Microfluidic device for chemical and mechanical manipulation of suspended cells

Microfluidic devices have proven to be useful and versatile for cell studies. We here report on a method to adapt microfluidic stickers made from UV-curable optical adhesive with inserted permeable hydrogel membrane micro-windows for mechanical studies of suspended cells. The windows were fabricated by optical projection lithography using scanning confocal microscopy. The device allows us to rapidly exchange embedding medium while observing and probing the cells. We characterize the device and demonstrate the function by exposing cultured fibroblasts to varying osmotic conditions. Cells can be shrunk reversibly under osmotic compression.


Introduction
Microfluidic devices are increasingly used for biophysical experiments, both to study bio-macromolecules, for example in x-ray probes of protein structures [1] or to study cells, for example in optical stretcher experiments designed to investigate cell viscoelasticity [2]. A major advantage of microfluidic devices for cell biophysics is the ability to perform high-resolution optical microscopy in the device, and to achieve high throughput for screening or sorting applications. Fast response of cells can be probed by rapid change of conditions or addition of chemicals [3].
Microfluidic stickers (MFS) are a kind of microfluidic device constructed using UV-curable optical glue. They have been shown to be suitable for culturing cells and tissues [4,5].
The UV-curable Norland Optical Adhesive NOA-81 is a versatile liquid photopolymer for low-cost microfluidic chip production [6]. Compared to polydimethylsiloxane (PDMS), the more widely used polymer for microfluidics, NOA-81 is entirely transparent and non-scattering in the visible range (refractive index of 1.56). It was developed for fast and transparent bonding of optical components and fiber optics. Additionally, its low auto-fluorescence and high transmission in the near UV range makes it the most promising optical glue among the members of the NOA series [7]. In comparison to PDMS, however, the fluorescence emission of NOA-81 is slightly higher or comparable. Nevertheless, fluorescence cell imaging in micro-devices made of NOA-81 shows a relatively high signal to noise ratio, such that it looks to be an ideal material for live cell imaging (see supplementary figure (stacks. iop.org/JPhysD/51/045403/mmedia)).
NOA-81 shows good chemical resistance to organic solvents, impermeability to oxygen and water vapor, persistence against swelling upon contact with fluids, and stability under S Supplementary material for this article is available online (Some figures may appear in colour only in the online journal) Original content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
surface treatments e.g. by oxygen plasma [6,8]. A droplet of NOA-81, exposed to UV light, cures within a few seconds to a hard elastic polymer. The peak of sensitivity is around 365 nm, and curing time depends on light intensity and thickness of the liquid layer that has been applied [9]. A further advantage of NOA-81 resin is its high Young's modulus (1 GPa), that allows for channels with large aspect ratios while still maintaining straight rigid walls. After curing, it shows very good adhesion to glass, metals or plastics without swelling or leakage for up to months. In contrast, swelling due to water absorption is still problematic for PDMS-based devices [4].
It is challenging to isolate cells from mechanical perturbations in microfluidic devices while rapidly changing chemical conditions. One method addressing this problem is the insertion of porous windows in microfluidic flow chambers that allow small molecules to rapidly pass between channels without producing substantial hydrodynamic forces [10,11].
We have here adapted a method to study cells suspended in such a chamber while exposed reversibly to changes of osmotic conditions. We report on the design and fabrication of a versatile microfluidic device with a three-channel configuration that allows us to temporally and spatially control the concentration of osmolytes in the central channel from the side channels. We fabricated MFS of NOA-81 and inserted hydrogel membrane micro-windows (HMM) using scanning confocal microscopy for optical projection lithography.
The microfluidic device was designed such that a rapid exchange of small molecule solutes is possible between the channels while decoupling the rapid fluid flow in the outer channels from the inner channel. The shielding from flow makes it possible to, for instance, study changes in mechanical response of individual cells during rapid changes of solvent conditions without exposing the cells to additional forces from fluid flow. By flushing the side channels, we could also rapidly reverse the medium changes.

Microfluidic sticker design
A three-channel microfluidic device configuration (figure 1) with a central sample channel located between two parallel outer channels with permeable windows inserted in the walls is an established design to generate concentration gradients across the central channel or introduce symmetrical concentration profiles in the sample channel by diffusion [10]. Windows in the wall are formed by hydrogel membranes. We designed chambers with three channels of 150 µm width and a pair of 50 × 100 µm 2 rectangular windows on opposing sides of the central channel in 40 to 80 µm high channels (figure 1).
The geometry of the windows plays a role for both, the hydrogel polymerization and for the transport of solutes through the windows. Windows form more uniformly when the channel depth is low and exposure for polymerization is shorter. Diffusive flux through the windows scales with their cross sections and decreases with increased windows thickness.
Height and width of the central channel needs to be adapted to the dimensions of the cells. Typical cell diameters are between 10 and 15 µm, and since cells swell or shrink under different osmolarity, devices with different channel heights were fabricated (ranging from 35 to 70 µm). The width of the side channels was chosen to be 150 µm, and the central channel width was between 150 µm. Using NOA-81, we could fabricate large aspect ratio devices with a length up to 4 cm, without stickers getting torn or deformed while peeling them from the PDMS molds.

Microfluidic sticker preparation
The preparation of stickers followed an established procedure [4,5] and is illustrated in figure 2. MFS were fabricated by soft imprint lithography using a negative PDMS stamp, UV-polymerizable resin NOA-81 (Norland Products Inc., Cranbury, NJ, USA), and a flat PDMS mold. The negative PDMS stamp was made by replica molding of a SU-8 photoresist mold produced by standard photo-lithography (using a patterned chromium on glass mask). A drop of UV curable resin NOA-81 (500-700 µl) was poured on a flat PDMS mold (figure 2) and sandwiched with the structured PDMS stamp. After 2-3 min of UV exposure (365 nm), the PDMS stamp was peeled off and after punching inlet/outlet holes at the end of the channels, replaced with a cover slip cleaned with isopropanol (IPA). The flat PDMS mold together with the NOA sticker on the coverslip, was exposed to UV light for the second time (2 min) through the flat PDMS. In this step, NOA further polymerizes so that the PDMS mold could be detached from the flat PDMS substrate, while the sticker remains tightly bound to the coverslip. The NOA sticker attached to the cover slip was flipped over and exposed to UV exposure for the last time (2 min). UV radiation time was optimized based on the aspect ratio and geometry of the channels. After this step, the inlet/outlet connections were fixed with small acrylic rings and a drop of 5 min epoxy glue to the sticker.

Photo-polymerization of HMM
Photo-polymerization is a relatively new method of lightbased lithography. Photo-initiated polymerization has become a popular technique for hydrogel formation in situ, owing to its adjustable temporal and spatial resolution [12]. Minimum feature resolution, depth of focus (DOF), photo-initiator, hydrogel precursor response, light source, oxygen contribution and sample thickness are the key parameters in such photo-lithography process.
The smallest feature size of a given pattern that the projection system can print faithfully is proportional to λ/NA, where λ is the illumination wavelength and NA the numerical aperture of the projection lens.
DOF scales like λ/(NA) 2 , and determines the focus error that still gives an acceptable lithographic result. It limits how much the fabrication plane can be displaced while resulting structural features remain sufficiently sharp. High resolution and large DOF are mutually exclusive, creating an optimization problem. Since in our case structural details are not extremely fine, but the window height is large, we opted for low resolution and large DOF using a low NA objective (NA = 0.45).
Oxygen is a major inhibitor of photo-polymerization. Since the NOA-81 sticker and the glass cover slip are both impermeable to oxygen, long UV exposure was not an issue in our fabrication process [13].
We used poly-ethylene-glycol-diacrylate (PEG-DA) as material for photo-polymerization with 405 nm UV light. Chemical composition, concentration (volume ratio) of photoinitiator and intensity of UV light will change the gelation time [12,14,15]. Fabrication of hydrogel membrane micro windows (HMMs) in a microfluidic device, has been introduced previously [10]. Here, we essentially followed the published procedure using UV photo-polymerization to insert HMMs in the gaps left in NOA walls (windows in figure 1) of the channels of our microfluidic sticker.
First, we prepared an aqueous hydrogel precursor solution made of 95% v/v polyethylene glycol diacrylate (PEG-DA, n = 400, 01871-250 Polysciences Inc.) and 5% v/v photo-initiator (2-hydroxy-2-methylpropiophenone, 405655, Sigma-Aldrich). Then, the whole chamber was filled with the hydrogel solution. UV induced polymerization of PEG-DA was performed with a Leica TCSSP5 scanning confocal microscope, defining a scanning pattern such that windows were fabricated in the appropriate size and locations.
The intensity provided by the 405 nm diode laser in the focal spot of the 10x objective (NA = 0.45), was 400-500 mW cm −2 (50 mW of laser power passing through the objective). The sample was mounted on a motorized stage while the objective was focused on the mid plane of the channels. As shown in figure 3, the scanning mirror was programmed to sweep a rectangular region of interest (ROI) of 70 × 130 µm 2 that was positioned to cover the gap in the chamber wall (extending slightly further to anchor the edge of the 50 × 100 µm 2 windows). 1-3 s of illumination with 95% of maximum intensity was enough to polymerize the HMMs to the full height of 40 to 80 µm deep channels. Immediate rinsing of uncured precursor solution, after illumination was important. This was done by 10-20 min continuous wash with water.

Bio-compatibility
Neither PEG hydrogels, nor NOA-81 are poisonous to the cells. Biocompatibility of microfluidic devices made of NOA-81 has been reported previously [5]. MFS have also been used for tissue culturing [4]. PEG-DA hydrogels have been explored in the form of 2D substrates, scaffolds or 3D capsules for cells and tissue engineering. The materials was found to be inert and non-toxic to cells [15][16][17][18][19]. To prevent cell adhesion to the channel walls, we here additionally silane coat the channel surfaces using Dichloro-dimethyl-silane (DDS) vapor, which is not toxic for cells. This silane compound in liquid form has been used successfully to coat glass cover slips for suspended cell assays in previous work [20]. To confirm cell viability in the MFS, we repeatedly measured the mechanical properties of cells suspended in the device, such as stiffness, using dual optical traps, for more than 2 h and found no significant changes (~6% variation) in the results in this period. The supplementary figure shows fluorescent labelled 3T3 fibroblast cells suspended in such a micro-device.

Surface passivation
Surface passivation of the inner surfaces of the device is crucial to prevent adhesion of cells. Accumulation of cells can rapidly clog the channels when left untreated. Polymerized NOA-81 produces hydrophilic surfaces [8]. To prevent cell adhesion, we rendered the glass and NOA surfaces hydrophobic by vapor silanization with dimethyldichlorosilane (DDS).
In order to not interfere with the adhesion of the sticker to the glass substrate and to obtain well sealed chambers, silanization has to be done on assembled devices. We used negative pressure to flow silane vapour through assembeled channels for about 10 min. This treatment kept chambers non-adhesive for cells for at least a few weeks.
Hydrophobicity of NOA devices using vapor silanization has been reported to last for 40 d [6]. To perform the silanization, a sticker with open inlet/outlet holes was faced down on top of a small container of 5% DDS-in-Heptane solution (85126, Sigma-Aldrich, Germany) and placed into a desiccator. Evacuating the desiccator vaporizes the DDS solution. The vapor was then sucked through the central channel using a syringe.

Recycling the MFS
Our microfluidic devices could be re-used multiple times, as long as they were properly maintained. Since NOA is rather rigid with very good adhesion to glass after UV curing, we did not experience any swelling or leakage of (E) Liquid NOA-81 is sandwiched between the stamp and a flat PDMS mold to create the sticker. (F) PDMS-NOA-PDMS sandwich is exposed to 365 nm UV. (G) PDMS stamp is peeled off from the sticker and replaced by a pre-cleaned cover slip (24 × 50 mm 2 ). (H) Flat PDMS support is peeled off and sticker is again exposed to UV. 365 nm is the peak sensitivity of NOA-81 and exposure time, depending on radiation intensity, was varied between 2 to 4 min. the device even when the channels were kept wet for a few weeks. Bacterial growth, however, needed to be avoided in all three channels. Due to the presence of the culture medium for cells, bacterial growth is expected and indeed was observed. To clean the chamber before and after use, micro-channels were washed and filled with IPA for 10-20 min. Then, they were washed with distilled water for the same time. After cleaning, the devices were ready to re-use and can be filled with the cell medium. Moreover, we used micro filters (0.2 µm pore size) for all solutions before infusing into channels.

Fluid flow control
In order to drive and control flow through the microfluidic channels, we used positive pressure provided by a compressed-air source, following established procedures [21]. In principle, each of the channels can be fed from an individual reservoir. With the aid of a pressure regulator, channels were filled by pressurizing solutions from sealed centrifuge tubes connected to each of the three channels.
Once inlet of the central (sample) channel was closed off, the pressure gradient along this channel largely vanished. However, due to the presence of the permeable windows between channels, flow can occur through the windows from the pressurized side channels. If flow in the side channels is kept running, flow is driven through the membrane windows due to the pressure gradient (trans-membrane flow). However, when the outlet of the central channel is also closed, pressure equilibrates rapidly and only diffusion through the HMMs will occur [11].

Hydrogel membrane fabrication using confocal microscopy
Confocal scanning hydrogel fabrication is a relatively simple and rapid method for 3D printing of structures in microfluidic chambers. We here used single-plane laser scanning hydrogel fabrication with a simple straight-line pattern design in combination with UV-polymerizable pre-cursor fluids. In principle, one can also print more complicated patterns as well as performing multi-layer and multi-step fabrication by scanning confocal.
We wrote the patterns with a focused 405 nm UV laser (see methods). Up to 80 µm high hydrogel windows could be printed with 2 s of scanning with the UV laser, shown in figures 4(A)-(C). This improves the previously used method, where 500 ms of UV illumination through a photo mask from a mercury lamp with 30 mW cm −2 intensity was used for window fabrication in a 10 µm high device [10].
HMM thickness, pore size and exposure time HMM thickness and its pore size will change with the duration of UV exposure and laser power. Longer UV exposure results in wider, more bulging membranes and smaller pores. The same is true for increased light intensity. The permeability of Figure 3. Photo-polymerization of permeable hydrogel micro-membrane windows using laser scanning confocal microscopy. A 405 nm diode laser was focused by a 10× air objective into the mid plane of the microfluidic sticker, which was pre-filled with photo-curable PEG-DA solution. The confocal scanning mirrors are used to scan a rectangular pattern covering the open windows in the NOA walls of the sticker (ROI). Laser intensity in the focal spot was 400-500 mW cm −2 . Scanning for 1-3 s scanning was sufficient to polymerize an HMM window up to 80 µm high. the HMM for macromolecules (i.e. sugars, drugs or proteins) depends on these two parameters.
We tested the permeability of our laser-fabricated HMMs, in two separate ways. First as shown in figures 4(D)-(I), we flowed fluorescent micro-spheres into the central (sample) channel and ran solutions containing buffer through the outer (side) channels and followed permeability of HMMs by watching the displacement of the particles.
Second, using live cell imaging, we introduced suspended 3T3-fibroblasts cells to the sample channel and flowed a solution of sorbitol (5-10%) in the outer channels while the central channel was closed on both ends. Diffusion of sugar through HMMs changed the osmotic pressure differences across the cell membrane and resulted in changes in cell size ( figure 6).
Hydrogel membrane pores are small enough to effectively decouple fluid flow in the side channels from flow in the central channel, but the pores are large enough to allow for rapid diffusion of solute/solvent molecules. Longer UV exposure time results in wider HMMs with smaller pores and therefore less trans-membrane flow is expected [10]. The practical parameter to vary in order to control the thickness of the windows was the exposure time during the scan. Increasing the exposure time from 1.5 s to 2.6 s increased the maximal thickness of the 40 µm high hydrogel window from ~100 µm to ~150 µm when all other parameters such as volume fraction of photoinitiator (5% v/v), laser intensity (500 mW cm −2 ) and the scanning area (70 × 130 µm 2 ) were kept constant.  Fluorescein solution with high intensity is kept running through side channels. The initial intensity in the central channel was zero. Snapshots were recorded at indicated time after filling the side channels with dye. Diffusion through HMM windows increased the fluorescence intensity in the central channel. HMM windows were 50 × 100 µm 2 , all three channels are 150 µm wide, and walls were 40 µm high.

Diffusion through HMM
Diffusion time τ D over a distance L for a solute or molecule with diffusivity D scales like τ D ∼ L 2 /D. Typical diffusivity for small organic molecules like glucose and sucrose, or small solute molecules like NaCl in water is on the order of D ~ 10 µm 2 s −1 [22]. Therefore, the time for diffusive delivery for such molecules over 100 µm, will be τ D ~10 s. According to [10], the time needed for diffusing into a micro-channel through a HMM becomes τ D ∼ (2ω m + ω s ) 2 /π 2 D, where ω s is the width of sample channel, ω m the thickness of the membrane and D is the diffusion coefficient of molecules passing through. Experimentally, we investigated the diffusion of fluorescein sodium salt through our laser fabricated hydrogel membrane windows. In this experiment, the central channels was closed at both ends, such that flow through the HMMs was suppressed. The results are shown in figure 5. Figure 5(A) shows snapshots after consecutive time intervals of ~5 s. Figure 5 Considering the dimensions of the device (150 µm sample channel and ~50 µm thick HMM windows) and diffusion from both of the side channels to the central channel, the results are consistent with the theoretical estimate of τ D < 10 s.
Step-wise and reversed osmotic pressure changes By rapid exchange of the medium flowing in the side channels, it is possible to switch back and forth between different medium conditions within a few seconds. The speed of this process is limited by the diffusion through the membrane windows. When flow is maintained, solutes will, of course, continue to diffuse in or out of the central channel until complete equilibrium between channels is eventually reached.
We performed experiments under varying osmotic conditions for suspended fibroblasts loaded into the central channel with waiting periods of 30-40 s. Channels had been filled with the cell culture medium, resulting in an isotonic environment for the cells. Then, they were exposed to rapid exchanges of medium from extreme hypotonic (100 mOsm kg −1 ) to hypertonic media (500 mOsm kg −1 and 800 mOsm kg −1 ) and back again. Figure 6 show how radius and cross-section area of a cell altered in response to these medium changes (S5).

Conclusions
We have described the fabrication and application of a threechannel microfluidic device with integrated hydrogel micromembranes (HMMs) for suspended cell studies.
We demonstrated how photo-polymerization of the hydrogel windows can be performed using a standard scanning confocal microscope. We applied the device to study rapid osmotic swelling and shrinking of fibroblast cells while observing them with a high-magnification microscope.
This works opens a new window to efficient studies of single-cell response to mechanical manipulations, as well as biochemical perturbations and osmotic changes.