# Electrochemical measurement of serotonin by Au-CNT electrodes fabricated on microporous cell culture membranes

Sep 7, 2020

### Characterization of carbon nanotube (CNT) coatings

Before full assembly and testing of the 3D-printed platform, SEM analysis was performed to evaluate the nanostructure of the CNT film on a Au-coated porous PETE membrane (Fig. 2). Both bare PETE and Au-PETE clearly showed 1-µm pores in the membrane, demonstrating that the thin 100-nm Au coating does not block the pores (Fig. 2a, b). The Au coating added a slight granular texture to the flat polymer surface (Fig. 2b). Drop-casting a CNT film on the Au-coated PETE (12.5 µL of the described single-walled CNT [SWCNT] solution) (Fig. 2c, d) produced a thick, dense mesh structure with nanoporosity that would be expected to provide a significantly higher surface area than bare Au. The width of the individual SWCNTs appears to be ~20–50 nm, with lengths in the micron range, demonstrating very-high-aspect-ratio nanostructures that contribute to the electroactive surface. Defects in the layer, including impurities and holes, can potentially act as added binding sites for 5-HT or other molecules40. The CNT-film morphology seen here is expected to be highly electroactive. The electroactivity of these CNT films was assessed across a range of drop-cast volumes on both microdisk and membrane Au electrode substrates (Supplementary Fig. S3). From this, we continued testing 2 µL CNT coatings on microdisk electrodes, and membrane electrodes were tested with the thin 2 µL film and the thick 12.5 µL film (Supplementary Fig. S4).

Figure 3a shows a comparison of each electrode type, with and without CNT coatings, for their cyclic voltammetric (CV) current response to 2 mM ferrocene dimethanol (FDM). The anodic and cathodic peak voltages were consistent between the Au and Au-CNT microdisk electrodes (Epa ~ 0.29 V, Epc ~0.23 V), but shifted left when measured at the Au membrane electrodes (Epa ~0.22 V, Epc ~0.16 V) and continued to shift with increased CNT coating on the membrane (Epa values: 0.2 V and 0.19 V for 2 µL and 12.5 µL, respectively). Nernstian behavior is indicated by near-ideal anodic and cathodic peak separation (Epa-Epc ~0.060–0.068 V, where the ideal value is 0.059 V for a one-electron-transfer reaction) and peak current ratio (Ipa/Ipc ~0.81–1.16, where the ideal ratio is ~1 for a perfectly reversible reaction41). The current peaks (Ipa and Ipc) increased due to the CNT coating in both microdisk and membrane electrodes, where 12.5 µL of CNT produced a higher current than 2 µL of CNT on membrane electrodes.

The scan-rate dependence was assessed across uncoated and CNT-coated Au membrane electrodes (Fig. 3b, c) to characterize the mode of FDM detection and the influence of the CNT-coating thickness. The Randles–Ševčik Eq. (1) describes the linear correlation between the peak current (Ipa) and the square root of the scan rate ((sqrt nu)) in diffusion-limited reversible systems41.

$${mathrm{Ipa}} = 0.446,nFACsqrt {frac{{nFDnu }}{{RT}}},$$

(1)

where n is the number of electrons transferred, F is Faraday’s constant, A is the electrode surface area, C is the solute concentration, D is solute diffusion constant, ({it{upnu }}) is the scan rate, R is the ideal gas constant, and T is the temperature. This linear correlation was seen in Fig. 3b for each electrode, confirming that the ferrocene redox reaction was diffusion limited and reversible. The current response over the scan rate takes the form of the following linear equations: Au membrane: ({mathrm{Ipa}} = 5.52sqrt nu + 0.52) (R2 = 0.9994), Au-CNT (2 µL) membrane: ({mathrm{Ipa}} = 7.96sqrt nu + 2.89) (R2 = 0.9999), Au–CNT (12.5 µL) membrane: ({mathrm{Ipa}} = 7.84sqrt nu + 0.85) (R2 = 0.9998).

As described by the Randles–Ševčik Eq. (1), the slope of the line is proportional to the effective surface area (ESA) of the electrode42,43. It can be seen that the presence of the CNT coating on the Au membrane electrode increased this slope, thereby showing an increased ESA. However, no difference can be seen between the 2 µL and 12.5 µL coatings, except when a 1-min accumulation time was applied between each CV cycle. By varying the accumulation time between CV cycles during 2 mM FDM detection (Supplementary Fig. S5), it was seen that ferrocene molecules needed ≥15 s to fully diffuse into the 12.5 µL CNT film to saturate the signal. Indeed, applying a 1-min accumulation time during the scan-rate sweep resulted in a linear relation with an increased slope (({mathrm{Ipa}} = 16.5sqrt {upnu} – 31.8), R2 = 0.9816), demonstrating a higher ESA than that of the 2 µL film (Fig. 3b). However, these data better fit an adsorption-controlled process where Ipa varies linearly with scan rate (({mathrm{Ipa}} = 1.075{upnu} + 16.5), R2 = 0.9994). The transition from diffusion-controlled to adsorption-controlled behavior indicated the increased capacity for molecular binding and retention at the 12.5 µL CNT electrode, as has been shown to occur for other electroactive films44.

### 5-HT detection at CNT-modified membrane electrodes

The CNT-coated membrane electrodes were then characterized for their sensitivity to 5-HT, although it was determined that the adsorption-limited nature of 5-HT electrochemical detection20,21 creates a time dependency in electrode sensitivity. Figure 4 demonstrates the time-dependent nature of Au-CNT membrane electrode detection of static 5-HT solutions at constant concentrations, without the input of stirring or heating. Over increasing accumulation times between CV cycles, Ipa increased until saturation was achieved, indicating that 5-HT molecules adsorb on the electrode over time. As shown in Fig. 4a, b, respectively, membrane electrodes coated with 2 µL CNTs saturated after ~10 min, regardless of the 5-HT concentration, while 12.5 µL CNT-coated electrodes saturated after ~2 h. This trend can be modeled as a first-order system response over accumulation time tacc:

$$I_{pa}left( {t_{{mathrm{acc}}}} right) = Aleft( {1 – e^{ – frac{{t_{{mathrm{acc}}}}}{tau }}} right),$$

(2)

where A is the maximum Ipa signal achieved at steady state and τ is the time constant of the curve, equal to the time required to reach 63.2% of the steady-state value45. Fitting this equation to the data in Fig. 4a, b allows estimation of A and τ for each electrode: Au-CNT (2 µL) membrane, [5-HT] = 1, 5, 10 µM: A = 0.47, 1.06, 1.44 µA and τ = 2.76, 2.38, 1.58 (R2 = 0.834, 0.969, 0.965); Au–CNT (12.5 µL) membrane, [5-HT] = 5 µM: A = 20.3 µA and τ = 52.58 (R2 = 0.990). These signal saturation rate constants are approximately similar for the Au-CNT 2 µL electrodes, although an increase in 5-HT concentration correlates with higher A values and smaller τ values, resulting in saturation at higher signals over ~10 min. These rate constants are increased by more than one order of magnitude for the Au-CNT 12.5 µL electrode, indicating saturation at a significantly higher signal over ~2 h. Considering the relation of this process to an RC circuit, τ can be approximated to equal the product of the resistance and capacitance45. It stands to reason that the increase in τ associated with the Au-CNT 12.5 µL electrode is due to an increased capacitance for 5-HT binding within the thicker CNT film. Further, if we compare the linear range of the two Au-CNT electrodes, the 12.5 µL CNT electrode has a linear range of ~0–30 min accumulation time, a 15 times increase over the ~0–2-min linear range of the 2 µL CNT electrode. However, when comparing 5 µM 5-HT detection at both electrodes, the slopes of their linear ranges are found to be approximately the same (0.2 µA/min and 0.3 µA/min for the 2 µL and 12.5 µL CNT-coated electrodes, respectively). The slopes of the signal over time may be the same for these films due to a constant binding rate of 5-HT on the SWCNTs, which would control the kinetics of accumulation. However, a higher overall signal can be achieved at the thicker film because there are more binding sites available for 5-HT.

To support the theory introduced in Fig. 4a, b that 5-HT detection is adsorption controlled, the scan-rate dependence was assessed by measurement of 500 nM 5-HT at the Au-CNT (2 µL) membrane electrode (Fig. 4c, d). In this case, Ipa varied linearly with the scan rate according to the following equation: ({mathrm{Ipa}} = 0.004nu + 0.092) (R2 = 0.9928). This follows the standard Eq. (3), which describes that adsorption-controlled reactions are linear with respect to Ipa and v:

$${mathrm{Ipa}} = frac{{n^2F^2}}{{4RT}},nu ,Gamma _{eq}.$$

(3)

In this scheme, Ipa is related to both the scan rate and the equilibrium surface coverage (Gamma _{eq}), which is a function of the 5-HT concentration and electrode ESA.

Concentration-dependent detection of 5-HT was then compared using both Au-CNT membrane electrodes, as shown in Fig. 5. Compared with Au membrane electrodes, which showed a broad CV peak and very low current response to 5-HT, CNT coatings increased peak sharpness and height while maintaining an Epa of ~0.22–0.23 V (Fig. 5a). The sensitivity of each Au-CNT membrane electrode was evaluated in Fig. 5b, c, in which the 2 µL CNT electrode was used to detect higher 5-HT concentrations (500 nM–100 µM) with a shorter accumulation time (10 min), and the 12.5 µL CNT electrode was used to detect lower 5-HT concentrations (10 nM–1 µM) with a longer accumulation time (2 h). Figure 5b shows that the 2 µL-coated electrode attained a linear range of 0.5–10 µM, where the slope of the linear region denotes sensitivity: 0.6 µA/µM (R2 = 0.9976). The resolution of detection can be calculated by 3*σ (σ: standard deviation of the lowest concentration): 3*0.0068 µA = 0.024 µA. From this, the limit of detection (LOD) can be calculated as resolution/sensitivity = 30 nM. The limit of quantitation (LOQ) can be calculated as 10*σ/sensitivity = 100 nM. In comparison, Fig. 5c shows that the 12.5 µL-coated electrode attained a linear range of 0.1–1 µM and a sensitivity of 4.5 µA/µM (R2 = 0.9993). The resolution calculated as 3*σ is 3*0.046 µA = 0.138 µA, resulting in an LOD of 30 nM and LOQ of 100 nM. Per these calculations, both electrodes achieve the same LOD and LOQ; however, in practice, the Au-CNT 2 µL electrode does not produce noticeable peaks below ~500 nM 5-HT. The sensitivity of the Au-CNT 12.5 µL electrode is 7.5× that of the Au-CNT 2 µL electrode when using a 2-h accumulation time (12× longer than 10 min for the 2 µL electrode). This demonstrates that thicker CNT films can be used to measure low 5-HT concentrations with low time resolution, while thinner CNT films are useful for high-concentration and high-time-resolution applications. However, the low-concentration measurements in Fig. 5c vary more in peak height and voltage than the high-concentration measurements, as indicated by large error bars, and so it may be difficult to distinguish very low concentrations.

### Dynamic 5-HT monitoring

5-HT is released from ECCs via vesicle exocytosis, resulting in a burst of 5-HT near the site of release and then diffusion to surrounding areas24. To demonstrate the ability of our device to monitor this type of dynamic release, we simulated burst release by injecting known concentrations of 5-HT just above the porous membrane and monitoring the change in 5-HT concentration with CV electrodes over the course of diffusion through the membrane and into the bulk volume (Fig. 6). A schematic of this cell-release simulation via 5-HT injections is illustrated in Fig. 6a. Five 17.5 µL injections of 100 µM 5-HT were monitored in this process with 5-min accumulation time per measurement. Each injection, when fully diffused into 3.5 mL Dulbecco’s modified Eagle medium (DMEM) bulk volume, was expected to reach a final 5-HT concentration of 0.5 µM, wherein each subsequent injection would increase the final bulk concentration by 0.5 µM (ranging from 0.5 to 2.5 µM over the course of the experiment). A 2 µL CNT-coated electrode was used for this experiment to obtain a fast time resolution. The timeline of Ipa values detected from these injections shows spikes at 10 min after the injection, followed by an exponential decrease in Ipa until the next injection event, when 5-HT detection spikes again. Previous work by our group46 showed that monitoring FDM diffusion through the membrane yielded a peak in Ipa at 1 min after injection, indicating that small-molecule transport across the membrane occurs within this 1-min timescale. 5-HT diffusion should occur at a similar or faster timescale, as it is a smaller molecule (5-HT: 176 Da, FDM: 246 Da); however, it resulted in a peak in Ipa at 10 min after injection. As demonstrated in Fig. 4a, this may be due to the time dependency of 5-HT adsorption and reaction at the Au-CNT electrode, where it was shown that molecular saturation occurs at ~10 min in a constant-concentration solution. The exponential decrease seen after the Ipa peak follows the expected decrease in concentration due to 5-HT diffusion away from the electrode. A linear trend can be observed by plotting the final Ipa values per injection against the expected final concentration (Fig. 6c, inset) with a slope of 0.156 µA/µM. There was significant variation in the peak Ipa values for each injection, despite using a constant concentration of 5-HT. However, since detection of 5-HT at constant concentrations has been demonstrated to be highly repeatable (denoted by R2 values in Fig. 4, 5), this may be due to experimental error when placing the pipet at the membrane surface for each injection or because our measurement frequency is much slower than the diffusion dynamics. The consistent shape of each injection curve provides confidence that dynamic monitoring of 5-HT burst release is feasible within this system, owing to the close proximity of electrodes to the cell culture membrane where release events would occur.

### RIN14B enterochromaffin cell-released 5-HT

Detection of 5-HT secreted from a standard T75 flask culture of an immortalized enterochromaffin-model cell line (TRPA1-expressing RIN14B cells) was also achieved, demonstrating that cell-secreted 5-HT is detectable within the linear range of our electrode. Figure 7a depicts the detection of 5-HT in the supernatant of RIN14B cells cultured in a separate T75 flask, as measured at 2 µL Au-CNT membrane electrodes. Cells were stimulated to release 5-HT by 1 h incubation with 100 µM sodium butyrate, a short-chain fatty acid produced by commensal gut bacteria that is known to stimulate enteric 5-HT production in vivo and in RIN14B cells2. Figure 7b, c shows that the CV signal increased by ~0.33 µA at Epa ~0.26 V, indicating that the 5-HT concentration detected in the supernatant increased after butyrate stimulation. Background 5-HT is detected in the cell supernatant before stimulation, potentially due to basal 5-HT release from cells or due to the presence of 5-HT in fetal bovine serum (FBS), which varies from batch to batch. A calibration curve was performed using the same electrode with 5-HT standards (1–5 µM 5-HT in DMEM) as an internal control to estimate the concentration of cell-released 5-HT (Fig. 7d). When compared with this calibration curve, the approximate concentrations of 5-HT detected from the supernatant before and after stimulation were ~3.1 µM and ~4.1 µM, respectively, suggesting that ~1 µM 5-HT was released from the cells. This cell culture contained 25 × 106 cells in a 5 mL volume, corresponding to 0.2 nmol 5-HT released/106 cells. This result demonstrates that RIN14B cells can be stimulated to release 5-HT using a known microbial metabolite and that this cell-released 5-HT can be detected by our Au-CNT membrane electrode in the cell supernatant. Equally importantly, no other redox active molecules were found in this cell supernatant within the ranges of the applied CV scan. This greatly enhances our ability to attribute biological function to 5-HT concentration. In future experiments, it could be tested whether other biological redox molecules (e.g., ascorbic acid, uric acid) interfere with 5-HT detection by producing overlapping CV peaks. The use of nanostructured electrode coatings has been shown to separate these peaks to distinguish between 5-HT and other contaminants. The CNT film used here may provide this capability38 or may be further optimized in combination with graphitic structures19, nanostructured platinum47, and ionic polymers such as Nafion38 and chitosan48,49.

RIN14B cells were plated and cultured directly on the electrode-integrated membrane and characterized by optical microscopy and CV monitoring, as shown in Fig. 8. Neither the 3D-printed platform in which cells were plated nor the PETE membrane used here was optically transparent, so assessment of the cell culture was performed with endpoint live/dead fluorescent staining and confocal microscopy. Figure 8a, b shows representative images of RIN14B cells plated on polystyrene T75 flasks (Fig. 8a), imaged with bright-field microscopy, and RIN14B cells plated on an electrode-integrated PETE membrane (Fig. 8b), imaged with fluorescence microscopy. The cells plated on polystyrene have epithelial-like morphology with many long processes and are clearly well attached to the surface. Comparatively, the cells plated on the porous membrane are rounded and sparse and appear to lack strong attachment to the membrane, despite collagen treatment.

In Fig. 8c, d, an Au-CNT (2 µL) membrane electrode was used to monitor cells over the course of molecular treatment of the cell culture, including injections of 1 µM 5-HT to calibrate the electrode and spike the solution above the LOD of the electrode (denoted by the arrow). In Fig. 8d, compared with the cell-only baseline over t = 0–20 min, which showed no peaks, the 1 µM 5-HT spike at t = 20 min induced an increase in Ipa over t = 20–35 min that then gradually decreased until t = 60 min. The increased time to reach a peak in Ipa, compared with the no-cell measurements in Fig. 6c, can be attributed to many factors, including the cell barrier and the collagen barrier. A subsequent injection of 100 µM butyrate did not lead to an appreciable increase in Ipa, although the shape of the graph does not decrease exponentially as it does in Fig. 6c. This may indicate a small, immeasurable release of 5-HT. Similar results are seen in Fig. 8e, f, in which RIN14B cells were monitored on Au-CNT (12.5 µL) membrane electrodes that were either coated with collagen (Fig. 8f) or not (Fig. 8e). Cells were measured at 6, 4, and 2 h timepoints before treatment with butyrate to establish a baseline and then 20 min and 2 h after butyrate treatment to measure any stimulated 5-HT release. All CV curves before and after butyrate treatment showed no appreciable 5-HT secretion despite using 2-h accumulation times, which would maximize the signal and allow for a sub-100-nanomolar detection range. These results suggest that RIN14B cell attachment to the 1-µm pore PETE membrane was not strong enough to produce a sustainable, healthy cell culture, nor is it conducive to the cell pathways required for 5-HT secretion. However, these results indicate that healthy cell cultures are sustainable on polystyrene plastics, and the electrodes are still capable of 5-HT detection in the presence of collagen and cells.