Temperature and Electrolyte Optimization of the α-Hemolysin Latch Sensing Zone for Detection of Base Modification in Double-Stranded DNA

Abstract

The latch region of the wild-type protein pore α-hemolysin (α-HL) constitutes a sensing zone for individual abasic sites (and furan analogs) in double-stranded DNA (dsDNA). The presence of an abasic site or furan within a DNA duplex, electrophoretically captured in the α-HL vestibule and positioned at the latch region, can be detected based on the current blockage prior to duplex unzipping. We investigated variations in blockage current as a function of temperature (12–35°C) and KCl concentration (0.15–1.0 M) to understand the origin of the current signature and to optimize conditions for identifying the base modification. In 1 M KCl solution, substitution of a furan for a cytosine base in the latch region results in an ∼8 kJ mol⁻¹ decrease in the activation energy for ion transport through the protein pore. This corresponds to a readily measured ∼2 pA increase in current at room temperature. Optimal resolution for detecting the presence of a furan in the latch region is achieved at lower KCl concentrations, where the noise in the measured blockage current is significantly lower. The noise associated with the blockage current also depends on the stability of the duplex (as measured from the melting temperature), where a greater noise in the measured blockage current is observed for less stable duplexes.

Introduction

The wild-type pore-forming toxin α-hemolysin (α-HL) has been studied extensively over the past decade as a platform for ion-channel recordings of single-stranded DNA (ssDNA) (1–7). By applying a potential difference across an α-HL pore that is embedded in a lipid membrane, one can drive DNA electrophoretically from one side of the pore to the other. The current is recorded as a function of time, and translocations of the individual DNA strands are observed as events in which the current momentarily decreases (8,9). The extent of this current change is dependent on the sequence of the DNA near the tightest constriction of the protein channel, which is comparable to the diameter of ssDNA.

Although ssDNA can translocate through α-HL, double-stranded DNA (dsDNA) is not able to because its diameter (∼2.0 nm) is larger than the narrowest constriction of the pore (∼1.4 nm) (10). However, it is possible to capture dsDNA in the α-HL vestibule, and this technique has been used to interrogate dsDNA hairpins within the vestibule of α-HL to reveal structural composition (11,12), to study escape kinetics (13), and to probe the electrical potential distribution within the α-HL protein pore (14). With an appropriate applied voltage (120 mV), dsDNA will unzip (denature) into its constituent components (11,15–19) (Fig. 1 A). In a typical experiment, dsDNA with an additional single-stranded poly-T tail is driven into α-HL from the cis (vestibule) side of the channel. The duplex is driven down to the 1.4 nm constriction (15) that separates the vestibule from the β-barrel, through which the duplex cannot pass. The electrophoretic driving force causes the double-stranded section to unzip into its constituent components.

Figure 1.

(A) Unzipping of dsDNA in the α-HL nanopore. DNA denatures or unzips into its constituent components when a voltage of 120 mV is applied across the pore. (B) Upon capture, the duplex section is driven down to the central 1.4 nm constriction and pauses momentarily before unzipping. (C) Replacing a cytosine base by a furan group in proximity to the latch constriction results in an increase in measured current before unzipping. The α-HL structure was taken from PDB 7AHL, and the DNA structure is shown on a 1:1 scale with α-HL (10). To see this figure in color, go online.

The time it takes for the unzipping to occur, which is on the order of milliseconds, is dependent on the length and composition of the DNA (19,20) and correlates with the stability of the duplex (15,19,20). In previous studies (15,21), we used this feature to identify the presence of damage sites in DNA that destabilize the duplex.

The decrease in current while dsDNA is captured in the pore, relative to the current measured through an open channel, is a result of the blocking contributions from both the single-stranded and double-stranded sections of DNA. Although the majority of the current is blocked by the poly-T tail that resides in the β-barrel during unzipping, the double-stranded section that resides in the vestibule also contributes to the current blockage. In our previous work (15,21), we showed that removing a single base in the duplex section (by replacing it with a furan group or abasic site) can result in an increase in measured current through the pore. The precise change in current is dependent on the position of the missing base in the sequence relative to the latch constriction of the α-HL protein pore (Fig. 1 C).

In this article, we demonstrate that the difference in observed current can be attributed to the differences in activation energy for transport of the electrolyte (K⁺ and Cl⁻) through the latch region of α-HL during dsDNA residence inside the pore. Through temperature- and KCl-dependent measurements, we show how identification of a furan group at the latch can be optimized at lower temperatures and lower KCl concentration. In particular, the noise associated with the current measured during unzipping increases with increasing KCl concentration. Finally, we demonstrate for the first time (to our knowledge) that the noise associated with the current measured during unzipping is dependent on the stability (as measured from the melting temperature) of the DNA duplex. We ascribe this finding to breathing of the duplex within the vestibule. Whereas previous efforts to obtain nanopore measurements of DNA have focused on the translocation of ssDNA (1–7), the ability to effectively exploit the latch sensing zone offers exciting possibilities for characterizing and sequencing the more biologically relevant dsDNA along with its modifications.

Materials and Methods

DNA preparation and purification procedures

DNA was prepared from commercially available phosphoramidites (Glen Research, Sterling, VA) by the DNA Core Facility at the University of Utah. Afterward, DNA oligomers were cleaved from the solid support and deprotected according to the manufacturer’s protocol, followed by purification using an ion-exchange high-performance liquid chromatography column running a linear gradient of B from 25% to 100% over 30 min and monitoring of UV absorbance at 260 nm (A = 20 mM NaP_i, 1 M NaCl, pH 7 in 10% CH₃CN/90% ddH₂O, B = 10% CH₃CN/90% ddH₂O, flow rate = 3 mL/min). Sequences used in this study are shown in Table S1.

Chemicals and materials used for nanopore measurements

All buffer solutions used were prepared as 10 mM phosphate (pH 7.5), with a KCl concentration as indicated. Wild-type α-HL was purchased from List Biological Laboratories in the monomer form of lyophilized powder and dissolved in water at 1 mg/mL. 1,2-Diphytanoyl-sn-glycero-3-phosphocholine (DPhPC) was dissolved in decane at 10 mg/mL and used to form the bilayer. The bilayer was supported by a glass nanopore membrane (GNM), the fabrication of which has been described previously (32). Glass nanopore membranes were modified with 2% (v/v) (3-cyanopropyl) dimethylchlorosilane in acetonitrile to create a moderately hydrophobic surface. The DNA duplexes were annealed by mixing the 41-mer and 17-mer at a 1:2 mol ratio, followed by heating in a 90°C water bath for 5 min and then cooling to room temperature over 3 h.

Current-time recordings

Current-time (i-t) recordings were performed using the low-noise Nanopatch system (Electronic BioSciences, San Diego, CA), which is also capable of controlling temperature to an accuracy of ±0.5°C. Temperature control was achieved by a thermoelectric peltier embedded underneath the solution reservoir (outside the capillary) and measured by a K-type thermocouple in contact with the solution. The temperature was permitted to equilibrate for 5 min before measurement. The KCl solution was used as the electrolyte to fill the solution reservoir and the GNM capillary. A voltage was applied across the GNM between two Ag/AgCl electrodes placed inside and outside of the capillary. As previously described, a lipid bilayer was deposited across the GNM orifice as indicated by a resistance increase from ∼10 MΩ (associated with the open GNM) to ∼100 GΩ (32). A pressure of 60–80 mmHg was applied to the inside of the GNM capillary via a syringe, allowing the lipid bilayer to be functional for the protein channel reconstitution (32). Next, 0.2 μL of α-HL monomer solution at 1 mg/mL was added to the cis side of the GNM (a volume of 350 μL). The duplex DNA (15 μM) was added to the solution reservoir after protein reconstitution into the lipid bilayer, which was indicated by a single jump in current by ∼1 pA /mV at 25°C. A voltage of 120 mV was applied trans vs. cis (i.e., cis negative). The i-t traces were filtered at 10 kHz and sampled at 50 kHz.

Data collection

Based on previous reports (19), I-t blockades that lasted longer than 2 ms were identified as DNA unzipping events. Shorter events were attributed to translocation of excess single-stranded DNA (ssDNA) and/or collision of the DNA with the protein surface. The current amplitude of each blockade was used to determine the identity of the duplex, as described in the Results and Discussion. Events were extracted using QuB (version 1.5.0.31). Histograms of current, noise, and unzipping durations were generated and plotted using Origin Pro (version 9.0). Details of error treatment are given in the Supporting Material.

Results and Discussion

The DNA sequence chosen for study was a portion of the KRAS gene, 5′-(T)₂₄-TGGAGCTGCTGGCGTAG, with the poly-T tail added to assist in threading through the nanopore. This sequence is of interest because damage that gives rise to point mutations within this gene can lead to uncontrolled cell growth and human carcinoma (22). The 41-mer sequence of interest was hybridized to a complementary 17-mer to form a 17 bp duplex section. The capture and unzipping of these DNA duplexes within α-HL were continuously monitored and the measured current of each event was used to construct histograms of the blocking current and noise.

When a single base was removed from the DNA sequence at the latch region and replaced with a furan (Fig. 1 C), the measured current through the pore increased relative to the fully complementary sequence (Fig. 2). The measured current difference between the complementary duplex and the duplex containing a furan at the latch region was ∼2 pA.

Figure 2.

Resolving the presence of a single furan, an abasic mimic, in a DNA duplex. Current histograms and representative current-time traces illustrating the observed unzipping event currents for (A, red line) the duplex containing a furan group at position 9F, (B, blue line) the reference duplex (containing no modifications), and (C) both the furan-containing and reference duplexes. (D) The approximate position of the furan group relative to the latch region in α-HL is shown for the reference and furan-containing duplexes. An extended current-time trace is shown in Fig.S1. To see this figure in color, go online.

Defining the extent of the latch sensing zone

In an earlier publication (21), we demonstrated that moving the position of the furan altered the observed current relative to the reference. These preliminary experiments were performed in a buffer containing 150 mM KCl, which we chose because the initial studies focused on monitoring the uracil-DNA glycosylase (UDG)-catalyzed conversion of a uracil base to an abasic site, and the UDG enzyme is catalytically inactive at higher KCl concentrations (19). Such a low KCl concentration is unusual in nanopore measurements employing α-HL, because the rate of capture of DNA is significantly reduced by repulsive electrostatic interactions between negative amino acid residues at the vestibule opening and the phosphate backbone of DNA (9,23). These interactions are much more effectively screened at higher KCl concentrations.

We studied the difference in blocking current, ΔI, as a function of sequence position between a DNA duplex modified with a single-furan site and an unmodified dsDNA reference containing standard Watson-Crick basepairs (Fig. 3). The position of the furan modification in the DNA was varied between 6 and 13 bp from the 3′ end of the 17-mer (i.e., 6–13 bp cis to the central constriction). The position of the furan site within the latch region of α-HL before unzipping determines the blocking current, defining a sensing zone that spans ∼6 bp (∼2 nm). At the center of this sensing zone, the differences in blocking currents between the furan-containing and reference duplexes reaches a maximum of just over 2 pA, corresponding to a readily detectable 1.6% change when the unzipping event current is normalized relative to the open-channel current (ΔI/I_o). On either side of this maximum, the current difference tends toward zero over 2–3 bp.

Figure 3.

Mapping the resolution of the latch sensing zone within α-HL in a 1 M KCl solution. The current was monitored for a series of duplexes in which the position of the furan group was moved systematically through the sequence while maintaining a guanine opposite to the furan. The differences in the blocking current ΔI for the furan-containing duplex relative to the (fully complementary) reference duplex, normalized to the open-channel current, I₀, are plotted versus the position of the furan. Current histograms from which these values were calculated are shown in Fig S2. The approximate position of the furan substitution relative to the α-HL vestibule is shown. Error bars are given as the standard error of the mean. The α-HL structure was taken from PDB 7AHL (10), and the DNA structure is shown on a 1:1 scale with α-HL. To see this figure in color, go online.

ΔI values recorded in 1 M KCl (Fig. 3) are similar to values previously measured in 0.15 M KCl (21). This is an important finding because it demonstrates the general applicability of the latch sensing zone for interrogating DNA structure at electrolyte concentrations commonly used for ion channel recordings.

Assuming that the distance between basepairs remains ∼0.34 nm in the vestibule under an electric force (24), this range of basepairs corresponds to a distance of 2.0–4.4 nm, placing the furan group at a narrowing of the protein vestibule during unzipping, as shown in Fig. 1 B. This latch site, as first described in detail by Song et al. (10), is ∼2.6 nm in diameter, which is comparable to the diameter of dsDNA (2.0 nm).

Vercoutere et al. (11) previously identified structural changes in DNA hairpins using the vestibule of α-HL. In one experiment, they observed a 2 pA decrease in measured current when a dsDNA hairpin was altered only by adding an additional T to the loop section. This loop section is situated at a distance equivalent to 6–7 bp from the central constriction of α-HL, placing it at the edge of the latch constriction during vestibule residence. Their observation is in agreement with our results, which showed that placing an additional T in the latch region of α-HL contributed to the decrease in measured current, whereas the presence of a missing base (furan group) resulted in a current increase.

Effect of temperature

We performed temperature-dependent measurements to better understand the origin of the increase in current when a furan was placed into the proximity of the latch region. As the temperature was increased from 12°C to 35°C, the currents of both the open channel and the unzipping events increased, in accordance with the expected increase in ion mobilities (23). Over this temperature range, the tertiary structure of the protein is believed to be stable (25). However, the current difference between the duplex containing a furan at the latch (position 9F) and the duplex with a furan situated outside the latch (position 13F) became smaller with increasing temperature, as shown in Fig. 4.

Figure 4.

Effect of temperature on the observed current during unzipping of two duplexes: one that has a furan group situated at the latch region of α-HL during unzipping (position 9F) and one that does not (position 13F). (A–E) Representative temperature-dependent current histograms illustrating the change in the current (and current difference) measured during unzipping. (F) Change in the current during unzipping as a function of temperature for position 9F (squares) and position 13F (circles), and change in current for the open channel (inset). (G) I-t traces as a function of temperature. Events <2 ms in duration are attributed to excess DNA and/or collision of the DNA with the protein surface.

We explored the origin of the dependence of the current on the position of the furan by measuring the activation energy (E_A) for electrolyte transport through the nanopore. The activation energy reflects the resistance to electrolyte transport when dsDNA is introduced into the vestibule. We determined E_A by measuring I as a function of temperature and analyzing the data using the linearized version of the Arrhenius equation:

$$\ln \left(I\right)=-\frac{E_A}{RT}+\ln \left(C\right)$$ (1)

where R is the gas constant, T is the temperature, and C is the preexponential factor.

Arrhenius plots for the open channel, as well as the two duplexes with the furan group situated inside and outside of the latch region (positions 9F and 13F, respectively), were constructed from the data in Fig. 4 and are shown in Fig. 5. The calculated activation energy for the open channel measured at an applied potential of 120 mV was 17.1 ± 0.6 kJ mol⁻¹. This is comparable to the activation energy for diffusion of KCl in bulk solution (17.7 kJ mol⁻¹) (26).

Figure 5.

Determining the activation energy of electrolyte transport from an Arrhenius plot. Changes in current as a function of the reciprocal temperature during unzipping of DNA duplex containing a furan at position 9F (squares) or position 13F (circles), and for the open-channel current (inset) are shown. Details of the sequences studied are given in Fig. 3.

A significant increase in the activation energy for electrolyte transport (relative to the open channel) was observed when dsDNA was situated inside the pore (i.e., during unzipping). This reflects a decrease in the mobility of ions within the pore, caused by the presence of DNA that occupies a significant volume of the vestibule and/or specific interactions of the ions with the DNA duplex. The activation energy was also highly dependent on the position of the missing base within the sequence. When the furan group was situated inside the latch region of α-HL, the activation energy at (120 mV) was significantly lower (19 ± 1 kJ mol⁻¹) than when it was situated outside the latch region (27 ± 1 kJ mol⁻¹). The activation energy under the same conditions for the perfectly complementary reference duplex (containing no furan sites) was found to be identical (29 ± 2 kJ mol⁻¹) to that of the duplex with the furan positioned outside the latch region (Fig. S3). We speculate that the decrease in activation energy reflects less steric hindrance at the constriction site, where removal of a base opens a passage through which the ions can move more freely. Our data predict that at higher temperatures (>35°C), the difference in the measured currents for the two duplexes will be inverted. That is, a duplex with a furan at position 9F would become more blocking than a duplex with a furan position 13F. Although this effect is interesting, measurements beyond 35°C are challenging because of the decreasing residence time of the DNA at higher temperatures. The exponential increase in the unzipping rate results in a much poorer signal/noise ratio and prohibits reliable determination of the blockage currents.

It is also clear that measurements obtained at lower temperatures have a distinct advantage for resolving the presence of a missing base in the latch region because they show the greatest difference between currents of the furan-containing duplex and the reference (Fig. 4 F), as well as lower noise (Fig. 4 G). However, this advantage is balanced against the increased length of the unzipping event time, which increases exponentially as the temperature decreases (Figs. 4 G and S4).

Effect of KCl concentration

A series of unzipping experiments using duplexes containing a furan positioned in the latch (position 9F) and distal to the latch (position 13F) of α-HL, as shown in Fig. 3, were conducted over a range of KCl concentrations between 0.15 and 1 M to further optimize detection of abasic/furan sites. The objective of these experiments was to identify how the difference in blocking current between two duplexes varies as a function of KCl concentration.

It is known that the current of an open α-HL channel increases as a function of electrolyte concentration (21). Although the effect of KCl concentration on ssDNA translocation has been studied previously (23), the dependence of blockage current on electrolyte concentration before dsDNA unzipping has not been reported.

The measured current during unzipping for both duplexes increased with an increase in KCl concentration up to ∼0.3 M, at which point the current remained fairly constant (Fig. 6). Below 0.15 M (outside the measurable range), the current is expected to decrease toward a limiting value determined by the counterions associated with the fixed charges on the wall of the protein and/or DNA duplex. The observed leveling-off in the current occurred at a concentration similar to that previously observed for ssDNA translocation (23). The number of current-carrying ions inside the pore during DNA unzipping is only weakly dependent on the external KCl concentration between 0.3 and 1 M. We speculate that this due to exclusion of anions from the pore by the negative charge of the sugar-phosphate backbone. This is in contrast to the behavior of the open-channel current, which increased linearly with increasing KCl concentration over the range studied (inset of Fig. 6).

Figure 6.

Effect of KCl concentration on the blocked channel current during dsDNA unzipping of a DNA duplex containing a furan at position 9F (squares) and position 13F (circles). Inset: the open-channel current increases linearly as a function of KCl concentration. Currents were recorded at 120 mV in a 10 mM phosphate (pH 7.5) solution containing 0.15–1 M KCl. Current histograms for each KCl concentration are shown in Fig. S5.

Further evidence for a constant number of ions in the α-HL channel above 0.3 M KCl may be inferred by analyzing the time τ taken to unzip the DNA duplex inside the α-HL channel. We find that τ remains constant as a function of KCl concentration between 0.3 and 1 M (Fig. S6 and S7). If the KCl concentration inside the protein pore increases with increasing bulk KCl concentration, and assuming that the stability of DNA as a function of KCl concentration in the confined geometry of α-HL is similar to that in bulk solution, then an increase in unzipping times with increasing KCl concentration is to be expected. This is because the unzipping time is related to the stability of the DNA duplex and the latter increases as a function of KCl concentration due to more effective screening of adjacent negatively charged phosphate groups (27).

Effect of electrolyte concentration on current noise

A single blockage current event has an associated noise. For example, if a blockage current is reported as −21.6 ± 0.8 pA, the noise associated with that one event is 0.8 pA. The noise from many events can be used to construct a histogram of blocking current noise (Fig. 7, A and B). In general, we find that the noise distribution for unzipping events is skewed right toward higher noise. The average noise associated with the unzipping of a duplex can be estimated from the median of this distribution.

Figure 7.

Representative histograms of the noise in the measured unzipping currents for a duplex with a furan at (A) position 9F and (B) position 13F in 0.75 M KCl. (C) Open-channel noise distribution (the noise of the current measured between unzipping events). The dashed red lines denote the median of the noise distribution. (D) Effect of KCl concentration on the median of the noise for a furan at position 9F (squares), position 13F (circles), and open-channel current events (inset). Histograms of the event noise at each KCl concentration are shown in Fig. S8.

The noise associated with unzipping was examined as a function of KCl concentration (Fig. S8) for the two duplexes with a furan situated outside of the latch region during unzipping (position 13F) and inside the latch region (position 9F). The noise associated with open-channel events (i.e., the events between DNA capture) was also measured (Fig. 7, inset). The event rate for ssDNA is greatest at high KCl concentrations (23). However, this is not necessarily the optimum electrolyte concentration in which unzipping experiments should be performed. This is because the median of the noise was found to increase with the increasing KCl concentration (Fig. 7).

At 150 mM KCl, the median of the blockage event noise (for DNA with a furan at position 13F) is 1.23 ± 0.07 pA. This increases to 1.9 ± 0.1 pA at 1 M KCl. The open-channel current event noise in this particular experiment increased to 2.25 ± 0.09 pA. Although the results presented earlier show that the latch region is capable of resolving the presence of a furan at 1 M KCl, it is clear that optimal resolution is obtained in KCl concentrations ≤ 0.5 M, where the noise in the measured current is lower. Clearly, it is important to consider both the KCl concentration and temperature when choosing the measurement conditions for identifying abasic sites in DNA. Optimal conditions are anticipated at low electrolyte concentration and low temperature, where current-peak separation is at a maximum and the noise in the current is at a minimum. At high electrolyte concentration and high temperatures, it would be extremely challenging to resolve the presence of a furan.

The noise in the measured current during unzipping also appears to be dependent on the stability of the duplex. In Fig. 7 D, the duplex with the furan at position 9F (black line) exhibits a consistently higher noise than the duplex with the furan at position 13F (red line) between 0.15 and 1 M KCl. The melting points of these two duplexes are 60.7°C ± 0.6°C and 64.6°C ± 0.6°C, respectively, suggesting that the noise level in the unzipping event current is higher for less stable duplexes.

To verify this observation, we measured the melting temperatures of all the duplexes used in the mapping studies presented above and analyzed the noise of the unzipping events for each one (Fig. S9). Fig. 8 shows the change in the median of the unzipping event noise normalized with the respect to the median of the open-channel event noise as a function of duplex stability.

Figure 8.

Effect of melting temperature on the noise of the unzipping event currents. The median of the noise associated with each unzipping event current, normalized to the noise of the open-channel current, increases as the stability of the duplex decreases. A constant temperature of 25°C was used for all nanopore measurements. Error bars on the y axis are the standard error of the median, and error bars on the x axis are twice the standard deviation of three averaged melting-point experiments. Histograms of the noise of the unzipping events for each duplex studied are given in Fig. S9.

An increase in the noise of the unzipping events is observed as the duplexes become less stable. DNA is a dynamic structure and it is reasonable to assume that during occupation of the α-HL channel, the DNA undergoes local conformational changes. Such conformational changes are important for DNA breathing, which leads to the breaking of localized basepairs and the formation of ssDNA regions (DNA bubbles) with a lifetime of ∼100 μs (28–30). For less stable duplexes (as measured by the melting temperature), these conformational changes are likely to be larger and more frequent when the DNA is situated inside the α-HL channel before unzipping. We speculate that these fluctuations in local conformation contributed to the measured current noise.

Conclusions

We have demonstrated that the latch region of α-HL constitutes a sensing zone that is capable of detecting a furan group in dsDNA over a range of 4–5 bp. The presence of a furan group in the latch region during unzipping gave rise to an increase in current relative to the unzipping of a duplex with a fully complementary sequence.

The results presented here demonstrate that one can optimize the detection of mutation-causing abasic sites in the KRAS sequence via nanopore unzipping by using low-electrolyte-concentration (150 mM) and low-temperature (12°C) conditions, in which the noise associated with individual events is at a minimum and the current signature resolution between an abasic site-containing duplex and the reference is at a maximum. A potential approach to further optimize detection would be to reduce the DNA breathing effect, which leads to noise in individual events. Potentially, DNA breathing could be decreased with the use of a DNA probe constructed from an analog such as peptide nucleic acid, which forms a more stable structure (31). Clearly, the latch constriction of α-HL, which is specific to dsDNA, offers the potential to study and detect site-specific changes in duplex structure of biological relevance, such as mutations and damage to individual bases.