As a basis for sensor development, we employ an 18-nucleotide mutated high-gain aptamer sequence (high-gain parent—
Table 1) that specifically binds to aminoglycoside antibiotics [
6]. We recently demonstrated that this modified sequence, adapted from the original aptamer sequence reported by Wang and Rando [
35], exhibits
increased sensitivity and binding affinity when employed in electrochemical, aptamer-based (E-AB) sensors [
6,
35]. The high signal gain is afforded by the large conformation change of the aptamer structure from the target-free to the target-bound state. Unfortunately, sensors using the high-gain parent aptamer exhibit a high affinity for tobramycin, such that the sensor saturates well before the therapeutic levels of the antibiotic tobramycin (7–18 µM) [
33]. The 18-nucleotide sequence exhibits a dissociation constant of 80 ± 10 nM and saturates at a tobramycin concentration of ~5 µM, precluding sensitive measurements in the therapeutic window (
Figure 1). Motivated by this result, we explore two strategies to
reduce the observed binding affinity of E-AB sensors while maintaining the magnitude of the signal change (read conformation change) and thus sensitivity. Furthermore, while the motivation here was to push sensor performance into the therapeutic window, the strategies we outline below should represent a general approach to reducing the observed binding affinity of E-AB sensors. Of note, in this report, we use “observed binding affinity” and “binding affinity” interchangeably. Both terms refer to the binding affinity displayed by the fabricated sensors which is not to be confused with the intrinsic binding affinity of the original RNA aptamer to tobramycin (~12 nm) [
32].
We employ two strategies to engineer aptamers capable of supporting E-AB signaling with reduced affinity towards tobramycin. Naively, our first strategy was to mutate a nucleotide involved in binding interactions with tobramycin to reduce binding affinity while minimally perturbing the predicted secondary structure of the parent aptamer. Hypothetically, maintaining similar secondary structure to our high-gain parent sequence would ensure that the magnitude of the conformation change would be similar. Our second approach was to modify the aptamer sequence in order to stabilize an alternatively-folded structure or target-free state, such that target binding would have to overcome a larger energy barrier to force the aptamer to the target-bound state. This approach is similar to the three-state equilibrium model reported by Kang
et al. [
17,
18] in the development of electrochemical DNA hybridization sensors. The strategy again was to minimally perturb the predicted secondary structure in order to maintain similar sensor sensitivity.
To quantitatively characterize the E-AB sensors developed in this manuscript, we fit the sensor calibration curves to a binding model adapted from the Langmuir isotherm [
36]. The calibration relies on the equilibrium reaction between the aptamer (A) and target (T) where A + T ↔ A:T and K
A = [A:T]/[A][T] (M
−1) and K
D = 1/K
A (M). With the assumption that each non-interacting binding site (aptamer) binds one tobramycin and binding does not appreciably alter the concentration of free target ([T]) in solution. The binding isotherm is given by Equation (1):
where S and S
max are the percent signal change at a given [T] and at saturating target concentration, respectively.
3.1. Disrupting the Aptamer–Target Interaction for Reduced Affinity Sensors
To design a mutated binding site aptamer sequence with a reduced affinity towards tobramycin, we set two design parameters to maintain sensitive signaling ability. Our goal was to disrupt a polar contact (e.g., hydrogen bond) between the aptamer and tobramycin by mutating one of the nucleotides involved in binding (
Figure 2). Our first criterion was that the altered nucleotide should only have one polar contact with tobramycin in order to
weaken the interaction rather than eradicate it. The second criterion was that the secondary structure of the new sequence must be similar to that of the 18-nucleotide parent sequence as predicted by
MFOLD [
37,
38] (
Figure 1). This would ensure that, upon target binding, the signal change of the E-AB sensor (and thus sensitivity) would be similar to the original sensors.
Figure 2.
Secondary structure predictions for the parent sequence suggests an internal loop that can potentially keep the redox label (MB—methylene blue) distal from the electrode surface. The introduction of tobramycin forces the aptamer to fold bringing the methylene blue close to the 5'-terminus. The lowest energy secondary structure prediction is calculated using
MFOLD [
37,
38]. This prediction was based on the parent aptamer in a 1 M NaCl solution at 25 °C and has a folding energy of −0.66 kcal/mol.
Figure 2.
Secondary structure predictions for the parent sequence suggests an internal loop that can potentially keep the redox label (MB—methylene blue) distal from the electrode surface. The introduction of tobramycin forces the aptamer to fold bringing the methylene blue close to the 5'-terminus. The lowest energy secondary structure prediction is calculated using
MFOLD [
37,
38]. This prediction was based on the parent aptamer in a 1 M NaCl solution at 25 °C and has a folding energy of −0.66 kcal/mol.
We examined the solved NMR structure of the aptamer-target complex to determine possible polar contacts between the aptamer and tobramycin [
32] (
Figure 3). Upon analysis, we find 15 hydrogen bonds between 10 different nucleotides in the aptamer sequence and tobramycin. Of the 10 nucleotides, only eight have one polar interaction with the target of interest and only six of the nucleotide interactions involve the base (in contrast to interactions with the sugar or phosphate in the backbone). As a result, we identified six possible sites for mutation. We then explored the effects of iterative mutations using
MFOLD to ensure that the secondary structure of the mutant sequence was similar to that of the parent sequence [
6]. We find that modifications to the uracil-7 site (5'-CUUGGU
UUAGGUAAUGAG-3') to adenine, guanine, or cytosine all exhibited similar structures (
Figure 3), as did folding free energies to the parent high-gain sequence. Moving forward, we chose to design two sequences in which the uracil-7 was changed to either a cytosine or guanine (
Table S1). The uracil was replaced with cytosine (sequence 7UC) with the prediction that the 4’ nitrogen would inhibit the hydrogen bond interaction. Alternatively, the uracil was replaced with a guanine (sequence 7UG) to sterically hinder the hydrogen bond between the aptamer and tobramycin (all sequences listed in
Table 1).
Figure 3.
The uracil-7 site is a likely candidate for mutation to disrupt aptamer binding with target, thus reducing sensor affinity. We chose this site based on our criteria that it only has one polar contact (dashed lines) with tobramycin as determined via the NMR structure. This figure is generated from the previously reported NMR structure (PDB ID 2TOB) by Jiang and Patel [
32].
Figure 3.
The uracil-7 site is a likely candidate for mutation to disrupt aptamer binding with target, thus reducing sensor affinity. We chose this site based on our criteria that it only has one polar contact (dashed lines) with tobramycin as determined via the NMR structure. This figure is generated from the previously reported NMR structure (PDB ID 2TOB) by Jiang and Patel [
32].
Unfortunately, the sensors employing both of the new sequences (7UC and 7UG) did not function as expected (
Figure 4). For example, no appreciable or specific signal changes were observed with sensors fabricated using the 7UC and 7UG sequences. At 30 µM, tobramycin sensors employing 7UG displayed a −9% ± 1% signal change and 7UC exhibited a 3% ± 1% percentage signal change. As such, these sensors exhibited no quantitative binding to tobramycin. It is likely that the alterations we made to the aptamer sequence rendered the aptamer unable to bind tobramycin.
Figure 4.
(Left) The aptamers with mutated binding sites, 7UG and 7UC, did not produce functioning electrochemical sensors. Unfortunately, both constructs exhibit largely variable signals with changing tobramycin concentrations suggesting that disrupting the interaction of uracil-7 abolished any specific interaction with the target molecule; (Right) Illustrates the signal for sensors employing 7UG and 7UC at <35 μM tobramycin. These data represent titration curves calculated as percentage signal change using voltammetric peak currents as described in the experimental section. Each data point represents the average and standard deviation of at least three independently fabricated sensors.
Figure 4.
(Left) The aptamers with mutated binding sites, 7UG and 7UC, did not produce functioning electrochemical sensors. Unfortunately, both constructs exhibit largely variable signals with changing tobramycin concentrations suggesting that disrupting the interaction of uracil-7 abolished any specific interaction with the target molecule; (Right) Illustrates the signal for sensors employing 7UG and 7UC at <35 μM tobramycin. These data represent titration curves calculated as percentage signal change using voltammetric peak currents as described in the experimental section. Each data point represents the average and standard deviation of at least three independently fabricated sensors.
3.2. Stabilizing an Alternative Fold for Reduced Affinity Sensors
As an alternative approach to design an aminoglycoside aptamer with a reduced binding affinity towards tobramycin, we aim to stabilize an alternative aptamer fold by stabilizing a stem-loop structure internal to the aptamer sequence. Stabilization of the unbound structure will make it more difficult for the oligonucleotide to bind target and thus lower binding affinity [
17,
18]. This technique has been used before in the development of electrochemical DNA sensors in order to tune the linear range and sensitivity of the resulting sensors [
18]. Specifically, Kang
et al. utilized a DNA sequence that forms a stem-loop structure when the oligonucleotide is not bound to its complementary target and altered the stability of the DNA probe sequence by modifying nucleobases not involved in interacting with the target. The stability was improved by increasing the GC content in the stem, which is in the stem-loop of the unbound DNA probe, to reduce the affinity of the DNA sequence to its complementary target [
18].
We took two approaches to stabilize the target-free aptamer structure as a stem-loop. First, we made various mutations at the 3'-end of the aptamer sequence to bases that are not involved in binding with tobramycin such that the 3'-terminus possessed internal complementarity. Alternatively, we extended the sequences at the 3'-end to self-fold into a stem-loop structure (
Figure 5). The secondary structures of the various sequences were predicted by
MFOLD to ensure that they formed a stem-loop structure where the 5' and 3' ends are distant from one another (
Figure 5) with a more favorable free energy for folding than the parent aptamer. It was necessary to ensure that the mutated aptamer sequences formed a stem-loop with distant 5' and 3' ends so that the probe will be forced to undergo large conformational changes upon target addition.
Figure 5.
We aimed to alter the aptamer sequences to develop stabilized target-free states while maintaining a similar secondary structure to the high-gain parent sequence. The sequence mutations (highlighted in red boxes) attempt to stabilize the internal loop structure at the 3'-distal end. Free energies of each structure were calculated using
MFOLD as described above [
37,
38].
Figure 5.
We aimed to alter the aptamer sequences to develop stabilized target-free states while maintaining a similar secondary structure to the high-gain parent sequence. The sequence mutations (highlighted in red boxes) attempt to stabilize the internal loop structure at the 3'-distal end. Free energies of each structure were calculated using
MFOLD as described above [
37,
38].
In analyzing the parent aptamer structure, it was determined that there is a hydrogen bond interaction between uracil-8 and guanine-16 (
Figure 5). Mutating guanine-16 to uracil causes an interaction between adenine-9 and uracil-16, which stabilizes a stem-loop from −0.66 kcal/mol for parent, to −2.13 kcal/mol (
Figure 5). This new sequence is named 16GU. Extending the parent sequence with nucleotides UAC at the 3'-end results in predicted interactions between guanine-11 and cytosine-21, uracil-12 and adenine-20, and adenine-13 and uracil-19. This sequence, here named 3-UAC, also stabilizes a stem-loop structure with a free energy of −1.60 kcal/mol (
Figure 5).
The sensors fabricated with the new aptamers, 16GU and 3-UAC, were successful in creating reduced affinity sensors. For example, sensors prepared with the parent aptamer exhibited a dissociation constant of 0.08 ± 0.01 μM (
Figure 1), while sensors fabricated with the 16GU aptamer exhibited a dissociation constant of 3.0 ± 0.4 μM (
Figure 6) and sensors employing 3-UAC displayed a 0.17 ± 0.03 μM dissociation constant (
Figure 6). Consequently, the limits of detection (LOD) for each sensor are also affected. Specifically, the LODs increase as the dissociation constants for the aptamers increase. Sensors employing the parent, 3-UAC, and 16GU aptamers exhibit LODs of 1.99 nM, 14.8 nM, and 114 nM, respectively, calculated as three times the standard deviation of the blank. In addition, the E-AB sensors fabricated with the 16GU aptamer exhibited a maximum percent signal change at 30 μM of 112% ± 22%, which is comparable to that of the parent sensors (117% ± 12%). The 3-UAC sensors, however, only exhibited a maximum percent signal change of 69% ± 8%, which is smaller than that exhibited by the 16GU sensors and the parent fabricated sensors (
Figure 7). It is still unclear as to why the 3-UAC sequence exhibits lower sensitivity, but is likely due to the difference in the secondary structure with respect to the high-gain and 16GU sequences.
Figure 6.
Sensors fabricated with mutated sequences stabilizing the target-free state exhibit reduced affinity (as indicated by the increase in dissociation constants—Kd). (Top) For example, the sensors fabricated with the 16GU sequence exhibit a dissociation constant of 3.0 ± 0.4 µM and a 112% ± 22% signal change at 30 µM tobramycin; (Bottom) Similarly, the sensors fabricated with the 3-UAC sequence exhibit a Kd of 0.17 ± 0.03 µM, but with a lower overall signal change of 69% ± 8% at saturating conditions. In agreement with the predicted stabilities of the target-free structure, the more stable 16GU (−2.13 kcal/mol) exhibits the highest Kd, followed by 3-UAC (−1.6 kcal/mol), both of which are higher than the high-gain parent sequence (−0.66 kcal/mol) with a dissociation constant of 0.08 μM. These data represent titration curves calculated as percentage signal change using voltammetric peak currents as described in the experimental section. Each data point represents the average and standard deviation of at least three independently fabricated sensors.
Figure 6.
Sensors fabricated with mutated sequences stabilizing the target-free state exhibit reduced affinity (as indicated by the increase in dissociation constants—Kd). (Top) For example, the sensors fabricated with the 16GU sequence exhibit a dissociation constant of 3.0 ± 0.4 µM and a 112% ± 22% signal change at 30 µM tobramycin; (Bottom) Similarly, the sensors fabricated with the 3-UAC sequence exhibit a Kd of 0.17 ± 0.03 µM, but with a lower overall signal change of 69% ± 8% at saturating conditions. In agreement with the predicted stabilities of the target-free structure, the more stable 16GU (−2.13 kcal/mol) exhibits the highest Kd, followed by 3-UAC (−1.6 kcal/mol), both of which are higher than the high-gain parent sequence (−0.66 kcal/mol) with a dissociation constant of 0.08 μM. These data represent titration curves calculated as percentage signal change using voltammetric peak currents as described in the experimental section. Each data point represents the average and standard deviation of at least three independently fabricated sensors.

Figure 7.
(Left) A comparison of the sensor responses when employing the parent, 16GU, and 3-UAC aptamers. The grey box is representing the therapeutic window of tobramycin from ~7–18 µM; (Right) A zoom in of the 0–1 µM window to show the difference in sensor function for the parent, 16GU and 3-UAC aptamer sequences. These data represent titration curves calculated as percentage signal change using voltammetric peak currents as described in the experimental section. Each data point represents the average and standard deviation of at least three independently fabricated sensors.
Figure 7.
(Left) A comparison of the sensor responses when employing the parent, 16GU, and 3-UAC aptamers. The grey box is representing the therapeutic window of tobramycin from ~7–18 µM; (Right) A zoom in of the 0–1 µM window to show the difference in sensor function for the parent, 16GU and 3-UAC aptamer sequences. These data represent titration curves calculated as percentage signal change using voltammetric peak currents as described in the experimental section. Each data point represents the average and standard deviation of at least three independently fabricated sensors.
Our initial goal was to design an aptamer sequence that would support sensing of tobramycin in the therapeutic window. Sensors utilizing the mutant aptamer 16GU provided better sensitivity towards tobramycin in the therapeutic window (
Figure 7). Specifically, 16GU sensors exhibit a ~20% signal change between 7 and 18 μM tobramycin, whereas the high-gain parent and 3-UAC (which both saturate before 7 µM) exhibit essentially no signal change in that window. We were able to successfully reduce the binding affinity of the E-AB sensors and significantly improve the sensitivity in the therapeutic window of tobramycin.