Application of Directly Coupled HPLC NMR to Separation and

Feb 15, 2001 - DK-2100 Copenhagen, Denmark, and Division of Analytical Sciences, ... etiology of several disease states such as coronary heart disease...
0 downloads 0 Views 83KB Size
Download PDF
Anal. Chem. 2001, 73, 1084-1090

Application of Directly Coupled HPLC NMR to Separation and Characterization of Lipoproteins from Human Serum Clare A. Daykin,† Olivia Corcoran,† Steen H. Hansen,‡ Inga Bjørnsdottir,‡ Claus Cornett,‡ Susan C. Connor,§ John C. Lindon,† and Jeremy K. Nicholson*,†

Biological Chemistry, Biomedical Sciences Division, Sir Alexander Fleming Building, Imperial College of Science, Technology and Medicine, University of London, South Kensington, London, SW7 2AZ, U.K., Department of Analytical and Pharmaceutical Chemistry, The Royal Danish School of Pharmacy, Universitetsparken 2, DK-2100 Copenhagen, Denmark, and Division of Analytical Sciences, GlaxoSmithKline Pharmaceuticals, New Frontiers Science Park (North), Third Avenue, Harlow, Essex, CM19 5AW, U.K.

Disorders in lipoprotein metabolism are critical in the etiology of several disease states such as coronary heart disease and atherosclerosis. Thus, there is considerable interest in the development of novel methods for the analysis of lipoprotein complexes. We report here a simple chromatographic method for the separation of highdensity lipoprotein, low-density lipoprotein, and very lowdensity lipoprotein from intact serum or plasma. The separation was achieved using a hydroxyapatite column and elution with pH 7.4 phosphate buffer with 100-µL injections of whole plasma. Coelution of HDL with plasma proteins such as albumin occurred, and this clearly limits quantitation of that species by HPLC peak integration. We also show, for the first time, the application of directly coupled HPLC 1H NMR spectroscopy to confirm the identification of the three major lipoproteins. The full chromatographic run time was 90 min with stopped-flow 600-MHz NMR spectra of each lipoprotein being collected using 128 scans, in 7 min. The 1H NMR chemical shifts of lipid signals were identical to conventional NMR spectra of freshly prepared lipoprotein standards, confirming that the lipoproteins were not degraded by the HPLC separation and that their gross supramolecular organization was intact.

Lipoproteins are macromolecular complexes of lipids and globular proteins held together by nonpolar and electrostatic forces. The particle core comprises hydrophobic lipids, e.g., cholesteryl esters (CE) and triglycerides (TG), while the surface is structurally similar to the outer part of a cell membrane, consisting largely of proteins (apoproteins or apolipoproteins) and phospholipids. Cholesterol molecules are of intermediate polarity and enter both the surface and the core regions of lipoprotein * Corresponding author: (tel) +44 (0) 207 594 3195; (fax) +44 (0) 207 594 3221; (e-mail) [email protected]. † University of London. ‡ The Royal Danish School of Pharmacy. § GlaxoSmithKline Pharmaceuticals.

1084 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

particles. The primary function of lipoproteins is to transport lipids through the vascular and extravascular body fluids.1 Lipoproteins are usually divided into broad categories defined historically by their density based on isolation by ultracentrifugation.2 These categories include very low-density lipoprotein (VLDL), lowdensity lipoprotein (LDL), and high-density lipoprotein (HDL), although additional classes such as intermediate density lipoprotein (IDL) have been invoked. Disorders in lipoprotein metabolism are critical to the development of atherosclerosis, coronary heart disease (CHD), liver dysfunction, and cancer, among others, and a variety of analytical methods have been devised to characterize and quantify them. At present, the most routinely used method for isolation and quantification of plasma or serum lipoproteins is ultracentrifugation, followed by specific enzymatic assays to estimate their composition.3 There are, however, practical difficulties associated with the ultracentrifugal isolation of the lipoproteins. Even isolation of the main categories from a serum sample may take up to one week, requiring frequent handling of the sample including tube slicing and adjustment of the density by addition of salt solutions. In ultracentrifugation methods for lipoprotein separation, the high g-forces (>100000g) and high salt concentrations used can cause structural modification to the lipoprotein particles. It is, therefore, preferable to develop a method that avoids these limitations especially for studies on supramolecular structure and dynamics by, for example, NMR spectroscopy. Additionally, the study of molecular mobility and phase transitions in lipoprotein particles may be enhanced by adopting an alternative approach. Another more recently developed method for the separation and quantification of microplasma samples is SMARTFPLC,4 which allows separation of lipoproteins in less than 2 h, but this cannot readily be adapted for use with NMR detection. (1) Voet, D.; Voet, J. G. Biochemistry; John Wiley and Sons Inc.: New York, 1990. (2) Gofman, J. M.; Lindgren, F. T.; Elliot, H. J. Biol. Chem. 1949, 179, 973979. (3) Ala-Korpela, M. Prog. Nucl. Magn. Reson. Spectrosc. 1995, 27, 475-554. (4) Hennes, V.; Gross, W.; Edelmann, A. Sci. Tools 1992, 36, 10-12. 10.1021/ac0011843 CCC: $20.00

© 2001 American Chemical Society Published on Web 02/15/2001

The use of 1H NMR spectroscopy as a technique for detection and accurate quantification of lipoprotein lipids directly from plasma samples is now a widely accepted method3,5-8 and is available as a commercial service (e.g., NMR Lipoprofile, Lipomed Inc., Raleigh, NC).8 The direct NMR-based analysis of blood plasma does not require sample pretreatment, or isolation, or decomposition of the lipoprotein particles. Quantification of the lipoprotein classes and subclasses is achieved by mathematical deconvolution of the heavily overlapping characteristic lipoprotein methyl or methylene resonances in 1H NMR spectra of blood plasma. A complete analysis can be done in less than 1 h and samples can then be used in further biochemical analyses. However, this type of line shape fitting analysis requires prior knowledge of the number and shape of the lipoprotein resonances, which can only be obtained from spectra of pure lipoprotein fractions. Thus, there is the need for a separation method for lipoproteins that is compatible with NMR spectroscopy. Directly coupled HPLC NMR spectroscopy has the potential to achieve this. Because the chromatographic and the spectroscopic methods are nondestructive, the separated lipoproteins can be collected for further biochemical analysis if required. However, standard reversed-phase chromatography is unsuitable for this purpose because the solvents used for analyte elution, e.g., methanol or acetonitrile, will denature the lipoprotein complexes. We report here, a modified HPLC method for the separation of lipoproteins from whole human serum in ∼90 min, which we have validated using directly coupled HPLC NMR spectroscopy. This method has been developed based on the modification of a previously published procedure, which achieved isolation of LDL and VLDL from serum in 10 h, using a Tiselius-type hydroxyapatite column and four stepwise elutions with potassium phosphate buffers.9 The principal aims of optimization described here were to shorten the separation time, to produce a method that would allow NMR study of the separated lipoproteins by directly coupled HPLC NMR, and to validate the use of this novel chromatographic system for lipoprotein separation. In this study, we present a new approach to lipoprotein analysis with the application of directly coupled 600-MHz stopped-flow HPLC 1H NMR spectroscopy to separate and identify the major lipoprotein classes, HDL, LDL, and VLDL. EXPERIMENTAL SECTION Chemicals. Potassium phosphate buffer was prepared using dipotassium hydrogen phosphate and potassium dihydrogen phosphate salts purchased from Fisons (Loughborough, U.K.). Deuterium oxide was HPLC grade from Goss Scientific Instruments (Ingatestone, U.K.). All reagents used for the preparation of the lipoprotein standards were purchased from Sigma (Gillingham, Dorset, U.K.) unless otherwise stated. (5) Ala-Korpela, M.; Korhonen, A.; Keisala, J.; Ho¨rkko ¨, S.; Korpi, P.; Ingman, L. P.; Jokisaari, J.; Savolainen, M. J.; Kesa¨niemi, Y. A. J. Lipid Res. 1994, 35, 2292-2304. (6) Hiltunen, Y.; Ala-Korpela, M.; Jokisaari, J.; Eskelinen, S.; Kiviniitty, K.; Savolainen, M.; Kesa¨niemi, Y. A. Magn. Reson. Med. 1991, 21, 222232. (7) Otvos, J. D.; Jeyaraajah, D. W.; Bennett, D. W.; Krauss, R. M. Clin. Chem. 1992, 38, 1632. (8) Nicholson, J. K.; O’Flynn, M.; Sadler, P. J.; Macleod, A.; Juul, S. M.; So ¨nksen, P. H. Biochem. J. 1984, 217, 365-375. (9) Shibusawa, Y.; Miwa, N.; Hirahima, T.; Matsumoto, U. J. Liq. Chromatogr. 1994, 17, 1203-1217.

Preparation of Human Serum. Blood was collected from a healthy female volunteer by venepuncture. The blood was allowed to stand for 3 h at room temperature until agglutination was complete. Serum was collected after centrifugation at 1000g for 20 min. Preparation of Lipoprotein Standards. Blood was collected into standard clinical vials using EDTA as an anticoagulant, and plasma was obtained after centrifugation at 1000g for 10 min at 4 °C. A protease inhibitor cocktail containing aprotinin, pepstatin A, leupeptin (25 µL/10 mL), and 20 µM 2,6-di-tert-butyl-p-cresol was added to plasma. The plasma density was adjusted to 1.21 g/mL by addition of solid KBr and transferred to centrifuge tubes (Beckman Ultraclear). The density-adjusted plasma was then overlaid with 1.063 g/mL KBr, followed by 1.019 g/mL and finally 1.006 g/mL KBr. The tubes were spun for 67 h at 8 °C and 25 000 rpm, using a Beckman J6-B centrifuge with SW28 rotor. Before fractionation, 3 × 875 µL was carefully removed from the top of each tube (the VLDL layer) and the other lipoproteins were fractionated by piercing the bottom of the centrifuge tube and pumping through the high-density salt solution. The pump flow rate and fraction collector were set to give 875-µL fractions. A 10-µL aliquot from each fraction was assayed for protein, and a graph of fraction number against protein concentration was used to obtain the lipoprotein profile. Fractions were then pooled to give VLDL, LDL, and HDL. HPLC Conditions. Human serum (100 µL) was injected onto an analytical column (12 × 0.46 cm) packed at the Royal Danish School of Pharmacy with Tiselius-type hydroxyapatite Bio-Gel HTP (DNA grade, purchased from Bio-Rad Laboratories). A flow rate of 0.15 mL/min was used, and a gradient of potassium phosphate buffer (pH 7.4) was changed stepwise from an initial concentration of 75 mM to 200 mM after 30 min and to 650 mM after 60 min. Directly Coupled HPLC NMR Spectroscopy. The HPLC system comprised a Bruker LC22C pump (Rheinstetten, Germany), a Milton Roy Spectromonitor D variable-wavelength UV detector (operating at 280 nm), and a Bruker BSFU-0 flow control unit. The column temperature was maintained at 29 °C throughout. The outlet from the UV detector was connected to the HPLC NMR flow via an inert polyether (ether) ketone (PEEK) capillary (2.5 m × 0.25 mm i.d.). HPLC NMR software (Hystar v1.2) controlled the flow dynamics of the system and stored the chromatographic data obtained. The mobile phase was as described above, but prepared with 10% 2H2O. The HPLC 1H NMR spectra of eluting chromatographic peaks were obtained after an injection of 100 µL of blood serum. Spectra were acquired using a Bruker Avance UltraShield 600-MHz NMR spectrometer equipped with a 1H-13C inverse-detection HPLC z-gradient flow probe (cell of 4-mm i.d. with a volume of 120 µL). 1H NMR spectra of individual components, separated from whole human serum, were obtained in stopped-flow mode at 600.13 MHz and a probe temperature of 37 °C. Solvent suppression of the residual HOD signal was achieved by using the presaturation pulse sequence: RD-[90°-t1-90°-tm-90°-]acquire FID with saturation of the water resonance during the relaxation delay (RD) and mixing time (tm). Relaxation and mixing delays (tm) were 2.0 and 0.1 s respectively: t1 is a very short delay, set to 3 µs. A total of 128 FIDs were summed into 64K data points with an acquisition Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

1085

Figure 1. UV-detected (280 nm) chromatogram of lipoproteins separated from whole blood serum. The points indicated, 1-8, show where the flow was stopped so that NMR spectra could be acquired.

time of 2.74 s and spectral width of 20 ppm. Prior to Fourier transformation (FT), an exponential apodization function was applied to the FID, corresponding to a line broadening of 5 Hz, and these data were zero-filled by a factor of 2. Chemical shifts for the HPLC NMR-derived lipoprotein spectra were referenced internally to the resonance from the choline methyl groups in the phosphatidylcholine moieties at δ 3.21. RESULTS Chromatographic Separation of the Lipoproteins. The original published HPLC method9 for plasma lipoproteins allowed efficient isolation of both LDL and VLDL on a preparative scale with a 10-h separation time. Although this compares favorably with the time required for separations based on ultracentrifugation, the original method was only really suitable for the preparative separation of lipoproteins and could not be readily applied to a quantitative lipoprotein analysis. Thus, we modified the HPLC method to improve the efficiency of separation and reduce the separation time. In the unusual case of lipoproteins, which have a complex supramolecular organization, simply increasing the flow rate to reduce separation time was not deemed practical because of the possibility of damage to the lipoprotein structure by increasing the shear forces on the column. Therefore, the column length was simply reduced from 25 to 12 cm to effect reduction of elution time.10 Application of this HPLC method to a whole serum sample showed the separation of three major and at least two minor peaks based on UV detection at 280 nm (Figure 1). Examination of the chromatographic retention times for lipoprotein standards (freshly prepared by ultracentrifugation) measured under identical conditions indi-

cated the elution order was HDL, LDL, and VLDL with approximate retention times of 21.9, 58.9 and 86.6 min, respectively (data not shown). It was, therefore, possible to assign the lipoprotein peaks based on retention times (Figure 1), which were then confirmed using directly coupled HPLC NMR spectroscopy (see below). In the case of whole plasma, the HDL was in the tail of the broad peak from serum proteins (mainly albumin) but the LDL and VLDL peaks were well separated from each other and other plasma components (Figure 1). The chromatographic separation for UV-detected peaks 1, 2, and 3 was shown to be highly reproducible based on the standard deviation of the percentage peak area after three injections of the same serum sample. The mean peak area ( standard deviation for each of the three peaks were 79.0 ( 1.6, 14.5 ( 1.4, and 6.0 ( 0.5 for peaks 1, 2, and 3, respectively. HPLC 1H NMR Spectra of the Chromatographic Peaks. A typical partial 600-MHz 1H NMR spectrum (δ 0.0-4.5) of a 500µL aliquot of whole human blood serum with 50 µL of 2H2O added is shown in Figure 2. The characteristic broad, strong signal complexes in the δ 0.8-1.3 region of the spectrum are assigned to the terminal CH3 and CH2 groups of “mobile” fatty acid components from all of the lipoprotein classes. There are also a number of other lipidic signals from lipoproteins together with sharp resonances from numerous low molecular weight endogenous metabolites that have been assigned in previous studies (Figure 2).11 Expansions of the low-frequency region of the 600-MHz 1H NMR spectra (δ 0.5-2.0) of lipoprotein standards and HPLC NMR spectra of HDL, LDL, and VLDL acquired at 21.9, 58.9, and 86.6 min, respectively, are illustrated in Figure 3. This illustrates the

(10) Li, J. B. Waters Column Int. 1997, VI (5), 6-9.

(11) Nicholson, J. K.; Foxall, P. J. D.; Spraul, M.; Farrant, R. D.; Lindon, J. C. Anal. Chem. 1995, 67, 793-811.

1086

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

Figure 2. Typical partial (δ 0.5-4.5) 600-MHz 1H NMR spectrum (with water presaturation) of whole blood serum with assignments of NMR signals (from ref 11); 3-HB, 3-D-hydroxybutyrate.

Figure 3. Comparison of lipoprotein standards and HPLC-resolved lipoproteins. Lower three traces: Partial 600-MHz 1H NMR spectra of (δ 0.5-2.0), standard (std) HDL, LDL, and VLDL. Upper three traces: Partial 600-MHz HPLC 1H NMR spectra (δ 0.5-2.0) of the HDL, LDL, and VLDL fractions separated using directly coupled HPLC NMR at 15.5, 58.9, and 86.6 min, respectively. Abbreviations: VLDL1, LDL1, and HDL1 indicate the resonances from the terminal methyl groups of the mobile fatty acid chains bound in the various lipoproteins; VLDL2, LDL2, and HDL2 indicate resonances from the methylene groups of the mobile fatty acid chains. The lactate signal is broad due to protein binding.

characteristic 1H NMR chemical shift variation with lipoprotein size,6 where larger lipoproteins, e.g., VLDL, are shifted to higher frequency than the smaller lipoproteins, e.g., HDL. Further, the lipoprotein methyl/methylene resonance area ratios of ∼1:4 for

VLDL and ∼1:1 for LDL are retained. These area ratios are not representative of the proton stoichiometry because of line width variations between the two signals. The presence of lipid resonances other than the dominant methyl and methylene peaks, Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

1087

Table 1. Comparison of the Chemical Shift and Line Width Values between Lipoprotein Standards and HPLC NMR Isolated Lipoproteins

chemical shifts of standard (δ) chemical shifts determined by HPLC NMR (δ) line width of standard (Hz)b line width determined by HPLC NMR (Hz)b a

HDL

LDL

VLDL

0.86 ( 0.01a (CH3) 1.26 ( 0.01 (CH2) 0.85 (CH3) 1.25 (CH2) 26.0 ( 1.43 (CH3) 47.8 ( 2.63 (CH2) 27.8 (CH3) 46.9 (CH2)

0.88 ( 0.01 (CH3) 1.28 ( 0.09 (CH2) 0.87 (CH3) 1.27 (CH2) 28.2 ( 0.0 (CH3) 47.0 ( 0.09 (CH2) 28.0 (CH3) 42.8 (CH2)

0.89 ( 0.01 (CH3) 1.29 ( 0.01 (CH2) 0.89 (CH3) 1.29 (CH2) 19.9 ( 0.38 (CH3) 34.7 ( 6.45 (CH2) 21.2 (CH3) 32.4 (CH2)

Values are quoted as mean ( standard deviation. b Full peak width at half-height.

Figure 4. Partial 600-MHz HPLC 1H NMR “time slice” spectra of peak 1, acquired after stopping the flow at points 1, 2, and 3 indicated in Figure 1. (A) HPLC flow stopped at 11.57 min. The NMR spectrum shows no presence of lipoprotein, but the broad peaks contributing to the uneven baseline indicate the presence of plasma proteins. (B) HPLC flow stopped at 15.5 min. The NMR spectrum shows characteristic lipoprotein resonances with chemical shifts corresponding to HDL. (C) HPLC flow stopped at 21.93 min. The NMR spectrum shows increased intensity of HDL over endogenous low molecular weight contaminants (e.g., alanine and lactate) as compared to (B).

albeit with a poor S/N ratios, provides further confirmation of the separation and identities of the lipoproteins and that they maintained their quaternary structure during chromatographic separation. A comparison of the chemical shifts and line widths at half-height from these spectra with previously acquired 600MHz 1H NMR spectra of lipoprotein standards obtained by ultracentrifugation, is shown in Table 1. Further NMR spectra were acquired between the UV-detected peaks, at about 37 and 72 min (points 4 and 6, indicated in Figure 1). These spectra showed an absence of lipoprotein NMR resonances, indicating a genuine baseline separation of the lipoproteins (data not shown). Coelution of HDL with serum proteins led to a single broad UV-absorbing HPLC peak. However, by monitoring the 1H NMR spectra with “time slices” (incremented stopped-flow analysis of the chromatographic peak) at 11.57, 15.50, and 21.93 min, as shown in Figure 4, it was possible to show a real-time resolution of HDL from serum proteins. It can be seen that, at 11.57 min, no lipoprotein resonances were present, but there were strong signals from lactate and free valine and isoleucine and weak but broad unresolved signals from albumin (appearing as slight baseline curvature, e.g., at δ 1.7 and 0.8-1.0). Lipoprotein resonances due to HDL became evident at ∼15.5 min continuing through to ∼22 min. The lactate signals 1088 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

observed in the HDL fractions shown in Figure 3 are broadened due to binding and chemical exchange with residual albumin resulting in shortened T2 relaxation times.12 DISCUSSION In recent years, the application of directly coupled HPLC NMR to the isolation and identification of xenobiotic metabolites from biofluids has become widespread.12 This, however, is the first example of its use to validate a chromatographic method for the separation of intact lipoproteins in whole blood serum and to prove that the supramolecular organization of the lipoproteins was maintained after chromatographic separation. Hydroxyapatite (HA) chromatographic columns were originally introduced for protein separations13 but have since also been successfully applied to the separation of nucleic acids. The mechanism of separation is poorly understood. However, in attempts to explain the operation of hydroxyapatite chromatography, a number of mechanisms have been proposed. These include (i) protein affinity for calcium ions on the column,14 (ii) (12) Lindon, J. C.; Nicholson, J. K.; Wilson, I. D. Prog. Nucl. Magn. Reson. Spectrosc. 1996, 29, 1-49. (13) Tiselius, A.; Hjerten, S.; Levin, O. Arch. Biochem. Biophys. 1956, 65, 132. (14) Gorbunoff, M. J. Anal. Biochem. 1984, 136, 425-432.

the interaction between polar side groups on the protein and positive sites on the crystal,15 and (iii) the interaction between ionized residues on the protein and positive (Ca2+) or negative (PO42-) sites on the column, depending on the net charge of the protein.16 In the case of lipoprotein separation, an observation that the lipoproteins elute by increasing size, as shown in Figure 1, and that the stationary-phase particle size is critical to the quality of separation9 suggests that separation may at least partly be due to the size differences of the lipoprotein classes. Other chromatographic separations of lipoproteins have been reported. These include, countercurrent chromatography, which was reported to yield two fractions from whole blood plasma, one containing HDL and LDL and a second containing VLDL and serum proteins,17 ion-exchange chromatography, which was carried out on lipoproteins previously separated by ultracentrifugation,18 and gel permeation chromatography, which was also carried out on lipoproteins previously separated by ultracentrifugation.19 Several reversed-phase chromatographic methods for the separation of apolipoproteins, cholesterol, esterified cholesterol, and triacylglycerols have also been reported. 20-22 On the other hand, here a method has been described that allowed separation of the intact lipoprotein classes LDL and VLDL, although the UV detector was unable to resolve the serum proteins from HDL. However, by “time-slicing” through the UV-detected peak, to acquire NMR spectra at 11.57, 15.50, and 21.93 min, it was revealed that the initial proportion of the UV-detected peak contained only serum proteins, whereas HDL eluted later. This highlights an important advantage of acquiring NMR spectra on-line: while the UV detector is unable to resolve the serum proteins from the lipoprotein, NMR can allow a more detailed study of such peaks. A previously reported approach to obtain serum protein-free lipoprotein fractions is to use countercurrent chromatography, followed by hydroxyapatite chromatography.17 The use of NMR detection allows the observation of separation which cannot be observed by the UV detector. This reduces the requirement to carry out a preliminary separation to isolate HDL from the proteins and, thus, leads to a reduction in experimental sample loss. 1H NMR spectra of intact plasma samples are extremely complex, and one-dimensional NMR spectra contain heavily overlapped resonances, even at 750 MHz 1H observation frequency.11 Nonetheless, the use of 1H NMR spectroscopy to study various dyslipidemic conditions in whole blood plasma has become increasingly popular and NMR-based diagnostic procedures for lipid and lipoprotein measurements are now available commercially. To obtain information on the lipoprotein profiles, two approaches have been used. One such method is line shape fitting (deconvolution) applied to the 1H NMR spectra of whole blood (15) Gorbunoff, M. J. Anal. Biochem. 1984, 136, 433-439. (16) Gorbunoff, M. J.; Timasheff, S. N. Anal. Biochem. 1984, 136, 440-445. (17) Shibusawa, Y.; Mugiyama, M.; Matsumoto, U.; Ito, Y. J. Chromatogr., B 1995, 664, 295-301. (18) Haginaka, J.; Yamaguchi, Y.; Kunitomo, M. Anal. Biochem. 1995, 232, 163171. (19) Vercaemst, R.; Rosseneau, M. J. Chromatogr. 1983, 276, 174-181. (20) Meyer, B.; Kecorius, E.; Barter, P.; Fidge, N.; Tetaz, T. J. Chromatogr. 1991, 540, 386-391. (21) Tetaz, T.; Kecorius, E.; Grego, B.; Fidge, N. J. Chromatogr. 1990, 511, 147-153. (22) Perona, J. S.; Barron L. J. R.; Ruiz-Gutierrez, V. J. Chromatogr., B 1998, 706, 173-179.

plasma. However, this approach requires prior knowledge of the chemical shifts and line widths of the individual lipoprotein components. In some cases, for example, when the lipoproteins from transgenic animal models are studied, these data may not be available. In these studies, the 1H NMR lipoprotein chemical shifts and line widths may vary from those found in humans due to species differences such as alterations in the cholesterol/ triglyceride ratios within individual lipoprotein classes. Second, artificial neural network (ANN) analysis has been used. This does not require a priori chemical shift information but does require a large data set, including a “training set”.23 In the case of many clinical investigations, it is often difficult to obtain the large numbers of samples required. Hence, in cases such as animal studies or restricted sample sizes, there may be insufficient information to allow analysis of the 1H NMR spectra of whole blood plasma. Chromatographic separation of the lipoprotein classes is presented here as an alternative method, with NMR spectroscopy used to confirm the integrity of the lipoprotein peaks. Furthermore, because NMR is a nondestructive method, the chromatographic peaks can be collected after NMR spectroscopy for further conventional analysis of the cholesterol and triglyceride concentrations. To allow facile detection of lipoproteins using NMR spectroscopy, a 100-µL injection of serum was used. Assuming a typical concentration of LDL in adult human serum of 250 mg/dL,24 this injection volume would provide ∼14 µg for detection by NMR and is, therefore, well above the detection limits for stopped-flow HPLC NMR.12 In cases where only small quantities of blood serum are available, for example in the case of small animal studies, this injection volume could be reduced further. By injecting less serum, the chromatographic peak widths would be reduced with a consequent increase in resolution. Together with a calibration curve, this would allow a quick and simple method for the quantification of LDL and VLDL directly from the UV-detected chromatogram. The small sample volume would also allow a subsample of the whole serum or plasma to be retained for further analysis, if desired. Thus, this HPLC approach to lipoprotein measurement may be useful in assessment of patient samples, particularly for LDL, as this is an important risk factor in coronary heart disease. From an analytical viewpoint, the chromatographic method would compare favorably with the currently widely used practice of precipitating HDL, followed by application of the Friedewald approximation to estimate LDL cholesterol, which though quick and inexpensive may not always be reliable.25,26 In routine clinical chemical analysis, it is envisaged that the simple HPLC-UV methodology would be applied as the additional information provided by the NMR spectrometer is primarily to provide proof of lipoprotein integrity following chromatographic separation. However, it is possible that in dyslipidemic states with unusual lipoprotein compositions that the directly coupled HPLC NMR approach may provide valuable additional information on the lipoprotein speciation. (23) Ala-Korpela, M.; Hiltunen, Y.; Bell J. D. NMR Biomed. 1995, 8, 235-244. (24) Eisenberg S. J. Lipid Res. 1984, 25, 1017-1058. (25) Schechtman, G.; Patsches, M.; Sasse E. A. Clin. Chem. 1996, 42, 732737. (26) Schechtman, G.; Sasse E. Clin. Chem. 1993, 39, 1495-1503.

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

1089

CONCLUSIONS A novel HPLC method using a hydroxyapatite column was applied to the on-line separation and identification of three major lipoproteins after a single injection of 100 µL of whole human serum. Given the relative fragility of the lipoprotein particles, directly coupled HPLC NMR was applied using identical chromatographic conditions and showed that the lipoprotein 1H NMR signals were unaffected by the chromatographic process and that, therefore, the supramolecular structures of the lipoproteins remained intact. The relative efficiency of the chromatographic method suggests that it may find future use in the quantitative determination of plasma LDL and VLDL.

1090

Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

ACKNOWLEDGMENT The authors acknowledge the EPSRC, UK, and GlaxoSmithKline Pharmaceuticals, UK for the funding of C.A.D.. This work was also supported by the EU Biomed 2 program “Hyphenated analytical techniques”, Grant BMH4-CT97-2533 (DG 12-SSMI). We also thank Martin G. Benson (GlaxoSmithKline Pharmaceuticals) for provision of the lipoprotein standards.

Received for review October 4, 2000. Accepted January 23, 2001. AC0011843