Anal. Chem. 1995,67,3211-3218
Immunoassay for
Capillary Electrophoretic Digoxin in Human Serum Xuhui Liu and Yan Xu* Department of Chemistfy, Cleveland State Univesify, Cleveland, Ohio 44 115
Michael P. C. Ip Department of Pathology, MetroHealth Medical Center, Cleveland, Ohio 44 109
The combined use of capillary electrophoretic (CE) separation and homogeneous enzyme immunoassay for analyzing drugs in hemolyzed, lipemic, or icteric serum samples was investigated. An FDA-approvedEMlT assay kit for digoxin in human serum was used. After the enzyme immunoassay, the enzymatic reaction product (NADH) and remaining substrate (NAD+),together with internal standard (p-nitrophenol, NP),were electrokinetically injected into a polyacrylamide-coatedelectrophoresis capillary and separated under applied potential. Detection was made by monitoring the UV adsorption at wavelengffiof 260 nm. The digoxin level in human serum was determined by comparing the peak area ratio of NADH and NP to the ratios established by the known digoxin standards. In this study, the factorsthat influence the CE separation were also investigated. Under the optimum conditions, NADH, NAD+, and NP were separated at electric field strength of 438 V/cm in the coated capillary (100 pm x 57 cm)with 200 mM Tris-borate b d e r (pH 7.9) containing 0.2% hydroxypropyl methylcellulose. CE analyses of serum samples spiked with NADH standards at concentrations of 100 and 400 pM resulted in detection variabilities of less than 2% and analytical recoveries of 98-102%. Both an internal calibration plot for NADH and a dose-response curve for digoxin in serum were constructed. Calibrator serum, patients’ sera with hemolyzed, lipemic, and icteric interference factors, and other pigmented blood components (e.g., serum albumin, bilirubin, hemoglobin, uric acid, coproporphyrin, melanin, protoporphyrin IX, and uroprophyrin) demonstrated no interference in this method. The authors believed that this method is useful for analyzingdigoxin in hemolyzed, lipemic, and icteric blood samples that are known to create problems in conventional EMIT assays and may be applicable to other EMITbased assays for monitoring drugs in complex biological matrices. Since its advent in the late 1 9 5 0 ~immunoassay ,~~~ has become a primary analytical tool not just in clinical diagnostics but also in many other applied scientificfields such as food and environmental ana lyse^.^ The methodology is mainly based on the inherent (1) Berson, S. A; Yalow, R S.; Bauman, A; Rothschild, M. A; Newerly, K. /. Clin. Invest. 1956,35, 170-90. (2) Yalow, R. S.; Berson, S. A Nature 1959,184, 1648-9.
0003-2700/95/0367-3211$9.00/0 0 1995 American Chemical Society
chemical specificity of an immunologicalreaction and the exquisite sensitivity in detecting labeled antigens or labeled antibodies, which allow fast, accurate, and precise quantitation of a variety of analytes at very low concentrations in complex sample matrices. Quantitation of an immunological reaction often uses label molecules. Immunolabels can be divided into two classes: isotopic and nonisotopic. Most early immunoassaysused isotopic labels such as 1251,57C0, 3H,14C, and 32P. Due to health and environmental concerns, nonisotopic labels have become more The nonisotopic labels include widely used in enzymes, chemiluminescenceand bioluminescence compounds, metal complexes, and electrochemically active molecules. In the case of enzyme immunoassay, the enzyme molecule is used as the label that converts enzyme substrate into its product at a relatively high reaction rate. As the result of enzyme amplification, a signiscant amount of product can be produced for final detection. Therefore, assays that use enzymes as labels usually have excellent limits of detection. One of the assay formats used frequently in therapeutic drug monitoring is known as the homogeneous enzyme immunoassay. It requires minimum sample handling and is well suited for measuring low molecular weight molecules in a concentration range of micrograms to milligrams per liter. The principle of homogeneous enzyme immunoassay can be illustrated by the technique known as the “enzyme-multiplied immunoassay technique” (EMIT),which is based on the competition between antigen in the sample and enzyme-labeled antigen for limited antibody-binding sites. Upon binding of enzyme-labeled antigen to the antibody, enzyme activity is inhibited or altered through steric exclusion of the enzyme substrate or conformational change of enzyme m ~ l e c u l eso , ~ the antigen concentration in the sample can be measured in terms of enzyme activity. One of the challenges in developing homogeneous enzyme immunoassays is to detect the enzyme product in whole blood or in samples that are hemolyzed, lipemic, and icteric6 Previous (3) Van Emon, J. M.; Lopez-Avila,V. Anal. Chem. 1992,64, 79A-88A 1359-60. (4) Petterson, K. Clin. Chem. 1993,39, (5) Kricka. L. J. Clin. Chem. 1994,40, 347-57. (6)Xu, Y.; Halsall, H. B.; Heineman, W. R In Immunochemical Assays and Biosensor Technologyfor the 1990s;Nakamura, R M., Kasahara, Y., Rechnitz, G. A, Eds.; American Society for Microbiology: Washington, DC, 1992; pp 291-309. (7) Schuurs, A H. W. M.; Van Weemen, B. K. Clin. Chim. Acta 1977,81, 1-40. (8) Thompson, S. G. In Clinical Chemisty; Kaplan, L. A,, Pesce, A. J.. Eds.; The C. V. Mosby Co.: St. Louis, MO, 1989 pp 191-206.
Analytical Chemistry, Vol. 67,No. 18, September 15, 1995 3211
effortsgJOhave been made by using chromatographic separation with amperometric detection. In spite of the achievements in circumventing the matrix effect of biological samples, the method required a switching between size exclusion and reversed-phase columns. Therefore, it complicated the routine operation. Recent advances in capillary electrophoresis (CE)11~12allow the separation of both charged and neutral molecules with very high efficiencies. Several studies13-15have demonstrated the applicability of capillary electrophoresis for separation of antibody-antigen complexes from unbound antibodies and antigens by direct injection of reaction mixtures into the separation capillaries. With laser-induced fluorescence detection, both Schultz and Kennedy and Chen and Sternberg were able to achieve the detection limits needed for monitoring insulin14and digoxin15in the therapeutic ranges. However, the direct injection of protein mixture onto the separation capillary often caused adsorption on the inner surface of the capillary, which affects the reproducibility of the methods and seriously compromises their reliability in quantitative analysis. The present study was aimed at coupling CE with homogeneous enzyme immunoassay for quantitative analyses of drugs or drug metabolites in hemolyzed, lipemic, or icteric serum samples. Digoxin, a widely used cardiac active drug for treatment of heart failure, atrial fibrillation, and supraventricular tachyarrhythmias was chosen as the model analyte. Because of its narrow therapeutic range (0.5-2.0 pg/L) ,I6 it is clinically important to precisely monitor its level in sera. In our study, an FDAapproved EMIT assay kit for digoxin was used. After the enzyme immunoassay, the product of enzymatic reaction (NADH), together with the remaining substrate (NAD+) and an internal standard (p-nitrophenol, NP), was electrokinetically injected into a capillary and separated by CE. Detection was accomplished by monitoring W absorbance at a wavelength of 260 nm. The digoxin concentration in human serum was determined by measuring the peak area ratio of NADH and NP using internal calibration. The factors that influence CE separation in both coated and uncoated capillaries were also investigated and compared. Calibrator serum, patients' sera containing hemolyzed, lipemic, and icteric interfering factors, and other blood components were tested and showed no interference in this method. The authors believe that this approach is valuable for quantitation of drug molecules in complex biological matrices. EXPERIMENTAL SECTION
Reagents. The EMIT digoxin assay kit (Lot 6H419ULF1) was obtained from Syva Co. (SanJose, CA), which consisted of reagent A, reagent B, serum pretreatment reagent, buffer concentrate, and digoxin calibrators. p-Nicotinamide adenine dinucleotide (NADt), reduced ,&nicotinamide adenine dinucleotide (NADH), and calibrator serum and diluent (Catalog No. 620213) were purchased from Boehringer Mannheim Corp. (Indianapolis, IN). fi-Nitro(9) Eggers, H. M.; Halsall, H. B.; Heineman, W. R Clin. Chem. 1982,28,184851. (10) Wright, D. S.; Halsall, H. B.; Heineman, W. R Anal. Chem. 1986,58,2995-
8. (11) Xu, Y. Anal. Chem. 1993,65, 425R-33R. (12) Monnig, C. A; Kennedy, R. T. Anal. Chem. 1994,66, 280R-314R (13) Nielsen, R G.; Rickard, E. C.; Santa, P. F.; Sharknas, D. A; Sittampalam, G. S.J. Chromatogr. 1991, 539, 177-85. (14) Schultz, N. M.; Kennedy, R. T. Anal. Chem. 1993, 65, 3161-5. (15) Chen, F. T. A.; Stemberg, J. C. Electrophoresis 1994, 15, 13-21. (16) Chen, I. W. In Clinical Chemistry; Kaplan, L. A,, Pesce, A J., Eds.; The C. V. Mosby Co.: St. Louis. MO, 1989 pp 1088-91.
3212
Analytical Chemistry, Vol. 67, No. 78, September 75, 7995
phenol, bovine albumin, human hemoglobin, proptoporphyrin E, coproporphyrin, uroprophyrin, and melanin were purchased from Sigma Chemical Co. (St. Louis, MO). Hydroxypropyl methyl cellulose (HPMC) was a product of Fluka Chemical monkonkoma, NY). Bilirubin was obtained from the Hartman-Leddon Co. (Philadelphia, PA). Bovine hemoglobin (Certain tHb level 2) was from Ciba-Coming Diagnostics (Meaeld, MA). Uric acid was from Mallinckrodt Chemical Works (St. Louis, MO). y-(Methacryloyloxy)propyltrimethoxysilane @ind-silane), ammonium perW M E D ) , and sulfate, N,N,W,N-tetramethylethylenediamine acrylamide were obtained from Pharmacia Biotech (Piscataway, NJ). All other reagents were of analytical grade or better and purchased from standard reagent suppliers. Aqueous solutions were prepared with deionized water (Barnstead NANOpure system, Boston, MA). Phosphate run buffers (11.5 mM) with pH values between 6.5 and 8.1 were prepared by dissolving the appropriate amount of disodium hydrogen phosphate and sodium dihydrogen phosphate in deionized water. All buffers were filtered through 0.45pm cellulose acetate membrane filters (Alltech Associates, Inc., Deerfield, IL) before use. Trisborate buffers (200 mM) with pH values between 7.2 and 8.0 were prepared by mixing appropriate amounts of tris(hydroxymethy1)aminomethane and boric acid in 250.00 mL of deionized water. Tris-borate/HPMC buffer was prepared by dissolving 0.2%(w/ w) of hydroxypropyl methyl cellulose (0.2 g) in 100.00 mL of the Tris-borate buffer at the specified pH value. This solution was sonicated for 15 min to allow complete wetting of hydroxypropyl methyl cellulose and then stirred in an ice bath until it clariiied. Filtration was performed using a 0.8 pm cellulose acetate membrane filter (Alltech Associates, Inc.). NADH, NAD+, and NP standard solutions were prepared in 20 mM Tris-HC1 buffer @H 7.4). All buffers and solutions were stored at 4 "C until use. Serum Samples. Hemolyzed, lipemic, and icteric serum samples used in the interference study were obtained from MetroHealth Medical Center (Cleveland, OH). Universal infectious precautions were followed during the handling of these specimens. Analyses. The EMIT digoxin assay is a homogeneous enzyme immunoassay. The assay procedure as described by Syva Co. was followed until the conversion of NAD+ to NADH by glucos&phosphate dehydrogenase (GGPDH). The reaction mixture was separated by CE after the addition of NP (1.1 mM) as internal standard. Compared to the EMIT procedure, this method used only one-tifth of the reagents and sample volumes for each analysis. The volume of reaction mixture used in the CE separation can be further reduced to one-thirtieth of that of the EMIT procedure if a pipeting device with good precision is used. Capillary Columns. The fused-silica capillaries (360-pm 0.d.) was purchased from Polymicro Technologies Inc. (Pheonix, AZ). The uncoated capillary was 57 cm in length (50 cm to detection window) with 75pm i.d. This capillary was mounted in a P/ACE cartridge that was connected to a temperature control system. New capillaries were conditioned by rinsing sequentially with 0.1 M NaOH, deionized water, and phosphate run buffer before use. Polyacrylamide-coated capillary was prepared from the uncoated capillary with l W p m i.d. as described below. The inner surface of a 1-m-longcapillary was first pretreated with 1M NaOH for 30 min and then flushed with HzO for 30 min. The pretreated capillary was rinsed with a silane solution (PH 3.5) for 3 h at room temperature. The silane solution was prepared by adding 50 pL of acetic acid (glacial) and 50 pL of y-(methacryloy1oxy)propylt-
rimethoxysilane to 9.90 mL of HzO; the resultant mixture was stirred until it clarified (-15 min). After sequentiallywashing with methanol and HzO, the capillary was blown dry with Nz. A monomer solution was prepared by dissolving 0.4 g of acrylamide in 10 mL of deaerated Tris-borate buffer (200 mM), followed by addition of 100 yL of 10%freshly prepared ammonium persulfate and 10 yL of TEMED. This solution was then pushed through the silanized capillary for 1h using a Nzpressurized capillary wash tube (AlltechAssociates, Inc.). After polymerization,the capillary was flushed with HzO and ready for use. This procedure produces a bonded layer of linear polyacrylamide (non-cross-linked)on the inner surface. The on-column detection window was generated by either razor pilling or etching with a drop of hot (-130 "C), concentrated (96-98%) sulfuric acid to remove a small section of polyimide coating on the outside of the capillary. If sulfuric acid is used, it should be applied using the tip of a mercury thermometer as it retains heat longer. Capillary Electrophoresis with UV Detection. A Beckman (Fullerton, CA) P/ACE 2050 system and an IBM personal computer with System Gold software were used. On-column UV detection was performed at 260 nm, and the separation temperature was maintained at 25 "C. Samples were introduced into the capillary by electrokinetic injection at 5 kV for 3 s. The optimal conditions of electrophoretic separation were 20 kV/34 pA at pH 8.0 in the phosphate run buffer using uncoated fused-silica capillary and 25 kV/27 pA at pH 7.9 in the Tris-borate/HPMC buffer using polyacrylamide-coatedcapillary. Between runs, the capillarywas rinsed by run buffer for 2 min. If uncoated capillary was used, the cathode was placed in the outlet side of the capillary and the anode was in the inlet. If coated capillary was used, the polarity of the electrodes was reversed. RESULTS AND DISCUSSION
NADH and NAD+ Separation. In homogeneous enzyme immunoassay for digoxin, the active glucose-&phosphate dehydrogenase simultaneously converts glucose &phosphate (the substrate) to glucono-&lactone &phosphate and NAD+ (the coenzyme) to NADH. The amount of NADH formed is proportional to the digoxin level in the sample. Resolving of NADH from NADt and other potential interferents is-criticalto the quantitative measurement of digoxin. Figure 1 shows the separations of NAD+,NADH, and NP (internal standard) in both coated (A) and uncoated (B) capillaries. In both cases, NAD+, NADH, and NP were completely resolved within 8 min. Peak identities were tentatively assigned on comparison to the electropherograms generated by single components and subsequently coniirmed through stepwise addition of known components to the mixture. The small unidentified peak close to the NADH peak was probably associated with the impurity found in the NADH preparation. Due to the suppression of electroosmotic flow, only electrophoretic migration existed in the coated capillary. Accordingly,NADH (net charges of 2-) migrated toward the anode (outlet) at a faster rate relative to NAD+ (net charge of 1-) (Figure lA).In the uncoated capillary, the electroosmotic mobility was greater than the electrophoretic mobilities of NADH and NAD+. As the result of combination of two opposite forces, NADH migrated slower toward the cathode (outlet) compared to NAD+ (Figure 1B). The migration rate of the internal standard (NP) pKa of 7.15 at 25 "C, may be modulated by adjusting the buffer pH (Figure 2).
0.09 O'l1I
(I)
'
A
NADH
NAD
0.07
0
C
ge
0.05
ln
r)
a
0.03
0.01
-0.01'
2
"
"
"
"
5
4
3
"
6
"
7
"
"
10
9
8
Time (min)
"'"[
B
NAD
I
0.13
1
0'05 0.03
.0.01'
2
1I "
3
"
4
h
"
5
"
6
"
7
"
8
"
9
1
"
0
Time (min) Figure 1. Electropherogramsof NADH, NAD+, and NP in both the coated (A) and uncoated (B) capillaries. Experimentalconditions: (A) coated capillary, 100 pm (i.d.) x 57 cm; separation potentiakurrent, 25 kV/27 pA; run buffer, 200 mM Tris-borate/HPMC buffer at pH 7.9; injection potentialhime, 5 kV/3 s; UV filter, 260 nm; sample concentration, NADH 1.OmM, NAD+ 1.O mM, and NP 1.5 mM in 20 mM tris-HCI buffer at pH 7.4; anode, outlet; cathode, inlet; temperature, 25 "C. (B) uncoated capillary, 75 pm (i.d.) x 57 cm; separation potentiakurrent, 20 kV/34pA; run buffer, 11.5 mM phosphate buffer at pH 8.0; anode, inlet; cathode, outlet; the other were the same as
(4. The conditions that allow complete separation of NADH and NAD+ were best obtained between pH 6.5 and 8.5 using an uncoated capillary. At a pH higher than 8.5, the electroosmotic flow became too fast to achieve good resolution. At a pH below 6, NADH was not stable. It is also demonstrated that separation using polyacrylamide-coated capillary was best accomplished at pH 6.5-8.0. At a pH higher than 8.0, the coating becomes unstable. NADH Quantitation. Quantitations of NADH using coated and uncoated capillariesby external and internal calibrations were investigated and compared. Peak areas were plotted against the known concentrations of NADH in external calibration (Figure 3), where it was assumed that the volume of NADH injected remained constant from run to run. Quantitation was based on the direct comparison of the peak area of NADH in the test sample with those of the NADH calibrators. However, our results indicated that the above assumption was not valid, as small deviations in timing and applied potential caused variations in the injection volumes. Between two different types of capillaries, a much improved linearity was observed in the coated capillary (Figure 3A). This marked improvement in linearity may be Analytical Chemistry, Vol. 67, No. 18,September 15, 1995
3213
14[
A NADH
\
0.30
NP
A
1
/
0.00L"
'
100
0
'
'
200
'
I
300
'
400
'
'
500
600
INADHI (vM) 8 7
/
7.2
7.0
7.4
7.6
7.8
8.0
8.2
0.16
PH
0.14
r
. a
r
*/
1
10 r
NADH
I B
9
,z
a
.EI-
7
1&
NP
n
2
NAD
h
Y
6
,
5
,
,
,
, , , ,
attributed to the suppression of the electroosmotic flow and the creation of better surface uniformity with polyacrylamide coating. The internal calibrations of NADH with NP are shown in Figure 4 using coated (A) and uncoated (B) capillaries. In internal calibration, a known amount of NP was added to each NADH calibrator and unknown sample prior to electrophoresis. Peak area ratios @NADH/&) were plotted against the NADH concentrations to obtain a calibration plot. The NADH concentration in the test sample was determined in the calibration plot by knowing the area ratio of the unknown vs NP. The use of the relative area obviated the need to maintain constant injection volume and detector response during the analysis. As a result, excellent Iinearities of the internal calibration plots were obtained. Although the upper limit of NADH was not explored in either the coated or uncoated capillaries, the l i t of detection for NADH, defined as the ratio of the triple average area of 30 baseline peaks to the area of NP (signal/noise 3), was calculated as 11.5 pM or 68.4 fmol for the coated capillary (Figure 4A) and 9.53 pM or 42.9 fmol for the uncoated capillary (Figure 4B). Precision and Recovery Studies of NADH Detection. To evaluate the accuracy and reproducibility of NADH detection in hemolyzed, lipemic, and icteric serum samples, where common endogenous interfering substances can often create analytical 3214
I
Analytical Chemistry, Vol. 67, No. 18, September 15, 1995
'
0.00 0
1
50
100
150
200
250
300
350
400
450
[NADH1 (PM)
Figure 3. Calibration plots of NADH in coated (A) and uncoated (6)capillaries without internal standard. Experimental conditions were the same as in Figure 1.
problems, samples of each type of interference spiked with NADH standards at concentrations of 100 and 400 pM were analyzed and compared with NADH standards dissolved in a buffer free from such interferences (20 pM tris-HC1 buffer at pH 7.4). As seen in Table 1,the interday variability was less than 2%,and the analytical recoveries ranged from 98 to 102%. These recovery values are within the range reported for most of the CE applications. The detection variability of 2% in the presence of various interfering factors is a good indicator of the analytical dependability of this method under less than ideal analytical conditions. capillatyEleclrophoreticEnzyme Immunoassayfor Digoxin. Electropherograms for the separation of enzyme immunoassay mixtures using both coated and uncoated capillaries were given in Figure 5. The peaks were identified by sequentially spiking the immunoassay mixtures with NADH, NAD+, and NP. The results were further verified by comparing the electropherograms of the assay mixture with those of each individual immunoreagent (results are not shown) for correct peak assignments. The unidentified peaks that appeared in Figures 5 were reagent constituents present in the assay kit. Since these unidentified peaks did not overlap with the NADH and NP peaks, no effort was made to reveal their identities. Figure 6 shows the dose-response curves for digoxin, which were constructed by plotting the area ratios of NADH generated by the enzyme reaction and internal standard, NP, against the digoxin concentrationsin human serum. In either coated capillary
~~-0.00994+0.004121 x
0.80
NAD
0.70 -
a
m:%
N
2 a
0.50
-
0.40
-
0.30 0.20
100
0
200
300
400
NP NADH
1
-0.10'
600
500
I
0.60 -
'
0
'
'
2
'
(NADH1 (IJW
0.90 -
0.80
B
0.18 r
y=0.024013+0.002007x
0.16
-
