2708
Anal. Chem. 1987, 59, 2786-2789
Rapid Heterogeneous Competitive Electrochem ca Immunoassay for IgG in the Picomole Range Uditha de Alwis a n d George S. Wilson* Department of Chemistry, University of Arizona, Tucson, Arizona 85721
The use of an immunosorbent microreactor for a competitive heterogeneous immunoassay is demonstrated. An Fab' fragment of an an#body Is attached covalently, largely via the -SH molety in the hinge region, to a polymeric support (Trisacryi GF-2000). An lmmunosorbent results which retains about 75% of the antlgen blndlng capacity of the soluble moiety. The flow injection Immunoassay for human IgG can be carried out wHh f2-3% accuracy and precidon in the subplcomole range. The total t h e taken for the assay Is 12 mln and the between-assay interval is 8 min. The technique is compatible with total automation.
Immunoassays enjoy widespread popularity in the clinical determination of drugs, metabolites, steroids, hormones, proteins, and a host of other animal and plant products. Heterogeneous enzyme-linked immunosorbent assay (ELISA) techniques involve the use of a solid-phase adsorbent to separate the bound and unbound analyte in order to determine the extent of reaction. The use of immunosorbents for such an assay has proven to be a problem in many cases. Typically, the immunosorbent is prepared by adsorbing an antibody or antigen onto polystyrene or polypropylene multiwell plates. These plates are treated by the manufacturer and are optimized for protein adsorption. Unfortunately there are variations in the plastic surface characteristics which cause variable adsorption in different wells within the same plate as well as in different plates. The protein loading can vary from lot to lot on thew plates (1). In addition, since this is a physisorption process there is no control over the overall orientation of the antibody on such surfaces. The binding site of the antibody occupies only a fraction of its total area. Binding to a large protein would require unhindered access to the target site. The use of randomly oriented physisorbed antibodies in the determination of relatively large protein molecules can prove disadvantageous under these conditions. Conventional ELISA assays frequently require long incubation periods. This is not very surprising because the multiwell plate involves a nonstirred system depending primarily on diffusion to bring the sample in contact with the immobilized reagent. The plastic can adsorb approximately 0.2-4 pmol/cm2 (2). The small number of binding sites and the dilute nature of the sample can further lead to slow or incomplete reaction. The column methodology described in this paper involves much higher effective concentration and much faster reaction rates are obtained. In addition, the same forces that promote the immobilization of the reagent on the plate can cause other proteins to stick nonspecifically unless care is taken to block such active sites after the initial adsorption. Protein has been found to leak from plates during long incubations and displacement of one protein by others has also been observed (3).
We have shown ( 4 ) that both antibodies and antigens covalently attached to chromatographic supports, packed into *Present address: Department of Chemistry, University of Kan-
sas, Lawrence, KS 66045.
reactors, and coupled to flow injection analysis (FIA) system with electrochemical detection can be used to carry out immunoassays with great precision and accuracy. In this report we wish to present the use of antibody fragments in conjunction with covalent coupling to prepare immunosorbents with reproducible and high biological activity. By use of these imunosorbents as previously mentioned (4) a competitive assay could be carried out for large molecules in the subpicomole range in 12 min. The accuracy and precision of these assays are far superior to conventional assays.
EXPERIMENTAL SECTION Materials. Human IgG (HIgG) was purchased as "Gammastan" (Cutter Laboratories, Berkeley, CA) at a local drug store and was separated from stabilizers on a Sephadex G-25 column prior to use in coupling procedures and in preparation of standards equilibrated with 0.15 M NaC1. Anti-human IgG (goat) was a generous donation from American Qualex, La Mirada, CA. Reactigel 6X was obtained from Pierce Chemical Co., Rockford, IL, and used according to the manufacturer's instructions. Glucwe oxidase type X-S and pepsin (porcinemucosa) 1:60000 were obtained from Sigma Chemical Co., St. Louis, MO. Sephadex G-25 and Sephacryl S-300 SF were obtained from Pharmacia Fine Chemicals,Piscataway, NJ. Tris Acryl GF-2000 (LKB Produkter AB, Broma, Sweden.) was a donation from Bio-Probe International, Tustin, CA. 2,2,2-Trifluoroethanesulfonyl chloride (Tresyl Chloride) was purchased from Fluka Chemical Corp., Ronkonkoma, NY. Dithiothreitol (Cleland's reagent) was obtained from Research Organics, Inc., Cleveland, OH. All other reagents mentioned are of analytical grade unless stated otherwise. All water used in the FIA buffers was triply distilled, filtered through a 0.45-fimfilter, and protected from dust. Preparation of Tresyl Activated Trisacryl GF2000. Activation of this gel was carried out according to the method of Mosbach and Nilsson (5). The gel was then stored in 1mM HC1. It was washed with excess 0.1 M phosphate-bufferedsaline (PBS), pH 6.0, containing 10 mM ethylenediaminetetraacetic acid (EDTA). Preparation of Fab' Fragments. Anti-human IgG antibody was subjected to pepsin cleavage (6) and affinity purified against human IgG immobilized on Reactigel6X. Concentrated affinity purified fractions of F(ab)zfragments were dialyzed against 0.1 M PBS pH 6.0,lO mM EDTA overnightand the sample was made 10 mh4 with respect to dithiothreitol. After incubation in the dark for 1 h at room temperature in an atmosphere of nitrogen, the Fab' fragmenta formed were separated on a SephadexG-25 column equilibrated with the nitrogen saturated PBS/EDTA buffer mentioned above. The fragments were then rapidly transferred to a vial containing prewashed Tresyl-activatedTrisacryl GF-2000. The vial was purged with nitrogen, sealed and rotated end over end at 4 "C for 24 h. After this period, the gel was washed with 0.2 M Tris-HC1pH 8.2 and allowed to stand overnight at 4 "C in the Tris buffer to block any unreacted sites. The gel was then packed into a stainlesssteel reactor 0.2 X 4 cm fitted with standard 1/2-in. Swagelok fittings and connected to the FIA system. The reactor was cycled back and forth between 0.1 M PBS pH 7.4 and 0.1 M phosphoric acid pH 2.0 for 5 cycles, each cycle being 10 bed volumes. This treatment removes any trapped and unstable antibody from the gel matrix rendering the gels stable for long periods of use. This immunosorbent has been stored at 4 O C for over 2 years with no appreciable loss of immunological activity. Preparation of Human IgG-Glucose Oxidase Conjugates. The conjugate of HIgG-GOD was prepared by using a modifi-
0003-2700/87/0359-2786$01.50/0 0 1987 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 59, NO. 23, DECEMBER 1, 1987
2787
a
t X
e
a
pH 2.0
1.00
3. 00
5.00
7.00
9. 00
11.00
pH6.8
Nanogram8 Conjugate Injected
Figure 1. Reactor capactty determination using a loading curve.
cation of the method of Ternynck and Avrameas (7). Glucose oxidase (20 mg/mL in 0.1 M phosphate buffer (PB) pH 6.8) was activated with 200 pL of p-benzoquinone (PBQ) (15 mg/mL EtOH) (recrystallized from a medium boiling petrolium ether fraction, yellow crystalline, mp 112-114 OC) in the dark at 37 O C for 1h. The molar ratio of PBQ:enzyme at this point is 117:l. It was then mixed with 240 pL of HIgG (unpurified Gammastan) and was passed through a Sephadex G-25 column (1.5 X 25 cm) equilibrated with 0.15 M NaC1. The reddish brown fractions that eluted first was collected and was mixed with 1M NaHC03 (101 v/v) and stirred for 72 h at 4 OC. The ratio of PBQ:enzyme was 3.41 at this step. Blocking of the u n r e a d sites was carried out with 1mL of 1M lysine pH 8.5 for 1h at room temperature. The fraction of interest was collected and concentratedafter passage through a Sephacryl S-3OOSF column (1.5 X 60 cm) at a flow rate of 0.33 mL/min. The final concentration of the conjugate was adjusted to 2 mg/mL. Determination of Reactor Capacity. The reactor capacity was determined in two ways. In the first case a minireactor was packed with 1mL of the settled immunwrbent. After the reactor was washed with 10 mL of 0.1 M PBS pH 6.8, 5 mL of HIgG solution (2 mg/mL) was passed over the reactor bed and the immunosorbent was washed with 10 volumes of the same buffer as before. The absorbance at 280 nm was tested in the final fraction to estimate protein concentration. If this was less than 0.005 mg/mL it was assumed that all unbound protein was removed. The material bound was eluted with 0.1 M phosphoric acid pH 2.0. The eluate was dialyzed against physiological buffer and the protein was determined by measuring the absorbance at 276 nm. The second method involved connection of the actual reactor used for the experiments to the FIA system and injection of progressively larger samples of the conjugate. The immobilized conjugate was determined by injecting a known amount of glucose and measuring the amount of hydrogen peroxide produced. The amount of conjugate needed to cause site saturation as shown in Figure 1was considered proportional to the number of binding sites in the reactor. Both methods indicate that the reactor capacity corresponded to approximately 65 nmol of binding sites/mL of settled gel. Preparation of Samples for Analysis. Serial dilutions of HIgG were made in buffer or control serum. To each standard, 10 nmol of conjugate was added. Fresh standards were prepared every day. Recovery studies were carried out by spiking control serum with HIgG. Apparatus. The apparatus used in this work is described elsewhere (3). A hydraulic pulse damper (Scient& Systems, Inc., Model LP-21 LO-Pulse) in addition to the small silica backpressure column was added to improve flow rate control. This, however added a 600-900 pL dead volume to the system, slowing the assays down by the time mentioned. The reactor was washed for 10-15 min at the start of each day with the assay buffer and the elution buffer (pH 2.0) was passed through the system for 2 min. The flow was switched to the assay buffer and the system was equilibrated for 9 min. The first sample was then injected (20 pL). The reactor was allowed to wash for 3 min followed by three injections of 0.1% glucose solution at 3-min intervals. As
- Gwt
I
Anti-Human IpG Fab'
- 7ris-Ocr"l
GF 2000
p@ +
Humon IpG
- Clu. Ox
-Humen IgC
Figure 2. Schematic for competitive ELISA.
the third injection of glucose was made the pH was changed to 2.0 and run for 2 min, after which the system was switched back to the away buffer. Since there is a 5-min dead time in the system, this method of pH variation resulted in a time saving of 4 min.
RESULTS AND DISCUSSION In this assay the sample and standards are mixed with the labeled anal* present in a constant and optimized concentration such that by itself the labeled molecule could saturate all of the immunwrbent binding sites. This therefore causes any analyte molecules in the sample to compete for available sites on the immunosorbent As expected this causes a decrease in the number of labeled analyte molecules bound as the number of unlabeled analyte molecules increase (Figure 2). The ideal immunosorbent should be lightly loaded such that it can easily be saturated with labeled molecules but yet should have sufficient sites to allow a reasonable dynamic measurement range. Trisacryl GF-2000 was selected for this assay primarily because it had a very high exclusion limit (2000 kdalton). This large pore size is important in reducing nonspecific interactions. Reactors packed with thismaterial showed a nonspecific signal in the order of 1%or less. Similar readors with smaller pore diameters showed nonspecific signals of 3-5% of the specific signal. This was measured by injecting a conjugate of goat IgG-GOD similar in concentration to the sample and carrying out a determination. The activation level was low compared to Fractogel and cross-linked Agarose. This has been observed by other workers (8). This however worked to our advantage. For this experiment a very highly loaded immunosorbent (8-10 mg/mL settled gel) was undesirable. From protein difference spectra at 280 nm before and after coupling it was calculated that there was 2.0 mg Fab' fragment/mL of settled support attached to the immunosorbent and it could retain 1.5 mg/mL of the antigen. This amounts to 75% activity of the immobilized antibody. These results seem to indicate that success is achieved in immobilizing most of the antibody fragments via the thiol group located in the hinge region of the antibody. Binding through an amine function has an excellent chance of blocking or modifying the antibody binding site. Careful pH control of the coupling reaction between the Fab' fragment and the tresyl activated support favors coupling through the thiol group. Holding the pH at 6.0 protonates the free amino groups rendering them unavailable for coupling while the thiols are completely accessible. It is extremely important to work under anaerobic
2788
ANALYTICAL CHEMISTRY, VOL. 59, NO. 23, DECEMBER 1, 1987
FLr
'1
T
4.0cm
a
t-,
I
0.2 cm C
-re
3. Schematic for competitive reactors: a, undiluted; b, uniform
dilution; c, sandwich packing. conditions. Low yields would be obtained if Fab' fragments dimerized through oxidative coupling to form -S-S- bonds. The addition of EDTA is believed to complex metals, thus inhibiting their ability to catalyze oxidation. It does not however completely stabilize the material. There are two methods for making a low-capacity reactor. The first is to use a extremely small reactor. We have tested a 0.2 X 0.2 cm reactor. The residence time of a 20-pL sample in this reactor is less than 2 s. This leads to a situation where the extent of reaction is less than 40%. The width of a substrate peak is less than 1 min. This includes dispersion in all elements of the system. This reactor seemed useful for the assays described. The reactor however is prone to shifts in flow patterns. The support swells and shrinks with pH and ionic strength changes. These microscopic changes in the support seemed to promote channeling. When channeling occurs the reactor loses its specific binding properties and there is a large loss in binding activity. The loss of activity is in the order of 90% or greater. If the reactor is unpacked carefully and repacked with the same material it is possible to achieve the same activity as before. An immunosorbent prepared as mentioned previously has about 2-10 mg IgG/mL of wet gel. This immunosorbent has a binding capacity too high for use in a competitive assay without being diluted. Dilution is carried out by taking a heavily loaded immunosorbent and mixing with the same support, which has not been activated. This can be easily accomplished by volume dilution of support suspensions. The typical dilution for a loading of 10 mg of antibody/mL of settled immunosorbent is in the range of 1:400 to 1:lOOO. The second method involves the preparation of very lightly loaded immunosorbent. Lightly loaded immunmrbenta do not function well. The low loading does not allow the law of mass action to favor the reaction. This is also the problem with mixing the immunosorbent with unactivated gel to prepare diluted reactors. In addition to showing weak interactions the diluted immunosorbentreactors show variable binding due to slight changes in the flow patterns. This problem can be easily overcome by laying down a layer of the highly loaded immunosorbent in the reactor between layers of unactivated gel. In this c o n f v a t i o n , since the immunosorbent is not near either the entrance or the exit of the reactor, it is insulated from the changes in bed volume and is more resistant to shifting. In addition the sample has dispersed sufficiently in the reactor to present an uniform front to the immunosorbent. Since the immunosorbent layer is a compact mass with a high local concentration of binding sites, mass action favors rapid and complete reaction. The work described in this publication was carried out in a 0.2 x 4 cm reactor prepared by laying down 10 pL of a 1.25 (v/v) diluted immunosorbent in a layer as described before. It is
I '
,
I 1
1
I
3
I
I
5
4
TIMC(~OUIIB)
Figure 4. Size exclusion chromatogram for conjugate: a, initial reaction mixture before purification (10 pL sample on Sephacryl S-300 SF, 1.5 X 60 cm column, 0.01 M PBS pH 7.4); b, after purification on the same column. Table I. Stoichiometry of the Different Conjugate Fractions fraction no. I I1 I11 IV
stoichiometrp free Ab and E
Ab-E Ab2-E Abs-E and Ab2-E2 "Ab = antibody. E = Enzyme.
re1 %
38.8 20.7 26.6 13.8
clear from a thermodynamic point of view (9) that the diluted, lightly loaded, and sandwich reactors have identical binding site concentrations. However kinetic considerations seem to favor the sandwich configuration. The different column configurations are depicted in Figure 3. The competition for the sites should be between the labeled analyte and the unlabeled analyte. The labeled analyte should not contain any unreacted analyte molecules. The presence of unlabeled analyte would cause the assay to lose its sensitivity. Removal of the unreacted analyte from the conjugate was achieved by size exclusion chromatography. Before the separation was carried out, fractions from the elution profile (Figure 4a) were collected and enzyme activity and total protein measurements were carried out according to the method described by Weibel and Bright (IO). From the ratio of AdSO:AZSO the enzyme:antibody ratio was calculated. The stoichiometries obtained are shown in Table I. Fractions I11 and IV were collected, concentrated, and rechromatographed on the same column. The amount of unreacted antibody and enzyme remaining was found to be less than 1% (Figure 4b). As expected, the assay was highly dependent on the conjugate concentration and its purity. The dynamic range of the assay is a function of the total number of binding sites on the reactor. Once this number is ascertained as mentioned previously, the required conjugate concentrationis established. The use of this amount of conjugate usually results in assays that have wide dynamic range but low precision. With an increase of the conjugate concentration, it is possible to increase the sensitivity of the assay. By doing so the assay loses some dynamic range. With a decrease of the conjugate concentration, it is possible to extend the range to higher analyte concentrations. This, however, deteriorates the precision and
Anal. Chem. 1987, 59, 2789-2794
COMPETITIVE 1.30
ASSAY
F O R HUMAN
falls off at higher concentrations as expected. The accuracy of the measurement is in the *2% region. The assays have been repeated for over 2 weeks a t the rate of 30 assays per day (600total)and the decrease in the immunological activity is less than 5%. The immunosorbent has been stored in the refrigerator for 2 years at 4 OC and the decrease in immunological activity is negligible. In summary the FIA system coupled to an immunoreactor with a thin-layer electrochemical detector forms a basis for carrying out immunoassays rapidly with a high degree of accuracy and precision.
IGG
I1
I
n 1s I. I -
0.00
2. 50
5 . 00
7. so
12.50
10.00
Plcomoler HlgQ Added
Flgure 5. Calibration curve for the human IgG competitive assay.
Table 11. Recovery of Human 1%: from Control Serum
amt taken, amt detected,” pmol pmol 0.5 2.0
0.525 (*1%) 2.1 (&2%)
amt taken, amt detected: pmol pmol 5.0 10.0
2789
5.11 (&3%) 10.5 (&5%)
a Unknown samples were prepared by spiking control serum with known amounts of human gamma globulin.
accuracy due to the flattening of the curve. A calibration curve for this assay is shown in Figure 5. A dynamic range of a factor of 20 is obtained. The competitive assay has an inherently much shorter range than the sandwich assay and is much more tedious to optimize. The results of a well-optimized assay are shown in Table 11. The precision of the assay
ACKNOWLEDGMENT We thank American Qualex International, Inc., for generous donations of materials. LITERATURE CITED Shekarchi, I . C.; Sever, J. L.; Lee, Y. J.; Castellano, 0.;Madden, D.L. J . Clin. Microbiol. 1984, .79(2), 89-96. Cantarero, L. A., Butler, J. E.; Osborne, J. W. Anal. Blochem. 1980. 105, 375-382. Kenny, G. E.; Dunsmoor, C. L. J . Clin. Microbiol. 1983, 77(4), 655-665. De Alwls, U.; Wilson, G. S.Anal. Chem. 1985, 57, 2754-2756. Nllsson, K.; Mosbach, K. Methods Enzymol. 1984, 704, 56-69. De Alwls, U.; Hill, 6. S.; Meiklejohn, B. 1.; Wilson 0. S. Anal. Chem. 1987, 59, 2688-2691. Ternynck T.; Avrameas, S. Ann. Immunol. (Paris) 1976, 727C, 197-208. Miron, T.; Wilchek, M. Appl. Biochem. Biofechnol. 1985, 1 1 , 445-456. Sportsman, R. J.; Wllson, G. S.Anal. Chem. 1980, 52, 2013-2017. Weibel, M. K.; Bright, H. J. J . Biol. Chem. 1971, 249(9), 2734-3739.
RECEIVED for review March 23,1987. Accepted July 15,1987. We thank the National Institutes of Health (Grant DK 30718) for financial support.
Magnesium Excitation Mechanisms and Electronic-State Populations in an Argon Inductively Coupled Plasma Tetsuya Hasegawal and Hiroki Haraguchi* Department of Chemistry, Faculty of Science, University of Tokyo, Bunkyo-ku, Tokyo 113, Japan The excitation mechanlsms of magnesium in an argon inductlvely coupled plasma have been investigated by using a coiildonakadlative process theory, which includes the Penning lonlzatkn and charge transfer reactlaw in addltion to the electron impact and radiative processes. The calculated population denslty distributions in the electronic states of atoms and ions showed markedly large deviations from local thennodynamk equlllbrlum, Le., overpopulations of ail atomic levels and lower lonlc levels. The overpopulatlons of these levels are interpreted by 8IgnWlcant spontaneous emisslon and negllgble radlathre absorption In the case of magnesium. The discrepancy between the present resuit and prevlous observations, Le., overpopuiatlon of ions, is attributed to the dlfference of temperatures used for the LTE calculation. From the comparison of the reaction rates for the ionization processes, R has been concluded that the electron impact processes are predominant for magnesium exclation/ionlzation rather than the Penning lonlzation and charge transfer reactions.
Present address: Department of Industrial Chemistry, Faculty of Engineering, University of Tokyo, Bunkyo-ku, Tokyo 113,Japan. 0003-2700/87/0359-2789$01.50/0
In recent years, an inductively coupled plasma (ICP) has been extensively used in analytical atomic spectroscopy as atomization (I), excitation (21, and ionization (3) sources. In particular, the ICP is an efficient excitation source in attomic emission spectrometry which provides the excellent analytical feasibilities such as good sensitivity, good stability and precision, less interelement interferences, wide dynamic ranges, multielement detection capability, and so forth (4-6). It is also well-known from the spectroscopic studies that ionic lines for most of analyks generally provide much better sensitivities than their atomic lines. According to the previous investigations (7), it has been considered that such spectroscopic characteristics of the emission lines of analytes in the argon ICP result in deviation from LTE (local thermodynamic equilibrium) and may be appreciated as the overpopulations of ionic species in the ICP, which are often called “suprathermal ionization”. Actually the non-LTE properties of the argon ICP, on the other hand, give analytical advantages of ICP-AES (inductively coupled plasma atomic emission spectrometry) mentioned above. In order to interpret the non-LTE phenomena of the ICP, various models for excitation and ionization mechanisms have been proposed by many research groups (8-19). 0 1987 American Chemlcal Soclety
