Deletion of the alternatively spliced fibronectin EIIIA domain in mice reduces atherosclerosis

Fibronectins (FNs) are best known as a family of ligands for the integrin family of adhesion and signaling receptors.^1-4  FNs are essential for heart and blood vessel development,^5,6  and their polymerization regulates extracellular matrix (ECM) composition and organization.⁷ 

FN variants are generated from a single gene by alternative RNA splicing of the V, EIIIA, and EIIIB segments, which are also known as CS-1, ED-A, and ED-B segments, respectively,⁸  and the (V + C) region.⁹  EIIIA and EIIIB are type III repeats that are included or excluded from the FN monomer. Gene targeting experiments have shown that plasma FN, which lacks EIIIA and EIIIB, but includes V/CS-1, reduces brain injury by promoting neuronal survival¹⁰  and also functions in thrombus growth and stability.¹¹  EIIIA-containing fragments are present in synovial fluid in arthritis¹²  and in plasma of patients with vascular¹³  and pulmonary injury.¹⁴  EIIIA peptide can induce expression of proinflammatory cytokines interleukin 1α (IL-1α) and IL-1β, and matrix metalloproteinases.¹⁵ 

High-sequence conservation of EIIIA and EIIIB (> 95% among mammals) and regulated expression^16-18  suggest that they are functional. In vitro, EIIIA-FN mediates wound healing in liver by inducing stellate cells to differentiate into myofibroblasts, which promote fibrosis.^19,20  EIIIA-FN is also up-regulated during cutaneous wound healing²¹  and may have a functional role.²²  Skin fibroblasts respond to EIIIA-FN in vitro by differentiating to a fibrotic phenotype.²³  EIIIA-FN is also secreted by infiltrating leukocytes and deposited in the myocardium of rejecting cardiac allografts.²⁴  FN alternative exons have also been shown to be binding sites for integrins. The alternative V/CS-1 segment is a binding site for leukocyte integrins α4β1 and α4β7.^25,26  Of particular interest, the alternative EIIIA domain is a ligand for leukocyte integrins: α9β1 and α4β1.²⁷ 

In the chronic inflammation that underlies atherosclerosis, development of lesions involves recruitment of blood monocytes to the arterial intima, differentiation to macrophages, lipid accumulation leading to foam cell formation, and smooth muscle cell migration from the arterial wall.²⁸  In thickened aortic intima, EIIIA-FN has been detected near smooth muscle cells.²⁹  In vitro, endothelial V/CS-1, like vascular cell adhesion molecule-1 (VCAM-1), binds α4 integrins on leukocytes.³⁰  In vivo studies using blocking antibodies in mice also implicate V/CS-1 in recruitment of leukocytes during formation of atherosclerotic lesions,³¹  although VCAM-1 appears to predominate.³²  Interestingly, recombinant EIIIA, but not other FN domains, binds and activates the Toll-like receptor 4 (TLR-4).³³  Mammalian Toll-like receptors are known to activate a signal transduction cascade that results in nuclear translocation of nuclear factor-κB (NF-κB)³⁴  and appear to be involved in atherogenesis.^35,36 

To investigate the in vivo roles of EIIIA-FN, we generated mutant mice by gene targeting in which the EIIIA alternative exon is deleted (EIIIA^–/– mice). Reduced atherosclerotic lesion area throughout the aortic tree of EIIIA^–/– alipoprotein E–null (ApoE^–/–) mice, compared with ApoE^–/– controls, demonstrates a functional role for EIIIA in atherogenesis. Expression of FN in both plasma and extracellular matrix suggest that this protective phenotype may be due to systemic mechanisms, involving plasma lipoproteins, or may be due to processes within the vessel wall, or both. Indeed, we found increased expression of EIIIA-FN in both the plasma and in vessel wall lesions in ApoE^–/– control mice. In EIIIA^–/–ApoE^–/– mice, total plasma cholesterol was reduced. Further characterization of this phenotype revealed decreased cholesterol in large lipoprotein particles only. To address EIIIA-FN function in the vessel wall, we assessed in vitro foam cell formation. EIIIA-FN mRNA was induced in macrophages in response to modified low density lipoprotein (LDL), and quantitative assays of foam cell formation showed that EIIIA^–/–ApoE^–/– macrophages accumulated less intracellular lipid than ApoE^–/– control cells. These studies suggest, by compelling genetic analyses, that EIIIA-FN is functional in both plasma lipoprotein metabolism and in macrophage foam cell formation.

Mutations within the single copy mouse FN gene were generated by homologous recombination in embryonic stem (ES) cells,³⁷  which was followed by Cre-loxP–mediated recombination.^38,39  The targeting vector contained 6-kilobase (kb) genomic sequence, flanking the EIIIA exon (Figure 1A). A selection cassette, with neo and thymidine kinase (TK) genes driven by a phosphoglycerate kinase promoter, was inserted between 2 loxP sites, with a third loxP site distal to exon EIIIA. J1 ES cells were cultured by standard procedures,⁴⁰  electroporated with linearized targeting vector, and selected in Geneticin (Life Technologies, Grand Island, NY). Resistant clones were screened by Southern blot analysis of EcoR1-digested genomic DNA with hybridization probe 1. A 4.5-kb band in addition to the wild-type 6.5-kb band indicated targeted clones (data not shown). Of 432 ES cell clones analyzed, 14 were targeted. Of these, 4 clones were cultured further and subjected to a second electroporation, with supercoiled Cre recombinase expression vector pMC-Cre. Seventy-two hours later, selection against TK began in 2 μM gancyclovir.⁵  Recombination between LoxP sites 1 and 3 (EIIIA^–/– mutation) versus 1 and 2 (EIIIA-flox mutation) was assayed by hybridization with probe 2 (Figure 1B) and occurred at a frequency of 66% and 9%, respectively. Heterozygous ES cells were injected into C57BL/6 and BALB/c blastocysts by standard procedures⁴¹  and bred to achieve germ line transmission of the mutant allele. Heterozygous progeny were detected by Southern blot hybridization with probe 2. Interbreeding of heterozygotes resulted in viable homozygous progeny from both mutations (Figure 1C). Mice were generated, bred, and housed at all times in a specific pathogen-free barrier facility. Mice (strain name EIIIA-null) can be obtained from the Mutant Mouse Regional Resource Centers (http://www.mmrrc.org/).

Primers from flanking constitutive exons that distinguish included versus excluded forms for each alternative exon were EIIIA (5′-CAAACTGCAGTGACC-3′ and 5′-CATGAGTCCTGACAC-3′), EIIIB (5′-CATGCTGATCAGAGTTCCTG-3′ and 5′-GGTGAGTAGCGCACCAAGAG-3′), and V/CS-1 (5′-GCTACATTATCAAGTATGAG-3′ and 5′-AATGATGTACTCAGAACTCT-3′). PolyA-plus RNA was reverse transcribed by standard methods, using random hexamers. First-strand cDNA was amplified using the above-mentioned exon-specific primers.

Newborn dermal fibroblasts were grown in serum-free Opti-MEM media (Life Technologies). Rabbit antiserum against rat FN (R61) was provided by Richard Hynes (Massachusetts Institute of Technology, Cambridge). Monoclonal antihuman EIIIA-plus FN (clone FN-3E2), rhodamine-conjugated goat antimouse immunoglobulin M (IgM) and fluorescein isothiocyanate (FITC)–conjugated goat antirabbit IgG were from Sigma (St Louis, MO).

Deoxycholate (DOC) was used to isolate soluble and insoluble FN in cultured fibroblasts, essentially as described.^42,43  After plating in media containing FN-depleted serum, cells were lysed and DOC-insoluble material was separated by micro-centrifugation. Protein was quantified by bicinchoninic acid (BCA) assay (Pierce, Rockford, IL). FN was visualized by Western blot analysis of reduced proteins, using affinity-purified rabbit antihuman fibronectin (Sigma). Chemiluminescent detection was performed by using horseradish peroxidase (HRP)–conjugated goat antirabbit IgG and the SuperSignal West Pico Substrate (Pierce).

ApoE^–/– mice⁴⁴  that had been backcrossed 13 generations to C57Bl/6 background were purchased from Taconic Farms (Germantown, NY). EIIIA^–/– mice (backcrossed 7 generations to C57Bl/6 background) were mated to ApoE^–/– mice to generate EIIIA^–/–ApoE^–/– and EIIIA^+/+ApoE^–/– controls. Genotypes were determined by PCR. For EIIIA, 5′-GTACGTAACCAATGCTCGGT-3′ and 5′-ATGGCTGTCAGGATGGTCAT-3′ yield a 2.3-kb band for the EIIIA^+/+ and a 1.7-kb band for the EIIIA^–/– allele, and a 2.5-kb band for the EIIIA-flox allele. For ApoE, primers IMR 180 (5′-GCCTAGCCGAGGGAGAGCCG-3′), IMR 181 (5′-TGTGACTTGGGAGCTCTGCAGC-3′), and IMR 182 (5′-GCCGCCCCGACTGCATCT-3′) yield a 184-bp product for wild-type and 250-bp product for mutant allele. For studies involving a high-fat and -cholesterol diet, animals were maintained ad libitum on Diet D12108 from Research Diets (New Brunswick, NJ), which provides 40% total energy from fat, 1.25% cholesterol by weight, and lacks cholate. After 8, 12, or 16 weeks on the high-fat diet, mice were maintained in a fasting state overnight and anesthetized with Avertin, blood was drawn by retroorbital puncture, and the whole animal was perfused with 10 mL phosphate-buffered saline (PBS) through the left ventricle. The entire aorta was dissected, beginning at the aortic root at the base of the heart, and ending at the branch to the iliac arteries. After fixation in 3.7% buffered formaldehyde, aortae were washed in PBS, briefly soaked in propylene glycol, and stained with oil red O for 4 hours while shaking. Oil red O was prepared as a 3% stock in 2-propanol, and 6 parts stock was diluted with 4 parts distilled water just before staining. Stained tissue was thoroughly differentiated from unstained tissue by soaking for 3 changes in 85% propylene glycol while shaking (the final change lasting overnight). Aortae were cut longitudinally and pinned en face while immersed in PBS, and the luminal side of the vessel was photographed at × 7 magnification with a Nikon SMZ dissecting microscope (Melville, NY), numerical aperture 0.2 using incident light. Photographs were scanned and evaluated for lesion area by morphometry using Image-Pro software (MediaCybernetics, Carlsbad, CA). The entire area of each aorta was calculated by the program using a hand-drawn outline. Stained areas, representing atherosclerotic lesions, were counted by the program, using consistent color parameters for all aortae. Lesion area is presented as a percentage of the total aortic area, for individual aortae.

Enzyme-linked immunosorbent assays (ELISAs) were performed essentially as described,¹³  using goat antibodies to a 29–amino acid synthetic peptide from human EIIIA. Human cellular FN was used for standards. Color formation was induced by addition of p-nitrophenyl phosphate (Sigma), and well optical densities were read in a Genios microtiter plate reader.

Sections of human carotid artery lesions were kindly provided by Frederick Schoen (Brigham and Women's Hospital, Boston, MA). Monoclonal antihuman EIIIA-plus FN clone IST-9 (from Sigma) was used, followed by biotinylated goat antimouse IgG and Vectastain ABC (Vector Labs, Burlingame, CA). EIIIA was localized in cryostat sections of mouse aortic lesions using the MOM (mouse on mouse) kit (Vector Labs) and monoclonal antihuman EIIIA-plus FN clone FN-3E2 (from Sigma) at 3 μg/mL. Biotinylated secondary antibody was used at one-fifth recommended concentration, followed by Vectastain ABC, and 3′-amino, 9′-ethylcarbazide as substrate for peroxidase.

Plasma from individual animals (100 μL) was subjected to fast performance liquid chromatography (FPLC) essentially as described.⁴⁵  Total cholesterol in each fraction was assayed by enzymatic methods (Wako Chemicals USA, Richmond, VA).

Three days after intraperitoneal injection of 3% Thioglycollate (Difco Laboratories, Detroit, MI) elicited macrophages were harvested and cultured overnight in Macrophage Serum Free media (Life Technologies). Native and acetylated human LDL were from Biomedical Technologies (Stoughton, MA). Foam cell formation assays were performed essentially as described.^46,47  Both native and acetylated LDL were used at 50 μg/mL. Sonicated cells were divided for protein and cholesterol assays. Lipids were extracted in hexane and isopropanol (2:1), the organic layer was dried down under vacuum,⁴⁵  and cholesterol was assayed as described in “Lipoprotein analysis.” For lipid staining, paraformaldehyde-fixed cells were stained with 0.25% oil red O and counterstained with hematoxylin.

Interbreeding of heterozygous mice demonstrates that both EIIIA mutations result in viable mice when homozygous (Figure 1C) on pure 129S6/SvEv, C57BL/6, and BALB/c backgrounds (back-crossed 7 generations). EIIIA^–/– and EIIIA-flox mice display no macroscopic abnormalities and have normal peripheral blood leukocyte counts and breeding characteristics. Although no systematic analysis of life span was performed, we detected no premature death of EIIIA^–/– mice up to age 2 years, which is average for mice in the laboratory.⁴⁸ 

In Figure 1D, RT-PCR shows that the EIIIA^–/– mutation affects expression of the EIIIA exon only. RNA from wild-type embryos is almost exclusively EIIIA-plus, EIIIB-plus, and V/CS-1-plus (includes each exon) with very low amounts of minus forms. In EIIIA^–/– homozygotes, all RNA is EIIIA-minus, whereas EIIIB and V/CS-1 expression is unchanged.

Immunofluorescence microscopy (Figure 1E) with a monoclonal antibody to the EIIIA domain of FN shows that wild-type fibroblasts (serum-free cultures) contain abundant EIIIA-FN. In contrast, EIIIA^–/– cells display no staining above background. Staining parallel cultures with a polyclonal antibody that recognizes all FN variants shows that EIIIA^–/– cells have abundant FN that lacks the EIIIA domain.

To determine whether the lack of EIIIA-FN affects assembly of the FN-containing ECM, we compared the rate of matrix assembly in wild-type and EIIIA^–/– fibroblasts using deoxycholate (DOC). DOC-soluble material represents cell-associated FN, whereas DOC-insoluble material represents higher molecular weight aggregates of FN that has organized into fibrils. At 4 and 24 hours after plating, we detected equal amounts of FN in wild-type and EIIIA^–/– cell lysates in both DOC-soluble and DOC-insoluble material (Figure 1F).

Immunohistochemistry with monoclonal antibodies revealed that EIIIA-FN is prominent in lesions from mice and humans. In Figure 2A, a section of ApoE^–/– mouse aorta shows that EIIIA-plus FN is present near endothelial cells within an early lesion but is absent from adjacent, uninvolved endothelium. Some staining is also associated with macrophages and/or early foam cells of the lesion. In a more advanced lesion (Figure 2B), endothelial EIIIA-FN is barely detectable. Instead, EIIIA staining is most prominent in macrophage foam cells. In human lesions (Figure 2C-D), EIIIA-FN is also present within the lesion but is not detectable in the endothelium. In a complex lesion, which has a fibrotic cap over a necrotic core, EIIIA-FN is present in scattered cells within the shoulder region. In less complex lesions, EIIIA-FN is detected near macrophages as well as smooth muscle cells on the luminal side.

In normal mouse plasma, which contains about 300 μg/mL FN,⁵  most FN lacks the EIIIA domain. In ApoE^–/– mice however, we found, by ELISA, that plasma EIIIA-FN was elevated. Immunopurified goat anti-EIIIA peptide was used as a capture reagent, and alkaline phosphatase-conjugated immunopurified antiplasma FN was used as a disclosing agent. In C57Bl/6 (ApoE^+/+) controls, EIIIA-FN was present at 1.5 μg/mL (n = 6). In ApoE^–/– mice, EIIIA-FN was increased to 2.3 μg/mL (n = 12) P < .005, Student unpaired t test. For these studies, mice were fed standard mouse chow.

EIIIA^–/– mice were bred with ApoE^–/– mice to generate EIIIA^–/–ApoE^–/– mice and ApoE^–/– controls, which were placed on a high-fat, high-cholesterol diet for 8, 12, or 16 weeks. To compare extent of atherosclerosis in the 2 strains, whole aortae were stained with oil red O, and lesion area was measured by morphometry. Figure 3A displays representative aortae after 16 weeks on diet. In both sexes, EIIIA^–/–ApoE^–/– mice have reduced lesion area, which was confirmed by quantification (Figure 3B). In males, EIIIA^–/–ApoE^–/– mice have a 51% reduction in lesion area after 16 weeks. Female EIIIA^–/–ApoE^–/– mice are protected at all time points: 67% reduction at 8 weeks, 60% reduction at 12 weeks, and 64% reduction at 16 weeks. In Figure 3C, regional lesion areas indicate that protection from atherosclerosis occurred throughout the aortic tree in EIIIA^–/–ApoE^–/– mice. In Figure 3D, cryosections of lesions were stained with oil red O. Compared with controls, EIIIA^–/–ApoE^–/– lesions were thinner, and contained areas that were lightly stained, indicating less lipid accumulation in foam cells. We detected no crystallized cholesterol clefts in either control ApoE^–/– or experimental EIIIA^–/–ApoE^–/– lesions, which would indicate extremely high concentrations of extracellular cholesterol.

Fasting plasma total cholesterol concentrations were determined after maintenance on high-fat diet. In both sexes, cholesterol concentrations in EIIIA^–/–ApoE^–/– mice were slightly lower than ApoE^–/– controls (Table 1). In males ApoE^–/– of 1053 mg/dL (n = 38) is compared with EIIIA^–/–ApoE^–/– of 867 mg/dL (n = 59), (P < .0001, t test). In females ApoE^–/– of 825 mg/dL (n = 40) is compared with EIIIA^–/–ApoE^–/– of 700 mg/dL (n = 56) (P < .001). To further characterize this difference, we separated plasma lipoproteins by FPLC and assayed total cholesterol in the fractions (Figure 4A). Lower cholesterol was observed in the VLDL fractions only. There was no difference in the LDL or HDL fractions compared with ApoE^–/– controls.

Total cholesterol concentrations were also determined in ApoE^+/+ mice. In both sexes, EIIIA^+/+ApoE^+/+ and EIIIA^–/– ApoE^+/+ concentrations were approximately equal (122 ± 32 mg/dL). To determine whether a difference in VLDL was present, despite the similar total concentrations, we assayed cholesterol on FPLC separated fractions. Figure 4B shows that cholesterol in VLDL fractions was too low to assay in both genotypes. Cholesterol in LDL and HDL fractions in EIIIA^–/– mice, compared with EIIIA^+/+ controls, however, was unchanged.

Expression of EIIIA-FN within lesions suggests that it may affect foam cell behavior in development of the lesion. To test this idea, we assayed in vitro foam cell development in peritoneal macrophages in response to modified LDL.^46,47  To determine whether EIIIA-FN expression is induced in macrophages by exposure to acetylated LDL, we performed RT-PCR on mRNA isolated from treated cells (Figure 5A), using specific primers in flanking exons 11b and 12a that distinguish mRNA that includes or excludes EIIIA. Nontreated ApoE^–/– macrophages have a relatively high amount of EIIIA^– mRNA, but EIIIA⁺ mRNA is only faintly detectable. On addition of acetylated LDL, splice variant accumulation shifts, and the amount of EIIIA⁺ mRNA increases significantly, whereas the amount of EIIIA^– mRNA decreases. After exposure to acetylated LDL for 24 hours, amounts of EIIIA⁺ and EIIIA^– mRNAs are equal. Comparison of ethidium bromide–stained band intensities revealed an approximately 10-fold increase in EIIIA⁺ mRNA and an approximately 5-fold decrease in EIIIA^– mRNA, compared with nontreated macrophages. EIIIA^–/–ApoE^–/– cells display only a marginal increase in EIIIA^– FN mRNA expression on exposure to acetylated LDL.

In Figure 5B, lipid accumulation in foam cells was visualized by staining macrophages with oil red O after culture for 72 hours either with or without acetylated LDL. In ApoE^–/– cells, many large lipid droplets accumulate in virtually every cell. In contrast, EIIIA^–/–ApoE^–/– cells display fewer and smaller lipid droplets, and some cells have none. To quantify this deficit in cholesterol accumulation in EIIIA^–/–ApoE^–/– cells, we assayed total cholesterol by enzymatic methods. After exposure to acetylated LDL for 24 hours, we compared the ratio of total cholesterol with protein in each culture (Figure 5C). Native LDL induced little accumulation of cholesterol above control in macrophages of either genotype. In contrast, cholesterol content in cells increased dramatically in response to acetylated LDL. However, EIIIA^–/–ApoE^–/– cells accumulate 31% less cholesterol per milligram of protein than ApoE^–/– control cells (P < .01, t test).

Although not directly related to the atherosclerosis studies presented here, we have also analyzed EIIIA^–/– mice for defects in skin and liver wound healing. Skin wound healing was assayed at 3, 7, 10, and 14 days after application of full-thickness wounds (4-mm diameter, by way of punch biopsy). In all groups, 5 to 7 each of wild-type and mutant mice were analyzed. Histologic analysis revealed no differences in re-epithelialization, granulation tissue, and neovascularization compared with EIIIA^+/+ controls (E.L.G., L. van de Water, unpublished results, August 2000). These results differ from those reported by Muro et al²²  in which the same alternative exon of FN was deleted. This lack of agreement may be due to differences in genetic background or degree of inbreeding in the 2 studies, or analysis of wounds generated by punch biopsy versus excisional wounds, or the pathogen status of the mouse colony. We also analyzed liver injury, in response to intraperitoneal injection of carbon tetrachloride, and found liver wound healing and subsequent fibrosis to be indistinguishable from control mice (E.L.G., S.L.O., unpublished results, October 2001).

These studies provide in vivo evidence for a role of the alternatively spliced EIIIA segment of FN in atherogenesis. We generated EIIIA^–/– mice in which the EIIIA exon is deleted. Analysis of mRNA and protein shows that the EIIIA^–/– mutation disrupts expression of the EIIIA exon only. On an atherogenic genetic background, EIIIA^–/– mice have decreased atherosclerosis (as much as 67% reduction), suggesting a proatherogenic role for EIIIA-FN in the pathophysiology of the disease. Studies to determine the underlying cellular mechanisms associated with this protective phenotype indicate that EIIIA-FN is involved in both plasma lipoprotein metabolism and in macrophage foam cell formation.

In ApoE^–/– plasma, we show that EIIIA-FN was increased, compared with control ApoE^+/+ plasma. This increased accumulation is consistent with playing a functional role by EIIIA-FN in lipoprotein metabolism, as is the protection from atherosclerosis in EIIIA^–/–ApoE^–/– mice. However, further analyses, especially localization of EIIIA-FN with respect to specific plasma lipoproteins, will be required to test this idea. To date, no interactions of FN with specific lipoproteins are known. Additionally, plasma EIIIA-containing fragments must be characterized, which have been observed in other inflammatory situations.^12,13 

Plasma cholesterol concentrations are slightly lower in EIIIA^–/–ApoE^–/– mice than in ApoE^–/– controls (up to 18% reduction). FPLC analysis revealed that the decrease in cholesterol is specific to the VLDL fraction. Cholesterol in the LDL and the atheroprotective HDL fractions is the same as in ApoE^–/– controls. The VLDL fraction includes several species of lipoproteins, and this modest decrease in cholesterol may reflect a significant disruption of a subset of those lipoproteins. The decrease in VLDL cholesterol was not observed in the ApoE^+/+ background. This may be due, however, to the very low concentrations of VLDL present in ApoE^+/+ plasma, as indicated by our analysis of separated lipoproteins. Further lipoprotein and in vivo analyses, using an alternative atherogenic background such as the human apolipoprotein B (ApoB) transgenic mouse strain,⁴⁹  will be required to determine whether this phenotype contributes to protection from atherosclerosis.

We also observed EIIIA-FN within atherosclerotic lesions, associated with macrophage foam cells, suggesting that EIIIA-FN plays a role in lipid accumulation within the lesion. Monocyte-derived macrophages, once in the arterial intima, are known to take up modified lipoproteins by way of scavenger receptors, becoming fat-laden foam cells.²⁸  The major scavenger receptors that have proatherogenic roles are CD36 and SR-A, which bind LDLs that have been extensively modified, either by acetylation or oxidation.⁵⁰  In our studies, acetylated LDL induced inclusion of the EIIIA alternative exon in mRNA in macrophages. In EIIIA^–/– cells, cholesterol accumulation, in response to acetylated LDL, was reduced by 31%. Our results suggest that EIIIA may play a role in scavenger receptor–mediated uptake of LDL. To date, FN has not been shown to interact with scavenger receptors. Preliminary results in our laboratory suggest that the extracellular matrix of foam cells in culture has very little EIIIA-FN. This raises the possibility of EIIIA acting in a soluble manner. Further FN localization and EIIIA peptide rescue experiments will be required to test this idea.

Our studies show that EIIIA^–/–ApoE^–/– females displayed significant protection at all time points assayed, whereas males are protected only after 16 weeks on diet. Sex-specific differences in atherogenesis have been observed previously on the ApoE^–/– background.⁴⁷  Further in vivo studies, using ovariectomized females, will be required to determine the significance of this observation. EIIIA^–/–ApoE^–/– mice (males and females) display relatively uniform reduction in lesion area throughout the aortic tree. In contrast, absence of von Willebrand factor appears to reduce lesions in particular aortic regions that are exposed to disturbed flow.⁵¹ 

We detected EIIIA-FN in endothelial cells of early lesions, which suggests a role for EIIIA in leukocyte recruitment. EIIIA has been shown to be a ligand for 2 leukocyte integrins, α4β1 and α9β1.²⁷  Endothelial expression of EIIIA is transient, however, and our studies have not ruled out a functional role. To definitively test for involvement of endothelial EIIIA, EIIIA-flox mice could be crossed to TIE2-Cre–expressing mice.⁵²  The V/CS-1 alternative domain of FN binds the α4β1 integrin on monocytes, and blocking this interaction in vivo, with V/CS-1 peptide, decreases leukocyte entry and fatty streak formation.³¹  In ApoE^–/– carotid artery, blocking α4β1–V/CS-1 interaction by antibody reduced monocyte adhesion and increased rolling velocities.³²  Thus, EIIIA may interact directly with leukocyte integrins also, or may influence V/CS-1 interaction with integrins.

Viability of EIIIA^–/– homozygotes demonstrates that EIIIA-FN variants, whereas widely expressed in embryos⁵³  are not required for embryogenesis. At least some FN variants are required for organization of cells in developing blood vessels and heart,^5,6  through interaction with α5β1 integrin.^54,55  The EIIIA segment is a ligand for α9β1 and α4β1 integrins,²⁷  and their absence results in perinatal⁵⁶  or embryonic⁵⁷  lethality. Viability of EIIIA^–/– mice demonstrates that these integrins interact with other ligands in embryos. Deletion of the EIIIB alternative exon also results in viable mice, but in vitro studies suggest that EIIIB is important for cell growth and matrix assembly.⁵⁸  In contrast, we show that EIIIA^–/– fibroblasts grow normally, and assembly of EIIIA-minus FNs into pericellular and fibrillar matrices is indistinguishable from that in wild-type cells.

A plausible explanation for the high-sequence conservation of EIIIA has not yet been found. We have shown that EIIIA is not essential for embryogenesis, and its proatherogenic roles are not likely to be evolutionarily advantageous. However, our results suggest that EIIIA may be important for host defense in vivo. Others have shown, in vitro, that EIIIA activates the TLR-4 receptor.³³  Scavenger receptor SR-A and TLR-4 are both macrophage pattern recognition receptors, which function in the innate immune response, and have both endogenous and exogenous ligands.^59,60  Recent preliminary results in our laboratory support this idea, suggesting a functional role for EIIIA in acute inflammation also, in accumulation of extravasated macrophages. These results are consistent with our present analysis of foam cell formation, in that EIIIA appears to be important for macrophage behavior after exit from the circulation. Studying the resistance of EIIIA^–/– mice to infection would be an interesting test of this hypothesis.

We thank Monty Krieger for valuable advice, Sharon Karackattu for assistance with FPLC, and Fred Schoen for human artery samples. We also thank Michael Gimbrone, Richard Hynes, Tanya Mayadas-Norton, and David Milstone for helpful discussions.