The internal vessel circumference of arteries taken from both ACTH-treated rats and SHR showed a trend towards a decrease, compared with WKY, although only saphenous arteries from SHR exhibited hypertrophy of the smooth muscle cell layers (Figure 1, Table 1). In the endothelium, there was a significant reduction in the area of endothelial cells in SHR, while area was significantly increased during ACTH-treatment (Table 1). There was no significant difference between experimental groups in the density of myoendothelial gap junctions per endothelial cell, which were rarely encountered in either WKY, SHR or ACTH-treated rats (Table 1).

The extent of RNA degradation was assessed by comparing the MAS5-derived 3'/5' and 3'/M' (middle) ratios of the housekeeping genes, GAPDH and hexokinase sample (Table 2). The 3'/5' ratios for GAPDH for all 12 samples were in the range 3–7, which is acceptable for doubly-amplified RNA, while the 3'/M' ratios for GAPDH were less than 2. Similar data were found for the 3'/5' and 3'/M ratios for hexokinase although some of the ratios tended to be greater than those for GAPDH, perhaps due to the longer hexokinase transcripts (3653 bp, versus 1268 bp for GAPDH).

Of the 15,878 probesets on each array, 9098 were present in all replicates of at least one experimental group (Affymetrix MAS5 Statistical Algorithm) and considered for further analysis. One-way ANOVA on each of the 9098 probesets showed significant differences in expression of 870 genes amongst the 3 experimental groups (F-test p value cutoff of 0.02) or 136 genes (p value cutoff of 0.001). The data discussed in this publication have been deposited in NCBIs Gene Expression Omnibus (GEO) at the following website 25 and are accessible through GEO Series accession number GSE8051.

The 50 most significant probesets found by ANOVA were superimposed on a GE-biplot of the 9098 probesets present in all arrays of at least one experimental group (Figure 2A). The locations of the probesets are shown as red asterisks with the 50 most significant genes indicated. The locations of the arrays are connected to the centre of the plot and labelled by their respective names. Figure 2A shows that the replicates within each experimental group are clustered together relative to the other groups, indicating consistency in measurement. In this genome-wide view the three experimental groups, WKY, ACTH and SHR groups are well separated, providing evidence of overall transcriptional differences between the groups.

The genes separating from the dense group in the centre of Figure 2A have relatively high variance. The labelled genes have a statistically significant difference in their means in the three groups. Generally the statistically significant genes on the left of the plot have relatively low expressions in SHR but high and approximately equal expression in WKY and ACTH. Those statistically significant genes on the right of the centre have relatively high expression values in SHR and low and approximately equal expression in WKY and ACTH. These patterns in gene expression are the dominant genome-wide patterns. However there are a number of genes which are relatively highly expressed in ACTH (eg. 9, 31, 48) compared to WKY and SHR.

Eleven out of 12 genes (6 in SHR, 6 in ACTH) which showed significant variation from WKY in the microarrays were validated by quantitative PCR (Table 3). The one variant gene (adenylate cyclase3) was significantly different in the SHR by microarray, but found not to differ from WKY by quantitative PCR. In contrast, four of 14 genes (7 in SHR, 7 in ACTH) which did not show any significant difference in expression in the microarrays relative to WKY, were found to be significantly different to WKY when assayed using quantitative PCR (Table 3). Three of these genes were found to be decreased in expression relative to WKY and one gene increased in expression (Table 3). These data concord with a recent analysis of microarray data which showed a high false-negative rate, or low sensitivity, especially in the case of genes showing small fold-changes 26.

Genes that were identified as significantly altered in expression by ANOVA in eight multiple comparisons of SHR and ACTH relative to WKY are given in Additional Files 1, 2, 3, 4, 5, 6, 7, 8: Tables S1–8. Genes identified in six dual comparisons of either SHR or ACTH with WKY are given in Additional Files 9, 10, 11, 12, 13, 14: Tables S9–14. Gene names, gene descriptions, biological function, Affymetrix IDs, accession numbers, expression levels and p-values are provided. The biological association networks of selected genes for each model of hypertension are shown in Figure 3A,B.

Table 4 summarises the changes in genes which relate to a range of processes or pathways which could impact on blood pressure regulation. The majority of changes reflect the specific morphological and physiological alterations occurring in the two different types of hypertension. Saphenous arteries from SHR showed upregulation of genes associated with renin-angiotensin signalling through MAP kinase ERK1/2 pathway and were characterised by upregulation of growth promoters and down regulation of growth suppressors (Table 4). In contrast, arteries from ACTH-treated animals showed upregulation of the p38MAP kinase pathway, diverse effects on growth promoters and suppressors (Table 4) and prominent involvement of the transcription factor Jun (Figure 3B). Differences were also found in expression of genes involved in mitochondrial activity, lipid metabolism, calcium handling, adenylate cyclase, ion transport and protein kinase C. In contrast, coordinate upregulation in both forms of hypertension was found for caveolin-1 and its interaction and modification by the membrane-associated tyrosine kinase Fyn (Figure 3, blue shape), and coordinate downregulation of Rgs2 and Rgs5, the regulators of G protein signalling (Tables 3, 4).

Expression of caveolin-1 was found in the membranes of both the endothelial and smooth muscle cells of saphenous arteries of WKY rats (Figure 4A, D, arrows). Expression varied both within and between cells in the endothelium (Figure 4A) and also throughout the muscle layers (Figure 4D). An increase in caveolin-1 expression was found in both endothelial cells and smooth muscle cells of arteries taken from both SHR and ACTH, compared to WKY rats (Figure 4B, C, compare 4A; Figure 4E, F compare 4D). Paired confocal images of either whole mounts or sections were taken with the same voltage and pinhole settings so that comparisons could be made of the extent of staining. Preincubation of the antibody with the immunogen abolished staining in endothelial and smooth muscle cells (Figure 4G).

At the ultrastructural level, caveolae were found along the membranes of both endothelial and smooth muscle cells of arteries from WKY rats (Figure 5A, B), although there were fewer on the abluminal than luminal endothelial surface. While the incidence of caveolae was increased in both cell types in the SHR (Figure 5C, D), the differential distribution in the endothelium was still apparent. In arteries from ACTH rats, a differential distribution in the endothelium was less obvious as caveolae were found in abundance on both sides of the endothelial cells (Figure 5E). Caveolae were also more abundant throughout the muscle layers of ACTH-treated rats, than of WKY rats (Figure 5F, compare 5B).

Ultrastructural studies were conducted on a minimum of 3 arteries each from a different rat, as previously described 4849. WKY, ACTH and SHR rats were anaesthetized (i.p., 44 and 8 mg/kg, ketamine and xylazine, respectively) and perfused via the left ventricle with 0.1% BSA, 0.1% NaNO₃ and 10 U/ml heparin and fixed with 1% paraformaldehyde, 3% glutaraldehyde in 0.1 M sodium cacodylate buffer, with 10 mM betaine, 0.2 mM CaCl and 0.15 M sucrose, pH 7.4, using standard procedures 48. The first lateral branch of the saphenous artery diverging towards the knee was removed and processed for electron microscopy 4849. Serial transverse sections (~100 nm thick) totalling ~5 μm of vessel length were cut from each vessel using 3–4 animals/experimental group. Myoendothelial gap junctions connecting the endothelium to the inner layer of smooth muscle were counted and photographed at ×3,500 to ×60,000 on a Phillips 7100 transmission electron microscope. Vessel circumference at the internal elastic lamina, the number of smooth muscle cell layers and endothelial cell areas were determined for each vessel as previously described 4849. Statistical significance was tested using one-way ANOVA followed by Student's t-test with Bonferroni correction for multiple groups. A p-value < 0.05 was taken as significant. The incidence of caveolae was assessed in montages of the vessel wall using a minimum of 3 arteries, each from a different rat, for each rat group.

Rats were anaesthetized as above and perfused with 2% paraformaldehyde in 0.1 M sodium phosphate buffer. The lateral branch of the saphenous artery was removed from both hind legs. Arteries were processed either as whole mounts or were sectioned at 10 μm. Preparations were incubated overnight at room temperature with rabbit antibodies against human caveolin-1 (Santa Cruz; 1:200 for whole mounts, 1:1000 for sections) followed by 1 hour in Cy3 conjugated anti-rabbit IgG (1:100, Jackson Immunoresearch Laboratories). For whole mounts, optical confocal series were taken throughout the thickness of the endothelium and recombined to a single image. For tissue sections, optical confocal series were taken throughout the section thickness and recombined to a single image. All images were collected using the same pinhole and voltage settings so that comparisons could be made of the intensity and extent of staining in arteries from the different rat groups (3 animals/group). Specificity of the staining was tested by preincubation of the antibody with a ten-fold excess of the immunogen.

Rats were anesthetized with ether and killed by decapitation. The saphenous artery branch was removed, stripped of extraneous material and stored in RNAlater (Ambion). Three different samples were prepared for each experimental group, i.e. WKY, SHR and ACTH. The choice of three biological replicates was based on the constraints of limited availability of mRNA and the high costs of microarray experiments. This is a common choice in the microarray literature 50. Each sample, which comprised arteries from 2–3 animals, was snap frozen in liquid nitrogen and processed using the RNeasy Micro Kit (Qiagen), including 20 min DNAse treatment. Concentrations of RNA were measured using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies). RNA quality was monitored using RNA 6000 Pico LabChip kit (Agilent 2100 Bioanalyzer). RNA integrity numbers for the 9 samples ranged from 6.4 to 8.1 51.

Sample preparation followed the Affymetrix GeneChip Eukaryotic Small Sample Target Labeling Assay Version II available at the website 52. After first round amplification, the yields of unlabeled cRNA ranged from 2.8–7.8 μg. Following the second round amplification, 600 ng of unlabeled cRNA was used for the second round of cDNA synthesis. Double-stranded cDNA was synthesized using oligo(dT) 24-anchored T7 primer, Superscript III (Invitrogen) and 100 ng total RNA for the first strand followed by DNA polymerase I and T4 DNA polymerase for the second strand. The double-stranded cDNA was precipitated with ethanol, air-dried and used as template for the first round of in vitro transcription of cRNA with MEGAscript enzyme mix (Ambion) at 37°C for 7 h. The cRNA was purified (RNeasy, Qiagen) and used as template for the second round of double-stranded cDNA synthesis with random primers (Invitrogen). The double-stranded cDNA product of second round synthesis was purified by ethanol precipitation and used as template for the second round of in vitro transcription and labeling of cRNA, using the ENZO BioArray HighYield RNA Transcript Labeling kit (Affymetrix) according to the manufacturer's protocol. The yields of labelled cRNA ranged from 58–193.8 μg. All components were mixed, incubated at 37°C for 7 h, and the biotin-labeled cRNA (1 μg for each sample) was purified (RNeasy, Qiagen) and fragmented prior to hybridisation to Affymetrix GeneChip RAE230A arrays. Hybridization, washing and staining followed the Affymetrix GeneChip Expression Analysis protocols.

Probeset signal values were calculated using the Statistical Algorithm in the Affymetrix MAS5 software. Raw data were normalised by scaling the global mean method to a TGT value of 150. Quality of all arrays was assessed using MAS5 quality control metrics and software procedures available in R/Bioconductor (Bioconductor version 1.8.0 53) and at the following website 54. Based on these metrics, the quality of all arrays was considered to be satisfactory (see Additional file 15).

One-way ANOVA was carried out on the log₂ transformed MAS5 signal values for each of the 9098 probesets which were present in all replicates of at least one experimental group (Affymetrix MAS5 Statistical Algorithm). Differences between experimental groups were examined using t-tests (post-hoc contrasts).

An ordination method, the GE-biplot 55, based on the singular value decomposition of the data matrix of scaled log₂ signal values, was obtained using an R function available at the website 54. The GE-biplot is an unsupervised visualisation tool that facilitates the detection of subsets of genes associated with subsets of microarrays. It provides a genome-wide view of the data, showing genes and microarrays simultaneously. Co-regulated (correlated) genes are located close to each other, with genes having reversed response profiles (negatively correlated) being located on the opposite side of the plot. Genes lying further from the centre have higher variance. Genes which lie close to, or lie in the direction of, a group of samples are relatively upregulated in that group compared with other samples, while those lying in the opposite direction are relatively down-regulated in that group. In a GE-biplot, arrays that have approximately the same relative values of gene expression across probesets cluster together. Further details are described elsewhere 55.

A heat map of the MAS 5.0 signal values of the 50 most significant probesets as evaluated by one way ANOVA was constructed 56. Hierarchical clustering with complete linkage was used to order the arrays in the columns of the heat map and the rank order of the p-values from the ANOVA was used to order the columns.

Quantitative PCR was used to determine the copy numbers of mRNA for 13 genes in the WKY, SHR, and ACTH samples (n = 3 for each). Details of the genes, primers and product sizes are given in Table 5. Nine of the genes were selected on the basis that their expression was significantly altered in either SHR or ACTH and that they were of biological interest to the processes involved in hypertension. The three vascular connexins and eNOS, although unaltered in the microarrays, were chosen because of their significance to vascular function and potential involvement in hypertension 57. Validation of the microarray data was considered to be a significant (p < 0.05) change in normalised PCR copy number in the same direction, or lack of a change in normalised PCR copy number which corresponded to that found in the microarray data.

Messenger RNA of all 9 saphenous artery samples was reverse transcribed to cDNA (42°C 1 h, 50°C 1 h, 90°C 10 min) using oligo dT (500 ng.μl^-1, Invitrogen) primers and Superscript II (200 U.μl^-1, Invitrogen). For every sample, control reactions from which either reverse transcriptase or RNA was omitted were run in parallel. Each 25 μl PCR reaction mixture contained 1 μl of each primer (final concentration 800 nM), 2 μl of 12.5 mM dNTPs, 3 μl of 25 mM MgCl₂, 0.125 μl of AmpliTaq Gold (5 U. μl^-1), 2.5 μl of 10× SYBR Green buffer (SYBR Green Core Reagents kit, Applied Biosystems), 10.375 μl H₂O and 5 μl of template containing either 10 ng of cDNA or diluted plasmid DNA containing the relevant cloned PCR product. All samples were diluted in water containing tRNA (2 ng. μl^-1). All reactions were performed in duplicate (ABI Prism 7700 Sequence Detection System, ACRF Biomolecular Resource Facility) at 95°C for 10 min, and 40 cycles of 20 sec at 95°C, 20 sec at 65°C, 45 sec at 72°C. Every experiment included duplicate control reactions and the expression of all genes was determined in every sample. The integrity of each PCR reaction was checked using dissociation curve analysis after every reaction.

For each gene, the mRNA concentration was determined by comparing the fluorescent signal at threshold (equal to 10 × S.D. of baseline fluorescence) to that generated by a standard curve using serial dilutions of a known concentration of purified plasmid DNA containing the relevant cloned and sequenced PCR product. Sequences were obtained using an ABI 3730 sequencer (ACRF Biomolecular Resource Facility, JCSMR, ANU). The concentration of each mRNA was normalized by determining 18S rRNA concentration, using primers complementary to a region of the 5' end of the molecule and a standard curve using serial dilutions of plasmid DNA containing the cloned fragment of the 18S rRNA gene 29. Results were analyzed for statistical significance (p < 0.05) using ANOVA and Student's t-test.

To examine differences in the major pathways affected in SHR and ACTH-induced hypertension, genes that were identified by t-tests as significantly altered in expression in SHR or ACTH compared with WKY (at the 0.015 level) and that were significant in ANOVA at the 0.02 level, were imported into a demonstration version of PathwayStudio 4.0 (Ariadne Genomics, Rockville, MD). This software contains the proprietary ResNet 3.0 database of functional relationships between genes, proteins and small molecules automatically extracted from PubMed and scientific journals using the MedScan natural language processing software. Functional links between genes and proteins were also investigated and cross checked using publicly available gene and genome databases: Online Mendelian Inheritance in Man (OMIM) and Genomic Biology and software provided by NCBI 58, and the protein database Human Protein Reference Database (HPRD; 59) and software 60.