Principle of NP detection with VIF. VIF introduces local perturbation in a micrometric space. By applying a fast and small oscillation, a second-order streaming flow with a steady component emerges around the solid microstructures. Hayakawa et al.39–41 showed that a local swirling flow was induced around a cylindrical micropillar by applying rotational vibrations with amplitudes and frequencies of a few micrometers and several hundred hertz, respectively. This phenomenon was characterized in detail both experimentally and numerically42,43.
We utilize this phenomenon to agitate the sample fluid suspending the target NPs and capture MPs with a specific affinity (Fig. 1). Initially, both particles are dispersed in microfluidic space. A local whirling flow around the pillars is induced when circular vibration was applied to the entire substrate, which stirs the NPs and capture MPs. Through this process, the NPs are adsorbed onto the surface of the large-affinity capture MPs. As the adsorption of the NPs proceeds, the capture MPs form aggregates as they collide because the NPs serve as a binder. The concentration of NPs in the sample is estimated from the extent of aggregation of the capture MPs51–53.
Effectiveness of VIF system. First, we examined the behavior of a mixture of particles in the VIF system. We fabricated a substrate with micropillars (100 µm in diameter and 50 µm in height) of poly-dimethylsiloxane (PDMS) using a standard micromolding process (Fig. 2a‒c). These pillars were arranged in a 10 × 10 matrix with center-to-center spacing of 200 µm. Spacers with a height of 50 µm were placed at the corners and sides to define the height of the fluid volume bounded by a cover-glass as a lid. After the oxygen-plasma treatment to make the substrate hydrophilic, the PDMS substrate was mounted on a piezo-drive stage connected to the power and signal sources using a double-sided tape (Fig. 2d, e).
We employed an avidin–biotin affinity binding pair, which is well known for its high specificity and binding strength (Kd < 10− 15 M), as a model NP-capture system. Avidin-coated capture MPs (3 µm nominal diameter) and biotin-coated NPs (diameter measured using an imaging analyzer was 152 ± 55 nm; see Figure S1, Supporting Information) were mixed by gentle pipetting and placed on the substrate with an array of micropillars. Subsequently, a cover glass was carefully placed on the substrate to define the boundary conditions of the fluid volume (Fig. 2e).
A few sample snapshots (bright-field microscopic images) of the particles in the micropillar substrate are shown in Fig. 3a–c. Here, the sample mixture contained 5.7 × 109 /ml NPs and 3.0 × 107 /ml capture MPs. The vibration condition was fixed as circular vibrations with a frequency f = 500 Hz and an amplitude A = 3.3 µm. Before the application of vibrations (0 min), the capture MPs were evenly dispersed in the observed field under all conditions, although NPs could not be observed because of their smaller size. This indicates that pre-mixing by pipetting in a microtube did not cause aggregation of capture MPs. After 15 min of applying the vibrations, several relatively large aggregates of capture MPs (above 25 µm in diameter in the image) were formed in the mixture containing NPs in addition to the capture MPs (Fig. 3a, lower panels). We conducted a control experiment without VIF and NPs. In the absence of VIF, no aggregate of the capture MPs was observed (Fig. 3b, lower panels). This result reflects the fact that Brownian diffusion alone is not sufficient to induce collision and aggregation of capture MPs. In the absence of NPs, relatively small aggregates (diameter less than 10 µm) were generated (Fig. 3c, lower panels). This result may be due to some intermolecular forces between the capturing MPs that caused nonspecific adsorption; however, the average area of aggregate was significantly smaller than that in the experiment where NPs were present in the sample. When the same microscopic observation was performed for the sample subjected to vortex mixing in a test tube at 500 rpm instead of VIF, no obvious aggregates were observed even after 1 h (see Figure S2, Supporting Information). The actual stirring of capture MPs and NPs with VIF is shown in Movie S1, Supporting Information. These results visualized local vortices generated between four adjacent pillars in the VIF device. On the other hand, we expect that vortices in a microtube are large in the case of conventional vortex mixing, meaning NPs and affinity capture MPs tend to move in parallel so that efficient collision does not occur. Thus, we suppose vortices induced by local flow in the VIF lead to the efficient aggregate formation of affinity capture MPs, rather than conventional vortex mixing often employed in commercial kits for NP detection using capture MPs.
This initial set of experiments provided the basis that only when there was an affinity between the target NPs and the capture MPs, the localized flow around the pillar (Fig. 3d) effectively induced aggregation of the capture MPs, and this phenomenon can be used for NP detection. Specifically, it is considered that microscale local swirling promoted the collision of capture MPs and NPs, and that the aggregates further encountered each other at the interference of the neighboring vortices, forming larger aggregates.
Relationship between flow field and capture-MP aggregation. We investigated the relationship between capture-MP aggregation and VIF driving conditions. To measure the flow field around the pillars, we used a micro-particle-image velocimetry (micro-PIV) technique. Figure 4a shows a two-dimensional (2D) vector plot of the net flow velocity at the center plane of the pillar (h = 25 µm). Swirling flow occurred around the pillar, with slightly lower velocities at the top, bottom, left, and right of the pillar. This deceleration was due to the interference with adjacent pillars that were separated by an interval of 200 µm.
In the case of VIF, vibration parameters such as frequency and amplitude influence the localized flow field40. In this study, the vibration frequency f was varied to 200 Hz, 500 Hz, 800 Hz, and 1000 Hz, and the amplitude A was varied to 1.5 µm, 3.3 µm, and 5.2 µm. Furthermore, Fig. 4b shows the representative velocity magnitude of the mean flow field (magnitude of the azimuthal velocity at r = 60 µm from the center or 10 µm away from the pillar wall) for each vibration condition (2D vector plots for each vibration condition are shown in Figure S3, Supporting Information). The induced velocities increased with the increase in vibration frequency and amplitude, suggesting that the induced velocity can be readily tuned using these parameters.
Subsequently, we analyzed the extent of aggregation induced by these flow fields. For simplicity, we considered the area of the MPs aggregates as an index, which could be readily obtained via image binarization and standard particle analysis (see Figure S4, Supporting Information). The dependence of the average area of capture-MP aggregates after 15 min on the velocity magnitude of the mean flow field is plotted in Fig. 4c. The concentrations of NPs and capture MPs were 5.7 × 109 /ml and 3.0 × 107 /ml, respectively. The average area of aggregates increased with velocity magnitude in the range below 300 µm/s. In this region, the higher the induced velocity, the more likely it is that the collision of MPs proceeds such that aggregates grow in size. However, at velocities above 300 µm/s, the aggregate size decreased. This result indicates that shear stress acted on the MP aggregates and caused them to break, resulting in smaller aggregates (the average area of aggregates for each vibration parameter is shown in Figure S5, Supporting Information). Although the largest aggregate formation was obtained at condition f = 500 Hz and A = 5.2 µm, the variability among triplicated experiments was relatively large. Thus, we employed the condition f = 500 Hz and A = 3.3 µm, which exhibited the smallest variability, in the following experiments.
Quantification of NP concentration. We examined the dependence of the average area of MP aggregates on the concentration of NPs to test the capability of the proposed system for quantitative measurements. The time variation of the average area of the capture- MP aggregates is shown in Fig. 5a, where the vibration conditions were: f = 500 Hz and A = 3.3 µm. In the presence of NPs, the average area of the capture-MP aggregates increased almost linearly with time in all cases (NP conc. = 5.7 × 109 /mL, 1.1 × 109 /mL, 5.7 × 108, /mL, and 5.7 × 107 /mL). Microscopic images and beeswarm plots of the area of the capture-MP aggregates after 15 min for each NPs concentration are shown in Figures S6 and S7, Supporting Information, respectively. These results support the hypothesis that the amount of NPs adsorbed onto the capture MPs increases with the concentration of NPs, promoting the formation of aggregates. In contrast, in the control experiment without NPs (gray line) and without VIF (black line), the increase in the average area of capture-MP aggregates remained small. These results quantitatively show that the concentration of NPs in the sample correlates with the average area of aggregates in the VIF field. Figure 5b shows the relationship between the NP concentration and the initial slope of the time-dependent increase in the area of aggregates. We chose this parameter rather than the size at the plateau, longer time mixing (~ 40 min) did not produce a clear and reproducible plateau (see Figure S8, Supporting Information). It is likely that, as the number of aggregates decreases, and statistically significant average was not obtained within a single frame. In addition, the total projected area of capture MPs remained nearly constant over time. This result indicates that the aggregations grew in a two-dimensional fashion. From this plot, NP concentration in the sample can be estimated. For the linear region, linear regression analysis was performed with the fitting equation of y = 1.8781 log x − 32.055 and an R2 value of 0.8388. The NP concentration of 5.7 × 107 /mL showed a significantly greater slope compared to the case without NPs, suggesting that the NP concentration can be quantified by determining the slope of the time-dependent increase in aggregate size above this concentration.
Application in EV detection. The abovementioned results demonstrate the effectiveness of the proposed VIF platform in detecting NPs using affinity capture MPs. However, its viability in diagnostic applications is still unclear, because the affinity between avidin and biotin is higher than that of other biological affinity-capture systems. Thus, we further examined the performance of the proposed system in detecting EVs, important biological NPs present in body fluids. We used EVs derived from bovine milk as the target NPs. To detect the EVs, Tim-4 conjugated magnetic MPs, which exhibit an affinity for phosphatidylserine on the EV membrane in a calcium-ion-dependent manner, were used. The diameter of the magnetic capture MPs provided by a manufacturer was 3 µm, and that of EV measured with an imaging analyzer was 219 ± 75 nm (see Figure S10, Supporting Information).
Figure 6a–c shows bright-field images at 0 min and 15 min of vibration. The conditions of the circular vibrations were: f = 500 Hz and A = 5.2 µm. We used an amplitude value that was larger than that in the previous experiment because the magnetic MPs were heavier, and therefore, it was more difficult to move along the flow compared to polystyrene MPs. Before applying the vibrations (0 min), the magnetic capture MPs were dispersed in the observed field of view, whereas after 15 min of vibration, many aggregates were observed (Fig. 6a, lower panels). As control experiments, we performed experiments without VIF and EVs (Fig. 6b, c). The trend of the results was similar to those of the experiments conducted with the avidin–biotin capturing model system; the aggregate of the affinity magnetic capture MPs were not formed in the case without VIF, and the aggregate size of the magnetic capture MPs was small in the case of without EVs (actual stirring of the affinity magnetic capture MPs and EVs with VIF are shown in Movie S2, Supporting Information). Furthermore, in the case of bulk vortex mixing in a test tube at 500 rpm instead of VIF, aggregates were not formed even after 1 h of mixing (see Figure S11, Supporting Information). These results suggest that even in the case of EVs, the local flow in the VIF effectively induced capture-MP aggregation only when there was an affinity between the target EVs and the capture MPs.
Figure 7a shows the time variation in the average area of the magnetic capture-MP aggregates. When vibrations were applied to the substrate on which the sample containing EVs and MPs was placed, the average area of the aggregates increased with time in all cases (EV conc. = 6.4 × 109 /mL, 1.28 × 109 /mL, 6.4 × 108 /mL, and 6.4 × 107 /mL). Microscopic images and beeswarm plots of the area of the capture-MP aggregates after 15 min at each EV concentration are shown in Figure S12 and S13, Supporting Information. The aggregate size did not increase much in the control experiments without NPs (gray line) and without VIF (black line), confirming that aggregation was induced by EV adsorption. Similar to the case with avidin–biotin MPs, the average area of capture-MP aggregates after 15 min of VIF increased with increasing EV concentration (Fig. 7b). We obtained a significantly higher EV concentration of 6.4 × 107 /mL over the negative control, indicating that the presence of EVs at this concentration could be detected. There was a linear relationship between the increasing slope and increasing EV concentration and the linear equation of y = 2.2861 log x – 39.511 and R2 = 0.9233.