Anal. Chem. 2005, 77, 200-207
Metal Ion Complexes of EDTA: A Solute System for Density Gradient Ultracentrifugation Analysis of Lipoproteins Brian D. Hosken, Steven L. Cockrill, and Ronald D. Macfarlane*
Department of Chemistry, Texas A&M University, College Station, Texas 77843
In the study reported here, we apply some of the features of coordination chemistry to solve a long-standing problem in the separation and characterization of lipoprotein particles. Lipoproteins are circulating micelle-like particles responsible for lipid transport. They exist in three major classes: very-low-density lipoprotein, low-density lipoprotein, and high-density lipoprotein in well-defined density ranges using the density gradient ultracentrifugation (DGU) method. The analytical instrumentation of DGU has improved over the years in response to clinical evidence that certain lipoprotein species are linked to a high risk for developing cardiovascular disease. A longstanding problem has been a lack of appropriate gradientforming solutes that can generate a useful gradient from a homogeneous solution. We have found that a new class of solutes based on metal ion complexes has the potential of providing a wide selection of compounds where the features can be modulated by choice of ligand, complexing metal ion, and counterion. In this study, we have chosen the cesium salt of BiEDTA (CsBiEDTA) and have investigated the dynamics of density gradient formation in the ultracentrifuge. We show that a useful density gradient can be formed within a few hours beginning with a homogeneous solution. We also present data on the migration behavior of lipoproteins under gradient-forming conditions and show that high-resolution density profiles can be obtained with good precision. The resolution of the CsBiEDTA profile is compared with those obtained using high molecular weight organic solutes. In 2001, the National Cholesterol Education Program-Adult Treatment Panel III (NCEP-ATP III) identified small, dense lowdensity lipoprotein (LDL) and high-density lipoprotein (HDL) subspecies as emerging risk factors because of their significant association with cardiovascular heart disease.1 Despite the fact that these particle populations can be separated and identified by density gradient ultracentrifugation (DGU), they were not formally adopted as accepted risk factors because no standard methods for their routine measurement have been validated. While there exist traditionally accepted density ranges for these species, the ability to measure lipoprotein particle densities on an absolute * To whom correspondence should be addressed. E-mail: macfarlane@ mail.chem.tamu.edu. Phone: (979)-845-2021. Fax: (979)-845-8987. (1) Circulation 2002, 106, 3143-3421.
200 Analytical Chemistry, Vol. 77, No. 1, January 1, 2005
scale would be an important step toward developing a primary density standard. These problems demonstrate the need for a widely available and accepted method for the profiling of serum lipoproteins in the clinical and research environment. DGU has been the standard method for lipoprotein profiling but is considered time-consuming and tedious compared to other profiling methods.2 Gel electrophoresis,3 capillary electrophoresis,4 gel permeation chromatography,5 NMR spectroscopy,6 and flotation centrifugation7 have been developed as alternatives, but all fundamentally rely on density gradient centrifugation as the “gold standard”, because the classical definitions of lipoprotein species are based on their hydrated densities.8 For this reason, DGU is the most logical method for lipoprotein profiling. In this study, we introduce a new concept in density gradient-forming solutes that addresses many of the problems associated with traditional lipoprotein density profiling by DGU. Ideally, the density gradient separation should be rapid, have high resolution, provide density information, and be reproducible in any laboratory that has an ultracentrifuge. These objectives can only be achieved with equilibrium density gradient ultracentrifugation (EDGU). The simplest method for performing an isopycnic separation begins with a homogeneous solution containing both the sample and the gradient-forming solute. During centrifugation, the solute forms a concentration gradient with a slope dependent on the relative strengths of opposing sedimentation and diffusion forces.9 As the gradient forms, different particles sediment or float to a position in the gradient that corresponds to their respective hydrated densities. The development of this method for lipoprotein profiling has been hampered by the lack of an ideal ultracentrifugation medium. Alkali metal halide salts such as NaBr have been used for isopycnic separations, but the dynamics of gradient formation are such that a pre-formed gradient (by layering solutions of different densities) or very long centrifugation times are required.10 Sucrose (2) Nauck, M.; Warnick, G. R.; Rifai, N. Clin. Chem. 2002, 48, 236-254. (3) Hoefner, D. M.; Hodel, S. D.; O’Brien, J. F.; Branum, E. L.; Sun, D.; Meissner, I.; McConnell, J. P. Clin. Chem. 2001, 47, 266-274. (4) Schmitz, G.; Mollers, C.; Richter, V. Electrophoresis 1997, 18, 1807-1813. (5) Kitamura, T.; Ito, S.; Moriyama, H.; Kato, Y.; Sasamoto, K.; Okazaki, M. Chromatography 1996, 17, 33-37. (6) Otvos, J. D.; Jeyarajah, E. J.; Bennett, D. W. Clin. Chem. 1991, 37, 376386. (7) Kulkarni, K. R.; Garber, D. W.; Marcovina, S. M.; Segrest, J. P. J. Lipid Res. 1994, 35, 159-167. (8) Havel, R. J.; Eder, H. A.; Bragdon, J. H. Circulation 1955, 7, 1345-1353. (9) Ifft, J. B.; Voet, D. H.; Vinograd, J. J. Phys. Chem. 1961, 65, 1138-1145. 10.1021/ac0490402 CCC: $30.25
© 2005 American Chemical Society Published on Web 12/02/2004
gradients have also been used but their viscosity at higher densities prevents lipoproteins from reaching their isopycnic points.11 In contrast, the nonionic, isoosmotic derivatives of triiodobenzoic acid (Nycodenz and Iodixanol) are capable of rapid gradient formation from a homogeneous solution.12,13 These solutes have been the best candidates for EDGU. An optimal density gradient medium should contain solute molecules with a high molecular weight and a large diffusion coefficient for rapid gradient formation. The ideal medium should also have a low viscosity so that lipoprotein migration to isopycnic density ranges is not impaired. These properties were achieved in this study by complexing a heavy metal ion to a highly soluble organic ligand to produce a compact but high molecular weight solute. A significant advantage offered by metal ion complex solutes is their flexibility. The metal ion, ligand, or counterion can be selected to provide the optimal gradient for a specific separation by “fine tuning” the molecular weight of the solute. This paper provides results for one member of this new class of solutes, the cesium salt of a bismuth-EDTA complex (CsBiEDTA). The crystal structure and coordination chemistry of CsBiEDTA crystals were reported by Davidovich et al.14 Here we present the relevant features of CsBiEDTA solutions as a density gradient-forming solute and their ability to form selfgenerating gradients. The equilibrium separation of serum lipoproteins in a CsBiEDTA gradient is also demonstrated and compared to separations performed using other solute systems. EXPERIMENTAL SECTION Materials. Iodixanol and Nycodenz were obtained from Accurate Chemical & Scientific Corp. (Westbury, NY). NBD C6-ceramide was purchased from Molecular Probes (Eugene, OR). Cs2CO3, (BiO)2CO3, and EDTA were purchased from Alfa Aeser Co. (Ward Hill, MA). Synthesis of CsBiEDTA. CsBiEDTA‚H2O was synthesized from Cs2CO3, (BiO)2CO3, and EDTA using a procedure similar to one described for the synthesis of NaBiEDTA‚3H2O.15 The reagents were combined in equivalent amounts, followed by refluxing for 2 h, and then filtration of the solution through a 0.20-µm filter. The solution was reduced in volume to the point of supersaturation and then allowed to rest undisturbed until crystal formation brought the solution back to its saturation point. The crystals were isolated by decanting the solution, rinsed with water, dried under vacuum, and stored in a desiccator. The solution volume was reduced again, and the crystallization procedure was repeated. Hydrodynamic Properties of CsBiEDTA Solutions. A series of six CsBiEDTA solutions were made for the construction of calibration curves necessary for relating density and refractive index to concentration. The density of each solution was determined gravimetrically with a calibrated 10-mL glass pipet. (10) Foreman, J. R.; Karlin, J. B.; Edelstein, C.; Juhn, D. J.; Rubenstein, A. H.; Scanu, A. M. J. Lipid Res. 1977, 18, 759-767. (11) Cruzado, I. D.; Cockrill, S. L.; McNeal, C. J.; Macfarlane, R. D. J. Lipid Res. 1998, 39, 205-217. (12) Rickwood, D.; Ford, T.; Graham, J. Anal. Biochem. 1982, 123, 23-31. (13) Ford, T.; Graham, J.; Rickwood, D. Anal. Biochem. 1994, 220, 360-366. (14) Davidovich, R. L.; Gerasimenko, A. V.; Logvinova, V. B. Russ. J. Inorg. Chem. 2001, 46, 1518-1523. (15) Jaud, J.; Marrot, B.; Brouca-Cabarrecq, C.; Mosset, A. J. Chem. Crystallogr. 1997, 27, 109-117.
Partial specific volumes were calculated from the solution densities. The refractive index of each solution was measured at 20 °C using an Abbe 60/DR refractometer from Bellingham + Stanley (Lawrenceville, GA). Gradient Formation and Measurement. Density gradients were formed in an Optima TLX ultracentrifuge, TLA 120.2 fixedangle rotor, and 1.5-mL, thick-walled, polycarbonate, ultracentrifuge tubes (Beckman-Coulter, Palo Alto, CA). All tubes contained 1000 µL and were centrifuged at 5 °C. Rotor speeds of 80 000, 100 000, or 120 000 rpm were used. For the TLA120.2 rotor, these speeds correspond to average relative centrifugal forces of 227000-, 355000-, and 511000g, respectively. Centrifugation periods ranged from 1 to 16 h. The rotor was accelerated to 5000 rpm and decelerated from 5000 rpm in 15 s. After centrifugation, gradient shapes were determined by removing 20-µL aliquots from discrete positions within the gradient and measuring their refractive index. Each tube was sampled sequentially from the top so that the gradient below each aliquot was not disturbed. Tubes were imaged while the sample was being removed so the exact location of each aliquot could be determined by digital analysis. Blood Draw. Blood from fasting donors was drawn into Vacutainer-brand serum collection tubes (Beckton Dickinson Systems, Franklin Lakes, NJ). The blood was allowed to clot for 15 min and then was spun for 15 min at 3200 rpm to separate the serum from the clotted red blood cells. The serum supernatant was divided into 250-µL aliquots and used immediately or stored at -86 °C until needed. Lipoprotein Profiling. Sample preparation consisted of diluting 6 µL of serum with 590 µL of water and 600 µL of a stock solution of gradient-forming media (either 20% w/v CsBiEDTA, 30% w/v Nycodenz, or 30% w/v Iodixanol). The sample volume was raised to 1200 µL with the addition of 4 µL of 1 mg/mL NBD C6-ceramide. The sample was mixed and then allowed to rest for 30 min so that the fluorescent probe could equilibrate with the lipoprotein particles. Afterward, 1000 µL of the sample was transferred to an ultracentrifuge tube and centrifuged at 120 000 rpm and 5 or 25 °C for 41/2 h. Reproducibility of the CsBiEDTA gradient and lipoprotein separation was demonstrated by dividing one sample between 10 centrifuge tubes. A 12-mL sample was produced by increasing the volumes of serum, stain, H2O, and CsBiEDTA by a factor of 10. As before, 1 mL of sample was transferred to each tube and centrifuged. Digital Imaging and Analysis. After centrifugation, the tubes were imaged with a custom-built fluorescence photography station located in a dark room. The main components consisted of a Plexiglas tube holder, a digital Optronics Microfire Camera (Goleta, CA), and a Dolan-Jenner (Lawrence, MA) MH-100 metal halide continuous light source fitted with a liquid light guide. The tube holder, camera, and light source were mounted orthogonal to each other on an optical breadboard. The camera lens and light guide were equipped with Schott Glass (Elmsford, NY) filters chosen for NBD excitation (BG12) and emission (OG515) wavelengths. BG12 is a band-pass filter centered at 407 nm with a width of 104 nm at 50% transmittance (fwhm). OG515 is a longpass filter that transmits wavelengths greater than 515 nm. For the analysis, image files were converted to an 1800 × 1200 matrix using a software application (Origin 7.0, Microcal Software Inc., Northampton, MA). The intensity values for 20 columns correAnalytical Chemistry, Vol. 77, No. 1, January 1, 2005
201
Table 1. CsBiEDTA Solution Properties (25 °C) concn (% w/v)
concn (M)
density (g/mL)
refractive index
0 10.02 20.04 30.04 40.05 50.03 60.03
0.000 0.150 0.290 0.422 0.546 0.661 0.771
0.99823 1.0643 1.1248 1.1809 1.2347 1.2836 1.3300
1.3330 1.3456 1.3563 1.3668 1.3767 1.3860 1.3948
sponding to the center of the tube were averaged. A tube coordinate scale was established where “zero” is the top of the tube and 34.0 mm is at the bottom of the tube. When the tube is filled with a 1.00-mL volume, the meniscus is located at 9.1 mm on the tube coordinate scale. The final step of the analysis was plotting the averaged intensity as a function of tube coordinate. Equilibrium Experiments. The time required for lipoproteins to reach their isopycnic point in a CsBiEDTA gradient was determined by centrifuging homogeneous solutions of stained serum and 10% w/v CsBiEDTA for 2, 3, 4, 41/2, 5, 6, and 8 h. Weighted mean densities of LDL and HDL were calculated at each time interval. Another experiment designed to test the equilibrium nature of this separation was based on the expectation that the initial sample state (layered or homogeneous) should not affect the outcome of an equilibrium separation. The serum sample was prepared by diluting 50 µL of serum with 40 µL of water and 10 µL of NBD C6-ceramide. After incubation for 30 min, 10 µL of the 100-µL sample was layered on top of 1000 µL of 10% w/v CsBiEDTA or mixed until homogeneous with 1000 µL of 10% w/v CsBiEDTA. Solute Comparisons. Nycodenz, Iodixanol, and CsBiEDTA solutions of differing concentrations but equivalent densities were centrifuged for 41/2 h at 120 000 rpm. Since all three solutions formed significant gradients during this time period, the experiment was repeated with the inclusion of 5 µL of NBD stained serum. RESULTS Relevant Properties of Aqueous CsBiEDTA Solutions. Large, well-formed crystals were obtained from the supersaturated CsBiEDTA solutions. An important characteristic of the CsBiEDTA solution is the high solubility and maximum density at 25 °C. The crystals were found to be soluble in water up to a concentration of 63% w/v (F ) 1.351 g/mL). This density is clearly above the maximum density of the lipoproteins and lipid-free serum proteins. To measure the density gradient formed in the ultracentrifuge, a link had to be developed between density and a physical property of the solution. Refractive index was chosen for this application due to the small sample volume requirements of the analysis. The density and refractive index values were measured for six CsBiEDTA solutions covering a density range from 1.06 to 1.33 g/mL. The results are given in Table 1. These values did not change when the product was recrystallized, indicating that the initial CsBiEDTA crystals were of high purity. Using these data, a linear calibration curve was obtained (R2 ) 0.9957). To calculate the density of CsBiEDTA from its refractive index, eq 1 was used. 202
Analytical Chemistry, Vol. 77, No. 1, January 1, 2005
F ) 5.3878η - 6.1838
(1)
CsBiEDTA solutions are transparent in the visible region but absorb strongly in the ultraviolet region. The complex has a maximum absorbance at 264 nm ( ) 10 000 cm-1 M-1). This intense absorbance is most likely the result of charge transfer between the ligand and the Bi3+ ion. Density Gradient Formation from Homogeneous CsBiEDTA Solutions. The density profile generated in the ultracentrifuge tube was measured by sampling the gradient at specific depths, as described in the Experimental Section. The ability of CsBiEDTA to form self-generating gradients was validated by centrifuging homogeneous solutions. The influence of three different parameters on gradient formation was studied. The first of these was the influence of spin time. A 10% w/v solution, (F ) 1.064 g/mL) was chosen for the initial homogeneous solution. Figure 1A depicts the evolution of the density gradient as a function of time over a 4-h period at 120 000 rpm (511000gAV). After 1 h, a detectible change from the homogeneous solution was observed. The density dropped to 1.04 g/mL near the meniscus and increased to 1.09 g/mL near the bottom of the tube. This concentration profile is indicative of a sedimentation process and was best described by a third-order polynomial. After 4 h, the gradient ranged from 1.00 to 1.25 g/mL and approached the exponential shape expected for equilibrium sedimentation conditions. Based on the study described above, we then investigated the influence of initial CsBiEDTA concentration on gradient formation setting a fixed spin time of 4 h. Gradients formed from 10, 15, and 20% w/v homogeneous CsBiEDTA solutions were examined. The results are shown in Figure 1B. In all three cases, nearexponential profiles were formed. The influence of concentration on the profile was an expected but important finding. Using this figure, it is possible to estimate the approximate locations of the LDL and HDL distribution and to control the separation between the two distributions. For our studies, where we are particularly interested in the separation of all the major lipoprotein classes in a single spin, we selected the 10% w/v CsBiEDTA solution. Finally, using the 10% w/v CsBiEDTA solution, we increased the centrifugation time to 16 h to approach near-equilibrium conditions for gradient formation and measured the influence of rotor speed. These results are shown in Figure 1C. Rotor speeds of 120 000, 100 000, and 80 000 rpm were used. As predicted from theory, the slope of the gradient increased with rotor speed. Based on these three results, we selected 10% w/v CsBiEDTA as the starting solution, 41/2 h as the spin time, and 120 000 rpm as the optimal parameters for obtaining the lipoprotein particle density profile. Migration of Lipoproteins in a CsBiEDTA Gradient. Two questions were to be answered in this part of the study, the first being whether lipoprotein classes can reach equilibrium in a dynamic CsBiEDTA gradient. For this experiment, stained serum was added to the 10% w/v CsBiEDTA solution and the resultant mixture centrifuged for increasing lengths of time. Figure 2A shows the lipoprotein distribution after a 2-h spin. Superimposed on the distribution is the density gradient profile. During that period, the buoyant very-low-density lipoprotein (VLDL) and LDL have floated to the meniscus (9.0 mm on tube coordinate scale)
Figure 1. Influence of centrifugation time, concentration, and radial acceleration on CsBiEDTA density gradient formation. Panel A shows the gradient formed after (9) 1, (b) 2, and (2) 4 h of centrifugation at 120 000 rpm. The solid line represents the initial solution density. Panel B demonstrates the gradient formed from (9) 10 (1.0643 g/mL), (b) 15 (1.0944 g/mL), and (2) 20% w/v (1.1248 g/mL) CsBiEDTA solutions after being centrifuged for 4 h at 120 000 rpm. Panel C shows the gradient formed after 16 h of centrifugation at (b) 80 000 (227000gAV), (9) 100 000 (355000gAV), and (2)120 000 rpm (511000gAV).
with the remaining HDL particles essentially evenly distributed throughout the tube. After 3 h (Figure 2B), the LDL separates from the meniscus and HDL particles begin to noticeably sediment toward the bottom of the tube. At 4 h (Figure 2C), the LDL and HDL distributions are now well resolved. Continuing to 5, 6, and 8 h, the LDL distribution is shifting to higher tube coordinates while the HDL distribution remains fixed at 25 mm. The second issue investigated was the time requirements for various lipoprotein particles to focus at their isopycnic positions within the density gradient. To answer this question, we measured the densities of LDL and HDL as a function of centrifugation time over a period of 3-8 h. The weighted mean densities of the LDL and HDL bands were calculated from their positions in the gradient. Table 2 shows the results of that analysis. Between 3 and 41/2 h, there is some variation in the density values, but after 41/2 h, both the LDL and HDL values did not change significantly. The lipoprotein bands had reached their isopycnic points in a gradient not yet at equilibrium. With a CsBiEDTA solute system, it is possible to achieve an equilibrium separation before the gradient reaches equilibrium. This finding also means that the absolute lipoprotein particle densities can be measured. Comparison of Two Initial States: Layered versus Homogeneous. Another experiment was designed to test the equilibrium nature of this separation. Figure 3 shows two lipoprotein profiles generated from the same serum sample. The only difference is the initial state of the sample as noted in the Experimental Section. With the exception of the VLDL band (which cannot reach equilibrium because it has a density less than water), the profiles from the layered sample and homogeneous samples are essentially the same indicating that the initial sample distribution had no affect on the final separation. This result is a requirement and further evidence that an equilibrium separation had been achieved.16 When this experiment was repeated with a Nycodenz solute system (result not shown), the separation demonstrated significant dependence on the initial sample distribution. Resolution and Precision of the CsBiEDTA Gradient. The 10 lipoprotein profiles superimposed in Figure 4 illustrate the precision and resolving power of the CsBiEDTA gradient. The position in the density gradient at which maximum intensity for each lipoprotein species occurred was the same for all profiles, and there were only minor differences in fluorescence intensity. The resolution of the lipoprotein density profile improved as the amount of serum applied to the gradient decreased. When serum comprised only 5 µL of the 1000-µL total volume, near-baseline separation of the major lipoprotein classes was observed. Importantly, buoyant and dense LDL subfractions (labeled LDLB and LDLD in Figure 4) were well resolved. The lipoprotein-free serum proteins also form an isopycnic band that can be visualized because the lipid-binding domains of serum albumin also bind NBD C6-ceramide. Carrying out the ultracentrifugation at the lower temperature reduces the effect of thermal diffusion and greatly reduced the amount of evaporation of the solution during the spin (
