DNA molecules and configurations in a solid-state nanopore microscope

A nanometre scale pore in a solid state membrane provides a new way to electronically probe the structure of single linear polymers, including those of biological interest in their native environments. Previous work with biological protein pores wide enough to pass and sense single stranded DNA molecules demonstrates the power of the nanopore approach, but many future tasks and applications call for a robust solid-state pore whose nanometre scale dimensions and properties may be selected, as one selects the lenses of a microscope. Here we demonstrate a solid-state nanopore microscope capable of observing individual molecules of double stranded DNA and their folding behaviour. We discuss extensions of the nanopore microscope concept to alternative probing mechanisms and applications including the study of molecular structure and sequencing. Probing,

conduction signals from voltage biased nanoscale biopores 6,7 . More recently, a voltage bias on an alpha hemolysin biopore has been shown to induce charged single-stranded DNA and RNA molecules to translocate through the pore [8][9][10] . Each translocating molecule blocks the open pore ionic current providing an electrical signal that depends on several characteristics of the molecule. This system has limits for studies of biological molecules: the pore is of a fixed size, and its stability and noise characteristics are restricted by chemical, mechanical, electrical, and thermal constraints. These difficulties may be overcome with the use of a suitable solid-state nanopore 11 . Here we report on a solid-state nanopore "microscope" capable of electronically characterizing single long chain polymers like DNA molecules. We also show the first observation of molecule induced quantized current blockades that reveal the folding configuration of single molecules as they pass through the nanopore.
At its heart the microscope consists of a voltage biased nanopore, fabricated in a silicon nitride membrane.
The membrane separates two chambers of conducting electrolyte solution. The only electrical conduction path from one chamber to the other passes through the nanopore. To resolve interesting molecular structure the nanopore dimensions must be small enough to avoid averaging over continuous single molecule configurations induced by thermal fluctuations and large enough to pass the smallest dimensions of the molecule to be probed. For double stranded DNA (ds DNA) this means a pore diameter and membrane thickness smaller than the molecule persistence length, 50 nm for ds DNA, and a pore diameter larger than the ~2 nm cross sectional size of the molecule. The recent discovery of ion beam sculpting 11 allows structures that meet these criteria to be fabricated with desired nanometer scale dimensions from solid state materials like silicon nitride. A transmission electron micrograph of an ion sculpted 3 nm nanopore in a 5-10 nm thick membrane is shown in Figure 1a, and a schematic of the experimental setup is shown in Figure 1b.
Open pore ionic conduction was first established with 120 mV bias across the nanopore. Then DNA was added to the negative cis chamber and current blockades appeared in the form of isolated transient reductions in current flow through the pore. Figure 1c shows part of a current trace recorded for 3kb DNA (~1μm long) and a 3 nm pore. Each event is the result of a single molecular interaction with the nanopore and is characterized by its time duration t d and its current blockage, ∆I b , ~120 pA. The expected current blockage from a single molecule blocking the pore is linearly dependent on the cross-sectional area of the molecule and independent of the area of the pore, although because the blockage current varies inversely with the thickness of the pore, different pores may produce different blockage currents for the same molecule.
Occasionally, the baseline level shifted for very long periods of time by a magnitude similar to that belonging to the discrete transient molecular event. This was likely due to a single molecule that became "stuck" in the nanopore. We do not present any molecular data here for times intervals where the open pore current had been reduced in such a way from its initial value. A visual study of individual events for the 3 nm pore plotted in Figure 2a shows them all to be simple single level current blockades of the type at the bottom of Figure   1c (see also inset Figure 4a). Approximately 60% of the events in 10 nm pore experiments are of this type but the remainder are more complex (see inset Figure 4b).
Selecting simple single level events from the 10 nm pore data significantly sharpens the distribution in both <∆I b > and t d . In Figure 3 we present a histogram of t d values for simple events in three experiments using 10 nm pores: 3 kbp ds DNA with 120 mV bias, 10 kbp ds DNA with 120 mV bias, and 10 kbp ds DNA with 60 mV bias. The 10 kbp DNA is seen to take slightly more than 3 times longer to negotiate the pore than the 3 kbp DNA at the same bias. Reducing the bias by a factor of two approximately doubles the translocation time. These observations provide strong evidence that each simple single level event corresponds to a DNA molecule translocating in single file order through the nanopore under the influence of electrophoretic forces. We shall see that the structure of the more complex signals confirms this interpretation. Figure 4a shows the density plot of the simple translocation events for 10 kbp DNA passing through the 10 nm pore. The main cluster of events is narrowly distributed in both <∆I b > and t d . We will discuss the second cluster later, but note its mean <∆I b > is twice that of the main cluster while its mean t d is half. Characteristic time recordings of events from these two regions of the density plot are shown in the inset. Figure 4b shows a density plot for more complex "multi-level" events that remain after the simple ones are subtracted. Examples of event time recordings in this group are shown in the inset. They look like simple events on which additional blockade structure has been superimposed. For ~85 % of the complex events the additional structure appears at the front of the event, ~5% at the rear, 1-2 % at both the front and rear, and 5% in the middle. Half of the events with structure in the middle have t d > 400 μsec. (More complex structures are also observed in longer t d events.) We attribute these remarkable additional features to DNA molecules that are folded on themselves as they pass through the pore. As overlapping folded parts of a molecule pass through the pore they enhance the current blockade during that part of the event. If the instantaneous current blockade is proportional to the number of strands of the same molecule in the pore (i.e. one or two), one calculates that the average current blockade for the event will be inversely proportional to the translocation time t d , where <I 0 > and t 0 are the mean current blockage and translocation time of a simple event. This simple model, plotted as the dotted line in Figure 4b, shows excellent agreement with the data. The smaller cluster in Figure 4a is thus interpreted as due to molecules that are folded nearly in the middle of the strand. Residual closed circle plasmid DNA in the sample preparation could presumably contribute to this peak.
More confirmation that complex nanopore signals correspond to events where folded DNA molecules translocate through the pore is provided by a study of the distribution of instantaneous blockade current magnitudes over all events. Assuming the instantaneous magnitude of the blocked current is in proportion to the instantaneous number of strands of ds DNA in the nanopore, we expect the distribution of blocked currents taken over many events, in time samples much smaller than an event duration, to show a quantization of local instantaneous I b values corresponding to 0,1,2,… strands of the folded molecule in the pore at any particular time. A histogram of these sampled values of I b for 10 μsec samples over ~9500 events (including 200 μsec before and after each event) is shown in Figure 5a for the 10 kb, 120 mV data and in Figure 5b for the 10 kb, 60 mV data. The expected quantization of sampled I b values is clearly seen corresponding to zero, one and two molecule strands occupying the nanopore (note the log scale). Experiments with 50 kbp DNA and a 15 -20 nm pore (data not shown) also show three level blockades.
A molecule microscope based on solid-state nanopore provides distinct differences and/or advantages over extant biopore detectors. All results presented here were for ds DNA molecules whose transverse size is ~2 nm. The hemolysin biopore used in previous studies cannot translocate such a large diameter molecule 8,12 where σ is the linear charge density on the molecule, η the viscosity of the solution, V bias the pore voltage bias and C is a factor of order unity accounting for the complex issues mentioned above. Setting a to the persistence length of DNA 18 (which implies statistical loss of effective drag force beyond that distance) and assuming a charge of e/3 per phosphate, we find that C~1/2 brings equation (2)  properties of solid-state nanopores will lead to dramatic improvements in nanopore microscopes' ability to probe important biological molecules.

Methods
Nanopores used in our "microscope" were fabricated in 25 μm x 25 μm free standing silicon nitride membranes supported by 3mm x 3mm x 0.3mm silicon substrate (100) frames. The 500 nm thick, low stress (~200 MPa tensile) silicon nitride membrane was deposited by low pressure chemical vapor deposition. Photolithography and anisotropic wet chemical etching of silicon were used to create the free standing SiN membrane.
An initial 0.1 μm diameter pore was created at the membrane's center using a focused ion beam (FIB, Micrion 9500) machine. The diameter of this large pore was then decreased to molecular size near one surface of the membrane using feedback controlled ion beam sculpting 11 . The final nanopore thus resides in a thin 5-10 nm thick membrane that covers an approximately 0.1 μm diameter cylindrical aperture that extends through the thick silicon nitride membrane below. Nanopore diameters were determined by transmission electron microscopy. We note that because the TEM projects a three dimensional structure on to a two dimensional plane, the image of the inner edge of the pore actually represents the minimum projected diameter of the pore wall at any height and may not correspond to the narrowest physical constriction. Also, because of the inherent inaccuracies of TEM for determining absolute size and the fact that our pores are not perfectly round, all sizes should be taken as estimates to within a nanometre of the actual pore size. Ionic current through the solid-state nanopore was measured and recorded using an Axopatch 200B integrating patch clamp amplifier system (Axon Instrument) in resistive feedback mode. Signals were preprocessed by a 10 kHz low pass filter.
Except for the data displayed in figure 1c, which is a live recording, all data was acquired in event driven acquisition mode, meaning analogue start and stop triggers were used to determine when data was to be recorded.
Professor Alec Kavcic, Dr Michael Burns, Albert Huang and Jiajun Gu assisted with software analysis.
Professor Qun Cai assisted with nanopore preparation and Christopher Russo provided assistance during preparation of this manuscript. Support for this research has been provided by DARPA, NSF, DOE,

AFOSR and Agilent Technologies
Correspondence and requests for materials should be addressed to J. A. G. (e-mail: golovchenko@physics.harvard.edu.).