Monoclonal antibodies (mAbs) that recognize and bind to specific antigens (Ags) have a wide range of applications in research, therapy, and diagnostics. However, many of these antibodies cannot bind well to the native Ags. In this study, based on the Chinese hamster ovary (CHO) cell display platform developed previously in our lab, we reported a novel artificial evolution procedure to improve the affinity of mAb against the native Ag directly using the plasma samples without purification of the native Ag. In this procedure, a pair of antibodies able to bind the Ag in sandwich manner are first confirmed (Ab1/Ab2) and the antibody (Ab) to be affinity-improved (Ab1) is displayed on CHO cells for Ab mutation. Then the cells were detected and sorted with flow cytometry in the form of Ab1-Ag-fluorescence labeled Ab2, which we named sandwich flow cytometry. Here, we used soluble isoform of suppression of tumorigenicity 2 (sST2) protein as model Ag, carried out "sandwich" maturation directly using the plasma samples containing the native sST2 protein and optimized a pair of antibodies with significantly improved sensitivity in the detection of the native sST2 in plasma. This method could be very useful in optimization of the diagnostic Ab pairs working in a "sandwich" manner if more antibodies were also successfully affinity-matured with this method.