Sequence-specific Single-molecule Analysis of 8-Oxo-7,8-dihydroguanine Lesions in DNA based on Unzipping Kinetics of Complementary Probes in Ion Channel Recordings

Introduction

Detection of chemical damage to genomic and mitochondrial DNA remains an important and challenging task.¹ In an ideal case, one would determine the precise chemistries and locations of all modifications occurring on individual DNA strands in healthy vs. diseased cells and be able to monitor such changes as a function of oxidative, alkylative, or micronutrient stress.^2,3 In the case of the key base oxidation product 8-oxo-7,8-dihydroguanine (OG), the most sensitive and accurate methods of quantification involve either the comet assay to produce strand breaks at each lesion or digestion followed by LC-MS analysis.⁴ Shortcomings of these procedures include the loss of sequence information surrounding the lesion and the difficulty in distinguishing between multiple lesions on one strand relative to the average of a set of strands. We therefore sought a single-molecule method that would overcome these limitations.⁵

As a step towards these goals, we report the ability of short oligodeoxynucleotides to interrogate a target strand of DNA in a sequence specific manner for the presence of OG vs. G. Our strategy is based on measuring the characteristic time for unzipping the duplex formed by the target and a short oligodeoxynucleotide probe in an α-hemolysin (α-HL) ion channel. Other laboratories have shown that the time required for electrophoretically-driven translocation of single-stranded DNA (ssDNA) oligomers through the α-HL nanopore is greatly increased when a terminal hairpin is present which must unfold before threading through the narrow constriction zone of α-HL.^6–11 Alternatively, addition of a small complementary DNA probe sequence also increases τ by providing a quasi-stable duplex which must unzip before translocation of the strand can be completed.^12–16 We chose the latter approach in designing a detection method for OG vs. G in a specific target sequence (Figure 1). We demonstrate that by monitoring the unzipping time constant τ, a specific target sequence containing G at a position of interest can be differentiated from a duplex containing OG at the same position. This capability results from the difference in duplex stability imparted by the presence of either G or OG and is reflected in τ. In concept, this method could be used to detect multiple DNA lesions within a given DNA target strand as each lesion will uniquely influence the duplex stability and thus τ. In the studies presented here, single OG lesions were examined to demonstrate the ability to detect single nucleotide damage and to determine the relative location of the lesion based on 3′ or 5′ unzipping.

A unique feature of OG is its ability to base pair to either C or A with nearly equal stability, compared to the parent base G, which forms a highly stable base pair only with C. NMR and x-ray crystallographic studies provide the base pair structures shown in Figure 2.^17–22 The Watson-Crick OG:C base pair, in which both bases are in the normal anti conformation about the glycosidic bond, is slightly destabilized compared to the G:C base pair because of steric interactions between the C8 oxo group and the C4′ oxygen of the same nucleotide; typical duplex melting temperatures are ~2 °C lower for the OG:C pair.²³ The adverse steric interaction may be relieved by rotating OG to a syn conformation which presents the Hoogsteen face of OG for pairing with A via two complementary H bonds rather than three. The OG:A base pair usually has a T_m lowered by another 1–2 °C compared to OG:C. In contrast, the presence of a G:A mismatch is significantly more destabilizing to a short duplex (ΔT_m = 6–8 °C) because two purines in anti conformations require a wider backbone spacing (Figure 2, red arrow).²⁴ Alternative structures for the G:A mismatch may be stabilized in tandem mismatches that provide greater π stacking or in other conformations that require low pH;^22,25–27 however, these conditions are not present in the single X:Y base (mis)pairs of the current study.

Results and Discussion

In the present work, we took advantage of the difference in stability of G vs. OG opposite C or A to design short complementary probes that would slow the translocation of ssDNA through the α-HL ion channel. Experiments were conducted initially with a set of 10 mer (10C or 10A) and 15 mer (15C and 15A) probes complementary to the central region of a 65 mer (denoted as G or OG) resulting in the following eight duplexes: G:10C or A, OG:10C or A, G:15C or A and OG:15C or A, as shown in Table 1. (Sequences were selected to minimize secondary structure.) T_m studies conducted in the same buffer solution (10 mM PBS, pH 7.4, 1 M KCl, 1 mM EDTA) as used for electrical measurements of the nanopore showed the general trends expected for stability of a single X:Y mispair compared to a native G:C pair. The order of stability was found to be G:C > OG:C > OG:A > G:A for both 10 mer and 15 mer probes annealed to the 65 mer target strand. The 10 mer probes were more sensitive to the presence of a single-base mismatch; the T_m for a G:A mismatch was 8 °C lower than the parent G:C in the 10 mer but only 5.5 °C lower in the 15 mer duplex.

Electrical measurements were employed to determine the duplex stability as a function of the unzipping time constant τ. To perform these measurements, a voltage was applied across an α-HL ion channel reconstituted into a lipid bilayer suspended across the orifice of a glass nanopore membrane (GNM).^5,28–33 The annealed duplex DNA was then electrophoretically driven into the α-HL channel where it unzipped for translocation (Figure 1). The analysis of duplex unzipping kinetics was performed by plotting histograms of log₁₀(events) vs. event duration time. From the slope of these histograms, the translocation time constant τ was determined for each duplex by assuming that unzipping follows a first-order kinetics law. Figure 3 shows an example of fitting for G:10C (or A) and OG:10C (or A) for +80 mV applied voltage (trans vs. cis), and the resulting values of τ. Histograms used for the determination of τ for the other DNA duplexes, and at different applied voltages, can be found in the Supporting Information. Probes of 25 nucleotides in length (5′-AAAAAAACATCTGA(C or A)GGCTCAAAAA) were also examined, but due to the very stable nature of the resulting duplex, an adequate sample of events could not be collected for voltages <140 mV; thus, 25 mer probes were not studied further. Measurements of the translocation of 65 mers G and OG in the presence of C or A-containing complementary probes indicated a correlation between the values of τ and T_m. Figure 4 shows the time constant for unzipping as a function of the melting temperature for 10 mer and 15 mer duplex sequences. As seen from this data, the unzipping time decreases in the order G:C > OG:C > OG:A > G:A, and correlates well with duplex melting temperatures, for both 10 and 15 mer probes. The duplexes generated with the 10 mer probes exhibited unzipping times that were distinct from one another at +80 mV, but at higher voltages (+100 mV and +120 mV) unzipping proceeded at similar rates, decreasing the measurement sensitivity, probably due to the inherent instability of 10 mer duplexes. For the 15 mer duplexes, use of a lower voltage (+80 mV) resulted in event durations that were very long (> 10 s for G:15C), making it difficult to collect an adequate population of events to distinguish between the different duplexes, and at higher voltages again the unzipping times show less sensitivity to the duplex stability. From this initial study it was determined that the probe design must balance the sensitivity, stability and the longer τ values of the 15 mer that limit event frequency. This process of probe refinement is described below.

Although the 10 mer probes displayed T_m values that were more sensitive to the presence of a single X:Y mispair, the τ values for the 15 mer probes were generally in a better event duration range for discriminating between G and OG. Duplexes with T_m values >50 °C provided sufficiently long translocation times that could be readily discerned from translocations of the 65 mer (~280 μs, see Supporting Information). In order to refine the probe design, we sought methods to enlarge the difference in event duration between OG:A and G:A mispairs. Previous studies have shown that locating the mismatch further away from the center of the probe led to a greater distinction in T_m values between correctly paired and mismatched duplexes.³⁴ Thus, three new sets of 12 mer probes were tested in which the X:Y base pair of interest was located in the middle or near either the 3′ or 5′ end of the probe, and care was taken to maintain the same number of G:C and A:T base pairs in each of the two sets so that a meaningful comparison could be made. As a further refinement, the G:C base pair immediately 5′ to the G/OG site was converted to a C:I base pair for all 12 mer probes (Table 2). Inosine (I) is the nucleoside form of the hypoxanthine base which lacks the 2-amino group of G and forms a weaker base pair with C. We predicted this substitution might destabilize the adjacent G:A mispair to a greater extent than the OG:A pair; G:A mispairs are known to be sensitive to the surrounding sequence, particularly on the 5′ side of the G site.²⁶ The T_m data for the resulting 12 mer probes annealed to 65 mer target strands are shown in Tables 2A–C, in which the probe is shifted in either the 5′ or 3′ direction or centered on the X:Y pair of interest.

Table 2.

T_m studies of 12 mer complementary probes with various positions of X:Y

Open in a new tab

^a

The probe strand is shifted 5′ (A), centered (B), or shifted 3′ (C) with respect to the interrogated base X. The position of Y is denoted as position 3, 6 or 10, accordingly.

^b

I = inosine

The data in Tables 2A–C confirm that the displacement of the probe sequences with respect to the potential mismatched base pair can influence the thermal stability of the duplexes. Importantly, shifting the probe sequence toward the 3′ direction relative to the target strand (Y in position 10) led to a greater difference between the relatively stable OG:A pairing (ΔT_m = −4.6 °C with respect to G:C, Table 2C) and the unstable pairing of G:A (ΔT_m = −8.3 °C) compared to other probe arrangements. When the probe was shifted in the opposite direction (position 3), the ΔT_m values were smaller (Table 2A).

Electrical measurement of the 12 mers annealed to G or OG-containing 65 mers unzipping and translocating through the α-HL ion channel were conducted at 80 mV to maximize the apparent differences in unzipping times due to the duplex stability. The unzipping time, τ, was determined in the same manner as described above for the 10 mer and 15 mer duplexes; the unzipping events were plotted as histograms of log₁₀(events) vs. event duration (individual duplex fittings and example i-t traces are available in the Supporting Information). A plot of τ as a function of the shift in probe position is presented in Figure 5 for each of the 12 mer duplexes shown in Table 2.

We again found that the unzipping times corresponded well to the melting temperatures; for each position (3, 6, and 10), the τ value reflected the expected duplex stability as shown in the melting temperatures (Table 2): G:C > OG:C > OG:A > G:A. Additionally, the largest difference in τ between the duplexes occurs for position 10 as predicted by the ΔT_m discussed above, demonstrating that the unzipping time τ is a sufficient measure of duplex DNA stability to allow the presence of a single DNA damage site to be detected.

However, these positional studies highlighted another feature of DNA translocation studies in the ion channel. DNA can enter the α-HL vestibule to unzip the duplex from either the 5′ or 3′ orientation. For the generation of the asymmetrical 12 mer probes (position 3 and position 10), the duplex unzipping times are expected to depend on whether the strand enters from the terminus closer to or farther away from the mismatched base pair, as suggested in recent studies of the kinetics of hairpin unzipping.³⁵ The melting temperature, a thermodynamic phenomenon, will not necessarily reflect the kinetics of 5′ or 3′ initiated duplex dissociation.

To examine 3′ vs. 5′ entrance into the α-HL channel, multi-dimensional plots were generated to examine the current blockage level as a function of event duration and event population density. Figure 6 illustrates how the unzipping events are dispersed based on the current blockage level and translocation time. Generally, the duplexes formed with an asymmetrical probe sequence (positions 3 and 10) displayed two well-resolved populations of events separated by both the current blockage level and the event duration, whereas the symmetrical sequences (position 6 probes), did not show any separate event populations based on the event duration. Symmetrical sequences did show some separation of populations based on the current blockage level, however. We interpret the multiple populations as being due to 3′ vs. 5′ entrance into the α-HL channel. It has been shown that for immobilized homopolymer sequences that 3′ and 5′ entrances into the α-HL channel have different current blockage levels, and these may change significantly based on the nucleotide identity within the channel.³⁶ Additionally, as shown above and by previous reports, duplex sequences of different stabilities have different unzipping event durations, with the less stable duplex exhibiting the faster time constant.^10,11,15,37

Because the 65 mer sequence is a heterosequence embedded within a poly-dT background sequence at each end, and based on the orientation of the molecule upon entry into the α-HL channel and the probe sequence, a variable amount of the heterosequence will be within the α-HL channel. Thus, assignment of entry via the 3′ or 5′ end was not made based on the current blockage level, but rather on the event duration. We assign 3′ vs. 5′ entrance of the 65 mer into the α-HL based on the expected stability of the duplex by looking at the first few terminal base pairs. (Note: all 3′/5′ designations are made referring to the orientation of the 65 mer sequence because it enters first.) For example, the position 3 G:C duplex is expected to be more stable for 3′ entrance relative to 5′ entrance as there are two G:C base pairs and one A:T base pair on the 3′ end of the duplex, but three A:T base pairs on the 5′ end (Table 2). Thus the 3′ entrance is expected to produce a longer unzipping time due to the higher G:C base pair content. Examining the position 3 G:C duplex density plot in Figure 6, the deeper blocking events have longer event durations than those of more shallow blocking, indicating that the deeper block corresponds to 3′ entrance whereas the more shallow blocking level with shorter event durations indicates 5′ entrance. The opposite is expected to be true for the position 10 G:C duplex, where the higher terminal G:C base pair content is located at the 5′ end of the duplex, which is expected to produce longer event durations relative to the 3′. From the density plot in Figure 6, the deeper blockage assigned as 3′ entrance possesses shorter event durations relative to the shallower 5′ blockage, consistent with the expected duplex stability of the position 10 G:C duplex. This convention of the deeper blockage corresponding to 3′ entrance and shallower blockages being 5′ entrance was applied to all the duplexes that could easily be separated into two distinct populations. The data for each duplex were separated into two populations, and each population was fit with a single exponential decay to determine the unzipping time constants, τ_3′ and τ_5′, for 3′ entrance or 5′ entrance, respectively. Histograms and data analysis are presented in the Supporting Information. The position 6 OG:C and OG:A duplexes could not be resolved into two populations, and the time constant was assigned to be τ.

Figure 7 compares the time constants for 3′ vs. 5′ entrance as assigned by the expected event duration from the orientation specific duplex stability. This τ_3′ and τ_5′ assignment agrees remarkably well with the predicted unzipping times for 3′ and 5′ entry. For the symmetrical probe sequence where the data can be resolved into two populations (G:C and G:A, position 6), there is little difference between τ_3′ and τ_5′. This is expected as both duplexes possess the same amount of terminal G:C base pair content. It is not understood why two populations are not resolvable for OG:C and OG:A in position 6. Further, for position 3 duplexes, as discussed above, the 5′ entrance is expected to reflect a more stable duplex relative to 3′ entrance. This is observed for position 3 duplexes G:C, OG:C, and OG:A, with τ_3′>τ_5′, but also for G:A, which is not well understood as the G:A mismatch should destabilize the duplex for 3′ entrance. Finally, for position 10 duplexes, the 5′ entrance is expected to be more stable than the 3′ entrance (τ_3′<τ_5′), except for the G:A duplex where the mismatch destabilizes the 5′ end of the duplex relative to the 3′. This prediction is reflected in τ_3′>τ_5′.

It has been previously demonstrated that the entry of ssDNA into the α-HL pore produced two event populations as a result of either 3′ entry or 5′ entry of the DNA strand,³⁶ separated in current blockage level and event duration.³⁸ Additionally, it has been reported that 5′ entry is less favorable relative to 3′ entry due to the tilt of the DNA bases toward the 5′ terminus, making 5′ entry, and subsequent translocation, occur at a lower rate relative to 3′ entry.^12,38,39 Based on this, it was not anticipated that distinct populations would appear as shown in Figure 6, separated by both the event duration and current blockage level, with an applied bias of 80 mV, as an 80 mV voltage bias is near the minimum electrophoretic force required to initiate DNA interaction with the α-HL channel.⁴⁰ Other duplex unzipping experiments have noted multiple event populations distinguishable by the event duration for a single duplex DNA sample,^{11,14,15,37,41} requiring a more complicated analysis than presented above. Unlike the previous duplex unzipping experiments in which the current blockage level was not emphasized in the analysis, and only the temporal separation was directly applied, the analysis presented here uses the current blockage to distinguish populations prior to analyzing the temporal separation. This allowed a simplified mechanism to be applied via single-exponential fits that generally correspond well to the relative 3′ vs. 5′ duplex stability. It is unknown if the duplex region influences the DNA entry into the α-HL channel in a manner different from ssDNA, allowing equal 3′ and 5′ entry rates.

Overall, the data presented in Figure 7 reveal that the relative position of a DNA damage site can be determined by how it influences the stability in an orientation-specific manner. Deviations present in the data from the expected duplex stability are attributed to possible interactions other than hydrogen bond strength between the base pairs and are not fully understood. However, further refinement of the probe design may lead to improved sensitivity to DNA damage, for example by chemical modification within the probe to further stabilize or destabilize the resultant duplex. This approach should be applicable to most sequence contexts surrounding OG; a summary of T_m data for four additional sequences containing these base pairs shows the same order of stabilities, often with higher ΔT_m values (see Supporting Information).

Conclusions

These studies demonstrate that a single oxidized damage site can be detected in a sequence specific location by annealing a probe to the sequence surrounding the damage site and observing the rate of duplex unzipping required for translocation through the α-HL ion channel. The presence of a single OG lesion will influence the duplex stability in a specific manner depending on whether OG base pairs with C or A, both of which produce event duration times distinct from the native G:C and G:A base pairs. Additionally, if the probe is designed so that the DNA damage site occurs on the 3′ or 5′ end of the probe, as opposed to the middle, the duplex stability will vary based on 5′ or 3′ entry into the α-HL channel and will be reflected in the current blockage level as a function of the event duration, producing two distinct populations of events. This ability to discern the relative location of the DNA damage site is a first step towards location-specific DNA damage detection. Further tailoring of the probe sequence with additional chemical modifications might be used to enhance the selectivity for the DNA damage site, providing a powerful tool in DNA damage detection and characterization.

Experimental Section

Supplementary Material