A total of 26,522 transcripts (73% of the reference list) were expressed above background and compared between segments. A total of 304 transcripts were differentially expressed (1.5 fold, p < 0.05) between supra and infrarenal aortic segments (53 upregulated within the suprarenal aorta and 251 upregulated within the infrarenal aorta). A full list of transcripts is given in Additional File 1 and examples of potentially relevant genes shown in Table 2. A number of these genes may be relevant to the predilection of the suprarenal aorta to aneurysm formation based on knowledge of their functional importance or their previous identification in human AAA association studies 9. Ucn3 for example, which was downregulated in the suprarenal aorta, has been demonstrated to antagonize angiotensin induced production of oxygen derived free radicals in vitro and play a vascular protective role in animal models 1112. Pik3ap1, Dapk1 and Cd59 have previously been identified to be upregulated within human AAA biopsies 9. Since vascular smooth muscle apoptosis and inflammation have been demonstrated to be important mechanisms in human AAA, upregulation of Dapk1 and Cd59 may be relevant to the predilection of the suprarenal aorta to aneurysm formation in ApoE^-/- mice 8. Perhaps more important to the predilection of the suprarenal aorta to aneurysm formation are the downregulation of a number of genes coding for structural extracellular matrix proteins including Lama3, Myo18b and the cytoskeletal protein Tuba8 (Table 2). Lama3 has been identified to be downregulated within human AAA biopsies 9.

Maximum aortic diameters measured in the aortic arch, thoracic, suprarenal and infrarenal aortas in saline controls and mice exposed to angiotensin II that did and did not develop aneurysms are shown in Table 3 for animals used in the microarray experiment (study 2). Aortic diameters were similar between non-aneurysmal angiotensin II perfused mice and saline controls. We used Volcano Plots to identify differentially expressed genes between groups using our pre-defined criteria, i.e. 2-fold differential expression and uncorrected p value < 0.05 by unpaired t-test. The numbers of transcripts differentially expressed in comparison between the 3 groups of mice are shown in Additional File 2. 531 transcripts were differentially expressed (two fold change, p < 0.05) between aortas of mice exposed to angiotensin II which developed aneurysms and those that did not develop aneurysms (aneurysm resistant) (446 up and 85 downregulated, a full list is given in Additional File 3). 1196 transcripts were differentially expressed (two fold change, p < 0.05) between aortas of mice exposed to angiotensin II which developed aneurysms and saline controls (749 up and 447 downregulated, Additional File 4). 654 transcripts were differentially expressed (two fold change, p < 0.05) between aortas of mice exposed to angiotensin II which did not develop aneurysms and saline controls (128 up and 526 downregulated, Additional File 5). To further analyse these results, we concentrated on the transcripts differentially expressed between the aortas of mice that had been exposed to angiotensin II and had or had not developed AAAs (n = 531 listed in Additional File 3). We considered that genes identified by this analysis would be most likely to represent those relevant to aneurysm formation or resistance rather than simply reflecting the effects of angiotensin II. We also wanted to incorporate some of the findings relevant to the saline control group as they enabled us to have a measure of the baseline level of expression of different genes. This group allowed us to better postulate genes upregulated in the AAA resistant aortas, and thus of potential protective value against AAA. We carried out hierachical clustering analysis incorporating expression levels from all three groups of mice for the 531 transcripts that were differentially expressed in angiotensin II treated mice with and without aneurysms. This analysis is illustrated in terms of a heatmap, with a gene tree showing clustering of individual transcripts, in Figure 1. Each line in the heatmap represents a transcript with mean upregulated (red) or downregulated (green) in expression within the three groups. No change in expression is indicated in black. Hierachical clustering analysis indicated four main patterns of transcript expression (Figure 1, patterns 1–4). Transcripts grouped in pattern 1 (104 transcripts) showed decreased expression in aneurysm resistant aortas relative to other groups and included genes such as Ccl5, and Mmp9 which have been noted to be upregulated in human AAA biopsies (Additional File 6) 9. Pattern 2 transcripts (325 transcripts) showed increased expression in aneurysms and included many genes associated with inflammation including matrix metalloproteinase encoding genes (e.g. Mmp2), genes encoding markers of inflammatory cells (Cd14, Cd68) and many cytokine and chemokine genes (e.g. Il1b, Il6, Ccl4, Ccl7, Ccl8, Ccl19, MMP2) (Table 4 shows a selected list and Additional File 7 a full list). Many of these genes, particularly those related to leukocyte chemotaxis, have previously been identified as upregulated within human AAA biopsies (highlighted in Table 4) 910. Pattern 3 transcripts (17 transcripts) were downregulated in the aortas of mice resistant to aneurysm formation. The difference between pattern 1 and pattern 3 transcripts is that in pattern 1, genes are most highly expressed in the AAA group, whereas pattern 3 genes are, in general, more highly expressed in the saline group. This can be clearly seen on the heatmap (Figure 1) where pattern 1 genes in the AAA group are mainly coloured red compared to red/black in saline group, whereas pattern 3 genes are red/black in the AAA group and red in the saline group. Although the clustering algorithim has separated these groups of genes, we believe that both pattern 1 and pattern 3 genes can be considered "pathogenic". Additional File 8 shows transcripts (out of those in pattern 3) that were associated with known genes and were significantly downregulated (2-fold, p < 0.05) with respect to both saline controls and mice that develop aneurysms. Two of these "pathogenic" pattern 3 genes, namely Apoc1 (Apolipoprotein C-I) and Stxbp2 (syntaxin binding protein 2), have been previously demonstrated to be upregulated within biopsies of human AAA 9. Pattern 4 transcripts (85 transcripts) were upregulated in mice resistant to aneurysm formation (examples shown in Table 5 and a full list in Additional File 9) and we considered that upregulation of these may be "protective" for aneurysm formation. It is interesting to note that a number of the genes identified in pattern 4 have been previously shown to be downregulated within human AAA biopsies supporting their potential protective role (see Table 5 for examples) 9.

Based on previous studies demonstrating the importance of inflammation in AAA we selected a number of inflammation associated markers to validate which were differentially expressed between aortic aneurysms and control aortas 9. Initially we confirmed the up-regulation of Il6, Ccl4, Ccl8 and MMP2 within aortic aneurysms using real time PCR (Table 6). These genes were highly upregulated in aortic aneurysms by comparison to aortas from both control groups in microarray studies and real time PCR (see Tables 4 and 6). We also confirmed the differential expression of the Tnfrsf11b (osteoprotegerin) and Tgfb1 (TGF-b1) genes across the three groups translated to differences in protein expression using samples from study 3. We measured the concentration of OPG and TGFb-1 in the suprarenal segments of angiotensin II treated mice that developed AAA (mean maximum diameter 2.67 ± 0.24, n = 9) and those that did not (mean maximum diameter 1.26 ± 0.25, n = 9, p < 0.01) and found that both proteins were expressed at higher concentrations in aortas of mice with aneurysms, p < 0.05 (Figure 2A). Using immunohistochemistry we demonstrated that OPG was found within the medial layers and remodeled adventitia of aortas with aneurysm, whereas little was detected in aneurysm-resistant aortas or saline controls (Figure 2B).

KEGG pathway analysis of transcripts upregulated in the AAA group compared to the no AAA group using Webgestalt, revealed that of the 302 known genes in the input list, 116 of these could be assigned to one or more KEGG pathways (Additional File 10). This pathway analysis revealed a significant enrichment for genes involved in leukocyte recruitment, transendothelial migration and B lymphocyte receptor signalling and identified a number of cell lineage specific genes associated with aneurysm (Table 7). Chemokine genes upregulated in aneurysms included Ccl21 and Ccl19 (which attract T cells and dendritic cells), Ccl7 (attracts monocytes), Cxcl9 (attracts T cells), Cxcl12 (attracts lymphocytes) and Cxcl5 (attracts neutrophils), suggesting that all these inflammatory cell types are recruited to mouse aortic aneurysms. Genes that can be considered as markers for the following cell types were upregulated in aneurysms (Additional Files 5 and 6): neutrophils (Csf3r); T lymphocytes (Cd3d) and B lymphocytes (Blnk, Cd72, various immunoglobulin genes); and monocyte/macrophages (Cd14, Cd68). We used IHC to confirm the recruitment of neutrophils, T and B lymphocytes, dendritic cells and monocyte/macrophages to aneurysmal but not aneurysm-resistant aortas (Figure 3).

Infusion of angiotensin II induces aortic dilatation particularly affecting the suprarenal aorta in ApoE^-/- mice 5618192021. Based on studies carried out in our and other laboratories the response to angiotensin II is variable, with some mice developing large aneurysms but other animals appearing resistant to aneurysm formation with aortic diameters similar to that of saline controls (Additional file 12). The infrarenal aorta is protected from aneurysm formation in these mice. To assess the likely signalling pathways relevant to aneurysm development and progression within this mouse model we carried out three studies (Table 1): 1) We compared RNA expression within segments of supra and infrarenal aortas from 13 week old male ApoE^-/- mice unexposed to angiotensin II (n = 10). 2) We compared RNA expression from whole aortas of 17 week old male ApoE^-/- mice which had been exposed to angiotensin II (1.44 μg/kg/min) for 4 weeks where there was clear evidence of aortic aneurysm formation (n = 5) with that of mice failing to develop aneurysms (n = 7) and those exposed to saline infusion (n = 6). 3) We selected 2 genes and 5 cellular pathways identified as upregulated within the aortas of mice with aneurysms for validation using protein assessments including ELISAs and immunohistochemistry (IHC). The selection of genes was based on their association with pathways we had identified from pathway analysis of the data and recognised ability to be able to assess them using ELISAs or IHC. For study 3, twenty-eight additional 17 week old male ApoE^-/- mice were infused with angiotensin II for 4 weeks. Ten mice in study 3 died prematurely due to aortic rupture and were excluded from further analysis in order to avoid post-mortem or post rupture effects confounding assessments, leaving 18 animals in this group.

This investigation conformed to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No. 85-23, revised 1996). Ethical approval was obtained from the local institutional committee prior to commencement of the study. Male ApoE^-/- mice were obtained from Animal Resources Centre, Canning Vale, Western Australia aged 6–8 weeks. Mice were fed standard chow, and at 13 weeks either their aortas were harvested (study 1, n = 10) or angiotensin II (n = 12 study 2; n = 28 study 3) or saline (n = 6 study 2) infusion commenced. Aortic harvesting and osmotic minipump placement (Model 2004, ALZET, Durect Corporation, Cupertino, California, USA) was carried out under ketamine (150 mg/kg i.p.) and xylazine (10 mg/kg i.p.) anaesthesia as previously described 22. The 28 day infusion period and dose of angiotensin II were selected as these have been routinely used in other studies within this model 5618192021. In studies 1 and 2, mice were euthanized by carbon dioxide asphyxiation, the aortas perfused with RNAlater^® (Qiagen, Doncaster, Victoria, Australia), and harvested from arch to iliac bifurcation and stored at -80°C for later analysis. For study 3 mice were euthanized, aortas were perfused with PBS and the suprarenal aortic segments divided into 2 and stored separately at -20°C for later IHC and cytokine assessment.

Aortas were placed on a black background and digitally photographed (Coolpix 4500, Nikon). Maximum diameters of the aortic arch, thoracic, suprarenal and infrarenal aorta were determined from the images using computer-aided analysis (Scion Image, Scion Corporation). Preliminary studies (n = 27) established that these measurements could be repeated with good intra-observer reproducibility (Coefficient of repeatability 0.98, 95% confidence intervals 0.975–0.982, and coefficient of variation 4%).

Total RNA was extracted from homogenised aortas using the TRI reagent (Sigma) and RNeasy Mini kits (Qiagen) and analysed on the HP 2100 Bioanalyzer (Agilent Technologies) for integrity using Eukaryotic total RNA nano chips (Agilent) (Additional File 11). All samples had RNA integrity scores between 6.9 and 8.1. For study 1, RNA extracted from the pooled suprarenal aortas of two mice was compared with that obtained from the infrarenal aortic segments of the same mice. A total of 10 mice (i.e. 5 pairs of pooled segments) were included in study 1. For study 2, RNA extracted from the whole aortas of mice with aneurysms (n = 5, Additional File 12A) was compared with that extracted from the whole of aortas of saline controls (n = 6, Additional File 12C) and mice exposed to angiotensin II which did not develop aneurysms (n = 7, Additional File 12B). No sample pooling was carried out in study 2. RNA hybridization was performed, and gene expression profiles determined, using CodeLink Mouse Whole Genome Bioarray chips (GE healthcare, Amersham, Bioscience) (study 1, n = 10) and Illumina Mouse Sentrix 6 version 1.1 Beadchips (study 2, n = 18). For study 1 total RNA from infrarenal and suprarenal aortas was supplied to GenUS BioSystems Inc. (Northbrook, Illinois USA) who conducted cDNA synthesis, hybridization and scanning using standard protocols. In brief, 10 μg of total RNA was used for each 'sample versus sample' comparison and first and second strand cDNA were prepared using the Codelink™ iExpress Expression assay Reagent kit (GE healthcare Amersham). Biotinylated cRNA target was prepared from the cDNA template by linear amplification using biotin-dNTPs and verified on a Bioanalyzer 2100. The cRNA was fragmented to uniform size and once again verified on the Bioanalyzer 2100. Each fragmented, biotin-labeled cRNA was added to a Codelink Mouse Whole Genome array which contained 34,957 probes targeting unique mouse transcripts. The bioarrays were washed, exposed to Cy5 streptavidin and scanned using GenePix 4000B laser scanner. Scanned image files were examined using CodeLink image and data analysis software (GE healthcare, Amersham, Bioscience). The image information was converted into spot intensity values using CodeLink™ Expression Analysis software (GE healthcare). The generated values were exported to GeneSpring GX 7.3.1 software (Silicon Genetic, USA) for further analysis. Study 2 labelling, hybridisation and scanning was done at the SRC Microarray Facility, University of Queensland, Brisbane, Australia according to the manufacturer's instructions. For study 2 total RNA was amplified in a single-round of in vitro transcription amplification that allowed incorporation of biotin-labeled nucleotides using the Illumina TotalPrep RNA amplification kit (Ambion, Inc., Austin, TX). mRNA samples were assessed for integrity and purity prior to hybridization using the Bioanalyser 2100 with mRNA Nano chips (Agilent Technologies). cRNA of each sample was hybridized to an Illumina Mouse WG-6 V1.1 BeadChip followed by washing, blocking, and streptavidin-Cy3 staining steps, and scanning with a high-resolution Illumina BeadArray reader scanner. The data extraction was performed by using Illumina Bead Studio V2.3.41 software with the output being raw, non-normalized bead summary values. The BeadStudio matrix contains the summarised expression values (Avg_Signal), standard error of the bead replicates (BEADSTDEV), number of beads used (Avg_NBEADS) and a detection score, which estimates the confidence limit of detection of a gene. The performance of the built-in controls that accompany each Illumina beadchip experiment was assessed as part of the Bead Studio V2 experiment performance report and was found to be satisfactory. Controls included Housekeeping controls for intactness of the biological specimens, negative controls to establish gene expression detection limits and hybridization controls including low and high stringency controls and biotin signal generation controls. The raw data matrix extracted from Beadstudio was uploaded into GeneSpring GX 7.3.1 (Silicon Genetics, Redwood City, CA) software for downstream analysis. Details of GeneSpring analysis including data transformations, per chip and per gene normalisations are described below in Analysis and Design.

Using RNA obtained for the mice employed in the micro-array investigations in study 2 we validated findings for 4 genes (Ccl4, Ccl8, Il6 and Mmp2) using real time PCR. RNA samples from mice exposed to angiotensin II that developed aneurysms (n = 5), mice exposed to angiotensin II that did not develop aneurysms (n = 7) and saline control mice (n = 6) were included in the analysis. The QuantiTect SYBR Green one-step RT-PCR Kit (Qiagen) and Quantitect Primer Assays (Qiagen) (Il6, QT00098875; Ccl4, QT00154616; Ccl8, QT00128548; Mmp2, QT00116116) were used according to the manufacturer's instructions with 25 ng of total RNA as template. Primers for mouse Gapdh were used to amplify the housekeeping gene (Qiagen, QT01658692). Standard curves for each gene including the housekeeping gene were constructed using duplicate sets of five (5) 10-fold serial dilutions of equal volumes of the pooled saline control RNAs. Negative and "minus-RT" controls were also included. All reactions were independently repeated in duplicate to ensure the reproducibility of the results. Cycling parameters were as follows: 50°C 30 min for RT; 95°C 15 min; 40 cycles of 94°C 15 sec, 55°C 30 sec, 72°C 30 sec. Data were viewed and analysed using the Rotor-Gene's real-time analysis software (Rotor-Gene 6000; Corbett Life Science, Sydney). The relative expression of the gene of interest in each sample was calculated by the Rotor-Gene software using the concentration-Ct-standard curve method and normalised using the average expression of Gapdh for each sample. SSPS statistical software was used to calculate median and interquartile ranges for expression of each gene of interest in each group.

Serial cryostat sections 7 μm thick were cut from suprarenal aortas with and without aneurysms prior to staining for inflammatory cells. Serial frozen sections were air-dried, fixed in acetone for 10 min at -20°C, air dried and rehydrated with PBS before being incubated in 3% H₂O₂/0.1% sodium azide/PBS to block endogenous peroxidase. For macrophage detection, sections were blocked in 2% normal goat serum in PBS followed by staining using pan-macrophage antibody (clone MOMA-2, Abcam AB33451), and goat anti-rat HRP (Millipore AP136P). Rat IgG (Sigma I4131) was used as isotype control. Detection of osteoprotegerin (OPG) was carried out using biotinylated goat anti-mouse OPG (R&D BAF459) diluted to 1 μg/ml in blocking buffer followed by Vectastain Elite ABC-HRP (Vector Laboratories). Biotinylated goat-IgG was used as isotype control (Vector Laboratories VEBI1001). T cells (CD3e FITC clone 145-2C11, BD Biosciences 553061), B cells (CD45R/B220 biotin clone RA3-6B2, BD Biosciences 550385), dendritic cells (CD11c biotin clone HL3, BD Biosciences 553800) and neutrophils (Ly6G/Ly6C(Gr-1) FITC clone RB6-8C5, BD Biosciences 553127) were demonstrated using either anti-FITC HRP (Invitrogen), biotinyl-tyramide (Perkin Elmer), SA-HRP (Perkin Elmer) for FITC-conjugated antibodies or SA-HRP, biotinyl-tyramide, SA-HRP for biotinylated antibodies. Appropriate isotype controls were added to other sections (all BD Biosciences). Slides were incubated in the peroxidase substrate 3, 3'-diamminobenzidine (ImmPACT DAB, Vector), counterstained in Mayer's Haematoxylin, dehydrated, cleared in xylene and mounted in Depex mounting medium. Sections were photographed using a Nikon Eclipse 50i microscope, Digital Sight camera and NIS-elements software.

Protein was extracted from individual frozen suprarenal aortic segments by homogenising in buffer (10 mM cacodylic acid, 60 mM L-arginine, 0.25% triton x-100 in PBS, pH 7.2) and centrifuging at 18,000 × g at 4°C for 20 min. Supernatant protein was quantified by the Bradford technique (Protein Assay, Bio-Rad, Hercules, California, USA). Concentrations of OPG and transforming growth factor beta-1 (TGFb-1) were assessed using commercial ELISAs (Quantikine, R&D Systems, MOP00 for OPG, MB100B for TGFb-1) and expressed as pg/mg of protein. Initial tests were performed to determine an appropriate amount of extracted aortic protein to load in each well to suit the ELISA standard curves (10 μg for OPG ELISA, 20 μg for TGFb-1 ELISA). Serum and extracted aortic proteins from OPG deficient mice were confirmed as having zero readings using the OPG ELISA (data not shown). Batch analysis of the samples was carried out to facilitate comparison between groups. We have previously reported excellent reproducibility of similar assays 36.

We designed the current study as an exploratory analysis expecting a minimal level of false discovery inevitable in analysis of large microarray gene probes. Sample size calculations were conducted using SPCalc 3738. For study 1 (which included paired samples) we aimed to detect a 1.5-fold difference at a power of 0.8 (alpha uncorrected 0.05), while for study 2 (which included unpaired samples) we aimed to detect a 2-fold difference. Based on the number of gene probes to be examined (26,522) we wished to control the false positives to a mean of 2 1718. We estimated the total number of arrays needed for study 1 and 2 were 10 and 18 respectively. Microarray data were imported into Genespring GX 7.3.1 (Agilent) for analysis. The data from the individual bioarrays have been deposited in NCBIs Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO accession number GSE7006 (

GEO Accession viewer