Experiments were carried out using the AX2 derived, LimE-GFP tagged strain of D. discoideum 6 and the RI9 derived, cAR1-GFP strain 7 . Cells were cultivated in HL-5 medium (ForMedium) at 22°C on polystyrene dishes (Primaria, Falcon), or shaken in suspension at 150 rpm. Cultivation of the cAR1-GFP mutant required addition of 5 μg/ml G418 (Geneticin, Gibco) to the HL-5 medium. To examine chemotactic cells, the culture was starved in a 20 ml phosphate buffer (PB, pH = 6) shaking cell suspension, prior to each experiment 8 . This suspension was supplied with 60 μl of 18 μM cAMP (Sigma) every 6 minutes to synchronize the cells 9 . The developed cells were harvested by washing the suspension three times in PB and then used in the experiments discussed below.

Bright field (BF) and TIRF microscopy were conducted on an Olympus IX81 setup, controlled with the software Cell^R. Observations of D. discoideum were carried out using a 60× PlanApo N oil immersion or a 20× LUCPlan FLN objective. The GFP linked to the DdLimE and cAR1 respectively was excited by the 488 nm line of a 20 mW Ar laser. BF and TIRF images were taken with a 16 bit, 1344 × 1024 pixel CCD camera (Hamamatsu Photonics). MATLAB (MathWorks) was used for data analysis. Brightness and contrast were adjusted using ImageJ 10 . Confocal laser scanning microscopy was conducted on the Olympus Fluoview1000 setup, using a 60× UPlanSApo objective. Excitation of the GFP was provided by the 488 nm line of a 150 mW Ar laser. All flattening techniques have been characterized by taking a confocal z-stack, using cAR1-GFP cells. Since cAR1 is a membrane bound G-protein coupled receptor, cAR1-GFP marks the membrane of the D. discoideum cell. The confocal z-stacks were depth corrected as previously described 11 .

In the agar overlay method, the cell suspension is sandwiched between a layer of glass below and a sheet of agarose above. As fluid is removed from between these layers, the agar overlayer is lowered, and presses on the cells. A detailed description of the agar overlay method can be found in 12 13 . Approximately 200 μl of a D. discoideum suspension in PB was placed on a cover slip (24 × 60 mm, No. 1, Menzel Gläser). An agar sheet (4% agarose, UltraPure, Invitrogen) of ca. 0.15 mm thickness was placed on top of the cell suspension. Cell flattening was induced by removal of PB with filter paper (Nr. 2, Whatman International). The degree of flattening could be controlled by removing different amounts of PB.

The second flattening approach is based on an oil overlay, instead of an agar sheet. Similar to the agar overlay, the underlying principles of this method have long been established. For an early reference from the Dictyostelium literature see 14 . Here, we proceeded as follows:

An amount of 100-200 μl of a D. discoideum cell suspension is placed on a glass bottom Petri dish (Matek). The dish is 50 mm in diameter, with a 14 mm circular well in the center. To maximize the hydrophilicity of the surface, we replaced the glass bottom with a fresh cover slip (No. 1, Menzel Gläser), and this modified dish was then plasma cleaned for ~30 seconds (PDC 002, Harrick Plasma) immediately before depositing the cells. After waiting 15 minutes, to allow the cells to settle, we removed most of the PB with a pipette, leaving a thin wetting layer on the glass surface. A wet cotton thread is then laid along the circumference of the well. To prevent the wetting layer from drying out, we coat it with an oil layer (Neutral oil, Euro OTC Pharma), leaving one end of the thread exposed to the air outside. As this portion of the thread dries, the aqueous fluid underneath the oil layer is gently wicked away, resulting in a thinning of the wetting layer. When the desired degree of flattening is achieved, the wick is removed. The setup is illustrated in Figure 1.

We used standard PDMS-based soft-lithography to fabricate microfluidic devices 15 . A chrome mask (Cornell Nanofabrication Facility, Cornell University) was used to create a photoresist micro-pattern (SU8-25 or SU8-50, Microresist Chem) on a Si-wafer (polished, diameter 100 mm, SiMat) by photolithography. See the individual sections below for the different mask geometries. The height of the SU-8 pattern was measured with a white light interferometer (Wyko NT 1100, Veeco). Polydimethylsiloxane (PDMS, Sylgard 184, Dow Corning) was molded against the patterned master wafer. Air bubbles were removed in a vacuum chamber and the PDMS was then cured for 2 hours at 75°C. A block of PDMS carrying the imprint of the microstructure was cut out using a sharp knife. Inlets and outlets were punched into the PDMS with a clean syringe tip (19 gauge stainless steel, McMaster). The microstructured PDMS surface as well as the surface of the cover slip (24 × 60 mm, No. 1, Menzel Gläser) were treated by an air plasma for 5 minutes. Afterwards, bonding between the microstructured PDMS surface and the cover slip was initiated by bringing the two surfaces into contact with each other.

The microfluidic cell flattening devices described below commonly rely on changes in the height of the flattening chamber by deformation of the PDMS ceiling. Prior to a cell flattening experiment, a suspension of D. discoideum cells was filled into the microchannel through PTFE tubing (Novodirekt) via a 23 gauge sterile and sharpened syringe (Kruuse). The device was left for 15 minutes to allow the cells to settle down on the glass surface.

To quantify the changes in channel height, a 100 μM solution of fluorescein (Acros Organics) in PB was filled into the channel, and a constant flow of 5 μl/h was applied to prevent photobleaching 16 . The fluorescence was excited with a 150 W Xe/Hg mixed gas arc burner (part of the MT20 system, Olympus) and observed with a 10× Plan N objective. Our imaging region was large enough to cover the middle of the channel, as well as a portion outside of the channel (to get I_back, used for background subtraction). We first measured the fluorescence intensity in the non-deformed flattening chamber and associate the corresponding intensity value I(0) with the maximal channel height h₀, measured as described above. Assuming that the intensity I scales linear with the height h, we estimated the height of the flattened chamber by

is averaged over at least 100 pixels and the corresponding standard deviation in I is used for calculating the error bars.

The single layer microfluidic device consists of a single PDMS layer with a simple through-flow micro-channel, see Figure 2A. The key feature of this channel is the high aspect ratio (width ≫ height) chamber in the middle section. A syringe pump is used to withdraw liquid through the channel. The higher the flow rate, the greater the pressure drop between the outer atmosphere and the chamber. This pressure difference lowers the elastic PDMS ceiling, causing the cells below to be flattened.

Figure 2B shows the layout of the chrome mask used to fabricate the single layer microfluidic device. The final height of the microfluidic channel after spincoating the SU8-25 photoresist was 14 μm (the taller the microfluidic channel, the higher the flow speed required to flatten cells inside the channel). The area of the chamber, where experiments were carried out was 2 × 2 mm in size. It was stabilized by two rows of pillars (Figure 2B, in red), to prevent uncontrolled collapsing. We used 45 g of PDMS and 4.5 g of curing agent (100:10), to create a ca. 3 mm tall microfluidic device. After the device was assembled, it was heated at 75°C for 10 minutes to prevent an untimely, irreversible collapse of the chamber ceiling. The cells were loaded as described above. Then, the outlet was connected to a reservoir of PB (see Figure 2C). To lower the PDMS ceiling, PB was withdrawn from the channel at a constant rate using a syringe pump (PHD 2000, Harvard Apparatus). The degree of flattening depends on the suction rate. It is crucial to remove bubbles from the microchannel, the tubing, and the syringes, since they may expand in volume, leading to an uncontrolled drop in the actual flow speed in the microchannel.

The closed-end actuator for microfluidic cell flattening was inspired by the PDMS-based fabrication of microfluidic valves introduced in 17 and relies on the same principle as the device described in 16 . This microfluidic device consists of two perpendicularly oriented arrays of channels in two PDMS layers, see Figure 3A and 3B. The channels of the lower layer contain the cells that are to be flattened. The actuation channels in the top layer are closed-end channels, i.e., they contain an inlet but no outlet. When pressure is applied to the inlet of an actuation channel, the thin PDMS film separating the upper and lower layers deforms, causing the height of the lower channels to decrease (Figure 3C).

The channels in the lower layer are 3 cm in length, 15 μm in height and 500 μm in width (Figure 3, PDMS I). This lower layer is made of a 100:9 mixture of PDMS to curing agent. Two cover slips (No. 1) are placed in the PDMS, on opposing sides of the structured pattern. These cover slips act as spacers, fixing the height of a third cover slip (24 × 60 mm, No. 1) placed across the microstructure (Figure 3D and 3E). After the PDMS cured, a rectangular block, containing the three channels is cut out. In this process, we must cut through the upper cover slip, so that the block consists of PDMS and the midsection of the upper cover slip. The ends of this cover glass, along with the spacers are left behind.

We take this block, and punch the inlet and outlet holes. Then, we carefully cut out the layer of PDMS above the embedded cover slip. It is critical that the corresponding cuts do not penetrate beyond the glass into the PDMS below; otherwise the channels will be damaged. The piece of PDMS, which was above the embedded cover slip, is put aside for use later. We remove all pieces of the cover glass from the main PDMS block. A fresh cover slip (22 × 40 mm, No. 1, Gold Seal) is placed on top of the thin midsection. The previously removed PDMS piece is replaced above the fresh glass (Figure 3F) and provides stability to the main block as it is plasma cleaned and bonded to a cover slip (24 × 60 mm, No. 1, Menzel Gläser). Finally, the stabilizing PDMS piece and cover glass are carefully removed from the block, and discarded. The lower layer is now assembled (PDMS I in Figure 3G).

The upper actuation channels are 1 cm in length, 30 μm in height and 500 μm in width (Figure 3, PDMS II). These closed-end channels are made from a 100:11 mixture of PDMS and curing agent. After curing, we cut out the PDMS block, and punch the inlets. We then plasma clean this upper layer with the previously assembled lower layer. We assemble the two layers such that the patterned side of the top layer (PDMS II in Figure 3G) is bonded on top of the mid-section of the bottom layer. The channels in the two layers are oriented perpendicular to each other. This concludes the construction of the closed-end actuation device.

We load the cells into the lower channel. After we allow the cells to settle and adhere to the glass substrate, we run a 5 μl/h flow of PB. The cells located in the region of overlap between the lower and upper channels (Figure 3G, marked in red) are the cells that can be flattened.

Next, to apply pressure to the upper channels, we use an infuse/withdraw syringe pump to drive a 500 μl syringe (Hamilton). The syringe and tubing are filled with water. The syringe pump is intrinsically a volume controlling device, not a pressure controlling device. We convert the volume change, induced by the syringe pump, into a pressure change by introducing a large air bubble (V₀ ≈ 200 μl) near the plunger end of the syringe. Note that this bubble provides most of the compliance in the system, so Boyle's law tells us

where P _actuation is the pressure in the actuation chamber, and V _pump is the volume injected by the syringe pump. By stopping the syringe pump, we can maintain a fixed pressure. To decrease the pressure, we simply run the syringe pump in reverse.

Note that, initially, the actuation channel is filled with air, but as pressure is applied, this air slowly permeates into the PDMS, and eventually will be displaced by water. This volume is negligible, when compared to V ₀, so the pressure equation above remains valid.

Similar to the previous device, the through-flow actuator is composed of two layers, an actuation channel on top and a channel for cell flattening below. The lower channel is filled with a D. discoideum cell suspension or a fluorescein solution for characterization. In contrast to the previous setup, the actuation channel is a through-flow channel connected to a continuous oxygen supply. The flow of oxygen is controlled by a metering valve. When the pressure in the actuation channel is increased, the PDMS sheet between the two layers is deformed, flattening the cells (see Figure 4B).

The microstructures were prepared by photolithography on two different master wafers. The layout of the first wafer, used for molding the actuation channel, was the same as for the single layer device (see Figure 2A) with a height of 28 μm. The actuation channel was made by covering the wafer with a layer of the PDMS (100:10) about 1 cm thick. To prepare the thin lower layer, we used a second wafer with a channel structure of 500 μm width and 24 μm height (Figure 4A). In order to distribute the PDMS uniformly, two spacers were used, each consisting of three cover slips (No. 1) that had been glued together. The spacers were taped to opposing edges of the wafer and a bubble-free PDMS (100:10) droplet was placed between them. The long edge of a thick glass slide was used to evenly spread the PDMS between the two spacers. Before curing the PDMS, the spacers were removed.

The actuation channel was cut out and holes were punched into the PDMS. After plasma cleaning, this PDMS block was bonded on top of the thin PDMS film on the master wafer, containing the lower channels. Then, the thin layer was cut along the edges of the PDMS block and the two bonded layers were carefully removed from the wafer. After this, holes for the lower channel were punched into the PDMS. The structured side of the device and a cover slip were plasma cleaned and bonded together as described. The assembled microfluidic through-flow actuator is displayed in Figure 4C. A suspension of D. discoideum LimE-GFP cells was filled into the lower channel. After 15 minutes, PB was supplied with a constant flow rate of 5 μl/h for the entire experiment. The pressure in the actuation channel was controlled by a metering valve, located between the channel inlet and the oxygen supply (0 to 1 bar above atmosphere) as shown in Figure 4D. The pressure was measured with a glycerin filled manometer (0 to 1 bar, Swagelok).

Following the protocol presented above, we flattened D. discoideum cAR1-GFP cells by an agar and an oil overlay. In the agar overlay method, the blotting of the fluid between the agar sheet and glass is difficult to control, and demands great care. Under the agar overlay it was possible to flatten the cells to less than 3 μm in height (Figure 5A and 5C). The temporal stability of flattening is limited by the drying of the agar. To overcome this limitation, a humid environment is required, as reported in 18 .

An exemplary D. discoideum cAR1-GFP cell, flattened to ~3 μm under an oil overlayer is shown in Figure 5B and 5D. We observed that the degree of flattening was prone to strong variations. In particular, flattening was strongly dependent on the distance between the cell and the wick. Flattening was further limited by the breakup of the aqueous layer, see Figure 6A. This break up of the wetting layer into isolated islands was frequently observed when liquid in the aqueous layer was reduced beyond a critical amount. Due to these limitations, the degree of flattening was harder to control, when compared to the agar overlay.

By adjusting the rate at which fluid is withdrawn from the single layer device, we can tune the degree of flattening. When we fix the flow rate, the channel height will eventually approach a constant value, which in turn allows us to conduct stable, long term microscopy on flattened cells. The device characterization with fluorescein is shown in Figure 7A. Here, we applied a suction speed of 250 μl/h, which reduced the channel height by 67% (from 14 to approximately 4.7 μm). After the fluid flow was stopped, the ceiling relaxed back into its initial position. During the first 5 minutes, 80% of the original height was recovered, but it took more than 16 minutes to relax back into its original height.

The withdrawal speed of 250 μl/h was also used to flatten cAR1-GFP cells. The confocal x-y scans of a cell before flattening, and during flattening are shown in Figure 7B and 7C. The height of the cell decreased from 11 μm (Figure 7D) to ~4 μm (Figure 7E) in this device.

In the double layer devices, we tune the height of the cell-containing channel by adjusting the pressure in an actuation channel above. The microfluidic closed-end actuator was characterized in Figure 8. The through-flow actuator was characterized in Figure 9. In these figures, subfigure A shows the temporal evolution of the lower channel height when flattened and unflattened. Subfigure B shows height profiles across the channel, for different degrees of flattening. Subfigures C-F show confocal images for a cAR1-GFP cell before (C, E), and during (D, F) flattening.

For the closed-end device, we saw that increasing the pressure in the actuation channel leads to a decrease of height of the lower channel from 15 μm to ~4 μm (Figure 8A) within approximately 70 seconds. The rates of pressurization and de-pressurization in the actuation channel are governed by the speeds at which the connected syringe pump runs. By running the syringe pump at the same rate, in reverse, we unflattened the lower channel in about the same amount of time it took to flatten. By using higher flow rates on the actuation syringe pump, or by decreasing the initial size of the air bubble in the actuation syringe (see the methods section above), we can decrease the time scales.

In the through-flow device, the time for the lower channel to flatten was even shorter. By quickly adjusting the oxygen pressure in the actuation channel, we were able to change the channel height from 24 μm to less than 1 μm in less than 10 s (Figure 9A). The reverse process had slightly longer time resolution, ~20 s. The limitations in speed and stability of this device are caused by the pressure supply. Pressure is applied by manually opening a needle valve on the supply end, and is released by closing this valve and opening an exhaust needle valve (Figure 4D). Replacing this system with a more sophisticated automated pressure control should improve the stability of the flattened state, and the speed at which flattening/unflattening occurs.

The confocal x-y scans of cAR1-GFP cells before and during flattening show a change from a rather irregular, ruffled geometry to a regular one, where the membrane is firmly pressed against the channel walls. This restriction improves TIRF imaging, since only the surface next to the glass can be visualized. A representative TIRF image of D. discoideum LimE-GFP cells for each method is shown in Figure 6B-F. The LimE-GFP label has been established as marker for filamentous actin 19 . Previously, the actin dynamics in these cells was investigated for cells flattened by an agar overlayer. For each of the setups, the actin bundles are visible, implying that the cells have been confined to the glass surface by the individual techniques. Furthermore, in the supplementary material, we show a time lapse TIRF movie of cells in the single layer device (additional file 1).