J. Lipid Res. 1984 Eisenberg 1017 58

7/28/2019 J. Lipid Res. 1984 Eisenberg 1017 58

1/42


2/42

tion of circulating HDL is about 250-500 mg/dl, equiv-alent to that of LDL. Because more HDL than LDLdistributes to extravascular spaces, the total body contentof HDL possibly exceeds that of LDL. In youngerhumans, when LDL levels are relatively low, HDL isthe major plasma lipoprotein, and HDL is the majorplasma lipoprotein in many animal species. When thenumber of circulating HDL particles is considered, thereis no doubt that HDL is indeed the predominant plasmalipoprotein class: there are 10-20-fold more HDL par-ticles in the human body fluids than the number of allother lipoprotein particles ( 5 ) .

In spite of their large number in plasma, it is difficultto define HDL as a vehicle for lipid transport. Chylo-microns and VLDL are undoubtedly lipoproteins thatcarry triglycerides from sites of absorption and synthesisto sites of storage and utilization. LDL, a degradationproduct of VLDL (6-9), is the major cholesterol-carryinglipoprotein in human plasma. HDL cannot be similarlyclassified and other considerations must apply to thislipoprotein. HDL is the site of plasma cholesterol ester-ification and could be defined as the machinery for thatimportant reaction. However, apparently only a smallfraction of the HDL is sufficient for that purpose, andthere is no indication of absence, or subnormal choles-terol esterification in patients with very low or evencomplete absence of HDL (10, 11). Another potentiallyimportant function of HDL is the so-called reversecholesterol transport, i.e., the transport of cholesterolfrom peripheral tissues to the liver, the main organ ofcholesterol utilization. This function, however, reflectsthe presence of HDL and cannot be used to define thelipoprotein. We prefer to define HDL by its majorpathway of origin. As described later in this review(Section IV), HDL can be viewed as a modified productof redundant surface lipids and proteins generated dur-ing the process of triglyceride transport. Generation ofsuch redundancies is an obligatory event during chylo-micron and VLDL lipolysis. Because surface lipids (freecholesterol and phospholipids) and apolipoproteins arepotent detergents that may, and probably are, harmfulto body cells, their packaging to spherical particles isof biological benefit. Thus, HDL appears to be anessential product of triglyceride transport (as is LDL),and is the major surface-remnant of the same process(12-1 6). This definition of HDL as a product of surface-remnants of lipolyzed triglyceride-rich lipoproteins isno t only convenient but also compatible with metabolicevents discussed in this review.

111. STRUCTURE A N D COMPOSITIONA. Lipids and proteins

Circulating HDL consists of spherical particles thatconform to the general core-lipid structure of lipo-

proteins (17). In normotriglyceridemicsubjects, the ma-jor core-lipid is cholesteryl ester. Variable amounts oftriglycerides are also present, depending on the presenceof lipid transfer activity in plasma and on plasma tri-glyceride levels. The outer shell (surface) of the HDLcontains phospholipids, free cholesterol, and apolipopro-teins. Reported values for the lecithin to sphingomyelinmolar ratio range between 4:l to 8:l (18). In thatrespect, HDL is similar to chylomicrons and VLDL andis different from LDL, a lipoprotein with relatively highcontent of sphingomyelin (about 30% of total phospho-lipid molecules). Assuming that the width of the outershell is 20 A, and the diameter of the HDL core is 40A (HDL3)or 60 A (HDL2), t is calculated that 8 0 4 5 %of the HDL volume is in the surface domain. Hence,even spherical HDL is essentially a surface lipoprotein.

The distribution of molecules between core and sur-face in HDL is not categorical. In analogy to triglyceride-rich lipoproteins (19), few cholesteryl ester and triglyc-eride molecules are apparently present at the surface,while -40% of the free cholesterol molecules are inthe core (20-22). This phenomenon may have importantbiological implications. Cholesteryl esters and triglycer-ides present at the surface may represent the moleculesthat are available for the lipid transfer reactions and thesurface may constitute the site of triglyceride hydrolysisby cell-associated lipases; the presence of free cholesterolin the core increases substantially the ability of HDL toaccept these molecules from other lipoproteins and fromcells.

The nature of protein-lipid associations in HDL hasbeen studied extensively (2-4, 17). In brief, it is believedthat helical segments of apolipoproteins interact withphospholipids and possibly other HDL lipids by hydro-phobic and ionic forces. Angles between phospholipidhead groups, determined by the large radius of curvatureof the lipoprotein, enable interdigitation of proteinhelices between the phospholipid molecules and betterinteraction of the hydrophobic face of the helix withphospholipid acyl chains and free cholesterol (23) oreven with esterified cholesterol molecules (24). It isinteresting to note that different apolipoproteins interactdifferently with the HDL particle. As discussed below(Section V), apoA-I, the major HDL apoprotein, isreadily displaced from HDL by other apolipoproteins,whereas apoC molecules are easily added or excludedfrom the lipoprotein.B. HDL subclasses

The presence of several discrete HDL classes hasbeen confirmed in many studies. In the studies ofGofman et al. (25), HDL was separated into threesubpopulations: HD Ll , HDL2, and HDL3. HDL2 andHDLB are undoubtedly the major HDL populations

1018 Journalof Lipid Research Volume 25 , 1984


3/42

present in the plasma of most, if not all, animal species.Th e main features of the tw o are shown in Fig. 1 andare best illustrated when the contentof lipid and proteinmolecules in each is considered. The core diameter ofHDL2 is about 50% larger than that of HDLS (60 and40 A, espectively), resulting in a core volume which is3.5-fold larger.Hence,HDL2contains 3-4-fold morecholesteryl ester and triglyceride molecules than HDLS.The surface area of HDLp is also larger than that ofHDLS, but the difference is only 2-fold; the differencein proteincontent is only 50%. When the molecularweight of the apolipoproteins is taken into consideration,the difference in protein content can be accounted forby one apoA-I and one apoC molecule. Thus, when thenumber of cholesteryl ester molecules carried in anHDL2 particle is expressed per one apoA (A-I and A-11) molecule, this ratio is twice as large as that in HDLS(1 5). If the capacity to carry lipids by proteins is regardedas the efficiency of a lipoprotein as a vehicle for fattranspor t, then HDL2 is twice as efficient as HDLS,at least in that respect. Micropolydispersity has beendemonstrated in either class of HDL, and both containparticles that vary somewhat in density, size, and lipidcontent (26). Yet, HDL2 and HDLS are almost completelyseparated on density gradients (Le., rate zonal centrifu-gation (26,27)), with adip between the two. Thisphenomenon indicates that the two HDL classes repre-sent thermodynamically favorable structures, with onlya relatively narrow margin of variation within each class.We have suggested that the addition of one molecule of

apoA-I may be responsible fo r hequantum umpbetween the tw o (1 5). Yet, this suggestion is not basedon data , and the question of why HDL2 and HDLS ar ealmost totally separable is unanswered.

Several other discrete HDL populations have beendescribed, and their presence in plasma has been con-firmed by several investigators. The designation HDL,refers to an HDL class tha t consists of particles rich inapoE and which is lighter and larger than HDL2. HDLlhas been identified in human plasma (28-30) and ispresent in relatively large quantit ies in rats,a specieswith very lo w levels of LDL (31-33). HDLI can beseparated from other HDL fractions by affinity chro-matography on heparin-Sepharose columns leaving be-hind apoE-free HDL2 and HDLs (28, 29). In the rat,HDLl spans the density of 1.04-1.09 g/ml, and is richin apoE (50-6096 of otalprotein) but also containsrelatively large amounts of apoC and small amounts ofapoA-I and apoA-IV. The dia mete r of the lipoproteinis 130-140 A, and hecore volume and thereforecholesteryl estercontent) is about 4-fold larger hanthat of HDL2 (33). T he predominant cholesteryl esterfatty acid is arachidonic acid, ndicating LCAT originof these molecules. A ipoprotein that is rich in apoEand separates at similar density intervals is consistentlyfound in the plasma of animals fed diets rich in choles-terol (for a review, see reference 34). This lipoproteinis designated HDL,, and is possibly an analogue of thenormally circulating HDL, .

An HDL population lighter and larger than normal

* b c t a l i p o p r o t e i n e m i aFig. I . Composition of high density lipoprotein subclasses. Data are numb er of prote in and lipid moleculesin average particles.

Eismhprg High density lipoprotein metabolism 1019


4/42

HDL2 has been recently identified in the plasma ofpatients with abetalipoproteinemia (35). Because themain apoproteins of this HDL are apoA-I and apoA-11,we refer to the lipoprotein as ABL-HDL2.Th e size andhydrated density of ABL-HDL2 are similar to those ofHD Ll , and the main difference between the two is theapoprotein profile. In that disease we have also identifiedan HDL population which is denser and smaller thanHDLS, referred to as HDL4 (35).The particles arespherical, contain mainly apoA-I and apoA-11, and rangein diameter between 65 and 70 A. Not dissimilar tinyHDL populations have also been found in other condi-tions, e.g., familial LCAT deficiency (36) nd cerebro-tendinous xanthomatosis (unpublished data). The bio-logical significance of these small-sized HDL populationsis unclear.

A totally different approach to separating HDL sub-populations was recently adopted using specific immuno-adsorption techniques. In two studies, HDL was sepa-rated into populations containing both apoA-I and apoA-11, and populations containing only apoA-I (37, 38).ApoA-I-HDL can be identified in HDL2 and HDLS butseems to prevail in the former. Reported metabolicstudies indicated that the plasma residence time ofapoA-I in either one of the two types of populations isidentical (38).

IV . THE ORIGIN OF HDLThe studies published by Marsh (39,40)nd Hamilton

et al. (41) n the nature of lipoproteins that accumulateduring isolated rat liver perfusion have demonstratedthat HDL is initially formed as a precursor lipoprotein,very different from the circulating particle. The initialparticle is a discoidal lipoprotein composed of phospho-lipids, apolipoproteins, and small amounts of free cho-lesterol. It has been suggested that the discoidal lipopro-teins enter the blood stream and are transformed to themature spherical particles after interaction with LCAT(41).The view that several metabolic steps are involvedwith the formation of mature HDL remained a corner-stone in all subsequent thinking on HDL formation. Yetit became clear that additional metabolic pathways con-tributed to the HDL system.A. Origin of HDL apolipoproteins

Intestine and liver cells are the major sources of Aapoproteins. Although data are mainly available forapoA-I in the rat, the existing information in humansindicates that those data can be extrapolated to the twoproteins (A-I and A-11) and that qualitatively similarpathways operate in the two species.

I . The intestine as a source of HD L and HD L apoproteins.

The mode of synthesis and secretion of HDL and HDLapoproteins by the intestine has been recently reviewed(42, 3). In intestinal lymph from the fasting and non-fasting rat, apoprotein A-I appears in chylomicrons andhigh density lipoprotein fractions (44-49).ApoA-I (andapoA-11) were also identified in chylomicrons isolatedfrom human chylous thoracic fluid (50), he urine ofhumans with chyluria (51), nd human thoracic ductlymph (52-55). n rat mesenteric lymph chylomicrons,apoA-I contributes about 50% of total protein (44, 6)and appears to be an important protein constituent ofnascent chylomicrons. The presence of apoA-I in intes-tinal absorptive cells was demonstrated in rats andhumans by immunofluorescence techniques (44, 56-59), and apoA-I is secreted by the cultured humanintestine (60-62). n studies carried out in humans,transport of apoA-I1 in lymph has been demonstratedas well (50-54).Windmueller and W u (48, 9, 3) aveshown that at least 50% of the apoA-I in rat plasmaoriginates from the intestine.

ApoA-I content in intestinal absorptive cells andapoA-I transport in the lymph increases considerablyduring fat absorption (44, 56-58, 64). Yet the basalrates of intestinal apoA-I synthesis and secretion areapparently independent of biliary lipid and are unaffectedby biliary diversion (63, 65). It is interesting to notethat, in rats with biliary diversion, substantial amountsof newly synthesized apoA-I are secreted via the portalblood circulation rather than with lymph (63).

The amount of apoA-I secreted with chylomicrons(and VLDL) as compared to that with HDL (and otherdense lymph fractions) is controversial. Imaizumi et al.(47)eported that 58-74% of the lymph apoA-I appearsat d


5/42

ers. Interactions of chylomicrons with such cellsor theirsecretory products could result in displacement ofapoA-I from the chylomicrons. Conversely, associationof secreted unbound apoproteins with the lipoproteinparticle may take place. Because of all these reservations,the finding that as much as 80% of lymph apoA-I isfound in fractions denser than chylomicrons should beinterpreted with care.Th e nature of HDL particles secreted by the intestineis unclear. Both discoidal and spherical particles wereidentified in rat mesenteric duct lymph (42-45). Discoidalparticles contribute about 50% of total lymph HDL infasting rats and 35% in fed rats. The relative numberof discoidal particles increases when LCAT inhibitor isadded, and it h a s been suggested that the discoidal HDLrepresents nascent lipoproteins, while the sphericalparticles represent HDL filtered from plasma. In humanthoracic duct lymph and in HDL isolated from the urineof humans with chyluria, however, discoidal lipoproteinswere not identified (51, 54). Human thoracic ductlymph contains a predominance of HDL2 particles (com-pared to the plasma HDL of the same subjects) (54). Aswell, discoidal HDL particles were not yet identified inthe enterocyte. More recently, Forester et al. (67) havesuggested that intestinal cells secrete spherical HDLparticles of small diameter. That suggestion agrees withconclusions derived from observations in patients withabetalipoproteinemia, that the intestine secretes a small-sized and dense HDL population (35). As discussedbelow, the finding of discoidal HDL in lymph cannotbe considered as a definite proof that such particles areindeed primary secretion products of the enterocyte; itis , however, possible that spherical HDL, containingcore-lipids, are secreted by intestinal absorptive cells. Itis noteworthy that intestinal cells contain a relativelyhigh activity of the enzyme system acy1:cholesterol acyltransferase (ACAT) (68, 69), and the intestine is themajor source of plasma cholesteryl esters in patientswith LCAT deficiency (70).

2. The liver as a source of HDL and HDL apoproteins.Biosynthesis and secretion of lipoproteins and apolipo-proteins, including HDL, by the rat liver was extensivelystudied in the late 1950s and during the 1960s (71).Later, individual HDL apoproteins were identified (39,72, 73), and characterized (40, 41, 74). Marsh (40)reported his investigations on the nature of newly syn-thesized HDL apoproteins identified in non-recirculatingrat liver perfusates in 1976. In livers from normal rats,apoE, apoA-I, and apoC comprise more than 95% oftotal HDL proteins. In contrast to serum HDL, perfusateHDL contains more apoE than apoA-I (weight ratio of1.7). To eliminate contamination of HDL with VLDL,the experiment was repeated in livers from rats treatedwith orotic acid. In these perfusates, VLDL release is

15% of normal, and HDL, 40%. The HDL containedan even higher ratio of apoE to apoA-I, 5.0. ApoCcontributed one-third of total HDL proteins with eitherpreparation. Addition of 14Glabeledamino acids revealedthat the major labeled apoprotein was apoE (64% ofradioactivity) followed by apoC (16% of radioactivity).ApoA-I contained only 3% of total radioactivity (specificactivity less than one-tenth that of apoE, and about one-fourth that of apoC). Hamilton et al. (41) also foundthat apoE is the major perfusate HDL protein, especiallywhen the LCAT reaction is inhibited during the perfu-sion period. Rates of apoE secretion increase with timeof perfusion whereas secretion of apoA-I decreases, oreven levels off (74). In HDL (separated as HDLp andHDLs), apoE to apoA-I weight ratios were as high as10.4 (in the presence of DTNB). Minimal amounts ofapoA-I were found in VLDL. Stimulation of apoE andapoA-I secretion was observed in rats with nephrosis(75). In livers from such rats, apoE secretion increasedby a factor of 1.8, and apoA-I by 8.4. The perfusateHDL from nephrotic rats contained more cholesterylesters (9.0% of lipid mass compared to 4.1% in controls)and was markedly enriched with apoA-I, 52.3% of totalprotein (compared to 15.7% in control rats). The incor-poration of labeled amino acids into apoA-I also increasedmarkedly in livers from nephrotic rats, by 441% (com-pared to 38% for apoE).

ApoA-I has been demonstrated in human hepatocytes(59), is secreted by perfused livers of several differentanimals (76, 77), and by rat hepatocytes in culture (78,79). ApoA-I secretion by hepatocytes is unaffected byadding fatty acids to the medium (79), but increases inliver perfusates from fed pigs (76). Cholesterol feedingdoubles apoA-I production in the guinea pig (77), butapoA-I production does not change when rats are madehypothyroid or hypothyroid-hypercholesterolemic78).Synthesis and secretion of apoA-I by other tissues havebeen reported in avians (80, 81). Yet, in mammals,mR N A for apoA-I is found only in the liver andgastrointestinal tract (82).B. Origin of lipids and formation ofHDL recursors

HDL precursors are defined here as apoprotein-phospholipid-free cholesterol complexes that are formedin vivo and can be transformed to spherical particles bythe LCAT reaction. This definition does not specify thenature of apoproteins associated with the HDL precur-sors, although apoA-I and apoA-I1 constitute the majorproteins of the circulating lipoprotein. Three processesmust be considered as sources of HDL precursors: a)direct secretion of discoidal high density structures fromhepatic and intestinal cells (nascent HDL particles);b) lipid and protein constituents released from lipolyzed

Eisenberg High density lipoprotein metabolism 1041


6/42

triglyceride-rich lipoproteins (surface remnants); andc ) phospholipid-apoproteinassociations, not dissimilar tothose formed in vitro.

The discoidal structures identified in liver perfusatesand intestinal lymph, were initially suggested to be themajor, if not the only source of nascent HDL. Thatview, however, was soon challenged by us (12, 14).Several considerations led us to doubt the view that thediscoidal high density particles are true secretory prod-ucts of cells. The first was the inability to identify suchstructures along intracellular secretory pathways (4 ,42, 74), although other lipoproteins are abundant inintracellular organelles and can readily be isolated fromthe Golgi apparatus. Because of their relatively smallsize, it was expected to identify nasceni HDL insidecells (especially in Golgi cisternae or secretory vesicles)even if the total amount of protein found in the lymphor perfusate is not very large. Could such discoidalstructures originate from the surface layer of secretedVLDL and chylomicrons? The liver is a major sourceof a triglyceride lipase (83-85) hat is highly activeunder the conditions of the perfusion procedure, i.e.,using medium devoid of plasma or plasma lipoproteins.[Plasma fractions, and in particular HDL, are knowninhibitors (competitive inhibitors?) of the hepatic lipaseactivity against triglyceride-rich lipoproteins (83, 86).]Indeed, we showed that VLDL triglycerides are readilytaken up by the liver during recirculating perfusion,and that the release of the lipase by heparin results inprompt hydrolysis of the VLDL triglycerides in theperfusate and uptake of the released fatty acids by theliver (86).Addition of plasma or of lipoproteins to theperfusion medium prevented these processes. As to theintestine, we postulated that chylomicrons could interactwith lipases when they traverse the lamina propria. Apotential source is lipoprotein lipase secreted by mac-rophages (87, 8).Comparison of the apoproteins asso-ciated with the hepatic and intestinal discoidal structuressupport our arguments. ApoE (and not apoA-I) is themajor apoprotein found with the hepatic particles,whereas apoA-I is predominant in the intestinal particles.These are the major apoproteins found with the triglyc-eride-rich lipoproteins secreted from the two organs,respectively. Finally, in the studies demonstrating thepresence of discoidal lipoproteins in liver perfusates orintestinal lymph, it was not possible to rule out thepossibility that the cells secrete free apoproteins thatbecome associated with phospholipids at a later stage ofthe secretory process or even after secretion occurred.With the former possibility, hepatic or intestinal cellmembrane would serve as a phospholipid source; in thelatter case, phospholipids would be derived from cellsand secreted triglyceride-rich lipoproteins.

The second potential source of HDL precursors is

the surface coat of lipolyzed triglyceride-rich lipoproteins.Transfer of VLDL and chylomicron constituents toHDL during the course of lipolysis was already demon-strated during the 1950s (89) nd confirmed in manysubsequent studies. However, it was not until the mid70s that the full significance of these observationsbecame apparent. Our earlier studies focused on thecomposition and structure of VLDL and post-lipolysisVLDL (7,90, 1).These studies demonstrated unequiv-ocally that when VLDL is lipolyzed in vitro, as much as80% of the phospholipids and free cholesterol, andpractically all of the C apoproteins, are excluded fromthe lipoprotein. In subsequent studies, we specificallystudied the form by which surface constituents leave theVLDL. The study included in vitro lipolysis experiments(92-96) and an investigation of VLDL-lipolysis in arecirculating isolated rat heart perfusion system (97).With all systems, lipolysis was carried out in mediadevoid of plasma lipoproteins. Albumin, however, hadto be included, and in some instances the albumin usedwas contaminated with bovine apoA-I (98) nd withphospholipids (92).Regardless of the system used, theexperiments showed consistently that phospholipid, freecholesterol, and apoC molecules are displaced from thesurface of lipolyzed VLDL and can be recovered withhigh density fractions. After isolation, the high densitybuffer fraction contained an abundance of discoidalstructures. [More recently, we observed heterogeneityof surface remnants, and some vesicular and sphericalstructures are identifiable as well (unpublished data).]These findings were subsequently confirmed by otherinvestigators (99-1 l), and essentially identical datawere reported for human thoracic duct chylomicrons(1 2).Thus, the data strongly suggest that constituentsoriginating from the surface coat of lipolyzed triglyceride-rich lipoproteins constitute the major, if not the only,source of HDL precursors (12-16).Yet, several differentmechanisms may be involved with formation of HDLprecursors. One mechanism is release, or exclusion offragments of surface redundancies from the shrinkingchylomicron or VLDL particles. Tall and Small (13)suggested that apoA-I is necessary for fragmentation ofthe surface coat of lipolyzed triglyceride-rich ipoproteins.This, however, has not been demonstrated experimen-tally and we have identified discoidal or vesicular highdensity structures even when lipolysis was carried out inthe absence of apoA-I (95).Other mechanisms mayoperate and even predominate. For example, it is possiblethat lipid and apoprotein molecules are excluded indi-vidually from the surface of lipolyzed lipoproteins andthat the discoidal (or spherical) structures are formed inthe water phase.

The third pathway by which HDL precursors can beformed is through apoprotein-phospholipidassociations.

1022 Journal of Lipid Research Volume 25, 1984


7/42

According to this view, shortly after an apolipoproteinbecomes dissociated, it forms a complex with phospho-lipids (mainly phosphatidylcholine) derived from cells orlipoproteins. This pathway is shown schematically inFig. 2. The apoA-I pool represents the amount ofprotein available for HDL formation. This pool is derivedfrom two sources: secretion of free apoproteins by cellsthat synthesize and secrete apolipoproteins, and apopro-teins released from lipolyzed triglyceride-rich lipopro-teins. ApoA-I-phospholipid complexes can form rapidlyif phospholipids are available. The figure suggests thatphospholipids are derived from either cell membranesand intact lipoproteins or the surface coat of lipolyzedtriglyceride-rich lipoproteins. True HDL precursors areformed when free cholesterol molecules are added tothe apoprotein-phospholipid complex. Again, cells, intactlipoproteins, and lipolyzed lipoproteins are the sourcesof free cholesterol.

Several features of the scheme are noteworthy. Firstly,the scheme is compatible with the previous two sourcesof the precursors. Formation of nascent HDL can beviewed as a similar metabolic sequence, occurring insidecells: apoproteins that are not part of VLDL or chylo-microns become associated with phospholipids alongtheir secretory path, and appear in intercellular spaces

FC. releasefrom l ipolyzed

FC ofcel l membranesand l ipoproteins

as protein-phospholipid complexes. Because such com-plexes have not been identified in the cells, the associationmust occur late in the secretory path, possibly evenwhen the proteins cross the plasma membranes. Withregard to the second source of HDL precursors, it hasbeen mentioned above that apoprotein-phospholipid-free cholesterol complexes may leave the surface oflipolyzed lipoproteins as surface remnants. In thesetwo instances, the nascent complexes or the surfaceremnants contribute directly to the HDL precursorpool. Conceptually, the three routes are indistinguishableand represent pathways that accommodate unassociatedapoproteins. Secondly, the scheme predicts that formationof HDL depends on two processes: secretion of freeapoproteins and lipolysis of triglyceride-rich lipoproteins.That seems necessary in view of the very strong rela-tionship that exists between lipoprotein lipase activitiesand HDL levels (103-106) and the presence of HDL inplasma even when lipolysis is totally absent (e.g., abetali-poproteinemia) or is extremely limited (e.g., lipoproteinlipase or apoC-I1 deficiencies).Thirdly, although apoA-Iis the only protein shown in the scheme, other apopro-teins can substitute for apoA-I, at least in part. Thus,apoC, apoE, or apoA-I1 can initially be part of theapoprotein-phospholipid complex. Finally, the scheme

OR IG IN OF HIGH DENSITY LIPOPROTEINI Lpo A -I release fromTG- rich 1 ipoproleinsApo A -I synthesisI and secret ionL I II J IApo A - I pool- +PL of PL rel eas ecel l membranes fr om l ipolyzedand l ipoproteins TG - rich lipoproteinsILPL

IpO A-I-PhospholipidI complexes

(d isco ida l LP )

Spherical HDLFig. 2. Schematic representation o f H D L formation. I t is suggested that apopro teins, phosph olipid, and freecholesterol become associated in body fluids and form H D L precursors (discoidal lipoproteins). The sourcesof apoproteins are synthetic sites (predominan tly liver and intestine) and the lipolysis process. Phosph olipidsand free cholesterol molecules are derived from the surface coats as a result of lipolysis of triglyceride-richlipoproteins and from membrane and lipoprotein lipids. Nascent H D L and surface remnants contributedirectly to the H D L precursor pool (see text). Spherical H D L is formed when sufficient cholesteryl esters aregenerated via L C A T .

Eisenberg High density lipoprotein metabolism 1043


8/42

predicts that formation of HDL precursors necessitatessupply o both apoproteins and phospholipids. Under physi-ological conditions, phospholipid supply apparently isnot limited. In many experimental models (e.g., lipolysis)however, hydrolysis of phospholipids (one action oflipoprotein and hepatic lipases, refs. 86, 93, 95, 97,107- 10) may obscure processes that during normalphysiology are followed by formation of HDL precursors.We have shown that, in the presence of excess lipase, asmuch as 70% of the VLDL phosphatidylcholine can behydrolyzed in vitro (95).Of interest, when VLDL tri-glycerides are hydrolyzed in vitro by a bacterial lipasedevoid of phospholipid hydrolase activity, the amountof surface remnants recovered in the test tube is 2-to 4-fold higher as compared to systems with the samedegree of triglyceride hydrolysis achieved by excessiveamounts of lipoprotein lipase (1 1).C. Formationof sphericalHDL

Transformation of HDL precursors to spherical HDLis dependent on LCAT activity. Hamilton et al. (41)have demonstrated that complexes isolated from ratliver perfusates are indeed substrates for LCAT, andsuggested that the cholesteryl esters formed at thesurface of the complexes are displaced to the hydropho-bic hydrocarbon domain of the phospholipid bilayer.Yet rearrangement of the complex seems mandatory.Hamilton et al. (41) ointed out that their complexescontain surplus phospholipids, some of which must beremoved during the disc - phere transformation. Aswell, apoproteins other than A-I that are present in thecomplexes must be replaced by apoA-I. This seemsobligatory inasmuch as in diseases where apoA-I isabsent (10, l),HDL is not found in plasma. In theseconditions, LCAT is active, and cholesteryl esters arepresent in both VLDL and LDL, although in reducedamounts. Tangier disease provides a unique opportunityto test the consequences of limited supply of apoA-I onHDL formation (1 12-1 15). In this disease, surfaceremnants are readily demonstrable after a fatty meal,especially in the patient who underwent splenectomy(1 14, 115).HDL levels, however, are exceedingly low,and disc -+ sphere transformations are severely limitedin spite of apparently normal synthesis and secretion ofall other apoproteins. Small amounts of apoA-I1 contain-ing spherical HDL are found in the patients plasma(1 14), but even apoA-I1 cannot effectively replaceapoA-I. The surface remnants and HDL precursorsare rapidly cleared from the plasma, predominantly bycells in lymphatic tissues (1 16).

The classic example of the consequences of the absenceof a disc- phere transformation is familial LCATdeficiency (70).The patients plasma contains abundantamounts of discoidal and vesicular structures (1 17- 19)

and, as in Tangier disease, surface remnants accu-mulate after a fatty meal (120).Both apoA-I and a p A -I1 have been identified in several HDL populations,while apoE was identified in others (36, 0, 121).Thepatients HDL contains about 10-15% of the normalamounts of apoA-I and apoA-I1 but almost twice asmuch apoE (36, 22, 123).In whole plasma, the amountof apoA-I is about 40% of normal, but apoA-I1 levelsare low, 10-30% (36, 119, 122, 123).After centrifu-gation, 70430% of the apoA-I is found in the plasmafraction of d > 1.25 g/ml, compared with only 3-5%in normal subjects (120).More apoA-I is found withHDL when the patients plasma is incubated with LCAT,while apoE moves to VLDL (120). n another study,disc - phere transformations were demonstrated forthe apoA-I-containing structure, but not for the apoEparticles (1 24).While it is extremely difficult to interpretall of the data in LCAT deficiency, several conclusionscan be drawn. Unlike Tangier disease, LCATdeficientplasma contains an abundance of HDL precursors, andtransformation to spheres is readily demonstrable whenLCAT is added (120, 124).The differences betweenthe two conditions are unclear, but possibly reflect thepresence of apoA-I in LCAT deficiency, and its nearlytotal absence in Tangier disease. Another difference isthe presence of multiple HDL precursors in LCATdeficiency and their absence in Tangier disease. Is itpossible that the presence of LCAT but absence ofnormal apoA-I causes accelerated catabolism of HDLprecursors or of spherical HDL containing abnormalapoprotein composition? These and other questions in-dicate that disc - phere transformation may reflect aprocess more complicated than cholesterol esterificationand displacement of cholesteryl ester molecules to thecenter of the phospholipid bilayer.D. Genetic control of HDL apoprotein synthesis

The mode of synthesis and secretion of HDL apopro-teins A-I and A-I1 and the genetic control of this processhave been clarified over the last 2 years (1 5).

ApoA-I can be separated by two-dimensional electro-phoresis techniques into several (six) isoproteins withthe same apparent molecular weight but different iso-electric points (62). he main circulating apoA-I formsare A-I4(79 k 7% of total apoA-I) and A-I5 and A-16(19 t 6%); A-I2 and A-Is constitute less than 2% (126).Examination of apoA-I secreted by either hepatic orintestinal tissues in culture (62, 27-133)or by human-derived hepatoma cells (1 4) revealed that the newlyformed protein is different from the circulating protein.ApoA-12 and A-Is are the major secreted apoproteins,78 f 5 and 21 -t 5% of otal protein, respectively, whileisoforms A-I4 and A-I5 constitute less than 1% (126).Hence, it appears that apoA-I is secreted as a proprotein

1024 Journal of Lipid Research Volume 2 5 , 1984


9/42


10/42

HQ h Density L ipopro te in : PHOSPHATIDVLCHOLINE

P h o s p h o l i p o l y s i sFig. 3. Intravascular dynam ics of H D L phosphatidylcholine (PC). Lipolysis appears to be the major sourceo f HDL-PC (see text). P C also enters H D L with "nascent" particles and is derived from the surface coat ofnascent chylomicrons. HDL-PC exchanges with PC in other lipoproteins and cell m embran es, a reaction thatis facilitated by phospho lipid exchan ge proteins (PLEP). PC is hydrolyzed by lipases and is consumed by theL C A T reaction. Whether there is net movement of PC molecules between H D L and cell membranes is notclear. Th e lifetime of PC molecules in H D L is less than 1 day.

latterhasbeen shown to begreatly acilitated by aplasma phospholipid exchange protein (154-1 56). Inour study (155), biosynthetically labeled rat plasmaVLDL served as the dono r of labeled phospholipids. Inthe presence of plasma, an almost complete exchangeof phospholipids between VLDL and HDL equal specificactivity of PC) was achieved afte r 60 min of incubation.The reaction, however, did not result in net movementof molecules in either direction. In analogy with otherPC exchangereactions (161, 162), it is expected hatsuchproteins may also facilitate exchange with cellmembranes. Under normalconditions, hisexchangealso doesnotcausenetmovementof PC moleculesbetween the l ipoproteins and the cells. Yet, when phos-pholipids are consumed at a rapid rate by cells (e.g.,rapidly growing tissues) or lipoproteins (e.g., LCATactivity), PC molecules may move in either direction.

HDL phospholipidsparticipate in metaboliceventsand are continuously consumed. Most impor tant is theLCAT reaction which forms cholesteryl esters from PCand free cholesterol molecules. About 5-10 mmol ofcholesteryl esters are formed in a day in adult humansby this eaction (70). Hence, he reaction onsumesequimolar amounts of PC, i.e., 5-10 mmol or 4-8 gper day. It shouldbe ealized that this amount isequivalent to total circulating HDL-PC mass, assumingthat 1 liter of plasma contains about 0.5-1.0 g of HDL-PC and that about 60% of the HDL is present in theplasma compartment. HDL-PC, moreover, is availabletoother enzymes with phospholipase activity, mainly

1026 Journal of Lipid Research V ol um e 25 , 1984

the lipoprotein lipase (108, 109) and the hepat ic lipase(163-166) systems. Th e exact contribution of phospho-lipid lipolysis to total consumption of HDL-PC is notknown.

The considerations presented above demonstrate thedynamic nature of he HDL-phospholipid system. PCmolecules continuously enterHDL particles andarerapidly exchanged or consumed. At equilibrium, whenthe rates of consumption of HDL-PC are considerablyshorter than a day, these molecules turn over in HDLat least five times faster than apoproteins.B. HDL free cholesterol

The dynamics of free cholesterol (FC) in H DL ar esimilar to those of PC (Fig. 4). As for PC, the sourcesof free cholesterol are nascent particles, the surface coatof lipolyzed triglyceride-rich lipoproteins and cell mem-branes. FC molecules exchange rapidly between lipopro-teins (167). At 37"C, more than 90% of 14C-labeled FCin HDLs is transferred to LDL in a first order processand tlI2 of 2.9 min. Forcomparison, he t 1 / 2 of PCexchange is 5.1 k 1 hr. FC exchange occurs by diffusionthrough the aqueous phase (167). Experiments carriedout with liposomes have documented rapid distributionof free cholesterol from structures with a high FC/PCmolar ratio, to structures with low ratio (1 68- 170). Thismovement of FC between particles also occurs throughthe aqueous phase (168, 170). Once in HDL, FC mole-cules exchange with others in other ipoproteinsandcell membranes (171). Under suitable conditions, free


11/42

nascc

in other lipoproteinsand ce l l membranes

[Liver)timulated b y )IAdrenal I [lCETIGonadcs)I otha IFig. 4. Intravascular dynamics of H D L freecholesterol (FC). T he main sources of FC moleculesare hesurface coat of intact lipoprote ins and lipolyzed triglyceride-rich lipo protein s, cell mem bran es, and nascent particles. With lipolysis, some free cholest erol molecu les becom e first associated with LDL but therea fter areutilized in HDL. HDL-FC exchan ges with FC n other ipoproteins and cell membranes, and is utilized byth e L C A T reaction (esterification stimulated by cholesteryl e ster transfer (C ET) ). Finally, FC molecules alsomove from H D L to cells. The lifetime of FC molecules in H D L is a few hours.

cholesterol may move out of HDL toFC-poor structures,but in biological systems thisprobablyoccurs nfre-quently.HDL-FC, like PC, is utilized in theLCATreaction.

Recently, w e have attempted to quantitate the amountsof FC transferred from lipolyzed triglyceride-rich l i p o -proteins to HD L (1 72). Specifically, w e wished to deter-mine amounts of free cholesterol that are transferredduring lipolysis to cell membranes, as compared to theamount ransferred oHDLandother lipoproteins.The experiments were conducted in vitro and in theisolated perfused rat heart. Rat plasma VLDL containingbiosynthetically labeled [3H]cholesterol was used as thetriglyceride-richipoprotein, and uman blood cells(RBCs, platelets, and WBCs) or the atheart werepotentialacceptors for cholesterol.Exchange of reecholesterol between the VLDL and thecells was relativelyslow, about 10 hr. With nductionof lipolysis, it wasfound hatnone of the cholesterol generated duringlipolysis was transferred to any of the cells tested. HDL,in contrast, ervedas an excellentacceptor for thischolesterol, eithe r in the presence or absence of cells.Thus, i t seems that all of the fre e cholesterol gener ated durin glipolysis remains in the lipoprotein system, and none is takenup by cells. Of nterest was the observation thatLDLalso serves as acceptor for lipolysis-generated FC. LDLin this cycle, however, seems to play a transient role asw e were able to show that FC in both VLDL and LDL

is consumed by HDL when theLCAT reaction isallowed to progress unpublished results). A imilarobservation was previously reported by Fielding andFielding (1 73). We therefore concluded that VLDL andLDL are important sources of FC fo r HDL, and thatthe process is accelerated by lipolysis. Whether phospho-lipids behave similarly has not been determined.

Another quantitative difference between PC and FCis the ease with which FC in cell membranes is trans-ported to HDL. Apparently, when the LCAT reactionis active and HDL-FC becomes esterified, FC moleculesmove from the membranes to the HDL. This phenom-enon is an essential part of thereverse cholesteroltransport process and is described in detail in SectionVIII.

HDL-FC is consumed by theLCAT eaction,andmay be used by cells. Ithasalreadybeenmentionedthat 5-10 mmol ofCE areformed in aday by theLCAT eaction. T h e whole HD L system (intra-andextravascular) contains at most 1 mmol of FC. Hence, ina day, the LCA T reaction uses$ve to ten times more FC tha nis actually present n HD L. In other words, all of theHDL-FC is replaced by new molecules within 2-5 hr .In the circulation, FC molecules consumed by the LCATreaction ar e replaced by molecules coming from othersources. Suchsources areother lipoproteinsand cellmembranes. Indeed, when plasma is incubated in vitroand heLCAT reaction is allowed to progress,both

Eis rnbrrg High density lipoproteinmetabolism 1027


12/42

V L D L and LDL become depleted of FC molecules(173). We have studied the kinetics of this process byusing [ 'C]cholesterol-labeled rat plasma supplementedwith human VLDL (unpublished results). During thefirst 6-12 hr of the incubation, utilization of FC wasevident only in VLDL and LDL, while the amount ofFC in HDL was not changed. Only when 30-50% ofthe VLDL-FC and LDL-FC was esterified, did theamount of HDL-FC begin to decline. Because the rat isan animal devoid of the core-lipid transfer reaction, wecould also show that more than 90% of the newlyformed I4C-labeled CE molecules remained in HDL.Hence, the CE molecules were indeed formed in HDL,and the source of FC was transfer of molecules fromVLDL and LDL to HDL.

Movement of FC molecules from HDL to cells occursunder special conditions. For example, in LCAT defi-ciency cholesteryl esters are not formed, and cell mem-branes (e.g., erythrocytes) are enriched with FC presum-ably originating from the abnormal lipoproteins presentin the plasma (174). Of interest, normal HDL enrichedwith FC also donates cholesterol to cells in culture (175).Other examples are tissues that utilize large amounts ofcholesterol for metabolic activities, i.e., liver, adrenal,and gonads. In the liver, HDL-FC may actually be amajor precursor for bile acids (176).

In summary, like PC, the free cholesterol moiety ofHDL must turn over at a very fast rate. This rate isprobably several-fold faster than that of PC, and 20- to30-fold faster than apoproteins. HDL, therefore, seemsto play a central role in plasma FC metabolism, assuggested by Schwartz et al. (177, 178).C. HDL cholesteryl esters and triglycerides

The metabolic behavior of the two core lipid moietiesof HDL is dependent on three processes: cholesterolesterification via the LCAT reaction, the plasma lipidtransfer reaction, and the activity of lipases. The mech-anism of the LCAT reaction has been summarizedrecently. It is appropriate, however, to describe brieflythe second reaction, i.e., the lipid transfer reaction.Nichols and Smith (179) were the first to show that,during in vitro incubation of human plasma, triglyceridesare transferred from VLDL to LDL and HDL, andcholesteryl ester molecules are transferred in the oppositedirection. Several years later, the reaction was demon-strated with isolated lipoproteins (180). In 1975, proteinsthat catalyze the reaction were found in rabbit lipopro-tein-free plasma (d > 1.2 1 g/ml) (18 1). Purification ofcore-lipid transfer proteins was reported by severalinvestigators (1 56, 182-1 88). Fielding and his associates(182, 189) suggested an identity between the lipidtransfer protein and apoD. This suggestion, however,was not corroborated in other studies, and it is currently

believed that the lipid transfer protein is an acidicglycoprotein with a molecular weight of about 6 1,000(185-1 87). Whenever studied, cholesteryl ester, triglyc-eride, and phospholipid transfer activities are co-purified(156, 184, 186).

In plasma, the transfer protein is associated withHDL, mainly HDLS, and almost no activity is foundwith lower density lipoproteins (190, 191). A plasmainhibitor of the transfer activity has been described(192). Cholesteryl ester and triglyceride transfer reactionswere demonstrated in many in vitro incubation systems:whole plasma, plasma fractions, and isolated lipoprotein(33, 156, 173, 179, 180, 193-205). While the mecha-nisms of the reaction are poorly understood, severalfeatures of the transfer process have been established.Transfer activity is present in some animal species,including humans and rabbits, but is absent in otherssuch as rats, pigs, and dogs. The bidirectional transferof triglycerides and cholesteryl esters is always relatedto the amount of transfer protein, the time of exposurebetween donor and acceptor particles, and the relativemass of the two. A t least two distinct processes can bedifferentiated: exchange (replacement of a molecule inone lipoprotein by the same molecule present in anotherlipoprotein) and unidirectional transfer of moleculesfrom one lipoprotein to another. This second reactionis responsible for the changed core-lipid compositionthat occurs during the exposure of two lipoproteins ofdifferent core composition to transfer activity. It hasbeen suggested that the mechanism of both reactions(exchange and unidirectional transfer) depends on in-teraction of the protein with molecules present at thesurface coat of the lipoproteins and is proportional totheir surface area (206). Th e magnitude of exchangeand transfer, however, is determined by the relativeproportion of cholesteryl ester and triglyceride moleculesin different lipoproteins (206). For example, incubationof LDL and HDL, two cholesteryl ester-rich lipoproteins,results mainly in exchange; incubation of LDL or HDLwith triglyceride-rich lipoproteins, in contrast, results inaltered core-lipid composition of the two lipoproteins(204, 206, 207).

Of particular relevance for understanding HDL core-lipid physiology is the interaction between the LCATand the core-lipid transfer systems. Removal of choles-teryl esters from HDL has been reported to acceleratethe rates of the LCAT reaction, possibly by relievingthe HDL of product inhibition (173, 208). I t has evenbeen suggested that both the LCAT protein and thecore-lipid transfer protein are present together on aspecialized HDL subpopulation fraction (189). Newlyformed (LCAT-derived) cholesteryl ester molecules canapparently be distributed to lipoproteins by two path-ways: a direct pathway, by which the molecules are

1028 Journalof Lipid Research Volume 25 , 1984


13/42


14/42

as well. Yet, in another study, it has been concludedthat significant loss of CE molecules from HDL withoutreplacement by other core molecules does not occur (213).

The lifetime of cholesteryl ester molecules in HDL isnot clear. HDL contains between 1 and 1.5 mmol ofcholesteryl ester in 1 liter of plasma. Th e total bodyHDL-CE pool then is about 4-8 mmol, an amountsimilar to that produced in 1 day by the LCAT reaction(70). In humans, most of the LCAT-derived cholesterylesters are transferred to other lipoproteins. As mentionedabove, some of these esters are possibly delivered directlyto lower density lipoproteins, while other molecules arefirst assimilated in the HDL system and are transferredlater. The contribution of each pathway to total CEtransport is not clear. When free cholesterol is allowedto become esterified in rat plasma, at least 90% of thenewly formed esters are recovered in HDL and lessthan 10% in othe r lipoproteins (33, 198). If this obser-vation can be extrapolated to the human, then all ofthe LCATderived CE molecules should turn overthrough the HDL system in a day. The residencetime of a cholesteryl ester molecule in HDL, therefore,is also 1 day or less.

Th e amount of triglyceride molecules in HDL isdependent on the presence and the rate of the lipidtransfer reaction. Small amounts of triglycerides, how-ever, may be present in nascent spherical HDL ofintestinal origin (67), but not in discoidal particles.Intestinal origin probably accounts for the presence oftriglycerides in the plasma of patients with abetalipopro-teinemia (35). With that exception, most or even allHDL-triglycerides must be derived from triglyceride-rich lipoproteins by the lipid transfer reaction. Variationof the content and activity of lipid transfer proteinscould be of importance in determining the number oftriglyceride molecules in HDL. That probably does notoccur. We have recently measured the activity of thecore-lipid transfer reaction in normal human subjectsand in patients with type IIA, IIB, and IV hyperlipopro-teinemia (S. Eisenberg, unpublished results). Cholesterylester exchange between HDL and LDL was used toestimate total plasma activity. Th e rates of CE exchangeranged between 50 and 150 Pmol/l/hr and were similarin all groups of subjects. Transfer activities were alsorecorded in one female subject with familial hyperal-phalipoproteinemia, in three patients with abetalipopro-teinemia, and in several samples of cord blood, althoughin the latter samples the activity was the lowest. It istherefore concluded that the most important singlefactor that determines the amount of triglycerides inHDL is the ratio between the mass of triglyceride-richlipoproteins (chylomicrons and VLDL) and of HDL inthe plasma. This conclusion is supported by the findings

that TG to CE ratios in HDL are highly correlated withplasma triglyceride levels (2 14-2 16).

The fate of HDL-triglycerides is not clear. In vitrostudies have demonstrated that lipoprotein lipases andhepatic lipases can hydrolyze HDL-triglycerides (203).T o what extent HDL-triglycerides are susceptible to theactivity of these enzymes in vivo is not known.D. HDL apoproteins

HDL contains a host of apolipoproteins (A-I, A-11,A-IV, C-I, C-11, C-111, D, E, and F) and carries severalother proteins, including the LCAT protein, the lipidtransfer protein, the serum AA amyloid protein (217),and the beta-glycoprotein-I (2 18). Th e biological behav-ior of apoA-I and apoC has been studied extensively. Inthe following section the behavior of apoA-I, the majorHDL apoprotein, will be discussed in detail, followedby a brief description of the behavior of other apopro-teins.

1. Apoprotein A-I. In normal fasting human plasma,most of the apoA-I is present in HDL. For example,when 251-labeledapoA-I is added to human plasma orto a mixture of lipoproteins, about 95% of the radioactiveprotein is re-isolated with HDL (2 19). Radioimmunoas-says of apolipoprotein A-I generally report that 80-90%of the immunoassayable apoA-I is present in HDL andthe rest is found mostly with the plasma protein fractionof d > 1.21 g/ml (220, 221). Very small amounts ofapoA-I are found with either LDL or VLDL. It hasbeen suggested that apoA-I has a preferential aviditytowards small particles with large radii of curvature(222-224). Yet, apoA-1 is a major apoprotein of nascentchylomicrons (42-47). These observations are contra-dictory, unless chylomicrons possess specific binding sitesfor the protein. Even then, it is difficult to understandwhy chylomicrons release most of their apoA-I oncethey enter the circulation. One explanation for thesecontradictory observations is that the apoA-I present inchylomicrons is different from that present in plasmaHDL. Thus, it is possible that nascent chylomicronscontain, in fact, the proprotein, and that conversion ofpro-apoA-I to apoA-I occurs after the protein is releasedfrom the chylomicrons.

The intravascular dynamics of apoA-I (and otherHDL apoproteins) are shown schematically in Fig. 6.Three potential sources for apoA-I are considered:nascent HDL particles, free apoA-I plasma pool, andapoA-I released from the surface coat of triglyceride-rich lipoproteins of intestinal origin. The first source isobvious and reflects the amounts of apoA-1 that areassociated with secreted spherical or discoidal HDL.The second is more complex. Small amounts of apoA-I(and other apoproteins) are present apparently as free

1030 Journal of Lipid Research Volume 25 , 1984


15/42

Exchange withother lipoproteins( C ' s . E )and among HDL( A - I . A - I t )

" C a t a b o l i c "i tes Disp lacern-ent b yapopro te lns I C ' s . A - I I )Fig. 6. Intravascular dynamics of HD L apoproteins. Apoproteins are part o f "nascent" complex es and aregenerated during the lipolysis process. A small p o o l of rapidly turning over free apoproteins probably alsocontributes see ext).Once in an HD L particle,apoproteins rapidly exchan ge with other ipoproteins,including HDL. Some apoproteins may become displaced from HD L by other proteins (e.g.,displacement ofapoA-l by apoC or apoA-11). Th e apoproteins in H DL are responsible for the binding of HD L to specificcells and som e of these interactions result in catabolic events (see Se ction VIII).

proteins (or are associated with very small amounts ofphospholipids) in the plasma. Th e main evidence forthe existence of a free apoprotein pool is the reportedsecretion of apoproteins in urine (225)and their presencein body fluids such as the cerebrospinal f luid (226). Thisfree apoprotein pool conceivably represents apoproteinsdissociated from lipoproteins and apoproteins secretedfrom cells that escape immediate association with lipo-proteins. The th ird source is the surface coat of triglyc-eride-rich lipoproteins of intestinal origin. T he mecha-nism(s) responsible for the release of apoA-I from chy-lomicrons (small or large) have not been completelyclarified. Incubation of chylomicrons in plasma appar-entlydoesnot esult in significant releaseof apoA-IalthoughbothapoCandapoE are transferred o hechylomicrons (46, 227). In a preliminary study, w e havenot oundappreciableexchange of apoA-Ibetweensmall chylomicrons (biosynthetically labeled with [3H]leu-cine) and rat HDL (T. Chajek-Shaul and S. Eisenberg,unpublished results). ApoA-I, however, is rapidly freedfrom the chylomicrons when the lipoprotein nteractswith lipoprotein lipases (52, 151) , and apoA-I contentin HDL increases dur ing clearance of alimentary lipemia(227, 228). The physicochemical basis for this processis notclear. Fo r example, it is not known whethertriglyceride hydrolysis is sufficient or whether omephospholipid hydrolysis is involved with this process.

Similar to other constituents, once apoA-I is presentin HDL, the molecule does not remain associated withindividual particles throughout their circulating lifetime.

Rather, apoA-I molecules seem to beeleased andtransferred between HDLparticles. This again is accom-plished by several different mechanisms. One mechanismis pure exchange process, i.e., replacement of an apoA-I molecule in anHDL particle by anothe r moleculefrom another particle. This path has been studied mainlywith HDLP and HDL3 using ultracentrifugation (229),and gradient gel electrophoresis (230) to separate HDLspecies. Both apoA-I and apoA-I1 do exchange betweenHDL2andHDLs (229). Initial rates of transfer ofradioactive proteins as high as 70%/hr have been re-ported. However, when the amounts of protein in t w oHDL fractions ar e similar, only 7-14% of the radioac-tivity is exchanged dur ing a 5-hr ncubation (230). Innone of these experiments has a full equilibrium (equalspecific activities of apoA-I in donor and acceptor par -ticles) been achieved. Such equilibrium has been dem-onstratedforapoC(155). We have used anotherap-proach o study apoA-I and apoA-I1 transfer betweenHDL particles. In the study we used the lack of antigeniccross-reactivity between human a nd rat HDL as a toolto determine bidirectional transfer of human apoproteinsto rat HDL and vice versa (231). Preferential transferof human apoA-I to rat HDL was found, although sometransfer of apoA-I1 was also detected. Of interest, onlyone-fourth to one-half of the human apoA-I was trans-ferable under the conditions of the experiment (incu-bations of human and rat HDL's for 30 min followedby the addition of antiserum).

A second phenomenon is the exchange of free apoA-

E:j.wrhprg High ensity lipoprotein metabolism 1031


16/42

I in solution with the same molecule in HDL particles.These investigations were car ried out to achieve labelingof the HDL with radio-iodinated apoA-I or apoA-I1(232, 33). For both proteins, exchange between freemolecules in solution and those associated with lipopro-tein particles has been documented. With apoA-I, how-ever, it was impossible to exchange all of the moleculespresent in HDL, and it has been concluded that part ofthe apoA-I in HDL is present as a discernible, unex-changeable pool. That conclusion agrees with our earlierobservations (23 ).

The third mechanism of exchange is replacement ofapoA-I molecules in HDL by other apoproteins. Thisphenomenon has been shown for apoC and apoA-11. Incanine (234, 35) and human (236)HDL, addition offree apoA-I1 causes considerable or even complete re-placement of apoA-I by A-11. The displaced apoA-Iappeared as lipid-free protein or, in the presence ofexcess of apoA-11, as an apoA-1:apoA-I1 complex (235).Similarly, addition of apoC to human HDL results inaccelerated displacement of apoA-I from the lipopro-tein (237).

The physicochemical basis of apoA-I exchange hasnot been clarified. Grow and Fried (229) uggestedformation of collision complexes. Pownall et al. (237)however, showed concentration-dependent dissociationof apoA-I from HDL when the lipoprotein is diluted in0.15 M NaCl solution. This observation suggests spon-taneous dissociation of apoA-I from HDL and agreeswell with the concepts presented by Osborne and Brewer(238). Such a dissociation process accounts for thepresence in aqueous solutions of a small pool of rapidlyturning-over unassociated apoA-I. Such an apoA-I poolmay adequately explain the different exchange phenom-ena described above, including displacement of apoA-Iby other proteins. In this latter instance, the differentaffinities of apoproteins towards the lipoprotein surfacewould be critical in determining the final apoproteinprofile of the lipoprotein. Finally, if such a pool existsin vivo, it may play an exceedingly important role inmetabolic events. HDL conversions, discussed in SectionV I , are examples where a changing apoprotein profileof HDL may occur.

2. Apoprotein C. Exchange and transfer of apoC amonglipoproteins has been extensively reviewed in the past(9, 39).The proteins exchange rapidly between VLDL(and chylomicrons) and HDL (6, 7,92, 95, 155, 219).This process is temperature-dependent (155), propor-tional to the mass ratio of the t w o lipoproteins (2 19),and is independent of phospholipid exchange (1 5).ApoC-I1 and apoC-111-1 behave similarly under a varietyof experimental conditions (94).Whether the apoproteinsexchange through a dissociation-association mechanismor lipoprotein collision has not been ful ly clarified.

Although about 50% of plasma apoC mass is present inHDL (in normolipemic humans), the proteins apparentlyprefer the triglyceride-rich lipoproteins, as the availablesurface of VLDL and chylomicrons is only a smallfraction of that of HDL. Y et C apoproteins displaceapoA-I from HDL. A changing mass ratio betweentriglyceride-rich lipoprotein and HDL apparently ex-plains the redistribution of C apoproteins during induc-tion of alimentary lipemia (150) and after initiation oftriglyceride hydrolysis by the injection of heparin tohumans (7) r rats (90). t is significant to note that Capoproteins return to VLDL when newly secreted par-ticles enter the plasma compartment (7). In that andother respects, therefore, the dynamics of apoC are verydifferent from those of apoA-I.

3. Other apoproteins. The behavior of apoA-I1 is ap-parently very similar to that of apoA-I. Free apoA-I1reassociates predominantly with HDL (2 9), can beincorporated into HDL (233-236), nd exchanges be-tween HDL2 and HDLs (229,230). adioimmunoassaysdemonstrated apoA-I1 predominantly in HDL (240).Unlike apoA-I, however, apoA-I1 reassociates with bothsmall and large lipid complexes (224) nd has a higheraffinity, to lipoprotein surfaces (including HDL) thanapoA-I. ApoA-IV is a prominent protein constituent ofrat plasma HDL (241). n humans, the protein entersthe plasma with nascent chylomicrons and is rapidlydisplaced to the lipoprotein-poor plasma fraction of d>1.2 g/ml (242-244).Other apoproteins are probablyassociated with specialized HDL populations and do notreflect the behavior of the HDL system as a whole.E. Dynamics of the HDL system

The data and considerations discussed in this sectiondemonstrate the dynamic nature of the HDL system. Inhumans, none of the HDL constituents is associatedwith the particle for a long period of time. Each con-stituent, moreover, turns over in HDL with a differentlifetime, ranging between a few hours to 1 day. In thatrespect, the HDL particle seems to play the role of atemporary station for lipid and proteins. Lipid con-stituents are transferred and exchanged into the particle,are used and modified by enzymes and transfer proteinsand their metabolic products, or the remaining intactmolecules are transported out of the HDL. The proteinsbehave similarly but remain unmodified. The role ofprotein movements is apparently to provide the HDLsystem with the necessary flexibility to accommodatedifferent amounts of lipids and /or to supply proteinsfor intravascular synthesis of new HDL. The integrityof the HDL system appears to depend on and reflectthe sum of all these movements and modifications of itsconstituents. In that sense, the HDL exists as a particlethat provides a center of intense biological activity. Th e



17/42

impact of this dynamic situation and the consequencesof over- or under-activity of one or more of the biologicalreactions discussed above on HDL formation have beendescribed in the previous section. The next sectionpresents the effects of these processes on the levels,nature, and distribution of HDL subpopulations.

VI. REGULATION OF HDLSUBPOPULATION DISTRIBUTION

The data presented in Fig. 1 (composition of HDLpopulations) suggest that HDL particles can becomelarger or smaller when, and if , lipid and protein mole-cules are added or deleted from different particles. Inthe present section, the evidence that such conversionstake place in physiologic situations, and their mechanisms,are discussed.A. Conversion of HDLS to HDLs

For this conversion to take place, it is necessary toincrease the number of CE molecules in HDLS by 2- to3-fold (from 40-50 to 100-120 molecules) and to pro-vide HDLS with one molecule of apoA-I and sufficientamounts of surface lipids (phospholipids and free cho-lesterol). Available evidence suggests that the activity ofthe extrahepatic lipoprotein lipase system is strongly

related to the levels of plasma HDL2 in humans. HDLconcentrations in plasma (as estimated by HDL-choles-terol) are in general proportional to the activity oflipoprotein lipase (103-106) and hence to rates oftriglyceride transport (245). Because most or even nearlyall the variation of plasma HDL concentration is ex-plained by the variation of HDLP (246, 247), the rela-tionship between HDL concentration 'and lipoproteinlipase activity reflects that of the HDL2. A mechanismthat explains these observations was proposed by us in1978 and is graphically presented in Fig. 7. In a seriesof experiments, we have investigated the effects ofVLDL lipolysis in vitro on the structure and compositionof HDLs (248). Th e study has demonstrated that HDLsreadily accepts phospholipids, free cholesterol, and apo-proteins (predominantly apoC) released from the surfacecoat of the lipolyzed VLDL. By this process, the contentof phospholipids and free cholesterol in HDLS increasesby nearly loo%, and that of apoproteins by about 50%.The capacity of HDLs to accept phospholipids is satu-rated at that level and, when more phospholipids arereleased from VLDL, the molecules become attachedto albumin (S. Eisenberg and J. Patsch, unpublishedresults). The capacity of HDLs to accept apoproteins isalso limited; when a great excess of apoC is releasedfrom the VLDL, apoA-I molecules are displaced fromthe lipoprotein.

H D L 3 -H DL 2 C O N V E R S ONHDL, " HDL2" HDL2

ILPL. CSllSP L , FC. APO -P L . FC, APOIHD L ~ .LP*S] FC +PC -+CE

Corevel: 33.510008

33.510 i-3+LP C

113.100 i-3-1008Fig. 7. A schematic representation of HDLS- DLp interconversion. The scheme suggests that theconversion of HDL s to HDL? is a two-step process. Lipolysis of triglyceride-rich lipoproteins, the surface coatof intact lipoproteins, and cell memb ranes supply HDL s with phospholipids, free cholesterol, and apoproteins.The se m olecules accumulate in existing HDLs particles and, if esterification d w s not occur, they maydistribute to other lipoproteins or cell membranes. This step, therefore, is reversible. True HDL? formationdepends on formation of cholesteryl esters via the LCAT reaction. "HDLr" denotes the phospholipid- andfree cholesterol-enriched HD Ls and is possibly identical to the HDLr, population described by Andersen etal . (246).An important feature of the HDLs -.+HDLp conversion process IS formation of thermodynamicallyfavored HD Lp particles (100 A diameter) containing four molecules o f apoA-I as discussed in the text.Eismberg High density lipoprotein metabolism 1033


18/42

Th e conclusion that HDL3 accepts lipids and proteinsreleased from the surface of lipolyzed triglyceride-richlipoproteins has been confirmed by other investigators(249) nd agrees well with results recorded after initiationof lipolysis by the injection of heparin (7 , 250) andduring the clearance of chylomicronemia (1 0-1 52,228).Yet, in our experiments when the LCAT systemwas either missing or inactive, we did not find a changein cholesteryl ester content or size of the HDL3. Hence,we suggest that incorporation of molecules releasedfrom the surface of lipolyzed triglyceride-rich lipoproteininto HDL3 is insufficient for complete conversion toHDL2. It is therefore important to note that, althoughwe have observed a shift of density of the HDLS towardthe HDL2 position, we considered that true HDL2cannot be formed unless the particles are enriched withcholesteryl ester molecules. The phospholipid- and freecholesterol-enriched HDL3 thus should be consideredas an intermediate in the path of HDLS- DL2conversion. Such HDL intermediates are possibly anal-ogous to the HDL2, population described by Andersonet al. (251) nd are not very different from HDLs thatbecomes enriched with phospholipids after in vitro in-cubation with phospholipid vesicles (252-254). Trueconversion of HDLS to HDLp occurs only after theparticle acquires cholesteryl esters, by the LCAT reac-tion. If that last reaction does not take place, the firstpart of the conversion path is fully reversible, as thesurface lipids (and proteins?) added to the HDLS equil-ibrate with other HDL particles (S. Eisenberg and J.Patsch, unpublished results). Because of all these consid-erations, HDL3- DLp conversion is described in Fig.7 as a two-step process, where the first step is potentiallyreversible.

The mechanism of HDLS- DL2 conversion illus-trated in Fig. 7 does not specify the source of apoproteinand lipids. In fact, lipolysis may be one of severalsources. Cell membranes must be considered as anotherpotential source of lipids that become associated withHDLS and contribute to the conversion process. Thecapacity of HDLs to accept cholesterol from cell mem-branes has been well established and is considered amajor part of the reverse cholesterol transport process(see Section VIII). Whether phospholipids may leavecell membranes and become associated with HDL is notknown. Finally, if apoprotein A-I and A-I1 are secretedfrom cells with no (or minimal amounts of) lipids, theymay also contribute to the HDL conversion path ratherthan to the pool of HDL-precursors. Transfer of lipids(and proteins?) from cells to HDL probably accounts forthe modifications that occur in HDL filtered throughextravascular spaces (255).

Th e evidence that HDL2 is formed in the plasmacompartment by pathways not dissimilar to those de-

scribed above is still indirect. The data that support theHDL3- DL2 conversion have been mentioned above.An experiment that conclusively shows in vivo conversionof HDLS to HDL2, however, has not been reported yet.Very recently, we decided to study this problem byfollowing up human HDLS injected into rats (D. Gavish,Y, Oschry, S. Eisenberg, unpublished results). Prelimi-nary results demonstrate conversion of the injectedHDL3 to HDL2 and, later, to HD Ll. If these resultscan be corroborated in humans, that will be sufficientto verify the essential features of the HDLS- DLzconversion process. It is also possible, although lesslikely, that some HDL2 is formed directly as a large-sized HDL population. As discussed below, mechanismsof conversion of HDL2 to HDLs are also present inhuman plasma.B. Conversion of HDLz to HDLl (apoE-HDL)

HDL, is an apoE-rich HDL population of lesserdensity and larger diameter than HDL2. HDLl is anormal plasma lipoprotein constituent in the rat (3 -33)and has been identified in human plasma (28-30).In rats, the contribution of apoE to total proteins inHDLl is estimated to be 60% and the mean diameterof the lipoprotein is 130-140 A (33).An analogouslipoprotein, HDL,, accumulates in the plasma of most,or even all, animal species when the animal is givencholesterol-rich diets (34). Several recently publishedexperiments indicate that HDLl is formed in plasma,presumably from HDL2 (or heavier HDL populations).The origin of apoE in HDL is most probably lipolysisof triglyceride-rich lipoproteins (256-259).That alonesuggests the intraplasma origin of HDLl. More directevidence for the plasma origin of HDLl is the observationthat in vitro incubation of human plasma results inpronounced increase of HDLl levels, presumably dueto accumulation of LCAT-derived cholesteryl esters inthe lipoprotein (30, 260). Similar shifts of HDLs toHDLl were reported during incubation of canine plasmawith Celite-coated cholesterol particles, and duringclearance of cholesterol from lipid-loaded macrophages(261).More recently, we have studied the intravascularfate of the cholesteryl ester moiety of rat plasma lipo-proteins (262).Because the rat is an animal whoseplasma lacks the core lipid transfer activity (33, 193),shifts of cholesteryl esters between lipoproteins stronglysuggest conversion of the lipoprotein itself. In the study,we consistently observed such shifts of HDL2cholesterylesters to HDLl. Very recently we found that humanHDL3 devoid of apoE also ends in the density range ofHDLl after injection into rats (S. Eisenberg, unpublishedresults). These observations ndicate that apoE-rich HDLlparticles are formed in plasma from apoA-I-rich HDL2or even HDLS particles. In further support of this

1034 Journal of Lipid Research Volume 25 , 1984


19/42

conclusion, are our earlier observations that the choles-teryl ester fatty acid profile of rat plasma HDLl indicatesan LCAT origin of these esters (33). Thus, it seems thatan HDL2- DLl conversion path exists in plasma andis possibly regulated by the same reactions that areresponsible for the HDLS- DL2 conversion.

Two important conclusions can be derived from thesestudies. The first concerns the ability of the HDL systemto carry cholesterol. Conversions of small HDL particlesto larger particles allow transport of several-fold morecholesterol in HDL, when needed. HDL2 to HDLlconversion may provide a much larger capacity to trans-port cholesterol than is achieved by HDLs- DL2conversions. The need to transport even larger quantitiesof cholesterol in plasma is possibly responsible for ac-cumulation of HDL, in cholesterol-fed animals (34).The second conclusion concerns the different apoproteinprofiles of HDL2 and HDLl . If the assumption thatapoA-I-containing HDL is converted to apoE-containingHDLl is correct, then during this conversion apoA-Imust be displaced from the HDL2 and replaced byapoE. Graphic illustration of a hypothesis that accountsfor such conversion is shown in Fig. 8. It is envisionedthat in the rat most of the LCAT-derived cholesterylesters remain in HDL. Hence, apoA-containing HDL inthe rats circulates as HDL2. With further accumulation

of cholesteryl esters, some of the HDL2 particles becomeeven larger, and particles of the size, density, andcomposition of HDLl are formed. During this metabolicconversion, however, apoA-I molecules dissociate fromthe lipoprotein (probably due to the poor association ofapoA-I with larger particles), and are replaced by apoEmolecules. Because apoproteins serve as markers forinteractions of lipoproteins with cells and enzymes, themetabolic significance of the changing apoprotein profileof HDL populations, if substantiated, cannot be over-emphasized.C. Conversion of HDL, to HDLS

Several publications have suggested that HDL2 par-ticles may become depleted of lipids and proteins andform the smaller HDL population, HDL3. Two pathwayswere proposed to cause this conversion. In the first, therole of the heparin-releasable hepatic lipase was empha-sized, and in the second, that of the core lipid transferreaction. The hepatic lipase has been shown repeatedlyto hydrolyze HDL phospholipids and triglycerides in invitro incubation systems (1 10, 163- 166). Although lipidsin other lipoproteins also serve as substrates for thisenzyme (83-86, 163), it has been proposed that HDLlipids are preferable. It should, however, be pointed outthat in in vitro systems HDL phospholipids can also be

VLDL, )

Fig. 8. A schematic representation of HDLp to HD Ll conversion. Th e scheme is based on observations inthe rat (33) and includes the events that are responsible for HDLp formation (lipolysis and cholesterolesterification). These same events cause further enrichment of the HDLp (diameter of -100 A) withcholesteryl esters and formation of large-sized HDLl (diameter - 40 A) particles. Along this path, however,apoE molecu les that originate from lipolyzed VLD L and chy lomicrons displace apoA-I mo lecules from thegrowing HDL particles. By this process, apoE becomes the predominant apoprotein in HDLl while thedisplaced apoA -I is re-utilized for further formation of HD Lp.Eisenberg High density lipoprotein metabolism 1035


20/42

hydrolyzed by lipoprotein lipase (108, 109). When wedetermined the activity of lipoprotein lipase and hepaticlipase against HDL phospholipids, we found that withthe hepatic lipase we were able to hydrolyze 2- o 3-foldmore HDL-phospholipids than with lipoprotein lipase(S. Eisenberg, unpublished results). In a recently pub-lished study, human serum was incubated with hepaticlipase and the change of lipoprotein profiles was followed(263).The enzyme hydrolyzed phospholipids and tri-glycerides in VLDL, LDL, and HDL with preferencetowards both VLDL and HDL. When HDL was frac-tionated, a marked decrease of HDLp and increase ofHDL3 mass was found. It has thus been proposed thatas a result of the activity of the hepatic lipase, HDL2 isconverted to HDLs (263). The conclusion that thehepatic lipase affects HDL metabolism (as well as otherlipoproteins) is supported by several additional unrelatedobservations. Firstly, administration of specific hepaticlipase antiserum to rats (264-266) or primates (267)results in marked increase of HDL phospholipid andpossibly triglyceride levels. Secondly, administration ofheparin to human patients with lipoprotein lipase defi-ciency releases hepatic lipase and reduces HDLz phos-pholipid and protein content, and increases HDL3 phos-pholipids (268).Thirdly, strong negative correlationsare found between the levels of HDL and those ofhepatic lipase (269, 70). Strong associations have alsobeen reported between the increase of the hepatic lipaseactivity and the decrease of HDL2 levels in womentaking progestins with androgeiiic activity, while estro-gens that decrease the hepatic lipase activity causedincreased HDL2 levels (27 -275).Estrogens, however,also increase apoA-I synthesis (see Section VII). Finally,remarkably high HDL2 levels were reported in twohuman patients with hepatic lipase deficiency (276).

The data described above strongly support the conceptthat the hepatic lipase plays an important role in HDLmetabolism and is involved with processes that causereverse conversions of HDL, Le., formation of smallerparticles from larger particles. Yet the main feature thatdistinguishes HDL2 from HDLs is their content ofcholesteryl esters. A true conversion of HDLp to HDL3,therefore, must be associated with loss of cholesterylesters from HDL2. In the study of Groot, Scheek, andJansen (263) or example, the shift in density of HDLptowards HDL3, observed after incubation of humanplasma with hepatic lipase, may simply reflect hydrolysisof lipids and not a true conversion of HDL2 to HDLs.Our approach was to seek metabolic pathways that areresponsible for removal of core lipids from large-sizedHDL populations. Our first study was carried out inplasma obtained from patients with abetalipoproteinemia,a disease where abnormally large-sized HDL particlesare found (35). Addition of V L D L to such plasma

induced transfer of cholesteryl esters from HDL toVLDL and of triglycerides from VLDL to HDL. Yetthe HDL retained its size and density. However, whenthe incubation system was supplemented with lipoproteinlipase and HDL lipolysis was allowed to take place (afterremoval of the VLDL), a decrease in the diameter andthe density of the lipoprotein was observed. This reactionwas studied by us more recently in greater detail whileusing human plasma HDLZ and a variety of in vitroincubation procedures (203).The study demonstratedthat particles similar to HDL3 (by composition, size, anddensity) can be formed from HDLz when the activity ofthe core lipid transfer proteins is combined with lipolysisof the transferred triglycerides. That conversion wasindependent of the source of triglyceride-rich particles(VLDL or Intralipid), the source of transfer activity(lipoprotein-deficient plasma or partially purified lipidtransfer proteins), and the source of lipolytic enzymes(lipoprotein lipase or hepatic lipase). In,vivo, however,it is reasonable to assume that the hepatic lipase plays adominant role in this process.Fig. 9 presents a hypothesis that combines the activitiesnecessary for HDL2+HDLs conversion. This hypothesisis applicable in general to situations that favor formationof small HDL particles from larger particles. A two-stepprocess is proposed. The first step involves the activityof the plasma lipid transfer protein and causes replace-ment of HDL-cholesteryl ester by triglyceride molecules.Triglyceride-rich lipoproteins serve as both cholesterylester acceptors and triglyceride donors. It is suggestedthat the cholesteryl ester-poor, triglyceride-rich HDLserves as intermediate in the conversion process. Thesecond step is triglyceride hydrolysis and deletion ofcore lipid molecules from the HDL. It is in this stage ofthe conversion process that the hepatic lipase plays amajor role, although lipoprotein lipases may also con-tribute. An essential feature of the second step is removalof surplus surface constituents, a process that may involvephospholipid hydrolysis by the hepatic lipase followedby exclusion of free cholesterol and apoprotein mole-cules. In the circulation, the two steps probably takeplace concomitantly,and as triglyceride molecules replacecholesteryl esters, they are hydrolyzed by lipases, andthe particle loses core lipid molecules. A quantumjump may take place with the exclusion of an apoA-Imolecule at a favorable thermodynamic status of theparticle, as described below.D. Regulation of HDL subpopulation distribution:role of enzymes and transfer proteins

In the foregoing discussion, the role of enzymes andof lipid transfer proteins in HDL conversion processeshas been described. Other metabolic pathways thataffect HDL conversion are generation and availability



21/42

HDL2 +HDL3 CONVERSIONHDL2C E - i c h

ICETI

HDL2TG - r i ch

HDL-CE VLDL-TGuTG

HDL3

TG +FAP L - F A + LLRemoval of surplussur face

- 0 11111111111 - 00 100 A 0 100A 0 80

Fig. 9. A schematic representation of HDL2 to HDL , co nversion. This again is a two-step process. Th e firststep depends on the lipid transfer reaction (GET) and replacement of part of the HDL2 cholesteryl estermolecules by triglycerides. VLD L (and chylom icrons) participate in this reaction as acceptors for cholesterylesters and donors of triglycerides. During this remodelling step, the size and density of the HDL2 does no tdecrease; a transient increase may occur. Hydrolysis of the transferred triglycerides (predominantly by thehepatic lipase (HL) but also by lipoprotein lipases (LPL)), however, results in decreased size an d density ofthe lipoprotein. Continuing replacement of cholesteryl ester by triglycerides and triglyceride hydrolysiscoupled with removal of surplus surface (? phospho lipolysis followed by displacem ent of ap oproteins) resultsin complete conversion of HDL? to H D L s .

of phospholipids and free cholesterol, cholesterol ester-ification, availability of apolipoproteins, and sources andrelative masses of triglyceride-rich lipoproteins that par-ticipate in the lipid transfer reaction. The nature ofinteractions and affinities of the enzymes and transferproteins with different lipoproteins is essential for theregulatory processes. Only when the coordinated effectsof all reactions are considered, can HDL subpopulationdistributions be properly analyzed. The activity of theLPL and LCAT systems, coupled with a proper supplyof phospholipids, free cholesterol, and apoproteins, re-sults in the predominance of large-sized HDL popula-tions. The activity of the hepatic lipase and the corelipid transfer proteins, in the presence of a relativelylarge mass of triglyceride-rich lipoproteins, acts in theopposite direction. The HDL system, therefore, reflectsthe equilibria of these opposing pathways, as suggestedoriginally by us while studying the plasma lipoproteinsin abetalipoproteinemia (35). In this condition, chylo-microns, VLDL, and LDL are absent from plasma. HDLlevels are low and there is an abnormally large-sizedHDL population (diameter of 130-140 A). Apoproteinturnover studies demonstrated normal catabolism ofapoA-I, with reduced synthetic rates (277). Lipoproteinlipase and hepatic lipase activities are moderately de-

creased (278) and the lipid transfer proteins are present(S. Eisenberg, unpublished results). LCAT is also present.The main determinant of the HDL subpopulation dis-tribution, therefore, seems to be absence of lipoproteinacceptors for transferred cholesteryl esters and donorsof triglycerides.

Fig. 10 has been prepared to illustrate the pathwaysthat regulate the HDL subpopulation distribution innormal subjects and patients with abetalipoproteinemia(35). The pathways included in Fig. 10 and in Figs. 7-9 also provide a physiologic background for the alterationof the HDL system in many other conditions. It isworthwhile to discuss briefly other examples that illus-trate different determinants of HDL subpopulation dis-tribution. In rats, HDL exists only as large-sized popu-lations, HDL2 and HD Ll , compatible with the absenceof the core-lipid transfer reaction (33). Another exampleis the HDL system in patients with severe hypertriglycer-idemia due to lack of lipoprotein lipase activity. HDLin such patients is exclusively in the HDL3 range; it isdenser than normal HDLS and is cholesteryl ester-poorand triglyceride-rich (279, 280). Limited availability ofsurface remnants and the presence in plasma of anextremely large mass of triglyceride-rich lipoproteinsthat accelerates core-lipid transfer together with normal,

Eiscnberg High density lipoprotein metabolism 1037


22/42

PROPOSED SCHEME OF HDL CONVERSIONS IN NORMALS ANDIN ABETAL IPOPROTEINEM A E x o g e n o u s c h o l e s t e r o l an d p h o s p h o l i p i d s f r o m l i p o l y z e d T G - r i c hl i p o p r o t e i n s an d c e l l m e m b r a n e sIIII CE ENRICHMENT (LCA T)I

4

HDL-2 CE REMOVALII CE t r a n s f er a n d l o rI C E d T G e xc h an g e a n dI TG h yd r o l ys i s

b

N O R M A L1Fig. 10. An integrated scheme of the regulation of HDL subpopulation distribution. Th e scheme is basedon observations in normo lipidemic and abetalipoproteinem ic subjects (35) and is reproduced (by permission)from the Journal of Lipid Research (1982. J . Lipid Res. 43: 1280). The scheme predicts that HDLsubpopulation distribution is determ ined by a balance of processes responsible for enrichment and removalof cho lesteryl esters from H DL , as detailed in Figs. 7-9.

or close to normal, hepatic lipase activity probablyaccounts for the abnormalities of HDL in these patients.Finally, the well-known inverse relationship betweenplasma triglyceride and HDL levels and the predomi-nance of HDL3 in subjects with milder forms of hyper-triglyceridemia ar e also determined by cholesterol andtriglyceride redistribution and lipolysis of the transferredtriglycerides (215, 216).

None of the metabolic pathways discussed in thissection and summarized in Figs. 7-10 explain the pres-ence of discrete, separable HDL subpopulations inplasma. It is thus not clear why HDL should notdistribute continuously and symmetrically along thedensity range of 1.063-1.21 g/ml. In fact, microheter-ogeneity of HDL2 and HDL3 has been documented (26)and it appears that particles exist within each populationof slightly higher and lower density, size, and composi-tion. We have suggested that addition or deletion ofone apoA-I molecule is responsible for transformationsbetween HDL2 and HDL3 (15, 16). Our assumption isthat t w o thermodynamically favored conditions occurwhen the HDL particle contains three (HDLs) or four(HDLp) molecules of apoA-I. According to this view,HDL3 acquires few cholesteryl ester molecules by theLCAT reaction. That accounts for the microhetero-geneity of the lipoprotein. Concomitantly, phospholipid

and free cholesterol are added to the surface layer. Ata certain stage, one apoA-I molecule is also added. Th ataccelerates accumulation of cholesteryl esters and resultsin the quantum jump to HDL2. HDL2 microhetero-geneity (towards larger (HDLI)or smaller (HDL3) par-ticles) can of course exist until a further conversion orreverse-conversion occurs.

VII. INTRAVASCULAR METABOLISMOF HDL APOLIPOPROTEINS

The considerations presented above indicate thatnone of the HDL lipids or proteins can be employed asa stable plasma marker to study the intravascularmetabolism of HDL particles in humans (or other specieswhose plasma contains core-lipid transfer activity). Thefate in plasma of apoA-I and apoA-I1 therefore will bedescribed with care in order not to equate the proteinkinetic parameters with those of the lipoprotein.

Blum et al. (281) were the first to report kineticparameters separately for apoA-I and apoA-11; in ear lierstudies, total HDL protein metabolism was investigated(282-284). In the study of Blum et al. (281), the wholeHDL particle (d 1.09-1.2 1 g/ml) was radioiodinated,and the decay of labeled apoA-I and apoA-I1 was

1038 Journal of Lipid Research V o l u m e 2 5 , 1984

7/28/2019 J. Lipid Res. 1984 Eisenberg 10

J. Lipid Res. 1984 Eisenberg 1017 58

Documents

Transcript of J. Lipid Res. 1984 Eisenberg 1017 58