Physiological Genomics

Liver gene expression associated with diet and lesion development in atherosclerosis-prone mice: induction of components of alternative complement pathway

Adrian Recinos III, Boyd K. Carr, David B. Bartos, Istvan Boldogh, J. Russ Carmical, L. Maria Belalcazar, Allan R. Brasier


Diet-induced changes in serum lipoproteins are a major risk factor for the development of atherosclerosis, the leading cause of mortality in Westernized countries. Atherosclerosis is now appreciated to be a systemic inflammatory disease where increased synthesis of inducible proteins by the liver, such as C-reactive protein (CRP) and others, may play roles in accelerating the disease process. To systematically investigate the genetic response of the liver to diet-induced atherosclerosis, we applied high-density microarray technology in a mouse model of atherosclerosis (LDLR−/− mouse). LDLR−/− mice and congenic (LDLR+/+) controls were placed on low-fat (LF) or high-fat (HF) Western-style diets. The Western diet produced sustained elevations in total cholesterol (2.5-fold for LDLR+/+, 5.0-fold LDLR−/−) relative to the respective LF groups. Tissues were harvested after 12 wk when significant atherosclerotic lesion development was first detectable by en face histomorphometry of oil red O-stained aortas. Diet, rather than genotype, was most highly associated with development of atherosclerotic lesions. Liver mRNA expression profiles of triplicate animals from each group were determined by high-density oligonucleotide microarrays; and genes with transcript levels influenced by genotype and diet were identified by two-way ANOVA. Approximately one-third of the 102 genes identified to be altered by diet [Pr(F) < 0.0005] were involved in lipid metabolism. In addition, we identified components of the alternative complement pathway, including C3, properdin, and factor D, for which mRNA levels were independently confirmed by quantitative real-time RT-PCR analysis, and C3 protein was demonstrated in aortic lesions by immunostaining. These findings suggest that induction of the alternative complement pathway may be an additional mechanism by which a high-fat/Western diet accelerates atherosclerosis.

  • LDL receptor knockout mouse
  • microarray
  • factor D
  • adipsin
  • properdin
  • C3

atherosclerosis is a multifactorial disease of the large arteries that represents the leading cause of morbidity and mortality in Western Countries (1, 8). In humans, atherosclerosis develops through three distinct phases (6). The fatty streak represents the first identifiable abnormality, formed by lipid-filled macrophages (foam cells) in the subendothelial space. The second phase, the fibrous plaque, is formed in the setting of infiltration of activated macrophages and T-lymphocytes, proliferation and migration of smooth muscle cells, and deposition of collagen into the intima. Finally, thrombus formation with aggregation of fibrin and platelets can result in acute vascular thrombosis (25). Intense study of this disease process has indicated that complex interplays of environmental and genetic influences determine atherosclerotic risk.

Enhanced dietary intake of saturated fat produces hypercholesterolemia, particularly of LDL and VLDL fractions, and is a well-established risk factor for the development/complications of atherosclerosis (23). Epidemiological studies in humans have shown that increases in total cholesterol of as little as 20 mg/dl produce a 12% increase in long-term risk of ischemic heart disease (44). The development of genetically modified atherosclerosis-prone mice has facilitated mechanistic studies of the relationship between diet and atherosclerosis. For example, the LDL receptor knockout mouse (LDLR−/−) on a low-fat (normal chow) diet has only twofold elevations in total cholesterol and develops no significant vascular lesions in the first 6 mo of life (26). When challenged with diets rich in saturated fats, LDLR−/− mice develop significant hypercholesterolemia and vascular (fatty streak) lesions throughout the aorta within 3 mo (19).

Despite strong association of diet and vascular disease, half of all myocardial infarctions occur in persons with normal plasma lipids (5), and most patients with coronary artery disease do not have elevated cholesterol (21). This suggests that mechanisms independent of dyslipidemia play roles in the development of atherovascular disease. Recently, large scale epidemiological studies have shown that vascular inflammation is an independent atherosclerotic risk factor (3840). We now appreciate that chronic elevations in circulating C-reactive protein (CRP) and other IL-6-dependent hepatic acute-phase reactants are independently predictive of atherosclerotic risk and mortality in humans (38). The hepatic acute-phase response alters patterns of plasma proteins by cytokine-induced changes in gene expression by the liver (15). In the acute-phase response, cytokines, such as IL-6, produced at the site of vascular injury induce expression of liver proteins that have significant cardiovascular actions (3, 4). For example, the complement factors (C3, C4, C9, factor B), coagulant/fibrinolytic factors (fibrinogen, plasminogen, tPA), antiproteases (α1-antitrypsin), components of the innate immune system (CRP), apolipoproteins (serum amyloid A), and the vasoactive precursor, angiotensinogen, are all established acute-phase reactants and have known cardiovascular effects (15). These findings have led to the hypothesis that liver-derived proteins may participate in the atherogenic process. However, the complete spectrum of genes induced by the liver and their potential roles in atherosclerotic mechanisms have not been systematically studied.

Here, we employed microarray gene expression analysis to explore changes in liver gene expression associated with early lesion development in atherosclerosis-prone mice. LDL receptor-deficient (LDLR−/−) and congenic C57BL/6J controls were challenged with Western-type high-fat diets and compared with the same animals on low-fat diets. Although only the LDLR−/− mice showed marked elevations in total plasma cholesterol and triglyceride levels on the high-fat diet, both strains developed significant early atherosclerotic lesions by 12 wk. We report here the 102 genes most significantly altered by the effect of diet. We identified genes involved in lipid metabolism, generation of reactive oxygen species, cytokines, and surprisingly, the alternative complement pathway. The alternative complement pathway has been implicated in the promotion of vascular inflammation and atherosclerosis. The effect of diet in influencing the expression of rate-limiting components of the alternative complement pathway suggests novel mechanisms for diet-induced atherosclerosis.



Six-week-old male mice, 25 LDL receptor-deficient (LDLR−/−) and 32 control (LDLR+/+), in the C57BL/6J genetic background, were obtained from the Jackson Laboratory (stock numbers 002207 and 000664, respectively). After acclimatizing 1 wk, half of the mice were fed normal rodent diet (4.5% fat by weight; PicoLab Rodent Chow 20, PMI LabDiet), and the other half were fed a Western-type high-fat diet (0.15% cholesterol, 21% milkfat by weight, 42% calories from fat; TD 88137, Harlan Teklad) (31), ad libitum, with animal care given in accordance with institutional guidelines. Mice were fed the low- and high-fat diets for 3, 6, or 12 wk before euthanasia and tissue harvest. For immunohistochemical studies LDLR-null and C57BL/6J mice were fed the low- and high-fat diets for 18 wk before euthanasia and tissue harvest.

Plasma lipid determinations.

Every 2 wk and at time of death, mice were anesthetized (intraperitoneal injection, 100 mg/kg ketamine and 5 mg/kg diazepam), blood samples were taken from the retro-orbital venous plexus, and serum total cholesterol and triglyceride levels were measured using the Infinity (Sigma) cholesterol and triglyceride reagents with known lipid standards (Sigma stock nos. C-0534, C-4571, C-4696).

Tissue harvest and quantification of atherosclerotic lesions.

Methods for aorta preparation and en face morphometry to quantitate lesions were modified from established techniques (22). Aortas were fixed, cleaned, stained with oil red O (Sigma) to visualize intimal fatty lesions from the proximal arch to the iliac bifurcation, and flattened onto slides (9). Photomicroscopy images were exported into Photoshop 6 (Adobe), atherosclerotic lesion areas were selected by contrast differences, and images were uploaded into the ScionImage (National Institutes of Health/Scion Corp) software for measurement of lesion areas. Total atherosclerotic lesion area (mm2) is expressed as percent positive staining relative to total aortic area.

Real-time RT-PCR gene expression analysis.

At time of death, animals were anesthetized, and livers were excised, flash-frozen, and stored at −80°C. Total liver RNA was extracted by acid guanidium-phenol extraction (Tri-Reagent, Sigma), and 30 μg samples were fractionated by electrophoresis on a 1.2% agarose-MOPS-formaldehyde gel to ensure RNA integrity. Real-time RT-PCR assays were designed for a subset of genes deemed statistically significant by microarray analysis. The coding sequence (CDS) and chromosome contiguous sequence for each gene were acquired through the NCBI public database and used to map exon-exon junctions. Probes were designed to transcend a junction to reduce the impact of potential genomic DNA contamination in the surveyed RNA samples. Gene-specific primers and TaqMan probes were designed with Primer Express 2.0 (Applied Biosystems) using default settings (primer Tm = 58–60°C, probe Tm = 10°C > primer Tm, GC content = 30–80%, length = 9–40 nucleotides). The resulting primers/probes are shown in Table 1. Assays were mixed at final concentrations of 900 nM primer and 250 nM probe. Reverse transcription was performed on 1 μg of total RNA with random primers, utilizing TaqMan reverse transcription reagents (Applied Biosystems) as recommended by the manufacturer. One-tenth of the reverse transcription reaction was used as template for the subsequent PCR reaction, which was carried out in triplicate. The reaction consisted of TaqMan universal master mix, template cDNA, and target primer and probe mixed as described above. Thermal cycling was carried out with an ABI Prism 7000 sequence detection system (Applied Biosystems) under factory defaults (50°C, 2 min; 95°C, 10 min; and 40 cycles at 95°C, 15 s; 60°C, 1 min). Threshold cycle numbers (Ct) were defined as fluorescence values, generated by cleavage of the probe, exceeding baseline. Relative transcript levels were quantified as a comparison of measured Ct values for each reaction to a designated control via the 2−ΔΔCt method (27). To normalize for template input, GAPDH (endogenous control) transcript levels were measured for each sample and utilized in these calculations. Fold change values were calculated as the ratio of the 2−ΔΔCt sample averages, and two-factor ANOVA with replication was employed to evaluate differences. Values of P < 0.05 were considered significant.

View this table:
Table 1.

Reagents for quantitative real-time RT-PCR gene expression analysis


At time of death, anterior portions of hearts with proximal aorta were carefully dissected free, embedded in OCT, flash-frozen in liquid nitrogen, and stored at −80°C. Aortic cryosections (transverse, 5-μm thick) on slides were prepared as previously described (32). Slides were fixed in cold acetone:methanol (1:1), blocked with 2% donkey serum, and incubated with fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mouse complement C3 polyclonal antibody (1:10) (Research Plus, Manasquan, NJ). Photomicroscopy was with a Nikon Eclipse TE 300 inverted UV microscope fitted with a Photometrix CoolSnap Fx digital camera interfaced with Metamorph software (Universal Imaging, Downingtown, PA).

Oligonucleotide probe-based microarrays.

The mouse genome U74Av2 GeneChip (Affymetrix, Santa Clara, CA) containing probe sets for ∼6,000 full-length genes and ∼6,500 expressed sequence tag (EST) clusters was used as previously described (47). Briefly, first-strand cDNA synthesis was performed using total RNA (10–25 μg), a T7-(dT)24 oligomer (5′ GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGG-dT24 3′), and SuperScript II reverse transcriptase (Life Technologies). The T7 promoter introduced during first-strand cDNA synthesis is then used to direct the synthesis of cRNA using bacteriophage T7 RNA polymerase. The biotin-labeled target RNAs are fragmented to a mean size of 200 bases and initially hybridized to a test array containing a set of probes representing genes that are commonly expressed in the majority of cells (actin, GAPDH, transferrin receptor, transcription factor ISGF-3, 18S RNA, 28S RNA, and Alu) to confirm their successful labeling. Hybridization is performed at 45°C for 16 h in hybridization buffer (0.1 M MES, pH 6.6, 1 M NaCl, 0.02 M EDTA, and 0.01% Tween 20). Four prokaryotic genes (bio B, bio C, and bio D from the Escherichia coli biotin synthesis pathway; and cre, the recombinase gene, from P1 bacteriophage) are added to the hybridization cocktail as internal controls. Arrays are washed using both nonstringent (1 M NaCl, 25°C) and stringent (1 M NaCl, 50°C) conditions prior to staining with phycoerythrin-streptavidin (10 μg/ml final). GeneChip arrays were scanned using a Gene Array Scanner (Hewlett-Packard) and analyzed using the GeneChip Analysis Suite 4 software (Affymetrix).

Data analysis.

Twelve microarray GeneChip experiments were performed involving three mouse livers in each of the four genotype/diet experimental groups (i.e., LDLR+/+ low-fat, LDLR+/+ high-fat, LDLR−/− low-fat, and LDLR−/− high-fat). For comparison of the fluorescence intensity (average difference) values among multiple experiments, the average difference values for each experimental GeneChip were scaled to that of the base GeneChip. This was done first by calculating the “2% trimmed mean” (a measurement of global signal intensity) for each probe set considered to be present on the GeneChip. The trimmed mean is obtained by calculating the mean signal intensity of the array after discarding the top and bottom 2% average difference values (representing the “outliers”). Scaling was performed by multiplying all the average differences of genes considered present in the experimental array by a global normalization factor defined as the ratio of the base trimmed mean to that of the experimental trimmed mean (the base GeneChip was considered to be the first control GeneChip). Because both genotype and diet can be considered experimental treatments, the scaled average difference values were then subjected to a two-way analysis of variance with replications (ANOVA; S-Plus 6, Insightful) to determine which genes were significantly influenced by either genotype or diet. The topmost 102 genes (probe sets) that were most significantly influenced by the diet variable, along with the top 66 that were most influenced by genotype, with “cutoff” P values [Pr(F)] at the 0.0005 and 0.003 level, respectively, were selected for further analysis. Agglomerative hierarchical clustering using the unweighted pair-group method with arithmetic mean (UPGMA, Ref. 30) was performed on the indicated genes (Spotfire Array Explorer, v. 3.01; Spotfire, Cambridge, MA). Visualization of the data is graphically presented as a heat map where the degree of fluorescence intensity is represented by a color gradient. For the heat map shown, green represents the minimum average difference value (5 scaled units), black represents the mid average difference value (5,000 scaled units), and red represents the maximum average difference value (10,000 scaled units). Investigators may obtain the primary data at our website ( or at the GEO database (; accession nos. GSM5403GSM5414, GSE363).

Other data are reported as means with standard deviation (SD). Differences were analyzed by ANOVA (single-factor or two-factor with replication). Values of P < 0.05 and P < 0.01 were considered significant.


Plasma lipid levels and diet-induced obesity.

Male LDLR+/+ (n = 32) and LDLR−/− (n = 25) C57BL/6J mice at 7 wk of age were each divided into two groups and fed either a normal chow diet or a Western high-fat diet (42% calories as fat) [group 1, LDLR+/+, low-fat diet (LF); group 2, LDLR+/+, high-fat diet (HF); group 3, LDLR−/−, LF; group 4, LDLR−/−, HF]. Serum total cholesterol and triglyceride levels were measured in blood samples collected at 2-wk intervals and death. By 2 wk on the diets, however, all serum lipid levels had equilibrated to steady-state values with no further significant changes. There were also no differences in lipids between mice in a given group. Mean total cholesterol and triglyceride values for each mouse group are shown in Table 2 (31–60 individual blood samples per group over the 12 wk). In the LDLR+/+ mouse groups, high-fat feeding elevated total cholesterol levels by ∼2.5-fold (P < 0.0001), reaching a moderate 143 mg/dl. Triglyceride levels were unchanged in these groups. In the LDLR−/− groups, high-fat feeding raised cholesterol levels ∼5-fold, from 162 to 853 mg/dl, and triglyceride levels ∼6-fold, from 82 to 497 mg/dl (P < 0.0001 for each), similar to previous reports (42).

View this table:
Table 2.

Effect of diet on body composition and serum lipoproteins

Mean body weights for each group of animals at 12 wk of dietary treatment are also shown in Table 2. The Western high-fat diet resulted in highly significant weight gains for both mouse genotypes (22%, P = 0.001 and 36%, P < 0.01, for LDLR+/+ and LDLR−/−, respectively). We also noted a small (10%) but significant (P = 0.03) unexplained weight difference at 12 wk between the two genotypes on the low-fat diet (with the LDLR−/− mice being smaller) but not on the high-fat diet. No weight differences between genotypes were seen at 3 or 6 wk.

Development of atherosclerotic lesions.

Mice were killed after 3, 6, and 12 wk on low- or high-fat diets, and aortas were evaluated for lesion development by en face morphometry. Representative oil red O-stained aortas at 12 wk are shown in Fig. 1. Percent aortic lesion area after 3, 6, and 12 wk for each experimental group is shown in Fig. 2. On low-fat diets, neither mouse genotype developed significant increases in atherosclerotic lesion area in 12 wk. On the high-fat diets, however, both genotypes developed increases in lesion area during the same time period (3–12 wk: LDLR+/+, 3.1% to 8.7% at P = 0.001; LDLR−/−, 3.9% to 11.1% at P = 0.0001). These experiments were designed to focus on early low-level lesion development, employing a mildly atherogenic mouse model. We were surprised to observe comparable small elevations in lesion area in both genotypes, especially since total cholesterol levels were much more highly elevated in the LDLR−/− mice. In view of this, the microarray data analysis (see below) focused primarily on the diet variable, with a secondary focus on genotype, as the diet variable was most highly associated with early lesion development. We anticipate, of course, that if our experiment had been carried out longer, the LDLR−/− group would have developed more substantial lesion increases over the LDLR+/+ group as has been reported by others (6, 42).

Fig. 1.

En face histomorphometry of atherosclerotic lesions. After 12 wk on diets, aortas were harvested, dissected en face from the arch to just beyond the iliac bifurcation, and stained with oil red O.

Fig. 2.

Quantification of atherosclerotic lesion burden. Percent aortic lesion areas were determined after 3, 6, and 12 wk on the diets for the 4 genotype/diet experimental groups. Values show means (with SD) of the percent lesion for all animals in each group. *P = 0.001 vs. LDLR+/+ HF at 3 wk, and P = 0.001 vs. LDLR+/+ LF at 12 wk. †P = 0.0001 vs. LDLR−/− HF at 3 wk, and P < 0.001 vs. LDLR−/− LF at 12 wk.

Changes in liver gene expression.

To explore changes in liver gene mRNA levels associated with early atherosclerotic lesion development, total RNA was purified from the livers of each of three mice from each of the four experimental groups for the 12-wk time point, and gene expression changes were profiled by oligonucleotide arrays representing ∼12,500 mouse full-length genes and EST clusters. Since lesion development in our mice most significantly associated with high-fat feeding in each of the two LDLR genotypic groups, we first employed two-way ANOVA to identify liver genes with expression levels significantly altered by diet. Diet significantly changed the abundance of 952 mRNAs represented by the GeneChip at the level of P < 0.05 and 413 mRNAs at the level of P < 0.01. The 102 of these genes affected by diet with highest statistical significance [Pr(F) ≤ 0.0005] were selected for further analysis. The functions of 86 genes and identified ESTs (16 remained unknown) were classified by the primary biochemical functions of their products and are presented in Table 3.

View this table:
Table 3.

Functional grouping of liver genes regulated by dietary fat

We also examined genes for which expression changes were do to the LDLR genotype as identified by the ANOVA. In contrast to diet, the difference in genotype associated with fewer changes in mRNA abundance, 525 genes at the P < 0.05 level and 162 genes at the P < 0.01 level. The 66 of these genes that were most significantly affected by genotype [Pr(F) < 0.003] were selected for further functional characterization as above. Table 4 shows 60 of these genes and identified ESTs (6 remained unknown), classified by primary biochemical functions of their gene products. Importantly, the LDLR was identified in this analysis as differentially expressed by genotype, an internal control for our microarray data.

View this table:
Table 4.

Functional grouping of liver genes regulated by mouse genotype

To further analyze the diet- and genotype-regulated gene sets, and to determine whether they contained distinct groups of gene expression profiles, hierarchical clustering was performed (47). In this technique, the mathematical proximity of each gene expression profile (or clusters thereof) is indicated by the height of a common line that connects two nodes. Figure 3 shows the hierarchical clustering analysis for the 102 diet-regulated genes alongside the primary data, which is visually represented as a heat map. Visual inspection of the dendrogram indicates that the gene expression patterns were divided into two large clusters (upregulated and downregulated with respect to the high-fat diet) and each of these were again separated into two additional clusters, giving four main groups of genes, overall. The largest of these four groups (heat map expression profiles 73 through 102) contains genes that were, for the most part, more highly expressed across the profile, with changes due to diet often being of downregulation in going from low fat to high fat. About half of the genes in the “general metabolism” functional class (Table 3) fell into this cluster. Another of the larger of the four clusters (heat map expression profiles 22 through 56) contains genes that were generally and notably upregulated by the high-fat diet within each genotype. About half of the genes in this cluster fell into the “cholesterol/lipid/steroid metabolism” functional group. A smaller group (expression profiles 57 through 72) describes genes that were generally downregulated by the high-fat diet (more than half of these clustered genes also fell into the “cholesterol/lipid/steroid metabolism” functional class); and a group at the top of the dendrogram (profiles 1 through 21) encompasses genes that were generally at low expression levels across the profile but were sometimes significantly upregulated by the high-fat diet. When hierarchical clustering was performed for the group of 66 genes most significantly affected by genotype [Pr(F) < 0.003], the two major clusters were 1) 35 genes that were more highly expressed in the wild-type mice vs. LDLR−/− and 2) 28 genes that were expressed at high levels only in the LDLR−/− fed the high-fat diet (data not shown). As would be expected, and as can be gleaned from Table 4, a number of these latter genes functioned in lipid metabolism or as receptors and cell surface transporters.

Fig. 3.

Clustering and heat map analysis of diet-regulated genes. Agglomerative hierarchical clustering was performed using the unweighted pair-group method with arithmetic mean (materials and methods). A heat map for the expression level of each gene for each of the four genotype/diet groups is shown at right (mean of 3 mice per group). In the color spectrum, green represents the minimum value of 5 scaled fluorescence intensity units, black represents the middle value of 5,000 scaled units, and red represents the maximum value of 10,000 scaled units. Top left: the number of nodes at each point on the horizontal scale. The Gene ID column cross-references genes that are listed in Table 3. Unknown ESTs shown here but not in Table 3 are, by Gene ID and GenBank accession number: 1, AA144469; 10, AI842065; 15, AV265258; 17, AA684456; 20, U30241; 21, X53929; 34, AA822174; 36, AC002397; 46, AI837679; 50, AI843399; 53, AW047688; 55, AI853098; 74, AI266885; 78, AI849615; 91, AA673769; and 101, X74856.

About one-third of the 102 most significantly altered gene expression profiles, by diet, in our study involved genes that function in lipoprotein, lipid, and cholesterol/steroid metabolism (Table 3). This, of course, is not surprising, given the plethora of genes that are transcriptionally regulated by cellular uptake of cholesterol and fatty acids. For example, in a recent study of mouse liver gene expression, SREBP-1, apolipoprotein A-IV, and adipsin (factor D) were all upregulated by a Western diet (16). Most of these cholesterolgenic genes, however, are downregulated with high-fat feeding to varying degrees in both control and LDLR-deficient mice (and hence cluster in the lower portion of Fig. 3). Another group of genes in Table 3 that are mostly moderately downregulated with high-fat intake are those that were categorized as functioning in general (or intermediary) metabolism.

In the microarray analyses, we observed multiple genes encoding components of the alternate complement pathway to be statistically significantly regulated by diet [Pr(F) ≤ 0.001]. Of particular interest in this statistical group are the genes for complement C3, properdin, and factor D (adipsin), members of the alternative pathway, and complement components C1q and C2, members of the classic pathway. Factor D, in fact, exhibited the highest fold change in expression level of any of the 952 genes surveyed at the Pr(F) ≤ 0.05 level (increased ∼88-fold in the high-fat LDLR+/+ mice; and upregulated to a lesser degree in the high-fat LDLR−/− mice).

To validate and more precisely quantitate changes in mRNA levels that were observed in the microarray analysis (as well as to assess complement genes that were not on the array), real-time RT-PCR analyses of the same liver RNA samples (in triplicate) were conducted for selected genes of interest related to findings concerning the complement pathways. As shown in Table 5, RNA expression level changes, regulated by diet, for four complement components and factors were positively correlated in the two types of analysis. We were able to confirm expression changes for C2, C3, factor D, and properdin, and we obtained new expression data for C4 and factor B, although these latter changes did not reach statistical significance. For C1q α-chain the expression changes between high- and low-fat for the LDLR-null mice varied in direction between the two types of analyses, but there was much less experimental variance in the data for the microarray analysis, allowing these data to reach statistical significance only in this one analysis.

View this table:
Table 5.

Comparison of diet-regulated expression level fold changes determined by microarray analysis with those quantitated by real-time RT-PCR, for selected genes of interest relative to complement pathways

Complement component C3 in atherosclerotic lesions.

Cryosections of proximal aorta in the region of the aortic valve from each of the four mouse experimental groups (3–4 aortas for each mouse group; 6–10 sections for each aorta) were immunofluorescently stained for complement component C3, the initiating protein of the alternative complement pathway. In some, but not all, sections containing atherosclerotic lesions from LDLR−/− mice on high-fat diets, FITC fluorescence was evident in lesion areas (Fig. 4). No C3 immunostaining was observed in any cryosections of aortas from other mouse experimental groups. These data are of interest in that they indicate alternate complement pathway activation in vascular lesions as suggested by microarray analysis of liver transcripts.

Fig. 4.

Immunofluorescence staining of complement component C3 in atherosclerotic lesion area. Adjacent 5-μm transverse cryosections of proximal aorta from an LDLR−/− mouse, fed a high-fat Western diet, were stained with either preimmune serum (A) or FITC-conjugated rabbit anti-mouse complement C3 antibody (B). A portion of the lesion area (L) in A is seen to fluoresce with complement C3 in B. CM, cardiac myocytes; VW, vessel wall. Original magnification = 100×.


Recent studies have supported the new paradigm that atherosclerosis is a chronic systemic inflammatory disease (4, 25, 38). In otherwise healthy patients, chronic elevation of hepatic acute phase reactants is a predictor of increased risk for the development and progression of ischemic heart disease and stroke. It currently remains controversial whether hepatic acute phase reactants are merely biomarkers for atherosclerotic processes or directly participate in lesion development (4). The observations that CRP is enriched in atherosclerotic lesions, where it induces monocyte chemotactic protein-1 (MCP-1) and adhesion molecule expression (34), and serum amyloid A functions as a apolipoprotein that enhances cholesterol uptake by smooth muscle cells (24) argue that hepatic proteins may participate in atherosclerotic lesion development. To better understand the participation of the liver gene expression in atherosclerosis, we first sought to identify the spectrum of hepatic genes expressed in mice developing early atherosclerotic lesions vs. those without lesions.

The major finding of this study was the discovery of differences in complement pathway component mRNA expression profiles between study groups that correlate with their atherogenic state. We have been able to independently confirm the changes in complement components C2, C3, factor D, and properdin expression identified by microarray with real-time RT-PCR analysis. We are presently performing plasma protein quantifications and activity assays, as well as more immunohistochemical studies on lesions, to further evaluate roles of complement components in atherogenesis. The mechanisms by which diet could induce the expression of these complement components are not presently known and will require further experimentation; however, because of the important activities of the alternative complement pathway on vascular biology, we will speculate as to its potential biological significance.

The complement system is the most archaic component of the mammalian immune system, first arising in prehistoric jawless fish, and is generally thought of as comprising two pathways, the classic, or antibody-dependent, and the alternative pathway (33). The alternative pathway, schematically diagrammed in Fig. 5, can be activated within atherosclerotic lesions containing enzymatically modified LDL (E-LDL), now recognized as a major atherogenic LDL form (2). E-LDL binds to factor C3b, forming the substrate for binding to factor B. The highly specific serine protease, factor D (43), then performs an essential rate-limiting enzymatic modification of C3bB, forming the C3bBb convertase (11). The C3bBb convertase amplifies the alternative complement cascade, cleaving C3 into additional C3b (eventually regenerating convertase), as well as producing the anaphylatoxin C3a. C3a, in turn, is rapidly degraded into the metastable C3a des-Arg in plasma [acylation stimulating protein (ASP)]. In addition to its ability to stimulate triacylglycerol synthesis in adipocytes, C3a/C3a des-Arg activates cytokine production by mononuclear cells and induces their chemotaxis (12, 18). For example, C3a des-Arg/ASP induces IL-6 production by LPS- or IL-1-stimulated monocytes through C3a receptor-dependent mechanisms (12). This activity is directly relevant to atherosclerosis, as IL-6 has been demonstrated to be expressed in atherosclerotic lesions (10), and chronic elevation of IL-6 is linked to long-term atherosclerotic risk in humans (39).

Fig. 5.

Possible involvement of components of alternative complement pathway in development of atherosclerosis. Shown is a schematic diagram of the alternative complement pathway in the atherosclerotic lesion (see text). The alternative complement pathway is initiated in an antibody independent manner, by association of complement factor C3b with enzymatically modified LDL (E-LDL). The activated C3 convertase complex amplifies C3b degradation, forming C3a. C3a is rapidly converted into the metastable C3a des-Arg (ASP), with inflammatory and immunomodulatory activities. B, factor B; D, factor D (adipsin). Red, liver gene expression increased by high-fat diet; green, liver gene expression decreased.

Others have shown that components of both the classic and alternative complement pathways, including C1q, C3, C3a, and C4, are enriched in human atherosclerotic lesions (45), and recently, component C3 deposition has been noted in the lesions of apoE-null mice overexpressing CRP (35). We were able to reproduce this observation for component C3 in mouse lesions. Importantly, LDLR−/− mice deficient in the C3 complement component have a defect in lesion progression from the fatty streak to the fibrous plaque stage (7). Together, these data suggest that in the E-LDL-rich fatty streak, deposition and activation of the alternative complement pathway may be an additional stimulus for recruitment of activated monocytes and B- and T-lymphocytes for progression to the fibrous plaque.

Factor D, originally named adipsin, based on its identification as an abundant transcript in adipocytes (41), was later noted to have expression changes in rodent genetic models of obesity (13). Although once thought to be expressed exclusively in adipocytes, adipsin mRNA is found in the gastrointestinal tract, spleen, and liver (16, 29). Factor D is unique among complement components in that it does not require proteolytic processing for activation; and it exhibits exceptional specificity for its substrate C3bB (43). Because it also circulates in plasma at rate-limiting concentrations, serum factor D activity is a linear function of its transcription level (43). Thus our observation of highly induced factor D expression with high-fat feeding suggests a mechanism for increased peripheral activity of the alternative complement pathway, i.e., the C3b:B:E-LDL complex assembled in the vascular wall is activated by enhanced circulating concentrations of factor D. An independent study, also recently observing dramatic inductions of liver adipsin mRNA, involved ICAM-1-deficient mice on a high-fat Western diet (16) and agreed with our findings here. As the rate-limiting enzyme of alternative complement pathway activation (14), factor D is potentially a therapeutic target for reducing vascular inflammation/lesion progression (14).

Factor B is the alternative pathway functional analog of C2 (from the classic complement pathway). Both factor B and C2 share numerous structural similarities and are encoded by genes located in close proximity in the MHC class III locus (46). Our study reveals downregulation of hepatic transcription of both C2 and factor B in animals on high-fat diets. Because of the physical proximity of the two genes, C2 and factor B may share common regulatory mechanisms. Downregulation of C2 can be explained by the unusual complement activation properties of atherogenic lipids; specifically, there is clear evidence that the alternative pathway is activated, but there is dependence on the classic pathway component, C1q, however, not involving antibodies (28). Further studies will need to be undertaken to address these findings.

Properdin transcription levels were modestly increased in the livers of the high-fat-fed animals in this study. Properdin functions to stabilize the C3 convertase formed by the action of factor D on C3bB (i.e., the C3bBb convertase, Fig. 5) (11), functioning as a positive regulator of the cascade (33). Properdin exists in plasma as cyclic polymers, in discrete ratios of dimers, trimers, and tetramers, and its plasma activity is affected dramatically by small changes in concentration (20, 36). Therefore, although the differences in transcriptional levels due to diet interaction observed herein were small, the physiological impact of these differences may be substantial.

In summary, high-density microarrays have identified changes in hepatic gene expression associated with high-fat diet in rodent models of atherosclerosis. In addition to the expected alterations in lipid and cholesterol metabolic gene networks, we have discovered that rate-limiting components of the alternative complement pathway are also coordinately induced early in the disease process. These findings are significant because the alternative complement pathway plays a pathogenic role in atherosclerotic lesion progression. Diet, therefore, may induce a pro-atherogenic state through several mechanisms, not solely through changes in circulating plasma lipids, but also through changes in serum proteins with pro-atherogenic activities. It will be of interest to determine whether recent novel alternative complement pathway inhibitors, such as anti-factor D and anti-properdin antibodies (17, 37), will have utility in attenuating diet-induced atherosclerosis.


This work was supported by National Heart, Lung, and Blood Institute Grant HL-70925 (to A. R. Brasier).


We thank the University of Texas Medical Branch (UTMB) Molecular Genomics Core Facility (Thomas G. Wood, Director) for performing the microarrays; John M. Doss, Jr., for technical assistance in the UTMB Animal Resource Center; and Lee DeForke, Jr., for technical assistance with photomicroscopy.


  • Article published online before print. See web site for date of publication (

    Address for reprint requests and other correspondence: A. Recinos III, Division of Endocrinology, MRB 8.138, The Univ. of Texas Medical Branch, 301 Univ. Blvd., Galveston, TX 77555-1060 (E-mail:


View Abstract