To study cell behaviour, experiments must be designed at the micrometer scale and enable the cellular microenvironment to be controlled in vitro. This is feasible via microfabrication technology, which allows ordered structures with dimensions in the micro- to nanometer range to be produced. These microstructures can be designed with topographical stimuli that mimics the native microenvironment of the cells. This allows studies on the biophysical signal effects on cellular functions.
With proper microfabrication design and material choice, the rigidity and roughness of the substrate, as well as the protein coating, can be simultaneously controlled. These properties can also be precisely controlled in 3D. This means that along with the standard cell culture conditions, other mechanical parameters affect the way cells interact with each other can be controlled.
Large scale production using microfabrication technology is also possible. This not only lowers the production cost per unit but allows for the deign of high-throughput devices that run experiments simultaneously. Test assays that used to be run in separate test tubes or vials can now be carried out in parallel using microvials or microfluidic channels.
Protocols commonly used to fabricate the microstructures are photolithography, micromolding and nanoimprinting. They are adapted from the established semiconductor fabrication processes and tailored to create the necessary microstructures for biological studies. With the ability to design miniature devices, it is now easier to probe and study cell behaviour at the cellular and molecular level.
The first step of microfabrication involves the production of a mask to serve as a stencil. This will allow for the repeated generation of a desired pattern on a photoresist-coated substrate. This pattern will determine the morphology of the resulting microstructure.
The mask is generally made up of a transparent substrate with an absorber pattern that is opaque to the light source. The absorber pattern is commonly made from metals such as chromium while the substrate is often comprised of an optically flat glass and quartz plate, which are transparent to near and deep ultra-violet (UV) regions, respectively.
A common procedure for mask fabrication (Figure 1) is as follow:
1) A layer of chromium is deposited onto a cleaned quartz substrate. This is often done by sputtering, where chromium atoms are ejected from the bulk target with argon ions (Ar+) and directed to the substrate.
2) A layer of photosensitive polymer called photoresist is deposited on the chromium layer. This can be done by spin coating where the substrate is rotated at high speed to spread the photoresist with centrifugal force.
3) A specific pattern for the mask is designed and this pattern (software mask) controls the lateral movement of the mechanical stage where the substrate is placed on.
4) The pattern is then directly written on the photoresist using either an electron or laser beam. The electron-beam focuses high energy electrons (10 – 100 keV) into a tight spot, where they react with the electron-sensitive resist layer. Two common electron-sensitive polymers are PMMA (positive resist) and SU-8 (negative resist). The system is operated in a vacuum and the smallest point of the beam can be less than 10 nm. On the other hand, the laser beam method can be performed under ambient pressure, making it both faster and cost effective. As laser beams are within the UV-visible range a wide variety of photoresists can be used. This system is applicable when the required line width resolution is 0.6 µm or larger.
5) A developer (solvent) is used to wash off the soluble photoresist which leaves behind a protective pattern of cross-linked resist on the chrome layer. The exposed chrome regions are then removed by wet-etching to expose the transparent regions of the mask.
6) The remaining photoresists protecting the chrome patterns are removed using stripping techniques such as oxygen plasma, acetone and ozonized water. The chrome photomask is ready for use after being inspected for pattern integrity, repaired and cleansed.
Using the mask, protocols such as photolithography, micromolding and nanoimprint lithography may be employed to create the desired microstructures.
Photolithography is a method that uses light to transfer patterns to a surface coated with photoresist. Although various methods of photolithography exist, they vary in efficiency. The direct beam writing method, for example, is slow and inefficient for bulk fabrication of microstructures. A more efficient protocol uses a photomask to transfer patterns.
Schematic of procedure for photolithography: Examples of post-lithography processing are (a) wet etching, (b) plasma etching, (c) electroplating, (d) lift-off and (e) micromolding.
The fabrication steps are as follows and illustrated in Figure 2:
1) A silicon substrate is cleaned with acetone, blow-dried with nitrogen and baked at 100 – 200 degrees C to remove adsorbed water and organic impurities. The surface is then primed with hexamethyl disilazane (HMDS, (H3C)3–Si–NH–Si–(CH3)3) to promote adhesion of the photoresist.
2) A layer of photoresist is then deposited on the primed surface. This can be done by spin coating, where the thickness depends on the spin speed and resist viscosity.
3) The photoresist-coated substrate is then ‘softbaked’ in an oven or hotplate to completely remove the solvent. The temperature must not be too high as this would decompose the photoactive compound of the photoresist and lower its sensitivity to radiation.
4) The photomask is aligned with the substrate in a mask aligner. The photoresist is then selectively exposed to irradiation through the photomask. The irradiation source is commonly in the UV range however when smaller features are required, X-rays may be used.
5) For photoresists that require post-exposure bake, the substrate can be placed in an oven. This heating process enhances diffusion of photogenerated molecules in the resist and smoothes out optical interference effects. For chemically amplified resists (CARs), this process is needed to complete the photopolymerization after catalyst molecules are generated by the UV exposure which lead to a solubility difference between the exposed and unexposed resist.
6) A developer solution is used to develop the photoresist. Here, the soluble parts of the resist are selectively removed. Exposed positive photoresists are soluble to the developer while negative photoresists will be insoluble.
7) If necessary, the substrate is subjected to a post-development hard bake in order to harden the resist and improve adhesion. This is useful when post-lithography processing is required. The temperature for this process must be below the photoresist glass transition temperature (Tg) to avoid flowing of the resist, which results in sloped features.
8) The resist structure is then checked for pattern and adhesion integrity before being used as the final substrate for experiments.
In other applications where further pattern transfer is required, the resist structure is subjected to post-lithography processing such as etching, deposition and micromolding.
Photolithography photoresists have three main components. These are the base resin, a photoactive compound (PAC) and solvent. These components determine the mechanical and thermal properties, the sensitivity to radiation and viscosity of the photoresist, respectively. It is important to tailor resist sensitivity for photons of specific wavelengths because this determines the time required for exposure and development.
Upon exposure, positive and negative photoresists become soluble and insoluble to the developer solution, respectively. Solubility of the positive photoresists increases as the inhibitor decomposes. Two commonly used positive photoresists are the single-component poly(methylmethacrylate) (PMMA), and the two-component phenolic Novolak and diazonapthoquinine (DNQ). PMMA acts as both the base resin and PAC, and becomes soluble through photo-induced chain scission upon deep UV illumination. In the two-component positive photoresist, the phenolic Novolak and DNQ are the base resin and PAC, respectively. Novolak is soluble in alkaline developers but with DNQ also acting as an inhibitor, the unexposed photoresist is rendered non-soluble. Upon UV exposure, DNQ decomposes to form carboxylic acid, thus making the exposed photoresist soluble in alkaline developers.
In contrast, negative photoresists become insoluble upon exposure. This is due to an increase in their molecular weight, which arises from photopolymerization or cross-linking. Examples of negative photoresists are azide-sensitized poly(isoprene) and epoxy-based SU-8 photoresist. The irradiation source for both photoresists maybe UV light (365 to 436 nm), electron beam or X-rays. Developers for most negative photoresists consist of solvents, which can cause the cross-linked resists to swell, affecting the pattern integrity and linewidth control. Depending on the cross-linking density, some negative photoresists are very stable, which poses a challenge in stripping the resists at the end of the fabrication process. These photoresists, however, are good to be used as permanent structural materials.
The most direct exposure method is known as contact or proximity lithography (Figure 3a). Here the mask and photoresist-coated substrate are in placed close contact (separated by 3 to 50 µm, respectively) and a 1x reproduction of the mask pattern is made on the substrate. The resolution is determined by the mask dimensions and diffraction at the mask edges. Mask damage may occur when the mask touches the substrate and therefore this method is cannot be used for mass production. An alternative method is known as projection optical lithography (Figure 3b). Here a smaller image of the mask is cast on the photoresist-coated substrate. This method produces a 4x reduction where, for example, a 200 nm feature, will be reduced to 50 nm on the substrate. With no physical contacts with the photoresist surface, the mask is less likely to be damaged and is thus reusable. Mass production is also possible with this system via a step-and-repeat exposure approach, where many small regions on a big substrate are exposed, one after another, using the same mask.
The linewidth and height of the structures created using photolithography can vary from 10s of nm to up to 200 µm. They are limited by various factors including the diffraction of the light source, exposure method, photoresist thickness and sensitivity to light, development, and subsequent pattern transfer process (e.g. etching, sputtering and electrode position). Requirements on feature resolution and material selection thus set the constraints on both the fabrication protocol and limitation achievable. An irradiation source with a shorter wavelength, for example UV, X-ray and electron beams, gives higher resolution than that created via visible light. Resist sensitivity can be tailored for the specific wavelength of irradiation source, which is important in determining the exposure and development time.
At the end of the fabrication process, or post-lithography processes, the photoresists which are no longer needed are stripped using oxygen plasma, ozone discharge, acetone, ozonized water, sulphuric acid, organic amines or hydrogen peroxide.
Replica micromolding allows for the production of polymeric microstructures by casting and curing a polymer against a mold. High throughput replication of feature sizes ranging from several centimeters to 30 nm is feasible using this technique. The master mold can be made using photolithographic techniques such as direct beam writing or irradiation through a photomask. To prevent damage from frequent usage of the master mold, an elastomeric template (e.g. polydimethylsiloxane, PDMS) is cast from the master mold. A common protocol for replica micromolding is as follow
1) A silicon substrate is cleaned and pre-treated to promote adhesion of the photoresist.
2) A layer of photoresist is then deposited on the primed surface.
3) Patterns are transferred onto the photoresist either by direct beam writing or exposure through a photomask.
4) The photoresist is developed using a developer solution, where the soluble parts are selectively removed. Shown in the figure is a negative photoresist, which become insoluble upon exposure.
5) Using the photoresist as a protective layer, the silicon is etched.
6) The protective photoresist is then stripped, giving a master mold.
7) To cast a master template, the master mold is firstly pre-treated to assist release of template.
8) A PDMS prepolymer mixture that contains curing agent is poured over the master mold and cured at an elevated temperature (70 – 80 degrees C).
9) The PDMS is slowly peeled off from the master mold. This will be the master template.
10) The template surface is then silanized and PDMS is poured over the template.
11) After curing the PDMS, the final PDMS cast is released from the template.
Nanoimprint lithography (NIL) is another technique used to produce polymeric micro- or nanostructures. While micromolding casts and cures the polymer on a mold, NIL works by pressing a hard mold onto a layer of polymer in its molten or solution state before curing. Therefore, both pressure and heat are typically necessary for the pressing and maintaining the polymer in its moldable state, respectively. The 3D master mold can be prepared via the photolithography and micromolding. NIL can provide linewidth resolutions ranging from several hundred microns down to about 5 nm. Depending on the type of polymer used, NIL can proceed via a thermal or UV curing protocol.
Thermal vs UV nanoimprint lithography: Schematic of procedure for (a) thermal NIL and (b) UV NIL.
1) A mold with inverse structures is cleaned before treated in oxygen plasma. The surface is then further treated with an anti-stiction layer to facilitate detachment of the mold. Depending on the pattern complexity, the mold can be made of silicon, quartz or nickel, and may also be available commercially.
2) A substrate is coated with a layer of polymer to be imprinted.
3) The mold is aligned to be directly on top of the polymer-coated substrate.
4) Both the mold and polymer-coated substrate are heated to a temperature above Tg of the polymer, where the polymer is fluid and amenable to the mold.
5) Imprinting is performed by pressing the mold onto the polymer film at the elevated temperature.
6) After the mold cavities are filled with a molten polymer, the temperature is lowered to below Tg of the polymer and the mold is detached.
7) The residual layer is etched away to leave behind the micro- or nanostructures of interest.
UV NIL works only for UV-curable polymers or resists. A typical procedure is as follow (Figure 4b):
1) The surface of the mold is cleaned and treated with an anti-stiction layer to facilitate mold detachment. Here the mold must be transparent to UV, for example quartz.
2) A prepolymer mixture of UV-curable polymer, which is less viscous, is dispensed onto a substrate.
3) The mold is aligned directly on top of the prepolymer mixture.
4) The mold and substrate are pressed together to fill the cavities of the mold with the prepolymer mixture.
5) The polymer is exposed to UV to cure and solidify.
6) After curing, the mold is detached from the substrate.
7) The residual layer is etched away to leave behind the micro- or nanostructures of interest.
Each process has its own advantages. UV-NIL can be performed at room temperature and at low pressure since the UV-curable polymer solution is already fluid and less viscous. Thermal NIL has less restrictions on the mold, where it can be non-transparent, and thus has lower production costs.
The surface cells grow on is called the cell substrate. To be applicable as a cell substrate, the microfabricated substrate must be bio-compatible. One way to achieve this is by using bio-compatible materials such as PDMS, which is also attractive for mechanobiological studies as it is elastic, which is useful for measuring forces exerted by cells on micropillars . It is also optically transparent to light of wavelengths down to about 300 nm. This allows for its use where cells must be observed using microscopes .
When used in biological studies, the substrate often needs to be coated with ECM proteins such as fibronectin and laminin  which helps cells to stick to the microstructure surfaces. The coating of the ECM proteins is generally done by immersing the microfabricated substrate in a solution that containins the protein for several hours. The substrate is then rinsed with water and dried . An alternative method, microcontact printing (µCP), is used when a layer of patterned ECM protein must be printed onto a flat substrate . Here, the stamps must be fabricated using similar methods to the those used in the production of the substrate.
In some experiments, the microfabricated surfaces must reflect the properties of the cell membrane. To achieve this the substrate can be functionalized with a layer of lipid membrane, which is more fluid and mobile than the ECM coatings. The effectiveness of such techniques was demonstrated in two recent studies exploring integrin-mediated adhesion. Each study employed substrates coated with lipid membrane separated by chromium lines . Using these substrates, processes that are carried out in the early stages of cell adhesion onto rigid substrates were revealed. Additionally, the formation of podosome-like adhesion was described following prolonged adhesion on fluid substrates. In these cases the chromium lines were fabricated on glass substrates using the NIL technique and thermal deposition. These lines served as physical barriers to the lateral movement of the lipid membrane layer.
To study the influence of substrate rigidity on cell function, cells can be grown and observed on microfabricated pillars of varying stiffness. Pillar substrates have the advantage over the use of hydrogels, in that substrate stiffness can be altered independently of the bulk material properties such as porosity, ligand density and wettability. Rigidity of the pillars is controlled by varying their dimensions. Shorter pillars are stiffer and taller pillars are softer . The tips of the micropillars are often functionalized with ECM proteins to promote cell adhesion.
Forces exerted by the cell can also be determined by using an array of elastomeric micropillars (Figure 5)  . In these experiments the cells adhere onto the ECM-coated tips and deflect the pillars as they move. This deflection can be measured as local force (F), determined by the equation shown in the figure where r, L, E and delta x are the radius, the length, the Young’s modulus, and the deflection of the pillar, respectively .
This approach was used, for example, to investigate collective cell migration along varying channel widths. Here, an alternating pulling and/or pushing mechanism was identified during directed, and ordered, cell migration. This was dependent on cell-cell interactions and extrinsic constraints . Similarly, collective cell migration during wound healing was investigated using micropillar substrates. Here maps of traction forces were obtained based on the deflections of micropillars. This study noted that wound closure initially occurs via a collective cell crawling however this is followed by processes where cells at the leading edge assemble supracellular actomyosin rings that compress the tissue underlying the wound. These contractile forces are transmitted to the substrate through focal adhesions, inducing displacement of the underlying substrate towards the wound area and speeding up wound healing.
Microchannels can be specially designed and fabricated as microfluidic devices for studying mechanical stimuli and blood flow in the body , as well as for developing medical diagnostic tools . PDMS is commonly used in the fabrication of microfluidics due to its bio-compatibility and optical transparency down to 300 nm, as well as its ability to conform to the microfabricated mold .
In one example study, microfluidics was used to separate malaria-infected red blood cells (iRBCs) from healthy RBCs in blood samples for further diagnosis . Due to the flow velocity gradient in blood vessels, the deformable RBCs will flow in the axial centre of the vessel, displacing the less deformable white blood cells (WBCs) to the wall in a phenomenon called margination. The iRBCs, which are also stiffer than RBCs, are found to behave in a manner similar to WBCs and undergo margination. Based on this difference in deformability, iRBCs can be successfully separated using a specially designed microfluidic device with a long microchannel and a 3-outlet system at the end (Figure 6). As shown in the figure, the iRBCs are randomly distributed at the microchannel inlet and marginalized to the sidewalls as the flow reaches the outlet. They are then filtered out from the blood and collected for diagnosis.
Microfluidics has also been used to isolate circulating tumor cells (CTCs) from blood for further analysis using basic physics . Here, blood is pumped through a spiral microfabricated channel (Figure 7). The inherent centrifugal forces cause the smaller cells, which include red and most white blood cells, to flow along the outer wall of the channel. Meanwhile, the larger cells, i.e. the CTCs and larger white blood cells, flow along the inner wall. At the end of the spiral the channel splits in two and the smaller cells flowing along the outer wall will go into a waste container, while the larger cells (CTCs) will flow to a collection chamber. These collected CTCs can be analysed further cancer diagnosis.
Cell functions, such as migration, development and tissue repair, are dependent on both biochemical and biophysical signals from their microenvironment.
To overcome the tedious need to analyse cell behaviour on one topography at a time, a multi-architecture chip (MARC) was developed for highly efficient high-throughput studies of cell-topography interaction  . This chip incorporates many architectures of different aspect ratios and hierarchical structures on a single chip, allowing simultaneous analysis of cell behaviour on varying topographies. The types of topographies can be customized according to the need of experiment.
This 2.2 cm x 2.2 cm chip can fit within standard cell culture plates or dishes and can hold up to 121 different nanopatterned surfaces of 2 x 2 mm each. Each area accommodates enough cells for statistical analysis and the whole chip will fit in standard microscopy specimen holder to allow microscopic inspection of cellular responses.
The MARC has already proved useful in finding the optimal combination of topographical and biochemical cues for stem cell differentiation to neurons  and for proliferation of human corneal endothelium cells . With fully customized topographies, the MARC provides a platform to study the ability to control the fate of stem cells, and realize the therapeutic potential of these cells.