Long dwell-time passage of DNA through nanometer-scale pores: kinetics and sequence dependence of motion

Abstract

A detailed understanding of the kinetics of DNA motion though nanometer-scale pores is important for the successful development of many of the proposed next-generation rapid DNA sequencing and analysis methods. Many of these approaches require DNA motion through nanopores to be slowed by several orders of magnitude from its native translocation velocity so that the translocation times for individual nucleotides fall within practical timescales for detection. With the increased dwell time of DNA in the pore, DNA-pore interactions begin to play an increasingly important role in translocation kinetics. In previous work, we and others observed that when the DNA dwell time in the pore is substantial (>1 ms), DNA motion in α-hemolysin (α-HL) pores leads to nonexponential kinetics in the escape of DNA out of the pore. Here we show that a three-state model for DNA escape, involving stochastic binding interactions of DNA with the pore, accurately reproduces the experimental data. In addition, we investigate the sequence dependence of the DNA escape process and show that the interaction strength of adenine with α-HL is substantially lower relative to cytosine. Our results indicate a difference in the process by which DNA moves through an α-HL nanopore when the motion is fast (microsecond timescale) as compared with when it is slow (millisecond timescale) and strongly influenced by DNA-pore interactions of the kind reported here. We also show the ability of wild-type α-HL to detect and distinguish between 5-methylcytosine and cytosine based on differences in the absolute ionic current through the pore in the presence of these two nucleotides. The results we present here regarding sequence-dependent (and dwell-time-dependent) DNA-pore interaction kinetics will have important implications for the design of methods for DNA analysis through reduced-velocity motion in nanopores.

Figure 1

Multi-nanopore DNA escape. Application of a capture voltage (+150 mV) leads to the capture and trapping of probes, resulting in gradual blockage of the pores and a decrease in the ionic current. After sufficient time to allow for >95% of the pores to be filled (each with a single DNA molecule), the voltage is reduced to the escape potential (+65 mV shown), allowing for thermally activated escape of DNA from the pore against the electrostatic energy barrier. After sufficient time for >99% of the probes to escape from the pore, the voltage is then reduced to a large negative potential (−100 mV shown), forcing any remaining probes out of the pore, followed by a return to 0 mV for ∼1 s and a return to the escape voltage (+65 mV shown) for ∼0.5 s to determine I_final. The current through the pore as a function of time during escape is used to determine the likelihood that DNA will be in the pore at time t upon application of the escape potential (i.e., the survival probability of DNA in the pore during escape).

Figure 2

Log-log plot of the survival probability versus time (+65 mV escape voltage). The survival probability of DNA in the pore (upon application of the escape potential) versus time for dA27 (>3000 single-molecule events) and dC27 (>5000 single-molecule events) at an escape voltage of +65 mV is shown. The large number of events comprising each data set ensures that long-dwell-time events are statistically significant. Also shown are the corresponding fits to the data from the Becquerel, single-exponential, and three-state model fits. In both cases (dA27 and dC27), the exponential fit is a poor fitting function to the data. Although the Becquerel decay law fits the dA27 data reasonably well, it is a poor fitting function to the dC27 data. In contrast, our three-state model fit forms an excellent fit, particularly at long timescales, to both the dA27 and dC27 data sets. All fits require four parameters for two data sets (in the case of the exponential fit, a prefactor is used, which is itself a free parameter).

Figure 3

Model of the DNA escape process from nanometer-scale pores. Model for the DNA escape process, adapted from Bates et al. (10), where S_F represents the free state of DNA (i.e., DNA not bound to the pore), S_B the bound state (i.e., when DNA is bound to the pore), and S_E the state in which DNA has escaped from the pore. Our model invokes a constant escape rate (k_e) and binding rate (k_b), and a uniformly distributed unbinding rate (k_u) ranging from 0 to k_{u max} (i.e., k_u ∼ U(0, k_{u max})), where k_{u max} is a sequence-dependent rate (as opposed to the binding rate and escape rate, which are both sequence-independent).

Figure 4

Activation energy distribution (g(E_b/k_bT)) for unbinding from the pore expressed as the distribution g(ln(τ)) at an escape voltage of +65 mV. Shown are g(ln(τ)) for both dA27 (dashed line) and dC27 (solid line). g(ln(τ)) and g(E_b/k_bT) differ from each other only by their location along the x axis (refer to text). The distribution, g(ln(τ)), for both dA27 and dC27 is a shifted exponential distribution (inset) shifted by the logarithm of the shortest timescale for unbinding (ln(τ_{u min}), where τ_{u min} = 1/k_{u max}). In the case of g(E_b/k_bT), the shift amount is the minimum energy barrier height for unbinding. Inset: The average energy barrier height for unbinding is ∼5.33 k_bT higher for dC27 relative to dA27, which suggests a greater binding interaction/strength of cytosine to the α-HL pore relative to adenine. The SD of each distribution is 1 k_bT.

Figure 5

Histograms of the ionic current through a single α-HL pore upon capture and trapping of the dC27 versus dCdmC27 polynucleotides. Each histogram is comprised of >2000 single-molecule events. The histograms are expressed as I/I₀ and are normalized to the peak of the distribution (I₀ is the open pore current; not shown). dC27 and dCdmC27 differ only by the presence of 5-methylcytosine at bases 10–14 (inclusive, measured from the 5′ to 3′ end) in the case of dCdmC27. Consistent with previous studies (13,14), these results show that dmC occludes the ionic current to a greater degree than does dC. However, in contrast to those studies, the results show that unmodified, wild-type α-HL can distinguish between dmC and dC by measurements of the blockage current alone.