Briefly, biotin (2.44 mg) together with EDC (19.14 mg) and NHS (11.49 mg) were dissolved in 0.01 M phosphate buffer saline (PBS, pH 7.4, 0.1 mL) and stirred for 15 min at room temperature (RT). This solution was then added dropwise to polylysine (PL) solution (polylysine: 450 mg in 1 mL of 0.01 M PBS at pH 7.2) so the PL: biotin molar ratio was 1:1. The mixture was stirred at RT for 1 h and excessive unreacted biotin and ions in aqueous solution were removed by size exclusion (5 KD) column chromatography against PBS buffer (0.01 M, pH 7.4). Then the biotin-labeled PL was labeled with Atto 647N according to the Atto 647N protein labeling kit protocol from Sigma. Firstly, the solution of biotin-labeled PL was added with sodium bicarbonate buffer solution (pH 9.5), adjusted to pH 8.5–9.0 and transferred to Atto 647N (1 mg Atto 647N dissolved in 10 μL DMSO), incubated at room temperature under gentle shaking for 2 h, followed by being separated conjugates from free dye by a size-exclusion column. Finally, the Atto 647N labeled polylysine (FLPL conjugate) fraction was collected and stored at 4 °C. The rabbit polyclonal antibody-Atto 647N conjugate (RIgG-FL) was obtained and purified with the same process.

Biotinylation of Mouse monoclonal antibody (mAb1) was performed as described in Section 2.3.1. The molar ratio of biotin to mAb1 was 2:1. After labeling, excessive unreacted biotin and ions in aqueous solution were removed by dialysis against PBS buffer (0.01 M, pH 7.4) for 2 days. Then, it was stored in −20 °C until use.

The biotinylated-mAb1, streptavidin and FLPL conjugate were mixed in a molar ratio equal to 1:1:3 or 1:2:6 (Table 1). The conjugating mechanism of FLPL-BSAS-mAb1 conjugates was shown in Figure 1. Finally, FLPL-BSAS-mAb1 conjugate and RIgG-FL (1:1, v/v) were dispersed in stock solution (1% BSA and 0.005% sodium azide in 0.01 M PBS pH 7.4), respectively, and kept at 4 °C until use.

As described in Section 2.3.3, the biotinylated-mAb1, streptavidin and FLPL conjugate were mixed in different ratios and two different mixing orders were tested (see Table 1). In order to obtained higher signal amplification conjugate, different conjugate procedures were used in preparation.

Five kinds of NC membranes (Millipore HF090, Millipore HF135, Sartourius CN95, Sartourius CN 140 and Whatman AE99) from different manufacturers were tested, respectively, according to the procedure described in Section 2.5. In this investigation, C line was coated with 1 mg/mL GAR, and T line was coated with 1 mg/mL pAb2. PBS buffer (10 mM, pH 7.4) was used as negative sample. After chromatography for 10 min, signals at the T and C line were measured.

Standard solution (1,000 pg/mL) was applied on the sample pad and allowed to migrate across the NC membrane. Signals of T line and C line were read every minute for 12 min.

The use of polylysine with a big amount of primary amino was vital for signal amplification in LFIA, which in turn would influence the loading amount of Atto 647N on the conjugate. The biotinylated-mAb1 and FLPL conjugate self-assembled via the streptavidin–biotin system. We compared the analytical performances of four types of conjugates prepared through different conjugation procedures (Table 1). In the presence of Cry1Ab (1,000 pg/mL), there was a very weak response observed with Type 1 conjugate (T1) and the highest responses were achieved by Type 3 and Type 4 conjugates (T3 and T4).

Figure 4 illustrates the sequential steps and structures of various conjugates prepared by chemical modification of mAb1. T3 and T4 had similar signal intensity and 2-fold higher than type 2 conjugate (T2), indicating they may have similar structures. In addition, T3 and T4 demonstrated a 60-fold higher intensity compared to T1. Therefore, we propose three kinds of structures for these conjugates. In these tests, the signal response of T1 accumulated on the bottom of the NC membrane, that may be caused by the steric hindrance effect of a polymer, as only a few tiny polymers could migrate across the NC membrane (Figure 4, panel A), so T1 might be structure 1. T2 might be structure 2, while T3 and T4 might be structure 3, because structure 3 potentially has 2-fold more FL molecules conjugated on mAb1 and would be expected to generate a 2-fold higher response compared to structure 2 (Figure 4, panels B, C and D, respectively). Thus, structure 3 was used to fix on conjugate pad for the following experiments. Furthermore, we have attempted to apply longer polylysine (70−150 KD) to construct higher bright conjugates, but the results indicated that conjugates with longer polylysine chains cannot for pass through the NC membrane, because they would clog in the membrane pores (data not shown).

Responses at the T lines on five NC membranes were compared in Figure 5. Responses of strips constructed with Sartorius CN 140 membrane were the highest, while the response of strips constructed from Whatman AE 99 membrane were the lowest in the presence of 1,000 pg/mL Cry1Ab. In detection of 0 ppt Cry1Ab, a little response could be observed on Millipore HF135, Sartourius CN95, Sartourius CN 140 and WhatmanAE99, because a few of the FLPL-BSAS-mAb1 conjugates would physically clog the small size of pores of these NC membranes, but almost no response could be observed on the Millipore HF 090 membrane in detection of 0 ppt Cry1Ab. With consideration both of the responses on 0 and 1,000 pg/mL, Millipore HF 090 was selected as the most suitable NC membrane for strip preparation in the following experiments.

The time-intensity curves were obtained by measuring the signals of the T and C lines every minute for 12 min after standard solution addition. The time-intensity curves and time-signal ratio A_T/A_C curve at one concentration of 1,000 pg/mL are shown in Figure 6, respectively.

The T line and C line signals increased rapidly during the first 25 min, between 25 and 45 min, the signals increased slowly and entered a plateau (data not shown). The time-signal ratio A_T/A_C curves of the first 12 min have been observed, and indicated that A_T/A_C would vary in the first 9 min, then gradually plateau, indicating the signal intensity was relatively stable for analysis (reading) over 9−12 min. With consideration of signal intensity for low concentration sample and the shortest time for detection, the detection time was considered as 10 min.

After optimizing the signal amplification system, a series of Cry1Ab samples were tested by the FLPL-BSAS-mAb1-based LFIA to obtain a standard curve. Quantitative detection data is illustrated in Figure 3(A). There was an obvious increase in signals at the T line with the increasing Cry1Ab concentration. A standard curve for Cry1Ab was plotted by the signal ratio A_T/A_C at different Cry1Ab concentrations (Figure 3(B)). It showed good linearity in the range between 0 and 1,000 pg/mL; the equation was: (1) [ y = 0.0013 x + 0.0219 ]while the correlation coeffcients were calculated to be 0.9958. The coefficients of variation (CV) values were both lower than 10%. The cut off level for all detection systems was calculated by the following equation: (2) [ 3 ∗ SD background + average background ]

A detection limit of 10 pg/mL was achieved, which was 100 times lower than that of the magnetic beads- based ELISA [13] and gold-based LFIA [5]. This suggested that the proposed method is highly sensitive, especially for detection of biomolecules at low levels.

The cross-reactivity (CR) and specificity of FLPL-BSAS-mAb1 based LFIA for Cry1Ab was assessed with available different samples: Cry1C, Cry2A, Cry3A. It was checked with different dilutions of above proteins but there was no cross-reactivity with those three transgenic proteins, indicating that this assay can be used for unique quantification of Cry1Ab protein.

As ingredients in food are more complicated than the PBS buffer, Cry1Ab recovery rates in flour matrix were calculated to evaluate the influence of matrix on detection of target cry1Ab. It was calculated using the formula: (3) [ ( observed concentration / spiked concentration ) × 100 ]and expressed in percent. As shown in Table 2, the recoveries were ranged from 80.05% to 109.69% with CVs of 3.8–9.5% in signal-amplification-based LFIA.