CLIN. CHEM. 41/7, 986-990

(1995)

#{149} Automation

and Analytical

Techniques

New Noncompetitive Immunoassays of Small Analytes Un Piran,’

William

J. Riordan,

and Laurie

A. Livshin

We developed a novel noncompetitive immunoassay format for monoepitopic analytes and describe here a model assay for triiodothyronine (13), performed on Ciba Coming’s ACS:180 analyzer. Acridinium ester (AE)-labeled bivalent anti-T3 was incubated with the sample, producing AE-anti-Tjr3 complexes and unreacted AEanti-T3. Controlled-pore glass particles (CPG) with immobilized diiodothyronine (13) were then added in excess, to bind AE-anti-T3 possessing two unoccupied binding sites but not AE-anti-T3 bound to one or two T3 molecules. Paramagnetic particles (PMP) with immobilized anti-AE were then added to the same cuvette to capture AE-anti-T3/13 complexes; AE-anti-T3 bound to the surface of CPG, however, was not captured, because of steric hindrance. After the incubation, the PMP was magnetically separated to remove the liquid phase and the suspended CPG from the cuvette. The chemiluminescence associated with the PMP remaining in the cuvette was then measured. This noncompetitive T3 assay exhibited a 10-fold lower detection limit than the equivalent competitive T3 assay, i.e., 0.3 vs 3 pg/test. Imprecision (CV) in the clinically significant range was 6% or less. The assay also displayed two- to sevenfold lower cross-reactivities and a wider dynamic range. Indexing Terms: paramagnetic particles/controlled-pore glass /triiodothyronirze/chemiluminescence/competitive immunoassays compared Immunoassays can be classified as either competitive or noncompetitive. In the former, the measured signal is associated with antibody that did not bind analyte, whereas the latter involves measurement of signal associated with antibody that bound analyte. Mathematical modeling has shown that, when label detectability is not limiting, noncompetitive assays are potentially more sensitive (i.e., having a lower analyte detection limit) than competitive assays, by orders of magnitude (1). Despite this potential advantage, however, analytes of small molecular mass, which have only one epitope, are usually assayed by the competitive format because of difficulties in separating and measuring the labeled antibody/analyte complex. Previous attempts at overcoming these difficulties produced noncompetitive formats that, because of their complexity, are difficult to automate (2-4). Other noncompetitive formats require unique antibodies and antiidiotypes that are difficult to obtain (5, 6), or in-

volve harsh chemical assay

986

CLINICAL CHEMISTRY, Vol. 41, No. 7, 1995

during

the course of the

We developed a novel noncompetitive format that does not require extra steps, and therefore can be easily performed manually or with existing automated instruments. It is also based on universal reagents that are easy to prepare. Materials

and Methods

Preparation of Reagents antibodies to triiodothyronine (T3) or to ester (AE) were produced in AJJ mice by immunizations and subsequent fusions of the splenocytes with Sp2/0-Ag14 myeloma cells, as described by Kohler and Milstein (8).2 The immunogen for producing anti-T3 was bovine serum albumin-T3 (BSA-T3), prepared as described by Burke and Shakespeare (9). The immunogen for producing anti-AE was keyhole limpet hemocyanin (KLH-AE), prepared with a molecular input ratio of 500:1 .AE:protein, as described by Law et al. (10). The mice were immunized three times with 0.1 mg of immunogen, the first time in complete Freund’s adjuvant and subsequently in incomplete Freund’s adjuvant. Four days before fusion, the mice were immunized intravenously with 0.01 mg of antigen. Spleen cells from the immunized mice were fused with myeloma cells at a ratio of 5:1. Cell culture supernates were screened for tracer binding at 7-21 days after the fusion, when macroscopic colonies were observed. The tracers used in screening for anti-T3 and anti-AE antibodies were [‘251]T3 and AE, respectively, and the solid phase was paramagnetic particles (PMP), to which was bound goat anti-mouse IgG. Hybridoma cells secreting the desired antibodies were injected intraperitoneally into pristane-primed CAF mice. Ascitic fluids from these mice were collected after 3-5 weeks. The antibodies were purified from the ascitic fluid by Protein A column chromatography with an Affi-Gel Protein A MAPS II kit (Bio-Rad Labs., Richmond, CA), according to the protocol provided with the kit. Bovine gamma globulin (BGG) was coupled to Nhydroxysuccininiide-activated N-hemisuccinate methyl esters of L-T3 and L-3,5-diiodothyronine (T2) as described for L-thyroxine (10); the resulting binding ratio was 1.5 iodothyronines per molecule of BGG. BGG-T2 was coupled to CNBr-activated Sepharose 6B (PharmaMonoclonal

acridiium

2

Ciba Corning Diagnostic Corp., 333 Coney St., E. Walpole, MA 02032. ‘Author for correspondence. Fax 508-660-4300. Received December 27, 1994; accepted April 11, 1995.

reactions

(7).

Nonstandard

abbreviations:

AR, acridinium

esters;

PMP,

paramagnetic particles; BSA, bovine serum albumin; BGG, bevine gamma globulins; T2, diiodothyronine; T3, triiodothyrornne; CPG, controlled-pore glass; RLU, relative light units; and TSH, thyrotropin.

cia Biotech, Piscataway, NJ) and used for affinity purification of anti-T3 according to the manufacturer’s instructions. PMP coupled to BGG-T2, BGG-T3, or anti-AE were prepared by the glutaraldehyde method (11). BGG-T2 was coupled to controlled-pore glass (CPG) particles by essentiafly the same method (11). CPG of -1 em diameter, derivatized with aminopropyltriethoxysilane groups, was prepared as described by Weetall and Hersh (12). The immobilizations onto both types of particles yielded -100 mg of protein per gram of particles. The Protein A-purified anti-T3 was labeled with AE as previously described (10). The anti-T3 used in this study is identical to the antibody used by Comeau et al. (13), rather than its lower-affinity counterpart used by Piran et al. previously (14, 15). For most of the experiments, AE-anti-T3 [0.25 mg with 2.5 X 1011 relative light units (RLU)] was affinity-purified by binding the labeled antibody to Sepharose-BG’G-T2 and eluting the purified antibody with 0.1 mol/L glycine buffer, pH 2.5. The affinity purification increased maximal binding to excess (0.25 mg) CPG-BGG-T2 from 90% to 97%. Calibrators were human serum-based calibrators obtained from Magic T3 RIA kits (Ciba Corning Diagnostics, Walpole, MA). Immunoassays

A. Magnetic Separatior Sample

I

4,

AE-anU-T3

CPG-BGG-T2 PMP-antl-AE

I 2.5mm I 4, 4,

Wash out CPG 2.5 mm

2.5 mm

B. Magnetic Separatior Sample AE-antl-T3

I

$

I

4,

2.5 mm

Buffer

4,

2.5mm

PMP-BGG-T2

I 2.5 mm

Wash

4,

YAE

Fig. 1. Schematic representation of the T3 immunoassay formats as performed by the ACS:180 instrument: (A) noncompetitive assay format; (B) competitive assay format. The reaction cuvette moves from left to right in a 37 ‘C water chamber, and the vertical anvws represent the tips of the sample and reagent pipetting probes. The sample probe adds serum sample and a releasing agent, followed by the first reagent probe adding labeled antibody. The cuvette then continuesto move to the right for 2.5 mm, as the binding ofT3 by the antibody approaches equilibrium. In (A), the second reagent probe then adds a large excess of CPG-BGG-T2, which rapidly binds essentially all of the AE-anti-T3 that has two unoccupied analyte-bmnding sites, leaving in solution the AElabeled antibody that bound one or two analyte molecules (14, 15). The third reagent probe adds PMP-anti-AE that, during the next 2.5 mm, captures the AE-anti-Tjr3 complexes in the solution but does not bind the AE-anti-T3 that is bound to CPG-BGG-T2, because of steric hindrance. The cuvette then arrives at the magnetic separation/wash station, where the PMP is attracted to the cuvette wall, and CPG and the aqueous medium are aspirated and discarded. After a water wash, the cuvette is moved to the readout position, where the chemiluminescent signal of AE associated with PMP is measured. In (B), the second reagent probe delivers buffer devoid of CPG-BGG-T2as a

The noncompetitive assay ofT3 was performed with the ACS:180 instrument (Ciba Coring Diagnostics Corp., Oberlin, OH). The sample probe delivered 0.01 mL of calibrator (or sample) and 0.05 mL of 0. 15 mol/L NaOH to the reaction cuvette. Reagent probe 1 delivered 0.1 mL of affinity-purified AE-labeled anti-T3 (activity of 2 X 106 RLU) in buffer A. Buffer A contains, per liter, 0.14 mol of sodium phosphate, 0.2 mol of sodium barbital, 0.04 mol of sodium chloride, 0.01 mol of EDTA, 0.15 g of 8-anilino-1-naphthalenesulfonic control, and the third reagentprobe delivers PMP-BGG-T2at a relativelysmall acid (ANS), 1 g of sodium azide, 0.02 g of BGG, and 2.5 concentration. The final readout involves AE-anti-T3 bound bivalently to the g of BSA, pH 6.6. After a 2.5-mn incubation at 37 #{176}C,PMP. reagent probe 2 delivered 0.25 mg of CPG-BGG-T2 in 0.1 mL of buffer B (per liter, 0.05 mol of sodium Reagent probe 1 delivered 0.1 mL of the same AEphosphate, 0.15 mol of sodium chloride, 1 mmol of labeled anti-T3 in buffer A (see above). After a 2.5-mn EDTA, 0.2 g of sodium azide, and 1 g of BSA, pH 7.4). incubation, reagent probe 2 added 0.1 mL of buffer B; After an additional 2.5-mn incubation, reagent probe 3 after another 2.5 mm, reagent probe 3 delivered 0.005 delivered 0.05 mg of PMP-anti-AE in 0.5 mL of buffer mg of PMP-BGG-T2 in 0.5 mL of buffer B (see Fig. 1B). B. After another 2.5-mm incubation, the PMP was Separation and washes of the PMP and chemiluminesmagnetically separated from the liquid phase, and two cence measurement were automatically carried out by washes were performed, each with 1 mL of deionized the ACS:180 instrument under the same conditions as water. Hydrogen peroxide (0.3 mL of 50 mLfL solution) for the noncompetitive assay. Both types of assay were in 0.1 moIJL HNO3, and 0.3 mL of 0.25 mol/L NaOH run in triplicate. Serum samples were obtained from containing 5 g/L of the catioic surfactant Arquad Ciba Coring Diagnostics Corp. employees, who signed (Akzo Chemical Co., Chicago, IL), were added to the informed consent forms, and passed screening tests for cuvette that had contained the sample, and the chemiHIV, hepatitis B antigen, hepatitis C antigen, syphillis, luminescence produced was detected in the instruand critical blood counts. ment’s photomultiplier tube, expressed as RLU, and recorded (see Fig. 1A for a scheme of the assay). Results The competitive assay of T3 was performed on the Noncompetitive and Competitive Formats Compared same instrument under similar conditions. The sample probe delivered 0.01 mL of calibrator (or sample) and Calibration curves. We ran the T3 assay in both 0.05 mL of 0.15 molIL NaOH to the reaction cuvette. formats and as described in Materials and Methods. CLINICAL CHEMISTRY, Vol. 41, No. 7, 1995

987

The differences between the competitive and noncompetitive formats are shown schematically in Fig. 1. We plotted both of the resulting calibration curves on a linear/linear graph to facilitate direct comparison (Fig. 2). The noncompetitive calibration curve had a low zero-dose signal, followed by an ascending curve that approached a plateau at the highest T3 doses. The competitive format exhibited a typical inhibition curve; i.e., the signal was progressively reduced by increasing amounts of analyte. The two calibration curves appeared to flatten out at about the same T3 concentration, implying that their dynamic range is limited at the high-dose end by comparable T3 concentrations. At the low-concentration end, however, the lowest T3 calibrator (0.25 j.tg/L) in the noncompetitive format caused a change of signal equal to 220% of the signal at zero analyte. In contrast, the competitive format gave only a 24% change of signal for the same T3 calibrators. These results predict a greater sensitivity of 10-fold in the noncompetitive format, provided that the CV of the signal RLU is the same in both formats. Thus, the calibration curves predict an overall broadening of the dynamic range in the noncompetitive format via extension of the low end of the assay. Detection limits. To compare assay sensitivities empirically, we measured the T3 concentration in serially diluted calibrators containing from 1.2 to 0.005 g/L T3. We measured in triplicate the calibrators in five paired sets of runs, using three different ACS: 180 instruments on four different days. The CVs of the signal RLU were similar, i.e., in the 1-2% range, for both assay formats. The results, plotted as a precision profile (Fig. 3), show that an analytical CV of 20% is achieved at concentrations of -0.03 and 0.3 g/L T3 for the noncompetitive and competitive formats, respectively. The corresponding minimal detectable concen-

300000

250000

200000 noncompetitive assay

60% 50% 40% >

30%

competitive 20% 10% 0.03 pg/L 0% 0.010

0.100

1.000

10.000

T3, pg/L

Fig. 3. Precision profile of the serially diluted T3 calibrator, obtained by the (0) noncompetitive and the (A) competitive assay formats. The detection limit is indicated as the concentration at which the CV is 2O%. trations defined as 2 SD of the zero calibrator result were 0.005 and 0.060 g/L. These results confirmed the prediction of a 10-fold enhancement of sensitivity for the noncompetitive assay, as derived from the shapes of the calibration curves. Compatibility with serum samples. Results by the noncompetitive format (y) correlated well (r = 0.99) with those obtained by the competitive format (x) for 46 serum samples, with T3 in the range of 0.1 to 1.8 g/L (mean = 0.57, SD 0.36): y = 0.960x 0.008 (S1 = 0.04). Analytical recovery of T3 added to a serum sample at 0.7, 1.25, 2.5, and 3.75 gfL was 103%, 104%, 93%, and 97%, respectively. Addition of bilirubin (0.1 g/L), conjugated bilirubin (0.2 gIL), intravenous lipid emulsion (30 g/L), or hemoglobin (5 g/L) did not significantly affect the T3 concentration measured in this sample. Assay specificity. Two cross-reactants, T2 and reverse T3, were tested in both assay formats. Surprisingly, the cross-reactivities of the two compounds were lower in the noncompetitive assay than in the competitive assay: T2, 0.02% vs 0.15%; reverse T3, 0.07% vs 0.15%, respectively. -

Alternative Formulations 150000

100000

50000

competitive

assay

2

4

0 0

6

8

T3, pg/L Fig. 2. Calibration curves for T3 immunoassays obtained by noncompetitive (0) and competitive (A) assay formats.

988 CLINICAL CHEMISTRY, Vol. 41, No. 7, 1995

Combining PMP and CPG into a single reagent. To further simplify the assay, we premixed the PMP and CPG reagents and programmed the instrument to add the combined reagents to the cuvette at reagent probe 3. The resulting calibration curve was similar to the one shown in Fig. 2, except that the sensitivity was slightly less because of a small increase in the zerodose signal. The success of this variation may be due to the large excess of CPG (which leads to complete binding of the antibody in -20 s) and a much slower binding by PMP-anti-AE (results not shown). The small increase in the zero-dose signal was probably due to capture of unoccupied labeled antibody by the PMP during the first 20 s of incubation.

250000

200000

-J

the supernate must be transferred to a separate cuvette for the signal development and readout. In addition, the label is detected in the presence of serum sample, assay buffers, and any other reagents transferred from the first reaction cuvette, all of which may cause interferences. Potentially interfering substances can be removed, albeit with additional assay complexity, by adding to the second cuvette a second solid phase designed to capture the labeled immune complex. This is followed by additional separation and wash steps and finally detection of the label attached to the second solid phase (16). We sought to develop a simpler noncompetitive assay format that would not require the transfer of supernate to a second cuvette but still would allow for removal of the sample and assay reagents before readout of the measurement. The novel elements in our approach include: (a) mixing two different particulate solid phases in a single cuvette, followed by efficient and rapid separation from each other, and (b) exploiting the inability of an antibody immobilized on one solid phase to bind labeled antibody that has been bound to another solid phase. In addition, the selection of a CPG-analyte derivative (T2) having low intrinsic affinity to the antibody but high avidity in the immobffized form, through use of bivalent binding (14, 15), obviated the need to prepare a monovalent labeled antibody and increased the functional stability of CPG-BGG-T2. The lower cross-reactivity of T2 and reverse T3 demonstrated in the noncompetitive format may be explained as follows: In a competitive assay, cross-reactivity is dictated by the relative equilibrium constants for the antibody’s binding of the cross-reactant and the analyte. In our new format, the same conditions are met for the first incubation period. However, after adding excess CPG, any unoccupied antibody is captured rapidly and practically irreversibly; therefore, the preformed antibody/analyte and antibody/crossreactant complexes dissociate without significant reassociation. Accordingly, the faster dissociation rates of antibody/cross-reactant complexes lead to proportionally greater loss of signal than in the antibody/analyte complexes. We observed lower cross-reactivities not only in the T3 assay but also in similar noncompetitive assays of digoxin (17) and testosterone (unpublished of

300000

150000

100000

50000

0 0

2

4

6

8

T3, pg/L

Fig. 4. Calibration curves for noncompetitive T3 assays using (A) PMP-immobilized T3 and AE-anti-T3 before affinity purification, and (0) PMP-immobllized

anti-AE with affinity-purified

antibody.

Replacinq PMP-anti-AE with PMP-T3. In another variation on the noncompetitive format we used 0.05 mg of PMP-T3 instead of PMP-anti-AE and did not affinity-purify the AE-anti-T3 on a column of immobilized T2; otherwise, the assay format was identical to that described in Fig. 1A. The rationale for this variation was that one unoccupied antibody-binding site might have sufficient affinity for immobilized T3 to form a stable bond. As shown in Fig. 4, we were able to generate a calibration curve for which zero T3 gave 7706 RLU and the 0.25 gfL T3 calibrator gave 44589 RLU, for a 480% change. Accordingly, the predicted sensitivity was enhanced -20-fold over that of the competitive assay. The disadvantage of this approach, however, is that the calibration curve obtained exhibited an earlier plateau and a “hook effect,” probably because analyte concentrations >1 gfL were occupying the second binding site of the labeled anti-T3. Results for this variation also support the mechanism we proposed for the assay because the increase in signal with the addition ofT3 can be explained only by the formation of a PMP-BGG-T3/AE-anti-T3/T3 ternary complex. Discussion

The feasibility of noncompetitive immunoassays based on a single epitope was first demonstrated by Miles and Hales, using insulin as a model (2). These workers incubated labeled anti-insulin with samples, then used immobilized insulin to remove the unreacted labeled antibody and subsequently measured the label remaining in the supernate. Similar assays were later developed to use analyte derivatives immobilized on a column or on particles (3, 4). The disadvantage of this approach is that, after the separation step, an aliquot

results). For direct comparison of the sensitivities of the two assay formats, we used the same labeled antibody and almost identical assay conditions. The 10-fold enhancement in sensitivity by the noncompetitive format directly supports the theoretical prediction of Jackson and Ekins (1). An even greater enhancement in sensitivity will require refinements in antibody purification, careful selection of immobilized analyte analogs, and optimization of the relationship between incubation time and dissociation rate constants of the labeled antibodies. The three variations we described for the new assay format demonstrate its flexibility and lend support for the mechanisms of action we proposed for the various CLINICAL CHEMISTRY, Vol. 41, No. 7, 1995

989

stages of the assay. We have also developed similar noncompetitive assays for digoxin, free T3, and thyrotropin (TSH), with CPG-immobilized digitoxin, T2, and an anti-TSH antudiotype, respectively (17), as well as estradiol, testosterone, free thyroxine, and melatonin (unpublished results); these findings imply that the basic format may be adaptable to a wide variety of analytes. The new assay format should allow routine measurement of a wide variety of small analytes present in body fluids at concentrations below those detectable by competitive immunoassays. This improved sensitivity allows use of a smaller sample size and yields higher precision. This format also increases the dynamic range of the assay, thus obviating the need for sample dilutions and repeated testing. In addition, cross-reactivities and potentially other interferences are apparently reduced. We thank E. Barlow for preparing the monoclonal antibodies, H. Lukinaky for preparing activated iodothyronines, and J. Unger for his support and helpful discussions. References

1. Jackson T, Ekins R. Theoretical limitations on immunoassay sensitivity. J Immunol Methods 1986;87:13-20. 2. Miles LAM, Hales C. Labelled antibodies and immunological systems. Nature 1968;219:186-9. 3. Freytag JW, Dickinson JC, Tseng SY. A highly sensitive affinity-column-mediated immunometric assay, as exemplffied by digoxin. Clin Chem 1984;30:417-20. 4. Grenier FC, Granados EN, Schick BC, Kolaczowski L, Pry TA. Enhanced sensitivity immunoassay for the TDx analyzer [Tech Briefi. Clin Chem 1987;33:1570.

990

CLINICAL CHEMISTRY, Vol. 41, No. 7, 1995

5. Self CH, Desse JL, Winger LA. High-performance assays of small molecules: enhanced sensitivity, rapidity, and convenience demonstrated with noncompetitive immunometric anti-immune complex assay system for digoxin. Clin Chem 1994;40:2035-41. 6. Barnard G, Kohen F. Idiometric assay: noncompetitive immunoassay for small molecules typified by the measurement of estradiol in serum. Clin Chem 1990;36:1945-50. 7. Ishikawa E, Hashida 5, Kohno T. Development of ultrasensitive enzyme immunoassay reviewed with emphasis on factors which limit the sensitivity. Mol Cell Probes 1991;5:81-95. 8. Kohler G, Milstein C. Continuous cultures of fused cells secreting antibodies of predefined specificity. Nature 1975;256: 495-7. 9. Burke CW, Shakespeare RA. Rapid purification of triiodothyronine and thyroxine protein conjugates for antibody production. J Endocrinol 1975;65:133-8. 10. Law SJ, Miller T, Piran U, Kiukas C, Chang S, Unger J. Novel polysubstituted aryl acridinium esters and their use in immunoassay. J Biolumin Chemilumin 1989;4:88-98. 11. Groman EV, Sullivan JA, Marinac JE, Neuringer IP, Finlay CA, Kenny FE, Josephson L. Bio-Mag-a new support for magnetic affinity chromatography. Biotechniques 1985;3:156-60. 12. Weetall HH, Hersh LS. Urease covalently coupled to porous glass [Short Commun]. Biochim Biophys Acts 1969;185:464-5. 13. Comeau L, Piran U, Leo-Mensha T, Huntress M, Hudson T. An automated chemiluminescence test for total triiodothyronine [Abstract]. Clin Chem 1991;37:941. 14. Piran U, Silbert-Shostek D, Barlow EH. Role of antibody valency in hapten-heterologous immunoassays. Clin Chem 1993; 39:879-83. 15. Piran U, Riordan WJ, Silbert DR. Effect of hapten heterology on thyroid hormone immunoassays. J Immunol Methods 1990; 133:207-14. 16. Baler M, Jering H, Klose S. Immune-chemical measurement process for haptens and proteins. US patent no. 4,670,383;1987. 17. Piran U, Riordan WJ, Livshin LA. Method for noncompetitive binding assays. Eur patent application no. 94306702.5, publication no. 0 643 306 A2; 1995.