While most nanopore-based, single-molecule sensing is performed using planar membranes, which have a well-defined pore length (i.e. membrane thickness), nanopipettes have a gradual taper length (Fig. 1a) that increases the sensing region of the device21. We fabricate nanopipettes by laser pulling glass nanocapillaries, producing two identical quartz nanopores. With this technique, we can achieve <10 nm inner pore diameters, Figure 1a. This process is fast, inexpensive, and does not require a clean-room environment22.
Current-voltage (I/V) analysis reveals that the conductance of the pores (Fig. 1b) varies between 0.58 and 5.35 nS; as well as the presence of ionic current rectification23. These conductance values are consistent with pore diameters between 5 (± 0.5) and 48 (± 4) nm, respectively. Specifically, the relationship between pore conductance (G) and inner diameter (di) allows us to estimate the size of the aperture24,25:
Where is the length of the conical pore (taper length), is the measured conductivity of the buffer, and is the diameter of the capillary at the beginning of the conical taper. The initial, inner capillary diameter is constant in our experiments (0.7 mm) and the buffer conductance depends on concentration and alkali chloride used. The taper length was measured using transmission electron microscopy and I/V analysis is used to measure the pore conductance. Pore sizes are also occasionally confirmed using transmission electron microscopy.
After retrieving I/V information, translocation experiments with λ-DNA at 500 pM were performed. When EPF dominates, the capture volume outside the nanopore assumes a nearly spherical shape surrounding the pore’s oriface26–30. As ionic strength decreases, EOF can dominate as the primary means for DNA entering the pore. According to the EOF streamlines, the capture volume adopts a shape confined along the sides of the pore31. There also lies a crossover concentration point in which EOF reverses direction, where EOF is generated along the glass surface and radiates away from the pore aperature31.
Finite element analysis was performed to determine the fluid flow rate at different voltages (Fig. 1c). As the applied voltage decreases from 0 mV, the mean fluid velocity increases into the glass pore. The same is true for positive voltages, however, the fluid flow direction switches (flow reversal) from towards the pore at negative voltages to away from the pore at positive voltages. Notably, these fluid velocities can influence DNA dwell time inside the pore and have been described using hydrodynamic drag32,33.
DNA proceeds to diffuse around the solution until it enters the EOF capture volume, where it is then transported through the pore. This method of translocation is further illustrated (Fig. 1d) with KCl as the electrolyte and K+ ions responsible for the movement of water carrying λ-DNA. Under these conditions, it is possible to see differing DNA configurations: linear, partially folded, and fully folded (Fig. 1d). Reports of different DNA configurations have been witnessed using high ionic strength conditions and with both planar nanopores34–37 and nanocapillaries24. The ability to discriminate folding states using DNA CEs does not directly help uncover the nature of CEs, but it is important to recognize the existence and understand the effects of having various DNA configurations upon translocation.
To show that this finding is not limited to low ionic strength phenomenon, we employed salt concentration gradients as previously described20. As shown in Figure 1e, our experimental set-up involved having a solution of 1 M KCl + λ-DNA inside the nanopore with 4 M KCl outside. With an applied voltage of -600 mV, λ-DNA was driven outside the pore through EPF, resulting in CEs. An additional salt gradient was implemented (Fig. 1f) with 4 M KCl inside the nanopore and 1 M KCl + λ-DNA outside. In this situation, EPF drove λ-DNA to translocate through the pore, again resulting in CEs. Based on these results, a working hypothesis was made that the existence of CEs stem from a flux imbalance between anions and cations. This is notably different than ion selectivity which is typically a characteristic of the pore itself. Rather, flux imbalances can be generated through externally applied conditions and parameters. EOF pumping of water into the pore, for example, can change the relative fluxes of ions. Since the electric field acts equally on both chloride and potassium ions, the net movement of water only provides a moving frame of reference which favors one ion over another. Nevertheless, total ionic current is constant regardless of EOF velocity. For the data shown where CEs are observed (Fig. 1 d-f) we speculate that there is a net flux that favors potassium ions. Figure 1g illustrates how K+ ions are pumped into the pore under low ionic strength (EOF; Fig. 1d) and concentration gradient conditions (EPF; Fig. 1e).
Validation using finite element methods was undertaken to further explain the potential impact that EOF has on DNA sensing and the unique capture dynamics of EOF-driven events. A 20 nm pore was modeled with 10 mM and 2 M salt to demonstrate the difference in capture kinetics of DNA. At low ionic strength conditions, fluid velocity is plotted with colored streamlines indicating where the fluid entering the pore is coming from within the bath solution. Since DNA events only occur anti-EPF, mapping the fluid motion is indicative of the capture zone. To experimentally validate the finite element analysis (Fig. 2a), λ-DNA was tagged with YOYO-1 and the nanopipette tip placed in the focal plane of a water immersion objective (Nikon, NA=1.2). A stacked timeseries of images allowed us to observe λ-DNA capture at -700 mV (Fig. 2a inset reveals that fluid motion along the sides of the pore is responsible for λ-DNA translocation). At high ionic strength conditions, DNA travels through the pore via EPF and so the electric field lines are plotted (Fig. 2b) and represent the capture zone.
Realizing that the capture volume in EOF-driven translocations surrounds the outer walls of the nanopipette, we chose to expand and shrink the capture volume via a depth-dependent study to witness any changes in event frequency (Fig. 2c). By submerging varying lengths of the taper length inside the salt solution containing λ-DNA, the capture volume is controlled (Supplementary Fig. 1). The nanopore was suspended at 0, 0.26, 0.53, 1.1, and 4.0 mm below the electrolyte solution surface containing DNA. For exact measurements, the nanopore was suspended from a micrometer. Translocations were obtained for voltages between -100 and -1000 mV, in increments of 100 mV. Recording at -600 mV yielded the most consistent translocations without clogging the pore. Events were recorded at -600 mV and the I/V relationship yielded a 2.53 nS pore. Capture rate was calculated at each depth. As nanopore depth increases, capture volume also increases, leading to higher event frequency with larger depth values. As more of the nanopore is exposed to the λ-DNA solution, the capture volume enlarges, leading to an increase in event frequency.
In order to understand how electro-hydrodynamics influences ionic flux, particularly at low salt conditions, three monovalent salts were modelled by altering the cation diffusion coefficient and electrophoretic mobility (Fig. 2d and e). Although the pore’s total ionic flux was not altered significantly by EOF since K+ flux increased and Cl- flux decreased by the same amount, EOF does significantly impact the flux imbalance between cation and anion. This finding was particularly noteworthy since CEs have been observed at high asymmetric salt conditions which would also change a pore’s ionic flux imbalance. These results predict that a flux imbalance in favor of Cl- transport leads to resistive events and a flux imbalance in favor of K+ leads to conductive events. This is based on the experimental results that the 10mM KCl electrolyte always produces CEs. In Figure 2d, anion-dominant flux only occurs with small pore sizes, 20 nm and less, and an applied negative voltage between -300 and -400 mV. It is important to note that no events could be recorded at these conditions to find out whether resistive events are observed. For a nanopore suspended in LiCl, we observed more opportunities for the pore to be Cl- selective, which we predict will result in REs upon translocation of λ-DNA. As the pore increases in size or an increasingly negative voltage is applied, the pore can become cation selective, which we speculate can give rise to CEs.
Although the ionic diffusion coefficients and electrophoretic mobilities encapsulate basic transport properties, all the while being utilized as variables in the finite element simulations, they neglect the geometric size of the ions and therefore the packing density/strength on oppositely charged surfaces. In order to understand the link between electro-hydrodynamics and Debye layer screening of the quartz surface charge, streaming current measurements were used as a proxy for cation mobility within the diffuse ion layer. Contrary to EOF, where mobile ions drag fluid, streaming currents measure the fluid’s ability to drag along ions co-axial to the fluid motion38. A pressure bias was used to generate a streaming current and the resulting data can be seen in Figure 2f. Negative pressures generate a flow into the nanopore and in the same direction as EOF in our experiments. We see that larger pressures create larger streaming currents. Interestingly, LiCl has significantly higher streaming currents compared KCl and CsCl at negatively biased pressures. Overall, these results indicate LiCl-filled nanopores can be Cl- flux dominant at low voltages, and secondly, Li+ has a higher flux under pressure biased fluid flow compared to other cations (K+ and Cs+). The same pore (1.30 nS in 10 mM KCl) was used for all measurements to reduce variability due to different pore sizes.
DNA and Neutral Polymers in Potassium Chloride
Under high ionic strength conditions, pores with a diameter slightly larger than the analyte molecule yield greater SNR values when compared to larger pores19. Because of this, we were motivated to explore SNR values under low ionic strength conditions. A typical conductive DNA event can be seen in Figure 3a (bottom) in 10 mM KCl. Potassium chloride was chosen as the electrolyte because it is most frequently utilized in nanopore research due to similar ion mobilities of anions and cations. We incorporated differently sized pore diameters to witness any effect that pore size may have on event shape and size. The depth of each nanopore was kept consistent for all recordings as well as the voltage (-600 mV).
As λ-DNA translocates through the nanopore, we witness a current-enhancing event. For all SNR calculations, we omitted all configurations except for linear DNA translocations (Fig. 3b). DNA has the ability to translocate linearly, folded8,24, or in knots34, in which the latter two increase the current change. To ensure DNA configuration had no effect on SNR, we applied only linearly translocating DNA to our calculation.
We witness an increase in SNR starting at 2.00 nS and saturating around 3.00 nS. To determine whether the current enhancement or the noise of the signal is the major contributor to the increase in SNR, we acquired the median current change of all events and the root mean square (RMS) noise of a data segment lacking events. We witness that the RMS noise maintaining values of 15 ± 7 pA, whereas the current change increases from 30 to 140 pA as pore size increases. For the left side of the graph, we witness a sharp increase in the SNR as the pore size decreases. This can be explained by a decrease in the noise associated with smaller pores. As seen in Supplementary Figure 2, the RMS noise is extremely low (< 10 pA) whereas the median current change is approximately 100 pA. Therefore, the higher SNR values for smaller pores stem from lower noise. On the right side of the graph, we speculate that the rise in SNR (and current enhancement) is a result of greater EOF pumping as a function of pore size (Supplementary Fig. 3). Owing to larger fluid velocities, the flux imbalance highly favors potassium rather than chloride.
The common hypothesis that DNA counterions are the sole mechanism of CEs led us to explore PEG under low ionic strength conditions with an applied negative voltage39. PEG 20,000 was diluted to 15% (w/w) in 10 mM KCl and voltage was applied from -100 to -1000 mV, in increments of 100 mV. Interestingly, PEG events could be observed at an extremely small pore size (0.43 nS); a pore size regime that we could not observe DNA events. Since EOF decreases with smaller pore sizes and EPF increases, we believe DNA energetically could not overcome the barrier at the pore entrance for translocations to occur. Since PEG is neutral, we were able to observe EOF-driven events at very small pore sizes (Supplementary Fig. 4). The results indicated that smaller pore sizes resulted in CEs whereas larger pore translocations yielded REs (Fig. 3c and d). SNR calculations showed that the smaller pore diameter yielded higher SNR values in comparison to larger pore diameters. In both pores, the median current change was 71 ± 1 pA; whereas the RMS noise increased from 7 to 18 pA as the pore size increased from 4 to 25 nm in diameter, respectively. Based on these results, the nature of the event (CEs versus REs) seems un-coupled from the analyte counterions (or lack thereof, in the case of PEG) but rather linked to the pore size and/or voltage in which translocations occur. Although the analyte counterions do not seem to play a significant role in generating CEs, extremely small, negatively charged pores may be more likely to generate CEs due to their cation selectivity. It is also not fully understood how transient or long-term interactions of PEG with the charged glass surface far from the pore would impact EOF pumping. Previous reports have used PEG to lessen or neutralize EOF9 and therefore could be impacting the pore’s flux imbalance via interactions with the nanopipette’s conical taper.
Voltage Dependence with Lithium Chloride
Lithium chloride was chosen as an electrolyte because it has been previously shown to “slow-down” DNA translocations under high ionic strength conditions40. This can be attributed to Li+ having a smaller atomic radius than K+ and therefore, Li+ binds to DNA stronger than K+ 40. Additionally, LiCl had a significantly higher streaming current (Fig. 2f) compared to both KCl and CsCl. Finite element simulations indicated a voltage and pore size dependence for flux imbalance that was within the voltage range: -400 to -1000 mV, where events are typically observed. The nanopore containing 10 mM LiCl was inserted inside a solution containing 10 mM LiCl + λ-DNA and current changes were recorded at various voltages (Fig. 4a). The same series of steps were repeated to calculate the SNR at each voltage.
Using the same pore (1.20 nS), we witnessed the crossover point that is independent of salt concentration, which is something not previously observed. At voltages of -300 and -500 mV, λ-DNA translocations resulted in REs and at voltages of -700 and -900 mV, DNA translocations resulted in CEs, as shown in Fig. 4b. Interestingly, at an applied voltage of -600 mV the event current shape assumes both a resistive and conductive spike (Supplementary Fig. 5). For this pore, we see an increase in the amplitude of the REs as the voltage applied is reduced to -600 mV. Less than -600 mV (i.e. more negative), the CE amplitude continues to increase as the voltage decreases to -900 mV. The events recorded at -900 mV and -500 mV yielded higher SNR values in comparison to -700 mV and -300 mV, respectively (Fig. 4c and d). Supplementary Figure 6 shows how the median current change is the main contributor to the SNR fluctuation, the RMS noise for each voltage remains relatively constant. The transition from REs to CEs can be understood by the pore being anion selective at low voltages and cation selective at higher voltages. As the applied voltages increase in negativity, the change in current switches to a CE. The biphasic nature of the events at the transitional voltages (-500 mV and -700 mV) suggests that there may be two mechanisms of current modulation (hydrodynamic flow and pore occupancy) that can occur when the DNA molecule is near or entering the pore. DNA entering the flow field of the pore during EOF pumping may cause current modulations that occur immediately prior to translocation.
Another comparison was done using two pores with inner diameters of 33 ± 3 nm. One pore contained 10 mM KCl and was suspended in 10 mM KCl + λ-DNA while the other contained 10 mM LiCl and was suspended in 10 mM LiCl + λ-DNA. Both had an applied voltage of -600 mV and we witnessed CEs for the pore containing KCl and REs for LiCl (Fig. 4e). At -600 mV with the aforementioned pore size, finite element simulations predicted that the nanopipette is cation selective in KCl and anion selective in LiCl, which may be a possible explanation for the event types observed. We also note that KCl and LiCl have similar event durations at these low salt conditions, however, KCl has a much larger variation in the degree of current-modulation (in the case of KCl: current enhancement). The current-reductions observed for LiCl are much more tightly clustered together compared to KCl CEs. The source of the variability observed in KCl CEs is still not fully understood and requires further investigation. The data seems to suggest that CEs are more variable regardless of the cation. The LiCl events in Fig. 4b, for example, show a much greater degree of scatter for CEs compared to REs.
Alkali Chloride Dependence on Event Characteristics
Recently, CsCl was shown to have an advantage over KCl in respect to sequencing using solid-state nanopores11. This publication used CsCl because it disrupts the hydrogen bonding between guanines, therefore denaturing the G-quadruplex into single-stranded structures. Although we are not working with ssDNA, we aimed to compare KCl event properties with another alkali metal chloride that holds promise in the nanopore community. Therefore, we performed experiments using 10 mM CsCl inserted into 10 mM CsCl + λ-DNA. The typical current trace and event signature is displayed in Fig. 5a.
Similar to KCl, we do not see a voltage dependence on event shape with CsCl, which is not surprising considering that K+ and Cs+ have nearly the same diffusion coefficient41. For confirmation, a pore with a conductance of 1.47 nS (14 ± 2 nm diameter) was used with λ-DNA. Under low ionic strength conditions, we applied voltages of -300 mV, -400 mV, -500 mV, and -1000 mV to witness any transition in event shape (Fig. 5b). All voltages resulted in CEs, which was predicted based on finite element analysis under the assumption that cation selective conditions yield CEs. Simulation results for CsCl can be found in the Supplemental Information, but were nearly identical due to the diffusion coefficients for KCl and CsCl being 2.02 × 10-5 and 2.00 × 10-5 cm2/s, respectively41.
To explore the difference that alkali chloride type has on event capture rate, we fabricated three pores with inner diameters 35 ± 4 nm to be used with λ-DNA at -400 mV. We calculated capture rate by methods previously described28 to yield capture rates for each electrolyte used. Experimentally, we saw that λ-DNA in LiCl resulted in the highest frequency of events, followed by KCl, then CsCl (Fig. 5c). COMSOL was used to describe how alkali chloride type and pore size affected EOP pump velocity (Fig. 5d). Based on the conductance values of each pore, we believe some of the differences observed in the capture frequency are related to the size of the pore which strongly impacts the EOF pump velocity since smaller pores yield higher intra-pore electric fields. Based on this rationale, the CsCl experiments yielded a lower capture frequency due to the larger pore size. The extremely high capture efficiency observed in LiCl experiments may be due to the higher charge screening of the DNA backbone. A reduction of DNA charge will reduce the energetic barrier to move anti-EPF. The reduction in EPF is cohesive with the idea that DNA translocations in LiCl generate longer event durations at high salt conditions40. Lastly, we calculated the SNR of each electrolyte (Fig. 5e). We witness an increase in SNR starting with the lowest (CsCl) to the highest (KCl). In this scenario, translocations in LiCl resulted in the lowest RMS noise and median current change: 10 and 69 pA, respectively. KCl and CsCl both resulted with median current changes of 116 ± 2 pA. However, the major difference between these two lied within CsCl having more noise, resulting in a lower SNR.
How a flux imbalance yields CEs has yet to be addressed. The working hypothesis currently is that stored charges can accumulate at the nanopipette tip effectively acting as a capacitor in series with the highly resistive nanopore. Since the voltage at the extreme ends of the fluidic reservoirs are clamped, charge build-up (i.e. potassium) tends to generate a voltage that, in turn, lowers the effective voltage at the pore. We speculate that a DNA-occupied pore transiently stops EOF pumping and thereby lowers the stored charge inside the nanopore and that the capacitor discharges current proportional to the blocked EOF. Finite element methods demonstrate the accumulation of charge inside the glass pore (Fig. 6a). The increase in stored charge with applied voltage is a characteristic trait of an ionic capacitor. Upon solving for the effective capacitance, we obtain a value of 4 ×10-17 Farads. The timescale of charging and discharging accumulated charge is also fast (3-5 µs to reach steady state space charge density; Fig. 6c). Ionic-generated potentials are typically named according to the principle in which they are generated. For example, diffusion potentials, streaming potentials, and exclusion potentials42. Nevertheless, charge separation is a commonality of these potentials as well as our capacitor model which ultimately could generate voltage and current transients. Data thus far supports the hypothesis that a flux imbalance plays an important role in the generation of CEs. The existence of CEs with PEG (e.g. using a 0.43 nS pore) further demonstrated that charged analytes are not a pre-requisite for CEs, but may indeed have an important role depending on the pore size. For example, a 2.63 nS pore filled with 10mM KCl produced CEs when DNA was the analyte, and REs for PEG at the same conditions. We speculate that the analyte and its concentration in the reservoir can transiently impact a pore’s flux imbalance via translocation, or indirectly via interactions with the glass surface (i.e. outside the pore). For example, adsorbed molecules on the glass surface will hinder EOF pumping velocities and therefore the flux imbalance. Nevertheless, the evidence here demonstrates the importance of the pore’s charged surface, voltage-bias, and associated electro-hydrodynamics in generating CEs.