- Email: [email protected]
Isolation and Measurement of Colloids in Human Plasma by Membrane-Selective Flow Field-Flow Fractionation: Lipoproteins and Pharmaceutical Colloids
+
+
Isolation and Measurement of Colloids in Human Plasma by Membrane-Selective Flow Field-Flow Fractionation: Lipoproteins and Pharmaceutical Colloids PING LI
AND
J. CALVIN GIDDINGSX
Received August 7, 1995, from the Field-Flow Fractionation Research Center, Department of Chemistry, University of Utah, Salt Lake City, UT 84112. Final revised manuscript received April 18, 1996 . Accepted for publication April 18, 1996X. Abstract 0 Using a modified flow field-flow fractionation (flow FFF) technique termed membrane-selective flow FFF, a capability is developed to isolate various colloidal constituents that are naturally present in or may be introduced into blood plasma. Once isolated, the colloids may be subject to quantitative measurement to provide relative amounts of the different constituents and the size distribution curve of each. The potential of the technique has been demonstrated by isolating and measuring (a) the lipoprotein fractions found in human blood plasma, (b) liposyn II pharmaceutical emulsion added at 2 mg/mL of plasma, and (c) amphotericin B colloidal dispersion (a drug delivery agent) added to plasma at 0.5 mg/mL.
Blood plasma (blood without cells) is a complex macromolecular-colloidal solution containing many components, principally numerous proteins and various classes and subclasses of lipoproteins.1-3 The high abundance of proteins (60-80 mg/mL of plasma) generally interferes with the measurement of other colloidal constituents, including lipoproteins (6-8 mg/mL) and various pharmaceutical colloids (such as liposomes and emulsions) that may be introduced into the blood stream as carriers of drugs or nutritional factors. The latter colloids, when present, require measurement because they significantly influence biological function and health status despite their fleeting existence and low concentration. Because of mutual interference, it is difficult to measure the concentrations of specific plasma colloids; it is especially difficult to measure distributions in their properties such as size and mass distributions. Most techniques for isolating subpopulations of colloids are laborious, lack good resolution, and require multiple experimental steps. Membrane techniques, for example, could be used to remove albumin and other low molecular weight proteins, but after this step a complex and intransigent mixture of colloids (large proteins and lipoproteins) remains. Ultracentrifugation can be used to isolate the lipoproteins, but the procedure consists of several stages and requires over 1 day for completion.4,5 We report here a technique based on flow field-flow fractionation (flow FFF) that is capable, in a single step, of isolating different size classes of plasma colloids >7 nm in diameter and measuring their size distribution. The procedure requires only 10-40 min and could probably be made significantly faster. The technique relies primarily on the resolving power of flow FFF to separate macromolecular and colloidal materials. The flow FFF process is assisted by membrane selectivity with no additional experimental steps or complications. Normally, the large abundance of small proteins hinders the flow FFF separation and detection of plasma colloids. As a remedy, our technique employs a membrane process that X
Abstract published in Advance ACS Abstracts, June 1, 1996.
© 1996, American Chemical Society and American Pharmaceutical Association
Figure 1sSeparation of three components, A, B, and C, along the main flow axis of a flow FFF channel. A secondary flow (the cross flow) drives the three components into cloud-like bands adjacent to the membrane. The clouds have different mean elevations because the constituent particles have different hydrodynamic diameters and thus different diffusion coefficients. The different elevations within the parabolic axial flow profile lead to differential displacement and thus separation. In membrane-selective flow FFF, the smaller particles (e.g., those in the C band) are gradually flushed through the membrane and the underlying support frit so as to avoid interference with bands A and B. Diagram shows edge view of ribbon-like channel.
purges most proteins <7 nm in diameter from the colloidal suspension. We term this technique membrane-selective flow FFF. It is a simple matter to combine flow FFF and membrane separation because flow FFF systems generally incorporate a membrane as a critical functional element.6,7 Perhaps surprisingly, the incorporated membrane is used for the retention but not the separation of macromolecular constituents. However, since blood plasma has such overwhelming concentrations of small proteins, the flow FFF membrane can be put to use to remove the bulk of these interfering proteins without lengthening or complicating the one-step procedure of flow FFF. Flow FFF is one member of the field-flow fractionation (FFF) family of techniques.6-11 The sister techniques of sedimentation FFF and thermal FFF (neither employing a membrane) are also used for separating a wide range of colloids and polymers. Although the FFF techniques have been described widely in the literature,6-11 a brief account will be given to provide proper context for this work. FFF is a family of chromatographic-like elution techniques applicable to macromolecules, colloids, and cell-sized particles. Separation is achieved in a thin ribbon-like flow channel. A field is applied across the channel in a direction perpendicular to flow. The field generates a driving force that induces unlike particles to occupy different stream laminae, which are flowing through the channel at different velocities governed by a parabolic flow profile. Separation takes place because of the differential occupancy of these laminae (see Figure 1). In flow FFF, the driving force originates with a secondary flow (termed a cross flow) of liquid moving perpendicularly to the main channel flow. Specifically, the cross flow stream is forced across the thin channel, and in the process it drives
S0022-3549(95)00335-2 CCC: $12.00
Journal of Pharmaceutical Sciences / 895 Vol. 85, No. 8, August 1996
+
+
entrained components into different mean positions in the parabolic flow profile so that separation will take place. To achieve this separative effect, the cross flow enters the thin channel by means of a frit enclosing the channel on one side and exits by means of a membrane (layered over a support frit) constituting the other channel wall (Figure 1). The liquid stream passes freely through the membrane, but the macromolecular components are held back due to the small size of the membrane pores. The rejected components form steadystate bands above the membrane whose thicknesses are determined by a balance between Brownian motion and the ongoing displacement of the cross flow. Because differentsized components end up with different mean elevations, they occupy different sets of stream laminae and they separate from one another in the primary channel flow. The process is illustrated in Figure 1. According to theory, the retention time tr of well-retained components in flow FFF is proportional to the hydrodynamic diameter dh of the individual component particles. The governing equation is12
tr )
πηw2(V˙ c/V˙ )dh 2kT
(1)
where η is solution viscosity, w is channel thickness, k is the Boltzmann constant, T is absolute temperature, and (V˙ c/V˙ ) is the ratio of the cross flow rate V˙ c to the channel flow rate V˙ . The latter flow rates are independently adjustable, giving great flexibility in achieving desired levels of speed and resolution. Equation 1 provides a relationship that makes it possible to determine particle diameter dh from measured retention times.6,13 For continuous distributions with a range of retention times, a size distribution curve can be obtained. Many kinds of membranes have been found suitable for use in flow FFF. The pore size of the membrane is usually chosen to be sufficiently small to reject (and thus allow the separation of) the smallest consitituent of the mixture of interest. In the present case, we choose a membrane with pores large enough to allow passage of the small proteins in the blood plasma but small enough to retain the larger proteins (e.g., immunoglobulins), high-density lipoproteins, and all larger colloids. The sensitivity of the flow FFF technique to larger (and scarcer) colloids is further enhanced by using a detector sensitive to light scattering; larger colloids scatter far more light per unit mass than smaller colloids in accordance with the principles of Rayleigh.14 Light-scattering detection is achieved here (as in most applications of FFF to colloids) by using a simple HPLC UV detector, designed for measuring UV absorption but equally effective at sensing small levels of light scattering (turbidity). While FFF has been used to characterize many colloidal mixtures, it has not previously been used in conjunction with membrane selectivity to “fish” dilute colloids out of a macromolecular soup. Previous applications of FFF to blood plasma either did not detect non-protein colloids15 or were limited to colloids (in the form of an emulsion) at high concentrations.16 An example of the separation of lipoproteins from blood plasma by flow FFF was given earlier, but membrane selectivity was not used.6
Experimental Section In order to determine the applicability of flow FFF (and particularly of membrane-selective flow FFF) to plasma colloids, a flow FFF
896 / Journal of Pharmaceutical Sciences Vol. 85, No. 8, August 1996
Figure 2sFlow FFF separation of (a) various proteins and vitamin B12 and (b) polystyrene latex beads of the indicated diameters: (a) a diameter (dh) scale is shown at the top based on eq 1; 10 µL injection of protein mixture into PBS carrier solution at pH 7.4; membrane is YM1, w ) 0.0163 cm, V˙ ) 0.7 mL/min, V˙ c ) 7.2 mL/min; detection at 280 nm; (b) 10 µL of latex beads injected into carrier liquid (5 mM Tris, 0.1 mM EDTA, 0.02% sodium azide, pH 7.4) at pH 7.4; V˙ ) 4.0 mL/min, V˙ c ) 1.0 mL/min; membrane is Amicon YM100, w ) 0.0210 cm; detection at 254 nm. Void time: (a) t0 ) 1.23; (b) t0 ) 0.28. channel system was assembled. The system utilizes a frit inlet to avoid flow interruption upon sample injection.17,18 The flow FFF channel is 28.5 cm in tip-to-tip length and 2 cm in breadth. The channel thickness (150-220 µm) is determined anew each time the membrane is replaced using the measured retention time of BSA (for which dh is known to be 7.0 nm) and eq 1. Auxiliary components of the flow FFF system included two SpectraPhysics Isochrom pumps (Spectra-Physics Inc., San Jose, CA), a septum injector, a Shimadzu SPD-6A UV detector (Shimadzu Corporation, Kyoto, Japan), and a PC compatible computer. The sample for each run consisted of 10 µL of plasma or other fluid mixtures injected by syringe. The membrane used is an Amicon YM regenerated cellulose hydrophilic membrane. Among the colloids added to blood plasma was amphotericin B colloidal dispersion19 (ABCD), prepared by reconstituting Amphocil (Liposome Technology, Inc., Menlo Park, CA) with preservative-free sterile water. The ABCD samples were prepared by diluting the ABCD (5 mg/mL) with tris-EDTA-NaN3 buffer or human plasma, ending with a 0.5 mg/mL concentration of ABCD. Another material added to blood plasma was intravenous fat emulsion (liposyn II 20%), obtained from Abbott Laboratories (North Chicago, IL) and mixed with phosphate buffer saline (PBS) or human plasma at a ratio of 1:100, giving a final concentration of 2 mg/mL. Polystyrene latex beads of 20, 54, 91, and 155 nm diameter (Duke Scientific, Palo Alto, CA) and various proteins (Sigma Chemical Company, St. Louis, MO) were used to standardize and to evaluate flow FFF resolution. The plasma samples were prepared following standard guidelines.20 PBS and tris-EDTA-NaN3 buffer were used as carrier liquids in the flow channel. The instrument was operated at room temperature, 23 ( 1 °C.
Results and Discussion The effectiveness of flow FFF in separating both proteins and larger colloids is illustrated in Figure 2: in Figure 2a, five protein constitutents (one a protein dimer) and vitamin B12 are separated, while in Figure 2b, four different-sized polystyrene (PS) latex microspheres up to 155 nm are fractionated. (Note that the cross flow rate V˙ c is adjusted 7.2 times higher in Figure 2a than Figure 2b to offset the almost
+
+
Figure 3sFlow FFF fractograms and size distribution curves of lipoproteins (with overlapping protein components) obtained using (a) normal flow FFF using a YM1 membrane and (b) membrane-selective flow FFF using a YM30 membrane. Both separations were carried out in PBS solution at pH 7.4 with V˙ ) 2.6 and V˙ c ) 10.9 mL/min; detection at 280 nm; void time t0 ) 0.32 min.
order of magnitude smaller diameter dh of proteins in order to keep the product V˙ cdh constant as suggested by eq 1.) Figure 3 illustrates the fractionation of human blood plasma into lipoprotein and protein components. An Amicon YM1 membrane is utilized in Figure 3a. This membrane, with a nominal cutoff of 1000 molecular weight, has pores so small that no proteins can pass. With 10 µL of injected plasma, the proteins overlap to form a single large peak that masks the smaller high-density lipoprotein (HDL) peak and strongly overlaps with the low-density lipoprotein (LDL) peak. When a YM30 membrane is used, as shown in Figure 3b, the pores are sufficiently large that most of the proteins are swept through the membrane by the cross flow. (While the nominal molecular weight cutoff for the YM30 membrane is 30 000 daltons (Da), we have found that albumin, at 67 000 Da, is almost entirely flushed through the YM 30 membrane in a flow FFF system at high cross flow rates.) The protein peak still overlaps the HDL peak because the larger proteins and the smaller HDL particles both have dh in the range 7-15 nm. However, the merged protein and HDL peak is now much better separated from the LDL peak. The very low density lipoprotein (VLDL) peak, although low and broad, is well separated from the LDL peak. A diameter (dh) scale calculated from eq 1 is shown on the upper horizontal axis in Figure 3. The diameter scale serves to convert the fractograms (signal-versus-time plots) into turbidity-weighted particle size distribution curves for the lipoproteins. The HDL, LDL, and VLDL peaks appearing in Figure 3 were confirmed using lipoprotein standards obtained from an ultracentrifugation procedure. Figure 4 illustrates the measurement and characterization of low levels of (a) liposyn II intravenous emulsion added at 2 mg per mL of plasma and (b) ABCD (designed as a carrier for the drug amphotericin B) present in plasma at only 0.5 mg/mL. Membrane-selective flow FFF was used in both Figure 4a and Figure 4b by employing YM30 and YM100 membranes (with nominal 100 000 Da molecular weight cutoff), respectively. The second curve from the top in these two figures shows that the peak representing the colloid (emulsion or ABCD) alone closely resembles that found in the presence of plasma. In fact, the bottom curve in each figure shows that the plasma background is negligible in the size range of interest, even though the lipoprotein content in plasma is 3-4 times higher than the emulsion content in Figure 4a and 12-14 times higher than the ABCD content in Figure 4b. The reason the lipoproteins do not significantly
Figure 4sFlow FFF fractograms and size distribution curves of (a) liposyn II intravenous fat emulsion and (b) ABCD. For both colloids, the upper curve represents the diluted colloid (at the concentration level stated) in 10 µL of human blood plasma, the middle curve was obtained for an equal amount of colloid without plasma, and the bottom curve represents the plasma background. Conditions: (a) carrier liquid is PBS, pH 7.4, V˙ ) 2.5, V˙ c ) 0.53 mL/min, YM30 membrane, detection at 254 nm, t0 ) 0.32 min; (b) carrier liquid is same as in Figure 2b; V˙ ) 4.3, V˙ c ) 1.5 mL/min, YM100 membrane, detection at 405 nm, t0 ) 0.16 min.
interfere with the measurement of the added colloids is that there is little overlap in the two size ranges which are separated by two fractionation mechanisms, flow FFF and membrane selection. Specifically, the emulsion droplets, dh ) 150-400 nm, are completely separated from the VLDL particles with dh ) 30-80 nm. There is, however, a small overlap of ABCD particles with the rather low VLDL peak since for the former dh ) 30-140 nm. The signals for the dilute emulsion and ABCD colloid in Figure 4 are sufficiently prominent that one could detect significantly lower levels, probably down to ∼0.1 mg/mL, particularly with background subtraction. This level is almost 3 orders of magnitude below the protein content. The excellent signal is partly a result of the enhanced light scattering (turbidity) of colloidal particles larger than proteins. However, the separation of these colloids into well-defined size classes by flow FFF is essential to the measurement. The effectiveness of the separation is further improved by membrane selectivity, which requires a membrane that will retain the colloids of interest while allowing smaller proteins to be purged from the system. Information on the reproducibility and recovery of proteins and lipoproteins will be reported in a future publication. With further development of the technique, the potential exists to examine changes in the particle size distribution of emulsions and liposomes in ex-vivo blood samples. However, the question of how blood cells in the ex-vivo samples may affect colloid detection needs to be considered because some interference with small-size colloidal particles by the blood cells may develop due to the steric effect. Because steric inversion is related to particle size, channel dimensions, and channel flow and cross flow rates, the size range of the particles subject to inference can be adjusted or controlled to some extent. In the analysis of particles in the submicron
Journal of Pharmaceutical Sciences / 897 Vol. 85, No. 8, August 1996
+
+
range, such as found in the ABCD and liposyn samples, suitable conditions can be sought to elute the red blood cells at or close to the void time to eliminate overlap with the colloidal particles. However, considerable work might be needed to identify suitable conditions because of the complexity of the steric effect for blood cells.
References and Notes 1. Nelson, G. L. Blood Lipids and Lipoproteins: Quantitation, Composition and Metabolism; R. E. Krieger: Huntingdon, NY, 1979. 2. Vance, D. E.; Vance, J. E. Biochemistry of Lipids and Membranes; Benjamin/Cummings: Menlo Park, CA, 1985. 3. Vander, A. J. Human Physiology: The Mechanisms of Body Function; McGraw-Hill: New York, 1990. 4. Rifai, N.; Warnick, G. R.; McNamara, J. R.; Belcher, J. D.; Grinstead, G. F.; Frantz, I. D., Jr. Clin. Chem. 1992, 38, 150160. 5. Warnick, G. R.; Dominiczak, M. H. Curr. Opin. Lipidol. 1990, 1, 493-499. 6. Giddings, J. C. Science 1993, 260, 1456-1465. 7. Ratanathanawongs, S. K.; Lee, I.; Giddings, J. C. In Particle Size Distribution II: Assessment and Characterization; Provder, T., Ed.; ACS Symposium Series 472; American Chemical Society: Washington, DC, 1991; pp 229-246. 8. Giddings, J. C.; Caldwell, K. D. In Physical Methods of Chemistry; Rossiter, B. W., Hamilton, J. F., Eds.; John Wiley: New York, 1989; Vol. 3B, pp 867-938. 9. Martin, M.; Williams, P. S. In Theoretical Advancement in Chromatography and Related Separation Techniques; Dondi, F.,
898 / Journal of Pharmaceutical Sciences Vol. 85, No. 8, August 1996
10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
Guiochon, G., Eds.; NATO ASI Series C: Mathematical and Physical Sciences, Vol. 383; Kluwer: Dordrecht, 1992; pp 513580. Beckett, R.; Nicholson, G.; Hotchin, D. M.; Hart, B. T. Hydrobiologia 1992, 235/236, 697-710. Levin, S. Biomed. Chromatogr. 1991, 5, 133-137. Liu, M.-K.; Giddings, J. C. Macromolecules 1993, 26, 3576-3588. Giddings, J. C. Anal. Chem. 1995, 67 592A-598A. Chu, B. Laser Light Scattering: Basic Principles and Practice, 2nd ed.; Academic Press: Boston, 1991. Giddings, J. C.; Yang, F. J.; Myers, M. N. Anal. Biochem. 1977, 81, 395-407. Li, J.; Caldwell, K. D.; Anderson, B. D. Pharm. Res. 1993, 10, 535-541. Giddings, J. C.; Benincasa, M. A.; Liu, M.-K.; Li, P. J. Liq. Chromatogr. 1992, 15, 1729-1747. Liu, M.-K.; Li, P.; Giddings, J. C. Protein Sci. 1993, 2, 15201531. Guo, L. S. S.; Fielding, R. M.; Lasic, D. D.; Hamilton, R. L.; Mufson, D. Int. J. Pharm. 1991, 75, 45-54. Manual of Laboratory Operations, Lipid Research Clinics Program. Lipid and Lipoprotein Analysis; DHEW Publication No. (NIH)75-628, Vol. 1; U.S. Dept. of HEW: Washington, DC, 1974.
Acknowledgments This work was supported by Public Health Service Grant GM1085138 from the National Institutes of Health.
JS950335S
+
Isolation and Measurement of Colloids in Human Plasma by Membrane-Selective Flow Field-Flow Fractionation: Lipoproteins and Pharmaceutical Colloids PING LI
AND
J. CALVIN GIDDINGSX
Received August 7, 1995, from the Field-Flow Fractionation Research Center, Department of Chemistry, University of Utah, Salt Lake City, UT 84112. Final revised manuscript received April 18, 1996 . Accepted for publication April 18, 1996X. Abstract 0 Using a modified flow field-flow fractionation (flow FFF) technique termed membrane-selective flow FFF, a capability is developed to isolate various colloidal constituents that are naturally present in or may be introduced into blood plasma. Once isolated, the colloids may be subject to quantitative measurement to provide relative amounts of the different constituents and the size distribution curve of each. The potential of the technique has been demonstrated by isolating and measuring (a) the lipoprotein fractions found in human blood plasma, (b) liposyn II pharmaceutical emulsion added at 2 mg/mL of plasma, and (c) amphotericin B colloidal dispersion (a drug delivery agent) added to plasma at 0.5 mg/mL.
Blood plasma (blood without cells) is a complex macromolecular-colloidal solution containing many components, principally numerous proteins and various classes and subclasses of lipoproteins.1-3 The high abundance of proteins (60-80 mg/mL of plasma) generally interferes with the measurement of other colloidal constituents, including lipoproteins (6-8 mg/mL) and various pharmaceutical colloids (such as liposomes and emulsions) that may be introduced into the blood stream as carriers of drugs or nutritional factors. The latter colloids, when present, require measurement because they significantly influence biological function and health status despite their fleeting existence and low concentration. Because of mutual interference, it is difficult to measure the concentrations of specific plasma colloids; it is especially difficult to measure distributions in their properties such as size and mass distributions. Most techniques for isolating subpopulations of colloids are laborious, lack good resolution, and require multiple experimental steps. Membrane techniques, for example, could be used to remove albumin and other low molecular weight proteins, but after this step a complex and intransigent mixture of colloids (large proteins and lipoproteins) remains. Ultracentrifugation can be used to isolate the lipoproteins, but the procedure consists of several stages and requires over 1 day for completion.4,5 We report here a technique based on flow field-flow fractionation (flow FFF) that is capable, in a single step, of isolating different size classes of plasma colloids >7 nm in diameter and measuring their size distribution. The procedure requires only 10-40 min and could probably be made significantly faster. The technique relies primarily on the resolving power of flow FFF to separate macromolecular and colloidal materials. The flow FFF process is assisted by membrane selectivity with no additional experimental steps or complications. Normally, the large abundance of small proteins hinders the flow FFF separation and detection of plasma colloids. As a remedy, our technique employs a membrane process that X
Abstract published in Advance ACS Abstracts, June 1, 1996.
© 1996, American Chemical Society and American Pharmaceutical Association
Figure 1sSeparation of three components, A, B, and C, along the main flow axis of a flow FFF channel. A secondary flow (the cross flow) drives the three components into cloud-like bands adjacent to the membrane. The clouds have different mean elevations because the constituent particles have different hydrodynamic diameters and thus different diffusion coefficients. The different elevations within the parabolic axial flow profile lead to differential displacement and thus separation. In membrane-selective flow FFF, the smaller particles (e.g., those in the C band) are gradually flushed through the membrane and the underlying support frit so as to avoid interference with bands A and B. Diagram shows edge view of ribbon-like channel.
purges most proteins <7 nm in diameter from the colloidal suspension. We term this technique membrane-selective flow FFF. It is a simple matter to combine flow FFF and membrane separation because flow FFF systems generally incorporate a membrane as a critical functional element.6,7 Perhaps surprisingly, the incorporated membrane is used for the retention but not the separation of macromolecular constituents. However, since blood plasma has such overwhelming concentrations of small proteins, the flow FFF membrane can be put to use to remove the bulk of these interfering proteins without lengthening or complicating the one-step procedure of flow FFF. Flow FFF is one member of the field-flow fractionation (FFF) family of techniques.6-11 The sister techniques of sedimentation FFF and thermal FFF (neither employing a membrane) are also used for separating a wide range of colloids and polymers. Although the FFF techniques have been described widely in the literature,6-11 a brief account will be given to provide proper context for this work. FFF is a family of chromatographic-like elution techniques applicable to macromolecules, colloids, and cell-sized particles. Separation is achieved in a thin ribbon-like flow channel. A field is applied across the channel in a direction perpendicular to flow. The field generates a driving force that induces unlike particles to occupy different stream laminae, which are flowing through the channel at different velocities governed by a parabolic flow profile. Separation takes place because of the differential occupancy of these laminae (see Figure 1). In flow FFF, the driving force originates with a secondary flow (termed a cross flow) of liquid moving perpendicularly to the main channel flow. Specifically, the cross flow stream is forced across the thin channel, and in the process it drives
S0022-3549(95)00335-2 CCC: $12.00
Journal of Pharmaceutical Sciences / 895 Vol. 85, No. 8, August 1996
+
+
entrained components into different mean positions in the parabolic flow profile so that separation will take place. To achieve this separative effect, the cross flow enters the thin channel by means of a frit enclosing the channel on one side and exits by means of a membrane (layered over a support frit) constituting the other channel wall (Figure 1). The liquid stream passes freely through the membrane, but the macromolecular components are held back due to the small size of the membrane pores. The rejected components form steadystate bands above the membrane whose thicknesses are determined by a balance between Brownian motion and the ongoing displacement of the cross flow. Because differentsized components end up with different mean elevations, they occupy different sets of stream laminae and they separate from one another in the primary channel flow. The process is illustrated in Figure 1. According to theory, the retention time tr of well-retained components in flow FFF is proportional to the hydrodynamic diameter dh of the individual component particles. The governing equation is12
tr )
πηw2(V˙ c/V˙ )dh 2kT
(1)
where η is solution viscosity, w is channel thickness, k is the Boltzmann constant, T is absolute temperature, and (V˙ c/V˙ ) is the ratio of the cross flow rate V˙ c to the channel flow rate V˙ . The latter flow rates are independently adjustable, giving great flexibility in achieving desired levels of speed and resolution. Equation 1 provides a relationship that makes it possible to determine particle diameter dh from measured retention times.6,13 For continuous distributions with a range of retention times, a size distribution curve can be obtained. Many kinds of membranes have been found suitable for use in flow FFF. The pore size of the membrane is usually chosen to be sufficiently small to reject (and thus allow the separation of) the smallest consitituent of the mixture of interest. In the present case, we choose a membrane with pores large enough to allow passage of the small proteins in the blood plasma but small enough to retain the larger proteins (e.g., immunoglobulins), high-density lipoproteins, and all larger colloids. The sensitivity of the flow FFF technique to larger (and scarcer) colloids is further enhanced by using a detector sensitive to light scattering; larger colloids scatter far more light per unit mass than smaller colloids in accordance with the principles of Rayleigh.14 Light-scattering detection is achieved here (as in most applications of FFF to colloids) by using a simple HPLC UV detector, designed for measuring UV absorption but equally effective at sensing small levels of light scattering (turbidity). While FFF has been used to characterize many colloidal mixtures, it has not previously been used in conjunction with membrane selectivity to “fish” dilute colloids out of a macromolecular soup. Previous applications of FFF to blood plasma either did not detect non-protein colloids15 or were limited to colloids (in the form of an emulsion) at high concentrations.16 An example of the separation of lipoproteins from blood plasma by flow FFF was given earlier, but membrane selectivity was not used.6
Experimental Section In order to determine the applicability of flow FFF (and particularly of membrane-selective flow FFF) to plasma colloids, a flow FFF
896 / Journal of Pharmaceutical Sciences Vol. 85, No. 8, August 1996
Figure 2sFlow FFF separation of (a) various proteins and vitamin B12 and (b) polystyrene latex beads of the indicated diameters: (a) a diameter (dh) scale is shown at the top based on eq 1; 10 µL injection of protein mixture into PBS carrier solution at pH 7.4; membrane is YM1, w ) 0.0163 cm, V˙ ) 0.7 mL/min, V˙ c ) 7.2 mL/min; detection at 280 nm; (b) 10 µL of latex beads injected into carrier liquid (5 mM Tris, 0.1 mM EDTA, 0.02% sodium azide, pH 7.4) at pH 7.4; V˙ ) 4.0 mL/min, V˙ c ) 1.0 mL/min; membrane is Amicon YM100, w ) 0.0210 cm; detection at 254 nm. Void time: (a) t0 ) 1.23; (b) t0 ) 0.28. channel system was assembled. The system utilizes a frit inlet to avoid flow interruption upon sample injection.17,18 The flow FFF channel is 28.5 cm in tip-to-tip length and 2 cm in breadth. The channel thickness (150-220 µm) is determined anew each time the membrane is replaced using the measured retention time of BSA (for which dh is known to be 7.0 nm) and eq 1. Auxiliary components of the flow FFF system included two SpectraPhysics Isochrom pumps (Spectra-Physics Inc., San Jose, CA), a septum injector, a Shimadzu SPD-6A UV detector (Shimadzu Corporation, Kyoto, Japan), and a PC compatible computer. The sample for each run consisted of 10 µL of plasma or other fluid mixtures injected by syringe. The membrane used is an Amicon YM regenerated cellulose hydrophilic membrane. Among the colloids added to blood plasma was amphotericin B colloidal dispersion19 (ABCD), prepared by reconstituting Amphocil (Liposome Technology, Inc., Menlo Park, CA) with preservative-free sterile water. The ABCD samples were prepared by diluting the ABCD (5 mg/mL) with tris-EDTA-NaN3 buffer or human plasma, ending with a 0.5 mg/mL concentration of ABCD. Another material added to blood plasma was intravenous fat emulsion (liposyn II 20%), obtained from Abbott Laboratories (North Chicago, IL) and mixed with phosphate buffer saline (PBS) or human plasma at a ratio of 1:100, giving a final concentration of 2 mg/mL. Polystyrene latex beads of 20, 54, 91, and 155 nm diameter (Duke Scientific, Palo Alto, CA) and various proteins (Sigma Chemical Company, St. Louis, MO) were used to standardize and to evaluate flow FFF resolution. The plasma samples were prepared following standard guidelines.20 PBS and tris-EDTA-NaN3 buffer were used as carrier liquids in the flow channel. The instrument was operated at room temperature, 23 ( 1 °C.
Results and Discussion The effectiveness of flow FFF in separating both proteins and larger colloids is illustrated in Figure 2: in Figure 2a, five protein constitutents (one a protein dimer) and vitamin B12 are separated, while in Figure 2b, four different-sized polystyrene (PS) latex microspheres up to 155 nm are fractionated. (Note that the cross flow rate V˙ c is adjusted 7.2 times higher in Figure 2a than Figure 2b to offset the almost
+
+
Figure 3sFlow FFF fractograms and size distribution curves of lipoproteins (with overlapping protein components) obtained using (a) normal flow FFF using a YM1 membrane and (b) membrane-selective flow FFF using a YM30 membrane. Both separations were carried out in PBS solution at pH 7.4 with V˙ ) 2.6 and V˙ c ) 10.9 mL/min; detection at 280 nm; void time t0 ) 0.32 min.
order of magnitude smaller diameter dh of proteins in order to keep the product V˙ cdh constant as suggested by eq 1.) Figure 3 illustrates the fractionation of human blood plasma into lipoprotein and protein components. An Amicon YM1 membrane is utilized in Figure 3a. This membrane, with a nominal cutoff of 1000 molecular weight, has pores so small that no proteins can pass. With 10 µL of injected plasma, the proteins overlap to form a single large peak that masks the smaller high-density lipoprotein (HDL) peak and strongly overlaps with the low-density lipoprotein (LDL) peak. When a YM30 membrane is used, as shown in Figure 3b, the pores are sufficiently large that most of the proteins are swept through the membrane by the cross flow. (While the nominal molecular weight cutoff for the YM30 membrane is 30 000 daltons (Da), we have found that albumin, at 67 000 Da, is almost entirely flushed through the YM 30 membrane in a flow FFF system at high cross flow rates.) The protein peak still overlaps the HDL peak because the larger proteins and the smaller HDL particles both have dh in the range 7-15 nm. However, the merged protein and HDL peak is now much better separated from the LDL peak. The very low density lipoprotein (VLDL) peak, although low and broad, is well separated from the LDL peak. A diameter (dh) scale calculated from eq 1 is shown on the upper horizontal axis in Figure 3. The diameter scale serves to convert the fractograms (signal-versus-time plots) into turbidity-weighted particle size distribution curves for the lipoproteins. The HDL, LDL, and VLDL peaks appearing in Figure 3 were confirmed using lipoprotein standards obtained from an ultracentrifugation procedure. Figure 4 illustrates the measurement and characterization of low levels of (a) liposyn II intravenous emulsion added at 2 mg per mL of plasma and (b) ABCD (designed as a carrier for the drug amphotericin B) present in plasma at only 0.5 mg/mL. Membrane-selective flow FFF was used in both Figure 4a and Figure 4b by employing YM30 and YM100 membranes (with nominal 100 000 Da molecular weight cutoff), respectively. The second curve from the top in these two figures shows that the peak representing the colloid (emulsion or ABCD) alone closely resembles that found in the presence of plasma. In fact, the bottom curve in each figure shows that the plasma background is negligible in the size range of interest, even though the lipoprotein content in plasma is 3-4 times higher than the emulsion content in Figure 4a and 12-14 times higher than the ABCD content in Figure 4b. The reason the lipoproteins do not significantly
Figure 4sFlow FFF fractograms and size distribution curves of (a) liposyn II intravenous fat emulsion and (b) ABCD. For both colloids, the upper curve represents the diluted colloid (at the concentration level stated) in 10 µL of human blood plasma, the middle curve was obtained for an equal amount of colloid without plasma, and the bottom curve represents the plasma background. Conditions: (a) carrier liquid is PBS, pH 7.4, V˙ ) 2.5, V˙ c ) 0.53 mL/min, YM30 membrane, detection at 254 nm, t0 ) 0.32 min; (b) carrier liquid is same as in Figure 2b; V˙ ) 4.3, V˙ c ) 1.5 mL/min, YM100 membrane, detection at 405 nm, t0 ) 0.16 min.
interfere with the measurement of the added colloids is that there is little overlap in the two size ranges which are separated by two fractionation mechanisms, flow FFF and membrane selection. Specifically, the emulsion droplets, dh ) 150-400 nm, are completely separated from the VLDL particles with dh ) 30-80 nm. There is, however, a small overlap of ABCD particles with the rather low VLDL peak since for the former dh ) 30-140 nm. The signals for the dilute emulsion and ABCD colloid in Figure 4 are sufficiently prominent that one could detect significantly lower levels, probably down to ∼0.1 mg/mL, particularly with background subtraction. This level is almost 3 orders of magnitude below the protein content. The excellent signal is partly a result of the enhanced light scattering (turbidity) of colloidal particles larger than proteins. However, the separation of these colloids into well-defined size classes by flow FFF is essential to the measurement. The effectiveness of the separation is further improved by membrane selectivity, which requires a membrane that will retain the colloids of interest while allowing smaller proteins to be purged from the system. Information on the reproducibility and recovery of proteins and lipoproteins will be reported in a future publication. With further development of the technique, the potential exists to examine changes in the particle size distribution of emulsions and liposomes in ex-vivo blood samples. However, the question of how blood cells in the ex-vivo samples may affect colloid detection needs to be considered because some interference with small-size colloidal particles by the blood cells may develop due to the steric effect. Because steric inversion is related to particle size, channel dimensions, and channel flow and cross flow rates, the size range of the particles subject to inference can be adjusted or controlled to some extent. In the analysis of particles in the submicron
Journal of Pharmaceutical Sciences / 897 Vol. 85, No. 8, August 1996
+
+
range, such as found in the ABCD and liposyn samples, suitable conditions can be sought to elute the red blood cells at or close to the void time to eliminate overlap with the colloidal particles. However, considerable work might be needed to identify suitable conditions because of the complexity of the steric effect for blood cells.
References and Notes 1. Nelson, G. L. Blood Lipids and Lipoproteins: Quantitation, Composition and Metabolism; R. E. Krieger: Huntingdon, NY, 1979. 2. Vance, D. E.; Vance, J. E. Biochemistry of Lipids and Membranes; Benjamin/Cummings: Menlo Park, CA, 1985. 3. Vander, A. J. Human Physiology: The Mechanisms of Body Function; McGraw-Hill: New York, 1990. 4. Rifai, N.; Warnick, G. R.; McNamara, J. R.; Belcher, J. D.; Grinstead, G. F.; Frantz, I. D., Jr. Clin. Chem. 1992, 38, 150160. 5. Warnick, G. R.; Dominiczak, M. H. Curr. Opin. Lipidol. 1990, 1, 493-499. 6. Giddings, J. C. Science 1993, 260, 1456-1465. 7. Ratanathanawongs, S. K.; Lee, I.; Giddings, J. C. In Particle Size Distribution II: Assessment and Characterization; Provder, T., Ed.; ACS Symposium Series 472; American Chemical Society: Washington, DC, 1991; pp 229-246. 8. Giddings, J. C.; Caldwell, K. D. In Physical Methods of Chemistry; Rossiter, B. W., Hamilton, J. F., Eds.; John Wiley: New York, 1989; Vol. 3B, pp 867-938. 9. Martin, M.; Williams, P. S. In Theoretical Advancement in Chromatography and Related Separation Techniques; Dondi, F.,
898 / Journal of Pharmaceutical Sciences Vol. 85, No. 8, August 1996
10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.
Guiochon, G., Eds.; NATO ASI Series C: Mathematical and Physical Sciences, Vol. 383; Kluwer: Dordrecht, 1992; pp 513580. Beckett, R.; Nicholson, G.; Hotchin, D. M.; Hart, B. T. Hydrobiologia 1992, 235/236, 697-710. Levin, S. Biomed. Chromatogr. 1991, 5, 133-137. Liu, M.-K.; Giddings, J. C. Macromolecules 1993, 26, 3576-3588. Giddings, J. C. Anal. Chem. 1995, 67 592A-598A. Chu, B. Laser Light Scattering: Basic Principles and Practice, 2nd ed.; Academic Press: Boston, 1991. Giddings, J. C.; Yang, F. J.; Myers, M. N. Anal. Biochem. 1977, 81, 395-407. Li, J.; Caldwell, K. D.; Anderson, B. D. Pharm. Res. 1993, 10, 535-541. Giddings, J. C.; Benincasa, M. A.; Liu, M.-K.; Li, P. J. Liq. Chromatogr. 1992, 15, 1729-1747. Liu, M.-K.; Li, P.; Giddings, J. C. Protein Sci. 1993, 2, 15201531. Guo, L. S. S.; Fielding, R. M.; Lasic, D. D.; Hamilton, R. L.; Mufson, D. Int. J. Pharm. 1991, 75, 45-54. Manual of Laboratory Operations, Lipid Research Clinics Program. Lipid and Lipoprotein Analysis; DHEW Publication No. (NIH)75-628, Vol. 1; U.S. Dept. of HEW: Washington, DC, 1974.
Acknowledgments This work was supported by Public Health Service Grant GM1085138 from the National Institutes of Health.
JS950335S