*2.3. MF Protocols and FL Detection*

MF protocol for PSA detection was as follows. PBS was first used to fill the MF paths. Second, the Cys-SA solution, adjusted to 20 μg/mL using the PBS, was flowed at 10–11 μL/min for 10 min on the metasurface areas. In a previous experiment measuring sensorgram for the Cys-SA [22], the immobilized amount reached the maximum at approximately 8 min, so that we set the time to flow the Cys-SA to be 10 min. The Cys-SA was rinsed with the PBS for 7 min; then, a background FL image was captured for 2 s on each channel under illumination by a green LED (M530F2, Thorlab, Newton, NJ, USA). The FL images were acquired using an uncooled CCD camera (Infinity3S-1URC, Teledyne-Lumenera, Ottawa, Canada). Subsequently, the biotin-labeled anti-PSA Abs of 2.0 μg/mL were flowed at 10–11 μL/min for 10 min on the metasurface areas, and then they were rinsed for 7 min with the PBS. The target PSA proteins diluted with the NS buffer or the human serum were flowed at approximately 8 μL/min for 20 min, and then they were rinsed for 7 min with the PBS. Due to the low concentrations, the PSA was flowed at the low flow rate. The HL555-labeled Abs were flowed at 10–11 μL/min for 10 min, and

then the final rinse was conducted at 19–20 μL/min for 8 min with PBS-Tween20 (PBS-T, 163-24361, FujiFilm Wako Pure Chemicals), pH 7.4. Following the MF-flow protocol just above, the green LED light illuminated on each channel, and each FL image was acquired for 2 s exposure time to detect the PSA. A custom-build software was used to control MF flows, liquid reagent changes, and the FL measurements in sequence. The automated setup was as compact as 40 × <sup>30</sup> × 60 cm3.

MF protocol for CEA detection differed from that for PSA. The sandwich complexes were incubated independently of the MF-flow system because we found that the step flows for the PSA were not suitable for CEA. This difference probably comes from smaller affinity between the CEA and Abs compared to that between PSA and the Abs. The target CEA was adjusted to particular concentrations for each experiment, typically, 0.04–25 ng/mL using the NS buffer or the human serum diluent (serum : NS buffer = 1 : 4 in volume). For the serum, the target CEA was first spiked in the human serum pool and the concentration was 200 μg/mL in the human serum. Afterwards, the target was diluted using the human serum diluent. The anti-CEA Abs were diluted to 10 μg/mL for the serum-diluted CEA and to 2 μg/mL for the NS-buffer diluted CEA using the PBS. Typically, the 50 μL CEA and the two 100 μL anti-CEA Abs were mixed and incubated at 299 K for 40 min at 400 rpm in the dark. After the incubation, the test liquid was flowed at 10–11 μL/min for 23 min on the metasurfaces that was already covered with the Cys-SA; then, the MF paths were rinsed with the PBS-T at 19–20 μL/min for 5 min. Subsequently, FL imaging was conducted on each channel for 3 s exposure time. When the FL images were analyzed, we used a free software, ImageJ [28].
