higher expression in C3H and clusters 4, 6, and 14 with higher expression in C57) and genes identified with atherogenic stimuli (Figure 4, K-means clusters 3 and 9 with higher expression in C3H and clusters 8 and 10 with opposite pattern) as well as genes which were differentially expressed at the baseline time point (Figure 2), were grouped and their expression was studied using SOM analysis. SOM analysis reveals diverse patterns of expression of these genes throughout the development of atherosclerosis in apoE knockout mice. Cluster 8 contains genes that are consistently increasing in expression with progression of atherosclerosis. Pie charts reflect the analysis group from which the genes populating each cluster were derived. The relative size of sectors of the pie chart indicates the relative number of genes that are derived from the various starting groups. Select genes from SOM clusters 4, 5, and 8 are listed in Table Xa. Please refer to Table Xb for full lists of genes in these clusters (available online at http://atvb.ahajournals.org).
Given C3H resistance and C57 susceptibility to atherosclerosis, as an initial hypothesis it was postulated that genes with higher expression in C3H mice confer resistance, whereas genes with higher expression in C57 mice may have a proatherogenic role. With this point of reference, gene clusters were further examined (Table X, available online at http://atvb.ahajournals.org). For example, limiting the list of genes in SOM cluster 8 (genes with increased expression with atherosclerosis) to those that also had higher baseline expression in C57 mice yielded an interesting set of genes that may be atherogenic. This group included inflammation related genes, such as H2-D1, Pdgfc, Paf, and Cd47 (Table Xa). Other compelling genes included Agpt2, Mglap, Xdh, Th, and Ctsc. Conversely, limiting the list of genes in clusters 4 and 5 to those with higher expression in C3H mice identified a group of genes with potential athero-protective function. Some of those genes included Ppar, Pparbp as well as Ptp4a1, and Mcd (Table Xa).
Lesion Analysis in the Genetic Models
To address whether some of the gene expression differences are related to presence of atherosclerotic lesion in C57 mice, we determined the total atherosclerotic burden in the aorta by calculating a percent lesion area in aortas of C57 (n=5) and C3H (n=5) mice. Comparisons were made at time 0 and 40 weeks on normal or high-fat diet. Non–cholate-containing high-fat diet was used to prevent caustic effects on the vascular wall. As expected, C57 and C3H mice on either diet did not demonstrate evidence of atherosclerosis throughout the course of the experiment, suggesting that observed gene expression changes cannot be explained by different cellular composition of the vessel wall (Figure XI, available online at http://atvb.ahajournals.org). Although minimal fatty infiltrates were noted on histological evaluation of the aortic root in C57 mice on high-fat diet, there were no obvious changes in inflammatory cell infiltrate.
Quantitative RT-PCR Validation of Expression Differences
To validate the array results with quantitative RT-PCR and assure that our statistical analyses were identifying truly differentially expressed genes, 10 representative genes were assayed by quantitative RT-PCR (Table XII, available online at http://atvb.ahajournals.org). We attempted to recapitulate the essence of the experimental design of the microarray experiment by using several genes from each group of significant genes. As can be seen in Figure XIII (available online at http://atvb.ahajournals.org), there is high degree of correlation between the 2 methodologies (Pearson correlation of 0.86), validating the results of the microarray analyses.
Discussion
In this study, we undertook a transcriptional profiling of aortic tissue from mouse strains that are resistant (C3H/HeJ) or susceptible (C57Bl/6) to atherosclerosis using a 22 000 gene Agilent microarray. We compared their gene expression at baseline, as well is in a longitudinal fashion during normal aging and during exposure to an atherogenic stimulus (high-fat diet). The hypothesis implicit in these studies has been that gene expression differences in the disease target tissue between mice resistant and those susceptible to atherosclerosis will identify genes that promote, and perhaps genes that inhibit, the development of disease. Gene expression differences before the disease promoting effects of age and high-fat diet may modulate the earliest disease-associated vascular wall signals, thus determining susceptibility, whereas gene expression differences during the course of disease may reflect ongoing molecular processes that mediate the progression of disease.
There have been previous studies using gene expression analysis, including microarray methodology, to identify genes differentially expressed in atherosclerosis.19–22 However, because of limitations of sample availability in humans and experimental design in rodent studies, these efforts have tended to identify those genes expressed at high levels at the terminal phase of the disease. In a broader approach to identification of differentially regulated genes, we have used a temporal experimental design and analytical approach that should identify genes that may be only transiently expressed during the disease process. Our AUC algorithm for data addresses the rigorous analysis of temporal microarray data, using an AUC calculation. The area is a natural measure of differential change over time between 2 comparison groups. It is large if there is a consistent rise or fall over time or a sudden change followed by a leveling off. This measure has been coupled with additional steps to help identify significant values and should have broad utility for a variety of applications to longitudinal array data. Additionally, in these studies, we have used statistically rigorous methods to identify terms and pathways that are reflected by the genes identified to be differentially regulated.
By validating the genes differentially regulated between the C3H and C57 mice in the apoE model of atherosclerosis, it has been possible to highlight those genes and pathways most likely to underlie the inherited disease susceptibility differences between these strains. PPAR and Pparbp, which bind PPAR-, RXR, and TR-?-1, were noted to have increased levels in C3H mice and decreased expression during progression of atherosclerosis in apoE mice. The role of peroxisome proliferator–activated receptors in atherosclerosis, inflammation, and insulin-resistance is becoming increasingly clear.23 Although Ppar has been shown to play a protective role in atherosclerosis, the role of Pparbp remains poorly understood. Ptp4a1 was noted to have a higher expression in C3H on high-fat diet and lower expression in apoE-deficient mice. Protein tyrosine phosphatases have been implicated in a variety of biological processes including cell-matrix adhesion and cell migration processes. They can also control the formation and maintenance of the actin cytoskeleton and regulate small GTPase molecular switches.24 Nevertheless, their roles have not been clearly defined in atherosclerosis. Mcd (malonyl-coenzyme A [CoA] decarboxylase), which showed a similar pattern of expression to Ptp4a1, catalyzes the conversion of malonyl-CoA to acetyl-CoA in the fatty acid biosynthesis pathway. A recent study has shown that increased levels of Mcd, which decreases circulating free fatty acids and liver triglcyceride content, is associated with improved insulin sensitivity.25 Hmgcs2 (3-hydroxy-3-methylglutaryl [HMG] CoA synthase) was noted to have lower expression in C3H mice as well. This enzyme condenses acetyl-CoA with acetoacetyl-CoA to form HMG-CoA, which is the substrate for HMG-CoA reductase. HMG-CoA reductase inhibitors are widely used to treat hypercholesterolemia and atherosclerosis. The relationship between HMG-CoA reductase inhibition with statin drugs and inflammation is a subject of intense clinical investigation.
Gene expression differences between the 2 strains suggest that C57Bl/6 mice have a higher genetic propensity to develop inflammation in the presence of appropriate atherogenic stimuli. A number of immune modulatory genes had higher expression in C57 versus C3H mice at baseline and were upregulated throughout the progression of atherosclerosis. Cd47 modulates a variety of cell functions, such as adhesion, spreading, and migration. Interaction of Cd47 with thrombospondin-1 has been shown to perpetuate inflammation in rheumatoid synovitis,26 and it may also contribute to platelet adhesion to tumor necrosis factor–stimulated vascular endothelial cells.27 Both major histocompatibility complex classes I and II proteins, which function in antigen presenting complexes, are of particular interest given their crucial role in modulation of inflammation. Our data demonstrate the differential expression of a number of these genes in C57 and C3H mice. H2-D1 (Histocompatibility 2-D1 antigen), H2-Eb1, Bat5, and major histocompatibility complex class II antigen A, beta (H2-Ab) demonstrated higher expression levels in the C57 mice. No prior role in atherosclerosis susceptibility has been reported for these genes. However, an intriguing recent study has implicated a potential role of H2-Ab in autoimmune mediated cardiomyopathy.28 Interestingly, hepatic expression of H2-Eb1 and H2Ab1 is upregulated in mice fed high-fat diet, which may suggest a relationship between lipid metabolism and immunity.29 Among the immune related genes with higher expression in C3H, Ifi202b (interferon-activated gene 202b) was noted to have a higher expression in C3H at baseline (data not shown). Ifi202b is a p202 protein whose overexpression is growth inhibitory and which can bind and inhibit the activity of numerous transcription factors including cJun, cFos, NFB, E2f1, E2f4, Myod, and myogenin.30
Whether these inflammatory pathways or 1 or more of the other fundamental cellular pathways outlined above are primarily responsible for the inherited differences in atherosclerosis susceptibility remains to be elucidated, but this work serves to focus future studies with these mouse models.
Acknowledgments
This work was supported by the Donald W. Reynolds Cardiovascular Clinical Research Center at Stanford University. We thank David Deng, Arnold Liao, Nick Sampas, David A. Sanan, and Peng Zhang for expert assistance.
Received May 10, 2004; accepted November 11, 2004.
References
Paigen B, Morrow A, Brandon C, Mitchell D, Holmes P. Variation in susceptibility to atherosclerosis among inbred strains of mice. Atherosclerosis. 1985; 57: 65–73.
Fazio S, Linton MF. Mouse models of hyperlipidemia and atherosclerosis. Front Biosci. 2001; 6: D515–525.
Glass CK, Witztum JL. Atherosclerosis. The road ahead. Cell. 2001; 104: 503–516.
Breslow JL. Mouse models of atherosclerosis. Science. 1996; 272: 685–688.
Darvasi A. Genomics: gene expression meets genetics. Nature. 2003; 422: 269–270.
Schadt EE, Monks SA, Drake TA, Lusis AJ, Che N, Colinayo V, Ruff TG, Milligan SB, Lamb JR, Cavet G, Linsley PS, Mao M, Stoughton RB, Friend SH. Genetics of gene expression surveyed in maize, mouse and man. Nature. 2003; 422: 297–302.
Mir AA, Myakishev MV, Polesskaya OO, Moitra J, Petersen D, Miller L, Orosz A, Vinson C. A search for candidate genes for lipodystrophy, obesity and diabetes via gene expression analysis of A-ZIP/F-1 mice. Genomics. 2003; 81: 378–390.
Shi W, Wang NJ, Shih DM, Sun VZ, Wang X, Lusis AJ. Determinants of atherosclerosis susceptibility in the C3H and C57BL/6 mouse model: evidence for involvement of endothelial cells but not blood cells or cholesterol metabolism. Circ Res. 2000; 86: 1078–1084.
Available at http://labanimals.stanford.edu/index.html.
Tabibiazar R, Wagner RA, Liao A, Quertermous T. Transcriptional profiling of the heart reveals chamber-specific gene expression patterns. Circ Res. 2003; 93: 1193–1201.
Available at http://mozart.stanford.edu/TQLab/lvad/tabibiazar/cross-strain/index.htm.
Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A. 2001; 98: 5116–5121.
Available at http://www-stat.stanford.edu/tibs/SAM/.
Chen MM, Ashley EA, Deng DX, Tsalenko A, Deng A, Tabibiazar R, Ben-Dor A, Fenster B, Yang E, King JY, Fowler M, Robbins R, Johnson FL, Bruhn L, McDonagh T, Dargie H, Yakhini Z, Tsao PS, Quertermous T. Novel role for the potent endogenous inotrope apelin in human cardiac dysfunction. Circulation. 2003; 108: 1432–1439.
Available at http://www.genedata.com/products/expressionist/.
Available at http://mozart.stanford.edu/heatmap.htm.
Hosack DA, Dennis G Jr, Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol. 2003; 4: R70.
Tangirala RK, Rubin EM, Palinski W. Quantitation of atherosclerosis in murine models: correlation between lesions in the aortic origin and in the entire aorta, and differences in the extent of lesions between sexes in LDL receptor-deficient and apolipoprotein E-deficient mice. J Lipid Res. 1995; 36: 2320–2328.
Archacki SR, Angheloiu G, Tian XL, Tan FL, DiPaola N, Shen GQ, Moravec C, Ellis S, Topol EJ, Wang Q. Identification of new genes differentially expressed in coronary artery disease by expression profiling. Physiol Genomics. 2003; 15: 65–74.
McCaffrey TA, Fu C, Du B, Eksinar S, Kent KC, Bush H Jr, Kreiger K, Rosengart T, Cybulsky MI, Silverman ES, Collins T. High-level expression of Egr-1 and Egr-1-inducible genes in mouse and human atherosclerosis. J Clin Invest. 2000; 105: 653–662.
Randi AM, Biguzzi E, Falciani F, Merlini P, Blakemore S, Bramucci E, Lucreziotti S, Lennon M, Faioni EM, Ardissino D, Mannucci PM. Identification of differentially expressed genes in coronary atherosclerotic plaques from patients with stable or unstable angina by cDNA array analysis. J Thromb Haemost. 2003; 1: 829–835.
Wuttge DM, Sirsjo A, Eriksson P, Stemme S. Gene expression in atherosclerotic lesion of ApoE deficient mice. Mol Med. 2001; 7: 383–392.
Glass CK. Potential roles of the peroxisome proliferator-activated receptor-gamma in macrophage biology and atherosclerosis. J Endocrinol. 2001; 169: 461–464.
Larsen M, Tremblay ML, Yamada KM. Phosphatases in cell-matrix adhesion and migration. Nat Rev Mol Cell Biol. 2003; 4: 700–711.
An J, Muoio DM, Shiota M, Fujimoto Y, Cline GW, Shulman GI, Koves TR, Stevens R, Millington D, Newgard CB. Hepatic expression of malonyl-CoA decarboxylase reverses muscle, liver and whole-animal insulin resistance. Nat Med. 2004; 10: 268–274.
Vallejo AN, Yang H, Klimiuk PA, Weyand CM, Goronzy JJ. Synoviocyte-mediated expansion of inflammatory T cells in rheumatoid synovitis is dependent on CD47-thrombospondin 1 interaction. J Immunol. 2003; 171: 1732–1740.
Lagadec P, Dejoux O, Ticchioni M, Cottrez F, Johansen M, Brown EJ, Bernard A. Involvement of a CD47-dependent pathway in platelet adhesion on inflamed vascular endothelium under flow. Blood. 2003; 101: 4836–4843.
Elliott JF, Liu J, Yuan ZN, Bautista-Lopez N, Wallbank SL, Suzuki K, Rayner D, Nation P, Robertson MA, Liu G, Kavanagh KM. Autoimmune cardiomyopathy and heart block develop spontaneously in HLA-DQ8 transgenic IAbeta knockout NOD mice. Proc Natl Acad Sci U S A. 2003; 100: 13447–13452.
Vergnes L, Phan J, Strauss M, Tafuri S, Reue K. Cholesterol and cholate components of an atherogenic diet induce distinct stages of hepatic inflammatory gene expression. J Biol Chem. 2003; 278: 42774–42784.
Wang H, Chatterjee G, Meyer JJ, Liu CJ, Manjunath NA, Bray-Ward P, Lengyel P. Characteristics of three homologous 202 genes (Ifi202a, Ifi202b, and Ifi202c) from the murine interferon-activatable gene 200 cluster. Genomics. 1999; 60: 281–294.(Raymond Tabibiazar; Roger)
Correspondence to Raymond Tabibiazar, Stanford Medical School, Division of Cardiovascular Medicine, 300 Pasteur Drive, Falk CVRC, Stanford, CA 94305 (E-mail rtabibiazar@cvmed.stanford.edu) or Thomas Quertermous, Stanford Medical School, Division of Cardiovascular Medicine, 300 Pasteur Drive, Falk CVRC, Stanford, CA 94305 (E-mail tomq1@stanford.edu).
Abstract
Objective— Different strains of inbred mice exhibit different susceptibility to the development of atherosclerosis. The C3H/HeJ and C57Bl/6 mice have been used in several studies aimed at understanding the genetic basis of atherosclerosis. Under controlled environmental conditions, variations in susceptibility to atherosclerosis reflect differences in genetic makeup, and these differences must be reflected in gene expression patterns that are temporally related to the development of disease. In this study, we sought to identify the genetic pathways that are differentially activated in the aortas of these mice.
Methods and Results— We performed genome-wide transcriptional profiling of aortas from C3H/HeJ and C57Bl/6 mice. Differences in gene expression were identified at baseline as well as during normal aging and longitudinal exposure to high-fat diet. The significance of these genes to the development of atherosclerosis was evaluated by observing their temporal pattern of expression in the well-studied apolipoprotein E model of atherosclerosis.
Conclusion— Gene expression differences between the 2 strains suggest that aortas of C57Bl/6 mice have a higher genetic propensity to develop inflammation in response to appropriate atherogenic stimuli. This study expands the repertoire of factors in known disease-related signaling pathways and identifies novel candidate genes for future study.
To gain insights into the molecular pathways that are differentially activated in strains of mice with varied susceptibility to atherosclerosis, we performed comprehensive transcriptional profiling of their vascular wall. Genes identified through these studies expand the repertoire of factors in disease-related signaling pathways and identify novel candidate genes in atherosclerosis.
Key Words: atherosclerosis ? vascular disease ? gene expression ? microarray ? inflammation ? oxidative stress
Introduction
Mouse strains with different genetic backgrounds have helped to advance the study of atherosclerosis and are an attractive model for investigating the intricate interactions between environmental and genetic factors that contribute to the development of this complex disease.1,2 Although there is great variation in disease susceptibility in a variety of mouse strains, the atherosclerosis-resistant C3H/HeJ (C3H) and the atherosclerosis-susceptible C57Bl/6 (C57) strains have emerged as popular models for cardiovascular research. Manipulation of expression of individual genes through transgenesis and gene targeting, such as the apolipoprotein E (apoE) gene knockout in C57Bl/6 mice, has added significantly to the repertoire of genetic models available for the study of atherosclerosis.3,4 Despite the availability of strains with differing susceptibilities to atherosclerosis, there remains little information regarding the underlying disease processes in the vessel wall that mediate atherosclerosis.
With the availability of genomic tools that allow genome-wide transcriptional profiling, individual environmental risk factors can be evaluated in a controlled setting to define gene expression differences that are associated with the risk, without a priori knowledge of specific candidate genes. Recent studies have shown that transcriptional profiling can be used to characterize gene expression patterns associated with inherited genetic differences.5 The integration of gene expression data with genetic analyses has been used to identify susceptibility loci for complex diseases, such as diabetes, obesity, and allergic asthma, and to prioritize candidate genes.6,7 Therefore, gene expression analysis of genetically distinct mice with different disease susceptibilities under controlled environmental conditions can potentially identify genes whose expression underlies differences in disease susceptibility.
Prior studies investigating genetic influences on atherosclerosis susceptibility in various mouse strains have provided evidence for involvement of the vascular wall and not peripheral blood cells or systemic cholesterol metabolism in the development of atherosclerosis.8 We thus hypothesized that genes differentially expressed in the vascular wall in genetically resistant C3H versus genetically susceptible C57 mice may play critical roles in the initiation and modulation of atherosclerosis. To identify these genes, we have performed genome-wide transcriptional profiling of mouse aortic tissue from these 2 mouse strains and compared their vascular wall gene expression. Using a comprehensive study design and novel analytical tools, we show that activity of specific biological processes and molecular functions distinguish disease-resistant versus disease-prone mouse strains. Further, the importance of these genes in development of atherosclerosis is validated with studies in the apoE model.
Methods
RNA Preparation and Hybridization to the Microarray
Three-week-old female C3H/HeJ, C57Bl/6J, and apoE knockout mice (C57BL/6J-Apoetm1Unc) were purchased from Jackson Laboratories (Bar Harbor, Me). At 4 weeks of age, the mice were either continued on normal chow or switched to non–cholate-containing high-fat diet, which included 21% anhydrous milkfat and 0.15% cholesterol (Dyets #101511, Dyets Inc) for a maximum period of 40 weeks. At each of the time points, including 0 (baseline), 4, 10, 24, and 40 weeks, for each of the conditions (strain-diet combination), 15 mice were harvested for RNA isolation, for a total of 450 mice. Following Stanford University animal care guidelines,9 the mice were anesthetized with Avertin and perfused with normal saline.10 The aortas from the root to the common iliacs were carefully dissected, flash frozen in liquid nitrogen (LN2), and divided into 3 pools of 5 aortas for further RNA isolation. Total RNA was isolated as described previously.10 First strand cDNA was synthesized from 10 μg of total RNA from each pool and from whole e17.5-day embryo for reference RNA in the presence of Cy5 or Cy3 dCTP, respectively, and hybridized to a mouse 60-mer oligo microarray (G4120A, Agilent Technologies), generating 3 biological replicates for each time point. For details, please see online Methods,11 available at http://atvb.ahajournals.org.
Data Processing
Array image acquisition and feature extraction was performed using the Agilent G2565AA Microarray Scanner and feature extraction software version A.6.1.1. Normalization was performed using a LOWESS algorithm, and Dye-normalized signals were used in calculating log ratios. Features with reference values of <2.5 SDs above background for the negative control features were regarded as missing values. Those features with values in at least two thirds of the experiments and present in at least 1 of the replicates were retained for further analysis. For significance analysis of microarrays (SAM), a K-nearest-neighbor algorithm was applied to impute for missing values.10
Data Analysis
Experimental design and analysis flow chart is depicted in Figure 1. SAM was used to identify genes with statistically different expression between the C3H and C57 mice at baseline.10,12–14 For partitioning clustering of the genes with K-means and self-organizing maps (SOMs), we used positive correlation for distance determination and required complete linkage, which uses the greatest distance between genes to ascribe similarity. SOMs and K-means analyses were performed using Expressionist software (GeneData Inc).15 Heatmaps were generated using HeatMap Builder.14,16 For enrichment analysis we used the EASE analysis software, which uses gene ontology (GO) annotation and the Fisher exact test to derive biological themes within particular gene sets.17 For time-course study a new statistical algorithm, the area under curve (AUC) analysis, was devised. For each sequence of 4 triplicate gene expression measurements over time, we first subtracted the measurement at time 0 from all values. We then computed the signed AUC. The area is a natural measure of change over time. These areas were then used to compute an F statistic for comparing C57 and C3H mice across the different diets. A permutation analysis, similar to that used in SAM, was performed to estimate the false discovery rate (Q value or "FDR") for different levels of the F statistic. For ease of presentation, genes which meet our FDR cutoffs will be referred to as "significant" throughout the remainder of this article. All microarray data were submitted to the National Center for Biotechnology Information Gene Expression Omnibus (GEO GSE1560; http://www.ncbi.nlm.nih.gov/geo/).
Figure 1. Experimental design and analysis flowcharts. A, Four-week-old female C3H/HeJ (C3H) and C57Bl/6 (C57) mice were fed normal chow vs high-fat diet for the maximum period of 40 weeks. Triplicate microarray experiments were performed for each time point using 3 pools of 5 aortas at 0, 4, 10, 24, and 40 weeks on either diet (total of 15 mice per time point). B, Data analysis overview. Of the 20 283 genes present on the array, 311 genes were found to be significantly differentially expressed between C3H and C57 mice at baseline (SAM FDR 10% and >1.5-fold change). Differential gene expression during aging was determined by comparing C57 vs C3H time-course differences on normal and atherogenic high-fat diets using the AUC analysis.
Aortic Lesion Analysis
For select time points within various experimental groups, 5 to 7 female mice were used for histological lesion analysis. Atherosclerosis lesion area was determined as described previously.18
Quantitative Real-Time Reverse Transcriptase–Polymerase Chain Reaction
Primers and probes for 10 representative differentially expressed genes were obtained from Applied Biosystems Assays-on-Demand. A Total of 90 reactions were performed from representative RNA samples used for microarray experiments. These included triplicate assay on 3 pools of 5 aortas. cDNA was synthesized and Taqman was performed as described previously.10
Results
Baseline Differences in Gene Expression Patterns Between the Mouse Strains
Differences in gene expression levels between the 2 strains at baseline, before effects of aging or diet become apparent, may identify genes that play a role in determining vascular wall disease susceptibility. To identify such genes we used SAM to compare the vascular wall gene expression of C3H versus C57 mice at 4 weeks of age, with all animals on normal chow diet. SAM identified 311 genes as being significantly differentially expressed (FDR<0.1 with >1.5-fold difference), and expression patterns of these genes provided a clear partition between C3H and C57 mice (Figure 2). A separate 2-class comparison (SAM, FDR<0.1) between C57 and apoE-deficient mice with a C57Bl/6 genetic background revealed only a few genes, including apoE, which were differentially expressed in the 2 groups of mice (data not shown).
Figure 2. Differential gene expression between C3H and C57 mice at baseline. The SAM analysis shown was associated with an FDR of 10%, and a total of 311 probes were identified as differentially regulated at this level of confidence. Lists represent a select group of genes (expressed sequence tags excluded) with higher expression in C3H (top 20 ranking genes) and C57 (top 45 ranking genes). The heatmap reflects normalized gene expression ratios and is organized with individual hybridizations for each of the 3 replicates for each mouse strain arranged along the x axis, with normalized expression ratios depicted by color intensity such that highest expression corresponds to bright red, and the lowest expression corresponds to bright green. See supplemental material for full list of genes (available online at http://atvb.ahajournals.org).
Comparison of C3H and C57 vascular wall gene expression at baseline provided a list of compelling candidate genes, which reflected differences in biological processes such as growth, differentiation, and inflammation, as well as molecular functions, such as cathecholamine synthesis, phosphatase activity, peroxisome function, insulin-like growth factor activity, and antigen presentation (Figure 2). These processes were exemplified by higher expression of genes such as Cdkn1a, Pparbp, protein tyrosine phosphatase-4a2, and Socs5 in C3H mice, compared with genes such as ABCC1, H2-D1, Bat5, IGFBP1, SCD1, and Serpine6b, which demonstrated higher expression in C57 mice (Tables I and II, available online at http://atvb.ahajournals.org). These fundamental baseline gene expression differences may determine disease susceptibility as the mice are exposed to age-related stimuli or dietary challenges.
Age-Related Differences in Gene Expression Patterns Between the Mouse Strains
To further examine the vascular wall gene expression differences between C57 and C3H mice, we performed analyses to identify genes differentially expressed in response to aging (Figure 3). Data were collected at 5 time points over a 40-week period. To identify such genes, we developed the AUC analysis. The AUC analysis relies on a permutation procedure to reduce the number of potential false-positives generated because of multiple testing but still uses the increase in statistical power of time-course experimental design. Comparing C57 versus C3H time-course differences on normal diet with a rigid cutoff (FDR<0.05) did not identify any genes. However, relaxing the AUC stringency (F statistic>10, FDR<0.45) allowed a large number of genes (413) to be included for pathway overrepresentation analysis using GO annotation (Table III, available online at http://atvb.ahajournals.org). Functional annotation and group overrepresentation analysis (Fisher test probability value <0.02) of the resultant differentially expressed genes revealed differences in a number of biological processes, including growth and development as well as a number of molecular functions, such as cell cycle control, regulation of mitosis, and metabolism (Figure 3B). Some of these processes are exemplified by genes with higher expression in C57 mice, such as Aoc1 (prooxidative stress), Bub1 (cell cycle check point), Cyclin B2, and genes with higher expression in C3H, including INHBA and INHBB (Table IV, available online at http://atvb.ahajournals.org).
Figure 3. Differential gene expression between C3H and C57 mice in response to normal aging. A, Response to aging was determined by comparing C57 vs C3H time-course differences on normal diet (AUC analysis F statistic>10). B, Functional annotation of the 413 differentially expressed genes reveal differences in various biological processes, including growth and differentiation. The probability rates provided are based on Fisher exact test (P<0.02).17 C, K-means clustering of the 413 genes reveals several profiles of gene expression. Clusters 1, 4, and 9 reveal increased gene expression in C3H vs C57 mice, whereas clusters 2, 6, and 14 reveal the opposite pattern. See Table VIII for lists of genes in the clusters (available online at http://atvb.ahajournals.org).
Temporally variable genes identified by AUC analysis were further characterized with K-means clustering to identify dynamic patterns of expression during the aging process (Figure 3C). Clusters 1, 4, and 9 revealed either higher overall expression or temporally increasing levels of expression in C3H mice compared with C57 mice. In contrast, clusters 2, 6, and 14 revealed the opposite pattern. Of the genes which were noted to be differentially expressed in the 2 strains during aging, 51 genes were also differentially expressed at baseline, suggesting that baseline differences of certain genes can further be affected with aging.
Diet-Related Differences in Gene Expression Patterns Between the Mouse Strains
Differential vascular wall response to atherogenic stimuli was determined by comparing temporal gene expression patterns in C57 versus C3H mice on high-fat diet (Figure 4A). Comparing C57 versus C3H time-course differences on high-fat diet with a rigid cutoff (FDR<0.05) identified 35 genes, including Hgfl and Tgfb4, which were downregulated in C57 on high-fat diet. Additional known genes as well as a number of expressed sequence tags were also identified (Table V, available online at http://atvb.ahajournals.org). Using a less stringent AUC cutoff allowed identification of a larger number of genes that could be evaluated with pathway overrepresentation analysis using GO annotation. At this level of stringency (F statistic>10, FDR<0.35), a total of 650 genes with temporally variable expression were identified (Table VI, available online at http://atvb.ahajournals.org). Genes that were also differentially regulated by the aging process (141 of 650 genes) were excluded from further analysis of this group. Thirty-eight of the remaining 509 genes were among those differentially expressed at baseline. Functional annotation and group overrepresentation analysis (Fisher test P<0.02) of these differentially expressed genes revealed differences in biological processes, such as catabolism, oxygen reactive species and superoxide metabolism, and proteo- and peptidolysis as well as molecular functions, such as fatty acid metabolism, oxidoreductase, and methyltransferase activities (Figure 4B). Interestingly, this analysis suggested important differences between the 2 mouse strains with respect to the activity of the peroxisome, microbody, and lysosome. Some of these processes were exemplified by genes with higher expression in C3H mice, such as Ccs, Ephx2, Gpx4, Prdx6 (antioxidants), Sirt3 (transcriptional repressor), PPAR, and Mcd as well as genes with higher expression in C57 mice, such as Lysyl oxidase and Cdkn1a (Table VII, available online at http://atvb.ahajournals.org). K-means clustering of these genes identified a small number of distinct expression patterns (Figure 4C), with clusters 3 and 9 revealing increased gene expression in C3H mice and clusters 8 and 10 showing the opposite pattern.
Figure 4. Differential gene expression between C3H and C57 mice in response to high-fat diet. A, Response to atherogenic stimulus was determined by comparing C57 vs C3H time-course differences on high-fat diet (AUC analysis F statistic>10). B, Functional annotation of the 509 differentially expressed genes reveals differences in various biological processes and cellular components. The probability rates provided are based on Fisher exact test (P<0.02).17 C, K-means clustering of the 509 differentially expressed genes revealed several patterns of gene expression with clusters 3 and 9 exhibiting increased gene expression in C3H vs C57 mice and clusters 8 and 10 with the opposite pattern. See Table IX for lists of genes in the clusters (available online at http://atvb.ahajournals.org).
Evaluation of Strain-Specific Differentially Regulated Genes in the ApoE Model
Using these techniques, we have identified a significant number of genes that are differentially expressed in the atherosclerosis-resistant C3H and susceptible C57 mice, some of which are likely involved in atherogenesis and some of which are likely irrelevant to the process. To further select genes most likely to be involved in atherogenesis, we examined their expression in apoE-deficient mice fed normal or high-fat diet over a period of 40 weeks (Figure 5). We used SOM analysis to visualize the expression profiles of these subsets of genes throughout the development and progression of atherosclerosis in the apoE-deficient mice. The analysis revealed several patterns of gene expression. For example, SOM cluster 8 demonstrated a consistently increasing pattern of expression which correlated with disease progression in the apoE-deficient mice (Figure 5). As evidenced by the pie chart, this cluster is enriched with genes that were identified as more highly expressed in C57 versus C3H mice at baseline (ie, potentially atherogenic). In contrast, clusters 4, 5, and 6 showed decreasing expression with disease progression. The decreased expression of genes in cluster 4 was somewhat attenuated with high-fat challenge of the apoE-deficient mice. This cluster was particularly enriched with genes that had revealed a higher expression in C3H mice (ie, potentially atheroprotective) with atherogenic stimuli and with aging.
Figure 5. Evaluation in the apoE knockout model of genes identified as differentially expressed between C3H and C57 strains. A, ApoE knockout mice (C57BL/6J-Apoetm1Unc) were fed normal chow versus high-fat diet for the maximum period of 40 weeks. Triplicate microarray experiments were performed for each time point using 3 pools of 5 aortas at 0, 4, 10, 24, and 40 weeks for regular and high-fat diet groups (total of 15 mice per time point). SOMs were used to visualize patterns of expression of genes of interest. Genes which were differentially regulated by aging (Figure 3, K-means clusters 1, 4, and 9 with higher expression in C3H and clusters 4, 6, and 14 with higher expression in C57) and genes identified with atherogenic stimuli (Figure 4, K-means clusters 3 and 9 with higher expression in C3H and clusters 8 and 10 with opposite pattern) as well as genes which were differentially expressed at the baseline time point (Figure 2), were grouped and their expression was studied using SOM analysis. SOM analysis reveals diverse patterns of expression of these genes throughout the development of atherosclerosis in apoE knockout mice. Cluster 8 contains genes that are consistently increasing in expression with progression of atherosclerosis. Pie charts reflect the analysis group from which the genes populating each cluster were derived. The relative size of sectors of the pie chart indicates the relative number of genes that are derived from the various starting groups. Select genes from SOM clusters 4, 5, and 8 are listed in Table Xa. Please refer to Table Xb for full lists of genes in these clusters (available online at http://atvb.ahajournals.org).
Given C3H resistance and C57 susceptibility to atherosclerosis, as an initial hypothesis it was postulated that genes with higher expression in C3H mice confer resistance, whereas genes with higher expression in C57 mice may have a proatherogenic role. With this point of reference, gene clusters were further examined (Table X, available online at http://atvb.ahajournals.org). For example, limiting the list of genes in SOM cluster 8 (genes with increased expression with atherosclerosis) to those that also had higher baseline expression in C57 mice yielded an interesting set of genes that may be atherogenic. This group included inflammation related genes, such as H2-D1, Pdgfc, Paf, and Cd47 (Table Xa). Other compelling genes included Agpt2, Mglap, Xdh, Th, and Ctsc. Conversely, limiting the list of genes in clusters 4 and 5 to those with higher expression in C3H mice identified a group of genes with potential athero-protective function. Some of those genes included Ppar, Pparbp as well as Ptp4a1, and Mcd (Table Xa).
Lesion Analysis in the Genetic Models
To address whether some of the gene expression differences are related to presence of atherosclerotic lesion in C57 mice, we determined the total atherosclerotic burden in the aorta by calculating a percent lesion area in aortas of C57 (n=5) and C3H (n=5) mice. Comparisons were made at time 0 and 40 weeks on normal or high-fat diet. Non–cholate-containing high-fat diet was used to prevent caustic effects on the vascular wall. As expected, C57 and C3H mice on either diet did not demonstrate evidence of atherosclerosis throughout the course of the experiment, suggesting that observed gene expression changes cannot be explained by different cellular composition of the vessel wall (Figure XI, available online at http://atvb.ahajournals.org). Although minimal fatty infiltrates were noted on histological evaluation of the aortic root in C57 mice on high-fat diet, there were no obvious changes in inflammatory cell infiltrate.
Quantitative RT-PCR Validation of Expression Differences
To validate the array results with quantitative RT-PCR and assure that our statistical analyses were identifying truly differentially expressed genes, 10 representative genes were assayed by quantitative RT-PCR (Table XII, available online at http://atvb.ahajournals.org). We attempted to recapitulate the essence of the experimental design of the microarray experiment by using several genes from each group of significant genes. As can be seen in Figure XIII (available online at http://atvb.ahajournals.org), there is high degree of correlation between the 2 methodologies (Pearson correlation of 0.86), validating the results of the microarray analyses.
Discussion
In this study, we undertook a transcriptional profiling of aortic tissue from mouse strains that are resistant (C3H/HeJ) or susceptible (C57Bl/6) to atherosclerosis using a 22 000 gene Agilent microarray. We compared their gene expression at baseline, as well is in a longitudinal fashion during normal aging and during exposure to an atherogenic stimulus (high-fat diet). The hypothesis implicit in these studies has been that gene expression differences in the disease target tissue between mice resistant and those susceptible to atherosclerosis will identify genes that promote, and perhaps genes that inhibit, the development of disease. Gene expression differences before the disease promoting effects of age and high-fat diet may modulate the earliest disease-associated vascular wall signals, thus determining susceptibility, whereas gene expression differences during the course of disease may reflect ongoing molecular processes that mediate the progression of disease.
There have been previous studies using gene expression analysis, including microarray methodology, to identify genes differentially expressed in atherosclerosis.19–22 However, because of limitations of sample availability in humans and experimental design in rodent studies, these efforts have tended to identify those genes expressed at high levels at the terminal phase of the disease. In a broader approach to identification of differentially regulated genes, we have used a temporal experimental design and analytical approach that should identify genes that may be only transiently expressed during the disease process. Our AUC algorithm for data addresses the rigorous analysis of temporal microarray data, using an AUC calculation. The area is a natural measure of differential change over time between 2 comparison groups. It is large if there is a consistent rise or fall over time or a sudden change followed by a leveling off. This measure has been coupled with additional steps to help identify significant values and should have broad utility for a variety of applications to longitudinal array data. Additionally, in these studies, we have used statistically rigorous methods to identify terms and pathways that are reflected by the genes identified to be differentially regulated.
By validating the genes differentially regulated between the C3H and C57 mice in the apoE model of atherosclerosis, it has been possible to highlight those genes and pathways most likely to underlie the inherited disease susceptibility differences between these strains. PPAR and Pparbp, which bind PPAR-, RXR, and TR-?-1, were noted to have increased levels in C3H mice and decreased expression during progression of atherosclerosis in apoE mice. The role of peroxisome proliferator–activated receptors in atherosclerosis, inflammation, and insulin-resistance is becoming increasingly clear.23 Although Ppar has been shown to play a protective role in atherosclerosis, the role of Pparbp remains poorly understood. Ptp4a1 was noted to have a higher expression in C3H on high-fat diet and lower expression in apoE-deficient mice. Protein tyrosine phosphatases have been implicated in a variety of biological processes including cell-matrix adhesion and cell migration processes. They can also control the formation and maintenance of the actin cytoskeleton and regulate small GTPase molecular switches.24 Nevertheless, their roles have not been clearly defined in atherosclerosis. Mcd (malonyl-coenzyme A [CoA] decarboxylase), which showed a similar pattern of expression to Ptp4a1, catalyzes the conversion of malonyl-CoA to acetyl-CoA in the fatty acid biosynthesis pathway. A recent study has shown that increased levels of Mcd, which decreases circulating free fatty acids and liver triglcyceride content, is associated with improved insulin sensitivity.25 Hmgcs2 (3-hydroxy-3-methylglutaryl [HMG] CoA synthase) was noted to have lower expression in C3H mice as well. This enzyme condenses acetyl-CoA with acetoacetyl-CoA to form HMG-CoA, which is the substrate for HMG-CoA reductase. HMG-CoA reductase inhibitors are widely used to treat hypercholesterolemia and atherosclerosis. The relationship between HMG-CoA reductase inhibition with statin drugs and inflammation is a subject of intense clinical investigation.
Gene expression differences between the 2 strains suggest that C57Bl/6 mice have a higher genetic propensity to develop inflammation in the presence of appropriate atherogenic stimuli. A number of immune modulatory genes had higher expression in C57 versus C3H mice at baseline and were upregulated throughout the progression of atherosclerosis. Cd47 modulates a variety of cell functions, such as adhesion, spreading, and migration. Interaction of Cd47 with thrombospondin-1 has been shown to perpetuate inflammation in rheumatoid synovitis,26 and it may also contribute to platelet adhesion to tumor necrosis factor–stimulated vascular endothelial cells.27 Both major histocompatibility complex classes I and II proteins, which function in antigen presenting complexes, are of particular interest given their crucial role in modulation of inflammation. Our data demonstrate the differential expression of a number of these genes in C57 and C3H mice. H2-D1 (Histocompatibility 2-D1 antigen), H2-Eb1, Bat5, and major histocompatibility complex class II antigen A, beta (H2-Ab) demonstrated higher expression levels in the C57 mice. No prior role in atherosclerosis susceptibility has been reported for these genes. However, an intriguing recent study has implicated a potential role of H2-Ab in autoimmune mediated cardiomyopathy.28 Interestingly, hepatic expression of H2-Eb1 and H2Ab1 is upregulated in mice fed high-fat diet, which may suggest a relationship between lipid metabolism and immunity.29 Among the immune related genes with higher expression in C3H, Ifi202b (interferon-activated gene 202b) was noted to have a higher expression in C3H at baseline (data not shown). Ifi202b is a p202 protein whose overexpression is growth inhibitory and which can bind and inhibit the activity of numerous transcription factors including cJun, cFos, NFB, E2f1, E2f4, Myod, and myogenin.30
Whether these inflammatory pathways or 1 or more of the other fundamental cellular pathways outlined above are primarily responsible for the inherited differences in atherosclerosis susceptibility remains to be elucidated, but this work serves to focus future studies with these mouse models.
Acknowledgments
This work was supported by the Donald W. Reynolds Cardiovascular Clinical Research Center at Stanford University. We thank David Deng, Arnold Liao, Nick Sampas, David A. Sanan, and Peng Zhang for expert assistance.
Received May 10, 2004; accepted November 11, 2004.
References
Paigen B, Morrow A, Brandon C, Mitchell D, Holmes P. Variation in susceptibility to atherosclerosis among inbred strains of mice. Atherosclerosis. 1985; 57: 65–73.
Fazio S, Linton MF. Mouse models of hyperlipidemia and atherosclerosis. Front Biosci. 2001; 6: D515–525.
Glass CK, Witztum JL. Atherosclerosis. The road ahead. Cell. 2001; 104: 503–516.
Breslow JL. Mouse models of atherosclerosis. Science. 1996; 272: 685–688.
Darvasi A. Genomics: gene expression meets genetics. Nature. 2003; 422: 269–270.
Schadt EE, Monks SA, Drake TA, Lusis AJ, Che N, Colinayo V, Ruff TG, Milligan SB, Lamb JR, Cavet G, Linsley PS, Mao M, Stoughton RB, Friend SH. Genetics of gene expression surveyed in maize, mouse and man. Nature. 2003; 422: 297–302.
Mir AA, Myakishev MV, Polesskaya OO, Moitra J, Petersen D, Miller L, Orosz A, Vinson C. A search for candidate genes for lipodystrophy, obesity and diabetes via gene expression analysis of A-ZIP/F-1 mice. Genomics. 2003; 81: 378–390.
Shi W, Wang NJ, Shih DM, Sun VZ, Wang X, Lusis AJ. Determinants of atherosclerosis susceptibility in the C3H and C57BL/6 mouse model: evidence for involvement of endothelial cells but not blood cells or cholesterol metabolism. Circ Res. 2000; 86: 1078–1084.
Available at http://labanimals.stanford.edu/index.html.
Tabibiazar R, Wagner RA, Liao A, Quertermous T. Transcriptional profiling of the heart reveals chamber-specific gene expression patterns. Circ Res. 2003; 93: 1193–1201.
Available at http://mozart.stanford.edu/TQLab/lvad/tabibiazar/cross-strain/index.htm.
Tusher VG, Tibshirani R, Chu G. Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A. 2001; 98: 5116–5121.
Available at http://www-stat.stanford.edu/tibs/SAM/.
Chen MM, Ashley EA, Deng DX, Tsalenko A, Deng A, Tabibiazar R, Ben-Dor A, Fenster B, Yang E, King JY, Fowler M, Robbins R, Johnson FL, Bruhn L, McDonagh T, Dargie H, Yakhini Z, Tsao PS, Quertermous T. Novel role for the potent endogenous inotrope apelin in human cardiac dysfunction. Circulation. 2003; 108: 1432–1439.
Available at http://www.genedata.com/products/expressionist/.
Available at http://mozart.stanford.edu/heatmap.htm.
Hosack DA, Dennis G Jr, Sherman BT, Lane HC, Lempicki RA. Identifying biological themes within lists of genes with EASE. Genome Biol. 2003; 4: R70.
Tangirala RK, Rubin EM, Palinski W. Quantitation of atherosclerosis in murine models: correlation between lesions in the aortic origin and in the entire aorta, and differences in the extent of lesions between sexes in LDL receptor-deficient and apolipoprotein E-deficient mice. J Lipid Res. 1995; 36: 2320–2328.
Archacki SR, Angheloiu G, Tian XL, Tan FL, DiPaola N, Shen GQ, Moravec C, Ellis S, Topol EJ, Wang Q. Identification of new genes differentially expressed in coronary artery disease by expression profiling. Physiol Genomics. 2003; 15: 65–74.
McCaffrey TA, Fu C, Du B, Eksinar S, Kent KC, Bush H Jr, Kreiger K, Rosengart T, Cybulsky MI, Silverman ES, Collins T. High-level expression of Egr-1 and Egr-1-inducible genes in mouse and human atherosclerosis. J Clin Invest. 2000; 105: 653–662.
Randi AM, Biguzzi E, Falciani F, Merlini P, Blakemore S, Bramucci E, Lucreziotti S, Lennon M, Faioni EM, Ardissino D, Mannucci PM. Identification of differentially expressed genes in coronary atherosclerotic plaques from patients with stable or unstable angina by cDNA array analysis. J Thromb Haemost. 2003; 1: 829–835.
Wuttge DM, Sirsjo A, Eriksson P, Stemme S. Gene expression in atherosclerotic lesion of ApoE deficient mice. Mol Med. 2001; 7: 383–392.
Glass CK. Potential roles of the peroxisome proliferator-activated receptor-gamma in macrophage biology and atherosclerosis. J Endocrinol. 2001; 169: 461–464.
Larsen M, Tremblay ML, Yamada KM. Phosphatases in cell-matrix adhesion and migration. Nat Rev Mol Cell Biol. 2003; 4: 700–711.
An J, Muoio DM, Shiota M, Fujimoto Y, Cline GW, Shulman GI, Koves TR, Stevens R, Millington D, Newgard CB. Hepatic expression of malonyl-CoA decarboxylase reverses muscle, liver and whole-animal insulin resistance. Nat Med. 2004; 10: 268–274.
Vallejo AN, Yang H, Klimiuk PA, Weyand CM, Goronzy JJ. Synoviocyte-mediated expansion of inflammatory T cells in rheumatoid synovitis is dependent on CD47-thrombospondin 1 interaction. J Immunol. 2003; 171: 1732–1740.
Lagadec P, Dejoux O, Ticchioni M, Cottrez F, Johansen M, Brown EJ, Bernard A. Involvement of a CD47-dependent pathway in platelet adhesion on inflamed vascular endothelium under flow. Blood. 2003; 101: 4836–4843.
Elliott JF, Liu J, Yuan ZN, Bautista-Lopez N, Wallbank SL, Suzuki K, Rayner D, Nation P, Robertson MA, Liu G, Kavanagh KM. Autoimmune cardiomyopathy and heart block develop spontaneously in HLA-DQ8 transgenic IAbeta knockout NOD mice. Proc Natl Acad Sci U S A. 2003; 100: 13447–13452.
Vergnes L, Phan J, Strauss M, Tafuri S, Reue K. Cholesterol and cholate components of an atherogenic diet induce distinct stages of hepatic inflammatory gene expression. J Biol Chem. 2003; 278: 42774–42784.
Wang H, Chatterjee G, Meyer JJ, Liu CJ, Manjunath NA, Bray-Ward P, Lengyel P. Characteristics of three homologous 202 genes (Ifi202a, Ifi202b, and Ifi202c) from the murine interferon-activatable gene 200 cluster. Genomics. 1999; 60: 281–294.(Raymond Tabibiazar; Roger)