18F-FDG PET Imaging of Murine Atherosclerosis: Association with Gene Expression of Key Molecular Markers

Aim To study whether 18F-FDG can be used for in vivo imaging of atherogenesis by examining the correlation between 18F-FDG uptake and gene expression of key molecular markers of atherosclerosis in apoE−/− mice. Methods Nine groups of apoE−/− mice were given normal chow or high-fat diet. At different time-points, 18F-FDG PET/contrast-enhanced CT scans were performed on dedicated animal scanners. After scans, animals were euthanized, aortas removed, gamma counted, RNA extracted from the tissue, and gene expression of chemo (C-X-C motif) ligand 1 (CXCL-1), monocyte chemoattractant protein (MCP)-1, vascular cell adhesion molecule (VCAM)-1, cluster of differentiation molecule (CD)-68, osteopontin (OPN), lectin-like oxidized LDL-receptor (LOX)-1, hypoxia-inducible factor (HIF)-1α, HIF-2α, vascular endothelial growth factor A (VEGF), and tissue factor (TF) was measured by means of qPCR. Results The uptake of 18F-FDG increased over time in the groups of mice receiving high-fat diet measured by PET and ex vivo gamma counting. The gene expression of all examined markers of atherosclerosis correlated significantly with 18F-FDG uptake. The strongest correlation was seen with TF and CD68 (p<0.001). A multivariate analysis showed CD68, OPN, TF, and VCAM-1 to be the most important contributors to the uptake of 18F-FDG. Together they could explain 60% of the 18F-FDG uptake. Conclusion We have demonstrated that 18F-FDG can be used to follow the progression of atherosclerosis in apoE−/− mice. The gene expression of ten molecular markers representing different molecular processes important for atherosclerosis was shown to correlate with the uptake of 18F-FDG. Especially, the gene expressions of CD68, OPN, TF, and VCAM-1 were strong predictors for the uptake.


Introduction
The use of positron emission tomography (PET) for visualization of atherosclerosis has been evolving over the last decade. The visualization of the vulnerable plaque using the tracer 18 F-FDG is promising [1][2][3][4], and also 18 F-FDG uptake as a surrogate marker for atherosclerotic disease activity shows potential [5,6]. The preclinical in vivo research of 18 F-FDG has mainly focused on rabbits [2,7,8]. As transgenic mouse models have shown their value in atherosclerosis research we have focused on developing the technique of small animal PET for in vivo imaging of atherosclerosis in mice. A study published in 2011 [9] used 18 F-FDG for in vivo imaging of mice and their results suggest that the method can be used to follow the development of atherosclerosis in murine models.
The aim of the study was to evaluate the uptake of 18 F-FDG in the aorta of apolipoprotein E knockout (apoE 2/2 ) mice and to correlate the tracer uptake with gene expression of the molecular markers mentioned above in order to test the hypothesis that 18 F-FDG can be used for in vivo imaging of key atherosclerotic processes.

Ethical Statement
All care and all experimental procedures were performed under the approval of the Animal Experiments Inspectorate in Denmark (permit number 2011/561- 14). All efforts were made to minimize suffering.
The mice were housed under controlled humidity, temperature, and light cycle conditions, and had free access to food and water throughout the course of experiments.
The mice were divided into nine groups. The characteristics of the groups are shown in Table 1. All animals were scanned once and then sacrificed. One group was scanned and sacrificed at the beginning of the experiment as a baseline group (0 weeks). Four other groups received normal chow for 8, 16, 24 or 32 weeks (8 weeks, 16 weeks, 24 weeks or 32 weeks) before scanning and sacrifice. The last four groups received a high-fat Western type diet for 8, 16, 24 or 32 weeks (8 weeks+diet, 16 weeks+diet, 24 weeks+diet or 32 weeks+diet). The high-fat Western type diet contained 21% fat and 0.21% cholesterol (diet #TD12079B, Research Diets, Inc., USA).

Experimental Protocol
The animals were fasted overnight prior to the scan [12]. Before injection, 30 minutes after injection and during the scans, the animals were anaesthetized using a mixture of 3% sevoflurane (Abbott Scandinavia AB, Sweden) mixed with 35% O 2 in N 2 by breathing through a nose cone. The mice were kept at a temperature of approximately 32uC from the time of the injection to the scans were executed. 18 F-FDG was obtained from our own production facilities (Rigshospitalet, Denmark). The exact concentration of the 18 F-FDG solution was measured in a Radioisotope Calibrator ARC-120 (Amersham, United Kingdom). 20.164.8 MBq in 0.3 mL physiological saline was administered i.v. (slow injection over several minutes) to the mice in a lateral vein using a vein catheter (BD Vasculon TM Plus, Becton Dickinson A/S, Denmark). Immediately after this, 0.2-0.3 mL of a long circulating emulsion formulation containing an iodinated triglyceride (Fenestra VCH, ART Advanced Research Technologies Inc., Canada) was administered through the same vein catheter. The mice remained anaesthetized for approximately 30 minutes after the injection to limit the up-take of 18 F-FDG in brown fat [12].
Three hours after injection, the animals were placed in a prone position on the acquisition bed and a 30 minutes PET scan was acquired, followed by a CT scan. The same acquisition bed was used for both scans, so the animals remained in precisely the same position during both scans. The animals were then sacrificed by decapitation. The blood was collected and centrifuged (3,200 RPM for 10 minutes) and plasma was transferred to a fresh tube and store at 220uC. The aorta was removed with care taken not to include any surrounding tissue and placed in RNAlaterH (Ambion Europe Limited, United Kingdom). Subsequently, the aorta was gamma counted and stored at 4uC. The following day, RNAlaterH was removed and the samples stored at 280uC until RNA extraction.

CT Protocol
CT data were acquired with a MicroCAT II tomography (Siemens Medical Solutions, USA). The X-ray tube with a 0.5 mm aluminium filter was set at 60 kVp, a tube current of 500 mA, and an exposure time of 310 ms per projection. A set of 360 projections was used for a full 360 scan. Images were reconstructed using the COBRA real-time reconstruction with the Sheep-Logan filter. The voxel size was 0.09560.09560.095 mm 3 .

PET Protocol
PET data were acquired with a microPET Focus 120 scanner (Siemens Medical Solutions). The energy window of the emission scan was set to 350-650 keV with a time resolution of 6 ns. The acquired emission dataset was automatically stored in listmode. All listmode data were post-processed into 1286144695 sinograms using a span of three and a ring difference of 47. The sinograms were then reconstructed using 3D Maximum a posteriori (MAP) algorithms [13]. The voxel size was 0.360.360.3 mm 3 and in the centre of field of view, the resolution was 1.2 mm full-width at half-maximum. The emission data were normalized and decay and dead time corrected. The system was calibrated to provide the quantification unity in becqerel per millilitre. 18 F-FDG uptake in the aorta was quantified for each animal using the image analysis software Inveon Research Workplace (Siemens Medical Solutions). Image fusion was achieved by doing rigid registration (feature of the Inveon Research Workplace software) and subsequently confirmed visually on basis of three distinct fiducial markers placed directly on the animal bed within the field of view but $5 mm from the regions of interest (ROIs). In Figure 1, a CT, a fused PET/CT, and a PET image of a mouse is shown. The 3D ROIs were drawn on the axial slices from the CT scan intending to cover the aorta, from the heart to the kidney arteries. The aorta appears white at the CT image ( Figure 2a) due to the contrast agent. In order to include the vessel wall, the ROIs were on average drawn with a radial increase of 0.13 mm (Figure 2b). About 70 slices (every fifth slice) per mouse were drawn to cover the aorta followed by ROI interpolation ( Figure 2C). During subsequent analysis, standardized uptake values (SUVs) were calculated. The mean value of the radioactivity in the ROIs was used i.e. SUV mean .

Gamma Counting
The tubes with the aortas were placed in the 2480 Automatic Gamma Counter, Wizard2TM3'' (Perkin Elmer, USA). The samples were counted for 120 seconds using a designated 18 Fprotocol. The counting efficiency of the gamma counter was tested to be 0.54. During subsequent analysis, SUVs were calculated.

RNA Extraction and Reverse Transcription
Total RNA was isolated with TRI ReagentH in accordance with the protocol of the manufacturer (Molecular Research Center Inc., USA). The quality of the isolated RNA was tested using the Agilent 2100 Bioanalyzer in conjunction with the Agilent RNA 6000 Nano Kits (Agilent Technologies Denmark A/S, Denmark). A RNA integrity number (RIN)-value above 5 was accepted [14]. The quantity of RNA was measured using the NanoDrop 1000 (Thermo Fischer Scientific, USA).
Total RNA (0.3 mg) was reversed transcribed (RT) using the AffinityScript TM QPCR cDNA Synthesis Kit (Stratagene, USA, cat.no. 600559) according to the protocol of the manufacturer. The RT was performed using the Eppendorf Mastercycler Gradient (Eppendorf AG, Germany) with the following protocol: incubation at 25uC for 5 minutes (primer annealing), 42uC for 15 minutes (cDNA synthesis) and 95uC for 5 minutes (termination of cDNA synthesis). Immediately after, the samples were cooled down and stored at 220uC.

Identifying the Optimal Reference Genes
The optimal reference genes for the study were selected from a panel of twelve common endogenous control genes (prefabricated panel of primer-mixes from TATAA Biocenter, Sweden). All candidate genes were tested by quantitative real-time PCR (qPCR) in tissue representing non-atherosclerotic and atherosclerotic mice. To analyze gene expression stability, the software geNorm [15] was used. Actin beta (ACTB), beta-2 microglobulin (B2M) and glucuronidase beta (GUSB) were found to be the optimal reference genes for normalization.

Quantitative Real-time PCR
Gene expression was quantified on the Mx3000PH and Mx3005 TM real-time PCR systems (Stratagene). Primers and dual-labelled hydrolysis probes for the genes of interest and references genes were designed using Beacon Designer TM (PRE-MIER Biosoft, USA). The design included a BLAST search for test of sequence homology, a test for secondary structures and optimization of multiplex setup. The genes, accession number, primers, probes and amplicon lengths are listed in Table 2. All primers and probes were purchased from Sigma-Aldrich (Sigma-Aldrich Danmark A/S, Denmark). For each gene, the optimal primer and probe concentration was established. All samples were run in duplicates for genes of interest and reference genes using 1 ml of cDNA.
The Brilliant Multiplex QPCR Master Mix (Stratagene, cat.no. 600553) was used. The thermal profile employed was 10 minutes of denaturation at 95uC followed by 40 cycles with denaturation for 15 s at 95uC and annealing/elongation at 60uC for 1 minute.
The quantification and normalization of results were based on the computation of target quantification cycle (Cq) values and reference gene Cq values in the qbase PLUS software (Biogazelle NV, Belgium). The method of qbase PLUS is a modification of the classic delta-delta-Ct method [16]. Instead of just normalizing to a single reference gene, multiple reference genes are taken into account and correction for gene specific amplification efficiencies can be done. The three reference genes, ACTB, B2M and GUSB, were included in the analysis. The reference target stability was 0.63 (M-value) and 0.27 (CV-value). This is a little higher than what is expected from homogeneous sample panels, however, as these samples include various cell types they are classified as heterogeneous and higher M-and C-values are expected. One default amplification efficiency (100%) for all targets was used, as the amplification efficiency varied little over the multiple runs. The data were reported as calibrated normalized relative quantities (CNRQs). As all the samples could not be included in a single run, inter-run calibration was done to correct for run-to-run differences. This was based on three inter-run calibrators included in all runs.

Plasma Measurements
Cholesterol level was measured in duplicate using the kit Cholesterol Chod-Pap (Roche Diagnostics GmbH, Germany) in accordance with the protocol of the manufacturer. Calibrator (C.f.a.s from Roche Diagnostics GmbH) and controls (Wako Control Serum I and II from Wako Chemicals GmbH, Germany) were included in the analysis. The intra-and interassay coefficients of variation were 3.6% and 6.0%, respectively.

Statistical Analysis
All statistical analyses were performed using SPSS 20.0 (IBM Corp., USA). Graphs have been constructed using GraphPad Prism version 5.0c for Mac OS X (GraphPad Software Inc., USA). All results were log-transformed to obtain Gaussian distribution as confirmed by one-sample Kolmogorov-Smirnov test. Comparisons between the different groups were performed by Student's t-test for independent samples and Bonferroni correction of p-values was applied by multiplying the acquired p-values. Univariate linear regression was performed between the molecular markers (gene expression) and SUV mean -values and p-values were Bonferroni corrected. Markers with significant correlation (R) were subsequently included in a multivariate linear regression model with stepwise backward elimination of the least significant marker. Data are reported as mean6SEM (standard error of mean) unless otherwise indicated and p,0.05 was considered statistically significant.

Uptake of 18 F-FDG in the Vessel Wall
The uptake of 18 F-FDG measured using PET and gamma counting, respectively, is shown in Figures 3a and 3b.
The uptake of 18 F-FDG measured using PET (Figure 3a) was not significantly different in the groups receiving normal chow for 24 and 32 weeks compared to the 0 weeks group. However, the 8 and 16 weeks groups showed a small decrease in the uptake compared to the 0 weeks group with a fold change of 0.84 (p = 0.013) and 0.78 (p = 0.0012), respectively.
The high-fat Western diet had a marked effect upon the uptake of 18 F-FDG measured by PET and from 16 weeks, SUV mean was significantly higher compared to 0 weeks group. The fold change was 1.73 after 16 weeks (p = 0.0011), 1.82 after 24 weeks (p,0.001) and 2.23 after 32 weeks (p,0.001) compared to 0 weeks. When comparing mice on high-fat Western diet with mice on normal chow of the same age, a significant higher 18 F-FDG uptake measured by PET was seen after 16 weeks of dieting compared to non-dieting (2.21 fold; p,0.001). The 18 F-FDG uptake was 2.02 fold higher at 24 weeks on diet compared to non-diet (p,0.001). At 32 weeks, a 2.29 fold higher level was seen (p,0.001).
The 18 F-FDG uptake measured by gamma counting showed the same pattern (Figure 3b) as measured by PET, except, the uptake did not differ significantly in the chow-fed groups compared to the 0 weeks group. The high-fat Western diet had a marked effect upon the 18 F-FDG uptake measured by gamma counting from 16 weeks. The fold changes were 2.19, 2.72, and 3.65 (p,0.001) for 16, 24, and 32 weeks, respectively, compared to the 0 weeks group.
The comparison of 18 F-FDG uptake measured by gamma counting of mice on high-fat Western diet to the mice of the same age on chow also showed a significant increase from 16 weeks. The 18 F-FDG uptake was 2.47 fold higher at 16 weeks, 2.83 fold at 24 weeks, and at 32 weeks the increase was 3.06 fold (p,0.001).

Correlation of PET and Gamma Counting
A correlation plot of PET and gamma counter data is shown in Figure 4.
A strong correlation between the two methods was seen with a R-value of 0.88 (p,0.001).   In Table 3 the R-values and their p-values are listed for all gene expression markers.
The gene expression of all markers of monocyte/macrophage recruitment exhibited significant correlations with 18 F-FDG uptake. All correlations were positive. VCAM-1 had the highest correlation (R = 0.61, p,0.001). The same was true for the next group, markers of macrophages/inflammation. Both CD68 and OPN were positively correlated with SUV mean . CD68 exhibited the strongest correlation (R = 0.70, p,0.001). LOX-1, the scavenger receptor, was also positively and significantly correlated to SUV mean (R = 0.53, p,0.001). The gene expression of all markers of hypoxia exhibited significant negative correlations with 18 F-FDG uptake. Also TF was negatively correlated with SUV mean and exhibited the strongest correlation (R = -0.71, p,0.001).

Multivariate Linear Regression Analysis of Gene Expression of the Molecular Markers Relative to 18 F-FDG SUV mean
As all molecular markers were significantly correlated with SUV mean , they were all included in a multivariate analysis.

Discussion
Our study showed increased uptake of 18 F-FDG in apoE 2/2 mice on high-fat diet compared to mice on normal chow and the uptake increased with duration of diet. This supports recent findings, that 18 F-FDG can be used to follow the progression of atherosclerosis in mice [9]. To further validate the image-based method, we also performed ex vivo gamma counting of all vessels that were PET-imaged. Ex vivo counting is devoid of any partial volume and spillover effects and may be considered a reference for tracer uptake. Indeed the two patterns of FDG uptake was in essence identical proving that when using our protocol for animal handling and imaging, spillover is of no concern. Accordingly, a strong correlation between the two methods was found. Secondly, we identified 4 genes involved in atherogenesis that together explained 60% of the 18 F-FDG uptake. 18 F-FDG is a glucose analogue taken up by metabolic active cells [17] and evidence suggests that uptake in atherosclerotic lesions correlates with vascular inflammation. Macrophages are thought to be the dominant cell type responsible for the uptake, although it has not been established conclusively [6,18].
To examine the molecular processes contributing to the uptake of 18 F-FDG in the aortas of apoE 2/2 mice, we investigated the gene expression of 10 molecular markers representing different molecular processes important in the atherogenesis. Surprisingly, we found that all of these markers had a significant correlation with the uptake of 18 F-FDG assessed as SUV mean .  [19,20]. We found a rather low correlation of SUV mean with CXCL-1 and MCP-1. Studies in knockout mice suggested that CXCL-1 and its receptor is important after early lesions have been established [21] which implies a delayed expression relative to the onset of atherosclerosis whereas 18 F-FDG uptake is increased from early in the atherogenesis (Fig. 3). Therefore, it seems reasonable that the expression of CXCL-1 does not correlate perfectly with SUV mean . MCP-1 on the other hand, has been shown to be crucial for the initiation of atherosclerosis [22]. The relative low correlation with Figure 5. Univariate linear regression analysis of gene expression relative to SUV mean . Univariate linear regression analysis of gene expression relative to SUV mean (N = 98). A CD68 relative to SUV mean . B OPN relative to SUV mean . C TF relative to SUV mean . D VCAM-1 relative to SUV mean . The 95% confidence interval is indicated by the broken lines. All p-values were Bonferroni corrected. doi:10.1371/journal.pone.0050908.g005 SUV mean may reflect that MCP-1 decreases with the progression of disease. A study of MCP-1 gene expression in mice showed a peak after only 10 weeks of high-fat dieting [23] further supporting this. VCAM-1 is an inducible endothelial cell surface molecule playing a role in mononuclear cell attachment, rolling, and transendothelial migration and is found to be important in both human and murine atherosclerosis [24,25]. VCAM-1 had a high correlation with SUV mean (Fig. 5d). This finding suggests that the expression of VCAM-1 is critical for metabolic activity of the tissue throughout atherogenesis. VCAM-1 also remained in the final model from the multivariate linear regression analysis, further emphasizing its independent importance for 18 F-FDG uptake. A mouse study showed increasing presence of VCAM-1 with high-fat diet and age [26]; which is in line with our findings.

Macrophages/inflammation
As expected, markers of macrophage/inflammation showed a very strong correlation with SUV mean (Fig. 5a and 5b). CD68 is a scavenger receptor expressed by macrophages, Langerhans cells, dendritic cells, and osteoclasts. It is widely used as marker for macrophage activity as it is significantly up-regulated in activated macrophages [27,28]. The strong correlation between CD68 and SUV mean was expected, as numerous studies, both clinical and pre-clinical, have found a high degree of correlation between macrophages and uptake of 18 F-FDG [1,7,8,29]. OPN is a secreted protein expressed by several cell types during inflammation. It is highly expressed at sites with atherosclerotic plaques and present in human atherosclerosis [30]. The expression of OPN has to our knowledge not previously been correlated with 18 F-FDG uptake. The high degree of correlation we found is supported by studies of knockout mice suggesting OPN as an important player in inflammation and atherogenesis [31,32]. Both CD68 and OPN had independent predictive value for 18 F-FDG uptake in the multivariate linear regression analysis.

Scavenger Receptors
LOX-1 is expressed by endothelial cells, macrophages, smooth muscle cells, and platelets and is present in human atherosclerosis [27,33]. We found that LOX-1 had a relatively strong positive correlation with SUV mean . Our finding is supported by previous studies of knockout mice. Here, a reduction of macrophages and a reduced expression of OPN when the mice were deficient of LOX-1 were found [34,35].

Hypoxia
Hypoxia has been demonstrated in advanced human atherosclerosis [36]. HIF-1a and HIF-2a are often used as markers of hypoxia and although the genes are similar, they are not identical, as the list of genes they regulate is only overlapping, not identical. HIF-1a is expressed ubiquitously, whereas HIF-2a is only expressed in certain tissues [37,38]. VEGF is regulated by both HIF-1a and HIF-2a. Whereas the expression of VEGF is low in normal vessel wall, it is up-regulated by hypoxia, inflammatory mediators, and certain growth factors. In atherosclerosis, VEGF is thought to contribute to inflammation, to intimal thickening, and to intra-plaque angiogenesis [39]. We found that all three markers of hypoxia correlated negatively with SUV mean . This is in contrast to a human study from 2008 [36], where the researchers demonstrated up-regulated expression of the same markers comparing stable plaques with early lesions and comparing thrombus-containing plaques with stable plaques, that is to say increasing expression over the development of atherosclerosis. The significance of the results was limited and as the authors only investigated a limited number of lesions (5 of each category), it could be argued that the results may not be representative. Also, we did not study distinct plaques but rather diffuse atherosclerotic disease. Another explanation of the discrepancy to the findings in our study could arise from the smaller lesions in mice compared to humans. It may be that hypoxia is limited in murine atherosclerosis; therefore, inferring a dissimilar gene expression of molecular markers of hypoxia compared to humans is logical [40].
Thrombogenicity TF is expressed both by cells in the vessel wall and by platelets. TF is not only an important player in thrombogenicity, but also functions as a membrane receptor which activation leads to cell proliferation, angiogenesis, and inflammation [41]. We found a strong negative correlation with 18 F-FDG SUV mean and that TF had independent predictive value for SUV mean in the multivariate linear regression analysis. The negative correlation with SUV mean implies that TF decreases with increasing uptake of 18 F-FDG indicating that TF expression decreases with the progression of atherosclerosis. This is in discrepancy with a human study from 1997 [42] where an increase in TF activity was seen during the progression of disease. However, as we were investigating mRNA expression instead of TF activity, this may at least in part explain the different results. Murine studies of TF in atherosclerosis have been sparse and not unanimous, suggesting all from no importance to influencing the degree of atherosclerosis [43][44][45]. Taken together, the role of TF in murine atherosclerosis requires further studies and studies of the mRNA levels in human atherosclerotic lesions will also be of importance.
The present study has some limitations for consideration. The use of PET as a technique for measuring uptake of 18 F-FDG in small areas such as the aorta of mice involves certain challenges, such as resolution and partial volume effects. As all the animals have been treated similarly, the problems are presumed to be alike and the relative comparisons we have made are therefore expected to be largely unaffected. This is also supported by the high correlation found between ex vivo gamma counting of the aortas and PET scans. We measured the expression of mRNA in preference to protein levels and it cannot be assumed that these levels are totally consistent as e.g. post-transcriptional regulation may be prominent [46]. The mRNA measurements were preferred as these are easily quantifiable compared to measurements using immunohistochemistry.

Conclusion
In this study, we have demonstrated that 18 F-FDG can be used to follow the progression of atherosclerosis in apoE 2/2 mice. The gene expression of ten molecular markers representing different molecular processes important for atherosclerosis was shown to correlate significantly with the uptake. A multivariate analysis showed CD68, OPN, TF, and VCAM-1 to be the most important contributors and a statistical model with these parameters explained 60% of the 18 F-FDG uptake.

Author Contributions
Conceived and designed the experiments: AMFH SFP AK. Performed the experiments: AMFH SFP CC TB MMJ JTJ DS RSR. Analyzed the data: AMFH SFP CC TB MMJ JTJ DS RSR AK. Contributed reagents/ materials/analysis tools: RSR AK. Wrote the paper: AMFH RSR AK.