3.1. Electrode Performance—Initial Characterisation
Characterisation of the clean electrodes was performed through electrochemical measurements and a physical examination of the surface. EIS and CV were employed to understand the initial electrochemical properties of each interface type. Figure 2
shows the cleaning CV as well as Fe[CN]63−
CV and impedance responses before any chemical modifications were made to the electrodes.
The cleaning CVs (Figure 2
A) show typical responses for gold immersed in H2
. AT and BT SPEs show different gold oxide formation and reduction profiles which can be attributed to the different crystal faces predominating on the gold particles employed in the ink, possibly resulting from the different curing temperatures [27
]. The CV voltage windows for each SPE type are also slightly different. AT electrodes produce a large oxygen evolution peak beyond +1.2 V, whereas BT electrodes produce this response above +1.4 V. The width of this potential window is related to the proportion of exposed gold, with higher proportions producing a narrower voltage window [28
]. The cleaning CVs for the PGEs and TFGEs are similar, which is to be expected, as these both exhibit a pure gold surface. Small differences possibly arise from the effect of the different cleaning methods employed.
CVs in the presence of Fe[CN]63−
show a typical response for the redox couple (Figure 2
B). The peak-to-peak separation of CV-cleaned AT SPEs was typically around 70 mV, whereas the BT SPEs showed approximately a 75 mV separation. The TFGEs exhibited peak-to-peak separations of around 90 mV, and the PGEs exhibited a separation of 73 mV. The SPEs and PGEs are close to the ideal peak separation of 59 mV at 298 K as described by the Nernst equation [25
], indicating excellent reversibility of the system. Systems employing the TFGE are slightly less reversible after initial cleaning.
The CV data correlates well with the EIS responses of these electrodes under the same solution (Figure 2
C). SPEs show rapid rates of electron transfer, indicated by the shape of the plot suggesting a high lambda value, and the fact that the Warburg impedance dominates even at high frequencies. The non-screen-printed gold surfaces show a small Rct feature, which shows that electron transfer is slowed at these surfaces. The TFGE exhibits a larger baseline Rct because the gold material is deposited thinly (approximately 10–20 nm) and therefore has a higher internal resistance than polycrystalline gold or screen-printed deposits.
Electrode surfaces were also visualized using microscopy and surface profiling to assess innate roughness. Scanning electron microscopy (SEM) images are shown in Figure 3
A–C, and atomic force microscopy (AFM) images in Figure 3
D–F. The TFGE has a highly smooth surface when viewed under both SEM and AFM. Of the five 90 µm × 90 µm areas examined on a fresh electrode, the median root mean squared roughness (Rrms
) was 187 nm.
The screen-printed sensors show a rougher profile, which is expected of a surface formed by the deposition of gold particles suspended in an ink. AT electrodes exhibit a reasonably smooth surface with isolated voids and raised regions on the micrometre scale. The median Rrms when examined under AFM was 712 nm, significantly higher than that of the TFGE. BT SPEs show an even more irregular surface than AT SPEs when examined under SEM, with deep voids of several micrometres in size throughout the surface and non-homogenous particle sizes. These electrodes could not be accurately imaged by AFM due to the large scale of the differences exceeding the capabilities of the measurement technique. It was not possible to examine the PGEs by SEM of AFM due to the size of the electrodes.
This initial characterisation of the electrodes suggests that SPE sensors would be the most suitable for DNA SAM formation in a point of care setting. Both SPE types showed very low initial Rct values and excellent reversibility of the Fe[CN]63−/Fe[CN]64− redox couple. The TFGE had the largest Rct and the widest peak separation despite possessing the least rough surface of all the electrodes examined, which may indicate that it will not perform as well as the other electrode types as a DNA biosensor.
3.2. Electrode Performance Following Cleaning and Immobilisation of SAM Layers
Once cleaned and characterised, electrodes were incubated with a mixed SAM solution containing thiolated DNA probe and a short chain alkanethiol molecule (either MCH or MCP). Previous work by this group has successfully detected specific DNA binding using the techniques detailed here [29
], and has optimised DNA target binding by examining the effect of DNA target length and overhang length [30
], and the effect of agents such as tris(2-carboxyethyl)phosphine (TCEP) and formamide for electrode preparation and target incubation [31
]. This work is therefore focussed on examining the role electrode surface conditions and SAM thickness have on biosensor performance to provide insight into how electrode and SAM selection can affect the sensitivity and specificity of DNA detection.
EIS was performed to measure impedance changes upon the formation of a DNA SAM and complementary target binding. The signal ratio is calculated as the percentage change in Rct following DNA hybridisation (therefore, a 0% signal ratio indicates no change with DNA hybridisation relative to the pre-hybridisation Rct measured), and is the mean of all electrodes in a group. This is a common way of reporting Rct changes in response to analyte binding [26
]. The signal increase after target hybridisation on each electrode type is presented in Figure 4
Typical EIS responses for each electrode type are presented in Figure 4
A. As expected, MCH SAMs produce a larger Rct than monolayers of MCP. This is due to the increased thickness of the MCH monolayer reducing the efficiency of electron tunnelling through the intact SAM, and is consistent across all electrode types [25
]. In most cases, the addition of target DNA to an electrode caused an increase in the impedance of the electrode. This is the expected response on macroelectrodes when using a negatively charged redox couple such as Fe[CN]63−
due to increased steric hindrance and negative charge accumulation at the electrode surface [26
]. The MCH SAM exhibited a smaller response increase than the MCP SAM, which is due to the lower initial Rct of the MCP SAM allowing a similar absolute change in Rct to produce a much larger signal ratio. This type of effect, i.e., improving sensitivity by reducing initial Rct, has also been reported through the use of PNA or morpholino-based probe molecules [29
]. The use of a shorter alkanethiol allows a similar effect to be achieved at a lower cost but has its own limitations; shorter alkanethiols are less stable due to reduced intermolecular attractions binding the film together [35
] and typically have more defects within the SAM [36
]. DNA-sensing layers containing short alkanethiols may also show an ageing effect when the sensor is washed or regenerated [37
]. Such an effect may be admissible in the laboratory, but for point-of-care testing and reusable sensors, this method of signal enhancement may not be appropriate.
The PGE and BT SPE both exhibit the expected behaviour in response to DNA target binding, with an increase in Rct observed independent of SAM composition. The BT SPE had significantly larger impedance responses than the PGE, which may be a result of the proportion of exposed gold being relatively low compared to other electrodes, resulting in a smaller conductive area [28
]. Both the AT SPEs and the TFGEs showed similar impedance responses with an MCH SAM. The AT SPE shows an unexpectedly large impedance with an MCP SAM relative to the MCH SAM. There may be some effect from the surface roughness or ink composition here, causing the mixed monolayer to form in a way which produces a greater impedance than expected. This may be supported by the TFGE, which shows a similar MCH impedance response and a more typical MCP response.
B shows the responses of each electrode and SAM combination to the addition of a complementary DNA target. The PGE produced consistent responses as expected. There are likely contributions to this consistency from the ability to clean these electrodes more thoroughly than the other types examined. The BT SPE showed a larger signal response than the PGE for both MCH and MCP SAMs, however in both cases this was highly variable. This is believed to be due to the complex surface topography allowing the DNA to take many orientations at the surface, rather than the highly organised and vertical SAM expected on more planar substrates. Although this surface could not be examined with AFM, it is expected that the surface will also be rough, which will increase the number of defects in the SAM layer and therefore increase variability.
The AT SPE produced very little response from baseline with both MCH and MCP modification, and in both cases, a mean decrease in signal was observed following DNA target incubation. The TFGE exhibits a similar but smaller decrease in signal with an MCH SAM, and an increase in signal when a DNA target is added to an MCP SAM. As these surfaces are more similar to the PGE than the BT SPE surface, we might expect that these electrodes would give responses more like a PGE. The AT SPE appears relatively insensitive to the addition of DNA target with the protocol used, which may be a result of this method not being optimised for screen-printed surfaces. Both the AT SPE and TFGE showed a consistent decrease with an MCH SAM which may result from single-stranded DNA remaining close to the surface as it has a low persistence length, but opening channels through which the redox mediator may move when in the more rigid double stranded hybrid [33
]. A positive signal ratio with MCP SAMs may result from reduced electrostatic attraction of the probe DNA to the electrode due to the DNA strand originating further out into solution from the SAM surface.
Based upon these results, for the simple transfer of protocols from traditional PGEs to a suitable PoC electrode, the TFGE appears to offer the best responses, as the responses of these electrodes are similar but smaller than the PGE. The BT SPE offers greater response sizes, but the variability in these electrodes was significant and is expected to remain high even with more electrodes tested due to electrode-to-electrode and batch-to-batch variation in the printing process. In contrast to this, the AT SPE offered acceptable reproducibility but showed minimal response to DNA target binding. The TFGE, especially with a shorter alkanethiol SAM, provided reasonable response sizes with enhanced reproducibility with only a small number of repeats. Whilst the results were not statistically significant, the experiments were self-consistent which allows comparison of the three electrode types (AT, BT and TFGE). In these comparisons, it was clear that the TFGE gave the response most similar to a PGE. In addition to this, the manufacturing process of the TFGE is much less variable and therefore batch-to-batch variability is expected to be low. While these SPEs may offer an excellent platform for biological measurement when using an optimised protocol [18
], for rapid translation from the benchtop to bedside we believe TFGEs may offer the simpler solution.
3.3. Specificity and Sensitivity of a TFGE-Based Biosensor
We next performed a specificity test on the TFGE biosensor to examine whether the performance of this system was similar to that of the PGE which we hope to replace. Electrodes were modified with an MCP SAM layer, as this showed the greatest signal-to-baseline ratio for both the PGE and TFGE, and then challenged with a non-complementary PCR amplicon of a similar length which was amplified from the same plasmid (the tetA sequence). Figure 5
shows the PCR outcome and the specificity results from this test.
A miniature thermocycler was used to amplify all PCR product used in these experiments. As shown in Figure 5
A, this successfully amplified the two products of interest with no non-specific amplification or cross-contamination visible. When the PGE was challenged with complementary target DNA, a mean Rct increase of 270% was observed (Figure 5B). The addition of non-complementary target produced a mean increase of 68%, which is believed to be due to the target adsorbing at defects in the SAM layer rather than non-specific hybridisation. Much of this signal increase was lost when the protocol was transferred onto a TFGE, which showed only a 103% mean increase in Rct with complementary DNA. When non-complementary DNA was added, the response size was low at only 13%. However, in both cases, the variability of the measurement is high relative to the signal size. This is believed to be due to the limited cleaning processes available for the TFGE resulting in surfaces which are less consistent than is possible with the PGE. While these surfaces most closely resemble those of a PGE, and the responses reflect those achieved on PGEs, there may be further optimisation of cleaning or SAM formation required to reduce the variability of the signal, especially with regards to filtering non-specific signal increase. However, the mean response sizes are promising for the accurate discrimination of complementary vs. non-complementary DNA, and for filtering non-specific signals in more complex media such as a PCR reaction mix. We also believe that, based on data previously presented, the translation of this protocol onto SPEs would not produce such results without extensive optimisation. We suggest that using surfaces more similar to a PGE, while not producing an identical response, will speed up the development of DNA biosensor platforms when methods are initially developed on classic benchtop electrodes.
Whilst this study did not achieve statistical significance for the OXA vs. tetA experiment, it is clear that differential hybridisation was achieved. It is also important to note that the PCR reaction adds an additional level of specificity, only amplifying genes which are present in the sample and primed for. This means that in a real-world sensor, specificity would come largely from the PCR step. However, it is reassuring that the OXA probe has a good level of specificity at room temperature.