Reprogramming cell shape with laser nano-patterning

The control of cell shape in vitro by the use of different designs of micropatterned substrates has been a useful tool to investigate the fundamental rules of morphogenesis (Théry, 2010). This method has revealed that in addition to shape, cell behavior is also sensitive to the spatial distribution of its extracellular adhesions. The cell adhesion pattern has notably been shown to regulate cell architecture (Brock et al., 2003; Parker et al., 2002; Rossier et al., 2010; Théry et al., 2006a), polarity (Desai et al., 2009; James et al., 2008; Lombardi et al., 2011; Pitaval et al., 2010), migration (Doyle et al., 2009; Pouthas et al., 2008), division (Fink et al., 2011; Samora et al., 2011; Théry et al., 2007) and differentiation (Dupont et al., 2011; Kilian et al., 2010; McBeath et al., 2004).

The dynamics of cellular responses to changes in the microenvironment is a fundamental property of living systems that ensures the functional and mechanical coherence of tissues during development or renewal (Lu et al., 2011). However, the manipulation of changes in the microenvironment in vitro is limited in conventional surface micropatterning methods because the design of the micropattern is fixed at the point of fabrication. Hence, cellular responses to the geometry of these micropatterns can only be observed at steady state; whereas cellular responses in real time to changes in the microenvironment cannot be measured. This has been a major limitation to the experimental investigation of the dynamic processes that support cell and tissue morphogenesis.

Several approaches have been used to overcome this limitation and alter the adhesive environment surrounding the micropatterns on which living cells are attached (Nakanishi et al., 2008). Electric potential has been used to detach cell-repellent coatings, either by detaching micropatterned electroactive groups (Raghavan et al., 2010) or by desorbing coatings on electrodes (Gabi et al., 2010; Kaji et al., 2006), thereby allowing constrained multicellular groups of cells on large micropatterns to specifically invade the activated regions. The minimum size of these regions was about 10 μm (Gabi et al., 2010). Alternatively, cell-repellent moieties have been chemically linked to the silane coating by photo-cleavable groups so that they are released in response to UV light. Similarly, loss of the cell-repellent coating could promote the local attachment of cells in suspension (Kikuchi et al., 2008b), trigger cell migration (Nakanishi et al., 2007) or promote the invasion of new areas by multicellular groups (Kikuchi et al., 2008a). With this method, substrate exposure to UV through a photomask placed in the optical plane of a microscope allowed the addition of new adhesive regions whose size could be as small as 5 μm (Nakanishi et al., 2007).

Here, we have developed a simple method to ablate the cell-repellent properties of the polyethylene glycol (PEG) coating in the vicinity of a live single cell already attached to a micropatterned substrate. The method uses a commercially available polymer to coat the cell culture substrate (poly-L-lysine–PEG) and a pulsed UV laser to introduce additional adhesive regions. The manipulation of the adherent properties in the microenvironment of a single cell with sub-micrometer resolution enables the precise control of intracellular architecture remodeling in real time.

Oxidation of a PEG layer on polystyrene (PS)-coated glass is an efficient and versatile micropatterning method to accurately define geometries that can stereotypically direct cell adhesion and cell shape. Oxidation of PEG can be achieved by deep UV (wavelength below 200 nm) exposure through a chromium mask (Azioune et al., 2010; Azioune et al., 2009). Deep UV creates ozone that oxidizes the surface and allows protein adsorption (Mitchell et al., 2004). To create new micropatterned regions in the presence of living cells, we used a Q-switched laser producing 300 picoseconds pulses at 355 nm and a high numerical aperture objective (Fig. 1A). The accumulation of pulses of energy in a highly confined volume induced the formation of localized plasma responsible for local oxidation and further destruction of irradiated materials (Colombelli et al., 2004; Colombelli et al., 2007; Pfleging et al., 2009; Vogel et al., 2005). We modulated the number and repetition rate of laser pulses as well as the laser power to control the size of individual spots. Laser patterning was conducted in the presence of Alexa-Fluor-546-labeled fibrinogen to detect protein adsorption on exposed regions (Fig. 1A). Surface modifications were also characterized by atomic force microscopy (AFM). Hexagonal arrays of spots separated by 160 nm were made using high or low laser power. The high-power beam did not allow homogeneous fibrinogen adsorption and resulted in a honeycomb-like topography within the glass slide, as seen by AFM, with holes corresponding to laser spots (Fig. 1B). The low-power beam resulted in superficial (4 nm) removal of the poly-L-lysine–PEG and PS layers, and efficient protein adsorption (Fig. 1C). Therefore, these conditions were further adopted for live-cell patterning in the rest of the study. However, the roughness of the PS layer induced a noisy AFM signal, preventing the measurement of single laser spot width. Therefore, we decide to measure single spot size with the intermediate power beam and a smaller polystyrene surface to induce detectable surface modifications without affecting the glass coverslip. In these conditions, the size of a single spot was 300 nm in diameter (Fig. 1D). However, it should be noted that the intermediate power used in this procedure led to an overestimation of the spatial resolution. In the regular, low-power conditions used for live-cell patterning, the spot size was probably smaller.

Initial cell adhesion and early spreading, when cells just contacted the substrate, depend on the micro- and nano-scale organization of adhesive ligands (Geiger et al., 2009). Integrins are transmembrane proteins connecting the ECM and the intracellular actin network. The engagement of actin filaments between individual integrins contributes to the clustering of integrins and the stimulation of membrane deformation and cell spreading. The distance between individual integrins needs to be smaller than 70 nm to allow actin filaments to connect their intracellular domains (Arnold et al., 2004). Integrin clusters must contain at least four integrin molecules within 60 nm to allow cell attachment (Schvartzman et al., 2011). The critical distance between these clusters that allows cell spreading depends on cluster size. It can be 25 μm on 9 μm² adhesion spots, but is reduced to 5 μm on 0.1 μm² spots (Lehnert et al., 2004). This suggests that, after the spreading phase, the subsequent cell extension and contraction onto the new adhesion sites could also depend on nano- and micro-scale organization of those adhesive sites.

The 300 nm width of single spots (Fig. 1D) allowed us to investigate the nano- and micro-scale dependency of the cell extension and contraction phase. RPE1 cells were first plated on micropatterns made with classical deep-UV exposure through a photomask) Azioune et al., 2010; Azioune et al., 2009). They were allowed to spread and contract until their shape adopted the convex envelope of the micropattern (Rossier et al., 2010; Théry et al., 2006a). Then, using galvanometric mirrors, the laser beam was scanned on the substrate to draw the regions to be oxidized (Fig. 2A). This scanning method was a versatile and modular way to design any kind of geometry anywhere around cells. Similar results could also be obtained by moving the sample rather than the laser beam. Local PEG oxidation allowed cells to form new adhesions and to extend on the irradiated regions. The distance between individual adhesion spots could be varied up to the formation of a contiguous adhesive line (Fig. 2A). Hence, this allowed us to identify the critical geometrical parameters allowing cell extension and contraction. Cells were attached to H-shaped micropatterns and adopted a square shape of about 900 μm². To test the requirements for the induction of a cell extension and the assembly of a new filament bundle, the two adhesive bars were extended at one side of the H with two new adhesive regions, made of parallel lines of adhesion spots. Cells did not extend on lines made of spots separated by 1600 nm (Fig. 2B). Only a few cells could initiate extension on lines made of spots separated by 800 nm. More cells extended onto these lines when the inter-spot distance was reduced to less than 400 nm (Fig. 2C,E). However, the cell extensions could not generate a substantial contraction between the new adhesive regions, as revealed by the low values of the cell edge curvature radius (Théry et al., 2006a) (Fig. 2C,E). Lines perpendicular to the longitudinal orientation of the H bar induced the same phenotypes (supplementary material Fig. S1). Interestingly, although cells could easily extend on dense square arrays of spots in which the inter-spot distance was 400 nm, they could only generate a substantial contraction between the new adhesive regions when this distance was reduced to 160 nm, i.e. when the region was almost continuously adhesive (Fig. 2D,E). These results show that RPE1 cells require the adhesion spots to be separated by less than 400 nm to stabilize the new cell extension and a continuous adhesive region to generate a substantial contraction force. Previous reports showed that mouse melanoma cells can spread on similar arrays of 300-nm-wide adhesion spots if their spacing is smaller than 5 μm (Lehnert et al., 2004). This suggests that the critical distance for extension and contraction in spread cells is one order of magnitude smaller than the critical distance for attachment and spreading.

Cell shape could be reprogrammed by adding the dense adhesive regions described above. For example, it was possible to remove the PEG from a region defined by a horizontal bar next to an apex of a triangularly-shaped cell constrained on a V shape (Fig. 3A). After this ablation, the cell also adhered to this region and adopted a square shape (Fig. 3B). Actin network reorganization during this cellular transformation was followed by monitoring Lifeact–GFP (Riedl et al., 2008). As the cell spread on the new bar, it formed many new actin cables, which connected the original and the new micropatterns. This showed that cells not only spread on to the laser-designed regions but also developed new internal cables during cell shape deformation from triangle to square (Fig. 3B). The tension in these cables was probably required to support cell shape changes (Rossier et al., 2010; Théry et al., 2006a). We further investigated cell shape changes by adding two bars, above and below the original V-shaped micropattern, and monitored cell shape extension in real time (Fig. 3C and supplementary material Movie 1). As the cell shape changed, some peripheral actin cables disappeared (arrowhead in Fig. 3C) whereas others were assembled (arrows in Fig. 3C). This suggested that cell shape reprogramming is supported by complex remodeling of intracellular structures.

Cell shape is supported by various structural elements made of actin filaments. They can be classified into branched meshworks and filament bundles (Michelot and Drubin, 2011). Both are highly dynamic and remodeled during cell shape changes (Rafelski and Theriot, 2004). We further tested whether laser-based patterning could be used to guide not only cell shape changes but also precise intracellular remodeling of these structural elements. Assembly of each structural element is dependent on local adhesion geometry (Brock et al., 2003; Parker et al., 2002; Théry et al., 2006a). When cultured on an H-shaped micropattern, cells adopt a square shape. For a given cell, branched meshworks were established along the adhesive bars and actin bundles across the gaps. Each type of actin-based structure could be induced, displaced or removed during square cell shape transformation into a rectangle by adding new adhesive regions of defined geometries. Extending the length of the juxtaposed bars on one side promoted the displacement of the actin bundle so that it remained situated between the tips of the two bars (Fig. 4A). Connecting the tips of two bars with a contiguous adhesive region favored actin bundle disassembly and the formation of a branched meshwork (Fig. 4B). Adding two small bars perpendicular to one of the H bars induced formation of an additional peripheral actin bundle (Fig. 4C). These results showed that in addition to controlling the global cell shape, the geometry and position of new adhesive regions can be used to finely control intracellular architecture remodeling.

In polarized cells, such as migrating cells or epithelial cells, the actin network is polarized into a branched meshwork on one cell side and contractile stress fibers on the other. The precise subcellular location of stress fibers and acto-myosin contractile activity are crucial to the determination of actin network polarity (Cramer, 2010) and internal cell polarity (Théry et al., 2006b). Live patterning could be used to precisely control and orient this actin network polarization step. A bar was added next to cells plated on discoidal micropatterns (Fig. 5A). Initially, actin networks did not display any significant polarized architecture. After live patterning, cells rapidly extended on the new bar and initiated the formation of contractile stress fibers along the edges connecting the disc and the tip of the bar (Fig. 5B and supplementary material Movie 2). Upon completion of this extension and contraction phase, cells ended up with a highly asymmetrical shape and polarized actin network.

The laser-patterning method we developed will find a broad range of applications in addition to its role in the study of living cells. The non-specific action of this method represents a versatile way to design micropatterns on various surfaces. It does not require specific photo-activatable substrates or photo-sensitive ligands, it simply ablates the protein-repellent coating. Therefore, it could be applied to any PEG-coated surfaces. In addition, it is a contact-less patterning method, which therefore offers the possibility of designing micropatterns on three-dimensional substrates or in close microfluidic devices. Finally, rounds of laser patterning and protein adsorption can be repeated at will to allow multi-protein patterning of substrates.

Here, we have demonstrated that this new and simple method for surface nano-patterning in live cell culture offers a precise control in real time of cell shape modifications and of intracellular architecture. This method should pave the way for further investigations of dynamic cellular responses to nano- and micro-scale changes in the microenvironment. It also opens new possibilities to adapt ‘on the fly’ the design of new geometrical constraints to the observed cell behavior. Therefore, it will enable the fabrication of micropatterned regions during the growth of multi-cellular colonies. This will enable new insights into tissue engineering.

Glass coverslip micro-patterning has been described elsewhere (Azioune et al., 2010). Briefly, coverslips were first spin-coated for 30 seconds at 3000 r.p.m. with adhesion promoter Ti-Prime (MicroChemicals), baked for 2 minutes at 120°C and then spin-coated with 1% polystyrene solution (Sigma) in toluene (Sigma) at 1000 r.p.m. for 30 seconds. Polystyrene-coated coverslips were oxidized through oxygen plasma (FEMTO, Diener Electronics) for 10 seconds at 30 W before incubating with 0.1 mg/ml poly-L-lysine (PLL)–PEG (Cytoo) in 10 mM HEPES, pH 7.4, for 15 minutes. After drying, coverslips were exposed to 165 nm UV (UVO cleaner, Jelight) through a photomask (Toppan) for 2 minutes. After UV activation, coverslips were incubated with a 20 μg/ml of fibronectin (Sigma) and 10 μg/ml Alexa Fluor 546 fibrinogen conjugate (Invitrogen) in phosphate-buffered saline (PBS) solution for 30 minutes. Coverslips were mounted in magnetic chambers (Cytoo) and washed three times with sterile PBS before plating cells.

LifeAct–mGFP plasmids were kindly provided by Wedlich-Soldner (Riedl et al., 2008). The lifeact–mGFP fragment was amplified by PCR using primers flanked with specific restriction enzyme sites (namely EcoRI and NotI). This fragment was subsequently cut and ligated with the pLVX lentiviral vector (Dupont et al., 2011) (Clontech), which was also cut with corresponding restriction enzyme. The virus carrying lifeact–mGFP was generated using the lenti-X packaging system (Dupont et al., 2011) (Clontech). hTERT-RPE1 cells (infinity telomerase-immortalised retinal pigment epithelial human cell line) were subsequently infected with the virus followed by antibiotic selection, according to the manufacturer's instructions (Clontech).

hTERT-RPE1 cells were cultured in DMEM F-12 (Gibco) supplemented with 10% fetal bovine serum (A15-551, PAA), 50 units/ml penicillin and 50 μg/ml streptomycin (Gibco). Cells were cultured in a 5% CO₂ incubator at 37°C. Cells were trypsinized, centrifuged, resuspended in fresh medium and allowed to spread on micropattern for 4 hours before the beginning of the experiment.

Laser patterning was performed using of a Laser illuminator iLasPulse (Roper Scientific) set-up on an inverted microscope (TE2000-E, Nikon). iLasPulse is a dual-axis galvanometer-based optical scanner that focuses the laser beam on the sample (diffraction limited spot size) on the whole field of view of the camera. It includes a telescope to adjust laser focalization with image focalization and a polarizer to control beam power. The laser used is a passively Q-switched laser (STV-E, TeemPhotonics), which produces 300 picosecond pulses at 355 nm (energy/pulse, 1.2 μJ; peak power 4 kW; variable repetition rate, 0.01–2 kHz; average power, ≤2.4 mW). Laser displacement, exposure time and repetition rate were controlled using Metamorph software (Universal Imaging Corporation). The objective used was a 100× CFI S Fluor oil objective (MRH02900, Nikon). The area to pattern was filled with different density of spot. Each spot was exposed for 20 mseconds at a repetition rate of 600 Hz. The polarizer was set to have an energy per pulse of 300 nJ.

To visualize the patterned zone, a polystyrene- and PLL–PEG-treated coverslip without cells was mounted in a magnetic chamber. This chamber was filled with a 20 μg/ml fibronectin (Sigma) and 10 μg/ml fluorescent fibrinogen conjugate (Invitrogen) PBS solution. Laser patterning was then conducted as described above and protein adsorption was allowed for 30 minutes. Coverslips were rinsed with PBS and fluorescent images were then taken using a 100× UplanSApo oil objective (Olympus) using an Olympus BX61 microscope and a CoolSNAP HQ2 camera (Photometrics).

Magnetic chambers containing the coverslips and filled with cell culture medium were placed on the microscope (TE2000-E, Nikon, France) in a stage incubator system at 37°C and 5% CO₂ (Chamlide WP, Live Cell Instruments). Epifluorescence images of cells were acquired through a 100× CFI Plan Fluor oil objective or 60× CFI Apo TIRF oil objective (MRH02900 and MBH76162, respectively, Nikon) and a QUANTEM:512SC cooled EMCCD camera (Photometrics). The whole system was controlled by Metamorph software (Universal Imaging Corporation).

Extension and curvature measurements were performed using ImageJ software. For extension measurements, the distance between the border of the initial pattern and the border of the cell extended on the new pattern was measured. Two measures were performed for each cell (one for each extension zone). For curvature measurements, a circle was manually drawn along the unattached edge of the cell joining the two new adherent zone and the radius of the circle was measured automatically. Only cells that had extended on both bars were measured. All the measurement series were compared using a one-way ANOVA comparison test. Means were considered as significantly different when the P value was below 0.05.

Laser-made micropatterns were observed and quantified by atomic force microscopy (AFM) to see the topographical effect induced by the procedure. AFM was performed on a 5500 LS AFM stage (Agilent) or a DI 3100 AFM stage (Veeco). Coverslips were attached to a glass slide and mounted in the AFM. Ambient tapping mode imaging was performed using a NSC19 cantilever (Mikromasch). Scan parameters were optimized to minimize the difference between the set point and the amplitude of the free cantilever while maintaining a stable image.

To estimate the size of a single spot, a polystyrene- and PLL-PEG-treated coverslip without cells was mounted in a magnetic chamber. Because the single spot margins could not be clearly seen in AFM owing to the small size of the topographical step (8 nm) compared with the polystyrene surface roughness (see Fig. 1), the laser beam intensity was increased to make small holes in the polystyrene layer. Therefore the width of 300 nm is an overestimation of the actual spot we used in the experiments in the presence of cells.

We thank David Peyrade and Patrice Baldeck for technical help during preliminary experiments and Simon Le Denmat for AFM analyses.