The device was used two different forms, which may be seen in Fig. 1; the PCB array itself is shown in Fig. 1a. Before use, the PCB was first cleaned, functionalised with 4-Aminothiophenol (4-ATP), and the 4-ATP self-assembled monolayer (SAM) was electropolymerised, using cyclic voltammetry (CV) between − 0.2 and 0.7 V at 50 mV/s, to improve the reversibility of the redox behaviour of the SAM. Once fully processed, depending on the experiment conducted, the PCB was secured in one of its two forms. Its ‘open’ form, where it was secured in a holder and a uniform well of fluid suspended over the array, is shown in Fig. 1b and c. The second form was its microfluidic ‘closed’ form, shown in Fig. 1d. In its open setup, samples were able to be freely extracted throughout the experiment. However, in its microfluidic form this physical removal of sample was no longer possible, and all data collection was required to be done without interrupting the system.
Before protein preconcentration trials using the PCB array in its open form, the optical response of different concentrations of bovine haemoglobin (bHb) was first calibrated between 100 µM and 1 µM. Initially, the absorption and transmission at 280 nm were measured using the Genova Nano Micro-volume Spectrophotometer (GNMSpec, Jenway, United Kingdom); this was performed as this wavelength maps to the maximum absorbance in proteins 14. Figure 2a and b show the results for the absorbance and transmission at 280 nm respectively and show the linear fits which have R2 values of 0.9989 and 0.9864. Full spectrum analysis was also conducted using the GNMSpec to provide insight into the optical properties of the solution between 300–500 nm (Fig. 2c). This analysis displayed a clear peak at 406 nm, the intensity for which may be seen in Fig. 2d, and was shown to have a strong linear relationship with the protein concentration (R2 value of 0.9982). The importance of this data lies in the ability of the full spectrometer system to record this wavelength accurately; this is because 280 nm falls outside of its working range of 350–880 nm and thus had little use when the device was sealed into its microfluidic form.
For the open configuration of the device, where the extraction of physical samples was still possible, the absorption at 280 nm was chosen as the calibration method due to its, slightly, higher R2 value. Preliminary experiments showed that physically extracting 2 µL samples should be done more than 2 minutes apart to minimise the effect on the environment, 5 minutes was selected, and the pH response over time when 0.4 V CA was applied may be seen in Fig. 3a. Samples were taken from above the pad on the array which was connected as the working electrode (WE) and transferred to the GNMSpec for analysis. In between uses, the machines’ receptacle was cleaned with ultrapure water (MilliQ) and dried with fibreless tissue.
The PCB was cleaned, functionalised and electropolymerised before being secured into its open setup, see Fig. 1c. A solution of 50 µM bHb and 100 µM 5(6)-Carboxynaphthofluorescein (CNF, Sigma-Aldrich, United Kingdom) was prepared in 10 mM phosphate buffered saline (PBS, Sigma-Aldrich, United Kingdom) (pH 7.1 ± 0.1), 1 mL of which was pipetted into the well to create a well with a uniform path length of 1.4 mm. This was to improve the consistency of the data collected using the USB spectrometer, however, it was noted that the surface tension at the edges of the PCBs’ holder caused imperfections in the uniform path length which the holder was designed to create. 0.4 V was then applied to the WE whilst a gold pseudo-reference on the PCB was used as the counter electrode (CE) / reference electrode (RE). Before the potential was applied, a 2 µL sample was taken from above the WE and analysed with the GNMSpec. 0.4 V chronoamperometry (CA) was then applied to the WE for 30 minutes and a 2 µL sample collected from just above the electrodes’ surface every 5 minutes. The results for this may be seen in Fig. 3b and shows a peak at a pH change of -0.4; this corresponds to a pH of 6.7 ± 0.1, where 6.8 is the pI of bHb 15, and a concentration factor (CF) of over 3. With the device working in its open configuration, microfluidics were subsequently designed using AutoCAD 2022 (Autodesk, United States) to affix to the PCB array, this may be seen in Fig. 1d.
Preliminary data with this microfluidic device showed issues with low concentrations of CNF as the number of molecules was insufficient to measure the pH optically. The concentration of CNF was increased ten-fold to 1 mM in 10 mM PBS and recalibrated in a more refined pH range, pH 6.0–7.4, the results for which may be seen in Fig. 4a. CA of different voltages were evaluated for 20 minutes to confirm whether 0.4 V remained the optimal potential for EGA; potentials between 0.1–0.6 V were tested. The results for this are shown in Fig. 4b and illustrate clearly that 0.4 V gives the most reliable change in pH, -0.31 ± 0.01, over a period of 20 minutes. Whilst 0.3, 0.5 and 0.6 V do give a higher pH change, the reliability is low compared to that of 0.4 V and thus were not used. The WE and CE/RE were pads within the same channel, and the channels were designed so that minimal optical scattering would occur with the microfluidic layers incorporated. The EGA was then performed for 60 minutes for direct comparison between the open and microfluidic systems, the graph for which may be seen in Fig. 5a. Despite showing a smaller response for the pH change in its microfluidic form, there is a clear, linear decrease in pH shown and the system was further characterised.
The behaviour of the microfluidic device was further characterised by defining the microfluidic systems’ spatial resolution, which we have defined in our published work 16 as “the distance at which the effects of EGA are no longer observed via our optical characterization method, using our control value as the benchmark threshold”; the detailed procedure for how this was achieved may also be found in that work 16. Here, CA of 0.4 V was applied for 20 minutes to pads in the microfluidic channels at different distances from the spectrometers’ position. In the non-microfluidic form, the spatial resolution was found to be 0.664 mm whilst in its microfluidic form we report that this property is at 0.171 mm, the data for which may be seen in Fig. 5b. Therefore, all data on our array with spacing of 0.9 mm may be considered as independent. This experiment also yielded visual data which may be seen in Fig. 6a; from top to bottom the pad used as the WE was changed from A5 to D2 whilst the optical analysis was held steady at column 5 throughout. CNF is known to change colour as the pH shifts from basic (blue) to neutral (purple) to acidic (pink/red). This coupled with the RGB analysis of the image, Fig. 6b, shows there is a clear pH gradient formed in the channel when EGA is performed, which was the prerequisite for performing microfluidic IEF (µIEF).
To perform protein preconcentration using µIEF on the individually addressed PCB array, the PCBs were stripped of thiols, cleaned using SC-1, functionalised with 4-ATP for more than 19 hours to form a well-ordered SAM, electropolymerised, as before, and the microfluidic ‘sandwich’ secured atop the array. For the bHb, a solution of 50 µM bHb and 1 mM CNF was prepared in 10 mM PBS (pH 7.1 ± 0.1) and the 0.4 V was applied to the WE for 30 minutes to achieve a pH change of -0.58 ± 0.03. A 200-point Savitzky-Golay (SG) smoothing filter was applied, and the pH change was compared against the normalised intensity at 406 nm; this method for assessing the concentration of bHb was used as it was not possible to physically extract a sample from the sealed channels. As the concentration is directly correlated to the absorption intensity at this wavelength, as shown in Fig. 2d, and the substrate is optically thick, the change in concentration over time may be equated to the change of reflective intensity over time. This data may be seen in Fig. 7a, and shows the averaged data and the standard error of the mean, along with the range in which a peak would be expected to lie, corresponding to 6.8 ± 0.1. Qualitatively, it is shown that the intensity at 406 nm, which directly correlates to the concentration of bHb, changes with the pH change, and shows a clear peak at a pH change of approximately − 0.4. This corresponds to a pH of 6.7 ± 0.1 which, as the pI of bHb is pH 6.8, demonstrates clearly that the device is capable of concentrating protein samples.
eGFP was then assessed using a process which only differed from that used to bHb by changing the protein and the concentration of that protein from 50 µM to 25 µg/mL, which equates to approximately 929 nM. As the pI of eGFP is pH 6.2 17, it was not expected that with a pH change of approximately − 0.6, a complete peak would be observed which is confirmed by the data shown in Fig. 7b. This graph shows the normalised intensity at 507 nm, which is the emission wavelength for this protein, and shows a clear increase in intensity as the pH decreases. However, unlike bHb, this wavelength falls significantly closer to the emission wavelength for CNF in acidic/neutral pH, 567 nm, which may impact the reliability of this data.