3.1.2. CEA

Figure 3a illustrates a protocol from incubation to immobilization of the CEA-sandwich bodies on the metasurface biosensor composed of a periodic array of silicon nanopellets, though the MF path around the periodic array is not drawn. The binding molecules, Cys-SA, were immobilized in advance on the silicon nanopellets. After the immobilization of the Cys-SA and the rinse of unbound molecules, the CEA-sandwich bodies flowed in the MF paths and effectively bound via the biotin–avidin interplay on the silicon nanopellets. Unbound CEA-sandwich bodies were rinsed with the PBS-T. Afterwards, the FL imaging was conducted from the top of metasurface biosensor; the optical configuration of the biosensor has been described in previous publications [20,21].

A detection curve for CEA in the sample diluent NS buffer is shown in Figure 3b, presented on a semi-log scale. The measured data are shown with orange closed circles associated with error bars. A dashed curve represents a fitted curve using the Hill equation (Equation (1)), which well reproduce the CEA detection data. Thus, the CEA was detected in a scaled manner. A lower concentration range is shown in Appendix (Figure A1a), and the CEA concentration at 0.008 ng/mL was detected in the measurement. From the crossing point of the Hill curve and 3*σ* level (horizontal bar) in Figure 3b, the limit of detection (LOD) of CEA in this measurement was found to be 0.002 ng/mL (or 11.1 fM). It is to be noted that the detection curve is scaled to the LOD; in other words, the dynamic range covers from 2 pg/mL to 25 ng/mL, exceeding four orders of CEA concentrations. Inset magnifies a concentration range near 0 g/mL on a linear scale.

Figure 3c shows FL signals emitted from the CEA-sandwich bodies in a human serum. The horizontal axis is logarithmic. When we conducted the CEA detection using the serum, the FL-signal level was about five-times lower than that using the NS buffer. It is considered that abundant proteins prevented the CEA and the Abs from forming the sandwich bodies. Accordingly, we conducted a more elaborate FL-signal analysis than that for Figure 3b. In each channel, the FL signal was evaluated in a criterion that intensity more than 3*σ* from the background level is counted as net signals. This statistical criterion does not output error bar; therefore, the FL signals in Figure 3c are shown only with orange closed circles. A horizontal bar indicates zero-signal level in the measurement. It is to be stressed that the CEA in human serums was detected even at 0.04 ng/mL in a scaled manner; the dynamic range is, at least, 0.04–25 ng/mL. An arrow indicates the clinical diagnosis criterion for CEA, that is, 5 ng/mL [2]. Thus, the metasurface biosensors are capable of detecting CEA in human serums around the diagnosis value.

**Figure 3.** CEA detection. (**a**) Schematic of incubation and immobilization of the CEA sandwich bodies on the metasurface of a periodic array of Si nanopellets. (**b**) Detection of CEA in the sample diluent buffer. Measured FL intensities are shown using orange closed circles with error bars. Dashed curves denote fitted curves by the Hill equation (Equation (1)). 3*σ* level from the zero level is indicated by a horizontal bar; the crossing point with the Hill curve means the limit of detection, indicating 0.002 ng/mL (arrow). The inset magnifies the detection curve around the zero concentration on a linear scale. (**c**) FL signals from CEA in the human serum. Horizontal bar indicates zero-signal level. Arrow indicates the clinical criterion of 5 ng/mL.

#### *3.2. Coexisting Target Detections*

Detection results under conditions that PSA and CEA coexist are shown in Figure 4. Presentation styles in Figure 4 are similar to those in Figure 2,3. We tested two configurations: (i) the target PSA concentrations were 0.8, 4.0, and 20 ng/mL, as shown in Figure 4a, while the CEA concentration was kept at a constant of 5 ng/mL, and (ii) the target CEA concentrations were changed from 0.04 to 25 ng/mL, as shown in Figure 4b, whereas PSA concentration was fixed at 4 ng/mL. In the case (i), CEA was an impeding biomolecule for the target; in (ii), the PSA could impede the detection of target CEA. We note that the clinical criteria for CEA and PSA are 5 and 4 ng/mL, respectively [2].

In both cases, even when the competing cancer markers existed, the target was successfully detected, similarly to the individual detections in Figure 2,3. In detail, the FL intensity in the CEA detection became low (Figure 4b), which suggests that the PSA affects the CEA detection; in contrast, there is no definite signature that the CEA affects the PSA detection because the FL intensity was not reduced in Figure 4a, in comparison with that in Figure 2d. Figure 4a is shown on a linear scale and the three data points around the clinical criterion value, 4 ng/mL, were fitted using a line, being well-reproduced (R2 = 0.999).

**Figure 4.** Detection of target cancer markers in coexisting conditions together with other markers. (**a**) PSA detection under a mixture with CEA of 5 ng/mL, presented on a linear scale. FL intensity coming from the PSA is shown using orange circles with error bars. Dashed line: a fitted line. Arrow: the medical criterion value of PSA, 4 ng/mL. (**b**) CEA detection under a mixture with PSA of 4 ng/mL. FL intensity coming from the CEA is shown using orange closed circles with error bars on a semi-log scale. Dashed curve: a fitted curve by the Hill equation (Equation (1)). A black horizontal bar represents the zero level, which is *y*<sup>0</sup> in the Hill equation. Arrow: the medical criterion value of CEA, 5 ng/mL.
