History of Nanopores and Nucleic acids

The concept of using small pores to analyze particles of biological relevance was pioneered by Coulter [1] who showed that blood cells in saline suspensions could be detected as they were moved through a pore 100 mm in diameter by a pressure gradient. As each cell enters the pore it changes the resistance to the ionic current that passes through the pore. The change is related to the volume of the cell by the equation Dg = 2pr3s/ L2, where r is the radius of the particle (with the volume of a spherical particle being a function of r3), L is the length of the pore, and s is the conductivity of the solution [2]. The resulting signals provide information about the cell number and cell volume. In a blood sample, cells such as red cells and leukocytes have different volumes, so that differential counts can be rapidly carried out. This information is useful in diagnosis of blood disorders, and commercial versions of the Coulter counter remain an important instrument in clinical laboratories.

Following the success of pore-based cell counters as analytical tools, interest turned to smaller pores that could be used to detect particles such as biological polymers. However, it is much more difficult to produce a pore in the nanoscopic size range, 2 – 3 orders of magnitude smaller that used by Coulter. Charles Bean [2, 3, 4] and his colleagues first used etched nuclear tracks for this purpose. When a high energy nucleus passes through a solid it typically leaves a "track" in which the structure has been chemically altered at the atomic level of structure. This track can later be etched by applying solvents that preferentially dissolve the damaged structures. Tracks are produced by natural radioactive decay in a substance, such as fission of a uranium nucleus present in silicate minerals, or by exposure to a radioactive source such as californium (Cf252) that undergoes spontaneous fission.

In the latter case the fragments can be collimated so that the resulting tracks are parallel. The tracks are then etched by exposure to solvents such as hydrofluoric acid or concentrated sodium hydroxide. (Nucleopore filters with pore sizes ranging from 50 nm to several micrometers are produced by this process and are commercially available.) If mica etched with 49% HF is used, the pore size at initial breakthrough is approximately 6 nm, which then increases to micron size ranges after 30 minutes of etching. Deblois and Bean [4] showed that mica nanopores were sufficiently sensitive so that particles as small as viruses could be detected by the change in resistance as the particle was driven through the pore.

Although mica and polycarbonate films produce pores of controlled size, the pores are in fact tunnels through the relatively thick materials. This limits their resolution, because a globular particle spends time negotiating a tunnel of varying dimensions. Ideally a nanopore will be present in a film that is no thicker than the analyte to be measured. This is particularly true of linear polymers such as nucleic acids, in which the desired resolution is at the level of a single nucleotide.

Biological Nanopores

In the 1970s, it became apparent that biological membranes of cells incorporate nanoscopic channels composed of proteins that are embedded in lipid bilayers. The function of the channels is to provide a gated pore that allows solutes such as sodium, potassium and calcium ions to cross the otherwise impermeable lipid bilayer barrier that surrounds all cells. Artificial bilayers are readily produced by self-assembly of amphiphilic compounds such as phospholipids. For instance, bilayers of phospholipid 5 nm thick can be produced as membranes covering a 100 micrometer diameter hole in a thin plastic sheet. Such membranes have dimensions appropriate for nanopore analysis, provided that a pore can be produced in the bilayer.

Hladky and Haydon [5] showed that the bacterial antibiotic gramicidin spontaneously inserted into lipid bilayers and could form continuous channels that conducted ionic current. Gramicidin is the smallest useful nanopore, having a pore diameter of roughly 0.4 nm. The gramicin pore contains a chain of approximately ten water molecules that form a single hydrogen-bonded strand across the bilayer. In the presence of a voltage, the pore conducts ionic current of a few picoamperes, the amplitude depending on the voltage and ion concentration. The gramicidin pore is too small to be used directly to detect solutes other than ions, but Cornell and co-workers [6] have shown that it can be used as a biosensor if it is coupled to a specific antibody. When a target antigen binds to the antibody, the ability of the gramicidin pore to conduct ionic current is inhibited. The reduction in current can then be used to estimate the presence and concentration of the antigen.

The fact that such pores could conduct polar and ionic solutes across a membrane suggested that a larger pore might be used to detect macromolecules in solution. This concept was first tested by Bezrukov et al [7] who showed that ionic current through the alpha hemolysin pore was affected by the presence of polyethylene glycol (PEG). The current fluctuations appeared only at certain molecular size ranges: PEG molecules that were smaller or larger than the pore had little effect, while PEG in the size range of the pore (0.5 - 1.5 nm) produced large current fluctuations. This allowed the authors to conclude that PEG could enter the pore and affect ionic current. Because PEG has no net electrical charge, the molecules enter the pore by diffusion, rather than by an energy-dependent process. ( See Bezrukov [8] for review.)

At about the same time, Bayley and co-workers [9] demonstrated that the hemolysin pore could be modified to serve as a metal biosensor. In this work, the hemolysin was modifed by introducing histidine for amino acids 130 - 134 in the central glycine-rich loop of the protein. The modified pore had a conductance that could be modulated by micromolar concentrations of ions such as zinc, suggesting that modified protein pores such as hemolysin could serve as sensitive metal ion sensors.

Nanopores as linear polymer sensors

The next advance in nanopore analysis was to take advantage of the fact that larger ionized solutes could potentially be driven through the pore by an electrical field. This would be particularly important for analyzing linear polymeric solutes such as nucleic acids. The pore diameter permits nucleic acids to be translocated as single strands, rather than duplexes, so that a given strand passes through the pore in strict linear sequence. This fact led to the concept of nanopore sequencing, which depends on the possibility that each base in a nucleic acid strand will modulate the signal in a specific and measurable way. If so, the base sequence in a nucleic acid molecule could be determined at rates far exceeding those now possible.

The hemolysin toxin produced by S. aureus is a suitable nanopore because it self-assembles in lipid bilayers to form aqueous pores with diameters just large enough to translocate the nucleotides in a single strand of DNA or RNA.The entrance to the head of the channel has a diameter of approximately 2.6 nm that contains a ring of lysine residues. The entry opens into a vestibule with an interior diameter of 3.6 nm that in turn leads to the pore’s stem which penetrates the lipid bilayer. The average inside diameter of the stem is about 2.0 nm with a 1.5 nm constriction between the vestibule and the stem. This limiting constriction is composed of alternating lysine and glutamate residues, and the remainder of the stem is lined primarily with neutral residues except for a single hydrophobic ring of exposed leucine residues. The opening at the other end of the pore has a 2.2 nm ring of alternating lysine and aspartate residues.

To produce a nanopore, a-hemolysin subunits are introduced into a buffered solution that is in contact with a lipid bilayer which separates the solution into two compartments, designated cis and trans. This apparatus, in conjunction with a standard amplifier and software, provides rapid response times on the microsecond time scale and substantially lower noise than the larger planar lipid membrane commonly used in single channel research. The open channel of a hemolysin pore carries an ionic current of approximately 120 picoamps (pA) with an applied voltage of 120 millivolts (mV).

Kasianowicz et al. [10] first reported that addition of single-stranded nucleic acid homopolymers produced large numbers of transient ionic current blockades, suggesting that individual molecules were being translocated through the pore and blocking ionic current. Formal proof for this conjecture followed when it was demonstrated that the number of blockades were correlated with the actual number of single-stranded DNA molecules translocated through the pore [10].

The experimental approach involved a comparison of the ability of DNA as single-stranded (ssDNA) and double-stranded (dsDNA) molecules to pass through the pore. Quantitative PCR was then used to make a direct measurement of nucleic acids appearing on the trans side of the membrane, and the numbers were compared with the number of blockade events that occurred during the same time interval. Only ssDNA appeared on the trans side, and the ratio of single-stranded DNA molecules appearing on the trans side to the number of blockade events was close to 1. This observation permitted the conclusion that single stranded nucleic acids were in fact translocated through the hemolysin pore, and that the resulting ionic current blockades could be used to gather information about single molecules.

Blockade mechanism

The process by which an ionic current blockade is produced must entail three phases: capture, entry, and translocation. When a voltage of 120 mV is imposed on the pore, the resulting ionic current consists of potassium and chloride ions. If single stranded nucleic acids are present, a given molecule will occasionally diffuse into a small volume near the mouth of the pore. When the polymer enters the biased region, three outcomes are possible. The first is that the diffusing nucleic acid molecule simply collides with the pore mouth. During this collision, the ionic current may be interrupted for a few microseconds, but the duration and extent of this transient blockade is variable and largely independent of chain length.

The second possibility is that one end of the molecule partially enters the vestibule and remains for periods ranging from tens of microseconds to several milliseconds, after which it either diffuses out of the pore or is drawn completely into the pore stem. While the molecule occupies the vestibule it causes a partial blockade having a characteristic amplitude about half that of the full blockade.

The third possibility is that one end of the nucleic acid is drawn completely into the pore stem, where it produces a full blockade with a duration that is a function of chain length. Translocation occurs as a result of electrophoretic force acting on the anionic phosphate groups of the chain. The pore stem is only 5 nm long, but the nucleic acid strand can be up to thousands of nucleotides in length. Experiments with very short polynucleotides show that the electric field acts primarily on the 10-14 phosphate groups within the stem [11]. The resultant force of about 20-40 piconewtons is sufficient to drive the molecule through the pore with velocities that depend on the composition and secondary structure of the strand.

Given the small aperture of the pore (1.5 nm), the polynucleotides must move as an extended linear polyanion. Meller et al. [11, 12] have undertaken detailed studies in which temperature effects on translocation are considered, as well as direct estimates of force acting on a DNA strand during translocation. (See Muthukumar [13] and Lubensky [14] for discussions of the physics of polymer translocation through nanopores.)

Parameters of ionic current blockades

Structural properties of the hemolysin nanopore are relevant to understanding the mechanism of ionic current blockades. Here we will assume that the major component of the blockade occurs when a nucleic acid occupies the 5 nm long stem of the channel, and neglect contributions by the larger vestibule. The pore volume within the stem is 18 nm3, and the average diameter is 2.0 nm, with a limiting aperture of 1.5 nm at the neck of the pore.
Three parameters provide information about the nature of the linear polymer passing through a nanopore. The first is blockade amplitude, which is best normalized and expressed as a fraction or percentage of the open channel current, I/Io, where I is the blockade current and Io is the open channel current. I/Io has a characteristic value for many homopolymers of RNA and DNA, suggesting that it will be an important analytical feature of nanopore technology.

Although several factors could conceivably contribute to blockades of ionic current, the simplest to test experimentally is that the fractional volume of a linear nucleic acid strand occupying a pore will reduce the number of ions available to carry current. The contribution of fractional volume of the molecule occupying the pore stem has been tested by investigating blockades produced by several polyanions that vary in molecular volume yet have approximately the same charge density of phosphate along the strand. These blockade amplitudes were then compared with the fractional volume of the pore stem occupied by the polymer, taking into account not only the molecular volume of the polymer but also water of hydration on the polymer and pore walls. The exact conformation of a single strand of DNA in the channel is unknown, but for the purposes of this calculation a minimal estimate is 0.34 nm per base, the repeat distance of bases in a double helix. About 15 nucleotides would then be in the pore at any instant. (The pore is defined here as the portion of the hemolysin channel that penetrates the lipid bilayer, and does not include the larger volume of the vestibule and channel mouth.) Assuming that each monomer of the nucleic acid has four waters of hydration (60 total), and that a single layer of water is bound to the interior surface of the pore (320 total), approximately 70% of the available pore volume would be occupied by a single strand of oligo(dA).

Similar calculations were made for oligo(dC), for an abasic strand of nucleic acid, and for polyphosphate, both of which also produce measureable ionic current blockades [15]. The experimental values are in approximate agreement with the calculated values. From the results of this simple experiment, it can be concluded that blockade amplitude is largely a function of the fractional volume occupied by a linear polymer traversing a nanopore. The difference in total volume between purine and pyrimidine deoxyoligonucleotides in the a-hemolysin pore is only 0.3 nm3, which represents a 6% difference in the volume occupied by the molecules in the pore after correcting for water of hydration. This difference is just barely detectable as an average signal over noise, and is produced by multiple nucleotides occupying the length of the pore. It follows that an improved pore having a smaller limiting aperture than that of the a-hemolysin channel will probably be required for single nucleotide resolution in nanopore sequencing applications.

Correlation of blockade parameters with polynucleotide composition

Howorka et al [23] found that it was possible to attach a single stranded oligomer of DNA to hemolysin so that the vestibule contained a short sequence that potentially could bind to traversing DNA if complementary base pairing was possible. This remarkable feat of nanoengineering produced a biosensor that could detect specific DNA sequences with single base resolution. The principle underlying the function of this biosensor is that a traversing DNA strand binds to the tethered DNA in the vestibule by complementary base paring and remains in the pore for a measureable amount of time. As a result, a distinctive two-part signal is generated, first by the duplex DNA that partially blocks the current, followed by a spike as the target DNA is released and passes through the pore stem. DNA that lacks the complementary base paring passes through very rapidly without the partial blockade. The duplex lifetimes allow discrimination between short DNA strands that differ by only a single nucleotide so that duplex formation is inhibited. Because of this sensitivity, it was possible to determine the sequence of a codon embedded in the sequence of a single DNA oligomer covalently linked to the vestibule. These authors also used this approach to detect a mutation in the reverse transcriptase gene of the HIV virus.
Nanopore analysis at single base-pair and single nucleotide resolution

From the measured velocities of single stranded nucleic acid translocation through the hemolysin pore, it is clear that a major factor limiting resolution is the time available to detect current modulation by a single nucleotide. For instance, a single nucleotide in a DNA strand passes through the length of the pore in a few microseconds, a time interval that does not provide sufficient ionic current to distinguish between a purine base and a pyrimidine base in the total ionic current of a blockade. In order to demonstrate that increased time intervals may permit the sensitivity required for single base resolution, DNA “hairpins” can be used to keep a single molecule of DNA in the a-hemolysin channel for relatively long (millisecond) time intervals.

Hairpins are sequences of single-stranded nucleic acids folded back on themselves to form hydrogen bonds between Watson-Crick base pairs [19, 20]. The entropy and enthalpy components of DNA hairpin formation are readily estimated, and the sequence can be designed so that only intramolecular interactions occur [21]. An extensively hydrogen bonded hairpin structure can enter the vestibule of an a-hemolysin channel, but would not be translocated through the pore until all the hydrogen bonds stabilizing the hairpin spontaneously dissociated.

The investigation reported by Vercoutere et al [22] used a well-characterized DNA hairpin with a six-base-pair stem and a four-deoxythymidine loop. When captured within an a-hemolysin nanopore, this molecule caused a partial current blockade (or ‘shoulder’) lasting hundreds of milliseconds, which was followed by a rapid downward spike. This signature was interpreted as an initial capture of a hairpin stem in the vestibule which produces the shoulder, followed by simultaneous dissociation of the six base pairs in the hairpin stem. The extended single-strand can then traverse the pore and produce the spike. With this assumption, a series of blunt-ended DNA hairpins with stems that ranged in length from 3 to 8 base-pairs.

It was found that each base pair addition produced a marked increase in median blockade shoulder lifetime, and a downward trend in shoulder current amplitude, from I/Io equal to 68% for a 3 bp stem to I/Io equal to 32% for a 9 bp stem. These results demonstrate that single base pair resolution can be achieved by monitoring reduction of ionic current as the hairpin stem extends further into the vestibule with each additional base pair.

Nanopore analysis using sequence-specific detection

Howorka et al [23] found that it was possible to attach a single stranded oligomer of DNA to hemolysin so that the vestibule contained a short sequence that potentially could bind to traversing DNA if complementary base pairing was possible. This remarkable feat of nanoengineering produced a biosensor that could detect specific DNA sequences with single base resolution. The principle underlying the function of this biosensor is that a traversing DNA strand binds to the tethered DNA in the vestibule by complementary base paring and remains in the pore for a measureable amount of time. As a result, a distinctive two-part signal is generated, first by the duplex DNA that partially blocks the current, followed by a spike as the target DNA is released and passes through the pore stem.

DNA that lacks the complementary base paring passes through very rapidly without the partial blockade. The duplex lifetimes allow discrimination between short DNA strands that differ by only a single nucleotide so that duplex formation is inhibited. Because of this sensitivity, it was possible to determine the sequence of a codon embedded in the sequence of a single DNA oligomer covalently linked to the vestibule. These authors also used this approach to detect a mutation in the reverse transcriptase gene of the HIV virus.

Synthetic nanopores for analyzing nucleic acids

Synthetic nanopores are now being tested that could have the potential to match or surpass a-hemolysin in detecting and analyzing nucleic acids in solution. In contrast to a labile protein in a delicate lipid bilayer, solid state nanopores should be mechanically robust, tolerate a broad range of temperatures, pH, and chemical conditions, and provide a low capacitance, low noise surface suitable for integrated electronics. Ideally, the fabrication process used to make the solid state nanopore should make it possible to easily control and vary pore dimensions so that these dimensions can be optimized for a particular application, e.g. probing single stranded vs. double stranded DNA.

Fulfilling these requirements is difficult because fabrication methods able to manipulate matter at nanometer dimensions have not been available. For example, single 3-5 nanometer holes in an insulating silicon nitride membrane can be made using an erosion process, such as reactive ion etching [25], but this method produced nanopores with a range of dimensions from 3-40 nm, only a few of which were of the desired dimension. Achieving greater reproducibility requires knowing precisely when to stop the erosion process. One approach that can reproducibly fabricate nanopores of a desired dimension is a feedback-controlled ion sputtering system that counts the ions transmitted through the gradually opening pore and extinguishes the ion sputtering erosion process at the appropriate time [26] With feedback control, reproducibility does not depend on precisely matching all conditions and starting dimensions. Using this system, it is possible to routinely produce robust single nanopores with 2, 3 or 4 nm diameters. Such nanopores are capable of registering single DNA molecules that produce characteristic curent blockade signals in much the same way as they do in the a-hemolysin pore [23].

A recent elaboration of this method was reported by Storm et al. (27) in which 20 nm holes in a silicon membrane were closed by exposure to an electron beam. It is clear that producing true nanopores on a routine basis is now feasible, and that numerous novel applications of single molecule analysis are now reasible.

Future applications of nanopore analysis

To summarize, single-base resolution in DNA or RNA remains the goal of investigators now working to establish nanopore sequencing, and several significant intermediate applications are now within reach. These include simple detection of DNA in microscopic volumes, as well as the targeted molecular bar codes. Although nanopore analysis is just emerging as a new research tool, the rapidly developing technology has immense promise as an analytical method for measuring and characterizing linear polyions such as nucleic acids. The ability to detect and identify single nucleic acid molecules in solution has numerous applications, but these will be realized only if a stable inorganic nanopore can be developed. Significant progress toward this end has been made, and as noted above the first such nanopore based on silicon nitride has detected long double-stranded DNA molecules as they pass through. Because single-stranded nucleic acids translocate through the pore in strict linear sequence, it is clear that in principle ultrarapid sequencing will be possible if single nucleotide resolution can be achieved. Sequencing the human genome required twenty five years of effort following the pioneering publications of Sanger and co-workers, and Maxam and Gilbert, who first described methods for sequencing DNA. The most efficient current methods can sequence about 30,000 bases per day per instrument, at an approximate cost of $0.50 per nucleotide for finished sequence. If it became possible to identify single bases as they traverse a nanopore, sequences could now be read at rates of 500,000 bases per second, and a base sequence equivalent to the human genome could be run through a single instrument in approximately two hours. More likely, of course, is that rates will be in the range of 1000 bases per second, still a marked improvement over the one base per second that is currently the state of the art.

Assuming that a nanopore instrument can be commercially developed, other significant applications appear to be feasible:

1. Measurement of nucleic acid concentration, for instance, during PCR amplification.
2. Identifying molecular species in solution. It seems reasonable to think that a nanopore in the size range of 4 - 6 nm diameter will be able to detect and perhaps identify individual soluble protein molecules.
3. Encoded polymers as molecular markers. Akeson et al. (unpublished results) have demonstrated that synthetic nucleic acids can be linked to a cleavable targeting agent and