Heterogeneous post-column immunoreaction detection using magnetized beads and a laboratory-constructed electromagnetic separator

The nature of immune reactors allows development of quantitative analytical methods that are highly selective and can often be used directly with complex biological matrixes such as blood, plasma or urine. A major limitation of immunoassay is that antibodies are sometimes unable to discriminate structurally similar species such as drug metabolites and synthetic analogs. The problem associated with the lack of discrimination can be circumvented by coupling immunoassay with liquid chromatography post-column. The most commonly used separation method in post-column immunoreaction detection is the affinity column. Affinity columns may create undesired effects such as a compromise of the chromatographic separation efficiency, the requirement for an antibody with fast reaction kinetics and the need for flushing the column. This paper reports a post-column immunoreaction detection system coupled with a laboratory-constructed on-line magnetic separation flow chamber that is designed to overcome these problems. The system uses disposable magnetic beads as a solid-phase support for separation that can be easily removed from the system. The model analytes chosen for this study were digoxin and its metabolites due to the commercial availability of monoclonal antibodies for these compounds. Digoxin was separated using a chromatographic method prior to being interfaced through a liquid handler system to the immunoreactor. Compatibility of the HPLC mobile phase was determined to be acceptable with a mixing ratio of 1:3 between the LC fraction and immunoreagent solution. The dynamic range of the calibration curve in digoxin-spiked phosphate buffer was found to be 0.25-12 ng/ml and a quadratic fit was found to provide the best fit to the data with a correlation coefficient of 0.9974. The residual error for all standards was less than 15%. The percentage RSDs for the two controls, 2 and 10 ng/ml, were 6.88 and 4.82% (n = 6) and the percentage errors were 7.07 and -6.89% (n = 6), respectively.