247

Genetic Modifiers of Atherosclerosis in Mice

JONATHAN D. SMITH, HAYES M. DANSKY, AND JAN L. BRESLOW

The Rockefeller University, New York, New York 10021, USA

ABSTRACT: Common atherosclerosis has a genetic component, but it is difficultto determine the specific genes that play a role in atherosclerosis susceptibilityin humans. We have used the apoE-deficient mouse as a model system to exam-ine the effects of candidate genes on atherosclerosis as well as to perform ge-nomic experiments to map and isolate other genes giving rise to atherosclerosissusceptibility. We have tested the effects of mutations in the MCSF and VCAM-1 genes on atherosclerosis, and in both of these cases mutations led to gene dos-age-dependent decreases in atherosclerosis. By successive back breeding, wehave established apoE-deficiency on the C57BL/6 and FVB/N inbred mousestrains. Lesions in C57BL/6 mice are about eightfold larger than those in FVB/N mice, and lesions in F1 hybrids are intermediate in size. We have performedquantitative trait locus mapping on two F2 cohorts and discovered atheroscle-rosis susceptibility loci on chromosomes 10, 14, and 19.

KEYWORDS: atherosclerosis; genome; susceptibility

INTRODUCTION

Atherosclerosis is a common disease in westernized cultures with both environ-mental and genetic components. Although there are examples of single gene defectsleading to atherosclerosis, such as LDL receptor deficiency, these monogenic causesof atherosclerosis are rare and cannot account for the prevalence of this disease. In-stead, it is proposed that atherosclerosis, like other common metabolic disorders, ismulitgenic with common variations in many genes determining one’s genetic sus-ceptibility to this disorder. However, it is exceedingly difficult to tease apart thegenes that affect atherosclerosis susceptibility in human studies because of the in-herent limitations of such studies, including the inability to control for environmen-tal effects, incomplete penetrance of each allele on the phenotype, and the smalleffect of any one allele on the phenotype. Therefore, we used a mouse model of ath-erosclerosis to test for the effects of candidate genes as well as to identify other ath-erosclerosis susceptibility genes by the use of genomic methods that do not assumea priori knowledge about the gene or the pathway by which it mediates atheroscle-rosis susceptibility.

It is possible to induce atherosclerosis in certain inbred mouse strains by the useof a diet containing high levels of cholesterol and cholic acid.1 Although this modelhas also been used to examine atherosclerosis susceptibility genes, the incompletepenetrance of lesion formation as well as the need to use an unphysiological diet that

Address for correspondence: Dr. Jonathan D. Smith, The Rockefeller University, 1230 YorkAve., New York, NY 10021. Phone: 212-327-7210; fax: 212-327-7165.

smithj@mail.rockefeller.edu

248 ANNALS NEW YORK ACADEMY OF SCIENCES

causes hepatic inflammation has dampened enthusiasm for this diet-inducedmodel.2,3 Several genetically modified mouse models that develop robust athero-sclerosis without the need for feeding a diet containing cholic acid are currentlyavailable; these include the apoE-deficient, LDL receptor-deficient, and apoB trans-genic mice.4 In this report, we describe the effects on atherosclerosis of mutations inthe MCSF and VCAM-1 genes upon breeding onto the apoE-deficient background.We also describe our initial results in attempting to isolate atherosclerosis suscepti-bility genes based on their genomic localization.

RESULTS

Testing the MCSF Gene as a Candidate for Atherosclerosis Susceptibility

The initial atherosclerotic cellular lesion in the arterial intima is the accumulationof monocyte-derived macrophages that can take up modified lipoproteins via a vari-ety of scavenger receptors and thus be transformed into lipid-laden foam cells. Wetested the effect of altering monocyte and macrophage number and function on ath-erosclerosis by breeding the osteopetrotic (op) mouse onto the apoE-deficient back-ground. The op mouse carries a spontaneously derived deletion in the gene encodingfor macrophage colony-stimulating factor (MCSF) that leads to a frame shift andcomplete absence of MCSF activity.5 The phenotype of these mice includes de-creased monocytes, decreased tissue macrophages, which is variable in different tis-sues, and an almost complete lack of osteoclasts. This latter deficiency leads todefects in bone remodeling such as decreased marrow space and lack of incisors,which makes it necessary to feed these mice a powdered or liquid diet.6 Breeding ofthis cohort was difficult, as the op mouse has decreased fertility and increased mor-tality. First we bred mice heterozygous for the op mutation (op1) onto the apoE-deficient background (E0) and then we brother/sister-mated these mice to derive E0mice with 0, 1, or 2 wild-type copies of the MCSF gene (op0/E0, op1/E0, and op2/E0, respectively, with the number after the op referring to the number of wild-typeMCSF alleles). We fed all the mice a powdered chow diet containing 4.5% fat and0.02% cholesterol and sacrificed the mice for study at 16 weeks of age. The op0/E0mice weighed ∼21 g, significantly less than their op1/E0 and op2/E0 littermates thatweighed ∼27 g.7 Surprisingly, the plasma cholesterol of the op0/E0 mice was ∼1,300mg/dl, about 2.5-fold higher than the ∼500 mg/dl levels in the op1/E0 and op2/E0mice.7 This excess cholesterol was due to increased cholesterol ester-enrichedβVLDL and LDL, without altering HDL levels.7 As expected, there was a gene dos-age- dependent effect on monocyte differential, with monocytes being 4.0%, 7.5%,and 9.4% of total WBCs in the op0/E0, op1/E0, and op2/E0 mice, respectively.7 Ath-erosclerosis was measured in the aortic root, and the op0/E0 mice had significantlysmaller lesions that had progressed less than those in their op1/E0 and op2/E0 litter-mates. Female mice at this age generally have larger aortic root lesions than do malemice, and in the female mice we saw a larger effect of MCSF deficiency on lesionsize. There was a gene dosage-dependent effect on the mean lesion area in females,which was reduced significantly by 89% and 44% in the op0/E0 and op1/E0 mice,respectively, compared to their op2/E0 littermates.7 In the males the op0/E0 meanlesion area was reduced significantly by 77% compared to that in their op2/E0 litter-

249SMITH et al.: GENETIC MODIFIERS OF ATHEROSCLEROSIS

mates, with intermediate lesion area in the op1/E0 mice.7 Two additional atheroscle-rosis studies were subsequently done in these mice. The first examined aortic rootlesions in 12-week-old mice fed a western type diet containing 20% fat and 0.2%cholesterol. Both male and female op0/E0 mice had about 67% smaller lesions thandid their op2/E0 littermates.8 The second study examined lesion surface area of theentire aorta (enface assay) in 1-year-old mice fed the chow diet, and the op0/E0 micehad 90% less lesion surface area than did the op2/E0 mice.8 Other labs have con-firmed the dramatic effect of the op mutation in the MCSF gene on atherosclerosislesion development upon breeding to either the apoE-deficient or LDL receptor-deficient mouse models of atherosclerosis.9,10

We also performed turnover experiments to determine if the increase in plasmacholesterol in the opo/E0 mice was associated with increased production or de-creased catabolism. We first looked at the catabolism of cholesterol oleyl ether-labeled AcLDL, a ligand for various scavenger receptors, and it was rapid and sim-ilar in the op0/E0 and op2/E0 mice (t1/2 ~1.5 min).8 We then looked at the catabolismof βVLDL isolated from apoE-deficient mice labeled with cholesterol oleyl etherand found that it was significantly delayed in the op0/E0 mice (t1/2 ~5 hours) com-pared to the op2/E0 mice (t1/2 ~2 hours).8 Thus, the increase in plasma cholesterolin the op0/E0 mice was associated with decreased clearance. It is still not clear whyMCSF deficiency decreases VLDL clearance and what receptors on which cells playthe major role in βVLDL uptake.

Testing the VCAM-1 Gene as a Candidate Gene forAtherosclerosis Susceptibility

VCAM-1 is expressed on aortic endothelial cells as sites of lesion predilectionand is induced in apoE-deficient hypercholesterolemic mice compared to wild-typemice.11 Since VCAM-1 is involved in the tight adhesion of monocytes to endothelialcells, we hypothesized that disruption of the VCAM-1 gene in apoE-deficient micewould lead to decreased monocyte entry into the arterial intima and decreased ath-erosclerosis. However, the VCAM-1 gene knockout has an embryonic lethal pheno-type;12 thus, our collaborator, Myron Cybulsky, created a domain 4 deletion (D4D)in the VCAM-1 gene. This deletion destroys one of the two α4 integrin binding sitesand also led to <10% expression of the VCAM-1 mRNA, presumably due to de-creased production or increased catabolism of the truncated mRNA. Although thismutation leads to viable mice, it is an incompletely penetrant embryonic lethal, asfewer homozygous mutants than expected are born from the mating of two heterozy-gous mutants. We bred the heterozygous D4D mutation onto the apoE-deficientbackground, and after brother/sister mating we studied the progeny with 2, 1, or 0wild-type VCAM-1 alleles (V2/E0, V1/E0, and V0/E0, respectively, with the num-ber after the V referring to the number of wild-type VCAM-1 alleles). The extent ofVCAM-1 expression in these mice was qualitatively assessed by immunohistochem-istry in the aortic arch prior to lesion formation, and we observed a gene dosage-de-pendent effect, with expression in the V2/E0 > V1/E0 > V0/E0. We also looked atmonocyte adhesion in situ to the aortic root in young prelesional mice. Again, wefound a gene dosage-dependent effect with monocyte adherence in the V2/E0 > V1/E0 > V0/E0. We then measured aortic root lesion areas in 16-week-old mice fed achow diet and observed a VCAM-1 gene dosage effect on atherosclerosis. Lesion ar-

250 ANNALS NEW YORK ACADEMY OF SCIENCES

eas in the V0/E0 and V1/E0 mice were significantly smaller with median reductionsof 74% and 54%, respectively, compared to those in the V2/E0 mice. A manuscriptdescribing this study in detail has been submitted by Dansky et al.

Quantitative Trait Locus Mapping (QTL) to Identify AtherosclerosisSusceptibility Genes

To identify atherosclerosis susceptibility genes we successively bred the apoEgene knockout onto two inbred mouse strains, C57BL/6 and FVB/N. When we ana-lyzed aortic root lesion areas in these chow diet-fed mice at 16 weeks of age, wefound that the lesions were seven- to ninefold larger in the C57BL/6 strain than inthe FVB/N strain.13 We then bred the F1 hybrid between these two apoE-deficientstrains, and lesion size was narrowly distributed and intermediate in size in these F1mice, which are all genetically equivalent and carry one allele at each locus fromeach parental strain.13 By brother/sister mating the F1 hybrids, we created an F2 co-hort in which each individual is a distinct genetic mosaic, inheriting at any locus ei-ther two C57BL/6 alleles, two FVB/N alleles, or one allele from each parental strain.Lesion size in the F2 cohort was very heterogeneous and spanned a large range cov-ering both parental strain values.13 We then examined if variation in lesion size inthe F2 cohort was associated with variation in biochemical parameters. We found noassociations with lesion area and the levels of total cholesterol, HDL cholesterol,VLDL + LDL cholesterol, apoAI, apoAII, or antibodies to oxidized cardiolipin.13

We then performed QTL analysis on two independent F2 cohorts of about 200mice each, one generated at Rockefeller University (RU) and one by our collabora-tor, Karen Moore, at Millennium Pharmaceuticals Incorporated (MPI). For eachmouse, lesion area was measured and a genome scan was performed using 194 mark-ers at <10 cM spacing that could distinguish between alleles derived from theC57BL/6 and FVB/N strains. The details of this QTL analysis will be described in aforthcoming manuscript. In brief, we discovered three loci that had significant ef-fects on lesion area within the F2 cohorts. The strongest locus was found on chromo-some 10, which yielded LOD scores for lesion area of 11.9 and 7.8 for the RU andMPI crosses, respectively. Interestingly, the FVB/N allele at this chromosome 10 lo-cus was associated with increased, not decreased, atherosclerosis. Although, this iscontrary to the finding in the parental strains, the QTL analysis only defines loci af-fecting variation in the F2 cohort, and it is possible that the affects of this allele aremitigated by other FVB/N alleles in the parental FVB/N mice. A locus on chromo-some 14 yielded LOD scores for lesion area of 3.3 and 2.4 for the RU and MPI cross-es, respectively. For this locus, the C57BL/6 allele was associated with increasedlesion area, as expected from the behavior of the parental strains. A third locus onchromosome 19 was found with a LOD score for lesion area of 3.8 in the MPI cross(this region of chromosome 19 in the RU cross was not genetically pure and thuscould not be analyzed). Using linear regression analysis, these three loci combinedcan account for ∼50% of the lesion variation in the F2 cohort.

We now are facing the arduous task of identifying the genes that are responsiblefor these effects on atherosclerosis susceptibility. We are using a variety of methodsto ascertain the identity of these genes, including analysis of secondary congenicstrains for each of these loci, fine mapping within these secondary congenics, highthroughput sequencing, expression arrays, and candidate gene testing. One aspect

251SMITH et al.: GENETIC MODIFIERS OF ATHEROSCLEROSIS

that we have started to address is which tissues are responsible for the strain effecton atherosclerosis susceptibility. We have performed bone marrow transplantationstudies and have concluded that both host tissues (including vessel wall and liver) aswell as donor tissues (including monocyte derived macrophages) play a role in thisstrain effect on lesion size. If we observe an effect on atherosclerosis in any of thesecondary congenic strains made from the three loci identified above, we expect thatbone marrow transplantation studies would provide conclusive evidence that the ath-erosclerosis susceptibility factor resides within either the host or the donor tissue.This information will aid us in our expression array and candidate gene testing toidentify the atherosclerosis susceptibility gene within each locus.

CONCLUSION

We have tested the effects of disruption of two candidate genes, MCSF andVCAM-1, that effect monocyte macrophage number or entry into the arterial wall,and both had dramatic and gene dosage-dependent effects on atherosclerosis withinthe apoE-deficient mouse model of atherosclerosis. These studies provide compel-ling preclinical support that the development of drugs intended to inhibit monocyteentry into the arterial wall may be an effective way to decrease lesion development.Furthermore, we have made significant progress in the elucidation of the genes thatconfer atherosclerosis susceptibility in an F2 cohort of apoE-deficient mice derivedfrom the C57BL/6 and FVB/N inbred genetic backgrounds. The identification ofthese atherosclerosis susceptibility genes may illuminate new pathways involved inatherogenesis and may also aid in the discovery of the common human genetic vari-ation that plays a role in determining one’s genetic susceptibility to atherosclerosis.The eventual aim of these studies is to gain insight into the genetic profile that pre-disposes humans towards atherosclerosis, allowing the physician to aggressivelytreat risk factors in certain individuals. For example, statin drug therapy may be ben-eficial to those with an atherogenic genetic profile even if they have normal levels ofLDL-cholesterol.

REFERENCES

1. PAIGEN, B., A. MORROW, C. BRANDON, et al. 1985. Variation in susceptibility to athero-sclerosis among inbred strains of mice. Atherosclerosis 57: 65–73.

2. PITMAN, W.A., M.H. HUNT, C. MCFARLAND, et al. 1998. Genetic analysis of the differ-ence in diet-induced atherosclerosis between the inbred mouse strains SM/J andNZB/BINJ. Arterioscler. Thromb. Vasc. Biol. 18: 615–620.

3. LIAO, F., A. ANDALIBI, F.C. DEBEER, et al. 1993. Genetic control of inflammatory geneinduction and NF-kappa B-like transcription factor activation in response to anatherogenic diet in mice. J. Clin. Invest. 91: 2572–2579.

4. SMITH, J.D. 1998. Mouse models of atherosclerosis. Lab. Anim. Sci. 48: 573–579.5. WIKTOR-JEDRZEJCZAK, W., A. BARTOCCI, A.W. FERRANNTE, et al. 1990. Total absence

of colony-stimulating factor 1 in the macrophage-deficient osteopetrotic mouse.Proc. Natl. Acad. Sci. USA 87: 4828–4832.

6. NAITO, M., S.I. HAYASHI, H. YOSHIDA, et al. 1991. Abnormal differentiation of tissuemacrophage populations in osteropetrosis mice defective in the production of mac-rophage colony-stimulating factor. Am. J. Pathol. 139: 657–667.

252 ANNALS NEW YORK ACADEMY OF SCIENCES

7. SMITH, J.D., E. TROGAN, M. GINSBERG, et al. 1995. Decreased atherosclerosis in micedeficient in both macrophage colony- stimulating factor (op) and apolipoprotein E.Proc. Natl. Acad. Sci. USA 92: 8264–8268.

8. DE VILLIERS, W.J.S., J.D. SMITH, M. MIYATA, et al. 1998. Macrophage phenotype inmice deficient in both macrophage-colony-stimulating factor (op) and apolipoproteinE. Arterioscler. Thromb. Vasc. Biol. 18: 631–640.

9. QIAO, J.H., J. TRIPATHI, N.K. MISHRA, et al. 1997. Role of macrophage colony-stimu-lating factor in atherosclerosis: studies of osteopetrotic mice. Am. J. Pathol. 150:1687–1699.

10. RAJAVASHISTH, T., J.H. QIAO, S. TRIPATHI, et al. 1998. Heterozygous osteopetrotic (op)mutation reduces atherosclerosis in LDL receptor-deficient mice. J. Clin. Invest.101: 2702–2710.

11. NAKASHIMA, Y., E.W. RAINES, A.S. PLUMP, et al. 1998. Upregulation of VCAM-1 andICAM-1 at atherosclerosis-prone sites on the endothelium in the ApoE-deficientmouse. Arterioscler. Thromb. Vasc. Biol. 18: 842–851.

12. GURTNER, G.C., V. DAVIS, H. LI, et al. 1995. Targeted disruption of the murineVCAM1 gene: essential role of VCAM-1 in chorioallantoic fusion and placentation.Genes Dev. 9: 1–14.

13. DANSKY, H.M., S.A. CHARLTON, J.L. SIKES, et al. 1999. Genetic background deter-mines the extent of atherosclerosis in ApoE-deficient mice. Arterioscler. Thromb.Vasc. Biol. 19: 1960–1968.

DISCUSSION

B. BERK (University of Rochester, Rochester, New York, USA): Two questions.First, were you not surprised that you actually got so few QTLs in your first cross,and do you think this was a consequence of the number of animals you used or theage at which you looked?

JONATHAN D. SMITH: I showed data for three QLTs that were highly significant,and we also obtained other QTLs that were moderately significant, which I didn’tshow. The QTLs shown were derived from two independent F2 crosses of 200 miceeach. We could have added the QTL scores from the two independent crosses andobtained even more significant loci, but we did not use this approach because therewere some methodological differences between these two studies. Our goal was tostart with the three strongest QTLs, work back towards identifying the genes in-volved in atherosclerosis susceptibility, and prove the concept that the QTL methodis useful for this aim. We are now breeding the necessary secondary congenics forthese studies. Additionally, we are starting to look at epistatic loci, which only canbe observed when we analyze each locus pairwise with every other locus.

BERK: My second question relates to clusters of genes. In the rat crosses that havebeen done, for example, for hypertension, we see very nicely that QTLs cluster as tovarious genetic loci in a variety of crosses. Interestingly enough, those clusters aredifferent between males and females in hypertension. Have you seen as profoundmale/female differences and have you detected any clustering in your atherosclerosiscrosses?

SMITH: Yes, we do see loci that are specific for certain genders. One of the threeloci that I discussed only had a significant association with atherosclerosis lesionsize in male mice. However, the strongest locus was equally significant for male andfemale mice. At the resolution of our current QTL map, it is certainly possible thatatherosclerosis susceptibility genes might be clustered together within one locus. In

253SMITH et al.: GENETIC MODIFIERS OF ATHEROSCLEROSIS

another field of study, Wakeland’s lab found multiple subloci within one large QTLassociated with the susceptibility to autoimmune lupus and nephritis in mice.

D. STEINBERG (University of California, San Diego, California, USA): My ques-tion concerns the MCSF-deficient mouse. First, data in the literature suggest thatwhite blood count is a risk factor in human atherosclerosis. Is it possible that that lowmonocyte count is what makes these animals resistant?

SMITH: Yes, that is one of our main hypotheses, that decreased atherosclerosis inthe MCSF-deficient mice may be due to a decreased number of circulating mono-cytes. Fewer monocytes may lead to less inflammation in the arterial wall. An alter-nate hypothesis is that monocyte-macrophage function is affected by MCSFdeficiency.

STEINBERG: A couple of them did differentials and indeed the monocyte count didcorrelate, but this is just a correlational risk factor. Have you looked at the macro-phages in the few lesions that the mice did get to see if the state of differentiation ofthe monocytes is less than it is in the mature wild-type lesion at the same site?

SMITH: We have only done a cursory examination of those macrophages in the le-sions, and we did find staining with an antibody to a macrophage marker. Incollaboration with Willem de Villiers and Siamon Gordon (1998. Atheroscler.Thromb. Vasc. Biol. 18: 631–640), macrophages in several locations were exam-ined. The Kupffer cells in the apoE/MCSF double-deficient mice were abnormallyfull of lipid deposits; however, this is likely a consequence of the highly elevatedplasma cholesterol levels in these mice

Y. SAKAKI (University of Tokyo, Tokyo, Japan): I believe this kind of modifierworks under a certain genetic background or is it a more general modifier?

SMITH: ApoE-deficiency has been bred onto seven different genetic backgroundstrains. We see strains with large and small atherosclerotic lesions. We do not knowif the same atherosclerosis susceptibility genes are involved in all of these strains,but it would be possible to find out if that is the case by testing all of these strainspairwise by QTL analysis. However, this would be very costly and time consuming.The QTL gene discovery process takes a long time, and that is one of the difficultiesin this type of project. We anticipate that it will take about 10 years for us to go fromthe planning stages of this project to the confirmation of the identity of atheroscle-rosis susceptibility genes.