A dynamically deformable microfilter for selective separation of specific substances in microfluidics

Abstract

To study an environmental or biological solution, it is essential to separate its constituents. In this study, a 3D-deformable dynamic microfilter was developed to selectively separate the target substance from a solution. This microfilter is a fine metallic nickel structure fabricated using photolithography and electroplating techniques. It is gold-coated across its entire surface with multiple slits of 10–20 μm in width. Its two-dimensional shape is deformed into a three-dimensional shape when used for fluid separation due to hydrodynamic forces. By adjusting the pressure applied to the microfilter, the size of the gap created by deformation can be changed. To effectively isolate the target substance, the relationship between the solution flow rate and the extent of microfilter deformation was investigated. The filtration experiments demonstrated the microfilter’s ability to isolate the target substance with elastic deformation without undergoing plastic deformation. Additionally, modification of the microfilter surface with nucleic acid aptamers resulted in the selective isolation of the target cell, which further demonstrates the potential application of microfilters in the isolation of specific components of heterogeneous solutions.

I. INTRODUCTION

We are surrounded and composed of heterogeneous solutions. For example, there are various cells in the living body, suspended particulate matter (SPM) in the atmosphere such as pollen, yellow dust, and PM2.5, as well as bacteria and viruses. Detection, isolation, and acquisition of these particles can be utilized for various purposes such as medical diagnosis, elucidation of pathological mechanisms, and treatment. Furthermore, environmental studies such as evaluating substances in the air and water and assessing their impact on living organisms can be performed (Lee et al., 2020; Kanaoka, 2019). Existing methods of capturing a target substance use fiber and porous filters that take advantage of the difference in filter pore size and target substance size. In conventional filters, substances larger than the filter pore size are collected, making it difficult to selectively capture target substances (Zhou et al., 2019; Sollier et al., 2014; Umezawa and Nuri, 1989; and Kunugi, 1990). Additionally, they capture non-target substances that are larger than or equal to the size of the target substances, and hence, it is difficult to separate the target substances from a variety of captured particles. Furthermore, methods that use dielectrophoretic force (Fabbri et al., 2013; Gascoyne and Shim, 2014; Jen, Chang, Huang, and Chen, 2012; and Chan et al., 2018) to separate molecules introduce the suspension containing the target substance into an inhomogeneous electric field to generate a difference in the dielectric constant between the target substance and the solution. This difference creates an electric field gradient that, in turn, gives rise to the dielectrophoretic force, allowing separation of the target substance. However, the problem with these methods is that the separation performance is low in a highly concentrated suspension, hindering continuous processing.

Meanwhile, methods based on antigen–antibody reactions such as ELISA (Enzyme-Linked Immunosorbent Assay) and immunostaining techniques are being widely used (Riethdorf et al., 2007; Wasniewski et al., 2013; Hammami et al., 2017; and Revzin, Maverakis, and Chang, 2012). These methods draw on the specific binding between antigen and antibody and can specifically detect the target substance if no cross-reaction occurs. The capture of a target substance using an antibody has been performed (Nagrath et al., 2007; Liu et al., 2020; Wu et al., 2019). In this study, we fabricated microfilters with the capacity to isolate target substances that cannot be otherwise isolated only by size via combining size-selective and affinity-selective isolation methods. Hence, this study aimed to develop a microfilter that can selectively isolate the target particle.

II. MATERIALS AND METHODS

A. Overview of the dynamically deformable microfilter and measurement of microfilter deformation

Figure 1 shows the schematic diagram of the microfilters. Unlike conventional filters that separate substances by pore size, these microfilters have slits. This structure allows them to be deformed from a two-dimensional to a three-dimensional form due to hydrodynamic forces. The microfilter in Fig. 1(a) has a comb-shaped structure in which deformation starts at the root by applying hydrodynamic forces, causing platelike elements to intersect with each other. The microfilter in Fig. 1(b) is a structure with multiple slits arranged in arcs, which are bent by hydrodynamic forces to form a shape resembling an insect net. These microfilters were fabricated using photolithography and electroplating techniques. Furthermore, the entire surface was gold-plated to enable modifications of molecules such as antibodies and DNA aptamers for the selective capture of the target substance. In this study, the microfilters were designed for target substances that were about the size of a cell (20 μm). Specifically, two slit widths (S) of 10 μm and 20 μm were used. The microfilter deformation was measured using a microscope USB (Universal Serial Bus) camera with a zoom lens (YCU-300 F, Yashima Optical Co., Ltd.) by comparing the filter shapes before and after deformation.

FIG. 1.

Overview of dynamically deformable microfilters. (a) Schematic of a microfilter with a comb-shaped structure. (b) Schematic of a microfilter with an arc-shaped multiple slit structure. The two microfilters with a 15 mm diameter were designed having 10 μm and 20 μm slit widths (S). The slit length (L) was 500 μm and the slit interval (P) 250 μm.

B. Isolating the target substance using the microfilter

The microfilters separate the target substance using both size separation, which results from microfilter deformation, and affinity separation by antibodies or aptamers, which modify the microfilter surface. Figure 2(b) describes the separation procedure of molecules using the microfilter. The principle of separation is as follows: the microfilter has slits finer than the target substance diameter. When a solution containing the target substance is poured into the microfilter at a constant rate, hydrodynamic forces influence the microfilter, causing it to be elastically deformed into a certain shape [Fig. 2(a)].

FIG. 2.

Target substance separation procedure using the microfilter. (a) Initial state of the microfilter. (b) Clogging of all substances to the microfilter. (c) The microfilter deformation and releasing of the non-target substance. (d) Selective isolation of the target substance.

The degree of elastic deformation can be adjusted according to the flow rate. Substances larger in size than the slit, including the target substance, are trapped by the microfilter, causing clogging of the slit [Fig. 2(b)]. If solution feeding continues, clogging increases, and simultaneously the hydrodynamic forces applied on the microfilter become progressively larger, causing the microfilter to deform further and the slit opening to become wider. In this process, the target substance remains on the microfilter surface by specific interaction with the antibodies or aptamers modified on the microfilter, while the non-target substances flow out from the slits [Fig. 2(c)]. Once the non-target cells flow out of the slits, the hydrodynamic forces applied on the microfilter become small, causing the microfilter to return to its original shape. In this way, the target substance is isolated and acquired [Fig. 2(d)]. In this respect, the presented microfilter is more advantageous than general filters (Huang et al., 2014) and filters described in other studies (Sarioglu et al., 2015).

C. Fabrication of the microfilter

The fabrication process of the dynamically deformable microfilter is shown in Fig. 3. The microfilter is a metallic structure developed using photolithography and electroplating techniques. In the photolithographic process, a spin coater (MS-A150, Mikasa Corporation) was used to apply a 30 μm-thick positive photoresist (AZ P4903, Merck Performance Materials Ltd.) on the copper substrate [Fig. 3(a)], which was subsequently exposed by UV light (1.75 J/cm²) using a mask aligner (MA-20, Mikasa Corporation) [Fig. 3(b)]. The photoresist was developed using the AZ 400K Developer (Merck Performance Materials, Ltd.) to form a resistant structure, which would serve as a mold [Fig. 3(c)]. Subsequently, the nickel structure was fabricated by an electroplating process [Fig. 3(d)]. The nickel structure was released from the copper substrate [Fig. 3(e)]. Finally, the entire surface was electroplated with gold [Fig. 3(f)].

FIG. 3.

Procedure of microfilter fabrication. (a) Photoresist coating on the Cu substrate. (b) Photoresist curing by UV exposure. (c) Microfilter mold creation by photoresist development. (d) Fabrication of the microfilter structure on the Cu substrate by nickel electroplating. (e) and (f) Removal of the nickel structure and Au electroplating to cover the structure surface.

III. RESULTS AND DISCUSSION

A. Device fabrication

Figure 4 shows images of the microfilters. Two filters (c) having different sizes of micropores (S = 10 μm, P = 30 μm; S = 20 μm, P = 60 μm) were fabricated as controls. The microfilters were designed with a diameter and thickness of 15 mm and 25 μm, respectively. The slit widths of (a) and (b) were set to 10 μm and 20 μm, respectively. The microfilters were metallic structures with a high aspect ratio (microfilter thickness/slit width = 2.5). These microfilters did not have any slit defects or surface irregularities and were fabricated with great precision. In the case of microfilters with a 20 μm slit width (designed width), the fabricated width was 18.3 μm (mean value for n = 12), and its error was only 6%. A smaller slit width is a desirable feature for separating fine target substances. The yield of the microfilters was 90%. Since the slit widths of dynamically deformable microfilters can be controlled by adjusting the flow rates of liquids passing through these microfilters, they are expected to be applied on target substances of various sizes.

FIG. 4.

Images of the fabricated microfilters. (a) Photographs of the fabricated microfilter with a comb-shaped structure. (b) Photographs of the fabricated microfilter with an arc-shaped multiple slit structure. (c) Photographs of the fabricated 2D filter having a micro-hole array. 3D microfilters have great precision and an elastic structure, which help them return to their original shape when deformed by hydrodynamic forces. A 2D filter is a filter manufactured with the same process and materials used for 3D filters. The 2D filter has holes of 10 μm or 20 μm in diameter arranged in an array and does not deform into a three-dimensional shape.

B. Measurement of filter deformation

The basic properties of the microfilters were evaluated. The microfilter was fixed to the filter holder, as shown in Fig. 5(a), and inserted into the housing. The experimental setup is shown in Fig. 5(b) and was used to measure the extent of filter deformation. It was connected with a peristatic tube pump, and the degree of deformation was measured at seven different flow rates (5, 10, 15, 20, 25, 30, and 40 ml/min).

FIG. 5.

Overview of the experimental design to separate the target substance. The microfilter is installed in the device housing (a), and a suspension containing the target substance is transferred to the device using a tube pump (b).

The relationship between the flow rate and the degree of microfilter deformation is shown in Fig. 6. In microfilter (a), the deformation was approximately 110, 350, and 820 μm at flow rates of 5, 20, and 40 ml/min, respectively. In microfilter (b), the deformation was approximately 40, 60, and 90 μm at 5, 20, and 40 ml/min, respectively. The deformation of both microfilters increased linearly with the increase in the flow rate. The deformed microfilters returned to their original state when the liquid flow was stopped, suggesting that the deformation was within the elastic range under the flow rate conditions set for this experiment (5–40 ml/min). A comparison between microfilter (a) and (b) revealed that the deformation was smaller in microfilter (b), suggesting that it was more appropriate for the filtration of fine particles. Microfilter (a) had a relatively simple comb structure, and the deformation started from the base of the comb structure. Therefore, the deformation can be calculated from the triangle similarity. For example, at a flow rate of 5 ml/min, if the total slit length was about 6 mm, the overall deformation was approximately 100 μm. This indicates that the slit gap was more than 20 μm at a 1.2 mm point from the deformed base.

FIG. 6.

Evaluation of deformation of microfilters [(a) and (b)]. The deformation of the microfilter increases linearly with an increasing flow rate.

For microfilters (a) and (b), finite element analysis (ANSYS 19.2, ANSYS Inc.) was used to analyze the microfilter deformation (Fig. 7). Conditions used in this analysis are shown in Table I. The deformation of the entire microfilter (a) at a 5 ml/min flow rate was approximately 120 μm. Almost no deformation was observed at the center of microfilter (b), though the outer slit of this microfilter was greatly deformed. The deformation of the entire microfilter (b) at the flow rate of 5 ml/min was approximately 40 μm, and the slit gap was at most 20 μm. These results suggest that the fluid passed at a flow rate lesser than 5 ml/min separates the target cell most efficiently. In addition, the measured deformation and simulation results were compared to each other; under the condition of a flow rate of 5 ml/min, microfilter (a) showed a deformation (simulation value) of about 120 μm, and the measurement result was 110 μm, whereas microfilter (b) showed a deformation (simulation value) of about 40 μm, and the measurement result was 40 μm. Thus, these values showed good agreement. Furthermore, as for microfilters (a) and (b), structural and fluid coupled analysis was performed to analyze the shear stress; no clear correlation was found between the deformation of the microfilter and shear stress (Fig. 2 in the supplementary material).

FIG. 7.

Analysis of the deformation of microfilters [(a) and (b)] by the finite element method using ANSYS. The amount of displacement from the reference plane when the hydrodynamic force applied to the microfilter from top to bottom is shown. Microfilter (a) deformed from the base of the comb structure, and the amount of displacement was 120 μm at most. In addition, the slit gap deformation of microfilter (b) showed almost no deformation in the range of φ6 mm from the center and was about 10 μm at φ8 mm and at most 20 μm at φ10 mm. Therefore, almost no deformation was noted at the microfilter center, and the slit gap deformation became greater toward the outer edge.

TABLE I.

Conditions for microfilter deformation analysis.

Item	Boundary condition
Analysis model	Axis-symmetric
Young's modulus	200 GPa
Number of nodes	1 049 655
Number of elements	414 361
Element type	PLANE182 (2D four-node structural solid)

C. Filtration efficiency of PMMA particles

Filtration experiments were conducted using monodisperse poly (methyl methacrylate) spherical particles (PMMA, Soken Chemical Co., Ltd.) with a mean particle size of 19.3 μm, which was about the same size as the target substance. The PMMA particles were mixed with ultrapure water and passed through the microfilter. The particles in the flow-through were counted using the hemocytometer, and the filtration efficiency was obtained using formula (1),

$$(n_{in}-n_{out})/n_{in}\times 100.$$ (1)

Here, n_in and n_out denote the number of PMMA particles before and after the solution was passed through the microfilter, respectively. The concentration of the PMMA particles was adjusted to 1.0 × 10⁶ particles/ml before the solution was passed through the filter. As a control, the same experiments were performed using filter (c). Three different flow rates, 1, 5, and 10 ml/min, were used. The experiments were performed in triplicate. The filtration efficiency for the PMMA particles in microfilters (a)–(c) by flow rate is shown in Fig. 8.

FIG. 8.

Evaluation of the filtration efficiency of PMMA particles.

In all filters, the filtration efficiency for the PMMA particles increased with a decreasing flow rate. In 3D filters (a) and (b), the filtration efficiency was thought to decrease as the slit gap widened with an increasing flow rate, allowing the PMMA particles to pass through the filter without being trapped. Furthermore, with respect to the slit width of microfilters (a) and (b), the efficiency for a slit width of 20 μm was higher than that for a slit width of 10 μm. The maximum filtration efficiency was about 80% at a flow rate of 1 ml/min. A possible reason for the low efficiency of the microfilter with a slit width of 10 μm is that the trapped PMMA particles clog the slit, causing the flow channels to become narrow and accelerate the fluid passing through the microfilter. Herewith, it is conceivable that the trapped particles were washed away, or the amount of deformation of the microfilter was increased and may not be trapped. Furthermore, with the slit width of 20 μm and a flow rate of 5 ml/min, microfilter (b) showed a filtration efficiency of about 40%, approximately twice that of (a). Microfilter (b) had a smaller degree of deformation than that of microfilter (a), and hence, was able to capture more PMMA particles. Figure 9 shows an SEM image of the three-dimensional filters (a) and (b) obtained after the PMMA particle filtration. It was confirmed that the dynamically deformable microfilters could capture the target substances with their filter structures. It is important to note that the PMMA particles trapped in the microfilters had fallen away from the slit space of the microfilters during our observations. Moreover, the filtration efficiency of the 2D filters was around 60% at the minimum flow rate of 1 ml/min, indicating that the 3D filters were more effective in trapping substances with the size of a cell.

FIG. 9.

Image of PMMA particles captured by microfilters. (a) Result of using the comb structure microfilter. (b) Result of using the arc-shaped multiple slit structure.

D. Target cell separation using the microfilter

The target particle size is 20 μm, which is similar to the size of a living cell. Separation based solely on size has the following problems: target substances may pass through the filter; non-target substances of the same or greater sizes may simultaneously be captured. Cancer cells were selected to demonstrate the feasibility of separating small target cells from a solution containing an enormous number of foreign substances. Cancer cells are known to circulate in small numbers in the blood of patients (Plaks, Koopman, and Werb, 2013; Cristofanilli et al., 2004; Scher et al., 2015; Hou et al., 2012; Bouab et al., 2019; Muhanna et al., 2015; Iliescu et al., 2019). Our study's technique can be applied to cancer diagnosis if the cancer cells can be specifically separated from the blood. Since it is known that epithelial cell adhesion molecule (EpCAM) is highly expressed on the surface of many cancer cells (Baeuerle and Gires, 2007; Went et al., 2004; Patriarca et al., 2012), microfilter (b) was modified with an EpCAM aptamer (5′-TTT TTT TCA CTA CAG AGG TTG CGT CTG TCC CAC GTT GTC ATG GGG GGT TGG CCT G-3′) that was capable of specifically binding to EpCAM. The point by point binding between the aptamer and EpCAM is not strong. However, EpCAM is abundantly expressed on the cancer cell membrane, and aptamers bind to it; thus, it has the same binding power as that of antibodies. Since multiple aptamers modified on the substrate and multiple EpCAMs on the cell membrane bind to each other, it has been shown that the binding force is strong when viewed in terms of the whole cell (Song et al., 2013). Subsequently, a self-assembled monolayer (SAM layer) was formed by adding 10 μM 6-hydroxy-1-hexanethiol solution to the microfilter to minimize non-specific absorption. This allowed the non-target cells, such as blood cells, to pass through the microfilter without being adsorbed onto the microfilter surface, whereas only target cells were specifically trapped and collected on the microfilter surface.

As a control, scrambled DNA (5′-TTT TTT TTT TCG TGG TGA GAA GTC GAG TGG TGC TGC TCC AGT TGC TGT CCA TCG TCC G-3′), consisting of the same nucleobase sequence as that of the EpCAM aptamer, was evaluated. MDA-MB-453 (human breast-cancer-derived cell) and HEK293 T (human embryonic kidney 293 T cells) were used as the controls for target and non-target cells, respectively. The cells were prepared at a 5.0 × 10⁴ cells/ml concentration and poured into the microfilter at a flow rate of 0.5 ml/min. Subsequently, the number of captured cells per unit area was evaluated by staining cell nuclei and observing them by fluorescence microscopy.

The fluorescence images of cells captured by the microfilter are shown in Fig. 10. When the microfilter surface was modified with the EpCAM aptamer, cancer cells were specifically captured [Fig. 10(a)], while non-specific absorption of normal cells was inhibited [Fig. 10(b)]. When the microfilter surface was modified with the scrambled DNA, only a small number of cancer cells and normal cells were absorbed by the microfilter, and non-specific absorption was inhibited [Figs. 10(c) and 10(d)]. In contrast, in a preliminary experiment in which the aptamer was not modified with a microfilter, almost no cancer cells could be captured.

FIG. 10.

Fluorescence images of cancer cells (MDA-MB-453) and normal cells (HEK 293 T) on the microfilter modified with the aptamer or scrambled DNA. In the case of using the microfilter modified with the EpCAM aptamer, (a) cancer cells were specifically captured, and (b) non-specific adsorption of normal cells was inhibited. In the case of using the microfilter modified with scrambled DNA, (c) cancer cells were not specifically captured, and (d) non-specific adsorption of normal cells was almost inhibited.

Figure 11 shows the number of cells captured per unit area by the microfilter. The number of captured cancer cells and normal cells were 64 and 1.7, respectively, and the number of cancer cells captured by the specific capture effect of the EpCAM aptamer was about 40 times that of the normal cells. In contrast, when the microfilter surface was modified with the scrambled DNA, the number of cancer cells and normal cells captured were 0.7 and 10, respectively, showing that the non-specific absorption was inhibited.

FIG. 11.

Cancer and normal cells captured by the aptamer-modified microfilter.

The objective is to separate the target substance from a large amount of foreign substances. To prove that this objective can be achieved, we carried out the cancer cell isolation experiments by mixing normal cells and cancer cells. The total number of cells was 5.0 × 10⁴ cells. The ratios of normal and cancer cells were 9:1, 1:1, and 1:9. The number of captured cells per unit area are summarized and compared between 2D filter (c) and 3D microfilter (b) modified with the EpCAM aptamer in Fig. 12. In the case of the microfilter with a slit width of 10 μm, the number of cancer and normal cells captured were 4.1 and 1.4, respectively, when the cancer to normal cell ratio was 1:1. These results indicate that approximately three times more cancer cells were captured as compared to normal cells. Similarly, at a 1:9 ratio of cancer to normal cells, the number of cancer and normal cells captured were 2.1 and 4.7, respectively. These results demonstrate that the filter could inhibit non-specific absorption and selectively separate cancer cells in a solution where the number of normal cells was nine times greater than that of cancer cells. Compared with the microfilter with a slit width of 20 μm, a slit width of 10 μm was able to capture more cancer cells. This result implies that as the microfilter slit width became narrower, the cancer cells were size-selectively trapped onto the microfilter surface, allowing an affinity-selective specific capture of cells. Besides, since the main component of cells is water, hydrodynamic forces deform the shape of cells on the filter, effectively increasing the contact area and yielding a more efficient capture of cells. These results indicate that the cancer cells were continuously captured without being released from the microfilter by the specific binding effect of the EpCAM aptamer on the microfilter surface, even when the cancer cells clogged the microfilter and the hydrodynamic force applied to the microfilter increased. In addition to the function to selectively capture the size of 20 μm or more, similar to conventional filters, experiments showed the microfilter applicability in the capture and separation of various substances. In this way, the optimization of the flow rate for specific binding of the EpCAM aptamer, the target substance, and changing the microfilter shape resulted in the specific separation of the target cells.

FIG. 12.

Comparison of cancer and normal cell separation using different three-dimensionally deformable microfilters. (a) Result of using the microfilter with a 10 μm slit width. (b) Result of using the microfilter with a 20 μm slit width.

IV. CONCLUSIONS

Dynamically deformable microfilters were fabricated to selectively isolate the target substance from a solution. The flow rate, the degree of deformation, and the filtration efficiency were evaluated as basic properties of the microfilters. It was demonstrated that no plastic deformation of the microfilters occurred at the set flow rate range of 5 to 40 ml/min, and the deformation occurred within the elastic range. The two microfilters were deformed linearly in accordance with the flow rate. The filtration efficiency at a flow rate of 1 ml/min was approximately 80% for both filters (a) and (b), and they successfully isolated the target substance in a size-selective manner. The filtration efficiency of filter (c) with pores was approximately 60% under the same condition. Furthermore, modification of the microfilter surface with nucleic acid aptamers, which specifically bind to cancer cells, demonstrated that cancer cells could be selectively separated from a pool of heterogeneous cell types.

SUPPLEMENTARY MATERIAL

See the supplementary material (Figs. 1 and 2) for a fluorescence image of cancer cells (MDA-MB-453) on the microfilter modified without the aptamer, and simulation analysis results of shear stress for microfilters (a) and (b), respectively. Please refer to Table 1 in the supplementary material for the conditions set during the microfilter shear stress analysis.

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author upon reasonable request.