Kinetic Model for Production and Metabolism

of Very Low Density Lipoprotein Triglycerides

EVIDENCEFORA SLOWPRODUCTIONPATHWAY

ANDRESULTSFORNORMOLIPIDEMICSUBJECTS

LORENA. ZECH, SCOTTM. GRUNDY,DANIEL STEINBERG, and MONESBERMAN,Laboratory of Theoretical Biology, Division of Cancer Biology and Diagnosis,National Cancer Institute, National Institutes of Health, Bethesda, Maryland20205; and Division of Metabolic Disease, Department of Medicine, La Jolla,California 92093

A B S T R A C T A model for the synthesis and degrada-tion of very low density lipoprotein triglyceride(VLDL-TG) in man is proposed to explain plasmaVLDL-TG radioactivity data from studies conductedover a 48-h interval after injection of glycerol labeledwith I"C, 3H, or both. The curve describing the radio-activity of plasma VLDL triglycerides reaches amaximum at about 2 h, after which the decay is biphasicin all cases; the late curvature becoming evident onlyafter 8-12 h. To fit the complex curve, it wasnecessary to postulate two pathways for the incorpora-tion of plasma glycerol into VLDL-TG, one muchslower than the other. A process of stepwise delipidationof VLDL in the plasma compartment, previouslyproposed for VLDL apoprotein models, was alsonecessary. Predicted VLDL-TG synthesis rates calcu-lated with this model can differ significantly fromthose based on experiments of shorter duration inwhich the slow VLDL-TG component is not apparent.The results of these studies strongly support theinterpretation that the late, slow component of theVLDL-TG activity curve is predominantly due to theslowly turning-over precursor compartment in theconversion pathway and is not due either to a slowcompartment in the labeled precursor, plasma freeglycerol, or to an exchange of plasma VLDL-TG withan extravascular compartment. It also cannot, in thesestudies, be attributed to a slowly turning-overVLDL-TG moiety in the plasma. The model was

This work was presented, in patrt, at the 49th ScientificSession of the American Heart Association, MIiami Beach,Fla., November 1976.

Received for publication 19 June 1978 ancd in revise(lformn21 February 1979.

tested with data from 59 studies including normalsul)jects anid patients with obesity and(or) various formsof hyperlipoproteinemia. Good fits were obtained in allcases, and the estimated parameter values and theiruncertainties for 13 normolipemic nonobese subjectsare presented. Sensitivity testing was carried out todetermine how critical various parameter estimationsare to the assumptions introduced in the modeling.

INTRODUCTION

A number of methods have been employed forestimating the transport of the triglyceride (TG)' (1, 2)moiety and the protein moieties (3, 4) of very lowdensity plasma lipoproteins (VLDL) in man. Eachinethod implies certain assumptions about the physi-ology of the processes, and the results can be highlysensitive to these assumptions.

One approach for estimating transport of VLDL-TGis to inject labeled TG precursors and to analyze theresulting plasma VLDL-TG specific radioactivitycurves. Some investigators (5, 6) have used labeledglycerol as precursor and concluded that the specificactivity curve of VLDL-TG, after an initial build-up,decayed essentially as a single exponential over theshort duration of their studies. They assumed that theexponenitial rate constant of this early (lisappearancecurve could be taken to represent the fractional

' Abbreviationts used itn this paper: apoB, apoprotein B;FCR, fractional catabolic rate; HDL, high density lipo-protein; IDL, initermediate density lipoprotein; LDL, lowdenisity lipoprotein; TG, triglyceride; VLDL, very low densitylipoproteini.

The Journial of Clinical Investigation Volume 63 Junte 1979 1262 -12731262

catabolic rate (FCR)2 of VLDL-TG. Eaton et al. (7),Shames et al. (1), and Quarfordt et al. (8) usedlabeled FFA as a precursor for VLDL-TG and kineticcurves were analyzed using a compartmental model.In these studies the disappearance curve after thepeak did not remain monoexponential and showed adecided decrease in slope beginning at about 8 h.Several compatible models were proposed andexamined to account for the observed kinetics.Alternative features considered in these modelsincluded two pathways for incorporation of FFA intoVLDL-TG, an extravascular compartment exchangingwith plasma VLDL-TG, and two VLDL-TG moietieswithin the plasma (Fig. 1 and [8]). The calculatedsecretion rates of VLDL-TG were shown to be highlydependent on the model used.

Subsequent to the FFA studies of Shames et al. (1),salient, adjunct studies of the turnover and heterogeneityof VLDL labeled in the apoprotein B and C moietieswere subjected to rigorous analyses by Phair et al.(9, 10) and Bermai et al. (3) with compartmentalmodels. Features contained in these models included achain of VLDL compartments to represent the progres-sive modification of these plasma lipoproteins (due tothe action of lipoprotein lipase) and a slowly turning-over plasma VLDL component, especially prominentin patients with Type III hyperlipoproteinemia (10).A somewhat less analytical discussion of the turnoverand heterogeneity of VLDLapoprotein B (apo B) moietyhas been presented by Reardon et al. (11).

In these studies, radiolabeled glycerol was used asa precursor for VLDL-TG and the kinetics were fol-lowed for 48 h, both in normal and hyperlipidemicsubjects. 13 normolipemic, nonobese subjects and 46obese and(or) hypertriglyceridemic subjects werestudied. In every case the radioactivity curveafter the initial peak showed at least a bi-exponentialdecay. These studies together with the VLDL-apoB

2Special terms used in this paper: bi, slope of a linetangent to the log of the tracer curve at region i (i = IPdesignates the inflection point, i = T designates the tail, andi = I, II, III designates the phase of the curve); FCRf,fractional catabolic rate of substance x relative to subsystemi. FCRf = 1ITx (h-'); D1,o delay of component 10 in the TGsynthesis pathway (h); Lij, fractional rate of transport ofactivity or mass to compartment i from compartmentj (h-');Ljj, fractional rate of irreversible and reversible trans-port of activity or mass out of compartment j (h-'); LOJ,fractional rate of irreversible transport of activity or mass outof system or subsystem from compartment j (h-'); R..,steady-state transport of substance x to compartment i fromcompartmentj (mg/h); R,, steady-state transport of substancex out of or into subsystem j (mg/h); Rfj, irreversible andreversible steady-state transport of substance x out of com-partmentj (mg/h); Rxj, irreversible steady-state transport ofsubstance x out of a system or subsystem from compart-mentj (mg/h); T,, average time substance x resides in sub-system i (h).

studies constitute a new body of data that per-mit a sharper discrimination between the previ-ously proposed models for the synthesis and degrada-tion of VLDL-TG. The proposed model is used toestimate parameter values for normal subjects. Resultsof this noninvasive technique are compared favorablywith more direct techniques, as discussed in ourcompanion paper (12).

METHODSPatients. All patients were studied as in-patients under

metabolic ward conditions on the Special Diagnostic andTreatment Unit of the Veterans Administration Hospital,San Diego, Calif. Informed consent was obtained from eachpatient. A total of 59 patients was studied. Of these, 13were normolipemic and nonobese and results for thesesubjects are presented here. The other 46 included obese andhypertriglyceridemic nonobese patients. Clinical character-istics of these patients and TG turnover results are pre-sented in our companion paper (12).

Study design. To insure steady state at the time of study,patients were admitted at least 7 d previously and put onmeasured weight-maintenance diets (40% of calories as fat[mostly lard], 40% carbohydrate, and 20% protein). Thesediets before study consisted of solid foods, liquid formula,or a combination of the two. Because studies had to extend48 h after injection of the labeled glycerol, it was not prac-ticable to have the subjects fasting and probably not de-sirable, because with prolonged fasting patients are not in atrue steady state with regard to lipoprotein levels, andplasma lipoprotein secretion will fall. On the other hand,continuation of their prestudy, fat-containing diet would intro-duce ambiguities due to the presence of dietary chylo-microns and their degradation products. The solution wasto establish a steady state on a regimen of fat-free liquidformula feedings given every 3 h around the clock beginning36 h before injection of labeled glycerol. The formula pro-vided 1.25 cal/g, 75% of calories as glucose, and 25% asprotein (RI-5, Ross Laboratories, Columbus, Ohio). In pre-liminary studies the formula feedings were started only 12 hbefore injection and in some of these subjects VLDL-TGlevels showed a slight rise during the study. With 36 h ofpreparation VLDL-TG levels were remarkably constant overthe study period. The weight of the patients and theirplasma VLDL cholesterol levels were also stable over the 48-hperiod of the study. The liquid formula intake was cal-culated to provide 60% of the calories needed to main-tain weight during the prestudy period, i.e., the calories fromfat were deleted for the reasons given above. This tech-nique, which was developed (13) has been employed re-cently by Reardon et al. (11) for measuring turnover ofapoB. As mentioned, this technique insured constant levelsof VLDL-TG throughout the study.

On the day of study, labeled glycerol (New EnglandNuclear, Boston, Mass.) dissolved in 0.15 M NaCl wasinjected rapidly via an antecubital vein: [1,3-'4C]glycerol(100 ACi), [2-3H]glycerol (300 ,uCi), or both. Blood samples(7 ml) were drawn into EDTA through an indwellingneedle in the opposite antecubital vein at 15 and 30 minand at 1.0, 1.5, 2.0, 2.5, 4, 5, 6, 7, 9, 11, 13, and 15h andthen every 3-48 h.

Analysis of samples. VLDL was isolated by preparativeultracentrifugation (14). A known volume of plasma (2-3 ml)was mixed with a solution of sodium chloride of density1.006 to a total volume of 6 ml in a 6.5-ml cellulose nitrate

Very Low Density Lipoprotein Triglyceride Synthesis and Metabolism 1263

tube (Beckman Instruments, Inc., Spinco Div., Palo Alto,Calif.). Samples were centrifuged for 16 h at 40,000 rpm and5°C in a Beckman L2-65B preparative ultracentrifuge (40.3rotor). The top 1.33 ml from each tube was removed quanti-tatively by tube slicing. This fraction was extracted with 20ml isopropranol, and phospholipids were removed by ad-sorption onto zeolite. A portion of the lipid extract wastaken for measurement of TG and cholesterol on an Auto-Analyzer II (Technicon Instruments Corp., Tarrytown, N. Y.).The remainder was evaporated and dissolved in toluenephosphor for analysis of radioactivity in a liquid scintilla-tion counter (Mark II, Searle Radiographics Inc., DesPlaines, Ill.)

Numerous experiments were carried out to develop amethod to maximize recovery of VLDL-TG for individualsamples. This procedure was found to consistently give>95% recovery on individual samples. Each individualvalue for VLDL-TG, in each set of 25 samples, was cor-rected by the mean ratio of VLDL-TG per total plasmaTG obtained from the 5 samples with the highest ratios,according to this equation: VLDL-TG (on an individualsample) = total plasma TG (on the same sample) x maximumVLDL-TG per total plasma TG. This value corrects forsmall losses that inevitably occur with some samples duringthe tube-slicing procedure. The correction assumes thatthe ratio of VLDL-TG per total plasma TG remains con-stant throughout the study. Careful examination of all ourdata has confirmed this assumption.

In several studies the FFA were examined and in nocase was there detectable radioactivity.

Analysis of data. The data were analyzed using a linear,first-order compartmental model.3 Such a model can bedescribed mathematically by a simultaneous set of ordinaryfirst-order differential equations vith constant coefficients(15). The solution of these equations and the adjustmentof the model parameters to yield a least-squares fit of thedata were carried out with the aid of the Simulation, Analy-sis, and Modeling (SAAM) computer program (16). For in-patients subjected to more than one study the parametervalues were determined with greater precision becausethey were arrived at by the simultaneous fitting of alltheir data.

Weighted averages for population values were determinedusing the equation

Xi 1XEXi /E -1 (1)

i Vi +v i jv +vwhere vi is the variance of the parameter x,, and v3 is theestimated population variance (17).

RESULTS

Model development. The model proposed for theanalysis of the data is based on the features of thekinetic data collected, on steady-state masses of plasmaVLDL-TG, and on information previously published(1- 10). A typical activity curve for VLDL-TG isplotted in Fig. 1 as fraction of the injected dose oftracer per milliliter of plasma. It contains three phases:

3For a single compartment model having an input u(t)and a first-order loss process characterized by a constant k,the implied differential equation is: df(t)ldt = -kf(t) + u(t),where f(t) is the amount of substance in the compartmentat time t.

E 10-4

:

a 0t5

; 10-7

0D-13 ,

24Time (h)

36 48

FIGURE 1 Typical VLDL-TG tracer curve after the injectionof labeled glycerol. Three phases are identified: initialrise (I), early fast decay (II), and late slow decay (tail) (III).

an early rising phase, a rapidly decaying phase after thepeak, and a late, more slowly decaying phase formingthe "tail" of the curve. The exact shape variedfrom subject to subject. The multicompartmiientalmodel finally arrived at to explain these data isshown in Fig. 2. It contains three subsystems: thefree glycerol precursor subsystem, an intermediatesubsystem where the conversion of glycerol to VLDL-TG takes place, and the plasma VLDL-TG suibsystem.The justification for the componenits of each of thesesubsystems is presented in the following paragraphs.In developing the model, advantage was taken of pre-vious studies of glycerol and plasma lipoproteinkinetics.

Malmendier et al. (18) studied the kinetics of intra-venously injected glycerol in plasma and found that atwo-compartment model fit the data. They showed thatabout 90% of the glycerol leaving the plasma couldbe accounted for by conversion to glucose and CO2,which is consistent with the results of these studieswhich require

0.37h-'(109 mg/h)

FIGURE 2 Detailed compartmental model for subsystems. Average fractional transport rates,Lij, (per hour) for nonnolipemic, nonobese subjects are given with the transport rates, Rij, (milli-grams per hour) in parentheses. UTG is VLDL-TG synthesis rate from nonplasma glycerol sources,and RGA2YXLDL is glycerol recycling from VLDL-TG to plasma glycerol.

Four compartments were necessary and sufficientto describe the delipidation chain. Each compart-ment is characterized by two parameters: L6,1 andL4,1, representing, respectively, the fractional rate oftransfer of TG to the next lower VLDL product andthe fraction leaving the VLDL-TG pool irreversibly.These parameters are assumed to be the same ateach of the four steps.

The residence time4 of VLDL particles in the chain,TcLDL, and the residence time of TG molecules in thechain, TcG, are related to the model parameters inthe following equationsi which have been derivedelsewhere (24):

4TCLDL = where L1, l= L6,1 + L4,1, (2)

and_T( 1 - (6lL )T- .(3)

L4,1

The total residence time of VLDL-TG in the plasmais a weighted average of times spent in the chain andin compartment 21. If TTG is the residence time of TGin compartment 21, and if P21 is the fraction of VLDL-TG diverted to that compartment, then

TTG = (1 - P21)TC + P21T21

I + [21 _ I P21 )TTG (4)

(TTG/TTG -1)P,, is thuis a correction term for theresidence time of plasmla VLDL-TG whein a fraction,P21, of the labeled TG contributes to the tail of theVLDL-TG curve in the form of compaitment 21.

4Residence tim11e as used here equals the reciprocal ofthe FCR.

Usually,5 T2GI/TcG 10, and therefore the correctionfactor is very sensitive to the value of P21.

The plasma glycerol to plasma VLDL-TG conversionsubsystem is a modification of the one proposed byShames et al. (1) for FFA. It consists of a fast pathway(Fig. 2, through delay component 10 and compartment14) and a slow pathway (through compartment 24).Channeling of isotope through the former fits the rapidinitial rise in VLDL-TG activity (after a short delay);channeling through the latter fits the final slope of theVLDL-TG curve. There are equivalent ways of model-ing this subsystem that cannot be distinguished kineti-cally. For example, compartment 24 (the slow pathwaycompartment) could be introduced as a compartmentin exchange with compartment 14 (the fast pathwaycompartment). Alternatively, compartment 24 could befed from plasma glycerol, as shown in Fig. 2, butrelease its material into compartment 14 rather thaninto plasma VLDL-TG. A variant of this has beensuggested by Nikkila and Kekki (25). The importantpoint is that these alternatives neither alter the fitof the experimental data nor the calculated frac-tional or absolute catabolic rates for VLDL-TG.

In a number of our studies a "hump" in theVLDL-TG activity curve appears between 12 and 30 h(Fig. 3). The model proposed does not account forthis hump. The hump has previously been noticed byNikkila and Kekki (5) in studies on normal adults,and they proposed that recirculation from liver glyco-gen may be the source. This, however cannot be anadequate explanation because the hump is also pres-ent in studies with 3H-labeled glycerol which wouldlose its 3H label in the glycogen pathway. The sameargument would also apply against recycling through

'The value of TTG is given by Eq. 3 and the value for1= i/L.24 which equals the reciprocal of the slope of the

tail of the VLDL-TG curve (b,11, Fig. 1).

Very Low Density Lipoprotein Triglyceride Synthesis and Metabolism 1265

CD

ur

CD 10'7-J

CDE

0o-8 D-500 12 24 36 48

Time lh)

FIGURE 3 A representative set of plasma VLDL-TG data(A) with a hump between 800 and 1,800 min. Two curvesare drawn through the observed values. The curve labeled"four" was generated with a delipidation chain in theplasma VLDL subsystem (Fig. 2) containing four compart-ments; the other, labeled "one", was generated by replacingthe four compartments with a single one. The parameters ofboth models were adjusted to best fit the data shown.

the glucose pathway. Because various recycling ordelay pathways, such as through phospholipid pools,could be proposed to account for the hump, and ourdata are not sufficient to differentiate between them,no effort is made to fit it at this time. Because thehump represents at most 5% of the area under theVLDL-TG activity curve, its neglect in modeling thedata introduces an error of

TABLE ITypical Parameter Values and Sensitivities*t

(1) (2) (3) (4) (5) (6) (7) (8) (9)Timile of

Fractional NMagniitude Late Early inflectioniParameter§ Value SD FCR,,, P21 of tail slope slope point

h-' h-' h-' h

RLDL 646 mg/h 0.14 0.18 7.68 0.97 0.01 0.21 0.57114TG4 23 mg/h 0.14 0.93 0.37 0.13 0.01 1.66 0.05124TG4 31 mg/h 0.16 0.92 10.02 1.80 1.06 0.02 0.50

UTG 591 mg/h 0.14 0.23 7.87 0.96 0.07 0.30 0.06FCRTG 0.18/h 0.14 0.18 7.68 0.97 0.01 0.21 0.57Li,, 0.31/h 0.67 0.14 7.58 2.97 0.34 8.51 0.80L4,1/Ll,l 0.52 0.83 0.05 6.48 3.25 0.43 6.75 1.63LIMG4 0.060/h 0.14 0.93 3.64 0.12 0.01 1.63 0.05L24,4 0.78/h 0.16 0.92 10.02 1.79 1.06 0.11 0.50LIMl4 1.08/h 0.32 0.27 2.52 0.18 0.11 3.30 0.46L1,24 0.018/h 0.47 0.15 0.23 0.81 1.83 2.57 0.26D1o 0.37/h 0.17 0.01 2.77 0.10 0.02 0.78 0.44RglyTVLDLi 625 mg/h 0.20 0.19 7.22 0.69 0.08 4.06 0.96

* All values are based on curves shown in Figs. 1 and 6. The approach is discussed in (27).Sensitivity = (percentage of change in paraimeter value/percentage of change in

perturbed parameters).§ The parameters perturbed (one at a time) were: column 4: FCR of plasma glycerol(h-1); column 5: Fraction of VLDL-TG shunted into slowvly-metabolized VLDL (compart-ment 21); column 6: Activity of VLDL-TG in the late part of the curve relative to thatat the peak; column 7: Slope of the late part of the curve (h-'); column 8: Slope of theearly fast decaying part of the curve (h-'); column 9: Time after injection at which theinflection point in decay curve occurs."IRT, VLDL = 1 RTG; i = 1, 6, 7, and 8.

calculating sensitivities, the curve in Fig. 1 and theparameters related to it were used as a base of refer-ence. Thus, for example if the FCR for free glycerol,FCRgiy, were to be increased by 1% an(d the data forFig. 1 refitted subject to this constraint, the calculatedTG transport would only be changed by 0.18% (TableI, column 4, line 1).

A similar procedure was followed for determiningthe sensitivities of calculated paraimeters to changesin the shape of the experimental curve (27). A givenchange was introduced in the curve and the modelparameters were allowed to readjust to fit the newcurve. The changes introduced into the curve of Fig.1 are shown to scale in Figs. 5A and B and areexpressed as a fraction of a slope, intercept, etc.Examination of Table I allows insights into thestrengths of "linkages" or interdependencies withinthe model.

Parameter values in normolipideinic subjects. Themean values and fractional standard deviations of themodel parameters for 13 normolipidemic subjects aregiven in Table II. The mean FCR for this total VLDL-TG pool was 0.19/h. Total transport (28) of TG aver-aged 806 mg/h. Only a small fractioni of plasmaglycerol (R TG4 plus RTG4) was used for VLDL-TG

synthesis. Input from unlabeled sources, UT6 (glyco-genolysis, glycolysis, aind glyceroneogenlesis) accountsfor about 90% of the total. Plasmiia glycerol usedfor VLDL-TG via the slow pathway (R TG4) iS aboutone-half of that via the frast pathway (RT,14).

Simultaneous 14C and 3H studies. In about a dozeniindividuals simnultaneous responses to labeled ([2-3H]- and [1,3-'4C]glycerol) were obtained as shown inFig. 6. The two responses were not identical (29).The initial VLDL-TG uptakes were comparable forthe two labels, but the slow exponential componeintwas about 40% lower for [3H]- compared to [14C]glyc-erol. These results are highly significant lbecause oftheir implication with respect to the slow synthesispathway which is disctussed below.

DISCUSSION

In a series of papers published by Shamlles et al. (1),Quarfordt et al. (8), and Berman et al. (3), a number ofmodels have been investigated for the study of VLDLmetabolismii anid the calculation- of its turniover. Datacollected over a 24-h interval showedl that specificactivity curves of VLDL-TG and VLDL-apoB con-sistently had a slow comnponient in the later portion of

Very Lows Density Lipo protein Triglyceride Synthesis and Metabolism 161267

-\ Late- Slope

-Magnitude >of Tail

24 36 48Time (h)

FIGURE 5 The tracer curve of Fig. 1 separated into an earlyportion (A) and a late portion (B). The dashed curves repre-sent 10% changes in: the magnitude of the tail, the lateslope, the early slope, and the time at which the inflectionpoint occurs. The parameter values for the solid and dashedline solutions were used to calculate the sensitivities shownin Table III (columns 3, 4, 5, and 6, respectively).

the curves. This response could be simulated withcompartmental models in a number of ways whichimplied various physiological interpretations. Un-fortunately, the value calculated for VLDL-TG syn-thesis greatly depends on the model chosen. It istherefore essential that the correct mechanisms forthe generation of the tail of the curve be known. Wepresent here a series of arguments for the existenceof a slow pathway for conversion of plasma glycerolto VLDL-TG which is a major contributor to the tailseen in most of our VLDL-TG tracer curves. Theremay be exceptions, however, particularly for patientswith relatively large amounts of slowly turning-overVLDL (,B-VLDL), or for subjects in which the tail ofthe VLDL tracer curve is very small. Although it isfully appreciated that a model is only one hypothesisand that many models may be proposed to explain

TABLE IIEstimated Parameter Values for Normal Population

(n = 13)

WeightedPhrase Parameter imean* FSDI

FCR (Chain) FCRTG 0.19/h 0.065Chain residence time TT 5.16 h 0.065VLDL-TG synthesis RTGVLDL 805.8 mg/h 0.15Recycling RTGVLDL 697.2 mg/h 0.20Fractional transport of

compartment 1 LI,, 0.48/h 0.051Delipidation fraction per

state L4,1/L1,l 0.22 0.16LIOA4 0.12/h 0.11L24,4 0.047/h 0.20L 1,14 2.32/h 0.25

Fractional transport rateof slow compartment Ll,24(bT) 0.025/h 0.062

Delay DIo 0.41 h 0.085Fast TG transport RLT4 50.2 mg/h 0.095Slow TG transport RTG4 18.8 mg/h 0.20Nonplasma TG transport UTG 666.8 mg/h 0.13Fast slope b,, 0.18/h 0.11VLDL-CH/VLDL-TG CH/TG 0.21 0.072VLDL-TG mass MTGDL 112.9 0.089M2/M1VLDL a 0.046 0.072

* The mean values are calculated using Eq. 1.t Fractional standard deviation.

data, we feel that the weight of the evidence makesthe conclusions about the major metabolic featuresindependent of the models proposed. A more completevalidation of the proposed pathways must come, asalways, from additional direct experimental evidence.

E l-,E~10-u

0

C

0

_1 1O-7

-J

aIE0

cL

t A

\k I

In general, the following are potential contributorsto a slow decaying tail of the VLDL-TG activitycurve:

Exchanges of VLDL with extra-plasma spaces, or ofTG with non-VLDL lipoproteins in plasma. Thestrongest argument against these mechanisms beingresponsible for most of the tail comes from thedoubly labeled experiments in these studies. Thesesubjects received both '4C- and 3H-labeled glyceroland, on the average, the 3H-labeled glycerol pre-cursors produced VLDL-TG tails about one-third aslarge as those produced by "4C-labeled glycerol,whereas the magnitude of the initial peaks werecomparable (Fig. 6). If the tails were largely due to aVLDL-TG exchange processes there is no reasonableexplanation for the discrimination between 14Cand 3H.

An argument against exchange of plasma VLDL withextravascular VLDL can be made from the publisheddata of Phair et al. (10) who studied the decayof exogenously labeled VLDL-apoB in man. Althoughthe decay curve exhibited a tail, these workersexcluded an exchange process mainly because radio-activity in the tail could not serve as a precursorfor intermediate density lipoprotein (IDL)-apoB ascould that in the early part of the curve. Reardonet al. (11) have also noted a slow component fromexogenously labeled VLDL-apoB. They suggested thata two-compartment exchange model could account forthis, but they did not claim it was the only explana-tion. From the work of Phair et al. (10) we suggestthat two metabolically distinct, intravascular moietiesof VLDL more likely explain their data. It shouldbe pointed out that the tail on the VLDL-apoBcurve caused by the slowly turning-over moiety inexogenously labeled VLDL is much larger than wouldbe produced by endogenous labeling. This is be-cause even a small pathway for generation of such amoiety with a slow removal produces a rather largesteady-state mass (24); the external labeling of thismass and its reinjection thus accentuates the size ofthe tail. WhenVLDL-apoB is labeled endogenously thesize of the tail due to the slow removal pathwayis much smaller (30).

Direct evidence for low concentrations of VLDL inextravascular spaces is contained in the studies ofReichl et al. (15, 31, 32) who found little or noVLDL in human peripheral lymph. In our laboratory,6studies have been carried out on pleural transudatesfrom eight patients, and we found essentially no VLDLdespite the presence of low density lipoprotein (LDL)and high density lipoprotein (HDL) at levels of _25%of their plasma concentrations. Therefore, it seems

6 Unpublished observations.

unlikely that a significanit portion of the tail could beexplained by exchange vith extravascular VLDL.

Independent exchanige of TG between VLDL andother plasmna lipoproteins could add to the tail of theVLDL-TG curve. The possibility for such exchangehas been demoinstrated by in vitro studies oIn rabbitplasma by Havel et al. (33). If such an exchaingewere to take place in man, LDL-TG, because of itsgreater mass compared to HDL-TG, would be the morelikely candidate. However, certain considerations sug-gest that exchange betweeni labeled VLDL-TG andLDL-TG probably dloes not contribute significantly tothe tail of the VLDL-TG activity curve. For instance,Havel et al. (33) founid in rabbits that in vivo ex-change of TG between VLDL and LDL was onlyabout 20-25% of the trainsfer noted in vitro. Further-mnore, Malmendier and Berman (34) found that over80% of TG that is transferred to LDL during degrada-tion of VLDL is rapidly hydrolyzed with recovery oflabeled FFA in plasmiia. Thus

in normal subjects and in most of the hyperlipidemicpatients presented in this study.

First, in the study of Berman et al. (3) patientswith Type III hyperlipoproteinemia had a much largertail, or slow component, than did other types of pa-tients (normal and Types I, II, and IV). In the latterpatients, the slow component was usually 0.3

C.)

0.2

0.1

0 0.10 0.18 0.30 0.50 0.700,

0.83

_ 0.90

0.50

F40.30 F

z

0.20 tE

0.15 20

0

_ 0.10

z

_ 0.07 D

0.05UL

_ 0.031 0z

0

0.025

_ 0.013_ 0.008

0.005

0.90

CH/TG = 0.83a + 0.18(1 - a),

CH/TG - 0.18a=-

0.83 -0.18

anda M2T P21 T2T1 CH/TG - 0.18

1- = . (7)1 -a Mc I - P2, TcG 0.83 - CH/TG

VLDLCHOLESTEROL/VLDLTRIGLYCERIDE

FIGURE 7 A plot of Eq. 7; fraction (a) of VLDL-TG mass incompartment 21 as a function of VLDL CH/TG ratio. Thefraction of VLDL-TG transport diverted through compartment21 (P21) is also plotted as a function of CH/TG ratio. Theordinate value for P21 is derived from Eq. 7 assuming(TTG/TTG) = 10 (notice the nonlinear scale for P21). Thedotted line indicates the CH/TG ratio of 0.41 with thecorresponding value of a which is 0.35, and P21 which is 0.051.

1270 L. A. Zech, S. M. Grundy, D. Steinberg, and M. Berman

hence

/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I

/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~I

/ *0III. S.I.* 1.

S~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~S~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

- 0

0

. .

ri.

glycerol phosphate for TG synithesis retains the 3H.In contrast, [1,3-14C]glycerol could be tranisformed togluconeogenic precursors and still retain its lal)el. Ifthere were onlv a sinigle rapid synthesis pathway, apreferential loss of 3H in it would result in a pro-portional decrease of the entire VLDL-TG curve,which is contrary to the observed respoinses as shownin Fig. 6.

Two labels cani generate different VLDL-TG curvesbecause of differences in the catabolismn of thelabeled TG, or differences in the secretion of VLDL-TG into the plasmiia. In the first case there wouldusually be a difference in the FCR and hence in thecalculated total VLDL-TG transport rates. In thesecond case the FCR and, hence, total transportrate would have the same values for both tracers,which implies that a change in the area under theVLDL-TG curve is accompanied by a correspondinigproportional change in the total plasma glycerollabel that is incorporated into VLDL-TG. The joint14C and 3H studies were fitted under the assumnp-tion that once in plasma, labeled VLDL-TG iscatabolized the samiie way for both labels.

The "slow" and "fast" pathways are derived strictlyfor glycerol which originated in the plasma comiipart-ment. The partitioning of input from other precursorsof glycerol (UTG) cannot be assessed but has Ino in-fluence on calculating VLDL-TG FCRand synthesis.The existence of the slow pathway, however, hasdefinite implicatioins for the interpretation of thekinetic curves based on plasma glycerol and FFAprecursors in that it affects the proposed model forthe VLDL-TG subsystem and the calculated valuesof VLDL-TG FCRand synthesis rate.

When any part of the tail of the labeled VLDL-TGcurve can be attributed to a slowly metabolizingVLDL (VLDL21), there will be a corresponding de-crease in the calculated VLDL-TG FCRand synthesisrate. Hence, when P21 is zero the calculated valuesrepresent maximal estimates. A correction of FCRTG(1 = 1/TTG) when P21 is not zero can be made usingEq. 4; and an estimate for the value of P21 can be madeusing the CH/TG ratio as given by Eq. 7 or using Fig. 7.

It is interesting to note that the presence of a slowconversion pathway can also be inferred from theresults of Havel and Kane (26). They reported studieson four subjects to whom both labeled FFA andendogeneously labeled VLDL-TG (obtained from twodonors) were administered. The kinetics of labeledplasma VLDL-TG were followed for 24 h. In all theirstudies the tails of the VLDL-TG curves were con-siderably higher for the FFA label than for the VLDL-TG label. This could only be explained by a con-tribution generated by the FFA precursor: either aslow component in the plasma or in the conversionpathway to VLDL-TG. An argument against the formerhas been advanced by Shames et al. (1), on the basis

that s;uch a comiiponenit would re(quire a nonplasmaFFA source for TG.

Somne (lirect support for the stepwise delipidlationmodel with newvly synthesized VLDL-TG enteringthe first stage comiies from recenit sttudies of Strejaet al. (39) an(d Steinier aindl Streja (23). They separatedplasmaii VLDL-TG into three density fractions anddetermiiined the specific activities of each fraction as aftunetion of time after the addministrationi of radio-labeled glycerol. WVe anialyzedl both sets of data fortwo patients (BAI. and B.V.) using a three stagemoldel as imnplied by them, with naseent VLDL-TGfree to eniter into each of the fractions. The resultsof this analy,sis showedl that abouit two-thirds of thenewly svnthesized VLDL-TG entered the lightest frac-tion, with most of the remainder entering the middlefraction. Only about 5% entered the heaviest frac-tion. The turnover rate for each fraction was foundto be the siame. We also fitted the suum of their datafor the three fractions using our model and found thatthe FCR calculated was nearly the saime as the onecalculatedl using the three stage model (B.V. 0.18/h vs.0.20/h acndl BA.L 0.24/h vs. 0.22/h).

The physiological site or sites of the model VLDL-TGsynthesis pathways remain unknown; however, the cal-culated size of the slow compartment for our normals is750 mg, well within the values for liver TGconcentration.

It has been the practice in the past to determinethe FCR of VLDL-TG by using the slope of the fastdecaying portion of the activity curve, (Fig. 1, phase

10-6

1O-7E

O 10'

.° 10-10> 10

0.4

J10-:E 10-'10

E10-'1

b1l= 0.03 bill=bT =0.010w~ ±

FCRVLDL= 0.05

t, I I,I D-57_1b11 = 0.14

b,III -bT = 0.021

TG

FCRvLDL= 0.15I- .D-13,

6b 11 = 0.20

7 t

8FCRVLDL

=

-0 _

0 12 24 36 48Time (h)

FIGURE 8 A comparison of three studies in which specificactivity of VLDL-TG in the late part of the curve, relativeto that at the peak, was high, medium, and low. The slopesof phase 11 (bl,) and of phase III (bill) are shown for com-parison with the FCR values derived by analysis of theentire curve using the model of Fig. 2.

Very Low Density Lipoprotein Triglyceride Synthesis and Metabolism

rI

8

6

7

8

1271

II). Fig. 8 illustrates three activity curves with valuesfor the fast slopes (b11), the slow slopes (b111), andour model-determined FCRs. As can be seen, thefast decay slopes, bll, can differ significantly from thetrue FCR values. It is clear from our model develop-ment that when labeled glycerol is used as a pre-cursor to determine VLDL-TG turnover and synthesisrates it is necessary to take into consideration thedetails of the kinetic behavior of the synthesis path-ways as well as the metabolic pathways of plasmaVLDL-TG.

ACKNOWLEDGMENTS

The authors wish to express their appreciation to Dr. F. K.Millar for her assistance with this project and to Mrs. SueSmith and Miss Sandra Taylor for their support in thepreparation of this manuscript. They are also grateful toElliott Groszek, Peter McGough, Suzan Butler, RichardEarl, Warren May, and Avon Drummond for their excellenttechnical assistance.

This research was supported in part by the MedicalResearch Service of the Veterans Administration and byNational Institutes of Health research grants HL-14197,awarded by the National Heart, Lung, and Blood Instituteand AM-16667 from the National Institute of Arthritis,Metabolism, and Digestive Diseases, U. S. Public HealthService, Department of Health, Education, and Welfare.

REFERENCES

1. Shames, D. M., A. Frank, D. Steinberg, and M. Berman.1970. Transport of plasma free faitty acids and triglyceridesin man: A theoretical analysis. J. Clin. Invest. 49: 2298-2314.

2. Farquhar, J. W., R. C. Cross, R. Ml. Wagner, and G. M.Reaven. 1965. Validation of an inicompletely coupled two-compartment nonrecycling catenary model for turnover ofliver and plasma triglyceride in iman. J. Lipid Res. 6:119-134.

3. Berman, M., M. Hall, R. I. Levy, S. Eisenberg, D. W.Bilheimer, R. D. Phair, andl R. H. Goebel. 1978. Metab-olism of apoB and apoC lipoproteins in man: Kineticstudies in normal and hyperlipoproteinemic subjects.

J. Lipid Res. 19: 38-56.4. Phair, R. D., M. G. Haimmond, J. A. Bowdes, M. Fried,

M. Berman, and W. R. Fisher. 1972. Kinetic studies inhuman lipoprotein metabolism in type IV hyperlipo-proteinemia. Fed. Proc. 31: 421. (Abstr.)

5. Nikkila, E. A., and M. Kekki, 1971. Polymerphism ofplasma triglyceride. Kinetics in normal human adultsubject. Acta. Med. Scand. 190: 49-59.

6. Reaven, G. M., D. B. Hill, R. C. Gross, and J. W.Farquhar. 1965. Kinetics of triglyceride turnover of verylow density lipoproteins of human plasma. J. Clin.Invest. 44: 1826-1833.

7. Eaton, R. P., M. Berman, and 1. Steinberg. 1969. Kineticstudies of plasma free fatty acid and triglyceridemetabolism in man. J. Clin. Invest. 48: 1560- 1579.

8. Quarfordt, S. H., A. Frank, D. M. Shames, M. Berman,and D. Steinberg. 1970. Very low density lipoproteintriglyceride transport in type IV hyperlipoproteinemiaand the effects of carbohydrate-rich diets. J. Clin.Invest. 49: 2281-2297.

9. Phair, R. D., M. G. Hammond, J. A. Bowdes, M. Fried,W. Fisher, and M. Berman. 1975. A preliminary modelfor human lipoprotein metabolism in hyperlipopro-teinemia. Fed. Proc. 34: 2263-2270.

10. Phair, R. D., M. Hall, D. W. Bilheimer, R. I. Levy, R. H.Goebel, and M. Berman. 1976. Modeling lipoproteinmetabolism in man. In Proceedings of the 1976 SummerComputer Simulation Conference, Washington, D. C.Simulation Councils, Inc., La Jolla, Calif. 486-492.

11. Reardon, M. F., N. H. Fidge, and P. J. Nestel. 1978.Catabolism of very low density lipoprotein B apoproteinin man.J. Clin. Invest. 61: 850-860.

12. Grundy, S. M., H. Y. I. Mok, L. Zech. D. Steinberg,and M. Berman. 1979. Transport of very low densitylipoprotein-triglycerides in varying degrees of obesityand hypertriglyceridemia.J. Clin. Invest. 63: 1274-1283.

13. Grundy, S. M., P. J. Nestel, R. Monell, H. Moh, K. vonBergmann, and D. Steinberg. 1975. Kinetics of very lowdensity lipoprotein-triglycerides (VLDL-TG) followingradioglycerol. Circulation. 52(Suppl. II): 39.

14. Havel, R. J., H. A. Eder, and J. H. Bragdon. 1955.The distribution and chemical composition of ultra-centrifugally separated lipoproteins in human serum.

J. Clin. Invest. 34: 1345-1353.15. Reichl, D., N. B. Myant, and J. J. Pflug. 1977. Con-

centration of lipoproteins containing apolipoprotein Bin human peripheral lymph. Biochim. Biophys. Acta.489: 98- 105.

16. Berman, M., and M. Weiss. 1977. SAAMManual. U. S.Govemment Printing Office, Department of Health, Edu-cation, and Welfare Publication No. 78-180. NationalInstitutes of Health, Washington, D. C. 200 pp.

17. Pettigrew, K. D. 1964. Estimation of a parameter fromobservations with unequal precisions in the presenceof neusance parameters. Master's Thesis, George Wash-ington University, Washington, D. C. 74 pp.

18. Malmendier, C. L., C. Delcroix, and M. Berman. 1974.Interrelations in the oxidative metabolism of free fattyacids, glucose, and glycerol in normal and hyper-lipemic patients. J. Clin. Invest. 54: 461-476.

19. Shore, V. G., and B. Shore. 1973. Heterogeneity ofhuman plasma very low density lipoproteins. Separationof species differing in protein components. Biochemis-try. 12: 502-507.

20. Barter, P. J., and P. J. Nestel. 1972. Precursor-product relationship between pools of very low densitylipoprotein. Biochim. Biophys. Acta. 260: 212-221.

21. Higgins, J. M., an(d C. J. Fielding. 1975. Lipoproteinlipase. Mech-anism of formation of triglyceride-rich rem-nant particles for VLDL and chylomicrons. Biochemistry.14: 2288-2292.

22. Hazzard, W. R., and E. L. Bierman. 1976. Delayedclearance of chylomicron remnants following vitamin-A-containing oral fat loads in broad-l8 disease (type IIIhyperlipoproteinemiiia). Metab. Clin. Exp. 25: 77-80.

23. Steiner, G., ancl D. Streja. 1977. Kinetics of VLDLsubfractions. In Atherosclerosis IV. G. Schettler, Y. Goto,Y. Heta, and G. Klose, editors. Springer-Verlag, Berlin,East Germany. 129-132.

24. Berman, M. 1979. Analysis of kinetic data. In Progressin Biochemical Pharmaeology: Lipoprotein Metabolism.S. Eisenberg, editor. S. Karger AG. Basel, Switzerland.

25. Nikkila, E. A., and M. Kekki. 1971. Measurement ofplasma triglyceride turnover in the study of hyper-triglyceridemia. Scand. J. Clin. Lab. Invest. 27: 97-104.

26. Havel, R. J., and J. P. Kane. 1975. Quantification oftriglyceride transport in bloo1( plasmla: a critical analysis.Fed. Proc. 34: 2250-2257.

1272 L. A. Zech, S. M. Grundy, D. Steinberg, and M. Berman

27. Mason, S. J., and H. J. Zimmerman. 1960. Electroniccircuits, signals, systems. John Wiley and Sons Inc.,New York. 580.

28. Brownell, G. B., M. Bermani, and J. S. Robertson.1968. Nomenclature for tracer kinetics. Int. J. Appl.Radiat. Isot. 19: 249-262.

29. Gurr, M. I., and A. T. James. 1975. Lipid Biochemistry.Chapman and Hall Ltd. London, England. 90.

30. Eaton, P. R., S. Crespin, and D. M. Kipnis. 1976.Incorporation of 75Se-Selenomethionine into human apo-proteins in man. Diabetes. 25: 679-690.

31. Reichl, D., N. B. Myant, J. J. Pflug, and D. N. Rudra.1977. The passage of apoproteins from plasma lipo-proteins into the lipoproteins of peripheral lymph inman. Clin. Sci. Mol. Med. 53: 221-226.

32. Reichl, D., L. A. Simons, N. B. Myant, J. J. Pflug, andG. L. Mills. 1973. The lipid and lipoproteins ofhuman peripheral lymph, with observations on thetransport of cholesterol from plasma and tissues intolymph. Clin. Sci. Mol. Med. 45: 313-329.

33. Havel, R. J., J. M. Felts, and C. M. Van Duyne.1962. Formation and fate of enidogenous triglycerides inblood plasma of rabbits. J. Lipid Res. 3: 297-308.

34. Malmendier, C., and M. Berman. 1978. Endogenouslylabeled low density lipoprotein triglyeeridle and apo-protein B kinetics. J. Lipid Res. 19: 978-984.

35. Quarfordt, S. H., R. I. Levy, and D. S. Fredrickson.1973. The kinetic properties of very low densitylipoprotein triglyceride in type III hyperlipoproteinemia.Biochim. Biophys. Acta. 296: 572-576.

36. Kushwaha, R. S., W. R. Hazzard, C. Gagne, A. Chait, andJ. J. Albers. 1977. Type III hyperlipoproteinemia:paradoxical hyplipemic response to estrogen. Ann. In-tern. Med. 87: 517-525.

37. Quarfordt, S., R. I. Levy, anid D. S. Fredrickson.1971. On the lipoprotein abnormality in type IIIhyperlipoproteineinia.J. Clin. Invest. 50: 754-761.

38. Pagnan, A., R. J. Havel, J. P. Kane, and L. Kotite.1977. Characterization of human very low density lipo-proteins containinig two electrophoretic populations:double pre-beta lipoproteinemia and primary disbeta-lipoproteinemia.J. Lipid Res. 18: 613-622.

39. Streja, D., M1. A. Kallai, and G. Stiner. 1977. Themetabolism heterogeneity of human very low densitylipoprotein triglyceride metabolism clinical and experi-mental. Metab. Clin. Exp. 26: 1333-1344.

Very Lowv Detnsity Lipoprotein Triglyceride Synthesis and Metabolism 1273