Anal. Chem. 1002, 64, 1973-1977
1073
Kinetic Chromatographic Sequential Addition Immunoassays Using Protein A Affinity Chromatography Scott A. Cassidy, Linda J. Janis,+and Fred E. Regnier' Department of Chemistry, Purdue University, West Lafayette, Indiana 47907
A new type of chromatographic immunoassay based on sequential addltlon Is descrlbed. On a proteln A column, the antlbody, the sample contalnlngthe antigen, and then a known amount of antlgen are sequentially Inlected. Thls assay Is derlgned to shorten analysls times and reduce compiexlty of duakdumn chromatographic Immunoassays, clrcumvent desorption buffer Interferences common to afflnlty chromatography, and ellmlnate the need for tagged molecules. This new technlque Is named kinetic immunochromatography sequential addltlon (KICQA). Because of Its klnetlc nature, flow rate wlil have a large effect on KICQA, and the Impact of changlng flow rate Is studled extenslvely. By use of varlous amounts of antlbody, the dynamic range of KICQA Is shown to be selectable over 2.5 orden of magnitude. Flnally, KICQA was usedto determlnetransferrin and albumin in humanserum. Both analytes show good agreement wlth thelr respectlve reference methods, and an albumln assay was performed In under 1 mln.
INTRODUCTION Analytical determinations based on antigen-antibody interactions generally require three steps. The first step is the reaction of the antigen with the antibody. The second involves separation of the immunological complex from the sample, and the third is quantitation of either the bound or free antigen. It is in the second and third steps of these assays that there is a need for improvement in current methodology. Chromatographicimmunoassays were introduced in 1978.' These assays were simply quantitative affinity chromatography. Using rigid supports containing a large excess of antibody, antigen was adsorbed from the sample and then desorbed from the immunosorbent into an absorbance detector. While this adsorption-desorption could be fast,2 problems arise with this technique. Slow desorption kinetics give broad peaks, nonspecificallyadsorbed proteins interfere, and desorption agent causes baseline perturbations. These problems compromise sensitivity. Some of these problems have been overcome by dualcolumn immunoassays, albeit at the expense of time.3+ In these techniques, the antigen is desorbed from the immunosorbent onto a reversed-phase column and separated from the aforementioned interferences. The purpose of this work was to devise an immunoassay that overcame the problems noted above while still being
* To whom correspondence should be addressed. +
Present address: Eli Lilly and Co. Indianapolis, IN 46285.
(1) Mosbach, K.; Ohlsson, S.;Hansson, L. FEES Lett. 1978, 93, 5-9. (2) Kalghatgi, K.; Varaday, L.; Horvath, C. J. Chromatogr. 1988,458,
207-215. (3) Janis, L. J.; Grott, A.; Regnier, F. E. J. Chromatogr. 1989, 476, 235-244. (4) Janis, L. J.; Regnier, F. E. Anal. Chem. 1989,61, 1901-1906. (5) Riggin, A.; Regnier, F. E.; Sportsman, J. R. Anal. Chem. 1991,63, 468-474. (6) Zdunek, D.; Kratzert, M.; Reh, E. Anal. Eiochem. 1991,196,104110. 0003-2700/92/0364-1973$03.00/0
simple, sensitive, and inexpensive. It is proposed that a sequential addition immunoassay (SAIA) will provide a solution. Traditional SAIA involves the sequentialincubation of an unknown amount of antigen (the dose) and then a known amount of pure labeled antigen (the label) with a known amount of antibody immobilized on a solid support. The terms dose and label will be used hereafter to distinguish these different quantities of antigen. Note that the amount of label is usually equal to the antigen-binding capacity of the antibody. The dose in SAIA can be a very complicated mixture such as urine or serum. Antigen in the dose is first allowed to come to equilibrium with the immobilized antibody; then the label is added and allowed to come to equilibrium. Finally, the supernatant liquid is removed from the antibody, and the remaining label quantitated. SAIA could be adapted to a chromatographic format. However, such an assay would require strict control of the amount of antibody immobilized on the column. Traditional immunosorbents could not meet this requirement because packing a column with the exact antigen capacity required would be difficult, and a new column would be required for different capacities. Antibody might also be denatured during repeated use. However, these problems could be circumvented by performing the assay on a protein A column. Protein A is a bacterial protein which binds most mammalian IgG molecules a t the constant (F,)region with differing affinities; this orientation maximizes the ability of the bound antibody to bind antigena7 Protein G has similar properties and could also be used in this role. The literature provides several examples of the use of protein A or G in chromatographic im'munoassays.4~5~s For an immunoassay, any desired amount of antibody (up to the column capacity) can be injected onto the protein A column. When the assay is finished, the antibody-antigen complex is desorbed from the protein A column, and the cycle is repeated. In addition, one column has the flexibility to carry out assays for many different antigens. This work will demonstrate a chromatographic SAIA using protein A affinity chromatography. After demonstration of the chromatographic SAIA, the effects of several variables on assay parameters will be explored. Finally, the determination of albumin and transferrin in human serum by this method will be demonstrated.
EXPERIMENTAL SECTION Columns. Experiments were conducted on five different columns. The 300-A-pore-diameter,10-pm-particle-diameter,50X 5-mm protein A column came from Pierce (Rockford, IL). The 500-A, 30-pm, 30- X 2-mm and lWO-A, 30-pm, 50- X 4.6-mm protein A columns as well as the 300-A, 30-pm protein G material that was packed into a 30- X 2-mm column were gifts of Chromatochem (Missoula, MT). Two 15-pm, 30- X 2-mm Poros columns were gifts of Perseptive Biosystems (Boston, MA). Instrumentation. Experiments were carried out on several different instruments. All experiments with the 500-A column (7) Langone, J. J. Adv. Immunol. 1982, 32, 157-252. (8)Mattiasson, B.; Borrebaeck, C. ACS Symp. Ser. 1979,106,91-105. 0 1992 American Chemical Society
1974
ANALYTICAL CHEMISTRY, VOL. 64, NO. 17, SEPTEMBER 1, 1992
were carried out on a Hewlett-Packard 1090Mworkstation (Palo Alto, CA) with a diode array detector and autosampler injection; injection volumes for antibody and antigen were 1 and 2 pL, respectively. Detection was by absorption at a wavelength of 225 nm with a bandwidth of 40 nm; integration was accomplished by the workstation. Data from the Poros column operated at 0.5 mL/min were taken with a Hewlett-Packard 1090Mworkstation equipped with a filter photometer. Detection was by absorption at 210nm witha 10-nmbandwidth. Integration was accomplished with a Hewlett-Packard 3390 integrator. Sample injectionswere made with a Rheodyne 7125 six-port injection valve (Couati, CA) fitted with a 20-pL loop. All other data using transferrin were taken on an apparatus consisting of a Waters 690 reciprocating pump (Milford, MA), a Waters WAVS valve station, a Hewlett-Packard 1040A diode-array detector with an HP85b computer, a Waters 740 integrator, and the same Rheodyne 7125 six-port injection valve and 20-pL loop. Detection was again by absorption at 225 nm with a 40-nm bandwidth. The albumin capacity experiments used two Rheodyne 7125 six-port injection valves and one Rheodyne 7120 six-port injection valve in series with the Waters 590WAVS and Hewlett-Packard 1040Adetector. Detection was at 220 nm with a bandwidth of 20 nm. Reagents. Human albumin and apo-transferrin were obtained from Sigma (St. Louis,MO). Rabbit anti-human transferrin and anti-human albumin were obtained from Boehringer-Mannheim (Indianapolis, IN). Mono- and dibasic phosphates, sodium chloride, Tris, and glycine were obtained from Aldrich (Milwaukee, WI) and used as received. HPLC grade acetic acid was purchased from Baker (Phillipsburg,NJ), and HPLC grade ethylene glycol, acetonitrile, and 2-propanol were procured from Mallinkrodt (Paris,KY). Phosphoricacid (85%) wasfromFisher (Fair Lawn, NJ). Procedures. Unless noted otherwise, all experiments used a 0.22-pm-filtered 0.01 M phosphate buffer (pH 7.0) containing 0.15 M sodium chloride for sample loading (load buffer). The antigen-antibody dissociation experiments were carried out in 0.2 M Tris (pH 7.4) on the protein G column. A minimum of four column volumes of 20% acetic acid or 20% acetic acid/50% ethylene glycol was used to desorb antibody from the columns at 2 mL/min. Antibody was diluted to the proper concentration (4pg antigen capacity) in load buffer before injection. Standards (i.e. the doses) of human transferrin were made up in load buffer to give injected amounts of 0, 1, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 25, 30, 35, 40, 45, and 50 pg. The label was also 4 pg. For all experiments, the antibody was injected first, followed by the dose and then the label. The antigen-antibody dissociation experiments used an incubation flow rate of 1 mL/min. The antibody-loading flow rate was 2 mL/min for all experiments. The 'incubation" flow rate (i.e. the flow rate during the dose and label injections) was varied from 0.1 to 2 mL/min during the flow rate experiments. The Poros column was also used at 3 and 4 mL/min. The 300- and 1000-8, columns failed to give peaks at zero antigen dose for 0.1 mL/min; the 1000-8, column also failed to give a peak at zero antigen dose for 0.5 mL/min. The antigen capacity experiments involved making calibration curves using antigen capacities of 4, 6, and 8 pg. The standards and label were the same as mentioned above. The 500-A column was used for this experiment at an incubation flow rate of 2 mL/min. The stationary-phase experiments proceeded exactly as the flow rate experiments. The albumin capacity experiments were carried out on a second Poros column with either 0.36, 9, or 36 pg of albumin capacity. A t 0.36-pg capacity, standards were 0, 0.04, 0.08,0.20,0.40,0.60,0.80,1.0,1.2, and 1.4 pg with an incubation for flow rate of 0.236 mL/min. At 9-pg capacity, standards were 0, 8, 12, 16,20, 24, and 40 pg with an incubation flow rate of 2 mL/min. At 36-pg capacity, standards were 0,20,30,40,50,and 60 pg with an incubation flow rate of 4 mL/min. In each case the label was the standard closest to the antgen capacity, i.e. 36-pg capacity used 40 pg as the label. The respective dilutions of serum were 20-, 80-, and 1500-fold with the load buffer. For the albumin experiments,the load buffer was 50 mM Tris-acetate pH 7.4. Desorption was with 2 mL (20 column volumes) of a mixture of 20 mM phosphoric acid, 10% acetonitrile, and 15% isopropyl alcohol.
Unretained Portion Rotein A
Calunn
Injection
r i
of Injection
Y Y Y
y Y y y
Y
a
A A
A A
* A " A
u u
Protein A on Column
0
Bntibody
Y
Antigen
A
Unretained Canponents
D
n n
O
Fburr 1. Schematic of KICQA. Row 1 represents injection of the antlbody. Row 2 represents injection of the dose. Row 3 represents injection of the label. The unbound portlon of the label (row 3)1s the signal for the experiment.
RESULTS AND DISCUSSION T o perform the chromatographic SAIA, one would sequentially inject antibody, the dose, and then the label onto a protein A column, as shown in Figure 1. After passing through the column, the unbound label, which is proportional to the antigen in the dose, would be quantitated by ultraviolet absorption. The antigen-antibody complex would then be removed from the immunosorbent under acidic conditions. Figure 2 represents an ultraviolet detector strip-chart recording of the process. Similar approaches using simultaneous addition and traditional antibody columns have been reported.8 Because the dose and label contact the antibody a t different times, the label would not need a fluorescent or radioactive tag to distinguish it from the dose. This is a distinct advantage over the simultaneous addition approach8 and perhaps the most useful feature of the chromatographic SAIA. Performing a SAIA in a flowing stream makes it a kinetic immunoassay. Because of its kinetic nature and use of sequential addition, the technique will be referred to as kinetic immunochromatographic sequential addition (KICQA). The KICQA technique using a protein A column can be described in terms of three consecutive second-order rate equations. The first equation describes the binding of antibody to the protein A column. [AI is the protein A concentration, [Bl is the antibody concentration, and kgA is the forward rate constant. Similar
ANALYTICAL CHEMISTRY, VOL. 64, NO. 17, SEPTEMBER 1, 1992
. fron Pnteru.tiust Antibody
10000
1
o
1. Desorption
2. Unbound Portion of Dote
o
o
o
o
A.
8000
P
n
e a
t
k
b 0
r
b
a
n C
a
1971
3 . Unbound Label
1
1
A r e a
2oooi - / 2000
o+0
0
I
I
I
I
10
15
20
10000
1 1
8000
P e a
k
6000
J
4000
A
r
2000
e
a
0 0
5
ug Transferrin Injected Figurr 9. KICQA calibration curves: (A) linear regression; (6) nonlinear regression Y = A/( 1 + B(exp(-CX))). The circles are raw data points; the lines are the different regressions. Table I. Comparison of the Dynamic Range of Linear and Nonlinear Regression Analysis useful dynamic actual dynamic regression type range,’ rg rangesbrg 4-a 4-a linear nonlinee 4-12 4-10 Defined as the range in which the points of the fit vary by more than 5 % . See text for details. b Defined as the range in which the points of the fit vary by more than 10%. See text for details. c The nonlinear function used waa Y = A/(1 + B(exp(-CX))). expected. The dynamic range of the plot is roughly 4-12 pg; since the antigen capacity used was 4 pg, this could also be thought of as a dynamic range of 1-3 times the capacity. T w o methods were used to construct calibration curves. First, a linear regression was applied over the linear portions of the cuve, as shown in Figure 3A. The second method involves fitting the data to some function that mimics the sigmoidalshapeof the data. Several functionswere examined, the most useful being
(9) Liddel, J. D.; Sporteman, J. R.; Wilson, G.S. Anal. Chem. 1983, 55,711-778.
Y = A/(1+ B(exp(-CX))) (4) Figure 3B shows a typical fit using eq 4. Table I presents results taken from data in Figure 3. The definitions of useful and actual linear ranges are as follows: the useful linear range assumes a difference in response of 5 % between points in the fit for both types of regression. This interval was chosen on the assumption that about 1% relative sample deviation was the optimum precision one might obtain. The 5% interval maintains a distance of five sample deviations between points of the fit and ensures the independence of each point in the useful range. The actual dynamic range has two criteria. The first criterion requires an interval of 10% between the
ANALYTICAL CHEMISTRY, VOL. 64, NO. 17, SEPTEMBER 1, 1992
1978
Table 11. Antigen-Antibody Dissociation Test delay time, min
dose delay, label-free area
label delay, label-free area 279 886 271 823 274 923 274 923 1.58
279 886 259 312 269 139 269 446 3.82
1 2 3
av RSD, %
Table V. Typical Assay Times for KICQA flow rate, mL/min
approx assay time,a min
flow rate, mL/min
approx assay time,a min
0.1 0.5 1.0
7.7 4.1 3.2
2.0 3.0
2.8 2.3
a
The times shown here are not optimized.
n.
Table 111. Effect of Flow Rate on Useful and Actual Dynamic Range. flow rate, mLimin 0.1 0.5 1.0 2.0 3.0
linear regression actual dynamic range, pg 2-6 2-8 1-8 0-10 1-16
useful dynamic range, r g 2-6 2-8 1-8 0-10 1-16
nonlinear regression useful actual dynamic dynamic range, r g range, rg 0-10 2-8 1-12 2-8 1-8 2-8 0-12 0-12 1-16 1-16
Data taken on Poros column. See Table I for details. 2.
Table IV. Effect of Flow Rate on Sensitivity column 500 A, 30 pm 30X2mm
300 A, 10 pm 50X5mm
lo00 A, 30 pm 50 X 4.6 mm Poros 30X2mm
3.
flow rate, mL/min
sensitivity,a pg-l
0.1 0.5 1.0 2.0 0.5 1.0 2.0 1.0 2.0 0.1 0.5 1.0 2.0 3.0 4.0
28.3 1.81 0.45 0.11 5.83 0.82 0.21 1.13 0.20 5.92 0.46 1.68 0.35 0.22 0.02
Data were normalized before comparison.
points of the fit due to the fact that most of the relative sample deviations in these studies are greater than 1%. To assure independence, the relative sample deviation of the actual data points was required to be less than half of the difference between the corresponding point in the fit and its neighbors. These standards may give conservative estimates of actual useful range for some of the data, particularly data from the 500-A column. Furthermore, the dynamic ranges from linear regression analysis were required to have a correlation coefficient of at least 0.98. These standards will be used throughout the rest of the paper. For the data in Figure 3 (Table I), the useful dynamic range of the nonlinear regression analysis is twice as large as the linear regression treatment. While this is an extreme example, the useful dynamic range of the nonlinear fit is usually the same or larger than that of the linear regression. The trend in actual dynamic ranges can vary because of the standards applied. One of the basic assumptions of the whole KICQA assay is that the protein A-antibody and antibodyantigen reactions are irreversible on the time scale of the experiment. To verify this assumption,the KICQA experiment was performed while either the dose or the label was delayed for times of 1, 2, or 3 min from the previous injection. The results of the experiments are summarized in Table 11. If bound antibody was being lost from the column, one would bind less antigen as the dose delay increased, and the free-label area would get larger. Table I1 shows that this is not the case. If bound
2.
c .
-
c
Figure 4. KICQA chromatograms. Chromatogram A Is from the 9-pg capacity, 2 mL/min experiment, and the total assay time (not shown) Is 1.6 mln. Chromatogram B is from the 36-1.18 capacity, 4 mLlmin experiment. In both chromatograms, peak 1 is antibody remnant, peak 2 is the unbound portion of the dose, and peak 3 is the unbound portion of the label.
antigen was being lost from the column, the free-label area would get smaller as the delay increased. A perusal of the label delay data of Table I1 shows no apparent trend of shrinking peak areas. The conclusion is that the assumption of irreversibility on the time scale of the experiment is valid. Effect of Flow Rate. As postulated previously,one might expect that dynamic range would increase with increasing flow rate. This hypothesis is confirmed in Table 111. A reduction in the time the antigen is in contact with the antibody requires more antigen to saturate the antibody. One would also expect sensitivity to decrease with increasing flow rate. Table IV, which lists the slopes of the linear treatments for the columns and flow rates used in this work, demonstrates this trend. Because the 5 0 0 4 column data was taken on a different integrator, the slopes are from the normalized data. The normalization process simply involvesdividing each peak area by the peak area at zero dose. Increasing flow rate also reduces assay time. Table V lists the average assay times of the different flow rates from Table 111. Assay times range from 7.7 min for a flow rate of 0.1 mL/min to 2.3 min for a flow rate of 3 mL/min. No attempt was made to minimize these assay times. See Figure 4 for optimized assay times.
ANALYTICAL CHEMISTRY, VOL. 64, NO. 17, SEPTEMBER 1, 1992
Table VI. Relationship of Flow Rate a n d Precision column
parameters
flow rate, dose, PU
mL/min
0
0.1 0.5 1.0 2.0 0.1 0.5 1.0 2.0 0.1 0.5 1.0 2.0
4
8
500 A RSD, % 11.31 6.94 0.67 1.93 12.13 0.70 0.32 0.25 1.15 0.29 2.40 0.74
Poros RSD, % 42.00 1.53 3.74 3.54 9.00 1.80 3.25 5.03 1.25 2.15 2.01 0.85
Table VII. Dynamic Ranges for Capacity Experiments. antigen capacity, Pg 0.36 9 36
linear regression useful actual dynamic dynamic range, Pg range, jtg 0.2-0.8 0.2-0.6 0-16 0-16 20-50 20-50
nonlinear regression useful actual dynamic dynamic range, fig range, r g 0-0.8 0-0.6 0-24 0-24 0-50 0-50
a Respective flow rates were 0.236,2, and 4 mL/min. Respective normalized slopes were 1.67, 0.32, and 0.14. ~
Flow rate also has some effect on assay precision. Table VI lists the relative sample deviations for the 5 0 0 4 and the Poros column a t several doses and flow rates. From Table VI, it is difficult to identify a definite trend based on flow rate. The only obvious features of Table VI are the high relative sample deviations a t 0.1 mL/min for doses of 0 and 4 pg. This phenomenon is interpreted to be a consequence of the antigen capacity used and the slow flow rate. At much slower flow rates, much more antigen can bind; a t zero dose, nearly all of the label binds. This leaves a relatively small amount of the label to reach the detector. In such a situation, small variations in other parameters and nonspecific binding can lead to large relative differences in the measured label peaks. As the relative amount of bound antigen decreases, this effect no longer operates and other variables control precision. This interpretation results from comparing the relative sample deviations of the two columns. The 500-A data were obtained with an autosampler; the Poros data were obtained with manual injection. The results at 4 pg might represent an intermediate case. Concentrations-Antigen Capacity. The dynamic range of a sequential addition immunoassay is determined entirely by the antigen capacity. In KICQA, increasing flow rate can expand the dynamic range, but the antigen capacity will still dictate the dynamic range. With this in mind, the KICQA assay was performed using albumin with anti-human albumin capacities of 0.36,9, and 36 pg. Table VI1 lists some relevant data, and Figure 4 shows actual strip-chart recordings of the 9- and 36-pg capacity experiments. A flow rate of 0.236 mL/ min for the 0.36-pg experiment was chosen to enhance the sensitivity for determination of the detection limit; the other flow rates were chosen to minimize assay time. As Tables I11 and VI1 demonstrate, the dynamic range of a KICQA assay can be selected from over 2.5 orders of magnitude. Detection Limits. A detection limit of 40 ng of human albumin was achieved in the 0.36-pg albumin capacity
1977
Table VIII. Accuracy of KICQA Assay for Human Transferrin and Albumin rate of colorimetric sample nephelometry, assay, KICQA, % antigen no. mg/mL mg/mL mg/mL error transferrin 1 0.95 1.10 15.6 2 0.98 1.00 2.0 3 1.55 1.56 0.6 4 1.53 1.44 5.9 5 1.75 1.76 0.6 albumin 1" 44 44.3 0.7 20 33 33.8 2.4 26 33 33.5 1.5 2c 33 36.3 10.0
-
*
a Albumin capacity 9 pg at 2 mL/min. Albumin capacity 0.36 jtg at 0.236 mL/min. Albumin capacity 36 pg a t 4 mL/min.
experiment. This should not be interpreted as the lower limit of detection for KICQA. This limit is determined by the system and operating conditions. In this case, the limit is subject to the limitations of the absorbance detector used. There are many ways to decrease this limit, including microcolumns, electrochemical, fluorescence, or absorbance labels for the label, and enzyme-linked methods. The role of antigen capacity in the determination of limits of detection should not be ignored. As the antigen in the dose becomes much smaller than the antigen capacity, sensitivity degrades severely (Figure 3). Decreasingthe antigen capacity and label should also serve to decrease the detection limit subject to detector performance. Applications. KICQA was used to assay transferrin and albumin in human serum (Table VI11 and Figure 4). The transferrin and albumin had been previously determined by rate nephelometry (transferrin) and colorimetric dye-binding assay (albumin). Both albumin and transferrin show good agreement with their respective reference techniques. Note that the albumin assay was conducted successfullya t dilutions ranging from 20-fold (at 36-pg capacity) to 1500-fold (0.36-pg capacity) in times as fast as 0.90 min. These data demonstrate the quantitative utility of KICQA. Efficient desorption of IgG is necessary due to the additional IgG from the serum. Reference 7 was helpful in designing desorption buffers.
CONCLUSION This work has demonstrated a new and useful type of immunoassay. The strengths of this technique are its simplicity, flexibility,and speed. The instrumental requirements of only a liquid chromatograph and an ultraviolet absorbance detector mean that it is easy to implement. The use of protein A or G columns and the availability of a wide variety of antibodies gives a remarkable flexibilityto perform many different assays with only one column. Finally, the short times required for KICQA assay compare favorably with other immunoassays.
ACKNOWLEDGMENT We thank Michael Wittrig for his assistance. We would also like to thank St. Elizabeth Hospital of Lafayette, IN, for the donation of the serum samples. We also thank Dow Chemical for financial support of this work. This work was also supported by NIH Grant GM 25431.
RECEIVED for review January 21, 1992. Accepted May 21, 1992.
