Capture antibodies of CK-MB (CK-MB-Ab₁), Myo (Myo-Ab₁), and cTnI (cTnI-Ab₁) are from Abcam (United Kingdom). The CK-MB protein is from EastCoast Bio (United States). The Myo protein and recombinant human cardiac troponin I protein are from Abcam (United Kingdom). Detection antibodies of CK-MB (CK-MB-Ab₂), Myo (Myo-Ab₂), and cTnI (cTnI-Ab₂) conjugated with HRP are from Abcam (United Kingdom). Supersensitive chemiluminescent substrate reagent kits are from Beijing Labgic Technology Co., Ltd. (Beijing, China). Bovine serum albumin fraction V (BSA) powder is from Tianjin Kangyuan Biotechnology Co., Ltd. (Tianjin, China). Phosphate-buffered saline (PBS) tablets and Tween-20 are from Amresco (United States), which are used to prepare 0.01 mol/L PBS solution (pH 7.4) and 0.5% PBST solution. The capture antibodies and antigens are diluted in PBS solution. The detection antibodies are diluted in 0.05% BSA solution. Polydimethylsiloxane (PDMS) and a curing agent (Sylgard 184) from Dow Corning Inc. (United States) are used to prepare the microfluidic chips. A 500*500*0.1 mm silicon film used to coat the capture antibody is from Shanghai Shentong Rubber and Plastic Products Co., Ltd. (Shanghai, China).

The CL image analysis system is purchased from BIO-OI Co., Ltd. (Guangzhou, China). A PT-10s Plasma Cleaner is produced by SANHOPTT Co., Ltd. (Shenzhen, China). A 202-00T electric thermostatic drying oven is purchased from LICHEN-BX Co., Ltd. (Shanghai, China). A DZF-6020A vacuum drying oven is purchased from LICHEN-BX Co., Ltd. (Shanghai, China). An F900 industrialization FDM printer is produced by Stratasys Ltd. (United States). A Lite 600HD 3D printer is produced from SHINING 3D Technology Co., Ltd. (Yangzhou, China). An IMM 3000 metallographic microscopy system is purchased from MEGA Instruments Co., Ltd. (Suzhou, China).

We designed the three-dimensional model of the SMC using SolidWorks^® 2016 software. The SMC (35 mm long, 57 mm wide) for the chemiluminescent reaction consists of three layers: a microchannel layer (4 mm thick) on top, a reaction layer (silicon film, 0.1 mm thick) in the middle, and a base layer (3 mm thick) at the bottom (Figure 1A).

The top microchannel layer comprises four capsule-shaped reservoirs (7 mm long, 3 mm wide, 3 mm deep, and with a capacity of 60 μl) for depositing samples, detecting antibodies, PBST, and Luminol-H₂O₂ CL substrates, respectively. To load reagent samples conveniently and maintain the pressure balance of the chip, a vent (1.0 mm in diameter) is drilled by a puncher on the top of the reservoir. To mix samples efficiently, an Archimedean spiral mixer is designed. At downstream of the Archimedean spiral microchannel, an S-shaped channel is devised. The sizes of Archimedean spiral and S-shaped microchannels are both 400 μm wide and 400 μm deep with a total capacity of 30 μl. All reservoirs converge into the Archimedean spiral microchannel and S-shaped microchannel through a press-type mechanical (PM) valve. The end of the S-shaped channel is a waste chamber. A negative pressure vent (a through hole with a diameter of 1.0 mm) in the waste chamber is used to connect a pump for fluid driving. The middle reaction layer is a silicon film patterned with three different capture antibody strips in advance (Supplementary Figure S1). After the fluid runs across the S-shaped microchannel on the surface of the reaction layer, a microarray zone with nine spots for multianalyte detection is formed by CL reaction (Figure 1B). The bottom base layer has the same size as the top microchannel layer, and they both contain a hole in the top to insert the PM valve.

The PM valve is used to connect different reservoirs with the Archimedean spiral microchannel. Each reagent in different reservoirs is injected into the microchannel of the chip by pressing the PM valve to a proper height which opens a peristaltic pump. To ensure the stability of the movement of the PM valve, we designed an external holder composed of two holder plates (a thickness of 2 mm) with a hole in the middle for inserting the valve and two pin holes in both ends for inserting pins to immobilize the chip (Figure 1B).

First, we select a 3D printer with a resin to produce the mold of the microchannel layer, base layer, PM valve, and external holder (Figures 1A,C). The material used for the mold should be easy to process and shape. In addition, it should be resistant to high temperature to avoid deformation in the oven and ensure that the size of the microchannel is uniform. We remove the unformed material from the surface via an ultrasonic machine and remove surface burrs by sanding. Then, the PDMS matrix and curing agent mixture are put into a vacuum drying oven to remove the bubbles. The bubble-free mixture is poured into the mold, and the tiny bubbles on the surface are blown away with airflow. Setting the temperature at 85°C for 30 min in a thermostatic oven, the formed microchannel and base layers are removed from the mold with tweezers. After removing the dust on the surface, a hole with a diameter of 1 mm is punched on the top of the reservoir. Under the condition of a 200 W power and 1.5 L/min oxygen flow, the microchannel and base layers are bombarded with plasma for 60 s. Hence, the hydrophobic surface was changed into a hydrophilic surface for the breakage of the silicon–oxygen bond.

After that, we place the silicon film between the microchannel layer and the base layer. Let the antibody strips of the silicon film intersect the S-channel of the top layer. The bubbles in the double PDMS layers were squeezed out. They were made tightly bonded using a plasma cleaner. Because the holder plates shade the reservoirs (Figure 1D), the detection antibody solution and CL substrate are added into the lower two reservoirs (Figure 1Ba,d) in advance. One holder plate is placed at the bottom of the double PDMS layer, and another one is installed at the top. Two short pins are inserted into pin holes on both ends of the holders to fasten the two holders. Last, the PM valve was loaded into the through hole, and the assembly of the chip is achieved (Figures 1B,D).

A PDMS microfluidic chip with three straight channels (500 μm wide and 300 μm deep) arranged in parallel with a spacing of 3 mm is vertically placed on the surface of the silicon film. CK-MB-Ab₁, Myo-Ab₁, and cTnI-Ab₁ are injected into the channels. After 1 h, the silicon film is cleaned with PBST buffer three times to remove the uncoated capture antibodies (Supplementary Figure S1). Then, the silicon film is assembled into the SMC. 50 μl of 5% BSA solution is added to any reservoir after SMC is assembled, and the PM valve is pressed down to the corresponding height. A negative pressure peristaltic pump is used to pump 5% BSA solution into the main microchannel until it is filled with 5% BSA solution. 5% BSA solution is incubated for 15 min to block the vacant sites on the silicon film and the microchannel to reduce nonspecific adsorption of proteins.

Fluid flow in microfluidic chips is a crucial factor for biochemical analysis (Terray et al., 2002). We employ a PM valve to switch reservoirs to the microchannel and a negative pressure system to control fluid flow. The PM valve comprises three group holes in different directions. Each group hole is 6 mm apart in an altitude interval (Figure 2F). When the hole of the PM valve connects to the reservoir, the liquid can be smoothly pumped into the main channel by the negative pressure system. When the nonhole part of the PM valve connects to the reservoir, the liquid is blocked and will not flow into the main channel. To characterize the fluid flow, we use COMSOL software to simulate the flow velocity in the microchannel (Supplementary Figure S2). Under a negative pressure of 3.5 kPa, the flow rate is stabilized.

As a mixing vessel, the Archimedean spiral microchannel can make samples, and detection antibodies released from different reservoirs mix efficiently and quickly (Viktorov et al., 2016; Lee and Fu, 2018; Raza et al., 2018) and then stimulate an immune response instead of affecting the reaction between the antigen and the capture antibody in the sample (Yin et al., 2016). To assess the blending effect of the Archimedean spiral microchannel in the SMC, we use COMSOL software to simulate the mixing effect of its 3D model under a steady state (Supplementary Figure S3A). The concentration of the two solutions at the inlet is set at 0 and 1 mol/L (Supplementary Figure S3B). The mixing efficiency at the outlet of the Archimedean spiral microchannel is 0.9999 (Supplementary Figure S3C), which can meet the mixing requirements of the next experiment.

Compared with lithography and other methods, 3D printing has a simpler process and a lower cost (Chan et al., 2017). It can solve the problems of a small size, high precision, and a complex shape of microfluidic chip channels. Its greatest advantage lies in the low production threshold; convenient, personalized design; and easy manufacture. In order to verify the reliability of this manufacturing method, we measured the deviation between the actual size and the designed size of the S-shaped microchannel (Figure 3A), the Archimedean spiral microchannel (Figures 3B,C), and the PM valve (Figure 3D). The deviations are all below 2.0%. The feasibility of this method to manufacture the precise product is verified. The uniform shape of the microchannel makes the fluid move smoothly in the microchannel, which is the prerequisite and guarantee of an automatic operation. The precise size is also beneficial to the reproducibility and stability of experimental results. SMC with high manufacturing accuracy lays a foundation for batch production in POCT.

To obtain an obvious signal, we optimize the concentration of capture antibodies on the silicon film. Silicon films are coated with different concentrations of capture antibodies (3.125, 6.25, 12.5, 25, 50, and 60 μg/ml cTnI-Ab₁; 5, 10, 20, 40, and 60 μg/ml CK-MB-Ab₁ and Myo-Ab₁). We found that the CL signal value of the sample increases upward at first and then gradually becomes a plateau with the increase of the concentration of the capture antibody when cTnI, CK-MB, and Myo are more than 25, 20, and 40 μg/ml, respectively (Supplementary Figure S4). The CL signal values of cTnI, CK-MB, and Myo show a downward trend when their capture antibody concentration is above 50, 40, and 40 μg/ml, respectively. This phenomenon might be caused by the steric hindrance effect, which decreases the sensitivity. Given both cost and sensitivity, 25 μg/ml cTnI-Ab₁, 20 μg/ml CK-MB-Ab₁, and 40 μg/ml Myo-Ab₁ are selected as the concentrations of capture antibodies.

30 μl of the detection antibody solution (10 μl of 7.2 μg/ml cTnI-Ab₂, 10 μl of 6 μg/ml CK-MB-Ab₂, and 10 μl of 9 μg/ml Myo-Ab₂) and 35 μl of the CL substrate are added into the lower two reservoirs (Figure 1Ba,d) before the assembly of the holder plates. After the chip is assembled, 30 μl of the sample (cTnI, CK-MB, and Myo, 10 μl each) and 50 μl of PBST are added into the upper two reservoirs (Figure 1Bb,c). The peristaltic pump is connected to the chip interface through a vent. The detecting process can be divided into the following steps: 1) push down the PM valve to the first stage, and let reservoir a (detection antibodies) and reservoir b (sample) connect to the main channel (Figure 1B). The peristaltic pump is opened and creates a negative pressure. The sample and detection antibody solution are simultaneously pumped into the Archimedean spiral microchannel and mixed thoroughly. Then, the solution flows into the S-shaped microchannel after the reaction (Figure 2B). 2) Turn off the peristaltic pump and incubate for 15 min. Let the mixed sample react with the capture antibodies on the surface of the silicon film sufficiently. Open the peristaltic pump and pump all the liquid into the waste chamber (Figure 2C). 3) Press down the PM valve to the second stage (Figure 1B), let PBST (reservoir c) go into the main channel for about 15 s of continuous washing until all PBST are pumped into the waste chamber (Figure 2D). 4) Press down the PM valve to the third stage (Figure 1B), and let the CL substrate (reservoir d) flow into the S-shaped microchannel, react with the HRP conjugated on the detection antibodies for 5 s, and then flow into the waste chamber (Figure 2E). A CL image analysis system is used to record the CL signal. Three S-shaped microchannels intersect with three capture antibody strips on the silicon film, yielding a total of nine (3*3) detection spots (Figure 1B). GEL-PRO Analyzer software is used for image processing to obtain the CL values of each spot. Their average CL value is subtracted from the background value to calculate the sample concentration.

The detecting performance of the chip is evaluated by the linear range and the limit of detection (LOD). Different concentrations of cTnI, CK-MB, and Myo are injected into the microfluidic chip, and the luminescence intensities are recorded. Detection curves, LOD, and the linear range of different cardiac biomarkers are obtained (Figures 4A–C). We find that cTnI has an ideal linear relationship in the range of 20–2,560 pg/ml ( Y = − 11305.16 + 13968.37 X ; R 2 = 0.98713 ). CK-MB has an ideal linear relationship in the range of 0.08–10.24 ng/ml ( Y = − 10969.63 + 11160.15 X ; R 2 = 0.98242 ). Myo has an ideal linear relationship in the range of 0.8–204.8 ng/ml ( Y = − 8378.74 + 5656.95 X ;   R 2 = 0.97872 ) (Figure 4C).

LOD is determined by 3 times the standard deviation of the blank. The LODs of cTnI, CK-MB, and Myo are 1.02, 1.37, and 4.15 pg/ml, respectively. They are far below the clinical diagnostic threshold (CK-MB 0.3–4.0 ng/ml, Myo 12–75 ng/ml, cTnI < 0.2 ng/ml). This method shows high sensitivity in detecting cardiac markers and can meet the needs of clinical diagnosis.

To evaluate the reproducibility of the chip detection, 10 chips are used to detect cTnI at 320 pg/ml, CK-MB at 640 pg/ml, and Myo at 12.8 ng/ml, which are selected in the linear range (Figures 5A,C,E). It is discovered that the inter-assay RSDs of cTnI, CK-MB, and Myo are 5.88, 5.54, and 6.73%, respectively, and the inter-assay RSDs are all less than 15%. It shows good reproducibility. This is one of the benefits of using a reliable manufacturing technique, where the precise size and stable shape of SMC allow reliable results to be obtained.

The storage stability of the chip at room temperature (25°C, 1 atm) will directly affect the chip’s performance. We test cTnI of 320 pg/ml, CK-MB of 640 pg/ml, and Myo of 12.8 ng/ml to evaluate the storage stability of SMC, which are stored for 1–7 days (Figures 5B,D,F). The results show that the coefficient of variation (CV) of cTnI, CK-MB, and Myo are 5.67, 5.14, and 5.72%, respectively, indicating that the coated capture antibodies and the preloaded reagents all maintain good activity. The chip has good storage stability, which provides a powerful guarantee for POCT applications.

In order to verify the reliability of SMC for triple detection of cardiac markers, we use SMC to detect 13 clinical serum samples at a dilution of 500 times and compare them with the results of the ELISA kit used in the hospital (Figures 6A,C,E). The results indicate that the SMC method was consistent with the ELISA method routinely used in hospitals, and four healthy samples (Nos. 1, 3, 7, and 11) and nine patient samples (Nos. 2, 4, 5, 6, 8, 9, 10, 12, and 13) are accurately detected and show a good correlation (cTnI, R 2 = 0.98729 ; CK-MB, R 2 = 0.98859 ; Myo, R 2 = 0.98243 ) (Figures 6B,D,F). It is noted that there is no significant difference in results between the two methods.

In normal human serum, the contents of cTnI, CK-MB, and Myo are below 200 pg/ml, 0.3–4.0, and 12–75 ng/ml, respectively. It is worth noting that the contents of CK-MB and Myo in serum are in the normal range for AMI positive sample 5. The SMC-CL method acquires a content of 2.79 ng/ml CK-MB and 42.6 ng/ml Myo, whereas ELISA obtains a content of 2.8 ng/ml CK-MB and 44.6 ng/ml Myo. In contrast, the result of cTnI shows a positive result for both SMC-CL and ELISA methods, which acquires a content of 242 pg/ml for the SMC-CL method and 264 pg/ml for the ELISA method. It means that the indicator of cTnI is more accurate than that of CK-MB and Myo for sample 5. There are two reasons accounted for this phenomenon. First, as an early indicator of AMI with the characteristics of early rise and fast disappearance, Myo is released into the blood after 1 h of the AMI onset and will decline at 12–24 h. When we do blood sampling after 12 h of AMI onset, Myo levels might have dropped below the diagnostic threshold. Second, when micromyocardial infarction occurs, there is not much damage in the area of the myocardial cell. The contents of CK-MB and Myo are increased above the normal level but do not exceed the diagnostic threshold. Fortunately, AMI can be confirmed by the elevation of cTnI and other clinical symptoms. Since a single indicator detecting cardiac biomarkers by the conventional method (ELISA KIT) cannot give an accurate and prompt diagnosis for AMI, multiplex detecting cardiac markers becomes crucial and essential. The SMC method we developed can detect three cardiac biomarkers simultaneously in 17 min, which meets the need of the clinical application for POCT.

To verify the advantages of SMC, we compare the colorimetric method, immunofluorescence analysis method, and surface-enhanced Raman scattering (SERS) method in terms of the number of markers, LOD value, and detection time (Table 1). It shows that SMC has higher sensitivity, a lower operating technology threshold, more portable detection equipment, and, most importantly, a significant advantage in detection time under the premise of simultaneous detection of multiple markers.