Reproducibility of microchamber structures constructed based on wax reflow
In our previous studies [35–37], we used hemispherical polymer beads (2 mm in diameter; SAYAKOBO, Yokohama, Japan) to create deep localized microchamber structures instead of the wax reflow process (Supplementary Fig. S2). The polymer beads were manually glued onto each SU-8 chamber pattern as a mold using an epoxy adhesive (Araldite; Huntsman Japan, Kobe, Japan) at room temperature for 12 h. This fabrication process led to low reproducibility of the microchambers, i.e., the low controllability of the adhesive layer thickness and excess adhesive being squeezed out at the bonding interface decreased the replication accuracy, particularly at the intersection of the microchannel and the entrance of the microchamber. Such process failures sometimes led to unexpected air bubbles being trapped in the microchamber (Fig. 2a). This unfavorable result was most likely caused by a non-uniform flow on a step or around an obstacle at the entrance of the microchamber. The probability of sample dispensing without air bubbles was 87.0% in experiments with 20 devices and a total of 100 chambers. By implementing the thermal reflow process using wax, the probability of dispensing without air bubbles was markedly improved up to 100% in experiments with 20 devices and a total of 100 chambers (Fig. 2b).
Sequential liquid dispensing using a double-faced stop valve
In this study, we improved the pressure resistance performance of valve S2 by changing its geometrical configuration. As shown in Fig. 3a, the gap distance of valve S2 is determined by two convex structures facing each other embedded on both sidewalls of the microchannel (referred to as a capillary stop valve [40]). In our devices, the valve structure of S2 is hereinafter referred to as a double-faced stop valve, whereas the S1 valve is a single-faced stop valve. The theoretical burst pressure of the double-faced stop valve can be derived as follows:
$$\begin{array}{c}P\left(g\right)=-\gamma \left(\frac{2\text{c}\text{o}\text{s}\left(\text{m}\text{i}\text{n}\left({\theta }_{m}+\beta ,180^\circ \right)\right)}{g+r\left(1-\text{cos}\beta \right)}+\frac{\text{c}\text{o}\text{s}{\theta }_{m}+\text{c}\text{o}\text{s}{\theta }_{f}}{H}\right)\#\left(4\right)\end{array}$$
The gap distance of the double-faced stop valve S2 was experimentally measured to be g2 = 21.4 µm (Fig. 3), resulting in a theoretical burst pressure P2 of 5.85 kPa for r2 = 6.4 µm in a microchannel (W = 202.1 µm and H = 59.0 µm). The theoretical burst pressure of the single-faced stop valve with the same gap distance (g2 = 21.4 µm) was calculated to be P2 = 4.47 kPa. Thus, the burst pressure can be increased by 1.3 times compared to that of the single-faced stop valve by implementing the double-faced stop valve for S2. It should be noted that five pairs of S2 were arranged in series (Fig. 3a) because we assumed that the gap distance (g2) and/or corner radius (r2) of one or more valves would not be the design value because of unexpected process failures in soft lithography. Thus, the one of the five valves that has the highest burst pressure (P2) determines the actual valve performance limit in microfluidic devices. For comparison, the gap distance g1 of 40.5 µm for the single-faced stop valve S1 resulted in a theoretical burst pressure P1 of 2.79 kPa. The microchannel lengths L1, L2, and L3 were set as 5.0 mm, 1.0 mm, 0.2 mm, as shown in Fig. 3a.
Figure 3b shows that an array of 10 microchambers was filled completely with water colored with blue food color (0.1% w/v) when we used the pressure-driven micropump at a flow rate of 10 µL/min. As expected, the possible number of microchambers dispensed when using the double-faced stop valve S2 was increased by 1.7 times as compared to the six microchambers that could be achieved at the same flow rate with the single-faced stop valve S2 used previously [37]. However, as predicted by the theoretical dispensing model, the possible number of microchambers dispensed decreased to five and two when the flow rate was increased to 20 µL/min (Fig. 3c and Video S1) or 50 µL/min (Fig. 3d), respectively.
Sequential sample dispensing using an air plug-in valve
According to the dispensing theory described in Eqs. 1–4, the maximal allowable flow rate is limited by an increase in the flow resistance ΔP(L1), which increases in proportion to not only the flow rate, but also the number of microchambers filled. When using the geometric dimensions of the microchannels and a pair of passive stop valves as designed in this study, the maximal allowable flow rate for dispensing 10 microchambers was limited to approximately 10 µL/min. It should be noted that a further reduction of the gap distance (g2 = 20 µm as a design value) of the permanent stop valve S2 would improve the pressure resistance (P2), but would negatively affect the patterning accuracy of the SU-8 mold by photolithography.
Therefore, we propose a newly designed valve configuration for the permanent stop valve, which we termed an “air plug-in valve.” As shown in Fig. 4a, a set of two double-faced stop valves S2 was placed downstream and upstream in the upper microchannel (hereinafter referred to as the air exhaust microchannel) for exhausting the air in the microchambers. Five pairs of S2 were arranged in series to avoid the risk of unexpected failures in the fabrication process as mentioned in the previous section. The most distinctive feature of the air plug-in valve proposed here is that air is trapped in the part of the air exhaust microchannel connecting two S2 valves of neighboring microchambers when the next microchamber is completely filled with liquid (Fig. 4a). For example, both S2 valves of the first microchamber are subjected to the total pressure given on the right side of Eq. 5a while the second microchamber is being filled with liquid. This condition corresponds to the case where m = 1 in Eq. 1, and the pressure applied to the S2 valves reaches the maximum right before the liquid reaches the S2 valves of the second microchamber. However, once the air trapping between the S2 valves of the first and second microchambers is completed, the pressure applied to the S2 valves of the first microchamber decreases to the pressure given on the right side of Eq. 5b because the liquids reaching the two valves push against each other through the air plug, resulting in the offsetting of the applied pressure.
$$\begin{array}{c}{P}_{2}>{P}_{1}+\varDelta P\left({L}_{1}\right)+\varDelta P\left({L}_{2}\right)+\varDelta P\left({L}_{3}\right)\#\left(5\text{a}\right)\end{array}$$
$$\begin{array}{c}{P}_{2}>\varDelta P\left({L}_{1}\right)\#\left(5\text{b}\right)\end{array}$$
It can be concluded that the burst pressure P2 of the permanent stop valves S2 should be designed only to satisfy the constraint given by Eq. 5a, i.e., they are required to have flow resistance only until the liquid has been fully dispensed into the next microchamber. This also means that the limitation to the possible number of microchambers dispensed has been removed by implementing the air plug-in valve for S2.
Figure 4b–d shows that all 10 microchambers were sequentially filled with water containing green food color (0.1% w/v) at flow rates of 10, 30, and 70 µL/min (Video S2). Thus, the air plug-in valve allowed successful dispensing into all 10 microchambers at flow rates below 70 µL/min. According to the constraint given by Eq. 5a, the theoretical maximal allowable flow rate is limited to approximately 70 µL/min for dispensing 10 microchambers given the geometric dimensions of the microchannels and the pair of passive stop valves in the design used in this study. Thus, the experimental results satisfactorily agreed with the theoretically predicted maximal allowable flow rate. The allowable flow rate achieved with the implementation of the air plug-in valve was seven times higher than that obtained with the previous valve arrangement using double-faced stop valves and 14 times higher than that obtained with the single-faced stop valves used in our previous study (the maximal allowable flow rate was 5 µL/min for 10 microchambers) [37]. Notably, the dramatic improvement in the pressure resistance of the permeant stop valves allowed us to introduce the liquid manually into the microfluidic device (Video S3; estimated mean flow rate: ∼40 µL/min).
Unexpected behavior of the air plug-in valve
Unexpectedly, we observed that the liquid–air meniscus, which had previously been pinned by the air plug-in valve, gradually leaked downstream and, at the first microchamber, both upstream and downstream, as the dispensing proceeded (yellow arrows in Fig. 4b–d). This unexpected valve leakage behavior was most likely due to the gas permeability of PDMS [41]. The leakage length (LL) seemed to increase not only with an increasingly upstream position of the valve in the device, but also with an increase in flow rate. Figure 5a shows LL at the air plug-in valve downstream of each microchamber at flow rates of 10–70 µL/min. The values of LL were estimated right after the 10th microchamber was completely filled. LL gradually decreased as the valves were located further downstream until it reached 0.8 mm, after which it decreased with a larger negative slope. Five pairs of S2 were arranged in series in the air exhaust microchannel, resulting in a total length of 0.8 mm between the back edge of the first valve convex structure (set to the zero position of LL) and the back edge of the fifth valve (Fig. 4a). The transition region near LL = 0.8 mm was most likely due to the geometric differences in the microchannel, i.e., periodic fluctuations in the microchannel width due to the facing convex structures embedded on both sidewalls of the microchannel.
To investigate the effect of air leakage at the interface of the PDMS device and the silicone-based adhesive double-sided tape, we prepared a modified PDMS device that was covalently bonded to a thin PDMS layer coated on a 4-inch glass wafer by air plasma surface treatment. The LL values measured for the modified PDMS device are also plotted in Fig. 5a. It can be seen that there was no significant difference between two types of devices. Thus, negligible air leakage occurred at the interface between the PDMS device and adhesive tape. In addition, although all air plug-in valves S2 were subjected to an applied pressure described in Eq. 5b right after the air trapping between adjacent valves was completed, the air plug trapped between the first and second microchambers exhibited a larger internal pressure than any other air plugs located downstream, i.e., the internal pressure increased with an increase in the number of microchambers that had been dispensed, according to Eq. 1. Therefore, LL can be considered to increase not only at air-plug in valves located further upstream, but also at a higher flow rate because of an increase in flow resistance that can be estimated by Eq. 2.
In Fig. 5b, the data for the PDMS device covalently bonded to a thin PDMS layer coated on a 4-inch glass wafer are replotted as a function of the product of the internal pressure (P) applied to each air plug and the natural logarithm of time (ln(t)). P was theoretically calculated according to Eq. 1, and t was experimentally determined as the total time elapsed between a certain air plug being completed and all 10 microchambers being completely filled with water. As shown in the graph, under the conditions of LL ≥ 0.8 mm, LL is expressed regardless of the flow rate and valve position as follows:
$$\begin{array}{c}LL\propto P\times \text{ln}\left(t\right)\#\left(6\right)\end{array}$$
where the coefficient of determination (R2) was estimated to be 0.823, indicating that the assumption presented in Eq. 6 is valid. Therefore, we can theoretically predict LL even during the design process of microfluidic devices. A fundamental way to overcome the leakage problem is to replace PDMS with engineering plastics with lower gas permeability, such as polymethyl methacrylate, polycarbonate, or cyclo-olefin polymer.
Multiplexed LAMP assay for the detection of food allergens
We explored the possibility of the rapid, simultaneous detection of multiple food allergenic substances i.e., specific DNA targets from wheat (T. aestivum), buckwheat (F. esculentum), and peanut (A. hypogaea), by the colorimetric LAMP method using a microfluidic device with a circular arrangement. With the aim of providing a microfluidic diagnostic device for the simultaneous genetic detection of seven allergens in processed foods, we compactly housed 10 microchambers in a circle instead of using a single row format. As shown in Fig. 6, in a device with a circular arrangement, green-colored water (0.1% w/v) could be successfully sequentially dispensed into all microchambers at a flow rate of 30 µL/min (Video S4). Our sequential liquid dispensing method provides substantial flexibility in the design of microfluidic devices, thus allowing the geometric design of 10 microchambers arranged in a circle and compactly housed within the 20-mm outermost diameter formed by the air exhaust microchannel. The geometric dimensions of this type of device were designed as L1 = 1.52 mm, L2 = 0.80 mm, and L3 = 0.89 mm. The theoretical maximal allowable flow rate for dispensing water into the 10 microchambers was estimated to be below 160 µL/min, the gap distances were g1 = 40.5 µm and g2 = 21.4 µm for valves S1 and S2, respectively, the corner radius at the back edge of the convex structures was r1 = r2 = 6.4 µm, and the width and height of the microchannel were W = 202.1 µm and H = 59.0 µm, respectively.
Three specific primer sets for the detection of wheat, buckwheat, and peanut were pre-spotted and dried in reaction chambers Nos. 2 and 7; Nos. 3 and 8; and Nos. 4 and 9, respectively. Additionally, universal primer sets to identify all plant species were pre-spotted in chambers Nos. 5 and 10 as a positive control, whereas no primers were pre-spotted in chambers Nos. 1 and 6 as a negative control. Figure 7a shows an experimental result for the detection of wheat-specific DNA (total DNA concentration: 1.0 ng/µL) by the colorimetric LAMP method using the microfluidic device. After the wheat DNA sample mixed with the LAMP reagents was introduced into the device, the LAMP assay was conducted in a hot water bath at 60°C for 60 min. As expected, positive reactions, manifested by a color change of HNB (150 µM) in the LAMP reaction solution from violet to sky blue, were clearly observed in chambers 2, 5, 7, and 10, without false-positive and false-negative results. LAMP assays with buckwheat DNA and peanut DNA yielded true-positive results in chambers 3 and 8 (Supplementary Fig. S3a) and 4 and 9 (Fig. S3b), respectively, as well as in chambers 5 and 10 used as a positive control.
Figure 7b shows the simultaneous detection of multiple food allergens in a mixture of wheat DNA and buckwheat DNA (total DNA concentration: 1.0 ng/µL each) by the LAMP assay; positive reactions occurred in chambers 2, 3, 5, 7, 8, and 10, without any cross-contamination. Multiplexed LAMP assays with other combinations of the DNA samples, i.e., mixtures of wheat DNA and peanut DNA (Fig. S3c) and buckwheat DNA and peanut DNA (Fig. S3d), also yielded true test results. Similarly, a mixture of all three DNA samples (1.0 ng/µL each) yielded positive reactions in all chambers except chambers 1 and 6, used as a negative control (Fig. 7c). In contrast, when DNA extracted from a tea plant (1.0 ng/µL) was introduced into the device, a color change from violet to sky blue was observed only in the positive control chambers (Nos. 5 and 10), without false-positive reactions in the other chambers (Fig. 7d).
In our previous study [37], we developed a method for quantitatively assessing color differences between a positive reaction (sky blue) and a negative reaction (violet) after the LAMP assays, based on a representation of the CIE L*a*b* color space (CIELAB). In the CIELAB color space, L* indicates lightness, and a* and b* are chromaticity coordinates. In brief, first, the RGB values for each microchamber were quantified using ImageJ (version 1.52a, National Institutes of Health, Bethesda, MD, USA) and subsequently converted to XYZ tristimulus values in the CIE XYZ color space. Then, the color parameter values L*, a*, and b* were calculated using the XYZ values. The calculation method has been previously thoroughly described [42]. Figure 8 shows the distributions of color differences plotted in the a*–b* chromatic plane of the CIELAB space from the LAMP results presented in Fig. 7 (similarly, the data presented in Fig. S3 are plotted in Supplementary Fig. S4). The result revealed that the colors of the positive and negative groups were clearly separated after the LAMP assays were run for 60 min, for any combination of the three food allergens and tea plant used as a negative control.
In conclusion, with the aim of improving the performance of PDMS-based microfluidic devices employed for the rapid and easy-to-use multiplexed genetic diagnostics, we proposed a newly designed valve configuration, termed an air plug-in valve. The pressure resistance performance of the air plug-in valve was remarkably enhanced after the air trapping between facing valves was completed. By replacing the previous valve arrangement using double-faced stop valves with the air plug-in valves, which was used as a permanent stop valve, the maximal allowable flow rate significantly increased by up to 14 times (~ 70 µL/min) for sequentially dispensing a liquid to fill an array of 10 reaction microchambers. Further, the limitation to the possible number of microchambers dispensed was removed by implementing the air plug-in valve. We demonstrated that the fabricated microfluidic device enables the simultaneous detection of three plant food allergens (wheat, buckwheat, and peanut) by colorimetric LAMP assays run at 60°C for 60 min in a single operation, without cross-contamination between the reaction microchambers. It should be noted that the diagnostic device could be designed with a geometrical configuration of 10 reaction microchambers arranged in a circle and compactly housed within a 20-mm outermost diameter because our sequential liquid dispensing method provides substantial flexibility in the design of microfluidic devices.
In future studies, we will develop a rapid and easy sample-to-answer diagnostic platform that allows the simultaneous detection of not only all seven allergenic ingredients that must be labeled under Japanese law, i.e., wheat, buckwheat, peanut, egg, milk, shrimp, and crab, but also foodborne pathogens (e.g., Salmonella enterica, S. aureus, Campylobacter jejuni, and E. coli O157:H7) using our microfluidic device, to ensure food safety and security. In principle, it is possible to flexibly customize the type of nucleic acid target of interest (DNA/RNA); thus, this highly versatile technology can be applied to the on-site multiplexed genetic detection of not only allergens, but also infectious diseases caused by various pathogens (viruses, bacteria, fungi, and parasites), poisonous plants, and illegal drugs, without a need for laborious and multiple operations.