2.1 Effect of immobilization method on the detection of CEA
Antibodies could be immobilized directly on the sensor surface or indirectly to the surface through binding to another immobilized molecule that is called a capturing molecule. The direct immobilization is a covalent binding of amine group of antibody with carboxylic group on the sensor chip. The advantage of the direct binding is that antibody molecules are close to the sensor chip, which is favorable for higher sensitivity, while the disadvantage is the orientation of the antibody molecules on the sensor surface being random. The indirect immobilization includes covalent binding of capturing molecules with the sensor chip and molecular interaction of capturing molecules with the antibodies. In this work, we used anti-mouse rabbit IgG polyclonal antibody, Protein A and Protein G as capturing molecules. Immobilization via Protein A and Protein G should allow most of the Fab portions of the immobilized anti-CEA IgG accessible to the antigen (CEA) [
18,
19].
Figure 1 shows the sensorgrams of immobilizing anti-CEA antibodies onto CM5 sensor chip via four different routes: (a) anti-mouse IgG polyclonal antibody, (b) Protein A, (c) Protein G and (d) direct covalent binding. The change in the resonance unit is shown on the
y-axis, with the time of the reaction on the
x-axis. The immobilized amounts of anti-CEA antibody via the four methods are 2,290, 50, 1,600 and 23,800 respectively. This indicates the direct covalent binding adsorb more anti-CEA antibodies onto the sensor chip, and anti-mouse polyclonal antibody and Protein G surfaces adsorb reasonable amount of anti-CEA antibodies, while these via Protein A are found to adsorb few amount of anti-CEA antibodies.
Immobilization via Protein A and Protein G are expected to provide well-oriented anti-CEA antibodies, which is beneficial for the further detection of antibody and antigen interactions. Considering that the introduced anti-CEA antibody belongs to IgG1 subclass, the extremely low capacity of Protein A should be attributed to its weak affinity to IgG1.
Interactions of CEA with anti-CEA antibody immobilized surface and control dextran surface are shown in
Figure 2(a). The resonance unit change on the two kinds of surfaces in response to the analyte solution of 100 ng/mL of CEA in HBS buffer was 56 and 23 RU for direct binding surface, and the anti-mouse polyclonal antibody, respectively. Considering that the amount of antibody immobilized onto the sensor chip via anti-mouse IgG polyclonal antibody is only one tenth of that via direct binding (resonance unit shift 2,290 to 23,800), and further signal shift to interact with antigen is two fifth (resonance unit shift 23 to 56), it is highly possible the antibody on anti-mouse IgG polyclonal antibody has better orientation, and also the crowded antibodies on both surfaces limit the binding capacity. As a result, the direct covalent binding immobilizes much more of the anti-CEA onto the surface, and shows higher sensitivity to the antigen solution.
The susceptivity to non-specific bindings of the two kinds of surfaces is shown in
Figure 2(b). Human serum diluted 10-fold in HBS buffer caused a high signal shift of around 600 RU on the sensor chip of anti-CEA on anti-mouse IgG polyclonal antibody, and a low signal shift on the anti-CEA surface through direct covalently binding.
The fact that the polyclonal IgGs surface gave a substantially high resonance unit increase is most probably due to the cross-reaction between immobilized anti-mouse antibodies and human IgGs present in serum in extremely high concentration. Likewise, anti-CEA antibody surfaces constructed through Protein A and Protein G immobilization also gave substantial resonance signal increases upon injection of human serum (data not shown). Considering that protein A and Protein G bind to IgGs of many different species including human, its application for the immunoassay using human serum is obviously problematic, since IgG is the second abundant protein in human serum with its average concentration in adult reaches as high as 12 mg/mL. From this point of view, covalent immobilization was found to be more advantageous for the immobilization of anti-CEA than indirect immobilization via rabbit polyclonal IgGs, Protein A and Protein G for the detection of CEA in serum sample, since it showed high response to CEA and low response to human serum.
For regeneration of the sensor surface, glycine·.HCl (pH 1.5) was used after antibody-antigen interactions. The antigen-antibody pair on the sensor surface through direct binding was broken, leaving antibody on the surface for use in further detections. In contrast, non-covalently immobilized anti-CEA antibody via anti-mouse IgG polyclonal antibody, Protein A and Protein G was removed from the surface by the regeneration process using glycine·HCl (pH 1.5), so the introduction of anti- CEA antibody is required for each analysis, which takes more steps. Considering the facts above, we decided to use direct covalent binding for the further study.
2.2. Kinetic analysis of interactions of carcinoembryonic antigens and antibodies
Kinetic analysis should provide important parameters in designing the experiment conditions to improve the detection sensitivity. Therefore, the interaction of CEA with immobilized anti-CEA antibody was analyzed under a series of concentration (12.5 ng/mL ∼ 800 ng/mL). The kinetic parameters were calculated based on the global analysis of the experimental data using a 1: 1 binding model accounting for mass transport.
The association rate constant (
ka), dissociation rate constant (
kd) and equilibrium constant (
KD) were calculated to be 4.22 × 10
5, 3.0 × 10
-5, and 7.11 × 10
-11, respectively. Based on the kinetic parameters, the change of resonance unit with the interaction time to the analyte solutions could be simulated by using the BIA evaluation software. As shown in
Figure 3, the experimental data and the simulated data of the interactions of immobilized anti-CEA antibodies with antigens of a series of concentrations agree very well. For a low CEA in HBS concentration of 6.25 ng/mL, the simulation indicates that the resonance unit increase is less than 1 RU in 300 seconds, while it could reach to 10 RU in 3600 seconds, which is a change from undetectable to a detectable range. Consequently the interaction time was extended to 3600 seconds, and the resonance unit shift was 8 RU, as seen in the experimental sensorgram of
Figure 4. This is quite similar to the simulated result of 10 RU. The above result indicates that even lower concentration of CEA (e.g. ∼1 ng/mL) can be detectable by extending the interaction time.
2.3. Method to increase the sensitivity --- sandwich assay
As an alternative to improve the detection sensitivity of CEA on SPR, a sandwich assay was used to detect CEA antigen with a low concentrations while maintaining the short analysis time of 180 seconds. In this work, three anti-CEA antibodies were selected: RDI detecting as the anchor antibody to be immobilized onto the sensor surface, and RDI coating and anti-CEA from HuChem as the second and third antibody for the sandwich assay.
Figure 5(a) shows the sensorgram of detecting 20 ng/mL of CEA in HBS, the antigen, second antibody and third antibody caused a shift of 30 RU, 50 RU and 20 RU, respectively, so the total signal shift is increased to 100 RU, 3.3 times the shift caused by antigen only. Therefore, it can be concluded that the three kinds of antibodies are able to bind to different epitopes of CEA antigen, and the sandwich assay amplifies the original interaction signal significantly.
Furthermore, the sandwich assay-based SPR experiment was conducted for the CEA antigen with a low concentration of 1.0 ng/mL, and the corresponding sensorgram is shown in
Figure 5(b). This figure indicates that the otherwise unclear signal change for the antibody-antigen interactions could become visible after the interactions of antigen with the second and third antibodies. Therefore, the sandwich assay is an efficient method for detecting CEA antigens at very low concentrations.
A trial of detection of CEA in human serum was conducted by analyzing human serum spiked with different amount of CEA (25-800 ng/mL). As shown in
Figure 5(c), as little as 25 ng/mL of CEA in serum was clearly detectable when serum was diluted to 1:10 with HBS. Though the detection sensitivity may need to be further improved several times for the detection of low ng/mL of CEA in serum, this approach clearly indicates that SPR could become rapid and informative alternative to detect small amount of CEA in serum. It has been reported that the glycosylation pattern of CEA changes upon malignant transformation [
20], which indicates that not only detecting CEA but also characterizing glycoforms present on CEA may help specifying the stage of diseases. In this regard, SPR may be a preferred analytical platform due to its ability to amplify responses using carbohydrate- recognizing molecules (e.g. lectins and anti-carbohydrate antibodies) without the need for labeling.
2.4. Influence of the sensor chip
All of the above experiments were conducted on the CM5 sensor chip. In fact, for detecting protein- protein interactions, several kinds of sensor chips are applicable based on the properties of proteins. Since the target analyte CEA has a large molecular weight of around 150 KDa, we also used C1 and a self-assembly monolayer of ethylene-glycol terminated alkanethiols as the sensor chip in order to find the best condition to detect CEA at low concentrations.
CM5 is the most versatile and general sensor chip in the BIACORE sensor chip family, where carboxymethylated dextrans are covalently attached to a gold surface as a thick matrix. The molecules could be covalently coupled to the sensor surface via amine, thiol, aldehyde or carboxyl groups. In this work, the amine group of the antibody was covalently coupled to the carboxyl group of carboxymethylated dextran. C1 is also a good option for antibody-antigen interaction, since C1 has a flat layer terminated with a carboxyl group and is suitable for detecting large molecules, such as antibodies or antigens. In addition, we also considered using a self-assembled monolayer of mercaptoalkanoic acid as the material for sensor chips, because they can form a thin layer with the functional carboxyl group on the gold surface. In order to avoid non-specific binding, the mixture of ethylene glycol-terminated mercapto-alkanol, (1-mercapto-11-undecyl)tetra(ethylene glycol) (HSC11EG4OH) and ethylene glycol-terminated mercapto-alkanoic acid, (1-mercapto-11- undecyl)hexa(ethylene glycol)carboxylic acid (HSC11EG6OCH2COOH) were used in this research and their self-assembly monolayer are abbreviated as EG-SAM in the following discussion.
The immobilization of anti-CEA onto the sensor chip caused a shift of 1,700 RU on C1 and 2500 RU on EG-SAM (the figures are omitted), while the shift was 23,800 RU on CM5 under the same condition. The immobilized amounts of anti-CEA antibody onto C1 and CM5 are much less than that on CM5, most probably because CM5 has a matrix containing more carboxyl group than C1 and EG- SAM. However, the further interactions of antibody-antigen caused quite similar signal shifts on the three kinds of sensor chip CM5, C1 and EG-SAM as shown in sensorgrams in
Figures 3 and
6. The response data summarized in
Figure 7 gives a more clear view of resonance unit shift with concentrations on different surfaces, namely, there is a linear relationship between resonance unit shift and concentration when the concentration is lower than 400 ng/mL and the shifts on the three kinds of sensor chip are similar to each other.
It is very interesting that the signal shifts for antigen-antibody interactions are very similar, even though the signal shifts for immobilizing antibody are quite different on the three kinds of sensor chips. To understand this phenomenon, the SPR principle and the structures of the sensor chips were examined. The structures of the three kinds of sensor chips are illustrated in
Figure 6(a). CM5 has a thick layer of around 100 nm of dextran, while C1 has a thin flat layer on gold and EG-SAM has a layer of 5 nm on gold. According to SPR principle, the SPR signals decay exponentially with the distance of analyte from the surface [
21], the change close to the gold film is easier to detect, so the interaction on the top of dextran layer may cause less shift compared to that at the bottom, and there is higher hindrance for the anti-CEA antibody at the bottom to bind with CEA. Therefore, the interactions of antibody and antigen would presumably cause similar signal shift even though there are more antibodies on CM5.