Cell spheroids culture array with modifiable chemical gradients

Abstract Cancer cell spheroids have been shown to mimic in vivo tumour microenvironment and are therefore suitable for in vitro drug screening. Microfluidic technology can provide conveniences for spheroid assays such as high‐throughput, simplifying manual operation and saving reagent. Here, we propose a concentration gradient generator based on microfluidic technology for cell spheroid culture and assay. The chip consists of upper microchannels and lower microwells. After partitioning HepG2 suspension into the microwells with concave and non‐adhesive bottoms, spheroids can spontaneously form. By controlling the fluid replacement and flow in microchannels, the doxorubicin solution is diluted automatically into a series of concentration gradients, which spanning more than one order of magnitude. And then the effect of doxorubicin on spheroids is measured in situ by fluorescent staining. This chip provides a very promising approach to achieve the high‐throughput and standardized anti‐cancer drug screening in future.


| INTRODUCTION
Cell spheroids are three-dimensional (3D) aggregations of cells. The compactness nature of cell spheroids endows them with some characteristics similar to tumour tissue in vivo. 1,2 Artificial in vitro cell spheroids provide similar metabolism and proliferation as in vivo. These properties make cell spheroids a more suitable tool for in vitro drug screening analysis than two-dimensional (2D) monolayer cell cultures. 3 For example, cancer cell spheroids have been shown to mimic in vivo tumour microenvironment and are therefore suitable for in vitro drug screening.
For cell culture, the traditional method of 96/384-well plates is labour-intensive. In contrast to traditional methods, microfluidic platforms can be used for long-term perfusion cell culture while maintaining high cell viability. Given its permeable nature, PDMS has been proven to be ideal for cell spheroid formation and long-term perfusion culture. Culturing cells and performing drug screening based on microfluidic chips is able to simplify the processes of cell loading, dispensing, medium exchange, and fluorescent labelling, and hopefully reduce the consumption of relevant reagents, which has great potential and promise for application. Current microfluidic cell spheroid formation chips can be divided into three main categories: (1) using emulsion technology, (2) utilizing U-shaped microstructures and (3) utilizing microwells. 4,5 Among them, microwells-based structures are more utilized due to their simplicity and ease of operation. However, the escape of spheroids from the microwells often influences the assay when high flow rates were used in the main channel. 6 A large number of cell analysis studies rely on testing the response of cells to a range of reagent concentrations. Especially, concentration gradient tests are widely applied in drug screening. Microfluidic chips have the advantages of miniaturization, automation, high throughput, easy adjustment and precise control, and therefore play a unique advantage in concentration gradient studies. 7 The most widely used microfluidic concentration gradient methods include tree networks and droplet microfluidics. Kim et al.
coupled a tree-like gradient generator design to an array of chambers that can be actively separated using valves. 8 This structure has a relatively large device footprint and generates a large shear force. And the geometry of the tree structure determines the achievable gradient. Droplet microfluidics can generate large numbers of water-in-oil droplets at high rates. The contents of droplets can be adjusted by regulating the chemical composition of the aqueous solution at different inlets as well as the relative flow rate. 9 However, the use of droplets brings complexity to droplet tracking and data collection. Therefore, various microfluidic techniques have been developed to divide samples into fixed positions. [10][11][12][13][14] Based on the droplet generation chip, droplets containing samples of different concentrations can be obtained by gradually adjusting the composition of the formed droplets through dilution during the droplet formation process. Sun et al. formed fixed-position droplet arrays with different sample concentrations by droplet dilution on a microfluidic device. 10 Further, the construction of concentration gradients by liquid dilution methods appears to have more development as well as wider application because of simple structure, high throughput and low demand for peripherals. Watanabe et al. implemented a highthroughput microreactor array with concentration gradients by simple structural design and liquid manipulation. 11  ing to design a simple microfluidic chip that enables the generation of concentration gradients for antibiotic susceptibility testing (AST) of Escherichia coli. 12 Zeng et al. used syringes to manually create different levels of negative pressure environment inside the chip to control the generation of concentration gradients and to measure the gradient by the area of the droplet. 13 Avesar et al. proposed a method for generating digital chemical gradients along 200 nL cell culture chambers and used computational and numerical models to predict the concentration in each chamber. 14 They demonstrated that the concentration gradient can be controlled by flow parameters. However, it has not been successfully applied to the detection of cell spheroids with large incubation chamber.
In this study, we propose a concentration gradient generator for cell spheroid culture and drug screening. With the PDMS concavebottom microwell structure, cell spheroid formation is achieved inside the chip. Combined with microfluidic technology, we construct independent units with concentration gradients to achieve rapid response to dynamic gradients and long-term stability of molecular gradients, and then use the chip for drug screening of cell spheroids.

| Chip and experiment design
The 3D structure of the chip is shown in Figure 1A

| Simulation of through-hole structure
To achieve microwell partitioning, we use through-hole structures in the chip to maintain the stability of the medium in the microwell. Due to surface tension, when the medium is removed from the channel, the medium inside the microwell is lost along with it, and cell spheroid may escape. At this point, the presence of the through-pore structure acts as a protection for the microwell contents, effectively stopping the loss of medium and attenuating the disturbance within the microwell. However, the structure may initially prevent liquids from entering the microwell.
We used a computational model to investigate the fluid entry and partitioning effect of the chip. The model included the complete geometry and the materials used were water and air. To study the effect of through-hole shear liquid to achieve sufficient retention of liquid in the concave bottom microwells, the structure was set to be filled with liquid, and air entered through the inlet to carry away the liquid in the channel. For all simulations, the results had good convergence. In addition, there is no significant difference in the results after refining the mesh density.  Figure 2C). The dark blue section in Figure 2D represents the loss of the solution. Figure 2E shows the amount of liquid left in the concave bottom microwell in the same situation without the through-hole structure. We found that the amount of liquid left in the concave bottom microwell with the through-hole structure was significantly more than that without the through-hole structure, which helped to expand the gradient range and, at the same time, the through-hole structure can avoid the loss of cell spheroids.
At the step of emptying the channel liquid, we also compared the functional effects of positive pressure operation and negative pressure operation. We found that negative pressure from the outlet provides a better liquid retention effect than pressure from the inlet. This is attributed to the fact that when positive pressure is applied from the inlet, the liquid in the microwell is simultaneously subject to In addition, the effects of different pore sizes of through-hole membranes were compared. It was found that the 100 μm pore size membrane was able to achieve liquid retention. The 200 μm pore size membrane showed a loss of liquid in the microwell and was unable to achieve greater liquid retention ( Figure S1). To maintain the function of the structure while enhancing the ease of cell inoculation, 100 μm pore size membranes were used.

| Simulation and experimentation of gradient formation and modification
To study the concentration of the substance in the concave bottom microwells, mass transfer was studied by coupling the study to a dilute substance transfer physical interface using the convection-   Figure 3B). The larger the microwell spacing, the larger the gradient range obtained. However, there is no great difference in the gradient range when the spacing is further increased (2400 μm). This suggests that the microwell distribution can be designed according to the gradient demand. But excessive spacing brings an increase in footprint and is not conducive to device miniaturization. Hence choosing the right spacing as needed is necessary.
Further, alterations in the channel shape also lead to modifications in the chip gradient effect. To investigate the gradient effect of channel shape, three different channel structures were investigated, S-shaped, narrow channel, and wide channel, as shown in Figure 3C.
We found that the gradients obtained with the S-shaped flow channel structure were closer to linear, which may be related to the adequate mixing time. The multiple contacts of liquid between channel and microwell increases the mixing efficiency.
To visualize and quantify the gradients, we used dye and 62.5 mg/mL of rhodamine B mixed with PBS, respectively, to simulate the process of generating gradient concentrations on the S-shaped channel chip ( Figure 3D). The insert in Figure 3E is a physical view of

| Design and fabrication of concentration gradient chip
The chip consists of three layers of PDMS structures bonded together, the upper PDMS (Dow Corning) channel structure, the middle through-hole structure, and the lower concave bottom microwell structure, as shown in Figure 1A The three PDMS layers were bonded together by plasma treatment to obtain the overall structure, followed by hydrophilic treatment with F127 (P2443, Sigma-Aldrich)..

| Generation of concentration gradients
The three-dimensional schematic is shown in Figure 1A.  In drug experiment, we used 1.5 μL of 40 mg/mL doxorubicin (D1515, Sigma-Aldrich) in each experiment.
In cell activity studies, live cells, as well as dead cells, were determined using the Dead Alive Dye LIVE/DEAD Cell Imaging Kit (R37601, Invitrogen) by incubation for 20 min. The cell nucleus dye Hoechst 33342 was used to help localize the cells. Fluorescent images were taken using a confocal microscope.

| DISCUSSION AND CONCLUSION
The incidence of early stage cancer has increased dramatically worldwide since 1990. Individual differences require individualized cancer treatment, and appropriate in vitro models are important for early cancer treatment. The cell spheroid model exhibits characteristics similar to those of in vivo tumour tissue. These characteristics make cell spheroids a more suitable tool for in vitro drug screening analysis than 2D monolayer cell cultures.
The chip we present here is useful for the in situ analysis of cell spheroid models. The chip avoids the loss of cell spheroids and provides a concentration gradients. Using numerical simulations, we showed that the value of shear stress in the wells of the chip is significantly low and will not affect the function and viability of cell spheroids. Using three independent gradient formation experiments, we find that the gradient formation matches the analytical model with high accuracy, demonstrating that the gradients can be reproduced experimentally. The gradient profile can be modified by adjusting the structure parameters including microchannel height, the space between wells and microchannel shape. Finally, we demonstrate the ability of the chip to study the effects of drug on the cell spheroids. Compared with microtitre plates, this chip can reduce the culture microwell volume by multiple orders of magnitude, thus greatly improving the sensitivity for cell analysis applications. In addition, microwell volumes can be easily customized by changing the concave bottom microwell size, resulting in different-sized cell spheroids.
The low-cost and simple approach is promising for highthroughput and standardized anti-cancer drug screening.

This work is supported by grants from The Strategic Priority Research
Program of the Chinese Academy of Sciences (XDA16021200) and The National Key R&D Program of China (2022YFA1104700).

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