Nicolas
Maïno
*,
Arnaud
Bertsch
and
Philippe
Renaud
Ecole Polytechique Fédérale de Lausanne (EPFL), Lausanne, Switzerland. E-mail: nicolas.maino@epfl.ch
First published on 11th April 2023
For over 30 years, carbon fiber microelectrodes have been the gold standard for measurements related to exocytosis and more generally to the processes taking place at the synaptic level. However, this method has a low throughput and molecules can escape detection due to the featureless nature of the planar microelectrodes it uses. Here we present a new electrochemical sensor that addresses these limitations. It is based on insulated protruding volcano-shaped tips of 2 μm in diameter housing two individually addressable microelectrodes. The sensor enables volume confined and parallelizable recordings of exocytosis from adherent cells. Exocytotic releases from PC12 cells measured by amperometry on our device have quantal size in agreement with commonly admitted values but happen on a much smaller time scale; mostly within half a millisecond. We demonstrate that this faster kinetics must involve a faster vesicle fusion mechanism and is plausibly due to perturbation of the plasma membrane brought by the topography of our sensor. This suggests that exocytosis kinetics may be manipulated by the adequate substrate geometry, which opens up promising new leads of investigation in the study of synaptic processes.
The attractiveness of the method is threefold. First it achieves unequaled temporal resolution. As an example, the presynaptic bouton of neurons from the central nervous system is home to small synaptic vesicles (SSV) that contains less than 50000 neurotransmitter molecules and give rise to release events on the sub-milliseconds timescale that could only be measured with the CFE technique so far.5 In contrast, optical methods like total internal reflection and super-resolution microscopy applied to vesicle imaging achieve acquisition periods of 1–10 ms at best.9 Secondly, amperometry directly relates the measured signal to the experimental quantity of interest (i.e., the number of molecules released) through the Faraday law:
where Q, the electrical charge in Coulomb, is determined by integrating the current i over time and relates to the number of moles detected n multiplied by z the number of electrons involved in the redox reaction and F the Faraday's constant. This considerably simplifies the experimental workflow as no calibration is needed. Lastly, amperometry of electroactive neurotransmitters (i.e., catecholamines) doesn't require labeling or mediator, which again is in contrast with optical methods involving either fluorescent antibodies,10 membrane bound fluorescent tags11,12 or fluorescent false neurotransmitters.13
Unfortunately, the outstanding performance of the CFE technique suffers from two main limitations. First, there is a chance for the released molecules to escape detection at the microelectrode by diffusion away from the electrode. While modeling using random walk has concluded that collection efficiency should be virtually 100% all across the electrode diameter,14 experimental data seems to indicate that the collection efficiency is worse than expected and leads to underestimation of the quantal size.15 Secondly, the use of CFE with micromanipulator is very work intensive and poses a true limitation on throughput.
Multi-electrode arrays (MEAs) fabricated using standard microfabrication techniques on the other hand have the advantages of allowing highly parallelizable experiments. MEAs have long been popular for electrophysiology investigations where measuring attenuated action potential across the intact cell membrane is sufficient, thereby sacrificing signal integrity for throughput. This figure of merit also applies in the scope of exocytosis investigations which have seen several implementations of MEAs dedicated to amperometric measurements of catecholamine releases.16–20 Unfortunately the use of MEAs still suffers from the same limitation as the traditional CFE technique in terms of collection efficiency. In the typical MEA configuration, cells are cultured on arrays of inlaid disk electrodes. Because of this loose interface, the molecules of interest released by the cell tend to spread as they diffuse to the electrode, which results in broadening of the signal registered or even loss of molecules.
In an attempt to reunite the merits of both approaches, we adapted volcano microelectrodes (VME) previously used for electrophysiology to the amperometric measurement of exocytotic releases. This new device was developed with the goals of (i) enabling high collection efficiency thanks to a volume confined cell/electrode interface and (ii) allowing parallelizable experiments by having cells cultured on chips with 28 sensing sites each. Furthermore, we built each sensing site to house two individually addressable electrodes instead of one to enable other measurements modality like redox cycling detection, which we use in this study to characterize the cell/VME interface. We benchmarked the performance of our devices with amperometric measurements of stimulated exocytosis from PC12 cells cultured on VMEs. Our devices record quantal sizes close to the expected value found in the literature.
Interestingly, the recorded amperometric spikes take place over a much smaller time scale compared to other reports; within a single millisecond. We hypothesize that this faster kinetics is the result of deformation of the plasma membrane brought by the topography of our sensor. Still, because the quantal size measured is preserved, the presented device is a well behaved sensor that reports the parameter of interest faithfully. Furthermore, it may be used as a novel tool to study the link between membrane mechanosensing ability and exocytosis opening new investigations opportunities.
Throughout most of this study, we used our devices in constant amperometry mode whereby the top and bottom electrodes are connected together and set to a positive potential with respect to the silver/silver chloride reference electrode, aimed at oxidizing the molecule of interest (Fig. 2A). Upon stimulation, cells undergo exocytosis and release catecholamines that diffuse to the electrode and are oxidized resulting in spikes in the amperometric trace (Fig. 2B). Spikes are then individually analyzed to retrieve their peak current, charge and full width at half maximum (FWHM) (Fig. 2C). When measuring exocytosis events, the contribution of the electrode's active surface within the cavity was found to be negligible from our simulation (see Materials and method section) probably because all catecholamine molecules were oxidized before reaching the cavity. In experiments measuring exocytosis events, the active electrode surface area can hence be approximated to the VME floor and inner wall (12.64 μm2 in total). However, the effective active electrode surface area seems to be lower as characterized by cyclic voltammetry and step-amperometry (ESI section 3 and ESI Fig. 4†) with a steady-state current equivalent to a recessed electrode of 3.14 μm2 surface area located 0.5 μm below the substrate level. Although the VME is not a recessed electrode strictly speaking because of its conductive inner wall, this result suggests that most electroactive molecules diffusing from the bulk to the electrode inside the VME during a potential controlled experiment are consumed before reaching the nanogap.
On the other hand, we also made use of the two electrodes being individually addressable to leverage a phenomenon known as electrochemical redox cycling to characterize the cell/VME interface. This detection scheme is not concerned with exocytosis but rather with the detection of electroactive molecules undergoing a reversible redox reaction and has been demonstrated down to the single-molecule detection levels.21 In the redox cycling mode, the potential of the two electrodes are set symmetrically apart from the formal potential of a redox mediator resulting in the mediator molecules undergoing oxidation and reduction up to several hundreds to thousands of times per seconds (Fig. 2D). When the two electrodes potential are scanned with an offset (Fig. 2E), the resulting current traces show anticorrelated peaks at the formal potential of the mediator molecule (Fig. 2F). We made use of this detection scheme below to assess whether molecules from the bulk can reach the confined detection volume created by the VME. In redox cycling mode, the electrode active surface area corresponds to the surface area of overlap between the floor and ceiling electrodes within the cavity (7850 μm2 corresponding to a disk 100 μm in diameter).
The wafers are diced into individual chips and a glass ring is glued on top of each chip to delimitate a culture chamber using polydimethylsiloxane (PDMS). The chromium sacrificial layer in between the gold electrodes is removed by potential-assisted wet etching as described in the ESI (section 2 and ESI Fig. 3†). We assessed the viability of cells cultured on VME after chromium removal (ESI Fig. 5†) and found a viability of 98.17 ± 0.19% (n = 10 field of views centered on different VMEs). We assessed the device yield by counting out the VMEs whose pair of electrodes were short-circuited (i.e., in physical contact) and found the yield was always over 85% with at least 24 out of 28 devices per chip operational.
Typical recording traces are presented at both time points over the initial 4 seconds of stimulation (Fig. 3A) and show strong, continuous amperometric spikes. Sample amperometric spikes from both time points are displayed over a 10 ms window (Fig. 3B). To compare exocytosis at the two time points, we averaged the spike features per cell which is more reliable than comparing pooled data from the different cells recorded from.22 The two conditions are compared through their distribution of spike feature means (mean quantal size, mean peak current, mean FWHM, Fig. 3C, D and E respectively). The distribution of quantal size at the two time points followed a log–normal distribution (two-sample Wald–Wolfowitz runs test, significance level 0.05) and can be compared qualitatively through their log–normal distributions (Fig. 3F).23 The solid lines superposed over the histogram correspond to the best fit of a normal distribution.
![]() | ||
Fig. 4 Redox cycling detection of a redox mediator diffusing inside free and cell-covered volcano microelectrodes (VME) (A) scheme depicting a cell covering a VME. A redox mediator (hexaamineruthenium) is added to the extracellular bath and subsequently detected by redox cycling within the VME. (B) Current traces recorded at a free VME compared to four other VMEs covered by PC12 cells. The solid line anodic and dashed cathodic traces corresponds to the currents registered at the top and bottom electrodes respectively as defined in Fig. 2E. (C) Comparison of the peak anodic and cathodic currents for free VMEs (n = 3) and cell-covered VMEs (n = 4). One-sided Mann–Whitney rank-sum test, *p < 0.05. |
We constructed a 3D geometry representing the VME and closed its top surface with an impermeable boundary representing the cell membrane (Fig. 5A cross-section). The bottom (1), top (2) and wall (3) domains as labeled on the cross-section represent electrode surfaces that will oxidize catecholamine molecules (see Materials and methods for detailed implementations of the electrodes boundaries). The nanogap was trimmed at a radial distance of 1.325 μm from the center of the VME as it was found that the electrode surfaces within the nanogap region past the VME wall projection had no impact on the simulated spike. The vesicle was modeled as a spherical domain with a fixed concentration of 0.6 M of a molecule with diffusion coefficient of 4.1 × 10−8 cm2 s−1 inside the vesicle and 6 × 10−6 cm2 s−1 everywhere else, which are typical of catecholamines in the vesicle matrix26 and aqueous medium respectively.27 The quantal size was taken from our pooled experimental results at DIC3 (Fig. 3F) as 125 zeptomole or 75814 molecules. The position of the vesicle was varied along the radial axis but kept at a constant height, with its hemisphere across the no-flux boundary modeling the cell membrane.
We first looked to ascertain whether or not the VME inner wall is conductive and participating in the electrode active surface area. Previous electrochemical impedance spectroscopy data28 and scanning electron microscope (SEM) images of VME cross-section29 have demonstrated that the inner wall of the VME is covered with metal redeposited by secondary ion sputtering during fabrication. However in the current study the thickness of metal etched that generated secondary ion sputtering is significantly thinner (50 nm gold + 10 nm titanium). It was therefore necessary to assess whether or not the inner part of the VME was part of the electrode active surface area to accurately simulate the diffusion processes.
We hence simulated amperometric spikes registered in the case of a conductive or insulating wall (Fig. 5A; domain 3 is set as an electrode or insulating boundary respectively) while keeping the vesicle at an arbitrary distance of 200 nm from the wall. The results are in favor of a conductive wall with the corresponding spike closely matching the experimental average spike (Fig. 5B). To further support this outcome we performed platinum electrodeposition over both electrodes of a VME simultaneously with the hypothesis that in the case of conductive walls the electrodeposited platinum would be localized both on the VME floor and walls. The SEM micrographs show a regular layer of platinum along the VME inner wall (Fig. 5C) providing further evidence of the VME inner wall being conductive.
Having established the electrode's active surface, we simulated the amperometric spikes resulting from vesicles at increasing distance from the wall (Fig. 5D). As the vesicle is positioned close to the wall, spikes become taller and narrower. In order to compare the experimental and simulated data, we shaded on the graph two areas that correspond to the range of the pooled experimental data observed (taken as the experimental mean ± 2 × standard deviation from Fig. 3D–E). Put in another way, any simulated data point within the shaded area of corresponding color is within two standard deviations of the experimental mean.
Secondly, our fabrication process is largely independent of residual stress in the different layers. This is achieved through the incorporation of SiO2 pillars in between the two electrodes that ensures that the electrodes remain separated after removal of the chromium layer (Fig. 1B). This fabrication approach and its advantages is discussed further in the ESI (section 5†).
In our data, the most striking features of this development are the increased quantal size and peak current after three days (Fig. 3C and D; p < 0.05 one-sided Mann–Whitney rank-sum test). In both cases, the FWHM observed were still well below the milliseconds mark (424 ± 1.9 μs DIC1 and 451 ± 91 μs DIC3; mean ± SEM), which is considerably smaller than the few/tens of milliseconds commonly reported in the literature for PC12 cells.30 Nonetheless, the quantal size measured were in line with the admitted range of 100000 molecules30 (30
669 ± 3341 catecholamine molecules per spike at DIC1 and 65
580 ± 8568 at DIC3; mean ± SEM) The conservation of a typical quantal size despite much shorter events is explained by the large peak currents observed (21.3 ± 1.1 pA DIC1 and 57.0 ± 15.8 pA DIC3; mean ± SEM) as compared to typical values around 10 pA obtained with a conventional CFE.
A plausible explanation of the reduced quantal size and peak current at DIC1, shortly after cell plating, is the lack of interactions with the extracellular environment. Loss of cell–cell and cell–substrate interactions have been linked to impairment of exocytosis for example in mice pancreatic β-cells and neocortex/brainstem neurons through impairment of NCAM (Neural Cell Adhesion Molecules) and α-neurexin respectively.4,31 In a wider context, integrin interactions with the extracellular matrix are pivotal in localizing exocytosis for cell migration and recruitment of specific synaptobrevins during neuritogenesis.32,33 While there is no similar report for PC12 cells specifically to our knowledge, it seems plausible that the decreased quantal size arises from the transient disruption of intra/extracellular communications shortly after passaging.
The large peak current and small duration of the spikes we measured are not an effect of using microelectrodes per se as measurements on MEAs by others also reported FWHM clearly above the milliseconds mark at inlaid microelectrodes16 and recessed disk microelectrodes.18,20,34 This however raises an interesting possible improvement of the micro-electrode array described in this study. Including standard inlaid micro-electrodes among the array of VMEs would provide a readily available control that can be used as reference against which the exocytosis features measured at the VME electrode could be compared.
The data in that experiment were obtained from 7 cells for a total of 5722 spikes analyzed. The use of the VME array made the experimental workflow already significantly faster compared to an approach using a carbon fiber electrode since there was no need to engage the electrode on the cells. However, the use of an off-chip single-channel amplifier (Axopatch 200B) still severely limits the experimental throughput. This issue should be addressed in a future study integrating the VME presented on top of a recording circuitry (e.g., CMOS) to take the last step toward high-throughput experiments.
It could be argued that the redox mediator finding its way in one direction supposes that catecholamine molecules could escape the confined detection volume by diffusing the same way but in the opposite direction during exocytosis. However, since the electrode oxidizes molecules almost instantly, the concentration of catecholamine within the VME remains negligible at all times; i.e., catecholamines do not accumulate at the electrode. Consequently, the flux of catecholamine escaping the VME through a nanometric slits necessarily has to be small in comparison to the unobstructed flux of molecules to the much larger electrodes of micrometric size.
It could be hypothesized that the VME conical geometry forms a funnel from the release site to the electrode and that the confined volume of the VME results in less dilution and hence a steeper concentration gradient from the vesicle to the electrode. This hypothesis is however in disagreement with experimental results using recessed cavity carbon electrodes which register amperometric spikes of similar duration to those from a conventional CFE.15 A significant difference of the VME however lies in the fact that the inner wall of the recess is itself conductive as evidenced by our simulated data (Fig. 5B) and in situ electrodeposition of platinum (Fig. 5C).
Under the assumption of a conductive inner wall of the VME, exocytotic release at varying distances from the wall results in a distribution of spikes of equal charge but varying peak currents and FWHMs (Fig. 5E) that captures well the variability of the experimental data (Fig. 5F). At large vesicle distance from the wall we see the experimental and simulated data converge to similar values of FWHM (436 μs) and peak current (35 pA). On the other hand, simulated data at vesicle distance from the wall below 500 nm predict faster, taller spikes, which are not found within the experimental data. Instead experimental data reach a maximum/minimum in peak current and FWHM respectively that could be attributed to diffusion time becoming small compared to the opening of the vesicle fusion pore. In absence of an appropriate modeling of the fusion pore opening, this hypothesis remains to be confirmed.
Additionally, the present simulation does not allow us to conclude whether exocytosis takes place preferentially from membrane portions close to the VME wall or rather at the center of the VME. In this scope, TEM imaging of the cell membrane conformation within the VME would prove very valuable and perhaps reveal preferential vesicle localization close to or far away from the VME wall rim. Still, the conclusion we draw from simulating the impact of vesicle position is twofold. First, simulated diffusion times within the VME are within the experimental bounds and hence do not challenge the plausibility of our data. Second, on top of diffusion being fast (sub-millisecond timescale), the kinetic of vesicle fusion needs to be faster as well since it usually takes place over a few milliseconds.7
A plausible hypothesis to reconcile our observations with admitted vesicle fusion kinetics could be the impact of membrane curvature and tension arising from the sharp VME wall (150 nm thick). It was observed by several investigators that hypotonic conditions result in faster, more frequent exocytosis while the converse is true for hypertonic conditions.7,42 The resulting spikes in hypotonic conditions are shorter in duration of both the rising and decaying phase of the spikes, which are associated with rate of opening of the fusion pore and diffusion of catecholamine out of the vesicle matrix7 yet with conserved quantal size.43 Our experiments did not involve osmolarity manipulation yet a possible role of the plasma membrane tension in the fast exocytosis kinetics we observe is plausible since it was demonstrated that membrane deformation brought by nanotopography induces membrane tension.44 Although the cell membrane was shown to be able to accommodate microscale vertical features from its substrate,45 the more subtle impact of local curvature on the nanoscale is believed to develop significant stress in the cell membrane even leading to membrane disruption.46 The fact that nanotopography can induce enough hoop stress to cause membrane failure whereas there was no such report in the classical hypotonic experiment suggests that the magnitude of the membrane stress induced by nanotopography is greater. Since hypotonic conditions did shorten exocytosis timescale by roughly 30%,7 it seems reasonable that even stronger membrane stress caused by nanotopography might lead to a more drastic shortening of the exocytosis duration; down to fractions of milliseconds as found in our experiments.
An ensuing argument is that the membrane deformation might alter the molecular composition of the membrane and/or cytoskeleton in the vicinity of the VME wall rim. This is supported by a vast literature on protein localization to curved membrane domains like α-synuclein,47 complexin48 and BAR domains-containing proteins in general.49 In particular, nanotopography was demonstrated to induce the recruitment of FBP17, a membrane curvature sensing protein, leading to filamentous actin assembly (F-actin) through action of the neuronal Wiskott–Aldrich syndrome protein (N-WASP) in U2OS cells.50 Interestingly, independent findings revealed the role of F-actin polymerization, also mediated by N-WASP, in providing enough membrane surface tension to enable the merging of vesicles with the chromaffin cell membrane.51,52 Taken together the impact of nanotopography on membrane tension and cytoskeleton rearrangement substantiates the hypothesis that the VME wall sharpness alters exocytosis kinetics. This hypothesis gives away some very interesting investigation prospects since to our knowledge there only exists one report of nanotopography impact on lateral vesicle movement53 but none about exocytosis itself.
A limitation of the current study is the small number of cells the data were collected from. Although our VME arrays like other MEAs allow a high number of simultaneous sensing sites, our experiment yield was limited by the random pairing of VME and cells and the use of a single channel amplifier. Others have addressed the former by defining zone of preferential attachment for cells around the microelectrodes,18,20 which could be implemented in the future together with other approaches like cell specific dielectrophoretic patterning.56 Finally, CMOS integration as implemented by others20 are a common strategy to parallelize recordings thereby increasing throughput.
The short time scale of the spikes observed led us to build a set of hypotheses relying on the increased membrane tension and cytoskeleton reshaping plausibly brought by the VME wall having sub-micron thickness. There exists a vast literature about the impact of nanostructures on cell mechanisms like differentiation,57 migration58 and signaling59 yet their impact on exocytosis remains to be explored. Accordingly, VMEs stand as an interesting platform to explore this question. A straightforward way to test our hypothesis could be to use pharmacological or genetic manipulations to perturb the hypothetical actin polymerization over the VME rim using latrunculin A or Actb knockout respectively. On the other hand, varying the VME geometry (e.g., wall thickness, height, diameter) could also be used to challenge or refine this hypothesis. Our current data offer a novel view of exocytotic behavior under membrane deformation and could foster further investigations of the impact of nanotopography on synaptic mechanisms and cellular mechanosensing.
A glass O-ring was glued on top of the individual chips using PDMS and cured overnight at 60 °C in a convection oven. The removal of the chromium sacrificial layer was carried out at this point by potential assisted wet etching followed by device yield assessment as detailed in the ESI (section 2 and ESI Fig. 3†). The device electrochemical response was characterized in cyclic voltammetry and step chronoamperometry (section 3 and ESI Fig. 4†).
Pheochromocytoma 12 rat cells were obtained from the European Collection of Cell Cultures. Cells from passage 10 to 15 were used. The catecholamine content of PC12 cells was determined by fluorometric measurement following reaction of catecholamines with 3-hydroxyphenyl boronic acid.60 The detailed lysis and assay protocol are detailed in the ESI (section 4 and ESI Fig. 6†).
During cell culture the cells were kept in RPMi-1640 supplemented with Glutamax (61870036, ThermoFisher), 10% heat-inactivated donor equine serum (26050070, ThermoFisher), 5% fetal bovine serum (F9665, Merck) and 0.4% penicillin/streptomycin (P4333, Merck) solution within a 37 °C incubator under 7% CO2 and 100% humidity atmosphere. Before plating on the devices, a 80% confluent culture was collected by trypsinization (1084440001, Merck) for 5 mn at 37 °C followed by mechanical dislodgement by repetitive pipette dispensing over the cell. Cells were centrifuged 2 mn at 0.3 RCF and resuspended in culture medium after the supernatant was discarded. We plated 150000 cells per chip (0.95 cm2) and conducted experiments on the third day after passaging.
In another experiment we cultured Human Embryonic Kidney cells on our chip in a similar way except for the medium (DMEM supplemented with Glutamax; 10566016 ThermoFisher) and no equine donor serum but 10% fetal bovine serum. On the third day after passaging we assessed cell viability by calcein-AM/ethidium homodimer-1 assay (L3224; ThermoFisher) according to the manufacturer protocol. We analyzed 10 fields of view of 610 × 460 μm centered on a single VME. Using the CellProfiler (CellProfiler™, Broad Institute) methods “identifyPrimaryObjects” we identified, segmented and counted individual cells to obtain the ratio of live cells to total cells in a given field of view (ESI Fig. 5†).
The cells were stimulated using an elevated potassium solution (same as recording buffer but with KCl elevated to 125 mM and NaCl reduced to 5.5 mM dispensed by a Nemesys Base120 syringe pump (Cetoni, Germany). The tubing outlet connecting to the elevated potassium solution syringe was positioned 3 mm above the MEA and dispensed at a rate of 20 μL s−1 while another tubing positioned 6 mm above the MEA withdrew an equal volume at an equal flow rate. After stimulation the culture chamber was exchanged in a similar way using a third syringe filled with standard recording solution and the next recording (typically from another cell) was started after a 5 mn break.
![]() | (1) |
In eqn (1), J is the flux of molecules, c the molecule concentration at the electrode surface in mol m−3, k the heterogeneous reaction rate at the formal potential was set to 0.025 cm s−1,63α the symmetry factor was set to 0.5, z the number of electrons involved in the reaction, F is the faraday constant, Θ the overpotential was set to 0.4 volt (i.e., 0.6 volt for the reduction of dopamine on a gold electrode), R is the gas constant, T the temperature was set to 298.15 Kelvin. Current at the electrodes was obtained by integrating eqn 1 over all surfaces defined as electrodes. Plotting the resulting current against time yields the simulated amperometric spike. The release of catecholamine from the vesicle was simplified to an instantaneous release and the quantal size was taken from the experimental results as 125 zeptomole localized within a vesicle of diameter determined by a fixed concentration of 0.6 M. The mesh elements were taken as quadrilaterals. Meshing of the geometry was carefully optimized with refinements along all electrodes surfaces, corners and edges. We performed a mesh refinement sweep until no change of the spike features were observed leading to meshes with about 500000 elements.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2an01779b |
This journal is © The Royal Society of Chemistry 2023 |