Simone
Tanzi
a,
Marco
Matteucci
a,
Thomas Lehrmann
Christiansen
a,
Søren
Friis
b,
Mette Thylstrup
Christensen
b,
Joergen
Garnaes
c,
Sandra
Wilson
b,
Jonatan
Kutchinsky
b and
Rafael
Taboryski
*a
aDepartment of Micro- and Nanotechnology, Technical University of Denmark, Building 345E, DK-2800 Kongens Lyngby, Denmark. E-mail: rafael.taboryski@nanotech.dtu.dk
bSophion Bioscience A/S, Baltorpvej 154, DK-2750 Ballerup, Denmark
cDanish Fundamental Metrology, Matematiktorvet, Building 307, DK-2800 Kongens Lyngby, Denmark
First published on 23rd September 2013
In this paper, we demonstrate recordings of the ion channel activity across the cell membrane in a biological cell by employing the so-called patch clamping technique on an injection-molded polymer microfluidic device. The findings will allow direct recordings of ion channel activity to be made using the cheapest materials and production platform to date and with the potential for very high throughput. The employment of cornered apertures for cell capture allowed the fabrication of devices without through holes and via a scheme comprising master origination by dry etching in a silicon substrate, electroplating in nickel and injection molding of the final part. The most critical device parameters were identified as the length of the patching capillary and the very low surface roughness on the inside of the capillary. The cross-sectional shape of the orifice was found to be less critical, as both rectangular and semicircular profiles seemed to have almost the same ability to form tight seals with cells with negligible leak currents. The devices were functionally tested using human embryonic kidney cells expressing voltage-gated sodium channels (Nav1.7) and benchmarked against a commercial state-of-the-art system for automated ion channel recordings. These experiments considered current–voltage (IV) relationships for activation and inactivation of the Nav1.7 channels and their sensitivity to a local anesthetic, lidocaine. Both IVs and lidocaine dose–response curves obtained from the injection-molded polymer device were in good agreement with data obtained from the commercial system.
Despite their enormous potential as druggable targets, the use of ion channels for drug discovery has been well behind the expectation, mainly due to the lack of adequate screening technology.8 Conventional manual patch clamping delivers high information content, but it is a complex technique that requires skilled personnel and suffers from extremely low throughput, while, on the other hand, all indirect screening methods suffer from low specificity and, thus, the associated risk of generating false negative or positive results.8,9 With the promise of removing this trade-off, the first automated parallel patch clamping system (APC) was made available in 2003.10 Since then, 5 companies (Molecular Devices, Sophion Bioscience, Nanion, Cytocentric, and Fluxion) have entered the market, establishing APC as an essential technology for secondary screening, lead optimization and cardiac safety testing.11 Performances of the different systems have been discussed in a few reviews.11–14 APC systems replace the use of conventional glass micropipettes with single-use, disposable devices fabricated using materials such as quartz15 and silicon/silicon dioxide.16 The majority of the available platforms share the so-called planar approach, where cells in suspension are blindly positioned by suction onto micro-apertures made as through holes in thin membranes.15 Alternatively, a lateral approach has been proposed, where suspended cells are trapped at lateral apertures generated at the junction of two microfluidic channels.17
The ability to form a high electrical resistance seal, preferably in the order of GΩ (the so-called gigaseal) between the aperture and the biological cell membrane, and the fabrication of such micro-sized through holes in thin substrates (10–20 μm) are the two most relevant benchmarks. These constraints have traditionally narrowed the range of usable materials and technologies to almost exclusively micromachining of silicon and glass, at least for commercial use.
The use of polymeric substrates has also been widely investigated, and planar microfluidic chips have been demonstrated in hybrid polydimethylsiloxane–polyimide (PDMS–PI),18 polyimide,19 oxygen plasma-treated PDMS20,21 and polyethylene glycol (PEG)–SU-8 mixture.22 Unfortunately, all the aforementioned polymer devices have so far failed to deliver relevant electrophysiological results. It has been more or less implicitly believed in the community that silicon oxide surfaces possess a unique surface chemistry for forming gigaseals, essentially through van der Waals interactions.23–25 The interaction between lipid membranes and other surfaces is, however, complicated and only poorly understood and, thus, leaves some leverage for challenging the uniqueness of these surfaces26–30 by employing polymers. Ionescu-Zanetti et al. implemented lateral junctions in a microfluidic device fabricated by casting of polydimethylsiloxane (PDMS).17,31 In 2009, this device was eventually commercialized by Fluxion, and it remains the only APC system relying on cornered apertures to date.32
Looking at all of the commercially available APC systems, the manufacturing cost per device is still prohibitive for establishing APC as a routine technique.33 Of the two materials employed nowadays, silicon remains extremely costly to purchase and machine, while PDMS is less suitable for mass production. Moreover, PDMS has issues when used for biological applications such as high permeability for small molecules, non-cross-linked oligomers, and surface diffusion of low molecular weight chains.34
The aim of this work is to propose a method for massive production of extremely cheap APC systems without compromising functionality. The trick lies in understanding the requirements to form tight seals between thermoplastic polymers and the cells. By shaping apertures that allow for a considerably larger effective interface area between low-energy polymer surfaces and the cell membrane, we demonstrate the formation of gigaseals, enabling high-quality ion channel recordings. The proposed device consists of only two parts: an injection-molded part in TOPAS cyclic olefin copolymer (COC) comprising the microfluidics and a polymer film of the same polymer grade, which is thermally bonded to the injection-molded part.35 For comparison, the commercially available QPlate™ from Sophion Bioscience comprises several materials, thermoplastic polymer, elastomer, ceramic, glass, and silicon. Moreover, TOPAS has none of the previously mentioned limitations of PDMS. Furthermore, the materials and fabrication platform fulfills all the requirements for cheap manufacturing in terms of “design for manufacturability”, “economy of scale” and production yield, and it avoids all drawbacks typical of prototyping methods.36 After an initial investment in master mold origination, injection molding delivers a cost per part that exponentially decreases with the number of parts, becoming the cheapest technology when production is greater than ~10000 units.36 After assembly into chips, the systems were tested with HEK cells expressing Nav1.7, a voltage-gated sodium channel appropriate for benchmarking studies. Experiments were designed to explore current–voltage (IV) relationships for activation and inactivation of Nav1.7 channels and their sensitivity to a local anesthetic, lidocaine. Both IVs and lidocaine dose–response curves obtained from the injection-molded polymer device were benchmarked against the commercially available QPatch™ system from Sophion Bioscience. In the following section, we provide the details of the device layout and functionality, benchmarking tests against existing state-of-the-art methods, and finally a surface topology characterization of the patching capillaries.
![]() | ||
Fig. 1 (a) Layout of the microfluidic device. (b) Close-up of the key part of the design with patching channel, channel for compound perfusion and channel for cell loading. (c) SEM micrograph of the Y-shaped patching channel connecting extra- and intracellular channels. (d) Optical microscope image of a captured HEK cell. Highlighted in red, the cell membrane is seen to protrude a distance of about 50 μm into the patching capillary shortly after being captured. (e) A single-use injection-molded device used for the experiments. (f) Illustration of the setup used for the recordings. The polymer device is connected to two intracellular solution reservoirs and then to the pressure controller. Ag/AgCl electrodes are connected to the head stage of the amplifier. Syringes are used to transport cells by convective flow from the inlet to the patch zone and to perfuse lidocaine through one of the lateral apertures, respectively. (g) Schematic of the patch clamping recording: a cell trapped to a lateral aperture. Electrodes are located across the orifice. |
Our customized polymer chip, 50 mm in diameter and 2 mm in thickness, contained 12 Luer fittings39 used to interface the microfluidic network. A single-use injection-molded device is shown in Fig. 1e. Prior to experiments, the device was primed with the electrolyte solutions. First, the capillary was filled with an intracellular solution, and then, the cell carrier channel was filled with an extracellular solution. The recording channel was connected to an external pressure controller, and reservoirs with the intracellular solution were employed to prevent the formation of air bubbles when applying suction. Ag/AgCl electrodes positioned across the recording channel and connected to the amplifier ensured electrical connection, as shown in Fig. 1f–g.
Cells were introduced into the inlet port and transported by the convective flow induced by a syringe. Before trapping one of the cells, a slight positive pressure was applied to the patching channel to prevent contamination of the aperture. Conveniently, one of the unused patching capillaries could be used for perfusion of lidocaine by using a syringe connected to the corresponding Luer port. During perfusion, the estimated average flow velocity in the carrier channel was approximately 1 mm s−1, and the average flow rate in the perfusion channel was estimated to be about 20% higher than the flow rate in the carrier channel. The lateral flow was applied for less than 10 seconds. A COMSOL® simulation, shown in Fig. 2, supports the premise that 100% of the lidocaine in solution was successfully delivered to the cell. The simulation considered a worst case scenario of lidocaine diffusion when a high lidocaine concentration of 1 mM was perfused.
![]() | ||
Fig. 2 Lidocaine (1 mM) delivery to the cell being patched, simulated in COMSOL®. A lateral flow (in red) perfused into the flow in the carrier channel (in blue). The simulation accounted for lidocaine diffusion in the extracellular solution. A sectional view across the channel shows how the concentration corresponds to 100% in the proximity of the cell. |
As previously reported,37 injection molding enables high replication accuracy of micro- and nano-features. Fig. 3a–c shows SEM micrographs of the orifice formed at the junction between the patching channel and the carrier channel of a polymer device after the three main steps of the fabrication process: etching, electroplating and molding. The uniformity of the patching channels was confirmed for several devices by the narrow range of chip resistances; the resistances were measured across the 45 μm long patching capillaries with the presence of only the electrolyte solutions. This was 9.4 ± 1.1 MΩ (mean ± standard deviation, n = 33) for type A devices and 10.5 ± 0.7 MΩ (n = 14) for type B devices. The devices with semicircular profile (type B) had a slightly larger chip resistance as expected from their smaller cross-sectional area. Both devices showed comparable resistance variation. Since the chip resistance is mainly a function of the capillary geometry, its variation is a good indication of variations in channel geometry.40 The chip resistances are comparable to those reported by Ionescu-Zanetti17 and Chen;38 however, the variation is significantly smaller, indicating a better reproducibility of the present method. However, the chip resistance is significantly higher than the typical 2–4 MΩ access resistance reported for planar patch clamping devices where the holes are tapered on the back side.41 In addition, the parasitic chip capacitance, that ideally should be as low as possible, was monitored. This was 11.5 ± 0.5 pF for both devices A and B. This capacitance was presumably dominated by the parasitic capacitance of external wiring, as a much lower estimated capacitance of about 10 fF can be obtained for the chip itself by employing a simple model of a pair of parallel wires, being effectively 150 μm long segments of the 50 μm deep micro-channels, separated by a gap of 100 μm (the length of the Y-shaped patching channel), and the dielectric constant of TOPAS (εr = 2.35). For comparison, the chip capacitances for the measurement plates for the QPatch™ are 55 ± 5 pF.
![]() | ||
Fig. 3 (a–c) SEM micrographs of the patching orifice after the main three steps of the fabrication process for type A devices: silicon master, nickel insert, and replicated polymer part (from left to right). (d) 3D AFM image of the patching orifice (type A). (e) 3D AFM image of the sidewall of the patching capillary (type A). (f) 3D AFM image of the semicircular patching orifice (type B). (g) 3D AFM image of the sidewall of the semicircular patching capillary (type B). |
In order to test the devices, whole-cell recordings were performed with human embryonic kidney (HEK) cells. Of 47 cells captured, 19 cells (40% of the total) allowed for whole-cell access with seal resistance of at least 100 MΩ, and of those, 7 (15%) were gigaseals. A success rate of 15% is thus attributed to our experiments. The rest of the cells showed either no whole-cell achievement or seal resistance below 100 MΩ. More precisely, tests were carried out on devices both having patching capillaries with rectangular (type A, n = 33) and semicircular (type B, n = 14) profiles. As shown in Table 1, both device types showed very similar sealing capability and ability to achieve the whole-cell configuration in a range of seal resistances. Both profiles also have similar gigaseal frequencies. The exact confidence interval for gigaseal frequency was calculated using the Clopper–Pearson method42 and is given at a confidence level of 90%. The confidence intervals for gigaseal frequency were also comparable for the two profiles. During cell experiments, it was noted that high-resistance seals could only be achieved on longer channels, and shorter channels were also tested with cells, but they all failed to deliver quality seals (data not reported). Importantly, for the 45 μm long channels, each cell showing a seal resistance above 250 MΩ also allowed for whole-cell recordings. It was observed that tight seals appeared gradually after the whole-cell configuration was established, eventually reaching gigaseal during a period of 10–20 minutes. Achievement of the whole-cell configuration was verified by activation of sodium currents through a depolarization of the cell membrane. Seals had an average lifetime of 30 minutes. Furthermore, it was also observed that whole-cell configuration with rupture of the patched cell membrane was obtained spontaneously upon cell capture without an intermediate cell-attached state, where electrical access to the interior of the cell goes through the patched piece of cell membrane inside the pipette in traditional patch clamping experiments. This seems to differ from other systems, both those using traditional pipettes and planar chips, where a drop in resistance when breaking the cell membrane is normally observed.
Device | Total of cells tested | R < 100 MΩ | R > 100 MΩ | R > 250 MΩ | R > 1000 MΩ | Gigaseal frequency | Confidence interval for gigaseal |
---|---|---|---|---|---|---|---|
Rectangular profile (A) | 33 | 14 (9) | 20 (13) | 8 (8) | 5 (5) | 15% | 6%–29% |
Semicircular profile (B) | 14 | 6 (4) | 8 (6) | 3 (3) | 2 (2) | 14% | 3%–38% |
In the literature, numerous discussions regarding the importance of various device parameters for high-quality seal formation exist, such as the aperture diameter,43 length of membrane protrusion into the microhole,44 smooth edges of the microholes,45 and hydrophilicity of the substrate.43 Additionally, these evaluations were done with different substrate materials and capturing geometries and none with thermoplastic polymers. Although the proposed technology allows for a valid device trade-off strategy based on shorter patching channels having lower seal and series resistance, but relying more on leak subtraction (see ESI†), our experiments indicate that long and smooth patching channels and apertures of approximately 4 μm2 with clean and smooth orifice edges allow for good seal quality. Furthermore, these parameters contribute to an increase in total contact area between the cell membrane and the polymer surface, and therefore, the effective contact area seems to be the most critical prerequisite to look at. To support this hypothesis, the topography of the orifices and the inside of the patching capillaries were thoroughly characterized by atomic force microscopy (AFM). AFM data from channels with rectangular and semicircular profiles are compared in Table 2 and discussed here. The profiles across the patching channels revealed the presence of scallops with an average periodicity of approximately 180 nm for type A profile chips with scallops generated by the Bosch-type reactive ion etching process employed for master mold formation.46 The amplitude and periodicity of the scallops defines the waviness of the surface and dominates the calculated roughness value. To overcome the dominating effect of the scallops and evaluate the surface roughness Rq between the scallops, images were filtered using a cutoff wavelength shorter than the average periodicity between the scallops. After filtering, the surface roughness of the sidewalls of the rectangular and semicircular channels are comparable and in the range of 1 to 2 nm. Therefore, the surface roughness Rq of the sidewall for both samples seems to be determined primarily by the smoothening steps that were the same for both. In general, the inside of the patching channel, including the bottom of the channel and the foil, exhibits very low surface roughness Rq of 1 to 2 nm, and it is also here, between the cell membrane and the polymer, that the seal is assumed to form. These findings suggest that the surface roughness inside the patching channel is a second prerequisite for good seal formation. In contrast, the waviness originating from the scallops does not seem to influence the seal quality. The cell membrane, which has a thickness of only 5–10 nm, is likely to conform to the shape of the surface waviness, which has a periodicity much larger than the thickness of the cell membrane. This effect has been studied before for pipette tip roughness.47 Hence, one can hypothesize that low surface roughness in addition to the patch channel length increases the effective contact area between the cell membrane and the polymer surface and, thus, facilitates the formation of tight seals. The shape of the cross section, either rectangular or semicircular, did not influence the capability of seal formation.
Device | Rq (nm) | Rq (nm) | Rq (nm) |
---|---|---|---|
Bottom of patching capillary | Sidewall patching capillary | Foil | |
a The image was line-wise corrected by a second-order least mean squares fit in order to remove the (semicircular) shape of the profile. b The surface roughness Rq was calculated from an image after it was filtered so that only shortwave profile components with a cutoff wavelength of λ = 0.18 μm are included. The filtering was done to avoid the influence of the Bosch process-induced scallops on the calculated surface roughness (only relevant for profile A devices). | |||
Rectangular profile | 2.0 ± 0.3 | 6.5 ± 1.3a | 1.2 ± 0.2 |
(2.4 ± 0.4)b | |||
Semicircular profile | 0.7 ± 0.2 | 1.0 ± 0.2a | 1.2 ± 0.2 |
(0.8 ± 0.1)b |
Electrophysiological cell experiments were designed to record activation and inactivation currents in whole-cell voltage clamp mode from voltage-gated sodium ion channels Nav1.7. Sodium channels play an essential role for the generation of action potentials in excitable cells, and they have a key role in pain perception.48 In order to benchmark the performance of the polymer devices, electrophysiology data taken using the devices were compared with data taken using the commercially available QPatch™ system. The cells, the electrolyte solutions and the compounds used for the QPatch™ recordings and for the polymer chip recordings were exactly the same. Raw data were collected and subsequently leak subtracted in order to compensate for the capacitance; Fig. 4a shows an example of a family of sodium activation currents. A +10 mV square pulse of 10 ms from −90 mV prior to depolarization was used to calculate the leak current that was subtracted from the recorded data.49 A description of leak subtraction is provided in the ESI.† Automated P/n leak subtraction50 was employed in the QPatch™ system. Activation currents were obtained after the start of depolarization pulses from −90 mV to 70 mV, while the holding potential was kept at −90 mV for all the experiments. Steady-state inactivation currents were investigated at +10 mV after conditioning potentials ranging from −90 mV to 70 mV were applied for 1000 ms. The resulting current amplitude represents the portion of sodium channels in the activated and inactivated states, respectively. For the polymer devices, the activation threshold was found to be between −50 and −40 mV, and the maximal current amplitude was obtained for voltages between −30 and −20 mV. At positive potentials, the current amplitude gradually decreased as the electrochemical driving force disappeared. The IV curves show good agreement with those obtained using the QPatch™ system. However, IV curves from the polymer device had a tendency to shift towards more negative potentials. Fig. 4b shows two representative IV curves for both systems. The inactivation graph in Fig. 4d shows the gradual transition from a state where the sodium channels were predominantly closed to a state where they were predominantly inactivated, above −50 mV for the polymer devices. The cutoff at −90 mV does not show at which potential the transition starts but is presumed to be at about −100 mV. As mentioned for the activation currents, the IV relationships obtained for the polymer devices, when compared to IVs from the QPatch™ system, had the tendency to shift towards more negative potentials. This shift was attributed to voltage drops across the higher access resistance of the polymer device, which was not compensated for, while series resistance compensation was employed for the QPatch™ recordings. Series resistance compensation for the used HEKA amplifier is tailored to the electrical properties of glass pipettes. Due to their different impedance properties, we did not expect the usual dynamic series resistance compensation to work properly for the polymer devices reported on here. Hence, in order to avoid amplifier oscillations, we did not employ the series resistance compensation for the polymer chip. An additional explanation for the shape of the IV relationship could also be rooted in the expected difference in the mechanical state of the cell membrane in the polymer chip and the QPatch™. Such effects are known to cause shifting of the IV relationships.51 As we shall see in the following, this difference did not influence the pharmacological studies.
![]() | ||
Fig. 4 (a) A family of Na+ activation currents in response to depolarization pulses from −90 mV to 70 mV, measured on the polymer device. The protocol used for determining both activation and inactivation is also shown. The membrane potential was held at a holding potential of −90 mV, subsequently shifted to potentials ranging from −90 to 70 mV for 1000 ms, and finally to 10 mV. To the left, the raw current responses are shown, and to the right, the data after leak subtraction. (b) The resulting IV relationship for peak Na+ channel currents. The activation threshold was −50 mV, and the maximal current amplitude was obtained at −30 mV for the polymer device. The threshold was −30 mV with a maximum at −10 mV for the QPatch™. (c) A family of Na+ inactivation currents from the same recording upon the step to 10 mV. To the left, the raw current responses are shown, and to the right, the data after leak subtraction. (d) The inactivation graph for the Na+ channel. At potentials more negative than −80 mV, the channels were predominantly closed, whereas at potentials above −30 mV, they were predominantly inactivated for the QPatch™. For the polymer device, the channels were predominantly inactivated at potentials above −50 mV. |
Dose–response experiments were carried out to further benchmark the polymer device with a pharmacological application. Lidocaine inhibition of whole-cell sodium currents was explored in voltage-clamped mode by application of lidocaine concentrations ranging from 30 μM to 1 mM. The ability of lidocaine to bind to sodium channels is state dependent, as lidocaine binds to the sodium channel in the inactivated state only.52 To explore state dependency, the cell was depolarized twice at 0 mV for 100 ms and 20 ms, respectively, from a holding potential of −100 mV with a temporal separation of 15 ms, see Fig. 5a. Depolarization was repeated every 5 seconds. Lidocaine response was studied at the start of the second pulse after the resting interval, during which only a portion of the sodium channels was able to recover from inactivation. Fig. 5b shows a typical recording in the absence of compound where the activation peak 2 was about 80% of the size of the activation peak 1, used as a reference. When lidocaine was applied, reference peak 1 exhibited a current that was almost unchanged, while peak 2 was substantially reduced. Fig. 5c shows activation currents from the same cell before and after 100 μM lidocaine was delivered to the cell. The current–time relationship for the peak sodium currents recorded in response to the first (empty circle) and the second (full circle) depolarization is shown in Fig. 5d. Inhibition is plotted against lidocaine concentration in Fig. 5e. The amplitudes of sodium currents immediately prior to lidocaine application were set to 100%. Complete inhibition was observed at 1 mM lidocaine, while the inhibition was only 20% at 30 μM lidocaine which was also the lowest concentration applied. From Fig. 5d, it can be observed that the inhibition produced by lidocaine was reversible within 60 seconds after its perfusion was stopped. For the polymer device, the half-blocking concentration IC50 was 119 ± 11 μM. This value is in excellent agreement with the IC50 of 152 ± 11 μM found with the QPatch™. The Hill coefficients were also very similar for the two systems.
![]() | ||
Fig. 5 (a) Protocol used for determining the concentration–response relationship of lidocaine inhibition on Nav1.7 channels consisting of two depolarizations to 0 mV from a holding potential of −100 mV with an interval of 15 ms. (b) Raw current response from a recording on the polymer device before lidocaine application. The amplitude of peak 2 is about 80% of the amplitude of peak 1. (c) Raw current response for peak 1 and peak 2 before (black) and after (red) 100 μM lidocaine was applied to the cell. (d) Activation currents for Nav1.7 channels for determining the lidocaine inhibition on Nav1.7 channels. The graph shows values of the current amplitudes of peaks 1 and 2 before and after 100 μM lidocaine was applied to the cell. Peak 1 is not affected, and peak 2 is reduced by 50%. (e) Concentration–response relationships of lidocaine inhibition on Nav1.7 channels. Data were fitted with a Hill-type equation. IC50 was 119 ±11 μM (Hill coefficient n = 1) for the polymer device and 152 ±12 μM (Hill coefficient n = 0.9) for the QPatch™. |
Footnote |
† Electronic supplementary information (ESI) available: Description of the leak current subtraction applied to the raw data obtained from the polymer devices; description of the simulation for the perfusion of lidocaine from one of the lateral channels into the deep channel during the dose–response experiments; description of the fabrication process (table and illustrations); description of the challenges faced during the development of the presented polymer device (and figures). See DOI: 10.1039/c3lc50760b |
This journal is © The Royal Society of Chemistry 2013 |