Microfluidics for cell biology #1 - PDMS

?? Ask for the full PDF in comments ??

Introduction

Microfluidics is a well-understood domain of physics and can now be used to develop tools for cell biology. By simply miniaturizing macroscopic systems and taking advantage of the possibility of massive parallel processing, some microfluidic chips enable high-throughput biological experiments. Specific effects of laminar flow at the micron-scale also allows spatial control of liquid composition at subcellular resolution, fast media changes and temperature changes, and single cell handling and analysis. Microfluidic technology enables studies of cell behavior from single- to multicellular organism levels with precise application of experimental conditions unreachable using macroscopic tools.

In the 80’s and 90’s, microfluidic devices were mainly fabricated on silicon substrates. These technologies required clean room facilities and strong know-how. In the late 90’s, the introduction of soft lithography using molding of polymers allowed the fabrication of cheap microfluidic devices with additional advantages due to the physical characteristics of those polymers.

The current most popular technology for the fabrication of microfluidic devices for cell biological application is based on the soft lithography of poly-di-methyl siloxane (PDMS). PDMS is an elastomer which, through simple molding procedures, can be turned into microfluidic devices. Its wide use as a material of choice is due to its mechanical properties, which make it amenable to the integration of fluidic valves, essential elements for major microfluidic applications. Further, PDMS is transparent, biocompatible, and permeable to gas, which explains the strong interest of the scientific community in using this material to fabricate microfluidic devices for cell biological studies.

The development of soft lithography gave a simple technology to fabricate devices that integrate channels at the scale of a cell. In most cases, the interest of biologists for microfluidics did not stem from an interest in new physical phenomena at the microscale, but instead from a practical experimental point of view and a favorable scaling of physical forces. At the microscale the laws of physics remain the same as in macroscopic systems, but the scale factor can give predominance to different forces. For example, in the case of fluid flow, the reduction in size reduces the influence of the inertial forces compared to frictional forces, leading to the formation of laminar flow in microfluidic channels. Furthermore, the reduction of size has a direct influence on the characteristic times of the system, such as the time required for the diffusion of a molecule, which decreases as the square of the characteristic length. Microfluidic devices give several advantages for cell biology applications. Some advantages come from the fast response of microsystems. The fast diffusive heat and mass transfer at the microscale (microscale characteristic times are approximately 10-3s-1s, compared to macro-scale time of 102s-104s) allows for fast media and environmental changes and fast temperature control. Laminar flow properties are also useful since they enable the formation of static and dynamic gradients at subcellular resolution.

Microfluidics also have a number of other positive practical aspects like low reagent consumption (nL), the opportunity to manipulate large number of cells simultaneously and independently, automatic generation of a large number of different individual conditions, and easy integration of numerous analytical standard operation and large-scale integration. From a technological point of view, soft lithography enables the integration of subcellular scale physical and chemical patterns to study cell behavior under a large spectrum of parameters. In addition, electrode integration inside the microfluidic devices can generate large localized electric fields using small voltages. Finally, the versatility of these devices partly enables the simulation of in vivo cellular microenvironments (vascularization, 3D, nutrient stress, etc…).

Microfluidics have some drawbacks. For example, laminar flows only produce relatively slow diffusive mixing, which can be a major limitation for some applications requiring fast homogenization of flow. This can be corrected using different types of integrated mixers [1], such as advective mixing in a microchannel [2]. In addition, the small reagent consumption theoretically reachable in microfluidic devices is generally not reached due to a current lack of methods for fluid handling. Changes in scaling can give further difficulties in the adaptation of biological protocols to fit experiments in microsystems (i.e., media and cell concentrations). Furthermore, PDMS has an affinity for small hydrophobic molecules and thus could lead to biomolecule absorption/adsorption from the medium, thus biasing the experimental conditions. The permeability of PDMS to water vapor can also lead to media drying and thus change its osmolarity. These differences require careful comparison between the data obtained in macroscopic experiments and data obtained in microsystems.

In this chapter, we discuss the use of microfluidics to fabricate research tools in cell biology with a particular focus on PDMS soft lithography.

?? Ask for the full PDF in comments ??

Soft-lithography and microfluidics 

PDMS casting and microcontact printing

We describe here the two major techniques, often referred to as soft lithography of PDMS, for fabricating microfluidic devices dedicated to cell biological research and for printing molecular micropatterns on cell culture substrates: PDMS casting and microcontact printing.

Figure 1 illustrates a typical procedure for making a microfluidic device. PDMS microfluidic devices are generally fabricated using molding methods [3, 4]. A silicon master mold, containing a photoresist pattern representing the channel design, is fabricated using photolithography. Once the master mold is fabricated, if carefully used and properly treated with an adhesive, it can be used hundreds of times to mold a PDMS replica of the channels.

Photolithography processes for mold fabrication, which will not be detailed here, require a spincoater and a dedicated UV lamp. For most of the biological applications described in this review, including patterned features as small as 1-micron, the photolithography equipment can be installed in a classical biological fume hood instead of a clean room, thus significantly reducing costs.

To fabricate a monolayer PDMS microfluidic device, liquid PDMS and a curing agent are mixed and poured onto the master mold and cured at 60°C for 2 hours. Then the mold and PDMS replica are disassembled, cut to the desired dimension before drilling of injection holes using generally a syringe needle. The PDMS microfluidic device and glass cover slip are plasma treated for 30s, and put into contact to produce a covalently bonded and sealed full microfluidic chip.

Since one PDMS layer can be bonded onto another using a similar plasma treatment, this technology enables the fabrication of multiplayer microfluidic devices. The possibility to fabricate multilayer devices, coupled with the low Young’s modulus of PDMS, enables the integration of microwaves [5]. The integration of elements like valves gives active control of the liquid inside the chip, leading to fully automated microsystems.

Applications requiring the integration of electrodes or resistors require the deposition of metal and dielectric layers inside the device. Integration of electrodes on a substrate is a common operation in microtechnology but requires clean room facilities. Furthermore, due to the low surface energy and high softness of PDMS, metal deposition is difficult to achieve on PDMS surfaces. Nevertheless, since glass substrate is compatible with most of the clean room techniques developed for silicon-based microelectronic industry, it is possible to coat metallic and dielectric layers directly on the glass substrate prior to plasma-bonding onto the PDMS device.

Figure 1: Fabrication procedure for a single layer microfluidic device. (See text for details.)

Microenvironmental control

The microenvironment of the cell is defined by chemical and mechanical parameters. Chemical environment is composed of soluble molecules around the cell which are related to the cell culture medium composition, and mechanical environment is composed of the extracellular matrix (ECM) which is related to the culture substrate composition. In conventional cell cultures, these environmental parameters are easily controlled for a population of cells, but cannot be addressed locally to individual cells. One of the major interests of microfluidics is environmental control at the scale of the cell. In addition to the ability to locally address parameters of the cell microenvironment, microfluidics also offers the ability to change these parameters dynamically and automatically due to the speed of the physical processes at the microscale and the different automation possibilities. Although an ideal comprehensive microenvironmental control device does not exist, individual environmental parameters can readily be controlled using soft?lithography and in some cases can be combined. We will review in the next part the different existing methods to control the chemical and the mechanical microenvironment of the cell.

Chemical microenvironment

One of the first applications of microfluidics in cell biology was to control the cell medium. There is a strong interest in producing chemical gradients to mimic natural stimuli which occur in biological processes such as cell migration, differentiation, or development. The study of cellular response to chemical gradients requires fine spatial control of local concentrations because cells can respond to concentration gradients localized in a region as small as 2% of their diameter [6]. The poor spatiotemporal resolution of macroscopic gradient generators (MGGs) led to an interest in fabricating microscale gradient generators (μGGs). Microfluidic devices can create multiple biochemical gradients with controlled spatiotemporal distribution and subcellular resolution. These microfluidic devices offer fast response times to study fast responsive systems such as immune cells. μGGs have been successfully used to study neural stem cells growth and differentiation [7], neutrophil chemotaxis [8-10] or migration [11], bacteria chemotaxis [12], endothelial cells migration [13], cancer cells chemotaxis [14], cellular response to viruses [15], and yeast gene expression under gradients of pheromones [16]. For more information, we refer the reader to the excellent review of Keenan and Folch [17].

Macroscopic methods to create concentration gradients are generally imprecise and unstable

MGGs have traditionally been developed using hydrogel made from fibrin, collagen or agars. Gradients are deposited on the hydrogel using an array of droplets containing biomolecules which then diffuse in the gel to form a gradient [18, 19]. A second method uses micropipettes to inject biomolecules into the gel at controlled rate to generate a gradient [19]. In general, hydrogel enables easy gradient production, but has poor reproducibility and spatiotemporal control over the gradient. In addition, the opaque optical properties of the gel can be a limitation for some applications. Other MGGs have been developed using chambers separated by membranes [20-22]. These methods cannot generate complex and stable gradients. An exception is the Dunn chamber [22] using modified micropipette techniques, which were able to generate stable gradients for several hours. Although these MGGs have helped address numerous questions, they are not useful for studies requiring gradients with precise spatiotemporal control and reproducibility.

Microfluidics allow for precise spatiotemporal concentration gradients

Microfluidics enables the creation of a large spectrum of gradients: time invariant gradients, subcellular resolution gradients, continuous or discrete gradients, fast response dynamic gradients. Most of the μGGs described here use simple technologies (i.e., single layer PDMS), and have been applied to generate concentration gradient of diffusible molecules to study bacterial chemotaxis [12], and cell migration in response to chemokines [13], and to generate surface gradients of adsorbed ECM molecules to study the dependence of axon growth of neurons on the surrounding ECM composition [23].

Figure 2 gives examples of gradient generation. One simple way to generate gradients in microfluidic systems is to use the properties of laminar flows. Laminar flow-based μGGs use diffusive mixing between two or more parallel laminar streams of different composition to generate molecular gradients. The shape of the gradient based on laminar flows depends on the flow rate and the time the streams are in contact. Gradients generated in these types of devices will maintain their shape at constant flow rate. The simplest μGGs of this type is the T?sensor [24], which is composed of two microchannels. These gradient generators have a small time constant and are theoretically able to establish or modify a gradient in 10-2s-10s. T-sensor devices are easy to fabricate and to describe mathematically. The constant perfusion permanently exchanges the medium and prevents the accumulation of cell waste products, thus enabling long cell culturing time. In contrast, T-sensor devices are reagent-consuming. They remove the autocrine/paracrine or other secreted signals of cultured cells, and submit cells to shear stress due to flow. Further, the useful region is limited to a short portion of the channel which generates only sigmoidal-shaped gradients in the direction perpendicular to the flow. To overcome this limitation, additional inlets can be added to generate more complex flow profiles.

Yet, the addition of needed controls can rapidly make these experiments tedious. T-sensor devices have been used in studies of bacterial chemotaxis [12], or endothelial cell migration [13].

An upgraded version of the T-sensor called premixer μGG [25] splits and recombines inlet fluids before merging them in the culture channel, and thus can generate more complex gradients (e.g. sawtooth and hill). By adjusting the inlet flows, premixer μGGs can generate smooth or step gradients. Dertinger et al. introduced the multiple premixer arrays which can generate overlapping gradients [26]. This type of device has been used in studies of neutrophil chemotaxis [8, 9], neural stem cell differentiation [7], and breast cancer cell chemotaxis [14]. Except for the wider range of gradient shapes, the premixer device shares the same issues as the T-sensor.

Another upgraded version of the T-sensor called universal μGG [27] includes a series of walls to split the streams. This configuration can generate many profiles of concentrations, and can reduce dead volumes compared to the T-shaped μGG, but also shares the same issues as the T-sensor and is more challenging to describe mathematically. Cookset et al. developed a μGG composed of an array of 16 multiplexed inlets, which gives 64 combinations of chamber feeding [28]. This device integrates a mixer which can be turned on and off with a bypass valve, and can produce gradients or homogenized mixtures. This device enables simultaneous formation of complex gradients of different biomolecules, with sub-second temporal resolution.

Major drawbacks of all laminar flow μGGs are that they require precise control of the flow rate. The shear stress produced by the flow can change the migratory behavior of the cells [29], and produce undesired mechanical stress on the cells, and flush away important factors secreted by cells [17]. Although possible, these μGGs are challenging for studies of nonadherent cells such as yeast or bacteria because of the movements generated by the flow [30].

The second type of μGGs is not based on the properties of laminar flow. The flow resistive μGG uses flow resistive elements to eliminate convection around the cells. This kind of device allows passive diffusion of biomolecules through a flow barrier to generate gradients. The flow barrier can be a hydrogel [31-33], nanopore membrane [10], or microchannel [32]. Hydrogel completely eliminates convection, whereas microchannels (which are easier to integrate) only minimize convection. In both cases, these devices can generate steady-state gradients, eliminate shear stress generated by flow, and preserve the autocrine/paracrine signals secreted by cells. In addition, they use less reagents than laminar flow μGGs, are possible for experiments with nonadherent cells like yeast [16], and some are able to generate gradients in hydrogel for 3D cell culture. Major drawbacks of these μGGs are their inability to create complex profiles, their large time constant compared to laminar flow μGGs, and, for the hydrogel-based μGGs, the greater difficulty to fabricate them than single PDMS layer-based laminar μGGs.

A miniaturization of the micropipette technique used in MGGs, called microfluidic multi-injector (MMI), can generate overlapping gradients using integrated valves for flow injection. This method gives better reproducibility and quantification of gradients than its macroscopic analogue, but is quite slow compared to other μGGs since it requires ~10 minutes to achieve a steady-state gradient.

In general, μGGs are able to generate gradients with much better spatiotemporal resolution than MGGs. The choice of the types of μGGs depends on the constrains of the biological experiment. Flow μGGs are the best candidate for experiments requiring fast response and/or complex gradient shapes. On the other hand, hydrogel- or microchannel-based flow resistive μGGs are best suited for experiments where shear stress, cell drift, and/or the unwanted flushing of secreted cellular factors are of concern.

Figure 2: Microfluidic gradient generator.

(A) Schematic representation of a T-sensor with two inputs. One can see the diffusion between the two laminar streams along the device [24].(B) μGG composed of an array of 16 multiplexed inlets which allow 64 combinations of gradient generation. This figure shows gradient shape modification in the central chamber depending on the valve state [28].

(C) A microfluidic cell culture array containing 100 cell culture chambers with integrated gradient generators [162].

(D) Zoom on the cell culture array with gradient generation demonstrated using red, blue and yellow dyes [162].

Laminar flows can address subcellular resolution and fast switching

Microfluidics, due to the small time constant and diffusion mixing, can be used to dynamically focus a drug stream on a given part of a cell. This ability allows dynamic observation of cell behavior immediately after drug treatment, or to study how local chemical stimuli propagate in the cell [34, 35].

This method is limited by molecular diffusion since small molecules can diffuse ~100 μm in less than 1 minute, and thus can diffuse throughout the whole cell during PARTCELL treatment. This effect smooths the molecular distribution and creates molecular gradient inside the cell instead of a straight molecular concentration step. Wheeler et al. showed that, by controlling the inlet flow rate, the displacement of the laminar mixing region can be achieved, and thus media switching can occur in less than 100 ms around a cell [37]; which is 10 times faster than standard perfusion chambers. Hersen et al. have used these kinds of chemical oscillating signals to extract kinetic information on the HOG MAP kinase pathway [38].

A second method to expose a part of a cell, called “hydrodynamic focusing”, uses flows from two sides to squeeze a central flow to widths as small as 50 nm [39, 40]. The position of the hydrodynamic focusing can be changed by simply adjusting the inlet flows; and stable focusing position and width are maintained by constant flow rate [41]. This method has low reagent consumption and good spatiotemporal resolution, but requires precise fluid handling systems and is also limited by molecular diffusion on the cell cytoplasm.

Starting from T-sensors, more complex devices have been designed to automate condition variations. For example, King et al. designed a method called “flow encoded switching” [42], to simultaneously deliver different temporal profiles of chemical stimuli, such as pulsed train of different widths or frequencies. Sabounchi et al. subjected Hela cells to biochemical reagents in a pulsatile manner, using external solenoid valves [43]. They were able to apply and remove a reagent from the cells in 100 ms.

In addition, the high permeability of PDMS to gas enables control of gas composition in cell cultures. By flowing gas of controlled composition in microchannels adjacent to the cell culture, it is possible to locally control the gas composition of the medium [44, 45]. This kind of device enables fast gas composition switching or gas gradient generation.

The devices presented above are adapted for fast changes in cellular microenvironment. However, in most cases, speed may be of less importance than control flexibility, cell seeding practicality, low shear stress, long-term culture possibilities, and large-scale integration. Substrate patterning / Substrate modification is required for adherent cells. In most of the devices described above, cells adhere on glass surfaces via incubation of the microchannels with ionic polypeptides like poly-lysine or proteins like fibronectin. In cases where cells have to adhere to PDMS surfaces, plasma curing promotes adhesion, but it is not stable over time [46]. To overcome this limitation, it is generally necessary to covalently bind the adherent molecules to the PDMS surface. We will describe techniques to control the chemical nature of the cell substrate.

In standard cultures of adherent cells, cells are randomly seeded on the surface of the culture substrate. This random organization, which is not representative of living tissues, does not interfere with the results of most biological experiments. However, recent advances in cell biology highlighted the need for controlling patterns of cell adhesion to ask questions concerning tissue morphogenesis and cell communication. This leads to the development of three major modes of spatial controls of cell adhesive substrates:

1. Mimicking signaling tracks which are naturally present in vivo on the ECM. For example, gradients of surface properties are used to study the chemotaxis behaviors of motile cells [47] or neuronal path findings [23].
2. Forcing cells to follow a given adhesion pattern in order to study interplay between the geometrical constrains and cell behavior such as changes in cell polarity [48, 49]. 3) Constraining the cell in a given location and shape to facilitate its analysis to produce “mean” cell maps of the position of cellular organelles [49, 50].

We will next discuss the existing methods to create patterned substrates of various chemical compositions to constrain cells, or to vary other physical properties of the substrate such as patterning of the substrate by molding [51] to study durotaxis behaviors [52].

Figure 3. Spatiotemporal drug control.

(A) Picture from a 2 minute movie showing successive perfusion with Trypan blue dye on a live cell, and subsequent methanol and Trypan after cell death.

(B) Picture showing the ability to change the stream in contact with a cell by changing inlet flow rate. This type of medium switching can be done in 130 ms [37].

(C) Schematic of PARTCELL principle [36]. Using laminar flow properties, partial treatment of a cell can be achieved. Picture shows treatment of a portion of a single cell with Latrunculin A and blue dye.

(D) Picture showing hydrodynamic focusing. Flows B (green) arriving from both sides of flow A (red) focused and maintained flow in a fine stream configuration [40].

Photolithography is efficient but not versatile

The oldest method to pattern biomolecules is photolithography using UV light to expose a mask containing the desired patterns onto photosensitive resists. It is then possible to transfer onto the resist patterns of biomolecules of interest by etching or liftoff. This method has been successfully applied to produce adhesion patterns for cell culture [53]. Nevertheless, it requires clean room facilities which are generally not easily accessible for biologists, and the chemicals used in the process may be harmful for some biomolecules. To avoid this last drawback, a water-soluble sacrificial layer like agar of PVA can be inserted for protection [54].

An alternative to photoresists is the use of UV to directly degrade molecules [46]. By using deep UV to degrade a repellant molecule like PLL-PEG, it is then possible to reactivate the surface for adsorption of the biomolecules of interest afterwards [55]. This method, which works on glass and PDMS substrates, is particularly robust and easy to process. Interestingly, variants of this method can also produce gradients of the surface concentration of certain molecules [56].

Photolithographic methods are probably the best methods in terms of pattern quality and patterns with high resolution (down to 1 μm). However, these methods have to be modified for each new substrate, and are thus not versatile.

Microcontact printing is the simplest method

Figure 4 gives examples of microcontact printing methods and applications. Microcontact printing (μCP) is a method which enables printing patterns of molecules on a substrate using a microstructured stamp (generally fabricated in PDMS by molding). The stamp is coated with the molecule of interest by dipping it into a solution which can contain a multitude of elements such as thiols, proteins, silanes or nanoparticles. Once the molecule of interest is adsorbed on the stamp, the stamp is temporarily put in contact with the substrate to allow transfer. After printing on a surface, the nonprinted adjacent surface can be made passive with another molecule to prevent cell spreading beyond the printed areas. The PDMS stamp can be used ~100 times over a period of several months without noticeable degradation of the quality of the printing [57]. When using classical PDMS, this technique can achieve resolutions below 500 nm [58]. μCP has been used extensively for substrate patterning of biomolecules for experiments such as axon guidance [59, 60], or cell culture on defined geometry [46]. μCP was initially used to print self-assembled monolayers of alkanthiolate on gold surfaces to perform hydrophobic patterning. This very efficient technique was soon extended to patterning of peptides, proteins, and a wide range of biomolecules on different substrates. Nevertheless, all molecules cannot be stamped using μCP since the “ink” has to be dried on the PDMS stamp to be patterned. To avoid drying, agar stamps can be used [61]; such methods have achieved resolutions down to 50 μm.

When printing of several different molecules is needed, it is possible to perform sequential functionalization of the substrate by using different stamps with different molecules. This method is easy to perform but requires an aligner to control the position of the successive printing. An alternative method consists in loading the stamp with different molecules simultaneously [58], but this method does not generally give precise spatial protein concentration. Crozatier et al. developed a method based on microaspiration to load multiple samples onto a stamp prior to transfer on substrate [62]. In contrast to single μCP methods, multi-molecules μCP requires strong technical know-how.

μCP also enables direct patterning of gradients onto a substrate. Stamps composed of arrays of high-resolution patterns with controlled spacing and density can generate gradients of biomolecules [59]. An original technique involving μCP directly patterned bacteria at cellular resolution on a substrate using structured PDMS stamp with bacteria as the ink [63].

In general, μCP enables high-resolution patterning with a large range of patterns. μCP can pattern planar or non-planar substrates and have been used on substrates such as glass, silicon, and polystyrene. μCP patterning is limited to molecules that are not altered when adsorbed on a substrate, and typically is limited to patterns containing only one or two types of molecules. During an experiment, the adsorbed biomolecules may also degrade or be replaced by other molecules in the medium.

Figure 4: Substrate patterning using microcontact printing (μCP).

(A) Schematic of μCP procedure [58].

(B) Bicolor μCP using successive stamping of molecules [58].

(C) Multicolor μCP using stamp pre-inked with molecular gradient [62].

(D) Influence of adhesive micropattern on cell cytoskeleton. This figure shows vinculin and actin repartition for different fibronectin patterns [50].

Stencil patterning helps to pattern fragile components

An alternative method called “stencil patterning” enables patterning of any component without altering it. This method requires covering the substrate with a membrane (typically PDMS) containing microholes (the stencil). Deposition of molecules is applied to all surfaces, but only the microholes are exposed. Subsequent removal of the membrane stencil produces treated surfaces at the  microholes’ positions. This method is often used to locally apply a harmful treatment on a delicate surface [64], or to directly pattern cells onto a homogeneous substrate [65]. Stencil patterning is a convenient method but manipulation and fabrication of micromembranes remains tricky.

Patterning using liquid flow in channel can achieve complex functionalization

Figure 5 gives examples of substrate patterning by flow and active elements. Liquid flow can also achieve patterned substrates. Patterns can be formed by restricting the flow area using microchannels directly on the substrate, or on an intermediary stencil. This method allows for simultaneous deposition of a large number of ligands by circulating different streams in parallel or in a gradient [66]. Liquid flow enables successive treatments on the patterned area, but often the shape and resolution of this method are limited.

To pattern the substrate by liquid flow, one method consists in sticking a microfluidic channel on the substrate and flowing it with a solution containing biomolecules. The microfluidic device can be permanently stuck if the biological experiment is performed in the same device [66], or temporarily stuck by simply putting down the device on the substrate [67], or by using vacuum aspiration [68]. Using this method, Delamarche et al. showed that it was possible to pattern biomolecules on various substrates like glass, gold or polystyrene with submicron resolution [67], and Folch et al. created protein templates of collagen and fibronectin allowing cells to adhere on selective surfaces [69].

By flowing different solutions of biomolecules on different channels in parallel, it is possible to create patterns of multiple compositions or to pattern molecular gradients on a substrate using a μGG [23, 66]. A simpler but less versatile method is to use the depletion effect of the solution while it flows along the channel to generate a gradient [70].

Using PDMS multilayer devices, it is also possible to indirectly pattern cells at desired locations on a substrate [71]. This device uses integrated structured valves to first restrict flows of blocking agents onto surfaces uncovered by the structured valves (Fig. 5). After releasing of the valves, flows containing ECM will adsorb ECM to the unblocked surfaces. This device produces adhesive islands on a microchannel for cell culture.

These described flow patterning techniques generally only pattern continuous shapes if no stencils are used. To overcome this limitation, Chiu et al. proposed a 3D microfluidic device to pattern discontinuous patterns [72]. This device has been used for classical chemical patterning and also for direct patterning of different mammalian cells on the same substrate. The major drawback of this device is the complexity of the fabrication process compared to the other flow patterning techniques.

Figure 5: Substrate patterning using flow or active elements.

(A) Structured valve based microfluidic device for substrate patterning. A1-A2: Schematics showing cross-sectional view of the device and deformation of structured membrane when applying pressure on the top channel. Flow of passive agent recovers only the area unprotected by membrane structure, allowing future adhesion of ECM protein. A3: Picture shows a scanning electron micrograph of the membrane structure. A4: Fluorescence picture of actin stained endothelial cell following ECM matrix protein shape [71].

(B) Direct cell patterning using reversible device with two interconnected channel layers. Up: Schematic representation of the method used to pattern different cell types on the same substrate. Down: Fluorescence picture of two cell types deposited on a tissue culture dish in a concentric pattern using this device [72].

(C) Dynamic cell patterning. Picture showing BCE cells attached to a surface patterned with specific thiols. Application of a cathode voltage pulse allowed release of the cells from the micropattern (time in minutes) [79].

Inkjet printing and microdroplet dispenser

To micropattern a substrate with chemicals and without contacting it, it is possible to use an inkjet printer or a microdroplet dispenser which can deposit a large number of different molecules onto the substrate [73]. This technique is also applicable to directly pattern the cells themselves [74]. Nevertheless, the resolution is limited to several tens of micrometers and the shape of small patterns is generally limited to a disk.

Active molecules allow dynamic substrate patterning

Methods for reversible substrate patterning have been developed to expose cells to a dynamically reversible surface chemistry. In traditional studies, a highly invasive method uses scraping away of cell monolayer to investigate cellular response to the newly exposed ECM [75]. Now, for these kinds of studies, the thermo-responsive polymer PNIPAAm, which changes from hydrophobic to hydrophilic with temperature can be used [76]. Coupled with collagen, this polymer enables control of cell attachment and detachment from the substrate. This polymer can be patterned on substrates using photolithography [77], and can also be coupled with growth factor and insulin to stimulate cells growth. Another promising method involves photosensitive molecules which enable easy switching of cell adhesion directly through the microscope. Nevertheless, sensitivity of photosensitive materials needs to be improved [78]. Electroactive polymers such as self-assembled monolayer (SAM), which can be switched by electrodes, can also be used to control the adhesion state of cells on surfaces [79, 80].

Topological patterning can also play its role

Figure 6 gives examples of topological patterning. In addition to substrate chemistry, substrate topology can also be patterned. The simplest method to integrate structures in PDMS devices consists in sticking a PDMS channel onto another PDMS layer structured with the desired pattern [81], or to directly mold a PDMS replica from a 3D mold containing both the channel and the structure [2]. These kinds of devices can be used to perform force measurement [82], or integrate microwells arrays which enable individual cell culture with easy cell docking by gravity sedimentation [81] or capillary force [83]. The capillary cell deposition technique is faster and more reproducible than sedimentation but exposes cells to air for several seconds which could be a major drawback for cell viability.

It is also possible to structure the glass substrate before bonding to PDMS channels [83]. Traditional approaches to pattern glass used etching procedures but those techniques require strong know-how, facilities and are not easily accessible to small microfluidic labs. An alternative method to structure the glass substrate is micromolding in capillary (MIMIC), which enables the fabrication of 3D structures on the substrate by injecting the material to be molded (sol-gel, salt, polymer beads, colloid, etc…) into a PDMS channel temporarily in contact with the substrate [51]. This method sometimes gives rise to problems of insufficient air evacuation to prevent air bubble in the channel and incomplete molding. Folch et al. solved this problem through injection by aspiration, but this method does not guarantee a complete filling of the channel [65]. Le Berre et al. integrated microaspiration networks around molding channels to completely release all remaining air bubbles [84]. A useful application of MIMIC is substrate patterning using hydrogel, which solidifies inside the microchannel to generate compartments for passive feeding of cells. Lee et al. used microfluidic moldings to fabricate 3D ECM structures of aligned collagen fibers for cultures of endothelial cells [85].

Other major applications of substrate physical patterning are durotaxis, cell guidance and cell growth control. Durotaxis studies use the ability of cells to sense the surface topography and stiffness. Solon et al. showed that fibroblasts tend to adapt their stiffness to the stiffness of substrate [86], and Engler et al. showed that stem cell differentiation is substrate stiffness dependent [87]. Soft-lithography can create PDMS substrate stiffness from 12 kPa to 1000 kPa depending on the ratio of elastomer to the curing agent [88]. In addition, hydrogel can be created with a gradient of crosslinkers to produce a substrate with a gradient of different mechanical stiffnesses, which can be used to study cellular response to substrate rigidity [89]. Substrate physical patterning can also be used to control cell adhesion and growth on a surface with a defined topology inside a microfluidic channel. Depending on the cell type, microstructure integration enables control of cell morphology and motility without the need for chemical stimuli or contact guidance [90, 91]. Such devices with integrated microstructures have been proposed to guide cell movements, cell separation [92], or guidance of neuronal axon growth [93].

Figure 6: Substrate physical patterning.

(A) Top: Schematic of cell docking in microwells using capillary force. Down: Picture of SG3 yeast docking in microwells [83].

(B) Cell guidance using physically modified substrate. Left: Picture of corneal epithelial cells on SiO2 substrate with 70nm wide ridges. Right: Cell on a smooth SiO2 substrate [91].

(C) Example of pattern made by micromoldling technique. From top to bottom: 150nm diameter pillar, fluorescent image of 40μm stripe of quantum dot, 100μm hole in a 20μm thick PDMS layer, optical image of 100μm width and 20μm height wall of agar gel [84].

?? Ask for the full PDF in comments ??