Development of Sandwich HPLC Microcolumns for Analyte Adsorption

A new class of columns is reported that uses only microgram quantities of active support and that provides for the retention of biological compounds a...
0 downloads 0 Views 171KB Size
Download PDF
Anal. Chem. 2001, 73, 1366-1373

Development of Sandwich HPLC Microcolumns for Analyte Adsorption on the Millisecond Time Scale William Clarke and David S. Hage*

Department of Chemistry, University of Nebraska, Lincoln, Nebraska 68588-0304

A new class of columns is reported that uses only microgram quantities of active support and that provides for the retention of biological compounds and other analytes on the millisecond time scale. This was accomplished by packing standard HPLC supports into layers as small as 60 µm in length and using only 90 µg of support material. This provided columns with effective residence times in the millisecond time range when routine HPLC flow rates and pressures were used. The retention of analytes by such columns was examined under both adsorption- and diffusion-limited conditions. The RPLC adsorption of hemoglobin (a system with diffusion-limited retention) was found to give 95% binding in as little as 4 ms. The adsorption of fluorescein by an anti-fluorescein antibody column (an adsorption-limited system) gave 95% retention in 100-120 ms. One application examined for these columns was their use in a chromatographic-based competitive binding immunoassay. This used bovine serum albumin (BSA) as the model analyte, and fluorescein-labeled BSA was used for detection. The resulting approach had a contact time of 180 ms between the sample and an anti-BSA immunoaffinity microcolumn and provided a signal within 5-25 s after sample injection. The columns developed in this work should also be useful in other situations that involve a small amount of a stationary phase or that require short column residence times. One topic of continuing interest in analytical chemistry is the development of smaller and faster methods of analysis. Examples in the field of separations include the production of microfabricated devices, as well as ongoing work in microbore HPLC and capillary electrochromatography.1-4 These last two areas generally use a decrease in column diameter with an accompanying increase in column length to produce more efficient chromatographic systems for the analysis of complex samples. However, another way in which column size can be decreased is through a reduction in length. This is possible in separations that do not require high * Corresponding author: (fax) 402-472-9402; (e-mail) [email protected]. (1) Afeyan, N. B.; Gordon, N. F.; Mazsaroff, I.; Varady, L.; Fulton, S. P.; Yang, Y. B.; Regnier, F. E. J. Chromatogr. 1990, 519, 1. (2) Novotny, M. V.; Ishii, D. Microcolumn Separations: Columns, Instrumentation, and Ancillary Techniques; Elsevier: New York, 1985. (3) Vonk, N.; Verstraeten, W. P.; Marinissen, J. W. J. Chromatogr. Sci. 1992, 30, 296. (4) Yang, F. J. Microbore Column Chromatography: A Unified Approach to Chromatography; Dekker: New York, 1989.

1366 Analytical Chemistry, Vol. 73, No. 6, March 15, 2001

efficiencies, such as those based on simple adsorption/desorption mechanisms. Examples include work that has been performed with immunoaffinity supports5 and studies involving the use of membrane supports in HPLC.6-10 One advantage of using a short column is that this requires only a small amount of stationary phase. This can be especially valuable when dealing with expensive or unique types of stationary phases, such as might be encountered in immunoaffinity chromatography or other affinity methods. Another advantage is the small sample residence time that can be obtained with short columns at even low-to-moderate operating pressures. This is illustrated in Figure 1, which shows how the void time for an HPLC column (packed with porous silica) will change as the column length and flow rate are varied. The column void time is used here to represent the minimum residence time that can be obtained for any analyte when no size exclusion effects are present. It can be seen from Figure 1 that the shortest possible analysis time for a 2 mm i.d. × 10 cm HPLC column will be on the order of a few minutes to a few seconds at flow rates up to 10 mL/min. Proportionally shorter analysis times are possible by decreasing the column length to 1 cm, but it is only when working with column lengths of 100 µm to 1 mm that millisecond analysis times become feasible at standard HPLC flow rates of 0.1-1.0 mL/min. The goal of this study is to develop a new class of HPLC columns that can work with small amounts of stationary phase and that can be used to perform affinity chromatography or other adsorption/desorption-based separations in the millisecond time domain. This will be done by using a sandwich packing method in which HPLC supports are placed into columns with effective stationary-phase layers that are 1 mm or less in length. The reproducibility of this packing method will be examined, along with the flow rates and sample residence times that can be employed with these columns. The binding of several model analytes to these columns will be evaluated under various kinetic conditions, including the retention of compounds by immunoaffinity chromatography and the adsorption of proteins to reversed(5) de Alwis, U.; Wilson, G. S. Anal. Chem. 1987, 59, 2786. (6) Hagen, D. F.; Markell, C. G.; Schmitt, G. A. Anal. Chim. Acta 1990, 236, 157. (7) Fernando, W. P. N.; Larrivee, M. L.; Poole, C. F. Anal. Chem. 1993, 65, 588. (8) Dombrowski, T. R.; Wilson, G. S.; Thurman, E. M. Anal. Chem. 1998, 70, 1969. (9) Zhou, D.; Zou, H.; Ni, J.; Yang, L.; Jia, L.; Zhang, Q.; Zhang, Y. Anal. Chem. 1999, 71, 115. (10) Podgornik, A.; Barut, M.; Jancar, J.; Strancar, A. Anal. Chem. 1999, 71, 2986. 10.1021/ac000870z CCC: $20.00

© 2001 American Chemical Society Published on Web 02/13/2001

Figure 1. Change in column void time with column length and solvent flow rate for 2-mm-i.d. HPLC columns packed with porous silica. These results assume an overall porosity of 0.80 within the column (i.e., 80% of the column volume is occupied by the mobile phase). Using a column with an inner diameter of 1 or 4 mm gives similar results but with the vertical position of the lines in this graph being lowered or raised by 4-fold, respectively.

phase media. The use of these columns in a chromatographicbased competitive binding immunoassay will also be considered. EXPERIMENTAL SECTION Reagents. The bovine hemoglobin, bovine serum albumin (BSA), fluorescein isothiocyanate (FITC)-labeled BSA, mouse antiBSA antibodies, and mouse anti-fluorescein antibodies were from Sigma (St. Louis, MO). Sodium fluorescein was obtained from Matheson, Coleman, and Bell (Cincinnati, OH). Reagents for the bicinchoninic acid (BCA) protein assay were purchased from Pierce (Rockford, IL). HPLC-grade Nucleosil Si-1000 (7-µm particle diameter, 1000-Å pore size) was obtained from Alltech (Deerfield, IL). All other chemicals were reagent grade or better and used without further purification. All aqueous solutions were prepared with deionized water from a Nanopure system (Barnstead, Dubuque, IA). Apparatus. All columns were packed using a CM3200 pump from Thermoseparations (Riviera Beach, FL) and a modified Valco N60 six-port valve (Houston, TX). Prior to packing, the support slurries were kept in suspension by a Thermolyne Rotomix 50800 mixer (Dubuque, IA). Samples for the BCA protein assay and turbidity measurements were analyzed with a Shimadzu UV160U absorbance spectrophotometer (Kyoto, Japan). Chromatographic studies were conducted using a CM3200 pump, a model 713 autosampler from Thermoseparations, and a Rheodyne 7126 sixport injection valve (Cotati, CA). An LDC SM3100X absorbance detector was used to monitor hemoglobin during the reversedphase studies. The adsorption of fluorescein and elution of FITClabeled BSA were examined with a Shimadzu RF-535 fluorescence detector. Microcolumn Preparation. Figure 2 shows a general schematic of the sandwich microcolumns that were used in this report. The outer casing and end fittings for these columns were the same as described in a previous report;11 however, other designs for these exterior components could also be employed. The interior (11) Walters, R. R. Anal. Chem. 1983, 55, 591.

of the column consisted of alternating layers of a nonadsorptive support and a support that contained the stationary phase of interest. The stationary phases that were used in this report for the central, active layer were either immobilized antibodies or a C18 reversed-phase material. The nonadsorptive support used for the upper and lower layers was diol-bonded silica, which was selected since it has little or no interaction with proteins and most biologically related analytes under the aqueous conditions that were employed in this work for sample application.12,13 The diol-bonded, aldehyde-activated, and immobilized protein supports were prepared from Nucleosil Si-1000, as described previously.12,13 The reversed-phase support was Nucleosil C18 Si100 (5-µm particle diameter, 100-Å pore size), which was purchased from Alltech. The microcolumns were packed with a calibrated 150-µL injection loop and a support slurry of known concentration. The upper and lower layers in the microcolumn were prepared using a 2 mg/mL slurry of diol-bonded silica in the packing solvent (pH 7.0, 0.10 M potassium phosphate buffer). The immunoaffinity and RPLC slurries were prepared in the same packing solvent at concentrations selected by using eq 2 (see Results and Discussion) along with the known packing densities of these materials and the final desired thickness of each active support layer. The concentration of each slurry was determined prior to packing by using turbidity measurements at 800 nm versus standard solutions that contained known concentrations of the support in the packing solvent. The bottom layer of diol-bonded silica in each microcolumn was applied in five 150-µL injections at 3 mL/min; this was followed by an increase in flow rate to 5 mL/min for 5 min to stabilize the packing in this layer. The flow rate was then returned to 3 mL/ min and a sufficient number of injections (as determined from eq 2) was made of the active support slurry to provide the required thickness of this support within the column. After this layer was stabilized at 5 mL/min for 5 min, the flow rate was readjusted to 3 mL/min and a sufficient amount of diol-bonded silica was added to fill the remainder of the column bed. After the column bed had been completely filled, the flow rate was increased to apply a back pressure of 3000-4000 psi across the column for at least 30 min. This pressure was then gradually released, the column was removed from the packing system, and a frit and end fitting were added to the column’s open end. Each column was then stored in pH 7.0, 0.10 M phosphate buffer until further use. The permeability studies were performed in triplicate using columns packed with diol-bonded silica of various pore sizes. The back pressure of these columns was measured as a function of flow rate between 0.0 and 3.0 mL/min. At each new flow rate, the system was allowed to stabilize before the back pressure was measured. Packing reproducibility was examined by optical microscopy and by performing protein assays on microcolumns containing immobilized hemoglobin silica. After these columns were packed, their contents were removed and examined under a microscope or placed into a fixed volume of pH 7.0, 0.10 M phosphate buffer. The hemoglobin content of this latter suspen(12) Larsson, P.-O. Methods Enzymol. 1984, 104, 2121. (13) Ruhn, P. F.; Garver, S.; Hage, D. S. Anal. Chem. 1994, 66, 9.

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

1367

Figure 2. General schematic of a sandwich microcolumn. The outer components of this column are described in ref 11. The construction of the inner portion of the column is described in the Experimental Section.

sion was determined by a BCA protein assay.14 After the color of this assay was allowed to develop for 60 min, the support was removed from the reaction slurry by using a 0.45-µm syringe filter and the absorbance of the filtrate was measured at 562 nm. Chromatographic Studies. All studies were performed at room temperature. The reversed-phase retention of hemoglobin was studied using a 2.1 mm i.d. × 1.0 cm column which contained a 1.1-mm layer of C18 Nucleosil Si-100; diol-bonded silica was used to fill the remainder of the column housing. A series of 10-µL injections of a 2 mg/mL hemoglobin solution were made onto this column in pH 7.0, 0.10 M phosphate buffer at effective residence times ranging from 1 to 600 ms. At the end of each series of injections, a fresh microcolumn was placed onto the system for convenience and to eliminate the possibility of any carryover effects between studies. The amount of nonretained hemoglobin that eluted during the sample application step was measured at 428 nm and compared to the signal obtained when only diol-bonded silica (and no reversed-phase support) was present in the column. A similar comparison was made between the hemoglobin that eluted nonretained from this diol-bonded silica column and the total amount of hemoglobin that passed through the chromatographic system in the absence of any column. This latter experiment indicated that no detectable adsorption was occurring between the hemoglobin and diolbonded support. (14) Smith, P. K.; Krohn, R. I.; Hermanson, G. T.; Mallia, A. K.; Gartner, F. H.; Provenzano, M. D.; Fujimoto, E. K.; Goeke, N. M.; Olson, B. J.; Klenk, D. C. Anal. Biochem. 1985, 150, 76.

1368

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

The immunoaffinity support for the adsorption-limited studies was prepared by coupling anti-fluorescein antibodies to diolbonded silica through the Schiff base method.12,13 Prior to coupling, these antibodies were incubated with fluorescein to protect their binding sites from inactivation during the immobilization reaction. The immobilized antibodies were later washed with pH 2.5, 0.10 M phosphate buffer to release the bound fluorescein. This immunoaffinity support was then placed into pH 7.0, 0.10 M phosphate buffer and used to pack a 545-µm-thick layer within a 2.1 mm i.d. × 1.0 cm column, with the remainder of the column containing diol-bonded silica. A series of 50-µL injections of 5.4 nM fluorescein were made onto this column at flow rates of 0.41.1 mL/min using pH 7.0, 0.10 M phosphate buffer as the application solvent. The amount of nonretained fluorescein was monitored by using on-line fluorescence detection with excitation and emission wavelengths of 488 and 520 nm, respectively. Between each injection, the column was washed with pH 2.5, 0.10 M phosphate buffer, followed by reequilibration with the pH 7.0 application buffer. The amount of adsorbed fluorescein was determined by comparing the nonretained peak areas of the antifluorescein column to peak areas measured for the same samples when applied at identical flow rates to a column that contained only diol-bonded silica. Control experiments similar to those used in the hemoglobin work indicated that no detectable adsorption occurred between the fluorescein and the diol-bonded support. The immunoaffinity support for the chromatographic competitive binding immunoassay was prepared by adsorbing anti-BSA antibodies to a 2.1 mm i.d. × 1.0 cm column that contained a 1.1-

mm-thick layer of an immobilized protein G support. Antibodies were placed onto this column by making two 50-µL injections of anti-BSA antiserum at a flow rate of 0.05 mL/min in pH 7.4, 0.10 M phosphate buffer. Samples containing FITC-labeled BSA and BSA were injected in the same buffer, and the amount of nonretained labeled BSA was monitored through its fluorescence at excitation and emission wavelengths of 488 and 520 nm. The retained BSA and labeled BSA were later eluted with pH 2.5, 0.10 M phosphate buffer. The column was then reequilibrated with pH 7.4, 0.10 M phosphate buffer before the next sample injection. RESULTS AND DISCUSSION Preparation of Sandwich Microcolumns. One goal of this work was to decrease the residence times that could be obtained with HPLC columns while still allowing the use of these columns at standard flow rates. It is difficult to pack an ordinary chromatographic column with a length of less than a few millimeters, so a sandwich-based method was used instead. This was accomplished by employing a conventional HPLC pump and injection valve to apply a slurry of one or more types of supports to a column. The exact procedure that was used to make such a column for this study is given in the Experimental Section. For convenience in handling and for protection of the stationary-phase layer, the column was first filled partially with a biologically inert material (i.e., diol-bonded silica). After this support had been placed into the column, the flow rate of the packing solvent was increased to ensure that this material was present in an even layer. A layer of the desired active support was next placed into the column in a similar manner to give a sandwich microcolumn. The length of this layer was determined by the concentration and volume of the injected support slurry and the number of injections that were made of this slurry. The remainder of the column was then filled with diol-bonded silica or some other inert support to remove any dead space at the head of the column. A picture of a typical microcolumn that was prepared for this study is shown in Figure 3. The effective length (Leff) and volume (Veff) occupied by the stationary-phase layer within such a column was adjusted by using the following equations,

Veff ) NinjVinjCs/Fs

(1)

Leff ) NinjVinjCs/πr2Fs

(2)

where Cs is the concentration of the support in the injected slurry, Vinj is the calibrated slurry volume applied per injection, Ninj is the total number of slurry injections that were made of the given support, Fs is the known packing density of the support within the column, and r is the internal radius of the column. A total of 32 or 64 injections for a particular support was usually employed, as will be discussed later. The length of the support layer was varied by changing the density of its injected slurry. In the initial stages of this work, the final lengths of the support layers in the microcolumns were determined by using a colored support (as shown in Figure 3) and optical microscopy to measure the thickness of this layer. In each case, the actual thicknesses of the stationary-phase layers were found to agree within experimental error to the lengths that were estimated by using eq 2. For the sake of convenience and ease of handling, the total length of the columns used in this particular study was 1.0 cm,

Figure 3. Example of a typical sandwich microcolumn. This column was packed with Nucleosil Si-300 silica and later removed from its housing for the acquisition of this image. Immobilized hemoglobin silica, shown as the dark center band, was used to visualize the stationary-phase layer after the column was unpacked. The total length of the column was 1.0 cm, and the effective length of the stationary phase layer was 620 µm. The column diameter was 2.1 mm.

which included both the stationary-phase-coated support and all inert materials. However, columns with total lengths as small as1-2 mm have been prepared with this approach. The thickness of the stationary-phase layer (referred to here as the microcolumn “length”) ranged from 1.1 to 60 µm. Since the columns were packed with 5-7-µm silica particles, the microcolumns had effective lengths equal to 9-160 particle diameters. The amount of active support required to make such columns was also quite small, being in the range of 90-1700 µg/column. Work with such small columns was possible because the separations performed in this work were essentially irreversible and involved only single contacts between the retained molecules and stationary phase under the time and solvent conditions used in this study. Several specific examples of such systems will be examined later in this report. Evaluation of Sandwich Microcolumns. Once a method for packing sandwich microcolumns had been established, the reproducibility of this method was tested. One way this was done was by comparing the total protein content of several columns that had been packed with small layers of an immobilized hemoglobin support. When two injections of a 4.8 mg/mL slurry were made of this material, this delivered a 620-µm-thick layer with a precision of ∼20% ((1 RSD). This reproducibility was much worse than the precision of only a few percent that was observed for the protein assay itself, thus indicating that this variability was mainly due to the packing method. Several possible sources of this variability were considered to improve the reproducibility of the packing procedure. These sources included (1) potential changes in the slurry density between injections and (2) variations in the slurry volume that was applied with each injection. To ensure that the slurry density remained constant over time, this density was routinely monitored by turbidity measurements and maintained by vigorous mixing of the slurry between injecAnalytical Chemistry, Vol. 73, No. 6, March 15, 2001

1369

Figure 4. Reproducibility of stationary-phase content in a sandwich microcolumn as a function of the number of injections used to apply a fixed amount of an immobilized hemoglobin support to a 2.1 mm i.d. × 620 µm microcolumn. These results represent the averages of triplicate analyses.

tions. This resulted in slurry densities that showed less than a few percent variation over even a large series of injections. An approach was also developed for minimizing the effects of variations in the injection volume; this was done by using a large number of injections of dilute slurry suspensions to average out such effects and reduce their overall impact. Figure 4 shows some results that were obtained by this technique for a 620-µm support layer. As can be seen from this plot, an increase in the number of slurry injections (with a corresponding decrease in slurry concentration) resulted in a significant improvement in the precision of the packing method. In this example, a variation of less than 5% in support delivery was obtained when 32 injections of a 0.3 mg/mL slurry were used. For longer microcolumns, an even better precision was obtained under comparable conditions; however, work with smaller column lengths required a greater number of injections to obtain similar results. Microcolumns containing a colored support, like the hemoglobin silica shown in Figure 3, were inspected by optical microscopy for their packing uniformity. It was found that columns containing support layers of 60 µm in length or greater gave level, well-defined boundaries. But attempts to pack even smaller columns gave rise to some edging effects in which the support layers were no longer packed evenly across the diameter of the column. In this case, the walls of the column contained a slightly larger layer of the colored support than the section in the center of the column. It is believed that this is due to the presence of some flow heterogeneity within the column during the packing procedure. Such an effect was not noticeable or significant in the longer microcolumns (effective lengths, 60 µm-1.1 mm) that were used in the remainder of this study. Another series of studies were performed to determine the flow rates and column residence times that could be employed with the sandwich microcolumns. This required that some information be obtained on the permeability of these columns. This was done by packing several microcolumns with 7-µmdiameter HPLC-grade silica that had nominal pore sizes of 3004000 Å. These supports were chosen because they are commonly used for proteins and other biological molecules in HPLC. For each type of support, a plot was made of back pressure versus 1370

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

Figure 5. Pressure versus flow rate for columns packed with Nucleosil Si-1000 ([) or Nucleosil Si-300 (b) supports. These measurements were performed at room temperature using pH 7.0, 0.1 M potassium phosphate buffer as the mobile phase. The error bars shown about each point represent a range of (1 SD for triplicate measurements. Table 1. Permeabilities of Several Types of Porous Silica Supports in Sandwich Microcolumns

a

nominal pore size (Å)

measured permeability, Bo (× 107 cm-2)a

300 500 1000 4000

4.8 ((0.6) 4.2 ((0.4) 3.3 ((0.1) 3.6 ((0.2)

The numbers in parentheses represent a range of (1 SD.

flow rate, as shown in Figure 5. This type of plot was then analyzed by using the Darcy equation,

∆P ) umpoηL/Bo

(3)

where ump is the linear velocity of the mobile phase, o is the column’s interstitial porosity, η is the mobile-phase viscosity, L is the total column length, ∆P is the change in pressure across the column, and Bo is the specific permeability of the support within the column.15 According to eq 3, the overall permeability of the column (Bo) can be determined from the slope of plots such as those in Figure 5 by converting from flow rate to linear velocity and using the known or measured values of o, η, and L. The permeabilities that were measured for the sandwich microcolumns are given in Table 1 for several porous silica supports. Based on these values, the maximum usable flow rates for microcolumns packed with these materials were calculated to range from 12 to 17 mL/min at a back pressure of 3000 psi; in practice, an upper limit of 9-10 mL/ min was actually observed because of some nonlinearity that occurred in plots of back pressure versus flow rate when working at high flow rates (see upper plot in Figure 5). However, work at flow rates below this range was more than adequate for this study since a typical 500 µm long × 2.1 mm i.d. microcolumn allowed (15) Poole, C. F.; Poole, S. K. Chromatography Today, 4th ed.; Elsevier: Amsterdam, 1995.

Figure 6. General scheme used to represent analyte retention by sandwich microcolumns. The forward and reverse mass-transfer rate constants k1 and k-1 represent movement of the analyte (A) from the flowing mobile phase to stagnant mobile phase within the support. The second-order rate constant k2 describes the actual adsorption of the analyte to the stationary phase. A reverse first-order rate constant for analyte desorption could also be included in this model (k-2); however, this was not necessary in this current study since analyte binding to the column was essentially irreversible on the time scale of the retention studies.

sample residence times of 40-80 ms to be obtained at 1-2 mL/ min. As will be shown later, residence times as low as even 1-2 ms could be obtained by using shorter columns and/or higher flow rates for sample application. Behavior of Sandwich Microcolumns under AdsorptionLimited Conditions. The next series of experiments considered the behavior of sandwich microcolumns in separations performed in systems with either “diffusion-limited” or “adsorption-limited” kinetics for analyte retention.16,17 This was based on a model in which analyte interaction with the stationary phase was viewed as consisting of two distinct steps (see Figure 6): (1) movement of the analyte to the surface of the stationary phase by diffusion and (2) adsorption or interaction of the analyte with the stationary phase. The term “diffusion-limited kinetics” is used here to refer to a system in which the overall rate of analyte binding to the stationary phase is dictated by how fast the analyte can get to the stationary phase from the flow mobile phase. For the model shown in Figure 6, this occurs when the rate constant k2 is much larger than k1 or k-1. The opposite situation (adsorption-limited kinetics) occurs when k1 and k-1 are much larger than k2. In this situation, the analyte gets to the stationary phase at a rate that is much faster than analyte-stationary phase binding, thus making this second process the rate-limiting step in retention. The first case which was examined was that of a system with adsorption-limited kinetics. This was accomplished by injecting a small analyte (fluorescein) onto a column that contained antifluorescein antibodies attached to HPLC-grade silica. Previous studies have shown that antibody-antigen systems such as this tend to display adsorption-limited kinetics when the antibody is part of an HPLC column.18,19 Each fluorescein sample was first injected in triplicate at various flow rates onto an inert column containing only diol-bonded silica, to which fluorescein does not bind, to determine the total expected peak area. The same sample was then applied to a sandwich microcolumn of the same overall (16) Hage, D. S.; Walters, R. R. J. Chromatogr. 1988, 436, 111. (17) Hage, D. S.; Walters, R. R.; Hethcote, H. W.Anal. Chem. 1986, 58, 274. (18) Hage, D. S.; Thomas, D. H.; Beck, M. S. Anal. Chem. 1993, 65, 1622. (19) Hage, D. S.; Thomas, D. H.; Roy Chowdhuri, A.; Clarke, W. Anal. Chem. 1999, 71, 2965.

Figure 7. Adsorption of fluorescein using an anti-fluorescein immunoaffinity support. Microcolumns varying in lengths from 124 to 620 µm were used for this study along with various flow rates to adjust the column residence times.

size as the inert control column, but that now contained a welldefined layer of anti-fluorescein antibodies. The degree of retention was then determined by comparing the size of the nonretained peaks for fluorescein on the microcolumn to those measured at the same flow rates on the control column. The results of this experiment are summarized in Figure 7. It was found that greater than 95% extraction of fluorescein could be obtained in as little as 100-120 ms with sandwich microcolumns. Previous studies with antibodies attached to low-performance agarose supports have reported greater than 95% binding in roughly 100 s or 40% extraction of analytes in 2 s,5,20 while antibodies attached to HPLC supports have been shown to give quantitative retention in as little as 6 s.21 However, this present study is the first report in which quantitative binding with immunoaffinity supports has been noted in the millisecond time range. Similar studies were conducted with sandwich microcolumns and antibodies directed against other analytes. For instance, the analytes L-thyroxine and warfarin have also been found to give quantitative retention in only 60-150 ms by such columns. When attempting this type of rapid separation, it is necessary to consider the amount of stationary phase that is actually present in the column for analyte retention. Increasing the amount of stationary phase per unit volume not only helps avoid column overloading, but in an adsorption-limited system, this also helps to speed the net rate of retention by increasing the probability that an analyte will encounter an unoccupied binding site as it approaches the stationary phase. Typically, immunoaffinity separations use a large excess of antibodies to help avoid these problems.22-25 But this can be more difficult to attain in microcolumns, where the space available for the stationary phase is minimal. To address this, the anti-fluorescein antibodies used in this study were immobilized under conditions that have previously (20) Pollema, C. H.; Ruzica, J. Anal. Chem. 1994, 66, 1825. (21) Hage, D. S.; Kao, P. C. Anal. Chem. 1991, 63, 586. (22) Ohlson, S.; Gudmundsson, B.-M.; Wikstrom, P.; Larsson, P.-O. Clin. Chem. 1988, 34, 2039. (23) Janis, L. J.; Regnier, F. E. Anal. Chem. 1989, 61, 1901. (24) Thomas, D. H.; Beck-Westermeyer, M.; Hage, D. S. Anal. Chem. 1994, 66, 3823. (25) Rollag, J. G.; Beck-Westermeyer, M.; Hage, D. S. Anal. Chem. 1996, 68, 3631.

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

1371

Figure 8. Adsorption of hemoglobin using a reversed-phase chromatographic support. Microcolumns varying in lengths from 62 to 620 µm were used for this study along with various flow rates to adjust the column residence times.

been shown to maximize the density of antibodies on HPLC-grade silica supports.26 This provided an amount of antibodies in the anti-fluorescein microcolumns that was roughly 2000-3000 times greater than the moles of fluorescein that were injected. As can be seen from Figure 7, this level was more then sufficient to allow quantitative retention of fluorescein to be achieved in the subsecond time domain. Behavior of Sandwich Microcolumns under DiffusionLimited Conditions. The second type of kinetic case considered in this report was the use of sandwich microcolumns to extract analytes under diffusion-limited conditions. An example of such a system is the retention of a protein like hemoglobin by a reversed-phase support.16,27-31 This was studied by packing a sandwich microcolumn that contained a 1.1-mm layer of C18 silica between layers of diol-bonded silica. The degree of retention of hemoglobin samples was estimated by making a continuous series of protein injections, with no elution step between, until the column was saturated with hemoglobin. All of the peak areas were then compared to those for peaks that were generated after column saturation. A similar comparison was made between the peaks measured on the C18 microcolumn and those that were obtained for the same samples on an inert control column that contained only diol-bonded silica. The data that were obtained with this system are illustrated in Figure 8. It was found that up until the point of column saturation (which occurred after ∼24 100-µL injections of 2 mg/ mL hemoglobin) greater than 95% of the injected hemoglobin was extracted at residence times as short as 4 ms. This was followed by a fairly sharp decrease in retention efficiency at shorter times (i.e., 1-2 ms). This retention rate was much faster than what was observed in Figure 7 for the fluorescein/anti-fluorescein system. (26) Clarke, W.; Beckwith, J. D.; Jackson, A.; Reynolds, B.; Karle, E. M.; Hage, D. S. J. Chromatogr., A 2000, 888, 13. (27) Karamushka, V. I.; Denisova, T. I.; Sklyarov, A. G.; Gurzina, T. G.; Meleshevich, S. I.; Ul’berg, Z. R. Zh. Prikl. Khim. 1989, 62, 561. (28) Norde, W.; Rouwendal, E. J. Colloid Interface Sci. 1990, 139, 169. (29) Ramsden, J. J.; Roush, D. J.; Gill, D. S.; Kurrat, R.; Willson, R. C. J. Am. Chem. Soc. 1995, 117, 8511. (30) Voegel, J. C.; De Baillou, N.; Sturm, J.; Schmitt, A. Colloids Surf. 1984, 10, 9. (31) Place, H.; Sebille, B.; Vidal-Madjar, C. Anal. Chem. 1991, 63, 1222.

1372

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

One reason for this is that the fluorescein system had adsorptionlimited kinetics in which retention was limited by the rate at which the analyte could reach unoccupied antibody sites with a proper orientation for binding. In contrast to this, the binding of hemoglobin to C18 silica is a much faster, but less specific, diffusion-limited process that involves a greater number of potential binding regions and in which many more collisions of hemoglobin with the stationary phase result in analyte retention.32-37 Use of Sandwich Microcolumns in Chromatographic Immunoassays. The last series of studies considered the use of sandwich microcolumns in a chromatographic immunoassay. This type of assay uses immobilized antibodies or antigens as part of a chromatographic system for the fast and selective determination of analytes.38 The most common format for such an assay is the competitive binding mode. In this format, analyte in the sample is incubated with a fixed amount of a labeled analyte analogue and is applied to a column that contains a limited amount of antibodies that can bind to both of these species. This is most often done by simultaneously injecting the analyte and its labeled analogue onto the column,19 but sequential injection can also be employed.18 The antibody-bound fraction of both compounds is then separated from the fraction that remains free in solution. From this, the amount of labeled analyte that was in the free or bound fraction can be determined, thus providing an indirect measure of how much analyte was in the original sample.39 Sandwich microcolumns are attractive for use in such assays because they provide a convenient way of placing a small amount of antibodies into a column while still allowing work to be performed at flow rates and column residence times that allow a competition to be established between the analyte and its labeled analogue.19 In addition, the use of microcolumns will minimize the surface area to which the analyte and label are exposed. This should help reduce nonspecific binding, which is often a limiting factor in determining the lower limit of detection that can be obtained in immunoassays. A chromatographic competitive binding immunoassay with microcolumns was developed by using the competition of labeled and nonlabeled BSA as a model. Anti-BSA antibodies were first adsorbed to a protein G microcolumn to form an immunoaffinity stationary phase. Various mixtures of FITC-labeled BSA and nonlabeled BSA were then injected onto this column, with the nonretained labeled BSA being monitored by an on-line fluorescence detector. Some typical chromatograms obtained by this method are shown in Figure 9. The sample on the left contained only a solution of the labeled BSA, while the second and third samples to the right contained the same amount of labeled BSA plus either a 10- or 50-fold excess of normal BSA. (32) Lok, B. K.; Bheng, Y.-L.; Robertson, C. R. J. Colloid Interface Sci. 1983, 91, 104. (33) Ramsden, J. J.; Roush, D. J.; Gill, D. S.; Kurrat, R.; Willson, R. C. J. Am. Chem. Soc. 1995, 117, 8511. (34) Voegel, J. C.; De Baillou, N.; Sturm, J.; Schmitt, A. Colloids Surf. 1984, 10, 9. (35) Van Dulm, P.; Norde, W. J. Colloid Interface Sci. 1983, 91, 248. (36) Young, B. R.; Pitt, W. G.; Cooper, S. L. J. Colloid Interface Sci. 1988, 125, 246. (37) Cheng, Y. L.; Darst, S. A.; Robertson, C. R. J. Colloid Interface Sci. 1987, 118, 212. (38) Hage, D. S. Anal. Chem. 1993, 65, 420R. (39) Hage, D. S. J. Chromatogr., B 1998, 715, 3.

Figure 9. Competitive binding immunoassay for albumin performed on a sandwich microcolumn that contained anti-albumin antibodies as the stationary phase. The column was prepared and sample injections were performed as described in the text. Sample injection took place at 0 min in each plot. The concentration of FITC-BSA in each sample was 1.6 µM. The concentrations of nonlabeled BSA were (from left-to-right) 0, 17.9, and 88.8 µM, respectively.

An 80% retention of labeled BSA was achieved in the sample containing only this labeled compound. Injection of this label plus a 10-fold excess of nonlabeled BSA yielded a 15% decrease in the binding of labeled BSA, and injection of the labeled BSA plus a 50-fold excess of nonlabeled BSA gave a 65% decrease in binding (see Figure 9). It is this inverse relationship between binding of the labeled analogue and the concentration of the unlabeled analyte that can be used for analyte measurements. For each injected solution, the amount of time the sample was in contact with the immunoaffinity support was ∼180 ms. However, a signal was not observed until about 5-25 s after injection due to the additional time that was required for the sample to pass out of the injection loop and through the column, connecting tubing and detector. Work is currently in progress to further minimize this time by decreasing the volume of these other system components. However, these results do clearly indicate that a sandwich microcolumn can be used to perform competitive binding immunoassays on very short time scales. This approach is not limited to BSA but could also be employed with any other compound for which antibodies and an appropriately labeled analogue are available. CONCLUSIONS This study examined the development of sandwich microcolumns for use with a small amount of stationary phase and the

creation of more rapid methods for analyte retention. This approach used standard HPLC supports and was capable of reproducibly packing these materials into layers as small as 60 µm in length and using as little as 90 µg of support material. This provided columns with effective residence times in the millisecond time range when used under standard HPLC flow rate and pressure conditions. Two applications examined for these microcolumns included their use in RPLC for the adsorption of proteins and in immunoaffinity columns for the selective retention of small molecules. These represented two contrasting situations that were controlled by either diffusion- or adsorption-limited rates of retention. The RPLC adsorption of hemoglobin, which has diffusion-limited retention, gave 95% binding in as little as 4 ms with the sandwich microcolumns. The adsorption of fluorescein by an anti-fluorescein antibody column (an adsorption-limited system) gave 95% binding in 100-120 ms. Similar results were noted with other analytes and demonstrate the relatively short time frame that can be obtained by microcolumns in such studies. Another specific application that was examined for these columns was their use in a chromatographic-based competitive binding immunoassay. This gave rise to a method in which a signal related to the analyte content of a sample could be obtained in 5-25 s after injection. This particular format could easily be expanded to many other types of analytes, such as drugs, proteins, or environmental agents for which suitable antibodies and labeled agents are available.39 The need for only a small amount of stationary phase per microcolumn is particularly appealing, since this minimizes the amount of antibody or biological ligand that is required in such assays. This opens up the possibility of using more exotic or expensive ligands in these columns. This, in turn, should be useful in a variety of fields that use ligand-based assays, including clinical chemistry, pharmaceutical testing, combinatorial screening, and analysis of proteomic libraries. ACKNOWLEDGMENT This work was supported by the National Institutes of Health under Grant R01 GM44931 and by the University of Nebraska Research Council. W.C. was supported in part by a fellowship from the University of Nebraska Medical Center for Environmental Toxicology.

Received for review July 27, 2000. Accepted January 9, 2001. AC000870Z

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

1373