Haixiang
Yu
a,
Juan
Canoura
a,
Bhargav
Guntupalli
a,
Xinhui
Lou
b and
Yi
Xiao
*a
aDepartment of Chemistry and Biochemistry, Florida International University, 11200 SW 8th Street, Miami, FL 33199, USA. E-mail: yxiao2@fiu.edu
bDepartment of Chemistry, Capital Normal University, Xisanhuan North Rd. 105, Beijing, 100048, China
First published on 29th July 2016
Sensors employing split aptamers that reassemble in the presence of a target can achieve excellent specificity, but the accompanying reduction of target affinity mitigates any overall gains in sensitivity. We for the first time have developed a split aptamer that achieves enhanced target-binding affinity through cooperative binding. We have generated a split cocaine-binding aptamer that incorporates two binding domains, such that target binding at one domain greatly increases the affinity of the second domain. We experimentally demonstrate that the resulting cooperative-binding split aptamer (CBSA) exhibits higher target binding affinity and is far more responsive in terms of target-induced aptamer assembly compared to the single-domain parent split aptamer (PSA) from which it was derived. We further confirm that the target-binding affinity of our CBSA can be affected by the cooperativity of its binding domains and the intrinsic affinity of its PSA. To the best of our knowledge, CBSA-5335 has the highest cocaine affinity of any split aptamer described to date. The CBSA-based assay also demonstrates excellent performance in target detection in complex samples. Using this CBSA, we achieved specific, ultra-sensitive, one-step fluorescence detection of cocaine within fifteen minutes at concentrations as low as 50 nM in 10% saliva without signal amplification. This limit of detection meets the standards recommended by the European Union's Driving under the Influence of Drugs, Alcohol and Medicines program. Our assay also demonstrates excellent reproducibility of results, confirming that this CBSA-platform represents a robust and sensitive means for cocaine detection in actual clinical samples.
However, target-induced conformational change is hard to control, especially for small-molecule-binding aptamers that have relatively high dissociation constants (∼μM KD).14 For example, the well-characterized cocaine-binding aptamer MNS-4.1 (KD ∼ 5 μM) is structurally stable and forms a three-way junction even before binding cocaine.15 To achieve an effective target-induced conformational change, Stojanovic et al. truncated the sequence to destabilize the aptamer, so that it exists in an equilibrium state consisting of both folded and unfolded structures.7 This aptamer exhibited cocaine-induced folding, but still retained some folding activity in the absence of target, resulting in a high background that significantly limited sensor sensitivity.7,16 This background can be reduced by splitting the aptamer into two17 or three18 fragments, which further destabilizes the aptamer such that the fragments are unable to assemble without target. This results in minimal background signal,17,18 while the fragments retain their capacity for target recognition and can successfully reassemble into a complex secondary structure in the presence of cocaine. However, aptamer splitting results in notably reduced target affinity.17 Thus, this approach still compromises the sensitivity of split aptamer-based sensors.
To improve the sensor's sensitivity, Zhang et al. reported an assay based on an enzyme-linked split aptamer to perform colorimetric cocaine detection.19 One of the fragments was conjugated to a plastic surface; in the presence of cocaine, this fragment would form a complex with the second fragment, which was modified with biotin in order to bind streptavidin-linked horseradish peroxidase for signal amplification. However, washing caused the dissociation of a subset of the assembled tripartite complexes, resulting in a limit of detection (LOD) of just 2.8 μM. To address this, Heemstra et al. incorporated a proximity ligation strategy into this enzyme-amplified split-aptamer-based assay.20 Specifically, cocaine binding facilitated the assembly of the two azide- and cyclooctyne-modified split fragments, bringing these two chemical groups into close proximity such that covalent bonds could be formed. This prevented dissociation of the assembled target–aptamer complex during washing and improved the sensor's detection performance. Although the signal was amplified by the enzyme, the sensitivity remained limited to 0.1 μM21 due to the low binding affinity of the split aptamer (KD = ∼200 μM).17 To overcome this limitation, we have developed a cooperative binding-based approach to generate split aptamers that retain high target affinity. Cooperative binding behavior is commonly observed in ligand-binding proteins that are highly responsive to ligand concentration, such as hemoglobin,22 ion channels23 and transcription factors.24 Those proteins generally have more than one ligand-binding site, where binding at one site increases the affinity of the other sites, resulting in a ‘switch-like’ binding curve.25 Breaker et al. initially found that some tandem riboswitches26,27 naturally employ such cooperative binding26 to control gene expression in response to subtle changes in ligand concentration. This cooperative behavior was further extended into artificial biosystems such as ribozymes,28 molecular beacons29 and DNA aptamers.30 Specifically, Plaxco et al. introduced the disorder into the parent aptamer to achieve cooperative binding, and focused on the demonstration of the “switch-like” response behavior of cooperative binding that occurred when ligand concentration approaches K1/2 (K1/2 represents the ligand concentration at which half of the binding domains are occupied). These engineered cooperative DNA aptamers could not be employed for a practical sensor platform because the introduction of cooperativity unavoidably reduced the target-binding affinity of the resulting cooperative aptamer (K1/2 = ∼3 mM).30
We have for the first time successfully incorporated two tandem target-binding domains into a split aptamer (termed cooperative binding split aptamer, CBSA) to achieve more sensitive detection of cocaine, even in complex biological samples. The initial cocaine-binding event stabilizes the structure of the split aptamer and facilitates subsequent target-binding at the second binding domain. Our CBSA exhibits higher target affinity and far more responsive target-induced aptamer assembly compared to the single-domain parent split aptamer (PSA) from which it was derived. Using a fluorophore/quencher pair, we have demonstrated that a CBSA-based fluorescence assay can achieve sensitive and reproducible cocaine detection in biofluid specimens, with a LOD of 50 nM in 10% saliva within 15 min. Given the simplicity of splitting and engineering a CBSA from an isolated aptamer and the excellent performance of the CBSA-based assay in biofluid samples, it should be straightforward to develop other CBSA-based assays from existing or future aptamers for rapid, sensitive and specific detection of various targets in clinical or field settings in a simple, low-cost assay format.
We anticipated that the CBSA fragments would remain separated in the absence of target, but would form two tandem cocaine-binding domains when fully assembled with cocaine. To confirm effective target-induced CBSA assembly, we developed a binding assay based on the fluorescent molecule 2-amino-5,6,7-trimethyl-1,8-naphthyridine (ATMND). It has been reported that ATMND can strongly bind to a thymine situated opposite a C3 spacer abasic site (AP site) within a DNA duplex (KD = 111 nM) via three-point hydrogen bonding.32 Although ATMND fluoresces brightly when free in solution, this fluorescence is greatly quenched when ATMND is bound to a DNA duplex in this fashion.33 We therefore replaced the adenosine (at position 10 from 5′) between the two binding domains of the short fragment with a C3 spacer to form an AP site with a thymine in the opposite position within the long fragment upon cocaine binding (Fig. 2A). In the absence of cocaine, the long and short fragments remain separated, with strong fluorescence produced by the free ATMND molecules (Fig. 2B, left). Upon addition of cocaine, the CBSA fragments undergo cooperative target-induced assembly and form a duplexed AP site that binds ATMND and quenches its fluorescence (Fig. 2B, right). The resulting CBSA-5325/cocaine complex contains four complementary base-paired segments (Fig. 2A, right; segments labeled (A)–(D)) and a dinucleotide bulge formed within each three-way junction.34
We subsequently confirmed the cocaine-induced assembly of CBSA experimentally. When we mixed 1 μM each of the long and short fragments with 200 nM ATMND in 1× binding buffer (10 mM Tris–HCl, 100 μM MgCl2, pH 7.4), we observed 10% background quenching (Fig. 2C, before cocaine addition). This quenching is most likely attributable to low levels of non-specific quench and non-target assembly of the CBSA. Upon addition of 250 μM cocaine, 72% of the ATMND fluorescence was quenched within 15 min, indicating rapid target-induced CBSA assembly (Fig. 2C, after cocaine addition). We also tested the performance of the AP-incorporating parent split aptamer (PSA) from which the CBSA was initially derived (Fig. 2D, PSA), which features only a single cocaine-binding domain, in our ATMND-binding assay. In the absence of cocaine, the PSA quenched 6% of ATMND fluorescence (Fig. 2C, before cocaine addition), and no measurable signal change was observed upon addition of 250 μM cocaine (Fig. 2C, after cocaine addition). This indicates that no cocaine-induced assembly is taking place, presumably because the shortening of stem 1 to four base-pairs with an abasic site in the middle results in poor thermodynamic stability of the cocaine–PSA complex at room temperature. We therefore conclude that the quenching observed most likely arises through non-specific interactions between PSA and ATMND.
Aptamer binding affinity7,35 and DNA hybridization efficiency36 are both strongly affected by magnesium concentration. We therefore used our ATMND-binding assay to optimize cocaine-induced CBSA assembly by varying the Mg2+ concentration from 0 to 1000 μM, and observed that maximum cocaine-induced ATMND quenching occurred in the presence of 100 μM Mg2+ (ESI, Fig. S1A†). In the absence of cocaine, we observed 9% quenching without Mg2+, which we attributed to non-specific CBSA assembly (ESI, Fig. S1A,† no cocaine). The presence of cocaine only generated an additional 7% quenching due to the absence of the Mg2+ counterion, resulting in strong repulsion between the CBSA fragments (ESI, Fig. S1A,† cocaine). At high concentration of Mg2+ (1000 μM), extensive CBSA assembly occurred in the absence of cocaine, producing 38% ATMND quenching, while also reducing the binding affinity of CBSA to cocaine, which only produced an additional 17% quenching upon addition of 250 μM cocaine (ESI, Fig. S1A†). We also varied the ATMND concentration from 50 to 1000 nM, and found that 200 nM ATMND produced low background and high target-induced signal change (ESI, Fig. S1B†). Under these optimized conditions (200 nM ATMND and 100 μM Mg2+), we examined the extent of target-induced CBSA assembly at different cocaine concentrations. We found a strong correlation between ATMND quenching from CBSA assembly and cocaine concentration in the range of 0 to 250 μM (ESI, Fig. S2†). We characterized the binding affinity of the assembled-CBSA for ATMND by titrating different concentrations (0–20 μM) of CBSA into 200 nM ATMND in the presence of 1 mM cocaine (ESI, Fig. S3†). The calculated KD was 365 nM, which is consistent with the reported value for ATMND binding to AP sites.32 We subsequently demonstrated that the presence of two binding sites makes CBSA-5325 far more responsive to the presence of cocaine than split aptamers containing a single binding domain. We produced a long split aptamer (LSA) from CBSA, in which we replaced one of the binding domains with fully complementary sequences (Fig. 3A, LSA). The LSA fragments quenched 75% of ATMND fluorescence in the absence of cocaine. Addition of 250 μM cocaine only induced an additional 10% signal change, indicating that most LSA fragments were stably pre-assembled even without target (Fig. 3B). In contrast, CBSA-5325 generated a large signal change (72%) upon addition of 250 μM cocaine because of its dual target binding domains, with far weaker background signal (9%) without cocaine.
Specific target binding at both domains is required for target–CBSA assembly. We tested the binding affinity of 38-GC and several point-mutated derivatives using isothermal titration calorimetry (ITC) (ESI, Fig. S4†). We found that replacing an adenosine at position 22 with guanine completely impaired cocaine binding (ESI, Fig. S4,† 38-GC-22G). Based on this finding, we created two CBSA mutants (CBSA-M1 and CBSA-M2) in which either of the two binding domains was disrupted by this single-nucleotide mutation, leaving only a single binding domain capable of binding cocaine (Fig. 3A). CBSA-M1 and CBSA-M2 were mutated at the 3′- and 5′-binding domain of the long fragment, respectively. We tested these mutants using the same ATMND-binding assay, and found that neither CBSA-M1 nor CBSA-M2 was capable of cocaine-induced aptamer assembly, with no significant ATMND quenching observed upon addition of cocaine (Fig. 3B). These results confirmed that both target-binding domains of the CBSA were required for target-induced aptamer assembly, and thus provide strong support for a cooperative-binding-based assembly mechanism.
We hypothesized that the binding affinity of the CBSA might be enhanced by further stabilizing the target/aptamer complex with additional base-pairs, based on prior findings that longer complementary stems surrounding the three-way junction increased the aptamer's target-binding affinity.37 Thus, we increased the total number of base-pairs between the two binding domains of CBSA-5325 by adding an additional one, two or three base-pairs to generate CBSA-5335, -5435 and -5445 (ESI, Fig. S5A†) and examined the extent of target-induced CBSA assembly at different cocaine concentrations (0–50 μM) using our ATMND-binding assay. Our results demonstrated that the number of additional base-pairs greatly affects both binding affinity and target-induced assembly. In the absence of cocaine, the quenching of ATMND fluorescence increased as the number of base-pairs between the two CBSA binding domains increased from 5- to 8-bp (ESI, Fig. S5B†). This is probably due to the increased thermo-stability of CBSA assembly, even without target. Upon addition of cocaine, we found a strong correlation between ATMND quenching from CBSA assembly and cocaine concentration in the range of 0 to 50 μM (ESI, Fig. S5C†). Compared to CBSA-5325, quenching saturation occurred at lower target concentrations for the other CBSAs, in keeping with the assumption that the CBSA binding affinity can be enhanced with additional base-pairs. However, considerable background assembly was observed for CBSA-5435 and CBSA-5445 in the absence of cocaine, reducing their target-induced signal gain. We thus found that CBSA-5335 is most responsive and exhibited the most extensive target-induced CBSA assembly, and we therefore used this construct for subsequent sensor development.
We then produced fluorophore/quencher-modified derivatives of CBSA-5325 and CBSA-5335 to achieve sensitive detection of cocaine. The short fragment was modified with an IowaBlack RQ quencher at its 5′ terminus and a Cy5 fluorophore at its 3′ terminus. In the absence of cocaine, the two CBSA fragments remain separate, and the flexibility of the unbound short fragment routinely brings the fluorophore into close proximity with the quencher, resulting in very low fluorescence (Fig. 4A, left). In the presence of cocaine, the two fragments assemble to form a rigid target/aptamer structure that separates the fluorophore/quencher pair, producing increased fluorescence (Fig. 4A, right). We used these two fluorophore/quencher-modified CBSAs to generate calibration curves for cocaine concentrations ranging from 0–1000 μM (Fig. 4B; spectra of CBSA-5335 shown in ESI, Fig. S6†) and used the Hill equation (eqn (1)) to fit the binding curve to calculate K1/2 and the Hill coefficient (nH):30,38
(1) |
We determined a K1/2 of 106 μM with an nH of 1.1 for CBSA-5325 and a K1/2 of 36 μM with an nH of 1.5 for CBSA-5335 (where K1/2 represents the cocaine concentration at which half of the binding domains are occupied and nH describes the order of binding cooperativity).38 An nH of 1.5 clearly indicates higher cooperativity between the two binding domains of CBSA-5335. We found that the measurable LOD was 500 nM (4.5 ± 0.8%) and 50 nM (4.3 ± 0.9%) for CBSA-5325 and CBSA-5335, respectively (Fig. 4B). Clearly, the target-binding affinity of CBSA can be affected by the cooperativity of the two binding domains. Note that CBSA-5335 has the highest cocaine affinity of any split aptamer described to date, and the LOD of CBSA-5335 is more than 200-fold lower than that of a previously-described single-domain, split aptamer-based cocaine assay using a similar sensing platform (LOD = 10 μM).17
We further used ITC to characterize the binding mechanism and affinity of CBSA-5335 for its target at both sites. Recognizing that the KD values of the aptamers were in the μM range, we set up a titration with a high cocaine:aptamer ratio to obtain accurate KD values for each binding scenario (the adjacent binding pocket empty, or occupied by cocaine), where KD represents the ligand concentration at which half of the receptor sites are occupied at equilibrium.39 The resulting two-phase titration curve confirmed the interaction of cocaine with the two binding domains of CBSA-5335 (ESI, Fig. S7A†). The binding stoichiometry between cocaine and CBSA was manually set as two, because previous studies have demonstrated that one cocaine-binding aptamer binds a single molecule of cocaine.31 ITC data of CBSA-5335 were then fitted with both independent-sites and cooperative-sites models40 with two binding sites (ESI, Fig. S7A,† black broken line represented independent-sites model and red solid line represented cooperative-sites model). We observed better fitting using the cooperative-sites model, and determined that the KD for the initial and secondary cocaine-binding events were 116 and 36 μM, respectively (ESI, Fig. S7A†). Based on these KDs, we calculated a K1/2 of 65 μM and an nH of 1.3 for CBSA-5335,41 which is comparable with the results obtained via the CBSA-5335-based fluorophore/quencher assay (K1/2 = 36 μM and nH = 1.5). Notably, the low K1/2 of CBSA-5335 represents a 56-fold higher target binding affinity relative to its single-domain PSA (KD = 2 mM, ESI, Fig. S7B†).
We believed that the target-binding affinity of CBSA could be affected by the intrinsic binding affinity of its parent split aptamer. To address this point, we replaced a G–C pair in the 5′ binding domain of CBSA-5335 with a wobble G–T pair to form CBSA-5335-GT. This alteration in stem 3 of the cocaine-binding aptamer reduces its binding affinity to cocaine31 (Fig. 5A), but we expected the resulting CBSA to still retain its cooperativity. Indeed, our fluorescence results demonstrated that both CBSAs have identical cooperativity (nH = 1.5) (Fig. 5B). However, since the intrinsic affinity of 38-GC is 4-fold higher than the parent split aptamer variant used for CBSA-5335-GT, CBSA-5335 yields a lower K1/2 (33 μM) and better sensitivity (LOD = 50 nM in buffer) than CBSA-5335-GT (K1/2 = 125 μM and LOD = 500 nM in buffer) (Fig. 5B).
We subsequently confirmed that our CBSA-5335-based fluorophore/quencher assay is capable of equally sensitive cocaine detection in saliva samples. The excitation wavelength for Cy5 (648 nm) does not induce auto-fluorescence in the saliva matrix, and thus produces minimal background fluorescence (ESI, Fig. S8†). Additionally, the high quenching efficiency of the IowaBlack RQ quencher42 allowed robust detection of cocaine at very low concentrations. To ensure that our CBSA-5335-based fluorophore/quencher assay can be reliably used in real-world (i.e., clinical or field) settings, we evaluated the assay's performance according to Scientific Working Group for Forensic Toxicology (SWGTOX) Standard Practices.43 To this end, we performed detailed experiments to investigate matrix effects, reaction time, limit of detection, interference effects and bias and precision in saliva samples.
To test the matrix effects on assay performance, we mixed eight different saliva samples collected from healthy and drug-free donors of diverse gender and ethnic backgrounds as a pooled matrix. The pooled matrix was then spiked with different concentrations of cocaine (0 to 500 μM) and diluted with binding buffer 1:1 (50%) or 1:9 (10%) before being applied to the CBSA-5335-based fluorophore/quencher assay. Our results showed that the 10% dilution resulted in a higher signal gain with a broader dynamic range (0–100 μM in the initial saliva sample) compared to the 50% dilution (0–25 μM in the initial sample) (Fig. 6A). Additionally, the 10% dilution compensates for variations (such as salt concentration and pH) in individual saliva samples, and we therefore used 10% dilutions for subsequent experiments.
We also monitored the time course of our CBSA-5335-based fluorophore/quencher assay in the 10% pooled saliva matrix. We found that the fluorescence signal greatly increased with the increase of reaction time upon the addition of 1, 5 or 10 μM cocaine and 85% of the maximum signal was obtained after 15 min (Fig. 6B). Clearly, the fast reaction time of our assay is suitable for on-site detection due to the rapid assembly of the cocaine–CBSA complex in saliva samples.
The CBSA-5335-based fluorophore/quencher assay also demonstrates excellent performance in target detection in complex samples. Studies have shown that cocaine concentrations are generally higher in saliva than serum within the first few hours of administration,44,45 and the European Union's Driving Under the Influence of Drugs, Alcohol and Medicines (DRUID) program identified 510 nM as the recommended cut-off sensitivity for road-side screening of cocaine in undiluted saliva.44 To determine the sensitivity of our assay, we generated a calibration curve in 10% saliva samples, obtaining a linear range from 0 to 10 μM and a measurable LOD of 50 nM (signal gain 3.3 ± 0.8%, Fig. 7A and ESI, Fig. S9†). This suggests that, accounting for the ten-fold sample dilution, our assay can meet the recommendations established by DRUID for on-site detection of cocaine, with a detectable LOD equivalent to 500 nM in undiluted saliva. In contrast, split aptamers containing a single binding domain have previously achieved LODs of 30 nM in 0.5% saliva,46 5 μM,47 and 3.8 μM48 in 25% saliva (ESI, Table S1†). Two factors contribute to the high sensitivity of our CBSA assay. First, the low thermo-stability of the split aptamer greatly suppresses non-specific assembly of the CBSA fragments. Second, the cooperative binding from the two target-binding domains significantly increases the CBSA's affinity.
We then investigated the specificity of the CBSA-5335-based fluorophore/quencher assay for cocaine versus structurally-similar and -dissimilar interferents in saliva. Benzoylecgonine (BZE), anhydroecgonine methyl ester (MEG), and cocaethylene (EC) are major structurally-similar metabolites of cocaine that are secreted into oral fluids.49 We tested our CBSA-5335-based assay with high concentrations of these metabolites as well as nicotine (NIC), since tobacco is widely used among cocaine users. We found that our CBSA assay showed excellent cocaine specificity: our results demonstrated no measurable signal from 50 μM of BZE, MEG, or NIC and only 19% and 3% cross-reactivity to 50 μM and 5 μM EC, respectively, in 10% saliva (Fig. 7B). The results are consistent with previously reported assays17,20 based on single-domain cocaine-binding split aptamers, demonstrating that CBSA retains excellent specificity for its target molecule in saliva. Finally, we tested the bias and precision of our CBSA assay by spiking cocaine at low, medium, and high concentrations into 10% saliva samples (final concentration of 1, 5 or 10 μM) from eight different individuals (ESI, Fig. S10A†). Using the pooled saliva as a standard, the average bias of signal gain obtained in these individual samples was 12.7%, −0.4% and −5.8% for 1, 5 and 10 μM cocaine, respectively. At 1, 5 and 10 μM, the coefficients of variation (CV) within samples were 7.1%, 5.2% and 9.0%, respectively, and the CV between runs were 7.3%, 5.1% and 8.5%, respectively (ESI, Fig. S10B†). Thus, the bias and CV were consistently below the acceptable cut-off (20%) for drug-screening methods,43 further demonstrating the immediate feasibility of our CBSA-based assay for on-site drug screening.
Our CBSA-based assay can be generalized in terms of both targets and sensor platforms. Stojanovic et al. recently isolated new aptamers for four different steroids via heterogeneous SELEX using partially randomized DNA libraries with a pre-designed, three-way-junction binding domain.50 We believe that such a selection strategy could similarly be applied for the isolation of aptamers for other drugs of abuse as well as clinically relevant targets such as small-molecule biomarkers, toxins and therapeutics. Based on Heemstra's general approach to split three-way-junction structured aptamers51 and the simplicity of engineering a CBSA from a split aptamer, it should be straightforward to develop other CBSA-based assays from the isolated aptamers for which the target-binding domain has been identified. Since many optical and electrochemical sensing strategies have been developed that employ target-induced split aptamer assembly,21,52–55 it should be feasible to integrate our CBSA into these platforms. We also foresee the potential to improve the performance of our CBSA-based assay by employing signal amplification techniques53–57 that could further enhance its sensitivity.
Footnote |
† Electronic supplementary information (ESI) available: Optimization of Mg2+ and ATMND concentrations for our CBSA-based ATMND-binding assay; ATMND-reported calibration curve for CBSA-5325 at various cocaine concentrations; ATMND binding affinity for the cocaine-assembled CBSA-5325; KD of 38-GC and different 38-GC mutants for cocaine as characterized by ITC; stem length effects on cocaine-induced CBSA assembly; spectra of CBSA-5335-based fluorescence detection of cocaine in 1× binding buffer; characterization of cocaine binding affinity of CBSA-5335 and PSA using ITC; fluorescence detection of cocaine in saliva with our fluorophore/quencher modified CBSA-5335; calibration curve of our CBSA-5335-based fluorophore/quencher assay in 1× binding buffer and 10% saliva at cocaine concentrations ranging from 0 to 10 μM; bias and precision of the CBSA-5335-based fluorophore/quencher assay; comparison of amplification-free split-aptamer assays for cocaine detection; sequence ID and DNA sequences used in this work. See DOI: 10.1039/c6sc01833e |
This journal is © The Royal Society of Chemistry 2017 |