Molecular networks in Dahl salt-sensitive hypertension based on transcriptome analysis of a panel of consomic rats

Mingyu Liang, Norman H. Lee, Hongying Wang, Andrew S. Greene, Anne E. Kwitek, Mary L. Kaldunski, Truong V. Luu, Bryan C. Frank, Scott Bugenhagen, Howard J. Jacob, Allen W. Cowley Jr.

Abstract

The Dahl salt-sensitive (SS) rat is a widely used model of human salt-sensitive hypertension and renal injury. We studied the molecular networks that underlie the complex disease phenotypes in the SS model, using a design that involved two consomic rat strains that were protected from salt-induced hypertension and one that was not protected. Substitution of Brown Norway (BN) chromosome 13 or 18, but not 20, into the SS genome was found to significantly attenuate salt-induced hypertension and albuminuria. Gene expression profiles were examined in the kidneys of SS and consomic SS-13BN, SS-18BN, and SS-20BN rats with a total of 240 cDNA microarrays. The substituted chromosome was overrepresented in genes differentially expressed between a consomic strain and SS rats on a 0.4% salt diet. F5, Serpinc1, Slc19a2, and genes represented by three other expressed sequence tags (ESTs), which are located on chromosome 13, were found to be differentially expressed between SS-13BN and all other strains examined. Likewise, Acaa2, B4galt6, Colec12, Hsd17b4, and five other ESTs located on chromosome 18 exhibited expression patterns unique to SS-18BN. On exposure to a 4% salt diet, there were 184 ESTs in the renal cortex and 346 in the renal medulla for which SS-13BN and SS-18BN shared one expression pattern, while SS and SS-20BN shared another, mirroring the phenotypic segregation among the four strains. Molecular networks that might contribute to the development of Dahl salt-sensitive hypertension and albuminuria were constructed with an approach that merged biological knowledge-driven analysis and data-driven Bayesian probabilistic analysis.

  • genomics
  • systems biology
  • kidney
  • blood pressure
  • diet

the dahl salt-sensitive (SS) rat develops hypertension and renal injury on exposure to high-salt intake and is a widely used model of human salt-sensitive forms of hypertension (1, 3, 21, 22). The precise identity of the genetic defects causing the enhanced blood pressure salt sensitivity in the SS rat is not clear (22). Moreover, the gene pathways participating in the development of salt-induced hypertension and renal injury in this rat model remain poorly understood.

One of the approaches to addressing these important yet complex questions is to study consomic or congenic rat strains that can be generated by substituting a chromosome or part of a chromosome from a salt-resistant strain of rat for the corresponding genomic region of the SS rat (1, 5, 6, 7, 11). The consomic or congenic rat strains have well-defined and often small genetic differences from the SS rat. Those strains that exhibit reduced blood pressure salt sensitivity, and/or ameliorated tissue organ injuries, compared with SS become informative model animals for studying genes and pathways involved in these disease processes (4, 9, 12, 18).

We previously used (14, 15) a highly focused microarray to study the expression profile of 1,800 mostly known genes in the renal medulla of SS and SS-13BN rats. SS-13BN/MCWi (SS-13BN) is a consomic strain in which chromosome 13 of the SS/JrHsdMcwi (SS) rat has been replaced by that of the salt-resistant Brown Norway (BN/MCWi; BN) rat, resulting in a significantly reduced blood pressure salt sensitivity (4). This focused approach generated interesting new insights into the molecular mechanisms underlying the SS phenotypes (14, 15), albeit with some limitations. For instance, only one consomic strain was used, making it difficult to determine which genes with expression changes were relevant to the phenotypes of interest. The focused microarray also did not allow an extensive interrogation of the transcriptome, including transcripts encoded by genes located on the substituted chromosome. In addition, only one region of the kidney, the renal medulla, was examined. The kidney is known to be important for long-term blood pressure regulation (2, 8). Not only are the cortex and the medulla both important in blood pressure regulation and renal injury, but they may be important in different ways. Studying the two regions may reveal insights not obtained from one region or the whole kidney.

The goal of the present study was to construct molecular networks involved in Dahl salt-sensitive hypertension and renal injury. Two additional consomic strains, SS-18BN/MCWi (SS-18BN) and SS-20BN/MCWi (SS-20BN), were analyzed together with SS and SS-13BN. SS-18BN mimics SS-13BN phenotypically, while SS-20BN resembles SS. The inclusion of all four strains of rat provided a powerful design that allowed us to identify strain-specific or phenotype-associated changes in gene expression. A microarray containing ∼28,000 expressed sequence tags (ESTs) was used. Both regions of the kidney, the renal cortex and the renal medulla, were examined in a time course design, requiring the use of a total of 240 microarrays. The extensive analysis revealed a series of new insights into the molecular networks underlying the complex hypertension and albuminuria phenotypes in the SS model.

MATERIALS AND METHODS

Consomic rats.

The consomic rat strains SS-13BN, SS-18BN, and SS-20BN were generated by substituting chromosomes 13, 18, and 20, respectively, from the salt-resistant Brown Norway (BN/MCWi) rat for the corresponding chromosome of the Dahl salt-sensitive (SS/JrHsdMcwi) rat of the Medical College of Wisconsin (MCW) colony. The consomic strains were generated and initially characterized as part of PhysGen, one of the National Heart, Lung, and Blood Institute (NHLBI) Programs for Genomic Applications (http://pga.mcw.edu). The chromosomal substitution was achieved through breeding as described previously (4–6). The resulting consomic rat strains contained homozygous SS alleles on every chromosome, as confirmed by total genome scans, except the substituted chromosome, which was homozygous BN as confirmed by a scan at a higher density. The consomic lines were maintained by brother-sister mating.

Phenotypic characterization.

Arterial blood pressure and urinary excretion of protein, and their responses to high-salt intake, were characterized in 8–17 male rats of each of the four strains, SS, SS-13BN, SS-18BN, and SS-20BN. All rats were maintained on a 0.4% NaCl diet (Dyets, Bethlehem, PA) from birth. The measurements were made in 10-wk-old rats and repeated over a period of 2 wk after the rats were switched to a 4% NaCl diet (Dyets). Arterial blood pressure was monitored daily from 9:00 AM to noon with a femoral arterial catheter, and urine samples were collected with metabolic cages as described previously (4, 23, 26).

Dietary protocol and sample preparation for microarray experiment.

A total of 128 rats, 32 for each of the four strains, were used for the gene expression study. Dietary treatment, tissue collection, and RNA extraction were performed as described previously (14, 15). Tissues were harvested from rats on the 0.4% NaCl diet or at 16 h (overnight), 3 days, or 2 wk after the rats were switched to the 4% NaCl diet. A total of 128 individual RNA samples were generated for each of the two tissues, the renal cortex and the renal medulla (4 strains × 4 time points per strain × 8 rats per group = 128 samples). Aliquots from every two rats were pooled. The pooled samples, four per group, were used for microarray experiments. The original individual samples, eight per group, were used in real-time PCR experiments.

Design of microarray experiment.

Samples were paired for cohybridization with a two-color hybridization method following a treelike design summarized in Fig. 1. Each cohybridization was repeated with dye switching (16), resulting in the use of 8 microarrays for each comparison and a total of 240 microarrays in this study (8 microarrays per comparison × 15 comparisons × 2 tissues = 240 microarrays).

Fig. 1.

Design of the microarray experiment. Each double-headed arrow represents a comparison between 2 groups of rats that was examined by 8 microarrays. SS, Dahl salt-sensitive rat; BN Brown Norway rat; HS, 4% NaCl diet; LS, 0.4% NaCl diet; 13, SS-13BN rat; 18, SS-18BN rat; 20, SS-20BN rat; d, days; w, week.

cDNA microarray.

cDNA microarrays containing ∼28,000 rat ESTs were constructed and hybridized as described previously (18). Approximately two-thirds of the ESTs represent genes coding for proteins with some known function.

Microarray data analysis and bioinformatics.

Gene expression data were analyzed as described previously (15, 18) with several tools including GeneSpring and ArrayStat for statistical analysis, GeneCluster for generating the heat maps, and B-Course for the Bayesian analysis. B-Course (http://b-course.cs.helsinki.fi/obc/) considers pairwise conditional dependencies among discrete variables and finds a model with the highest level of Bayesian probability given the data. The expression level of each gene in the group of SSLS (SS on the 0.4% NaCl diet) was fixed at 0. The expression level in each of the other groups was expressed as log2 ratios over SSLS. The treelike design shown in Fig. 1 allowed every group to be traced back to SSLS and adjusted through simple addition or subtraction. The threshold of differential expression was P < 0.05 and absolute log2 ratio > 0.263 (corresponding to a 20% change) in the analysis of 0.4% NaCl groups. Time course analysis was performed with analysis of variance and genes identified with a false discovery rate of 10%.

Real-time PCR.

Real-time PCR analysis was performed with the original individual samples (n = 8) as described previously (10, 17, 25).

Statistics.

Nonmicroarray data were analyzed with analysis of variance and are reported as means ± SE. P < 0.05 was considered significant.

RESULTS

Attenuation of hypertension and albuminuria in SS rats by introgression of BN chromosome 13 or 18, but not 20.

Mean arterial blood pressure in SS and SS-20BN increased by 47 ± 4 and 42 ± 5 mmHg, respectively, after 2 wk of exposure to a diet containing 4% NaCl. There were no significant differences between SS and SS-20BN. Mean arterial blood pressure in SS-13BN and SS-18BN increased by 28 ± 6 and 25 ± 5 mmHg, respectively, which was significantly (P < 0.05) less than the increase in SS (Fig. 2, A and B). At shorter periods of exposure such as 3 days after the initiation of the 4% NaCl diet, mean arterial blood pressure was significantly elevated in SS (P < 0.05) and nearly significantly elevated in SS-20BN (P = 0.08), but not elevated in SS-13BN or SS-18BN (Fig. 2, A and B).

Fig. 2.

Substitution of chromosome 13 or 18, but not 20, attenuated salt-induced hypertension (A and B) and albuminuria (C and D) in SS rats. B and D: differences between LS and HS 2 wk. MAP, mean arterial pressure. n = 17, 12, 8, and 9 for SS, SS-13BN, SS-18BN, and SS-20BN rats, respectively. Because of the large number of rats involved, the experiment had to be done in more than 1 batch. Several SS rats were included in each batch so that consomic strains were always analyzed in parallel with SS rats. *P < 0.05 vs. SS.

Renal injury, as reflected by urinary albumin excretion, followed a pattern similar to that of mean arterial pressure. The increase in urinary albumin excretion in response to the 4% NaCl diet was approximately fourfold greater in SS and SS-20BN than in SS-13BN and SS-18BN (Fig. 2, C and D). Changes in urinary total protein excretion also followed a similar pattern (data not shown).

We analyzed gene expression profiles in the renal cortex and the renal medulla in these four strains of rat (Fig. 1). We sought to answer two major questions. First, which genes might contain genetic variations contributing to Dahl salt-sensitive hypertension and albuminuria? Second, which genes might be involved in the broad biological pathways contributing to Dahl salt-sensitive hypertension and albuminuria? We focused the analysis of the expression data on three aspects: genes differentially expressed between each consomic strain and SS on the 0.4% NaCl diet, genes expressed in a unique manner in a protected consomic strain compared with all other three strains on the 0.4% NaCl diet, and genes that responded to the 4% NaCl diet in a way that mirrored the phenotypic segregation (i.e., SS-13BN and SS-18BN sharing one expression pattern, while SS and SS-20BN sharing another).

Genes differentially expressed between consomic strain and SS rats maintained on 0.4% salt diet were enriched for genes located on substituted chromosome.

In general, genetic differences should be reflected in changes in gene expression (abundance or other properties) if the genetic differences are to be physiologically relevant. Differentially expressed genes between SS rats and a consomic strain should include genes located on or affected by the substituted chromosome. One would expect this to be particularly true when the rats have not been experimentally challenged. Experimental challenges that cause substantial phenotypic divergence could result in overwhelming secondary changes in gene expression.

As shown in Fig. 3, the substituted chromosome in each consomic strain was overrepresented by 2- to 10-fold in genes differentially expressed between SS-13BN, SS-18BN, or SS-20BN and SS before the high-salt challenge. The overrepresentation was observed in both the renal cortex and the renal medulla. Similar overrepresentation was observed when either ESTs (Fig. 3) or nonredundant genes (data not shown) were analyzed.

Fig. 3.

Genes differentially expressed between a consomic strain and SS rats maintained on the 0.4% salt diet were enriched for genes located on the substituted chromosome. Representation index was calculated as (nchr/n)/(Nchr/N), where nchr is the number of differentially expressed expressed sequence tags (ESTs) located on a chromosome, n is the total number of differentially expressed ESTs with known chromosomal locations, Nchr is the number of ESTs located on a chromosome in the entire microarray, and N is the total number of ESTs on the array with known chromosomal locations. Of the ∼28,000 ESTs printed on the array, 24,876 (∼89%) could be mapped.

Genes or ESTs that were not only differentially expressed between a consomic strain and SS rats but also located on the substituted chromosome are shown in Fig. 4.

Fig. 4.

Genes differentially expressed between a consomic strain and SS rats on the 0.4% NaCl diet and located on the substituted chromosome. The data shown are means ± SE from 4 independent replicates. Gene symbols or GenBank accession numbers, when gene symbols are not available, are shown.

Genes with expression patterns unique to consomic strain.

One would expect a gene to be differentially expressed not only between a consomic strain and SS rats but also between the consomic strain and all other consomic strains if the gene's expression is entirely determined or controlled by the substituted chromosome.

Fifteen ESTs in the renal cortex and 39 in the renal medulla were found to be differentially expressed between SS-13BN and all of the other three strains (SS, SS-18BN, and SS-20BN) (Fig. 5). The differentially expressed genes included coagulation factor V (F5), serpin peptidase inhibitor, clade C, member 1 (Serpinc1), solute carrier family 19 member 2 (Slc19a2), and three other ESTs (representing unknown genes) located on chromosome 13.

Fig. 5.

Genes exhibiting expression patterns unique to SS-13BN on the 0.4% NaCl diet. A: renal cortex. B: renal medulla. The genes were differentially expressed between SS-13BN and SS, SS-18BN, and SS-20BN when the rats were on the 0.4% NaCl diet. Each column in the heat map represents an independent replicate, except for SS, which was fixed at 0. Data represented in the heat map are log2 ratio relative to SS (i.e., SSLS in Fig. 1).

Twelve ESTs in the renal cortex and 15 in the renal medulla were differentially expressed between SS-18BN and all of the other three strains (Fig. 6). The differentially expressed genes included 3-ketoacyl-CoA thiolase (Acaa2), β-1,4-galactosyltransferase 6 (B4galt6), collectin subfamily member 12 (Colec12), 17-β-hydroxysteroid dehydrogenase 4 (Hsd17b4), and five other ESTs (representing unknown genes) located on chromosome 18.

Fig. 6.

Genes exhibiting expression patterns unique to SS-18BN on the 0.4% NaCl diet. A: renal cortex. B: renal medulla. The genes were differentially expressed between SS-18BN and SS, SS-13BN, and SS-20BN when the rats were on the 0.4% NaCl diet. Each column in the heat map represents an independent replicate, except for SS, which was fixed at 0. Data represented in the heat map are log2 ratio relative to SS (i.e., SSLS in Fig. 1).

The genes exhibiting expression patterns unique to a consomic strain and located on the substituted chromosome likely contain genetic sequence differences between the SS and the BN alleles. For example, a comparison of the SS and BN strains reveals that the Serpinc1 gene harbors two single nucleotide polymorphisms (http://www.ensembl.org).

Chromosome 13 of the SS rat harbors a blood pressure quantitative trait locus (QTL) cluster, BpQTLcluster11, which spans nucleotide positions 57,419,350 to 97,977,732 and covers ∼36% of the chromosome (http://rgd.mcw.edu). Interestingly, of the six genes or ESTs located on chromosome 13 and exhibiting expression patterns unique to SS-13BN, five are located within BpQTLcluster11. They include F5, Serpinc1, Slc19a2, AI071832, and AA817956. Chromosome 18 of the SS rat also harbors a blood pressure QTL cluster, BpQTLcluster15, that spans nucleotide positions 62,486,497 to 79,558,280 and covers ∼20% of the chromosome. Of the nine genes or ESTs located on chromosome 18 and exhibiting expression patterns unique to SS-18BN, four are located within BpQTLcluster15. They include Acaa2, AI044306, AW142611, and AW917554.

Several of the genes described in the preceding paragraphs were found differentially expressed between the SS and the parental BN in the whole kidney in a previous study (18). They included F5, Slc19a2, AW916788, and AI071832 on chromosome 13 and Acaa2, Colec12, Hsd17b4, AW917554, and AI058545 on chromosome 18.

Overall, a substantial percentage of the genes differentially expressed in a kidney region between a consomic strain and SS on the 0.4% NaCl diet were found differentially expressed in the whole kidney between BN and SS. The percentages were 36%, 27%, and 33% for the cortex in SS-13BN, SS-18BN, and SS-20BN, respectively, and 18%, 16%, and 21% for the medulla. This is also consistent with the fact that the renal cortex is better represented in the whole kidney than the renal medulla.

Gene expression programs shared by SS-13BN and SS-18BN but distinct from both SS and SS-20BN.

SS-13BN and SS-18BN were protected from dietary salt-induced increases in blood pressure and albuminuria compared with SS and SS-20BN (see Fig. 1). Genes responding to the 4% NaCl diet in a way that mirrored the phenotypic separation would likely be involved in the development or progression of Dahl salt-sensitive hypertension and albuminuria.

We identified 184 and 346 ESTs in the renal cortex and the renal medulla, respectively, that responded to the 4% NaCl diet similarly in SS-13BN and SS-18BN rats but differently than SS and SS-20BN rats. These ESTs and their expression patterns are shown in Supplemental Figs. S1 and S2.1 Many of these array elements were up- or downregulated by the 4% NaCl diet in SS and SS-20BN rats but not changed or changed to a lesser extent in phenotypically protected SS-13BN and SS-18BN rats. Additional annotations for these array elements are available in Supplemental Data Sets S1 and S2.

Construction of regulatory networks based on both biological knowledge and probabilistic dependencies.

We performed both knowledge-driven and data-driven analyses to begin to extract information of regulatory networks from the identified genes. An EASE analysis (http://david.abcc.ncifcrf.gov/) indicated that several Gene Ontology terms were highly overrepresented in the genes identified in the renal medulla. These terms included extracellular matrix, basement membrane, protein tyrosine kinase signaling, and ion transport. Gene Ontology terms that were highly ranked for the renal cortical genes included wound response and inflammation, but the representations did not reach statistical significance after corrections.

Bayesian network analyses were performed to examine data-based, probabilistic dependencies among the identified genes. The analysis was performed for the phenotype-associated, salt-responsive genes (see Supplemental Figs. S1 and S2) with or without genes unique to a consomic strain (see Figs. 5 and 6). Each analysis evaluated up to 300,000 candidate models until additional evaluations did not produce a more probable model. The Bayesian networks are shown in Supplemental Fig. S3.

Alterations in extracellular matrix, basement membrane, protein tyrosine kinase signaling, and ion transport in the renal medulla could all be involved in biological pathways that contribute to the development of salt-induced hypertension and/or renal injury according to the current literature. The phenotype-associated, salt-responsive genes belonging to these groups are represented in Fig. 7. Other salt-responsive or strain-specific genes were added to Fig. 7 if their expression patterns exhibited highly probable dependencies with the genes that were already in Fig. 7.

Fig. 7.

A regulatory network based on both biological knowledge and probabilistic dependencies. The analysis started with renal medullary genes shared by SS-13BN and SS-18BN and belonging to 3 Gene Ontology categories that were highly overrepresented according to an EASE analysis. These genes are shown inside the dashed circles. Genes that exhibited a highly probable dependency (see Supplemental Fig. S3) with the genes in the circles were then added outside the circles. We considered a dependency in a Bayesian network “highly probable” if the removal of the dependency would make the model less likely by >100 times. Genes that were uniquely expressed in SS-13BN or SS-18BN are shown with square and arrow brackets, respectively. Red indicates that the gene was expressed at a higher level in SS compared with SS-13BN and/or SS-18BN. Blue indicates a lower level of expression in SS.

The result was a regulatory network that included genes probably involved in the development or progression of salt-induced hypertension and albuminuria. The regulatory network also included genes that might be coregulated with or contribute to the regulation of the genes in the pathways associated with hypertension and albuminuria.

Verification of microarray results.

The expression time course of nine genes, representing a broad range of fold change and level of statistical significance, was examined by quantitative real-time RT-PCR. All nine instances of differential expression between two strains over the time course identified by the microarray analysis were verified to be statistically significant according to real-time RT-PCR (Fig. 8).

Fig. 8.

Verification of microarray results with real-time PCR. The genes shown were considered differentially expressed between SS and SS-13BN, or between SS and SS-18BN, in the renal medulla over the time course. All of them were verified by real-time PCR to be significantly differentially expressed in the same direction as indicated by the microarray analysis (n = 8). See Fig. 1 for abbreviations.

DISCUSSION

Substitution of chromosome 13 or 18 from the BN rat partially corrected the enhanced blood pressure salt sensitivity in the SS rat. This suggests that chromosome 13 and 18 harbor genes involved in the determination or modulation of blood pressure salt sensitivity, and that one or more of those genes are defective in the SS rat. The fact that substitution of either chromosome was not sufficient to completely eliminate blood pressure salt sensitivity indicates that genes outside these chromosomes also contribute to the regulation of this phenotype.

While the salt-induced increase in blood pressure was reduced by ∼50% by the substitution of either chromosome, urinary albumin excretion was reduced by >75%. This suggests that albuminuria may be disproportionately more sensitive to large increases in arterial blood pressure. It is also possible that chromosomes 13 and 18 of the BN rat harbor genes that reduce the vulnerability of the kidneys to hypertension or high salt intake. Ongoing studies in our group are applying a chronic servo-control approach (19, 20) to study the effect of high perfusion pressure per se on renal injury in SS rats.

Common mechanisms by which a genetic defect leads to disease include altered abundance or functionality of the encoded mRNA and protein. A gene will be a good candidate for the hypertension phenotype in the SS rat if it is located in a substituted genomic region associated with attenuation of the disease, and at the same time differentially expressed between SS and the corresponding consomic or congenic strain. In this regard, it is interesting to see that the substituted chromosome was overrepresented in genes differentially expressed between SS and a consomic strain. These genes were identified in rats on the 0.4% NaCl diet and might represent an expression profile that was related to the inherent susceptibility to, instead of the consequence of, a high salt intake. A comparative study of only SS and a single consomic/congenic strain, however, is limited in its ability to rule out confounding, secondary effects of phenotypic differences on gene expression.

A feature in the design of the present study is the inclusion of the parental SS and three consomic strains, two of which (SS-13BN and SS-18BN) had a reduced blood pressure salt sensitivity while the third (SS-20BN) did not. This design allowed us to identify genes on the substituted chromosome that exhibited an expression pattern unique to the corresponding consomic strain and not shared by SS, SS-20BN, or the other similarly protected consomic strain. The differential expression of these genes would likely be the result of genetic characteristics of the substituted chromosome. As such, genes reported in Figs. 5 and 6 are particularly interesting candidate genes for hypertension and albuminuria in the SS model. The number of candidate genes nominated by the present study appeared to be small, which might reflect the high stringency of our selection criteria. It is noteworthy that several of these genes are located within blood pressure QTLs previously identified on the substituted chromosomes, supporting the possibility that these genes might contribute causally to the blood pressure phenotype in SS rats. Several studies of hypertension have previously utilized the approach that integrates QTL and transcriptome analyses (9, 12, 13, 27).

The same feature of the design also allowed us to identify genes likely contributing to the development of renal injury or secondarily contributing to the development of hypertension in the SS model. The logic was slightly different in that here we were interested in salt-responsive genes shared by both SS-13BN and SS-18BN, but distinct from SS and SS-20BN. The responses of these genes to the 4% NaCl diet (see Supplemental Figs. S1 and S2 and Supplemental Data Sets S1 and S2) mirrored the segregation of salt-induced hypertension and albuminuria among the four strains of rat. They are unlikely to be the direct consequence of the substitution of a specific chromosome. Instead, they might represent common biological pathways that are involved in the development of salt-induced hypertension and albuminuria.

It is interesting to note that the largest group of renal medullary genes that responded to salt similarly in SS and SS-20BN rats, but not in the two protected strains, was related to extracellular matrix formation. These genes generally exhibited progressively higher expression levels in SS and SS-20BN compared with SS-13BN and SS-18BN rats. This is consistent with the prominent fibrotic injury in the renal medulla of SS compared with SS-13BN rats (4). It is also consistent with the direct and significant effect of high perfusion pressure on renal interstitial fibrosis (19, 20).

Gene expression data can be analyzed with knowledge- or data-driven approaches. The knowledge-driven approach, such as displaying genes within known biological pathways, is intuitive to biologists but limited by current knowledge. The data-driven approach, such as clustering analyses, generates new clues for further studies but can be difficult to make sense of biologically. We merged the two approaches in the present study. We used the EASE analysis to identify highly overrepresented biological themes in the differentially expressed genes and the Bayesian network analysis to identify highly probable dependencies between genes. Known biological pathways, such as extracellular matrix formation, were connected to the phenotypes of interest. Meanwhile, the specific genes in those pathways and the probable dependencies between those genes and others provided abundant new possibilities for further study.

An interesting example is Col4a1, or collagen IV α1. Col4a1 is a key component of the basement membrane. Our analysis indicated that Col4a1 was upregulated in SS rats. Moreover, the expression pattern of Col4a1 exhibited highly probable dependencies with Tgfb1i4, or transforming growth factor β1-induced transcript 4. It suggests that Col4a1 and perhaps the other genes connected to Col4a1 in Fig. 7 could also be regulated by transforming growth factor β1 in the kidneys of the SS model, an interesting possibility to test in future studies. We recently confirmed that Col4a1 mRNA and protein were upregulated in hypertensive SS kidneys (20).

It should be pointed out that overrepresentation of a Gene Ontology term suggests, but does not prove, that the term is functionally relevant to the disease of interest. Biological pathways are often controlled by rate-limiting steps rather than the summation of all steps. Overrepresentation of genes in a pathway or a functional category only increases the chance that the rate-limiting steps are altered.

There are several interesting directions for future studies. First, the functional significance of the genes or pathways identified in the present study needs to be investigated in the SS model. Differential expression of a gene suggests, but not proves, that the gene might be pathophysiologically significant. The time course analysis provides useful information about temporal relationships between expression and phenotypes, but any cause-effect relationships suggested by the expression study would have to be validated. Second, it is conceivable that many additional new insights into the SS model or the regulation of gene expression in general could be obtained by further analyzing this large data set. The combination of four inbred rat strains, two tissues, four time points, and eight replicate arrays in each comparison makes this study particularly appealing for meta-analysis. Moreover, future analyses of consomic strains containing multiple introgressed chromosomes (e.g., SS-13BN/18BN) may help further refine and delineate our molecular networks. Finally, one needs to be cautious in extrapolating mRNA levels to protein abundance. Quantitative proteomic analysis will be a valuable complement. Proteins can be regulated posttranscriptionally by several mechanisms including microRNA (24). In addition, we have found in a proteomic study evidence for differential posttranslational modifications between SS and SS-13BN rats (25). Our ultimate goal is to construct the network of genes and pathways that underlie blood pressure regulation. The approach in the present study is one, but certainly not the only, approach that could contribute to achieving the ultimate goal.

GRANTS

The study was supported by NHLBI Grants HL-066579, HL-054998, HL-029587, HL-082798 (A. W. Cowley, Jr.), HL-066619, HL-075414 (N. H. Lee), and HL-077263 (M. Liang).

Acknowledgments

We thank Gina Tadisch, Jolean Morrison, Dani Didier, and Jason Evans for excellent technical assistance.

Present addresses: T. V. Luu and B. C. Frank, Department of Physiology and Pharmacology, George Washington University Medical Center, Washington, DC 20037; A. E. Kwitek, Department of Internal Medicine, University of Iowa Hospitals and Clinics, Iowa City, IA 52242.

Footnotes

  • * M. Liang, N. H. Lee, and H. Wang contributed equally to this study.

  • 1 The online version of this article contains supplemental material.

  • Address for reprint requests and other correspondence: M. Liang, Dept. of Physiology, Medical College of Wisconsin, Milwaukee, WI 53226 (e-mail: mliang{at}mcw.edu); N. H. Lee, Dept. of Physiology and Pharmacology, George Washington Univ. Medical Center, Washington, DC 20037 (e-mail: phmnhl{at}gwumc.edu).

    The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

REFERENCES

View Abstract