Soluble complex formation between low-density lipoprotein and glycosaminoglycans. A 2H and 31P-NMR, and quasi-elastic light scattering study

Chemistry and Physics o/Lipids, 49 (1988) 15--29 Elsevier Scientific Publishers Ireland Ltd.

15

Soluble complex formation between low-density lipoprotein and glycosaminoglycans. A 2H and 31p-NMR, and quasi-elastic light scattering study D a v i d B . F e n s k e a n d R o b e r t J. C u s h l e y Department of Chemistry, Simon Fraser University, Burnaby, B.C., V5A 1S6 (Canada) (Received April 14th, 1988; revised and accepted June 21st, 1988) Soluble complex formation between LDL and heparin (HEP) and chondroitin sulfate (CS) has been studied by ~H- and 3tp-NMR and light scattering. The 2H-NMR linewidths of [2I-I]HEPand [2H]C4S increase substantially upon binding to LDL, with the [2H]HEP linewidtbs broader at low glycosaminoglycan (GAG)flow density lipoprotein (LDL) ratios. Preliminary analysis of the bound C2H3 group correlation times suggests that the observed linewidtbs are determined by the complex size, and that both [2H]GAG-s have similar motions when bound to LDL. The 31p-NMR data demonstrate that large LDL-HEP complexes (diameter approx. 50 nm) are formed only over a narrow range of HEP concentrations, whereas the size of LDL-CS complexes increases continuously over the range of CS concentrations studied, reaching values of 32-35 nm for both C4S and C6S. At the lower protein concentrations studied by light scattering (~<1 mg/ml), the same trends are observed, although the mean diameters are less than those estimated by aIp-NMR. Soluble complex formation was unaliected by the presence of 2 mM Ca 2+. Dilution studies demonstrate that complex size varies with protein concentration. The binding of GAGs to LDL was also examined by HEP-CS competition studies. HEP has the higher affinity while no differences in binding could he detected between C4S and C6S.

Keywords: deuterium NMR; 31p-NMR; low density lipoprotein-heparin soluble complexes; low density lipoprotein-chondroitin sulfate soluble complexes; low density lipoprotein-glycosaminoglycan interaction; light scattering.

Introduction Human plasma LDL is involved in the transport of cholesterol, predominantly in the form of cholesteryl ester, to the peripheral tissues [1]. LDL is also the source of the majority of the cholesterol found in atherosclerotic lesions [1]. Thus, understanding the pathogenesis of atherosclerosis requires elucidation of the factors which contribute to deposition of LDL in the arterial matrix. The primary route by which LDL is removed from the plasma by the smooth muscle cells of the inner arterial wall is known as the LDL-receptor pathway [1,2]. However, in some

Correspondence to: Prof. R.J. Cushley.

human populations the mean levels of LDL in the plasma are too high for effective clearance to occur. In such cases, LDL can come into contact with the GAGs and proteoglycans (PGs) of the subendothelial matrix. LDL is known to form both soluble and insoluble complexes (the latter r e q u i r i n g C a 2÷ in concentrations exceeding 5mM) with GAGs and PGS in vitro [3--15,52,55]. Such an interaction in vivo has been implicated in the deposition of LDL cholesterol in the arterial wall and in the path0genesis of atherosclerosis [11,12,16---23]. In addition, it has been observed that insoluble complex formation leads to alterations in the structure of the LDL core [5,24--26], a factor which may enhance deposition in the arterial matrix. Much work has been directed towards identifying the G A G

0009-3084/88/$03.50 © 1988 Elsevier Scientific Publishers Ireland Ltd. Published and Printed in Ireland.

16 moieties of PGs which are involved in the extracellular binding of LDL. Of all GAGs, CS, particularly the C6S isomer, seems to be most closely associated with changes in the arterial wall and with its ability to sequester lipoproteins [12,17,18,22,23]. HEP, which has the highest affinity for LDL [9], is not usually present in the extracellular matrix, but is found in the form of a HEP-PG in mast cells within their cytoplasmic granules [27]. However, it has been demonstrated that LDL can bind to the negatively-charged H E P - P G of rat serosal mast cell granules [ 15], and HEP has been detected in L D L - G A G complexes isolated from human aortic fibrous plaques [ 17,18]. Thus, both CS and HEP may play a role in the sequestration of LDL within the extracellular space. The physiological significance of soluble complexes in the plasma has not received as much attention as has the role of insoluble complexes in the arterial matrix. At concentrations much higher than those found physiologically, the presence of heparin has been found to reduce uptake of LDL by cultured fibroblasts [28], and the uptake of LDL by aortic tissue in vivo can be inhibited by the presence of G A G [29,30]. Also, GAGs have been found to act as effective antiatherosclerotic agents [31--33], presumably via saturation of LDL binding sites. CS is the predominant G A G found in human plasma, in which it exists in both low-charge and high-charge forms in non-covalent association with plasma proteins [34]. Nakashima et al. [35] have observed that the high-charge plasma GAGs can modify LDL conformation, an effect that can be prevented by the low-charge species. This has led to the suggestion [35] that the rheological properties of lipoproteins may be maintained by interactions with plasma GAGs. These observations suggest the possibility of a role for soluble complex formation in the many factors which regulate the deposition of LDL cholesterol in the arterial wall. The physicochemical properties of LDL-HEP soluble complexes have been investigated by many techniques including quasi-elastic light scattering (QELS), theology, gel filtration, analytical ultracentrifugation, and steady-state fluorescence polarization [6,36,37,52]. However,

little is known about the nature of complexes of LDL with CS. In this paper, we report the first application of 2H and 31p-NMR spectroscopy to the study of L D L - G A G soluble complexes, and of QELS to complexes of LDL with CS. The purpose was to investigate the different properties of LDLHEP and LDL-CS soluble complexes, and to examine any differences in the binding behaviour of C6S and C4S. HEP and CS provide interesting contrasts in light of their different charge densities and molecular weights. Materials and methods

Heparin (Porcine Intestinal Mucosa, Grade 1, sodium salt), chondroitin sulfate Type A (C4S) (whale cartilage, sodium salt), chondroitin sulfate Type C (C6S) (shark cartilage, sodium salt), perdeuterated acetic anhydride ([2H6]acetic anhydride), and deuterium-depleted water were purchased from Sigma. The GAGs were used as supplied without fractionation. LDL was isolated from fresh (<3 days old) plasma by ultracentrifugal flotation in the density range 1.025--1.063 g/ml [38]. The LDL from several units was pooled and purified by reisolation in the same density range. The LDL was stored at 4°C under N2. Acetylation of the lysine residues of LDL was accomplished as described [9]. Protein was determined by the method of Lowry et al. [39].

Deuteroacetylation of GAGs Chondroitin 4-sulfate and heparin were partially N-deacetylated essentially as described by Hook et al. [40]. The preparation of [2H]C4S has been described [41]. [2H]HEP was prepared as follows. A homebuilt bomb containing 110 mg of HEP, 5.5 ml of hydrazine, and 55 mg of hydrazine sulfate was heated at 100°C for 1.5 h, after which it was cooled in an ice bath for 2 h. The flask containing the HEP was removed and placed in a water bath at approx. =40°C. Five millilitres of toluene was added, and the mixture was evaporated to dryness with stirring under a stream of N2. The toluene additions were repeated twice more. Approximately 2 ml of absolute ethanol was

17 then added to the deacetylated HEP, which was evaporated as above. The HEP was then dissolved in 5.5 ml of 0.2 M NH4HCO3 containing 0.02% NaN3 and re-isolated on a column (2.5 x 92 cm) of Sephadex G-50 eluted with the same buffer. Fractions containing HEP (assayed using Alcian Blue 8-GX (42)) were pooled and lyophilized. The HEP was dissolved in 20 ml of water and was lyophilized again. N-Deacetylated HEP (approx. 100mg) was N-deuteroacetylated essentially as described [40], with the exception that the addition of [2H6]acetic anhydride (1.1 ml) was performed at room temperature. The HEP was dissolved in 4 ml of 10% methanol containing 0.05 M Na2CO3, and the [2H6]acetic anhydride was added in I I 0.1-ml aliquots over a period of 3.5 h. The pH was maintained between 7 and 8 by the addition of 10% methanol saturated with NaECO3. The reaction was stopped by the addition of 20 ml of water, and the [2H]HEP was applied to a column (2.5 × 92cm) of Sephadex G-50 and isolated as described above.

LDL- GAG binding studies For the 2H-NMR experiments, LDL was first exhaustively dialysed against 0.05 M Tris--HCl (pH 7.5) containing 0.3 mM NaEEDTA, following which it was concentrated to approx. 23 mg protein/ml for NMR experiments using Millipore Immersible-CX Ultrafilters (CX-30). This and all subsequent dialyses and solvent exchanges were performed at 4°C. The sample volume was 1.5 ml. The residual 2HOH signal was reduced by several exchanges (usually six) with 0.05 M Tris--HC! (pH 7.5) in deuterium-depleted water (resulting in a dilution of the original solvent of 800-fold). The [2H]HEP and [2H]C4S were lyophilized at least once in deuterium-depleted water prior to addition as solids to LDL. For the JIP-NMR experiments in which the effect of increasing G A G was studied, LDL was concentrated to 14--20 mg protein/ml as above. The sample volume was 1.5 ml. Several exchanges (usually five) were performed with the desired buffer, either 0.05 M Tris--HC! 0.02% NaEEDTA (pH 7.5) prepared in 50% 2H20 or

0.05 M HEPES 0.02% Na2EDTA (pH 7.5) prepared in 50% 2H20. The GAGs were added as solids to the LDL and gently mixed just before NMR experiments. For the LDL-HEP concentration studies, the LDL was exhaustively dialysed against 0.05 M Tris--HCl 0.02% Na2EDTA (pH 7.5) prepared in 50% 2H20, and the HEP was added as a solution (2 mg/ml) in the same buffer. For the QELS experiments, LDL (4.9 mg protein/ml) was dialysed exhaustively against 0.05 M Tris--HCl 0.02% Na2EDTA (pH 7.5) or 0.05 M HEPES 0.02% Na2EDTA (pH 7.5). Stock solutions of HEP and CS (0.25 mg/ml) were prepared in the appropriate buffer. Aliquots of GAG were added to the LDL, and the solution was then diluted to 0.6--0.8 mg protein/ml. The complexes were filtered directly into the sample cell (Gelman Acrodisc polysulfone filters; 0.2/~m). In studies dealing with the effect of physiological concentrations of calcium, the buffer was 0.05M Tris--HCl 2 mM CaCI2 (pH 7.5), and the protein concentration of the complex was 1 mg/ml. The 2H-NMR experiments were performed at 38.8 MHz with a homebuilt spectrometer and a 5.9 Tesla Naiorac superconducting magnet interfaced to a Nicolet BNC-12 computer. Temperatures were controlled at 25 °C by a solid-state temperature controller with an accuracy of ±0.25°C built by the Simon Fraser University electronics shop, using a homebuilt variabletemperature probe. Spectra were obtained using a one-pulse sequence with phase alternation in order to minimize baseline distortion. Spectral parameters: pulse width= 14/xs (flip angle= 60°); delay between pulses = 0.1 s; sweep width = 5 kHz; number of data points = 1024 zero-filled to 4096. Further parameters are given in the figure legends. The 31p-NMR experiments were run at 40.SMHz on a Bruker SY-100 spectrometer using inverse-gated proton noise decoupling (decoupler gated on during FID acquisition). Field frequency stabilization was obtained from the 2HOH resonance of the solvent. Temperatures were controlled at 25 °C using a Bruker temperature controller. Spectral parameters: pulse width = 20 p.s (flip angle = 60°); delay between pulses = 2 s; data set = 2048 zero-filled to

18

a)

a)

2

z

~

HO H

./~

÷

,e

200 Hz

HOH ---~

200 Hz

÷

b)

b)

+ ,~+~+

÷

+ ++÷+~

c) c)

÷

+

÷

÷~-+

÷~-

-~

~+

+

++

÷+

~

+ ++

+÷+÷

+ +

÷

d)

d)

+ ~*÷+

÷+

Fig. 1. 2H-NMR spectra of [2H]HEP in the presence of LDL (22 mg protein/ml) in 0.05 M Tris---HCl (pH 7.5) at 25°C. CHE p (mg/ml)=0.4 (a), 1.1 (b), 2.5 (c), 5.4 (d). Spectral parameters: number of acquisitions--490,000 (a), 600,000 (b), 420,000 (c), 420,000 (d). Line broadening = 25 Hz (a), 20 Hz (b), 8 Hz (c), 5 Hz (d). , the best fit Lorentzian function to the 2HOH (low field (left)) and [2H]HEP resonances.

Fig. 2. 2H-NMR spectra of [2H]C4S in the presence of LDL (24 mg protein/ml) in 0.05 M Tris--HCl (pH 7.5) at 25"C. Ccs (mg/ml) = 0.3 (a), 0.9 (b), 2.1 (c), 4.8 (d). Spectral parameters: number of acquisitions = 200,000 (a), 300,000 (b), 116,000 (c), 53,000 (d). Line broadening = 20 Hz (a,b), 10 Hz (c), 5 Hz (d). ~ , the best fit Lorentzian function to the 2HOH (low field (left)) and [2I-I]C4S resonances.

19 4096. Further parameters are given in the figure legends. The 2H-NMR spectra in Figs. 1 and 2 were analysed using a seven-parameter iterative leastsquares fit of the 2HOH and [2H]GAG resonances to Lorentzian functions. The program, provided by Dr. K.E. Newman and modified by Dr. W.D. Treleaven, both of the Chemistry Department, Simon Fraser University, was run using an IBM 3081 GX computer in the Computing Services Department of Simon Fraser University. The 3Xp_ NMR spectra in Figs. 4 and 5 were simulated by computer using a two Lorentzian lineshape function chemically shifted by 0.6 ppm (25.4 Hz), corresponding to the sphingomyelin and PC resonances of LDL. The linewidths of the two phospholipids were assumed to be equal. All spectra are plotted with low field to the left. QELS measurements were performed using a Nicomp Model 270 Submicron Particle Sizer. The scattering angle was 90 °. The autocorrelation function was evaluated over 64 channels. The

channel width varied from 2.8 to 5.0/~s. The autocorrelation function can be evaluated by either a gaussian analysis or a bimodal distribution analysis. The gaussian analysis performs a leastsquares fit of the autocorrelation function to a gaussian distribution, using a x2-test to assess the goodness of fit. The distribution analysis makes no assumptions regarding the nature of the particle size distribution, but rather attempts to determine the set of exponentially decaying functions which will yield the measured autocorrelation function when added and squared. The sizes cited in the text, unless otherwise stated, were obtained from the Distribution analysis and are expressed as mean diameter+standard deviation, obtained from repeated runs (usually 7--20) on the same sample. For L D L - G A G complexes, a gaussian analysis of the data yielded X2 values in the 0.5--20 range, with the majority of values in the range 5m10. The assumption of a gaussian distribution of complex sizes does not, therefore, seem to be valid. In general, the sizes obtained

250

I 200 -

Ii "W

150 -

<1 100

-

~ . ~

50O .......

O ...................

"0

0

o

I'o ?s CGAG/CLD L x 10'

2'o

z5

2H-NMRlinewidthsArt/2of [2H]HEPin the presenceof LDL (Q) and acetylated-LDL(O), and of [2H]C4Sin the presence of LDL (i) at 25°C. CLDL--~23 mg protein/mlfor the native LDL studiesand 13 mg protein/mlfor the a c e t y l a t e d LDL study.The error bars represent an estimateduncertaintyin the linewidthof + 10%. Fig. 3.

20

a)

a) 50 Hz 50 Hz

b) 200 Hz

100 Hz

dl

Fig. 4. 3~p-NMR spectra of L D L (15.6rag protein/ml) in 0.05 M Tris--HCl 0.02% Na2EDTA (pH 7.5) in the presence of increasing concentrations of HEP at 25 °C. cxEp (mg/ml) = 0 (a), 0.35 (b), 3.0 (c), 5.6 (d). Spectral parameters: number of acquisitions = 2000 (a), 2007 (b), 2000 (c), 2002 (d). Sweep width = 2000 ~ (a,c,d), 3000 ttz (b). Line broadening = 2 Hz (a,c,d), 5 Hz (b). The low-field sphingomyin resonance appears on the left.

c)

Fig. 5. 3=P-NMR spectra of LDL (13.7mg protein/ml) in 0.05 M HEPES 0.02% Na2EDTA (pH 7.5) in the presence of increasing concentrations of C6S at 25°C. Ccs (mg/ml)= 0 (a), 0.32 (b), 7.6 (c). Spectral parameters: number of acquisitions = 1408 (a), 1663 (b), 2765 (c). Sweep width = 2000 Hz (a---c). Line broadening = 2 Hz (a---c).

21 from the distribution analysis are 2--7 nm larger than the gaussian sizes. However, it should be stressed that the same trends are observed with both methods of analysis, and the conclusions of our study are the same regardless of which analysis is used. Results

2H-NMR studies of L D L - G A G soluble complex formation Deuteroacetylated GAGs were chosen as a convenient means of monitoring the L D L - G A G interaction from the point of view of the polysaccharide component. The method used to obtain [2H]GAGs has been shown to result in minimal perturbation of the G A G structure [40]. Because our procedure was modified slightly from Ref. 40 (see Experimental), the effect of chemical modification on G A G structure and binding properties was also examined. Deacetylated CS was examined for degradation on a Sephadex G-200 column eluted with 1 M NaCI in 0.05M Tris--HCl (pH 8.0) containing 0.02% NaN3 (figure not shown). In addition, the binding properties of [2H]HEP and [2H]C4S were compared with the unmodified GAGs by QELS and 31p_ NMR, respectively. The results indicate that hydrazinolysis does not degrade the polysaccharides, and that deuteroacetylation does not significantly alter the binding behaviour of GAGs with LDL. 2H-NMR spectra of increasing concentrations of [2H]HEP in the presence of LDL were obtained at 25°C and are shown in Fig. 1. The solid line represents an iterative least-squares fit of the 2HOH and [2H]HEP resonances to single Lorentzian lineshape functions. The corresponding spectra of [2H]C4S are shown in Fig. 2. A plot of observed linewidth as a function of G A G / L D L ratio for [2H]HEP and [2H]C4S is shown in Fig. 3. The binding of both [2H]HEP and [2H]C4S to LDL is clearly evidenced by the large change in linewidth at lower G A G / L D L ratios. At a G A G / L D L ratio of 0.025, the linewidths of [2H]HEP and [2H]C4S are increased approximately 15- and 5-fold, respectively, over

their unbound values (13 and 32 Hz for [2H]HEP and [2H]C4S at approx. 10 mg/ml, respectively). Below a G A G / L D L ratio of approx. 0.1, the linewidths of the LDL-[2H]HEP complexes are broader than those of the LDL-[2H]C4S complexes. Above a ratio of 0.1, the [2H]HEP linewidths are slightly narrower than those of the [2H]C4S. In order to verify that the 2H-NMR data is monitoring soluble complex formation, [2H]HEP was added to acetylated LDL. It has been observed by others [9,43] that acetylation of the lysine groups of apo B abolishes G A G binding. As shown in Fig. 3, the linewidths of [2H]HEP in the presence of acetylated LDL are considerably narrower than in the presence of native LDL, and remain constant over the range studied, suggesting that complex formation has been greatly reduced.

3JP-NMR studies of L D L - G A G soluble complex formation In order to examine the behaviour of the lipoprotein component of L D L - G A G complexes, 31p_ NMR was the method of choice. Figure 4 shows representative 31P-NMR spectra of LDL in 0.05 M Tris 0.02% Na2EDTA (pH 7.5) in 50% 2H20 in the presence of increasing concentrations of HEP. The 3tp-NMR spectrum of native LDL (Fig. 4(a)) is characterized by two resonances, the less-intense sphingomyelin peak (downfield) and the dominant phosphatidylcholine (PC) peak (upfield). The PC/sphingomyelin ratio of 1.63 obtained from the 31p signal intensities compares well with literature values [44]. As the HEP concentration is increased, the linewidths broaden until resolution of the two resonances is lost (Fig. 4(b)). With increasing HEP, the linewidths narrow and once more are resolved (Fig. 4(c) and (d)). Similar results were obtained in HEPES buffer (not shown). The maximum linewidth of the LDLHEP complexes occurs at a G A G / L D L ratio of 0.024.03. Figure 5 shows representative spectra of LDL in 0.05 M HEPES 0.02% Na2EDTA (pH 7.5) in 50% 2H20 in the presence of increasing concentrations of C6S. The linewidths are seen to broaden slightly but resolution of the PC and

22 70

50-

40-

30I

on ~ ~ . - - T - - - _ - |- 7 - J ~rd,hr - ~

o

~

I

1'0

-'---

I

_ + ............+

l

........ '

2o

3o

go

50

60

70

C G A G / C L D L x I0 a Fig. 6. 31p-NMR linewidth Avl/2 of the PC and sphingomyelin resonances of LDL as a function of GAG/LDL ratio at 25°C. HEP (Tris) (C)),C4S (Tris) (El), C6S (HEPES) (O), [2H]C4S (HEPES) (I). The linewidths were obtained from simulated spectra formed by the addition of two Lorentz/an iineshap¢ functions, representing the PC and sphingomyelin resonances of LDL. The chemical shift of the two resonances was kept constant at 25.4 Hz (0.6 ppm), and the linewidths were assumed to be equal. The error bars represent an estimated uncertainty in the linewidth of + 10%. sphingomyelin resonances is n e v e r lost. Thus, the complexes f o r m e d with CS are strikingly different than those formed with H E P . In order to obtain values for the 31p linewidths of the PC and sphingomyelin resonances of L D L , the spectra were simulated by c o m p u t e r using two Lorentzian lineshape functions chemically shifted by 25.4 H z (0.6 ppm). T h e chemical shifts were o b s e r v e d to remain constant in all cases. T h e results are shown in Fig. 6, where the linewidth AVl/2 is plotted as a function of G A G / L D L ratio. T h e linewidths of L D L - H E P complexes are seen to increase until a m a x i m u m value of approx. 62 Hz is reached at H E P / L D L 0.02. A b o v e this ratio the linewidths decrease. In contrast, the linewidths of L D L increase only slightly in the presence of CS o v e r the concentration range studied. T h e 3~p linewidths of all the L D L - C S complexes are the same within experimental error, and thus no differences in binding can be detected between C4S and C6S. Similar curves are

obtained in both H E P E S and Tris buffer for both H E P and CS.

QELS studies of LDL-GAG soluble complex formation Q E L S was used to obtain information on the sizes of the different L D L - G A G complexes formed with C4S and C6S. H E P was also examined to allow comparison with [2H]HEP, the 31p results, and with previous studies [36]. T h e m e a n diameter of native L D L , determined from several samples, was 23.4 ± 0.4 nm. Figure 7 shows the changes in complex m e a n diameter as a function of the G A G / L D L ratio of H E P , [2H]HEP, C6S, and C4S in 0.05 M Tris---HCl 0.02% N a 2 E D T A (pH 7.5). Increasing concentrations of H E P result in an initial increase in the m e a n diameter of the soluble complex up to a value of 3 5 - - - 4 0 n m at a G A G / L D L ratio of 0.02---0.03 mg H E P / m g protein. As the H E P concentration is increased, the

23

5o

40-!~ 35-

N

•

l

_

11

-$ I

25

2O

o

lb

z.0

3'0

CGAG/CLDL x 108

4'0

5'o

60

Fig. 7. Mean diameter of LDL-GAG soluble complexes as a function of GAG/LDL ratio at 25°C as determined by QELS. HEP (Q), [2H]HEP (O), C6S ([~, C4S (ll). CLDL(mg protein/ml) = 0.59 (HEP and [2H]HEP), 0.59--0.66 (C6S and C4S). The error bars are standard deviations obtained from repeated runs on the same sample, and are smaller than the symbols where absent.

mean diameter decreases rapidly until a ratio of 0.15 is reached. Further additions of H E P result in small decreases in complex diameter, until, in the presence of excess H E P ( G A G / L D L = 0.57), the complex diameter is reduced to 2 7 . 9 ± 1.3 nm. Behaviour essentially identical to that of H E P was observed with [2H]HEP (Fig. 7), demonstrating that the deacetylation-acetylation procedures did not significantly alter the H E P binding properties. Complexes of both H E P and [2H]HEP reached their maximum diameter at a H E P / L D L ratio of 0.03, and the sizes were the same within experimental error. Furthermore, in the concentration studies in Table II, both H E P and [2H]HEP yielded complexes of the same size at L D L concentrations of 0.75 and 0.20 mg protein/ml. T h e breakdown of large L D L - H E P complexes is complete at a G A G / L D L ratio of =0.15 in the Q E L S studies, but not until a ratio of ~ 0 . 2 in the 31P_NMR studies. These differences may be due to concentration effects (see below).

In contrast to the H E P results, the addition of increasing concentrations of CS to L D L results in a gradual increase in mean diameter, giving values of 32.6 ± 0.4 and 29.3 + 0.3 nm for C6S and C4S at a ratio of 0.57, respectively. T h e mean diameters of the L D L - C 6 S complexes are slightly larger than the L D L - C 4 S complexes o v e r the entire range studied.

QELS studies of LDL-GAG soluble complex formation: effect of calcium L D L - G A G soluble complex formation was studied by Q E L S in 0.05 M T r i s - - H C ! 2 mM CaCI2 (pH 7.5) at 25°C (figure not shown). T h e binding of H E P , C6S, and C4S with L D L was unaffected by the presence of 2 mM Ca2+; at all G A G / L D L ratios studied, the mean diameters of L D L - G A G complexes were within experimental error of the values observed in the absence of Ca 2+.

24 TABLE I

TABLE II

Competition of HEP and CS for LDL binding sites as determined by QELS a in 0.05M Tris--HCl 0.02% Na2EDTA (pH 7.5).

3~p-NMR and QELS concentration studies of LDL-HEP and LDL-[2H]HEP soluble complexes in 0.05 M Tris--HC1 0.02% Na2EDTA (pH 7.5).

HEP/LDL b

0.025 0.025 0.025 0

CS/LDL b

0 0.57 0.57 0.57

Mean diameter ± S.D. (nm) C4S

C6S

37.3±2.3 39.6±3.4 c

37.3±2.3 42.4±2.1 c 40.1±3.4 d 32.6±0.4

29.3±0.3

HEP/LDL ~

{).1}20 0.021 0.022

aLDL concentration (CLDL) = 0.59 mg protein/ml. bmg GAG/mg protein. eHEP added first. dCS added first.

QELS studies of H E P - C S competition for L D L binding sites In order to assess the relative affinities of H E P and CS for L D L , and to p r o b e for differences in the binding of C4S and C6S, H E P - C S c o m petition experiments were p e r f o r m e d (Table I). T h e m e a n d i a m e t e r of L D L - H E P complexes at a H E P / L D L ratio of 0.025 was 3 7 . 3 ± 2 . 3 rim. When C6S was added ( C 6 S / L D L = 0.57), the m e a n d i a m e t e r increased slightly to 4 2 . 4 + 2.1 nm. Addition of the same a m o u n t of C4S to the L D L - H E P complexes yielded a d i a m e t e r of 39.6 ± 3.4 nm. With no H E P present, the m e a n d i a m e t e r of C 6 S - L D L and C 4 S - L D L complexes at a ratio of 0.57 was 32.6 ± 0.4 and 29.3 ± 0.3 nm, respectively. W h e n H E P was added to L D L - C 6 S complexes in the same proportions as a b o v e ( C S / H E P = 23), the c o m p l e x m e a n diameter was 40.1 ± 3.4, the same within experimental error as when the C6S was added to L D L - H E P complexes.

Effect of protein concentration on the mean diameters of LDL- HEP complexes T h e results of varying the L D L concentration at a fixed H E P / L D L ratio are given in T a b l e II. Most of these studies were p e r f o r m e d at a H E P / L D L ratio (approx. 0.02) which g a v e maxi m u m linewidth values for 31p-NMR and corn-

0.032

Concentration

Mean

A~I/2

of LDL (rag protein/ml)

diameter ±S.D. (nm)

(Hz)

0.68 0.20 0.79 0.20 15.6 4.4 2.2 0.59 0.16

33.4 + 0.8 30.4±0.8 (34.2 ± 0.5 h) (31.9 ± 1.0 b)

39.3 ± 4.3 40.1 ± 3.8 36.8±2.4

62 c 45 d 35 ~

"mg HEP/mg protein. bThe numbers in parentheses represent studies involving [~H]HEP. ~Number of acquisitions -- 2002. dNumber of acquisitions = 19,740. "Number of acquisitions = 18,459.

plexes in the 35 nm size range by QELS. Both native H E P and [2H]HEP were used for the light scattering investigation. From Q E L S , at a H E P / L D L ratio of 0.02, the m e a n diameter of L D L - H E P complexes was found to decrease f r o m 33.4 + 0.8 to 30.4 + 0.8 nm as the protein c o n c e n tration was decreased from 0.68 to 0.20 mg protein/ml. In like manner, the m e a n diameter of L D L - [ E H ] H E P complexes was observed to decrease from 34.2 ± 0.5 nm to 31.9 + 1.0 nm as the L D L protein concentration was changed from 0.79 to 0.20 mg protein/ml. At a H E P / L D L ratio of 0.032, a decrease was also o b s e r v e d (40.1 ± 3.8 nm to 36.8 + 2.4 nm as the protein was varied from 0.59 to 0.16 mg/ml), although in this case the errors were greater. At H E P / L D L = 0.02, the 31p-NMR linewidth of L D L - H E P was found to decrease f r o m 62 to 35 Hz as the protein concentration was decreased from 15.6 to 2.2 mg protein/ml. Q E L S investigations on the latter sample (2.2 mg protein/ml) gave a m e a n d i a m e t e r of 39.3 ± 4.3 nm from the distribution analysis and 39.6 ± 0.2 nm from the gaussian analysis.

25

Effect of ionic strength on the mean diameters of LDL- HEP complexes The effect of ionic strength on the L D L - H E P interaction was studied by determining the mean diameter of LDL-HEP complexes (HEP/LDL = 0.020) in Tris as a function of NaCI concentration. The interaction is clearly dependent on the ionic strength. In the absence of NaCI, the mean diameter was 33.4 + 0.8 nm. Addition of NaCI resulted in a monotonic decrease in complex size, with a value of 23.3+0.1 nm observed at [NaCI]= 0.15 M. Thus, at the latter concentration, the large complexes have been broken down into single LDL particles, Discussion

2H- and 3~p have been used to study the formation of soluble complexes of LDL with the glycosaminoglycans HEP, C4S, and C6S. This NMR approach allows each component of the complex to be monitored separately, and in detail, at protein concentrations in the range 15--24 mg/ml. As well, 3~P-NMR and QELS have been used to provide information on the mean diameter and approximate size distribution of the overall complex at protein concentrations ranging from 0.2 to 20 mg/ml. The results from the current study demonstrate significant differences between the types of complexes formed by HEP and CS. Soluble complex formation has been studied using unfractionated GAGs. Mitterer et al. [37] and Eigner et al. [36] have shown that fraction II HEP, the main component obtained from gel filtration on Sephacryl S-300-SF, has binding properties with LDL that is nearly the same as unfractionated H E P . Neither fraction I nor fraction II HEP were studied as the former, a higher molecular weight minor component, caused precipitation of LDL at some concentrations [37], an effect we wished to avoid, and the latter, a lower molecular weight component, had similar binding to fraction III, although the interaction appeared to be somewhat reduced [37]. 2H-NMR provides the clearest demonstration of L D L - G A G soluble complex formation, especially for the chondroitin sulfates. The binding of

both [2H]HEP and [2H]C4S to LDL is clearly evidenced by the broad resonances observed at low G A G / L D L ratios (Figs. 1 and 2). A plot of linewidthversus [2H]GAG/LDL is shown in Fig. 3. In solution, [2H]HEP and [2H]C4S have linewidths of 13 and 32 Hz, respectively, at concentrations of approx, l0 mg/ml. In the presence of LDL at 0.02 mg GAG/mg protein, these values increase 15-fold for [2H]HEP to 220 Hz and 5fold for [2H]C4S to 160 Hz. By way of contrast, the addition of CS to LDL increases the 3tp-NMR linewidths of LDL only by a factor of 1.6, even at high CS/LDL ratios. The 2H-NMR linewidths of [2H]HEP are broader than those of [2H]C4S at G A G / L D L < 0 . 1 3 . There are several factors which may contribute to the observed linewidth of the [2H]GAG. These include the statistical equilibrium constant for the dissocation of G A G and LDL (which determines the fraction of G A G bound to LDL), the overall size of the complex, and the presence of internal motions of the C2H3 group. It is difficult to separate these contributions on the basis of the 2H-NMR results alone, and hence it was found necessary to utilize 31p-NMR and QELS to supplement the 2H data. The addition of HEP to LDL (15--20 mg protein/ml) results in a rapid increase in 31p-NMR iinewidths (Figs. 4 and 6), with a maximum value of 6 2 H z observed at a H E P / L D L ratio of ~0.02 mg HEP/mg protein. Above this ratio, the linewidths decrease rapidly, levelling off at H E P / L D L ~0.2 (Fig. 6). The broad 31p linewidths observed at low H E P / L D L clearly result from complexes composed of several lipoproteins bound together by HEP, and it has been suggested that the stoichiometry (at much lower protein concentrations) is approximately 2--3 LDL particles per 1--2 H E P chains [37]. As further HEP is added, the binding sites on LDL are saturated and the large complexes break down until, in the presence of excess HEP (HEP/LDL > 0.20), the 31P-NMR linewidths are slightly broader (20 Hz) than that of native LDL (13 Hz). From the ~ P - N M R data, it is possible to arrive at an estimate of the complex diameters at higher protein concentrations. The 3~P-NMR linewidth Avon can be expressed in terms of the correlation time for phospholipid reorientation ~'c and the

26 residual second moment M2 by [45] • rAul/2 = MzT~ + C

B = [6D~r/M2]

(1)

The correlation time ¢c is composed of contributions from particle tumbling and phospholipid lateral diffusion, and is given by 11~-~ = liar, + l/ze

(2)

where ~-, and cd are 7t = 4 ~rTIRa/3 k T

(3)

~'a = R 2 / 6 D

(4)

where 7/ is the solvent viscosity, R is the LDL radius, and D is the lateral diffusion coefficient for phospholipids in the LDL monolayer. The residual second moment M2 is given by [46] M2 = (4/45)(2 "/I'Po)2Ao"2

(5)

where Ao" is the residual chemical shift anisotropy. In the native lipoprotein, the value of R in Eqns. 3 and 4 is identical. However, in the case of LDLG A G soluble complexes involving more than a single lipoprotein particle, the values may differ. Assuming isotropic motion for the large LDLHEP complexes (R~ = 25 nm; see below) complex reorientation occurs at a rate I/T,. Since soluble complex formation involves the protein component of LDL, it is reasonable to assume that phospholipid diffusion remains unchanged. Thus, phospholipid lateral diffusion is still occurring at a rate l/ca with R2 = 11.7 nm. Therefore, we assume that the soluble complexes are rigid enough that each LDL in the complex undergoes isotropic tumbling at a rate characterized by the overall complex radius RI, and that phospholipid reorientation due to lateral diffusion occurs at a rate characterized by the radius R2 of a single LDL particle. Then Eqns. 1--5 can be combined to give (RI) 3--- (A lVl/2- C ) A / [ ( R 2 ) 2 - {(AVl/2- C)B}]

(6) A = [3kTR2/4rIM2]

(7)

(8)

As Eqn. 6 is based on several approximations, its reliability was assessed by examining a single L D L - H E P sample (HEP/LDL = 0.022) by both 3~p-NMR and QELS at a concentration of 2.2 mg protein/ml. In order to estimate the complex sizes from Eqn. 6, it is necessary to know both D and Aofor phospholipids in LDL. We have recently used 3tP-NMR to obtain a value of D for phospholipids in LDL of 1.4 x 10 -9 cm2ls [47], and a value of Ao" of ~-50 ppm (D. Fenske, unpublished results). The value of the constant C in Eqn. 6 (4 Hz) was estimated from 31p-NMR of egg PC (63 mg/ml) in C2HCI3/CHCI3 (3:1) at 25°C. Substitution of these values into Eqn. 6 yields a complex diameter of 39.4 nm, which agrees well with the QELS values of 39.3 + 4.3 and 39.6 + 0.2 nm obtained from the distribution and gaussian analyses, respectively. The L D L - H E P complexes observed by 31p-NMR at 15--20 mg protein/ml are calculated to have mean diameters of 49 + 3 nm at HEP/LDL ratios of 0.02--0.03, and 3 1 + 2 n m in the presence of excess HEP. In our hands, a QELS study of the L D L - H E P interaction (Fig. 7) yields results which are in qualitative agreement with previous studies [36,37]. At the lower concentrations necessary for light scattering (~<1 mg protein/ml), complexes with diameters of 37---40nm are formed at 0.025----0.032 mg HEP/mg protein. In the presence of excess HEP (HEP/LDL = 0.57), the mean diameter (27.9+ 1.3 nm) is only slightly larger than that of native LDL (23.4 + 0.4 nm). At H E P / L D L = 0.02, the complex diameters increase by 6--10 nm going from protein concentrations of <1 mg/ml (QELS) to 15--20 mg/ml (31p-NMR). These differences are real, as complex diameter exhibits a dependence on LDL concentration (Table II). The sizes obtained by 3JP-NMR and QELS for excess HEP are the same within experimental error. In contrast to the behaviour observed with HEP, complexes formed with CS do not pass through a maximum diameter at low G A G / L D L ratios. This is shown by the 3~p-NMR linewidth behaviour (Figs. 5 and 6). While the 3~p linewidths of L D L - H E P complexes are much broader than

27 those of LDL-CS complexes at G A G / L D L <0.2, the linewidths of both complexes are comparable at a ratio of 0.3, and even at ratios as high as 0.6, the LDL-CS linewidths are only slighly broader. At these high ratios, the diameters estimated from 31p-NMR are 35 + 2 nm for both C4S and C6S. Similarly, the addition of CS results in monotonic increases in QELS diameter until, at a ratio of 0.57 mg CS/mg protein, the diameter reaches a value of 32.6 + 0.4 nm for C6S and 29.3 + 0.3 nm for C4S, somewhat larger than observed with LDL-HEP complexes at the same ratio (Fig. 7). The effect of physiological concentrations of Ca 2÷ on the L D L - G A G interaction was studied by QELS. The presence of 2 mM Ca 2+ had no effect on the binding of HEP, C6S, or C4S to LDL, and soluble complexes were formed in all cases (calcium concentrations greater than 5 m M are required to effect precipitation [55]). Similar results were obtained by Cardin et al. [52], who examined the binding of fluoresceinamine-labeled HEP to LDL in the presence and absence of 3 mM Ca 2+, and found no differences in stoichiometry, affinity, or solubility. By both 31p-NMR and QELS the binding of C4S and C6S to LDL is similar. Our binding results may be contrasted with a recent study in which C6S was found to be the only G A G retained on LDL-affinity columns in the presence of 50 mM Ca 2+, suggesting that the position of the CS sulfate groups may be of critical importance in the formation of insoluble complexes [4]. We find little evidence to support such a claim for soluble complex formation, both in the absence and presence of physiological concentrations of Ca 2+. With the information on complex size provided by 3tp-NMR and QELS, we may now attempt a very preliminary interpretation of the 2H-NMR data. For a quadrupolar nucleus (I = 1) undergoing isotropic rotation, and line width AlPl/2 the transverse relaxation rate T2 ~ = ~rAvu2 is given by [48]

T-21 = (4"tr2/160)(e2qQ /h)2{9J(O) + 15J(to0) + 6 J(2 tOo)}

(9)

where e2qQ/h is the static constant. The spectral

densities J(to) are J(to) = 2"re/(1 + toZ'r2)

(10)

where % is the effective correlation time for molecular motion. From Eqns. 9 and 10, we can calculate the correlation times for the C2H3 groups of [2H]HEP and [2H]C4S both free in solution and bound to LDL. The value of e2qQ/h used in the calculation was that of deuterated acetone (174.5 kHz [49]). Values of % of 9.4x 10-1~s and 2.2x 10-ms were obtained for [2H]HEP and [2H]C4S in solution, respectively, based on the free linewidths A vt. In order to obtain the values for [2H]GAG complexed to LDL, values of the bound linewidth Av~ were estimated by extrapolation of a plot of log (A v112) versus G A G / L D L to G A G / L D L = 0. This approach assumes that a single A~,d is physically meaningful, which may not be the case if complexes with different stoichiometries (and therefore different sizes) are formed. However, the fact that all of the [2H]GAG spectra, including the broad lineshapes observed at low GAG/LDL, could be computer-simulated with a single Lorentzian lineshape, suggests that the extrapolation was justified. For [2H]HEP, a value of ~'c = (2.4 + 0.5) x 10 -9 s was obtained from Av~ = 260 + 30 Hz. For [2H]C4S, %= 0 . 3 + 0.2) x 10 -9 s was obtained from Avba = 1 7 0 + 2 0 H z . Thus, the correlation times of [2H]HEP and [2H]C4S increase 26-fold and 6fold, respectively, when complexed to LDL. The correlation time of bound [2H]HEP is nearly twice that of bound [2H]C4S, implying slower motions of the former. This could originate from the larger size of the LDL-HEP complexes, or from slower internal motions ("tighter" binding). If the [2H]GAGs were tightly bound and rotating at the same rate as the overall complex, then the correlation time for motion would be given by Eqn. 3. We can calculate a mean ~-c for the overall complex using the sizes obtained from 31p and QELS, assuming that the complexes formed at low G A G / L D L ratios are approximately the same size as those observed at higher ratios, when all or most of the LDL is bound. For LDL-HEP complexes, this is the value observed

28

at H E P / L D L = 0.03 (49 + 3 nm), and for LDL-CS the size observed at CS/LDL --- 0.6 (35 + 2 nm). The size ratio R(LDL-HEP)/R(LDL-CS)= 1.4 is the same by both 31p and QELS. Based on the 31p values, Eqn. 3 gives 7c(LDL-HEP) = 1.5 x 10 -5 s and ~-c(LDL-CS) = 5.5 x 10 -6 s. A comparison of these values with those estimated for [2H]GAG bound to LDL reveals that the C2H3 groups of both GAGs possess considerable spin rotation. Indications of further motions in either G A G could be found from the ratio ~-,([2H]GAG~)/-r,(complex), which normalizes the correlation times with respect to complex size. Assuming a 10% uncertainty in ~-,(complex), the ratios are ( 1 . 6 + 0 . 5 ) x 10 -4 for [2H]HEP and (2.4 + 0.6) x 10 -4 for [2H]C4S. These are the same within experimental error. Therefore, the broader 2H-NMR linewidths observed for bound [2H]HEP at G A G / L D L < 0.13 can be explained on the basis of complex size. This implies that both GAGs have similar internal motions when bound to LDL. L D L - G A G soluble complexes are thought to form via an attraction between the negativelycharged sulfate and carboxyl groups of the G A G and the positively-charged protein groups of LDL [9]. The results in the current study are consistent with this view. Acetylation of LDL, which neutralizes the positive charge on the lysine groups, abolishes the interaction of [2H]HEP with LDL (Fig. 3). Similarly, increasing the ionic strength of the solvent results in the dissociation of large L D L - H E P complexes (see Results). The different behaviour observed with HEP and CS in soluble complex formation with LDL probably originates from the different charge densities and molecular weights of HEP and CS 14,37,46]. It is reasonable to speculate that a single HEP molecule (extended chain length 25 nm for mol. wt. 15,000 [4,37,50]) may not be long enough to interact with the 5--7 H E P binding sitesof apo B [51--53], or that to do so would bring the HEP chain into contact with regions of the LDL surface that possess a net negative charge. The formation of multiple-LDL complexes would allow maximal interaction between the limited number of HEP chains and the positively-charged apo B binding sites. At higher HEP concentrations, saturation of binding sites results in the dissociation of some of the corn-

plexes. In contrast to the behaviour observed with HEP, the lower affinity of CS for LDL results in the formation of fewer multiple-LDL complexes. It is also possible that the longer CS chain [4,50,54] may be able to interact with several binding sites on a single LDL, thereby decreasing the need to equalize surface charges by the 'sharing' of CS chains between lipoproteins. The observation of LDL-CS complexes in the 30---35 nm size range is consistent with our previous analysis [41] of LDL-[2H]C4S complex formation in which we predicted the presence of 2:1 and 3:1 LDL/CS complexes. However, the present study indicates that soluble complex formation, especially at higher protein concentrations, may be more complex than the assumptions of our earlier model allowed for. In our previous paper [41], we generated theoretical binding curves assuming the existence of only one type of complex in solution, and experimental binding curves using the same assumption coupled with an estimated value of the bound linewidth of [2H]C45 complexed to LDL. However, concentration effects may be occurring which would render some of those assumptions inaccurate. Also, the estimated value of AVba we obtain in the present paper for [ 2 H ] C 4 S (170+ 20 Hz), based on computer simulations of the spectra to Lorentzian functions, is considerably smaller than obtained earlier (400 Hz). Thus, considering the large number of assumptions involved in our earlier model, we feel that the accuracy of our previous Ka estimates is uncertain and that several different complexes may exist simultaneously. References 1 J.L. Goldstein and M.S. Brown (1977) Annu. Rev. Biochem. 46, 897--930. 2 R.W. Mahley and T.L. Innerarity (1983) Biochim. Biophys. Acta 737, 197-222. 3 P.A.S. Mourao and C.A. Bracamonte (1984) Atherosclerosis 50, 133--146. 4 P.A.S. Mourao, S. Pillai and N.D. Ferrante (1981) Biochim. Biophys. Acta 674, 178-187. 5 M. Bihari-Varga, G. Camejo, M.C. Horn, D. Szabo, F. Lopez and E. Gruber (1983) Int. J. Biol. Macromol. 5, 59---62.

29 6 7 8

9 10 11 12 13

14 15 16 17

18

19

20

21 22 23 24 25 26 27

28 29

Y.T. Pan, A.W. Krushi and A.D. Elbein (1978) Arch. Biochem, Biophys. 189, 231--240. S.R. Srinivasan, A. Lopez-S, B. Radhakrishnamurthy and G.S. Berenson (1970) Atherosclerosis 12, 321-334. S.R. Srinivasan, B. Radhakrishnamurthy and G.S. Berenson (1975) Arch. Biochem. Biophys. 170, 334---340. P.-H. Iverius (1972) J. Biol. Chem. 247, 2607--2613. P. Bernfeld, J.S. Nisselbaum, B.J. Berkeley and R.W. Hanson (1960) J. Biol. Chem. 235, 2852--2859. G. Camejo, F. Lalaguna, F. Lopez and R. Starosta (1980) Atherosclerosis 35, 307--320. G. Camejo, E. Ponce, F. Lopez, R. Starosta, E. Hurt and M. Romano (1983) Atherosclerosis 49, 241--254. S. Yla-Herttuala, O. Jaakkola, T. Solakivi, H. Kuivaniemi and T. Nikkari (1986) Atherosclerosis 62, 73--80. R.H. Steele, W.D. Wagner, H.A. Rowe and I.J. Edwards (1987) Atherosclerosis 65, 51-----62. J.O. Kokkonen and P.T. Kovanen (1987) Biochem. J. 241,583--589. S.R. Srinivasan, P. Dolan, B. Radhakrishnamurthy and G.S. Berenson (1972) Atherosclerosis 16, 95--104. S.R. Srinivasan, P. Dolan, B. Radhakrishnamurthy, P.S. Pargaonkar and G.S. Berenson (1975) Biochim. Biophys. Acta 388, 58--70. A.S. Kuznetsov (1982) in: E.I. Chazov and V.N. Smirnov (Eds.), Vessel Wall in Athero- and Thrombogenesis, Springer-Verlag, Berlin, Heidelberg, New York, pp. 91 --98. S.R. Srinivasan, B. Radhakrishnamurthy, E.R. Dalferes, Jr. and G.S. Berenson (1979) Atherosclerosis 34, 105118. S.R. Srinivasan, K. Yost, B. Radhakrishnamurthy, E.R. Daiferes, Jr. and G.S. Berenson (1980) Atherosclerosis 36, 25--37. K.W. Walton and N. Wiiliamson (1968) J. Atheroscler. Res. 8, 599---624. M.Z. Alavi and S. Moore (1987) Atberosclerosis 63, 65--74. H.F. Hoff and W.D. Wagner (1986) Atherosclerosis 61, 231--236. M. Bihari-Varga, S. Goldstein~ D. Lagrange and E. Gruber (1982) Int. J. Biol. Macromol. 4, 438-----441. M. Bihari-Varga, J. Sztatisz and S. Gal (1981) Atherosclerosis 39, 19--23. L. Mateu, E.M. Avila, G. Camejo, V. Leon and N. Liscano (1984) Biochim. Biophys. Acta 795, 525--534. L.-A. Fransson (1985) in: G.H. Aspinall (Ed.), The Polysaccharides, Vol. 3, Academic Press, New York, pp. 337---415. J.L. Goldstein, S.K. Basu, G.Y. Brunschede and M.S. Brown (1976) Cell 7, 85--95. C.R. Sirtori, A. Catapano, G. Ghiselli and W.R. Malino (1976) Artery 2, 39(I---399.

30 31

32 33

34 35 36 37 38 39 40 41 42 43 44 45 46 47

48 49 50 51 52 53

54 55

G. Weber, P. Fabbrini, L. Resi, V. Sforza and P. Tanganelli (1979) Pharmacol. Res. Commun. 11,341--348. C. Caruzzo (1974) in: G. Schettler and A. Weizel (Eds.), Atherosclerosis IV, Springer-Verlag, Berlin, pp. 751--757. L.M. Morrison (1971) Angiology 22, 165--174. B. Radhakrishnamurthy, H.A. Ruitz, S.R. Srinivasan, W. Preau, E.R. Dalferes, and G.S. Berenson (1978) Atherosclerosis 31, 217--229. I. Staprans and J.M. Felts (1985) J. Clin. Invest. 76, 1984--1991. Y. Nakashima, N. Di Ferrante, R.L. Jackson and H.J. Pownall (1975) J. Biol. Chem. 25(I, 5386--5392. W.-D. Eigner, A. Mitterer, J. Schurz and G. Jurgens (1982) Biosci. Rep. 2, 413----417. A. Mitterer, W.-D. Eigner, J. Schurz, G. Jurgens and A. Holasek (1982) Int. J. Biol. Macromol. 4, 227--232. R.J. Havel, H.A. Eder and J.H. Bragdon (1955) J. Clin. Invest. 34, 1345--1353. O.H. Lowry, N.J. Rosebrough, A.L. Farr and R.J. Randall (1951) J. Biol. Chem. 193, 265--275. M. Hook, J. Riesenfeld and U. Lindahl (1982) Anal. Biochem. 119, 236--245. D.B. Fenske and R.J. Cushley (1984) Biochem. Biophys. Res. Commun. 121,871--877. E.W. Gold (1979) Anal. Biochem. 99, 183--188. R.W. Mahley, K.H. Weisgraber and T.L. lnnerarity (1979) Biochim. Biophys. Acta 575, 81--91. G. Assmann, E.A. Sokolski and H.B. Brewer (1974) Proc. Natl. Acad. Sci. U.S.A. 71,549--553. A. Abragam (1961) The Principles of Nuclear Magnetism, Clarendon Press, Oxford, p. 424. A.C. McLaughlin, P.R. Cullis, J.A. Berden and R.E. Richards (1975) J. Magn. Reson. 20, 146--165. R.J. Cushley, W.D. Treleaven, Y.I. Parmar, R.S. Chana and D.B. Fenske (1987) Biochem. Biophys. Res. Commun. 146, 1139--1145. A. Abragam (1961) The Principles of Nuclear Magnetism, Clarendon Press, Oxford, p. 315. J. Seelig (1977) Q. Rev. Biophys. 10, 353--418. B. Chakrabarti and J.W. Park (1980) CRC Crit. Rev. Biochem. 8, 225--313. K.H. Weisgraber and S.C. Rail, Jr. (1987) J. Biol. Chem. 262, 11097--11103. A.D. Cardin, C.J. Randall, N. Hirose and R.L. Jackson (1987) Biochemistry 26, 5513--5518. N. Hirose, D.T. Blankenship, M.A. Krivanek, R.L. Jackson and A.D. Cardin (1987) Biochemistry 26, 5505--5512. R.A. Geiman and J. Blackwell (1973) Biochim. Biophys. Acta 297, 452----455. A.D. Cardin, S. Ranganathan, N. Hirose, L. Wallhausser, J.A.K. Harmony and R.L. Jackson (1986) Biochemistry 25, 5258--5263.