The terminal complement proteins C5b 9 augment binding of high density lipoprotein and its apolipoproteins A I and A II to human endothelial cells

The terminal complement proteins C5b-9

augment binding of high density lipoprotein and

its apolipoproteins A-I and A-II to human

endothelial cells.

K K Hamilton, P J Sims

J Clin Invest.

1991;

88(6)

:1833-1840.

https://doi.org/10.1172/JCI115504

.

Terminal complement protein complexes C5b-9 have been found in human atherosclerotic

lesions. Insertion of C5b-9 in the endothelial cell membrane alters permeability, induces

membrane vesiculation, and activates secretion. We hypothesized that complement might

also alter interactions of the endothelial surface with lipoproteins, particularly high density

lipoprotein (HDL), which is reported to inhibit C5b-9-induced hemolysis. We now

demonstrate that exposure to C5b-9 increases (by 2- to 50-fold) specific binding of HDL and

its apolipoproteins (apo) A-I and A-II to endothelial cells. Binding to cells exposed to

antibody, C5b67, and C5b-8 was virtually unchanged. Enhanced binding was also

dependent on the number of C5b-9 complexes deposited on the cells. Other agonists that

activate endothelial secretion did not augment binding. Calcium was required for full

exposure of new binding sites by C5b-9. The C5b-9-induced increase in binding was

independent of the increase observed after cholesterol loading. In addition, apo A-I and A-II

appear to compete for the same binding sites on untreated and C5b-9-treated cells. In

contrast to the data reported for red cells, we were unable to detect significant inhibition of

C5b-9-mediated endothelial membrane permeabilization by HDL (up to 1 mg/ml) or by apo

A-I (up to 100 micrograms/ml). These data demonstrate that the C5b-9 proteins enhance

endothelial binding of HDL and its apoproteins, suggesting that intravascular complement

activation may alter cholesterol […]

Research Article

The Terminal Complement Proteins

C5b-9

Augment

Binding

of

High

Density

Lipoprotein and Its

Apolipoproteins

A-I

and

A-Il

to

Human

Endothelial Cells

Karen K. Hamilton** and Peter J. Sims*uII

DepartmentsofMedicine*and Microbiology and

Immunology,"1

University ofOklahoma Health SciencesCenter,Saint FrancisHospital of Tulsa Medical Research InstitutetandCardiovascular BiologyResearch

Program,6

Oklahoma MedicalResearchFoundation, Oklahoma City, Oklahoma73104

Abstract

Terminal complement protein complexes C5b-9 have been found in human atherosclerotic lesions. Insertion of C5b-9 in the endothelial cell membrane alters permeability, induces membranevesiculation, and activates secretion. We hypothe-sized thatcomplementmightalso alterinteractionsof the endo-thelialsurface withlipoproteins, particularly high density lipo-protein (HDL), which is reported to inhibitC5b-9-induced he-molysis. We now demonstrate that exposure toC5b-9 increases (by2-to50-fold) specific bindingof HDL and its apolipopro-teins(apo)A-I and A-Il to endothelial cells. Binding to cells exposed to antibody, C5b67, and C5b-8 was virtually un-changed. Enhancedbindingwasalso dependent on the number of C5b-9complexesdepositedon the cells.Other agoniststhat activateendothelial secretion didnotaugmentbinding. Calcium was required for fullexposure of new binding sites by C5b-9. TheC5b-9-induced increaseinbindingwasindependentof the increase observed after cholesterol loading. In addition, apo A-I and A-II appear to compete for the samebinding siteson un-treated and C5b-9-treated cells. In contrast to the data re-ported for red cells, we were unable to detectsignificant inhibi-tion ofC5b-9-mediated endothelial membrane permeabiliza-tion byHDL(upto1mg/ml)orbyapo A-I(upto100,ug/ml). These data demonstrate that theC5b-9 proteinsenhance endo-thelial binding of HDL and its apoproteins, suggesting that intravascularcomplementactivationmay alter cholesterol ho-meostasis in the vessel wall.(J. Clin. Invest. 1991. 88:1833-1840.) Keywords: apolipoprotein A-I- apolipoprotein A-11

atherosclerosis * cholesterol *inflammation Introduction

The endothelial cell surface expresses high-affinity binding sitesfor

circulating

lipoproteins whichare believedto be im-portant inregulatingcholesterolhomestasis ofthevessel wall. Among these are low density lipoprotein (LDL) receptors, which mediate internalizationof the LDL particle and uptake ofcholesterol.High density lipoprotein(HDL; 1, 2) and apoli-poproteins (apo) A-I and A-IV (3) have also been shown to bindtoendothelialcells. However, membrane binding ofHDL

Addressreprintrequests toDr. Hamilton, Department of Medicine, BSEB Room 306, University of OklahomaHealth SciencesCenter,

P.O.Box26901, Oklahoma City, OK 73190.

Receivedfor publication15May 1991and in revisedform 14 Au-gust 1991.

does notappeartoresult in significant internalization ofthe HDLparticle(1).Thenatureof theinteraction ofHDL with cellsurfaces,and its functionalsignificance,remain poorly un-derstood.BothHDL and free apo A-I andA-Ilhavebeen dem-onstratedtoserveasacceptorsfor removal of cholesterol from several types of cells in culture (4-7),andHDL and apo A-I and A-IV binding isincreased significantly aftercholesterol loadingofsomecells(8, 9). However,it isnotknown whether HDL binds to a membrane protein receptor, or directly with membranelipid.AnHDL-binding protein,which increases in activity after cholesterol loading, has been reported in fibro-blasts and bovine aortic endothelial cells(10).However,given the high affinity of apo A-I and A-II for phospholipid(1 1-13), directbindingof these apoproteins to membrane lipid remains animportant possibility.

Membraneinsertion of the terminal complementproteins C5b-9 has been shown to disrupt membranephospholipid orga-nization, creating patches of nonbilayer phospholipid and faci-litating transbilayer movement of phospholipid (14-16). As-sembly of C5b-9 complexes has been shown to perturb the endothelial surface by inducingsheddingof membrane micro-vesicles andfusionof secretory granules with theplasma mem-brane(17, 18). The membrane vesicles express binding sites for factor Va and thus provide catalytic surface forassemblyofthe prothrombinaseenzymecomplex. It has been suggested that factorVa-bindingsitesinducedbyC5b-9 in platelet and endo-thelial membranes may result from C5b-9-induced surface

ex-pression of acidicphospholipids, whicharenormallya compo-nent of the innerleaflet ofthecell membranebilayer(18).

Recently,C5b-9complexeshave been identified in human atherosclerotic plaques (19, 20).Moreover, endothelial C5b-9 deposition has been demonstrated before lesion development in the hypercholesterolemic rabbit model ofatherosclerosis, suggesting an interrelationship between complement activa-tion and atherogenesis (21). The specific role of the terminal complement proteins in this process remains unknown. The knowninteraction oftheC5b-9proteins withplasma lipopro-teins, including theinhibitoryeffects of HDL andits apopro-teins onC5b-9-mediated hemolysis(22,23), raises the possibil-itythatC5b-9 depositionon theendothelial surfacealters the interaction of HDL with the vascular wall.

Asa firststep towards elucidating the effects of comple-mentactivation onendothelial interaction with plasma lipo-proteins, we have examined the effects of the C5b-9 complex on thebinding of HDL and itsmajor apolipoproteinsapo A-I andA-Il to theendothelialcellsurface.Theseexperiments dem-onstratethatcomplementactivationresults in asignificant in-crease in thespecific binding ofboth HDL and itsfree apolipo-proteinsto human endothelial cells, and that the exposure of these membrane binding sites requires incorporation of C9 intotheC5b-9complex.

J.Clin. Invest.

©The American Society for Clinical Investigation,Inc.

Methods

Cell culture. Endothelial cells were harvested from human umbilical vein and cultured as previously described (17). First-passage cells grownin 1-cm2 wells in 48-well plates were used in all experiments describedhere. Before HDL binding experiments, cells were grown for 24 h in media containing 10% lipoprotein-deficient human serum (LPDS)1 and 50

_,g/ml

25-hydroxycholesterol, rather than 20% fetal bovine serum. For some apolipoprotein binding experiments, cells werepreloaded with cholesterol by incubating for 24 h in medium containing 10-30

_,g/ml

25-hydroxycholesterol as indicated in the fig-urelegends.

Purification ofhuman complement proteins. Human C8 and C9 were purified as previously reported (24). Human serum deficient in complement component C8 was prepared by absorption of normal serumagainst rabbit antibody to C8 coupled to agarose (17). For some experiments, LPDS (prepared as described below using Liposorb, Cal-biochem-Behring Corp., San Diego, CA) was used to prepare serum which was both C8- and lipoprotein-deficient.

Preparation of polyclonal rabbit anti-endothelial cell antibody. Plasmamembranes partially purified from cultured human endothe-lial cells were used to immunize rabbits (17). IgG was prepared by absorption to protein A-agarose(17).

Purification and iodination of HDL. HDL was purified from hu-manplasmaanticoagulated with EDTA by sequential flotation ultra-centrifugation (d 1.063-1.21) (25). HDL was iodinated using IODO-GEN(Pierce Chemical Co.,Rockford, IL). Free iodine was removed by gelfiltrationfollowed by extensive dialysis against phosphate-buffered saline (PBS, 0.15 MNaCl, 9 mM Na2HPO4, 1.6 mM KH2PO4) con-taining 2% KI,followed by dialysis against PBS and activated charcoal, to reducethe amount of"25Iassociated with lipid. The specific activity was 860cpm/ng; > 98% of the radioactivity was precipitable with tri-chloroacetic acid.

PurificationofapoA-I and A-II. ApoA-IandA-IIforpreliminary experiments wereobtainedfrom Sigma Chemical Co., St. Louis, MO. Theapolipoproteins were later purified from HDL (prepared by se-quentialflotationultracentrifugationof human plasma).HDLwas dia-lyzedinto 5mMammoniumbicarbonate, 1mMEDTA, pH 8.0, and lyophilized. Apo HDL was prepared bytwo extractions withdiethyl

ether, and fiveextractions with ethanol/ether (3:1)asdescribed by Os-bqrne(26),dried undernitrogen, and dissolved in 3MguanidineHCl,

26

mMTris, pH 8.0. The HDLapolipoproteinswerechromatographed onSephacryl S200 (2.6 x 85cm)yieldinganelution pattern similarto

thatpreviously reported with Sephadex G200(27). ApoA-IandA-II

peakswerefurtherpurifiedbyreverse-phaseHPLC(SynChropakRP-P

C18,250X21.2 mm;SynChrom, Inc.,Lafayette, IN)usingagradient

of acetonitrile(inwatercontaining 0.1% trifluoroaceticacid)from 25% to58%at 1.0%/min,20°C,flowrate8ml/min(28). Apolipoproteins

werelyophilizedandstored at-70°C.Before use, theapolipoproteins

weredissolved in 3 Mguanidine, 0.02 MTris,pH8.0, anddialyzed

into PBS. ApoA-I and A-II were iodinatedusingimmobilized

lactoper-oxidase (Enzymobeads, Bio-RadLaboratories,Richmond,CA)to spe-cific activities of 500-2,660 and 1,700-4,900 cpm/ng, respectively.

Afteriodination, purity was assessed by SDS-PAGE followed by

autora-diography; in the absence ofdisulfidebondreduction,apoA-Iand A-II yieldedsingle bands of molecularmass28 and 17kD,respectively.

PreparationofLPDS. LPDS was prepared byultracentrifugationof human serum atdensity 1.23(adjustedby the addition of solidKBr);

the infranatewasdialyzedagainst HBSSto removeKBr.Alternatively,

human serumwasabsorbedwithLiposorbaccordingtothe manufac-turer's instructions. LPDSwas heat-inactivated at 56°Cfor 30 min beforeuse oncultured cells.

Exposureof endothelial cellstoactivatedcomplement. Thebuffer solution used for all endothelial cellexperimentsconsistedof HBSS

containing 10 mM Hepes and 1% bovine serum albumin (HHBSS-1.Abbreviations used in this paper:

[3H]DOG,

2-deoxy-D-[3H]glucose;

LPDS,lipoprotein-deficient human serum.

BSA). Cultured endothelial cells were exposed to various stages ofcom-plementactivation by sequential incubation at370Cwith antiendothe-lialIgG (4 mg/ml) for 15min,25%C8-deficient serum for 10min,and human C8 and C9 for 10 min at the concentrations indicated in the figure legends. In some experiments, the above buffer with 0.1 mM EGTA and nocalcium was used for the C8 and C9 incubations, as detailed in thefigure legends.

Lipoproteinbinding experiments. After two washes to remove com-plement components, cell monolayers were incubated with '25I-HDL

for 2 hat370C. Cellswerethenwashedsix times rapidly with HHBSS-BSA at4VC to remove unbound radioactivity, solublized in 4% SDS, and counted in a y counter(Beckman Instruments, Inc., Fullerton, CA).Nonspecific binding was determined by parallel incubations in the presence of 50-fold excess of unlabeled HDL, and was always

- 50%of total binding. Specific binding, defined as total bound

radio-activityminus nonspecific, is reported in the figures. Binding experi-mentsusing apo A-I and A-II were performed identically except that duration of binding was 10-40 min as indicated in the figure legends, andnonspecific binding was determined in the presence of a 15-fold

excessof unlabeled apolipoprotein.Attheconcentrations used in these experiments, nonspecific binding for apo A-I was - _{45%of total in}

untreated cells and decreased to - 30% inC5b-9-treatedcells;for apo

A-II,nonspecificbinding was - 45%of total in control cells and only

20%oftotal inC5b-9-treated cells.

Quantitation ofendothelial membrane permeabilization. Mem-brane permeabilization by C5b-9 was monitored using 2-deoxy-D-[3H]glucose([3H]DOG)release(29). Cellswereloaded by incubation with[3H]DOG10

_ACi/ml

(2.75gCiper well)for24h.After washingto

remove extracellular radioactivity, C5b-9wasdepositedasdescribed above.Mediawereremovedafter 10min, pooledwithtwowashes,and

subjectedto scintillation counting. Resultsare expressedaspercent reduction by potential lipoprotein inhibitors ofC5b-9-induced

[3H]-DOG release (counts per minute released from C5b-9-treatedcells minuscountsper minute releasedfrom untreated cells).

Results

HDLandapolipoprotein bindingtoCSb-9-treated endothelial cells.HDLaresmall, spherical or discoid particles containing about one halfprotein and one halflipidby weight.The protein components include apo A-I (MR 28,000; 60-70%), apo A-II (MR 17,000;20-30%), and small amounts of apo C and E (30). The role of theseapolipoproteinsin theinteraction ofHDL with cell surfaces is not known, but these apolipoproteins con-tain extensiveregionsofamphipathica-helix which have high affinity for phospholipids. Apo A-I dissociates readily from

HDL, and "free" apo A-I is found in both plasma and the interstitialspace(30).AsillustratedinFig. 1,binding of immu-noglobulin or complement activation through C8, does not alter thebindingof either of the HDLapolipoproteins (A-Ior A-IT)tothe endothelial surface.Bycontrast, addition of C9 (to membrane C5b-8)wasfound toincreasebindingof both apo A-IandA-II,suggestingaspecific effect ofthe assembledCOb-9 complex on the exposureofmembranebindingsites for these apolipoproteins.

Apolipoprotein binding

to C5b-9-treated cellsincreased from 2-to50-fold above

binding

tountreated cells. Thisvariabilitywas inpart due to thelargevariabilityin specificbindingobserved for untreated cells. In additionto a requirement forC9, theexposure ofspecific membrane bind-ingsites forapoA-IandA-IIwasalsofoundtoincrease as a

function of the numberof cellsurface C5b-9complexes, gener-atedbythe additionof

limiting

amountsof C8to

preformed

membrane C5b67(Fig. 2).Increasedbindingtohuman

0.1 0.2 0.3 0.4

1.

I0E on 0.

<:

E

Control IgG C8D C5b67 C5b-8 C5b-9

Figure1.Effects ofcomplementdepositiononbindingofapoA-I and

A-IItohuman endothelialcells. Complementwasactivated,as

de-scribedinMethods, by sequentialexposureofcellstoantiendothelial IgG, C8-deficientserum,and humanC8(0.2 Mg perwell) and C9 (1

Mg perwell).Aftercomplement activation,cellswereincubated with

either

1251-apo

A-I(10 ug/ml, A),or'251I-apoA-Il (10ug/ml,B)at

370C for 30min.Effects ofexposuretoantibodyonly (IgG),

C8-defi-cientserumonly (C8D), antibody plus C8-deficientserumto

assem-bleCSb67,C5b67 plus C8 (CSb-8), and C5b-8plusC9(Cb-9)are

shown.Dataarespecific apolipoprotein binding(mean±range,n

=2),andarerepresentative of four experiments performed. Note the

difference in scale for AandB.

antibody and complement activation through C8, but in-creased withaddition of C9tothe membraneCOb-8complex, suggesting aspecific requirement for the membrane-inserted C5b-9 complex. The similarity of the results obtained with HDLand withapoA-IandA-II (cf. Figs. 1-3)suggeststhat the binding ofHDLtotheCOb-9-inducedmembranesites is

me-diatedby the apolipoproteincomponentsofthese particles. Relationship ofexposureofapolipoprotein bindingsitesto

endothelialstoragegranulesecretion. Inadditiontoincreasing cell surface binding of HDL and its apolipoproteins (above), COb-9 has been showntoinduce secretion of endothelial

stor-agegranules (17), raising the possibility that the expression of

thesenewbinding sites is relatedtothefusion ofWeibel-Palade bodies with theplasma membrane. Several other agonistsare

also knowntoactivate endothelial cell secretion, including his-tamine, thrombin, the phorbol ester phorbol 12-myristate 13-acetate (PMA), and the calcium ionophoreA23 187 (31). By

contrast toCOb-9,when endothelialcellswereexposedtothese agonists at concentrations that activate maximal von

Wille-brand factorsecretion, noincrease in specific binding ofapo

A-II wasobserved (Fig. 4). Similar results wereobtained for

binding ofapoA-I(datanotshown). These resultssuggestthat neitherpermeabilization of the plasma membranetocalcium

CD

0

E

m

l 0

<

E

_CIL

o

O:

0.2

-/

B _ _ _ _ _

,-0.4-I

0.3-I

0

0.2-0.1

0.0

0.0 0.1 0.2 0.3 0.4

C8 Input

_(tg/well)

Figure2.Effect of number of membrane C5b-9complexesonbinding

ofapoA-IandA-II. Complementwasactivatedby sequential expo-sureof cellstoantiendothelialIgG,C8-deficientserum,andhuman

C8 and C9. Cellswerethen incubated with either'25I-apoA-I(10 ,gg/ml,A)or'251-apoA-II(10 Mg/ml,B)at370Cfor 20min.

Mem-branedepositionofC5b-9complexeswascontrolledby limitingC8

input,asindicatedonthe abscissa. C9inputwas2 Mg perwell(16 ug/ml). Specificbindingdata shownarethemeanofduplicate deter-minations;errorbars indicaterange.Of note,errorbarsarefrequently

obscuredby symbols.Dataarerepresentativeof threeexperiments performed.Note the differentscales forapoA-I andA-II.

(asoccursafter A23187), release of intracellular calciumstores

(as induced by histamine and thrombin),noractivation of

pro-tein kinaseC(by PMA) is sufficienttoexposetheplasma

mem-branebindingsitesforthe HDLapolipoproteins. These data

alsosuggestthat the increaseinbinding ofthe HDL apolipo-proteins observed inCOb-9-treated endothelial cells is

unre-latedtoinduced secretionof endothelial storagegranules.

Con-sistent with this interpretation, the C5b-9-induced binding sites forapoA-IandA-IIwereobservedtopersistonthe

endo-thelial surfaceduring prolongedincubationat 37°C (Fig. 5). This contrasts the transient cell surface exposure and rapid reinternalization of the granule membrane-derived protein, GMP-140, that isobserved in association with endothelial

se-cretion induced under these conditions(17).

Effect ofplasma membrane cholesterol loadingon

apolipo-proteinbinding.It has beenreported that thebindingofHDL and apo A-I and A-IV to bovine endothelial cells increases

markedly when the cellsarepreincubatedwithcholesterol,

ei-ther in the form ofLDL, acetylated LDL,or 25-hydroxycholes-terol(3, 8, 9).The increasedbindingdoesnotfollow accumula-tion of esterified cholesterol but requires an increase in free

cholesterolinthe membrane (32).Inourpreliminary

experi-ments,weconfirmedthiseffect ofcholesterolonHDLandapo C:

_i5>

0.2

l ₀

< E

0o1

a

0%.O

0Q

0.

0

o 0.08-0

<

EO0

a.0.044 o

0.0

A0

0

n6-5 Figure3. Increased HDL

bind-C

I

A

B

+c

ing to human endothelial cells

o

4X-

exposed to

C5b-9.

(A)

Endothe-mT hlialcellswereexposedtobuffer

C14

4-3t

only (Control),antiendothelial

I

Ig|GT

I

I

I

csz~r

i/

No C9

IgG followedby C8-deficient

2 I humanserumtodeposit

CSb7,

.S 2 or_IgG,C8-deficientserum,and

m I then humanC8 (0.2

Ag

per well,

U) 1.6

Ag/ml)

and C9(2(g per well,

C - 0 I _l16

_ug/ml)

to depositCb-9.After

Control

_Control

C5b67

_{C~~b67 C~~b-9}

C5b-9

0.0 0.1 0.2 0.3 0.4

_{washing to remove complement}

08

input

(azg/weil)

components,

cellswere

incu-bated for 2 h with'251I-HDL(30 ,Mg/ml)with andwithouta50-fold molarexcessof unlabeled HDL.Specific bindingisshown,andisrepresentativeoftwoexperimentsso per-formed. Data aremean±SD, n=3.(B) Endothelial cells were exposedtoantiendothelial IgG and C8-deficientserum todeposit C5b67. The numberofC5b9complexesdeposited was controlled by adding human C8 in the amounts indicated on the abscissa, in the presence of human C9 at 2Mugper well(-)or0MAgper well (o). Data shown aremean±SD, n=3.

A-Ibindingto human endothelial cells, and found a similar increase in apo A-II binding. In order to determine whether cholesterolloading and C5b-9 deposition might synergistically alter HDLapolipoprotein binding,weexaminedtheeffect of cholesterol on C5b-9-induced exposure of these endothelial binding sites (Fig.6).Asillustratedby these data, theeffects of cholesterol and C5b-9 areapproximately additive, suggesting thatplasma membrane cholesterolcontent does notinfluence the

magnitude

of C5b-9-induced

binding

of these

apoproteins.

Tobenoted, distinct morphologic changeswereobservedin the human endothelial cells incubated with 10 Mg/ml 25-hy-droxycholesterol,eventhoughconcentrations of40-100

ttg/ml

have beencommonly usedtoenhanceHDL and apoA-I bind-ing (3, 8). Moreover, C5b-9 assemblywaspoorlytolerated by thecholesterol-loaded cells, necessitatingadecreaseinC8 con-centration(0.05

gg

perwell)toavoidlysisanddetachment of the cells usedintheseexperiments.

Requirementfor calcium. C5b-9 induces the secretion endo-thelialvWFthroughapathwaythatrequiresextracellular

Ca2",

and can be abolished byEGTA(17). In order todetermine

-o

o E

m\

I

E

<3

a 0

0-1.0

0.0- I E .

0.5

0.0

r5-

r

m--

m-- ii-- _

Control PMA A23187 His Thr C5b-9

Figure4.Comparison of C5b-9tootheractivators of endothelial

se-cretion. Endothelial cellswerewashed,and treatedsequentiallywith antiendothelial IgGand C8-deficientserum(C5b-9-treatedcells)or

bufferonly (controland otheragonists).Afterwashing,125I-apo A-II (10 Mg/mlwithand without 15-foldexcessofunlabeledapoA-II)was

added,followedby agonistsatthefollowingconcentrations:PMA,

10-7M; A23187,

l0o`

M;Histamine(His),

lo-'

M;thrombin(Thr)

U/ml;C8(0.2 Mgperwell)and C9(1lgperwell)(C0b-9). Specific bindingwasmeasured after 30minat370C.Von Willebrand factor secretion intothesupernatantwasassayedandwasgreaterthan five times control for allagonists.Data shownaremean±range,n=2.

whethercalcium influx is requiredfor theCOb-9-induced gen-eration of newbinding sitesforapoA-IandA-II,wecompared binding to C5b67-bearingcells whenC8, C9,and

'25I-apolipo-protein wereadded in the presence of calciumorEGTA. As

0.2

-o

Q3

0.1±

o

o

E

1.0

NI-I

Q0

15 30 45 60

B C5b-9

0 ~~ ~

Control

.

----.--.

6 _1v.

Time

(min)

45 60

Figure5.Stabilityof theC5b-9-inducedapolipoprotein bindingsites.

Endothelial cellswereincubatedwith antiendothelialIgG,

C8-defi-cient humanserum,and then C8(0.2 Mgperwell) plusC9(1 Mg per

well) (Cb-9)orbufferonly(Control)for 8 min.Cellswerethen

washedtoterminatecomplementdeposition and remainedinbuffer

until additionof'251-apoA-I(A)or'25I-apoA-II (B)at0, 25,and50

minaftercompletionof C5b-9assembly.The'25I-apoproteinswere

usedat 10 Mig/mlinthepresenceand absence of 15-foldexcessof

unlabeledapoprotein.Duration ofbindingwas 1O min,and dataare

shownatthemidpointofthebindinginterval(mean±range,n=2).

Note the different scalesfor A and B.

A

I

-0

C5b-9

I

0

C:

cs-

2-O E

m

_OE

<0a

0_

0-0 10 20

Cholesterol

(,ag/ml)

Figure 6.Effect of cholesterolloadingonthe C5b-9-induced re-sponse.Endothelialcellswerepreincubatedfor24 hwith culture me-dium containing 25-hydroxycholesterolattheconcentrationgiven ontheabscissa. Cellswerethen washed five times and treated

se-quentiallywith antiendothelialIgG,C8-deficientserum,and C8(0.05

Mg perwell,0.4Ag/ml)andC9(1I Mgperwell,8,ug/ml).Control cells

wereincubated with buffer onlyateachstage.Aftertwowashesto

removeC8 and C9,cellswereincubated for 30minwith '25I-apoA-Il

(10,g/ml)for 30minat37°C.Data shownarespecific binding (mean±range,n= 2),andarerepresentativeoftwoexperiments

per-formed.

shown in Fig.7A, the C5b-9-induced increaseinapoA-I bind-ingwasreduced - 80% by EGTA. In general,_asmallereffect

of EGTAon the C5b-9-inducedbindingofapo A-IIwas

ob-served(Fig.7B). These datasuggestthat extracellularcalcium,

orthe influx of calcium into the cytosol, contributes to full expression of the C5b-9-induced apolipoprotein binding sites, particularlyinthecaseofapoA-I.Bycontrast, calcium does

notappeartoinfluence the interaction of these apolipoproteins with the untreated endothelial cell.

The calcium requirement for the C5b-9-induced apolipo-proteinbinding demonstrated in Fig.7maybe duetothe cal-cium influx required for cell activation by C5b-9ortoaneffect of calcium on _{the interaction of the apolipoprotein with the}

newly expressed binding site. In ordertoexaminethese effects,

weperformed C5b-9 assembly from C5b67 sites, andapoA-I bindinginseparate stages;theeffects ofEGTAon(a) cell

"acti-vation"byC5b-9, and (b)apoA-Ibindingcould then be

ob-served. As demonstrated by the data in Table I, calcium ap-pearstobe importantatthe time ofcomplement depositionon

the membrane. Once the complementporehas been formed in

thepresenceofcalcium, apolipoprotein binding is unaffected by removal of calcium. These data suggesta calcium

depen-dence of expression of the binding site and donot supporta

calcium dependence of ligand interaction with the site. Effect of C5Sb-9 on internalization of apolipoprotein. Our binding studieswereperformedat37°C, andmaypotentially

reflect both surface bound and internalized ligand.Inorderto determine whether the increases inapoA-IandA-II bindingto C5b-9 (vs.untreated)cells(demonstratedin Fig.1,4,5,and7)

reflected an internalization of the plasma membrane-bound

apolipoprotein, the cellsweresubjectedto atryptic digest,to distinguish surface-bound from internalized ligand. In the C5b-9-treated cellsincubatedwithligandat37°C,the

percent-ageofcell surface-boundapoA-Iremained essentiallyconstant with time (67% trypsin-releasableat10minvs.69%at40min, Table II). These datasuggestthat the 2-to30-foldincreases in

specific apo A-I binding is not accounted for by increased membraneinternalizationin the C5b-9-treatedcells.

:U

0~

0 ,

Control

C5b-9

Figure 7. Requirement for calcium.C5b67 was assembled by incu-bating endothelial cells with antibody and C8-deficient serum. Cells were then washedtwice in buffer containing either 2 mM calcium (openbars) or 0.1 mM EGTA (hatched bars), followed by addition of this same buffer without(Control)orwith C8 and C9(0.2and2 tg per wellrespectively, COb-9); '25I-apoA-I(20 tg/ml, A)or

'231-apo

A-Il (10

,g/ml,

B) was also included.Bindingproceeded for 20min

at370C. Data shown are specific binding, mean±SD, n = 3, and are representative of three similarexperiments.Notethe different scales for apoA-IandA-Il.

Competition betweenapo A-Iand A-IHfor the binding sites onCSb-9-treatedendothelial cells. Apo A-I and A-I1exhibit significant structural homology and both contain extensive

re-gions ofamphipathichelix whicharebelievedtobeimportant in the phospholipid binding and lipid transport function of these proteins (33). Because both apolipoproteins exhibited similar enhancement in bindingtoC5b-9-treated endothelial cells,wewishedtodetermine whether this binding occurredat

TableL CalciumRequirementforExpressionofC5b-9-induced ApoA-IBinding Sites:

Effects

ofEGTAonCell Activation by CSb-9 andonApo A-IBinding

C5b-9assembly Apo A-I

from C5b67 sites* bindinge Apo A-Iboundl

pmol/cm2

Ca Ca 0.159±0.006

Ca EGTA 0.151±0.020

EGTA EGTA 0.061±0.006

* C8 (0.2,ggperwell)andC9

(1

Mg

per well) ineither calcium2 mM orEGTA 0.1 mM wereadded to cells bearing preformed C5b67 sites andincubated for 10 min at 37°C to assembleCOb-9.Cells were then washedtwice with the calcium- or EGTA-containing buffer to

beusedfor apoprotein binding. tBindingof'25I-apoA-Iwas per-formed for 15 minat37°C in the presence of calcium 2mMor EGTA0.1 mM, after termination of C5b-9 assembly. § Results are mean±range,n =2.

* C5b-9

o Control

0,_---00

Table II. Accessibility to Trypsin ofApo A-I Bound to

CSb-9-treatedEndothelialCells

ApoA-I bound

Duration of Trypsin- Percentage of binding Total releasable trypsin releasable

min pmot/cm2

10 0.255±0.021 0.172±0.006 67 40 0.495±0.023 0.340±0.028 69

Endothelial cells were treated with antiendothelial IgG and C8-defi-cient serum, and then incubated with C8 (0.2

_Ag

per well, 1.6

_jug/ml),

C9(1 ug per well, 8Mlg/ml)for 10min to depositC5b-9. Afterwashing

twice, '25I-apoA-I (10 ,ug/ml, with and without a 15-fold excess of unlabeledapoA-I)was added and incubated for 10 or 40min.Cells were thentransferred to ice and rapidly washed six times. In some wells, cells weresolubilized in SDS to yield specific apo A-I bound. Inother wells,cells were incubated with trypsin(1 mg/ml)at4VCfor 75min;trypsinization was terminated and wells were washed with soybean trypsininhibitor(2 mg/ml). After microcentrifugation, supernatants werecounted to yield the trypsin-releasable fraction. Data shown are mean±range, n = 2.

thesamemembranesite.AsshowninFig.8 A, apoA-II effec-tively competed for the125I-apo A-Ibinding sites exposed in C5b-9-treated cells.Likewise,apo A-Ieffectivelycompeted for the

'25I-apo

A-IIbinding sites (Fig.8 B), althoughinall experi-ments, apo A-II was observed to competebindingofthe radio-labeledapolipoproteins (apoA-Ior apo A-II) moreeffectively thanapo A-I. Similar results were obtained forcontrol cells (datanotshown).To benoted,apo A-II has been reported to associate with lipidmorerapidlythan apo A-I,andtodisplace apo A-IfromHDLparticles (34).

Effect ofHDLandapo A-IandA-II ontheC5b-9pore. As notedabove, ithasbeenreported that apo A-I and A-IIinhibit C5b-9-induced lysis of human erythrocytes, suggesting that these

apoproteins

inhibit assembly of theC5b-9 pore(22, 23). Inorder todetermine whether theHDLparticleorits apolipo-proteins inhibits C5b-9assembly on theendothelialsurface,we examinedtheeffect ofHDLand apo A-I onC5b-9-induced

permeabilization

of the plasma membrane. In these

experi-ments,theinhibition ofC5b-9-induced release of

[3H]DOG

by HDLandapolipoproteinwasmeasured. Under theconditions ofour

experiments,

neitherHDL(at concentrations upto 1 mg/ml)norapo A-I(at concentrationsup to 100 ,ug/ml)were found toinhibit

significantly

C5b-9 pore formation: percent inhibitionbyapo A-I 100

_jtg/ml,

10±16%(NS);inthe presence ofHDL 1 mg

protein/ml,

there was noinhibition,but rather a smallincrease in [3H]DOG release (21±8%, NS).By contrast,it has been reported that HDL(at 30 ug/ml) and apo A-I (at 1.5-9.5

,g/ml)

causes50%reductionin

COb-9-induced

lysis of humanerythrocytes (22).

Discussion

Theseexperiments establish that membrane assembly ofthe

C5b-9

complexcanlead toa

significant

increase inthe

binding

ofHDLandapo HDLtotheendothelialcell

surface,

suggest-ingapossible mechanism by which cholesterol metabolism of the vessel wall

might

bealteredby intravascularcomplement activation.To our knowledge, this response to Cb-9 is the first

A

0 2

-C5b-9

4 6

Q-Eo

E.2-I

<aC

c

Control

LO

0.0

0 2 4 6

C:

~~~B

0

m

-AA-o

0.5---

_---

_&+A-l

<a

8-fol exes

Control

0.0

0 12 3

[Competing

Ligand] (MuM)

Figure8.ApoA-IandA-I1compete fora commonbindingsiteon

COb-9-treatedendothelial cells.(A)

12.31-apo

A-I(20

Mg/ml,

0.7

_MIM)

bindingtoC5b-9-treated cells is shown in thepresenceof 0-to8-fold

excessof unlabeledapo A-IandA-Il.(B)

1251-apo

A-Il(6

jig/in1,

0.35

MzM)bindingtoCOb-9-treatedcellsis shown in thepresenceof 0-to

8-foldexcessof unlabelledapo A-Iand A-II.Bindingtocontrol cells is shownfor comparison. Datashownaremean±range,n= 2. Note thedifferent scales forAandB.

evidence fora direct effect ofan inflammatory mediatoron lipoprotein bindingtothe endothelial surface.

Itis ofinterestto notethat theonlyreported mechanism for increasing the binding ofHDL(and its

apolipoproteins)

to en-dothelial cells is through increased concentration offree choles-terolinthe

plasma

membrane. Since

increasing plasma

mem-brane cholesterol concentration and membrane insertion of theC5b-9

proteins

areboth knowntoalter the arrangement of membrane

phospholipid,

it is

possible

thattheincreased

bind-ing

ofHDL and HDL

apolipoproteins

observed under these conditionsreflects formation of

lipid

domainsthat favor inter-calation of these

amphipathic proteins

into theplasma

mem-brane.

Alternatively,

C5b-9 may exposea

protein

receptor for apo A-I and A-IIontheendothelial

surface, perhaps by

induc-ing translocation ofsuch receptorstothe

plasma

membrane from an intracellular

pool

or

by inducing

a conformational changein

preexisting

receptors. Of note,an

HDL-binding

pro-tein with

affinity

for apo A-I andA-IIhas been described in fibroblasts and bovine aortic endothelial cells

(10), although

evidence

against

a

protein

receptorforHDLhasalso been pre-sented(35).

Areview ofthe known effects of the C5b-9

complex

on

of protein(s) derived from the granule membrane(17). How-ever, fusion of the secretory granules with the plasma mem-brane cannot befully responsible for the observed increase in HDLbinding,because otheragonists which stimulate secretion did not increasebinding (Fig. 4). COb-9 also induces endothe-lial membranevesiculationandexpression ofnewbinding sites for factor Va, an effect that has been related to the surface exposure ofphosphatidylserine andothernegatively charged phospholipidunder theseconditions (18). Althoughanaffinity of apo A-I orA-1I for acidicphospholipids mightunderlie the observed increase in binding to C5b-9-treated endothelial cells,it isofnotethatmostofthemembrane

binding

sitesfor factor Va(andpresumably, mostof theexposedacidic phos-pholipid)arevesiculated fromtheendothelialsurface, andare unlikelyto accountforthe

binding

ofHDL tothecell mono-layers that we report here. As HDL andits apoproteinshave beenreported toinhibit COb-9-induced hemolysis, direct bind-ing oftheselipoproteinstocomponentsoftheCOb-9 complex canalso beconsidered (22, 23).Asnotedabove, with endothe-lial cells we were not able to demonstrate aninhibitory effectof apo A-I or HDL onCOb-9pore

formation,

and wereunableto inhibitapo A-IIbindingtoCOb-9-treatedcellswitha polyclo-nalantibodytoC9(data not shown).Nevertheless,the possibil-ity that these lipoproteins bind directly to the membrane-boundC5b-9proteinscannotbe excluded.

Thefunction ofHDLisnotfully understood,andthe

physi-ologic and

pathophysiologic

consequencesof enhanced HDL binding to vascular endothelium remains tobe determined. HDLand apo A-I havereceived attentionbecause

epidemio-logicdata suggests aninverserelationship between theirplasma levels and the risk ofcardiovascular disease in industrialized countries (36,

37).

Whether HDL hasadirect

protective

effect againstatherogenesisorisamarkerforthe

efficiency

of other metabolicprocesses (e.g.,clearanceof triglyceride-rich lipopro-teins) is controversial

(for

discussion,

seeTall[30]). HDL are believedto

play

arole intransport of cholesterol from

periph-eralcellstotheliver (30),and thusenhanced

binding

to endo-thelialcells

might

resultin increasedcholesterolefflux. Alterna-tively, sinceapo A-Iisapotentactivatoroflecithin:cholesterol acyl

transferase,

thepossibility existsthat theCOb-9-induced apo A-Ibinding to theendothelial plasma membrane

might

result in theaccumulationofcholesterol ester inthecell. Fur-thermore, HDL are heterogeneous, andit isnow recognized that a significant population ofapo A-I-containing lipopro-teins exists outsidethe HDLdensityrange(38, 39).Theeffect oncholesterolhomeostasisorothercellularfunctionsmay de-pendupon the subclassofapo

A-I-containing

lipoproteinthat is actually boundtothe plasma membrane:forexample, there is evidence in vitrothatbinding ofHDLcontainingapo A-I but notA-II,Lp (A-Iwithout

A-II),

mayactivate cholesterol efflux, whereasLp (A-Iwith

A-II)

bindsbutfailstoactivate this pro-cess(5).Thus,theC5b-9-induced increasedHDLbindingmay serve to activate or inhibit cholesterol efflux, if a particular subclassis preferentiallybound.

Inadditiontoits potential role in reversecholesterol trans-port,apo A-Ihas recentlybeenshowntocirculatein associa-tionwithacomplementinhibitor(clusterin,complementlysis inhibitor, SP-40,40, sulfatedglycoprotein2,apolipoproteinJ;

40-42).

By augmenting apo A-I binding,

C5b-9

deposition could increase the local concentration ofan inhibitor which prevents further

C5b-9

assembly on the cell surface, perhaps

limiting injury.

Since HDL and apo

A-I/phospholipid

com-plexeshave been showntobindprostacylin (43, 44),an

inhibi-torofplatelet aggregationthat isproduced byendothelium in responseto C5b-9 and other stimuli(45), the possibility also existsthat bound HDL may alter the antithrombotic properties of theendothelial surface.

The interrelationshipbetween complement activation and atherogenesis remains unresolved. Detection of C5b-9 com-plexesinatherosclerotic plaques (19, 20)and the association of anti-HLA antibodies with thedevelopmentofaccelerated

coro-nary artery disease intransplanted hearts (46) suggests that the humoral immune response may playarole in the initiationor

progression oftheatherosclerotic lesion. Our finding that the C5b-9 proteinsinducebindingofHDLtothe endothelial

sur-face suggest one mechanism by which intravascular comple-mentactivation mightalter cholesterol homeostasis, and

con-tribute to the development of atherosclerosis.

Acknowledgments

Theauthorswishtothank Janet Heuser and Darlene Schwartzott for

their excellent technicalassistance,andTommieHoward for helpwith preparationof themanuscript.

This research was supported bygrantsHL-01749, HL-3606 1, and HL-36946 from theNational Heart, Lung and BloodInstitute,

Na-tionalInstitutesof Health.

References

1.Tauber, J.-P., D.Goldminz, I. Vlodavsky, and D. Gospodarowicz. 1981.

Theinteraction ofthehigh-density lipoprotein with cultured bovine vascular endothelium.Eur.J.Biochem. 119:317-325.

2. Havekes, L., D.Schouten,V.vanHinsbergh, andE.de Wit. 1984. Charac-terization ofthe bindingof apolipoprotein E-free high density lipoprotein to

cultured humanendothelial cells. Biochem. Biophys. Res. Commun. 122:785-90.

3.Savion,N.,and A.Gamliel. 1988. Binding of apolipoproteinA-Iand apoli-poprotein A-IV to cultured bovine aortic endothelial cells. Arteriosclerosis. 8: 178-186.

4.Brown, M.S.,Y. K.Ho, and J.L.Goldstein. 1980. The cholesterylester

cycle in macrophagefoam cells: continual hydrolysis and re-esterification of cyto-plasmiccholesteryl esters.J.Biol. Chem. 255:9344-9352.

5.Barbaras, R., P. Puchois, J.-C. Fruchart, and G. Ailhard. 1987. Cholesterol efflux from cultured adipose cells is mediated by LpA1 particles but not by LpA,:Al particles.Biochem. Biophys. Res. Commun. 142:63-69.

6.Slotte,J. P., J. F.Oram,and E.L.Bierman. 1987. Binding of high density

lipoproteins to cell receptor promotestranslocation of cholesterol from

intracel-lularmembranes to the cellsurface. J.Biol.Chem. 262:12904-12907. 7.Hara, H., and S. Yokoyama. 1991. Interaction of free apolipoproteins with macrophages: formation of high density lipoprotein-like lipoproteins and reduc-tion ofcellular cholesterol.J.Biol.Chem. 266:3080-3086.

8.Tauber, J.-P., D. Goldminz, and D. Gospodarowicz. 1981. Up-regulation in vascularendothelial cells ofbinding sites of high density lipoprotein induced by 25-hydroxycholesterol.Eur.J.Biochem. 119:327-339.

9.Brinton, E. A., R. D. Kenagy, J. F. Oram, and E. L. Bierman. 1985. Regula-tion of high density lipoprotein binding activity of aortic endothelial cells by treatmentwith acetylated low density lipoprotein. Arteriosclerosis. 5:329-335.

10.Graham, D. L., and J. F. Oram. 1987. Identification and characterization ofahigh density lipoprotein-binding protein in cell membranes by ligand blot-ting. J. Biol. Chem. 262:7439-7442.

1 1.Matz, C. E., and A. Jonas. 1982. Micellar complexes of human apolipo-protein A-I withphosphatidylcholines and cholesterol prepared from cholate-lipid dispersions. J. Biol. Chem. 257:4535-4540.

12. Swaney, J. B. 1983. Reconstitution of apolipoprotein A-I from human high density lipoprotein with bovine brain sphingomyelin. J. Biol. Chem. 258: 1254-1259.

13.Epand, R. M., J. P. Segrest, andG. M. Anantharamaiah. 1990. Thermody-namics of the binding of apolipoprotein A-I todimyristoyl-phosphatidylglycerol.

J. Biol. Chem. 265:20829-20832.

14.Esser, A. F., W. P. Kolb, E. R. Podack, and H. J. Muller-Eberhard. 1979. Molecular reorganization of lipid bilayers by complement: a possible mechanism for membranolysis. Proc. Natl. Acad. Sci. USA 76:1410-1414.

proteins C5b-9 induce transbilayer migration of membrane phospholipids.

Biophys.J.56:935-946.

16. Esser,A.F. 1990. The membrane attack pathway of complement. In Molecules andCells of Immunity.J. M. Cruse and R. E. Lewis,Jr.,editors. Year

Immunol. 1989-90. 6:229-244.

17.Hattori,R., K. K.Hamilton,R. P. McEver, and P. J. Sims. 1989. Comple-mentproteins C5b-9 induce secretion ofhigh molecularweightmultimers of

endothelialvonWillebrand factor and translocation ofgranule membrane

pro-tein GMP-140tothe cellsurface.J.Biol. Chem.263:9053-9060.

18.Hamilton,K. K., R.Hattori,C. T. Esmon, and P. J.Sims.1990. Comple-mentproteins C5b-9 induce vesiculationof the endothelial plasma membrane

andexposecatalytic surface for assembly of the prothrombinaseenzyme

com-plex.J.Biol. Chem.265:3809-3814.

19. Vlaicu, R.,F.Niculescu,H.G.Rus, and A.Cristea.1985.

Immunohisto-chemicallocalization of the terminal C5b-9 complement complex inhuman

aorticfibrous plaque. Atherosclerosis. 57:163-177.

20.Seifert, P. S., and G. K. Hansson. 1989. Complement receptorsand regula-toryproteins in human atherosclerotic lesions. Arteriosclerosis. 9:802-811.

21.Seifert, P. S., F. Hugo, G. K. Hansson, and S. Bhakdi. 1989. Prelesional complement activation in experimental atherosclerosis: terminal C5b-9

comple-mentdeposition coincides with cholesterol accumulation intheaortic intima of hypercholesterolemic rabbits. Lab.Invest.60:747-754.

22.Rosenfeld, S.I.,C.H.Packman,and J. P.Leddy. 1983.Inhibition ofthe

lytic action of cell-bound terminal complementcomponents by human high

density lipoproteins and apoproteins. J. Clin.Invest.71:795-808.

23. Packman, C. H.,S. I. Rosenfeld,and J. P.Leddy. 1985. High-density lipoprotein and its apoproteins inhibit cytolytic activity of complement. Studies

onthenatureof the inhibitory moiety. Biochim. Biophys.Acta.812:107-115.

24.Wiedmer,T., and P. J.Sims. 1985. Effect ofcomplement proteinsC5b-9

onbloodplatelets: evidence for reversible depolarization of membrane potential.

J.Biol. Chem. 260:8014-8019.

25. Havel,R.J., H.A.Eder, and J.H. Bragdon. 1955.Thedistributionand

chemical composition of ultracentrifugally separated lipoproteins in human

serum.J.Clin.Invest.34:1345-1353.

26. Osborne, J. C. 1986. Delipidation of plasma lipoproteins. Methods

Enzy-mol. 128:213-222.

27.Edelstein, C., C.T.Lim, andA. M.Scanu.1972.On thesubunitstructure

of theproteinof humanhigh density lipoprotein. I.Astudy of its major polypep-tidecomponent(Sephadex, Fraction III).J.Biol. Chem. 247:5842-5849.

28.Anantharamaiah, G. M.,T. A.Hughes,M.Iqbal,A.Gawish,P.J.Neama, M. F.Medley, andJ. P.Segrest. 1988. Effect of oxidationontheproperties of

apolipoproteinsA-Iand A-II.J.LipidRes.29:309-318.

29.Andreoli, S. P.,R. L.Baehner, and J.M.Bergstein. 1985.Invitro detec-tion ofendothelial damageusing2-deoxy-D-3H-glucose:comparison with chro-mium 51, 3H-leucine,3H-adenine, and lactate dehydrogenase.J.Lab. Clin. Med. 106:253-261.

30.Tall,A.R.1990. Plasmahigh density lipoproteinsMetabolismand rela-tionshiptoatherogenesis.J.Clin.Invest.86:379-384.

31.Hattori, R.,K. K.Hamilton,R. D.Fugate, R.P.McEver, andP. J.Sims. 1989.Stimulatedsecretionof endothelialvonWillebrandfactor isaccompanied

by rapid redistributiontothe cellsurface oftheintracellulargranulemembrane protein GMP-140.J.Biol. Chem. 264:7768-7771.

32.Oram,J. F., E. A.Brinton,and E. L.Bierman. 1983. Regulation of high density lipoproteinreceptoractivity in culturedhuman skinfibroblasts and hu-manarterial smooth muscle cells. J. Clin.Invest.72:1611-1621.

33.Segrest,J.P.,R. L.Jackson,J. D.Morrisett, andA.M.Gotto.1973.A

molecular theory of lipid-protein interactions in the plasma lipoproteins. FEBS (Fed.Eur.Biochem. Soc.) Lett. 38:247-253.

34. van Tornout, P., H. Caster, M.-J.Lievens,M. Rosseneu, andG. Assmann. 1981.Invitro interaction of humanHDLwith human apoproteinA-II.Synthesis ofanapolipoprotein A-Il-rich particle. Biochim.Biophys.Acta.663:630-636.

35. Mendel, C. M., S.T.Kunitake,J.P. Kane, andE.S. Kempner. 1988. Radiation inactivation of binding sites for high density lipoproteins in human fibroblast membranes. J.Biol.Chem. 263:1314-1319.

36.Maciejko,J.J.,D. R.Holmes,B. A.Kottke,A.R.Zinsmeister,D. M.

Dinh, and S.J. T. Mao.1983.Apolipoprotein A-Iasamarkerofangiographically assessedcoronary-arterydisease.N.Engl.J.Med.309:385-9.

37.Gordon, D. J., and B. M. Rifkind. 1989.High-densitylipoprotein-the clinicalimplications ofrecentstudies.N.Engl.J. Med.321:1311-1316.

38.McVicar,J.P., S.T.Kunitake,R. L.Hamilton, and J.P.Kane. 1984. Characteristics of humanlipoproteins isolated byselected-affinity immunosorp-tionofapolipoproteinA-I.Proc.Natl.Acad.Sci. USA.81:1356-1360.

39.Kunitake, S. T., K.J. LaSala, and J. P. Kane. 1985. Apolipoprotein A-I-containing lipoproteins with pre-beta electrophoretic mobility.J.Lipid Res. 26:549-555.

40. Jenne,D.E.,and J.Tschopp. 1989. Molecularstructureandfunctional characterization ofahumancomplementcytolysisinhibitorfound in bloodand

seminal plasma:identitytosulfatedglycoprotein 2, aconstituent ofrattestis fluid.Proc.Natl. Acad. Sci. USA. 86:7123-7127.

41.DeSilva, H.V., J. A.K. Harmony,W. D.Stuart, C. M.Gil,andJ.

Robbins. 1990.ApolipoproteinJ: structure andtissue distribution.Biochemistry 29:5380-5389.

42. deSilva,H.V.,W. D.Stuart, C.R.Duvic,J.R.Wetterau,M. J.Ray,D. B. Ferguson, H. W.Albers,W.R.Smith, andJ. A. K.-Harmony. 1990.A70-kDa

apolipoprotein designatedapoJ isamarkerfor subclasses of humanplasma high

density lipoproteins.J.Biol. Chem. 265:13240-13247.

43.Yui, Y.,T.Aoyama,H.Morishita,M.Takahashi,Y.Takatsu, and C. Kawai. 1988. Serumprostacyclinstabilizing factor is identicaltoapolipoprotein

A-I (apoA-I)Anovelfunction ofapoA-I. J.Clin.Invest.82:803-7.

44.Morishita, H.,Y.Yui,R.Hattori,T.Aoyama, and C. Kawai. 1990.

In-creased hydrolysisof cholesterolesterwithprostacyclinispotentiatedbyhigh density lipoprotein through the prostacyclin stabilization. J. Clin. Invest.

86:1885-1891.

45. Suttorp, N.,W.Seeger, S.Zinsky, and S.Bhakdi. 1987.Complement complex COb-8 inducesPGI2 formation in cultured endothelial cells.Am.J.

Physiol. 253:C13-C21.

46.Petrossian,G.A.,A.B.Nichols,C. C.Marboe,R.Sciacca,E. A.Rose, C.R.Smith,P.J.Cannon, K.Reemtsma, and E. R. Powers. 1989.Relation