Scalable large-area mesh-structured microfluidic gradient generator for drug testing applications

Abstract

Microfluidic concentration gradient generators are useful in drug testing, drug screening, and other cellular applications to avoid manual errors, save time, and labor. However, expensive fabrication techniques make such devices prohibitively costly. Here, in the present work, we developed a microfluidic concentration gradient generator (μCGG) using a recently proposed non-conventional photolithography-less method. In this method, ceramic suspension fluid was shaped into a square mesh by controlling Saffman Taylor instability in a multiport lifted Hele–Shaw cell (MLHSC). Using the shaped ceramic structure as the template, μCGG was prepared by soft lithography. The concentration gradient was characterized and effect of the flow rates was studied using COMSOL simulations. The simulation result was further validated by creating a fluorescein dye (fluorescein isothiocanate) gradient in the fabricated μCGG. To demonstrate the use of this device for drug testing, we created various concentrations of an anticancer drug—curcumin—using the device and determined its inhibitory concentration on cervical cancer cell-line HeLa. We found that the IC₅₀ of curcumin for HeLa matched well with the conventional multi-well drug testing method. This method of μCGG fabrication has multiple advantages over conventional photolithography such as: (i) the channel layout and inlet-outlet arrangements can be changed by simply wiping the ceramic fluid before it solidifies, (ii) it is cost effective, (iii) large area patterning is easily achievable, and (iv) the method is scalable. This technique can be utilized to achieve a broad range of concentration gradient to be used for various biological and non-biological applications.

I. INTRODUCTION

Microfluidics concentration gradient generators (μCGG) are widely used for various chemical and biological studies including chemical synthesis,¹ serial dilution,^2,3 sensors,^4,5 clinical diagnostics,⁶ chemotaxis,⁷ drug screening,⁸ biochemical assays, etc. μCGG provides precise control over the concentration of the active molecule in space and time.⁹ In such flow-based systems, mostly two fluid streams are combined and allowed to mix or diffuse, in “T” or “Y” shaped channels.¹⁰ Christmas tree design is another widely used flow-based device where a gradient is generated by continuous splitting and recombining fluid streams.¹⁰ The well-controlled shapes of μCGG are commonly fabricated using lithography-based techniques, including photolithography, soft lithography, e-beam lithography, etc. These methods comprise multi-step processes and need sophisticated instruments, thereby increasing processing cost and limits industrial scale-up.¹¹ Recently, to overcome these limitations, non-conventional methods have also been reported including paper-based^12,13 or thread-based¹⁴ gradient generation. However, in both paper and thread-based μCGG, it is difficult to collect the outlet streams as fluid is absorbed in the device.

In this work, we report the fabrication, optimization, and validation of a flow-based μCGG for drug screening applications. The device was made with poly-dimethyl siloxane (PDMS) based soft-lithography technique using a template which was fabricated using a non-conventional method of shaping ceramic fluid in a lifted Hele–Shaw cell as reported earlier.¹⁵ Controlled shaping of fluid allows customizing number of outlets, channel layout, and total area of pattern, thereby controlling gradient strength in μCGG. The optimum operating conditions to generate a concentration gradient were determined using COSMOL simulations and validated by experiments with fluorescent dye FITC. The outlet concentration profile was shown to be controlled by changing the flow rate ratios of both inlets. Furthermore, the inhibitory concentration (IC₅₀) of curcumin on cervical cancer cells (HeLa) was estimated using the fabricated μCGG (∼32 μM). The result matched well with the same estimated using conventional drug testing platforms, i.e., multi-well plates (34.9 ± 1.7 μM). We propose that this easy-to-fabricate and scalable μCGG can be used for the generation of concentration gradients for various biological and non-biological applications.

II. MATERIAL AND METHODS

A. Fabrication of square mesh microfluidic device

Fabrication was carried out in a parallel lifting Hele–Shaw cell custom-built at Suman Mashruwala Advanced Microengineering Laboratory, IIT Bombay. The details of the setup can be found in Ref. 15. The cell plates of the setup were modified to suit the proposed application.

1. Controlling source holes

Using a CNC-micro drilling machine, source-holes were drilled on one of the cell plates for fabricating the proposed μCGG structures. Diameter of source-holes in this study was kept equal to 0.5 mm for all the experiments. Locations of these holes were based on the guidelines given in Ref. 16 such that a square mesh structure could evolve after the process completion. The square mesh with several sizes and pitch dimensions were prepared.

2. Square mesh preparation

The measured amount of shear thinning suspension consisting of ceramic (alumina) particles suspended in photoresist (HDDA) [as reported by Islam and Gandhi¹⁶] [Fig. 1(a)(i)] was squeezed between two cell plates to the required thickness and radius [Fig. 1(a)(ii)]. The array of source holes on one of the cell plates was decided as mentioned in Sec. II A 1. During the squeezing step, all the source holes were properly sealed. Squeezing was followed by a small delay to neutralize the normal stresses within the fluid. The cell plates were separated with unsealed holes [Fig. 1(a)(iii)]. As the two plates started separating, air moved inside from the source holes, creating the air finger. Each air finger growth was shielded by adjacent air finger which finally gave a square mesh pattern due to pre-designed source hole locations.¹⁷ The Saffman–Taylor instability (viscous fingering) phenomenon was controlled by conducting fabrication with a proper capillary number and aspect ratio.¹⁵ Once the plate separation was completed, channel layout, i.e., square mesh pattern was formed in center of each plate and surrounded by connecting and continuous boundary [Figs. 1(a)(iv)–1(a)1(iv')]. This unwanted boundary was removed by wiping, without affecting the channel layout or dimensions [Fig. 1(a)(iv')]. The presence of the photo-initiator in the ceramic suspension allowed it to UV cure [Fig. 1(a)(v)] and finally a cross-linked square mesh pattern was formed. Figure 1(a)(vi) shows the digital image of a 9 × 9 square mesh template used as the template for the study. Changing distance between drilled holes and number of holes can be utilized to customize the pattern and patterned area. Using same parameters, we then fabricated various mesh patterns (as shown Fig. S1 in the supplementary material).

FIG. 1.

Schematic of preparation of (a) template and (b) microfluidic device.

3. Preparation of microfluidic device

PDMS (Sylgard 184 from Dow Corning) was mixed in 10:1 (w/w) with a crosslinker and then degassed under vacuum. This uncured PDMS was then poured onto square mesh pattern prepared [Fig. 1(b)(i)] and cured at 60 °C for 12 h. The casted cured PDMS was then peeled off from the template [Fig. 1(b)(ii)] and is referred here as the negative replica. The inlet and outlet holes of 1 mm diameter were created in the negative PDMS replica using biopsy punches [Fig. 1(b)(iii)]. All the outlet positions were made equidistant from the mesh layout. In order to make microfluidic device, this negative replica was plasma oxidized and bonded to glass slides by bringing two surfaces in conformal contact [Fig. 1(b)(iv)]. Figure 1(b)(v) shows photos of the fabricated microfluidic device for CGG using the mesh layout formed.

B. Template characterisation

The height and width of the channels and nodes were measured using white light interferometry (WLI). WLI resulted in a 3D profile of area scanned.

C. Flow setup

The square mesh-shaped microfluidic device can be customized in terms of inlet-outlet number and position. In this work, fabricated μCGG have two inlet ports and 10 outlet ports—alternate channels [Fig. 2(a)]. Capillary tubes of 1 mm diameter were used for connecting the syringe and the inlet ports of the device. Outlets were collected in 200 μl microtips [Fig. 2(b)]. A syringe pump (Era's NE-1000) was used for the experiment and programmed for the desired flow rate.

FIG. 2.

Schematic of (a) inlet–outlet positions and (b) experimental setup for μCGG.

D. Characterization of concentration gradient

1. COMSOL simulation

The 2D geometry of the device was made in a COMSOL Multiphysics geometry module. The meshing was done to discretize the domains into small elements, as is done in the finite element method. The concentration at the outlets was simulated using COMSOL (version 5.2) assuming laminar flow conditions. Other COMSOL parameters are mentioned in Table I. In addition to equal flow rates in both inlets, we also checked the effect of the inlet flow rate ratio on the outlet concentration profile (Table I).

TABLE I.

COMSOL simulation parameters.

Software	COMSOL Multiphysics 4.4
Physics type (in COMSOL)	Laminar flow and transport of diluted species
Material	Water
Viscosity	0.001 Pa s
Density	1000 kg m−3
Material	FITC
FITC diffusivity in water	4.25 × 10−10 m2/s
Boundary conditions
Inlet velocity condition (equal inlet velocity)	0.001–0.02 m/s
Inlet1:Inlet2 flow rate	50–100, 50–150, 50–200, 200–50, 150–50, 100–500 μl/min
Inlet concentration	0 and 0.05 mol/m3
Outlet condition	Constant pressure
Walls	No slip condition

2. Fluorescein dye gradient

A sodium fluorescein dye (FITC-376 Da) in MilliQ water was used to generate gradients at different flow rates in the fabricated microfluidic device. The collected outlet was imaged in a fluorescence microscope (EVOS FL-Auto). The concentration was determined by comparing fluorescence intensities of known concentration, i.e., standard curve of intensity vs concentration (Fig. S3 in the supplementary material) with intensities from outlets from device.

3. Curcumin gradient

The concentration gradient of model drug, curcumin (389 Da), was generated in cell growth media at a flow rate of 50 μl/min. The absorbance of the outlet solution was measured at 425 nm using a SpectraMax M2e microplate reader. To estimate outlet concentration, a standard curve with absorbance or known concentration was plotted (Fig. S4 in the supplementary material) and outlet concentration of curcumin from the device was determined.

E. Cell culture

The cervical cancer cells (HeLa) were cultured in high glucose Dulbecco's modified eagle medium (DMEM) supplemented with 1% anti-anti, 1% l-glutamine and 20% fetal bovine serum (HiMedia). Cells are trypsinized with 0.05% trypsin—EDTA(1×) (Gibco), incubated for 5 min at 37 °C. To get pellet of cells, it was centrifuged at 1000 rpm for 5 min. After centrifugation, cells were resuspended with a fresh medium and counted using a hemocytometer and seeded in a 96 well plate.

F. MTT assay

HeLa cells were seeded in a 96 well cell culture plate at 5000 cells per well. After 24 h, cells were exposed to different concentrations of curcumin drug (in growth media) prepared by (i) μCGG and (ii) manual dilution. Cells were allowed to proliferate for 24 h and cell viability was estimated by quantifying reduction of tetrazolium dye—MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] by monitoring absorbance at 590 nm using a SpectraMax M2^e Microplate reader. IC₅₀ value, i.e., 50% inhibition of cell viability—indicates cytotoxicity of drug on cells, was determined by nonlinear regression fitting of the concentration-dependent proliferation inhibitory data.¹⁸

G. Statistical analysis

All experiments were performed using three independent sets. Graphs were plotted using OriginPro 9.1 software.

III. RESULTS

A. Square mesh dimension

The dimensions of square mesh can be varied by changing the initial thickness of the fluid film and the pitch distance of the source holes in the plate.¹⁵ The optimized parameters for the preparation of the square mesh were selected from a source hole characterization graph.^15,16 The intersections of branches in the mesh are referred to as nodes. The shape of the channels and the nodes are shown in Figs. 3(a) and 3(b), respectively. The template used in this study has an average channel height of 58.9 ± 5.75 μm and width of 650.7 ± 38.14 μm. The average height of node is 109.2 ± 22 μm as measured by a white light interferometer (WLI). Detailed dimensions at various locations are provided in Fig. S2 in the supplementary material and in the associated table.

FIG. 3.

WLI image of the (a) channel and (b) node.

B. Concentration gradient characterization

Using COMSOL simulation, we analyzed the concentration at the outlet streams at various flow rates. Figure 4(a) shows the simulation result for the concentration gradient formed in the device with equal inlet flow rates, i.e., 50 μl/min. The simulation result was verified with further experiments using 50 μM FITC (fluorescein dye) in water as one inlet and pure water as other, with an equal flow rate of 50 μl/min [Fig. 4(b)] and 100 μl/min [Fig. 4(c)] in both the inlets. Furthermore, the flow rate of one inlet was varied keeping the other constant [Figs. 4(d)–4(e)]. Figure 4(f) compiles different outlet concentration profiles obtained with different inlet flow rate ratios showing that the concentration profile can be controlled by changing inlet flow rates. Figure 4(g) shows fluorescence variation across the device outlets for equal flow rate of 50 μl/min.

FIG. 4.

Gradient characterization from COMSOL for (a) inlet a: 50 μl/min, inlet b: 50 μl/min, i.e., equal inlet flow rate of 50 μl/min; comparison between outlet concentrations obtained from experiment and simulation at flow rate of (b) inlet a: 50 μl/min, inlet b: 50 μl/min (N = 3) and (c) inlet a: 100 μl/min, inlet b: 100 μl/min for both the inlets (N = 3). If both inlets have different flow rate, concentration profile changes as in (d) inlet a: 150 μl/min, inlet b: 50 μl/min and (e) inlet a: 50 μl/min, inlet b: 150 μl/min; (f) shows outlet concentrations with different combinations of inlet flow rates and (g) fluorescence intensity (green) of FITC gradient from experiments at inlet a: 50 μl/min, inlet b: 50 μl/min flow rate. (In COMSOL, red = 50 μM concentration of FITC, blue = 0 μM concentration of FITC in water.)

C. Drug gradient and its effect on cell metabolic activity

The potential application of concentration gradient is in determining the effect of various drugs/biomolecules on cells. We selected curcumin as the model drug to study its cytotoxic effect on HeLa cells, a commonly used model cell line for cervical cancer. First, the concentration gradient of curcumin in growth media was generated in the fabricated device. The concentration of curcumin at each outlet was determined using absorbance of the solution at 425 nm [Fig. 5(a)]. Using the MTT assay, we then tested the cytotoxicity of curcumin on HeLa cells by measuring the percent inhibition with respect to the drug collected from the device outlets [Fig. 5(b)]. As the graph shows, drug concentration coming from outlet 5 gives the IC₅₀, i.e., inhibitory concentration at which 50% cell are viable/dead. Comparing with Fig. 5(a), the concentration of solution from outlet number 5 is found to be ∼32 μM. Therefore, IC₅₀ of curcumin on HeLa cells is ∼32 μM. To note, our COMSOL simulation estimated that the concentration at outlet 5 is 28 μM which matches well with the experimental measurement [Fig. 4(b)]. Hence, drug concentration at different outlets can be estimated using COMSOL simulation by providing molecular weight and diffusivity of drug as input and thus eliminating the need of experimental measurement of drug concentration every time. Furthermore, to validate the device, IC₅₀ of curcumin on HeLa was estimated using a conventional 96 well plate in which concentration gradient was generated manually by serial dilution. Using this conventional method, IC₅₀ was found to be 34.9 ± 1.7 μM [Fig. 5(c)] which matched well with IC₅₀ estimated from our μCGG device (32 μM). This result indicates that the μCGG can potentially be used for concentration generation in drug testing and drug screening replacing the time consuming, labor intensive, and error-prone method of creating concentration gradients manually.

FIG. 5.

(a) Curcumin concentration at different outlets: determination of IC₅₀ using (b) fabricated μCGG and (c) conventional 96-well plate.

D. Scalability of the fabrication process

The fabrication process used for the mesh pattern generation is spontaneous and scalable. As the shaping of fluid is created by varying air entrance pathways, changing the initial experimental parameter like initial fluid volume, initial fluid film thickness, initial squeezed fluid film radius and pitch distance of the array of source holes, can change the micro-mesoscale structure of template. The parametric study was performed by Kanhurkar et al and the stability was expressed in terms of capillary number, initial non-dimensional plate separation, and initial ratio of the interface radii.¹⁹ This work demonstrates that the pattern can be predicted through numerical simulations,¹⁷ which can then be utilized while scaling up or down the grid size of template. In order to demonstrate the potential of the LHSCs in the customized channel layout, we fabricated 5 × 5 [Fig. 6(a)], 11 × 11 [Fig. 6(b)], and 9 × 9 [Fig. 6(c)] over different pattern area and thereby channels of varying sizes. As the number of channels/outlets can be varied, broader or narrower drug concentration gradient can be generated. Therefore, this multiscale fabricated mesh pattern based microfluidic device can also be scaled up or scaled down for concentration gradient generation.

FIG. 6.

Scalability of the fabrication process: (a) 5 × 5 mesh, (b) 11 × 11 mesh, and (c) mesh patterning over a large area.

IV. DISCUSSION

In this study, we used a mesh like template generated by shaping ceramic fluid controlled via lifted Hele–Shaw for developing the microfluidic concentration gradient generator (μCGG). The outlet concentration with respect to flow rate was determined using COMSOL simulations and validated using experiments with fluorescein dye-FITC. Next, the gradient of curcumin in growth media was estimated using the absorbance of curcumin. The drug testing application in fabricated μCGG was successfully demonstrated by measuring IC₅₀ of curcumin on model cell line—HeLa (cervical cancer cells). Further, consistent IC₅₀ value from both—conventional well-plate assay and microfluidic devices—depicts potential use of device in concentration dependent studies.

For generation of the complex concentration gradient profile, “Christmas-tree” based designs^20,21 are widely used. It has been used for biological studies including cellular migration,²² proliferation, and neuronal differentiation²³ and non-biological studies such as chemical synthesis,¹ serial dilution,^2,3 sensors,^4,5 etc. However, most of the lithography-based templates and respective μCGG are complex and involves multi-step fabrication,¹¹ thereby increasing cost and time of fabrication. To overcome these limitation, some non-conventional methods have been reported such as paper based high-throughput devices,¹² thread based microfluidics,^14–24 etc. However, in these devices collection of solution is difficult as the studies are majorly conducted inside the device. These shortcomings are addressed in the microfluidic device presented in this paper. We have demonstrated microfluidic devices with 10 outlets; however, number of channels/outlets can be easily changed by varying number and position of drilled holes, thereby controlling gradient strength and resolution. Therefore, fabrication of lithography-less template using lifted Hele–Shaw cell is advantageous in terms of scalability, cost of fabrication, and customizable inlet and outlet of fabricated μCGG. This study highlights the potential of lifted Hele–Shaw cell in fabrication of microfluidic devices for drug testing applications and can be further explored for other biological and non-biological applications.²⁵

V. CONCLUSION

Our work demonstrates that mesh-like shaping of ceramic fluids can be utilized to fabricate microfluidic devices for microfluidic concentration gradient generation. The application of fabricated μCGG was demonstrated by testing curcumin-drug gradient on cervical cancer cells (HeLa). We believe that this simple, customized number of outlets, cost effective, and scalable μCGG can be used in future for investigating various concentration gradient-based biological and non-biological studies.

SUPPLEMENTARY MATERIAL

See the supplementary material for different templates prepared using same parameters using LHSCs, positions at which dimension measurements were taken, calibration curve of FITC intensity with respect to the known FITC concentration and curcumin absorbance with respect to known the curcumin concentration.

AUTHOR DECLARATIONS

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

P.T. and K.B. have contributed equally to this work.

Shital Yadav: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Pratik Tawade: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Validation (equal); Writing – review & editing (equal). Ketaki Bachal: Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – review & editing (equal). Makrand A. Rakshe: Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Writing – review & editing (equal). Yash Pundlik: Formal analysis (equal); Investigation (equal); Software (equal); Validation (equal); Writing – review & editing (equal). Prasanna S. Gandhi: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Writing – review & editing (equal). Abhijit Majumder: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Supervision (equal); Visualization (equal); Writing – review & editing (equal).

DATA AVAILABILITY

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