Biological nanopore

A biological nanopore is a protein channel, about 1 nm wide at its narrowest point, set in a membrane and used to sense single molecules by the ionic current that flows through it. Most of the pores in use were taken from toxins and transport proteins, and the main ones are assembled from seven to twelve identical subunits, so that every pore in a preparation has the same sequence and the same structure. The best known is alpha-hemolysin, which in 1996 carried the first single strands of DNA and RNA to be detected as they crossed a membrane.^[1] Engineered descendants of a second pore, CsgG, read DNA in the nanopore sequencing instruments of Oxford Nanopore Technologies.^[2]

A biological pore has a fixed size, sits in a membrane that is harder to make than the pore itself, and needs a motor enzyme to slow DNA to a speed at which it can be read. Molecular Reality Corporation builds solid-state nanopores instead, and its public guide says that "protein pores are delicate flowers".^[3] Designed proteins and DNA-built pores have been added to the natural set, and the pores have been used on peptides and proteins as well as on nucleic acids.

A membrane separates two chambers of electrolyte and holds one pore. A voltage across the membrane drives an ionic current through the pore, and a molecule that enters the narrowest part of the channel changes the current for as long as it stays. One alpha-hemolysin channel in 1 M potassium chloride carried an open-pore current of 120 pA at 120 mV, a conductance of 1 nS.^[1] Most of the sensing happens at the constriction, which sets how much of a molecule is read at once.

Subunits assemble by themselves. Alpha-hemolysin monomers insert spontaneously into a lipid bilayer and form heptameric channels, and the pore is a mushroom-shaped heptamer whose channel runs 10 nm along the sevenfold axis and varies from 1.4 nm to 4.6 nm in diameter.^[1][4][a] A mutation in the gene of a pore built from identical subunits appears once in every subunit, so engineered pores carry seven, eight or nine copies of each change, which designers have used to alter the charge and shape of the channel.^[5]

Five of these are toxins or toxin relatives. MspA is the porin through which a mycobacterium takes in hydrophilic nutrients, and CsgG is the secretion channel for curli, the amyloid fibers that hold bacterial biofilms together.^[6][9]

Two architectures account for them. In a β-barrel pore, strands from every subunit join to form a barrel through the membrane. The transmembrane domain of alpha-hemolysin is the lower half of a 14-strand barrel in which each subunit adds two strands 6.5 nm long, CsgG builds a 36-stranded barrel from nine subunits, MspA has two consecutive barrels and aerolysin two concentric ones. ClyA and FraC are α-helical.^{[4][9][6][11][14]} Becoming a pore is a large change for the protein. ClyA starts as a soluble monomer, and on the way to a 400 kDa dodecameric pore more than half of its residues change conformation and parts of the monomer move by up to 14 nm.^[14]

In 1994 the channels of the peptide alamethicin in a lipid bilayer registered single polyethylene glycol molecules with gyration radii as small as 0.5 to 1.5 nm, in a paper that began with a description of the Coulter counter.^[21] For DNA, alpha-hemolysin was the natural choice, for practical reasons. Publications had suggested that its channel was wider than 1.2 nm, enough to take a single strand of DNA, and while most channels open and close with voltage, which would have confused the observation of nucleic acid passing through, a few large ones, alpha-hemolysin among them, can stay open for extended periods and carry a continuous current.^[1][5] The patent that its authors call the first for nanopore sequencing, filed in 1995, named alpha-hemolysin and maltoporin as preferred channels and suggested a channel whose limiting aperture is much shorter than its whole length, so that the current would reflect fewer bases at a time.^[22][5]

A pore with that property arrived with MspA, whose constriction is about 1.2 nm wide and no more than 0.6 nm long, nearly an order of magnitude shorter than the stem of alpha-hemolysin. A physicist at the University of Washington predicted that it would be the better detector of single bases, and wild-type MspA nevertheless refused DNA. Mutants in which negatively charged amino acids lining the pore were replaced by neutral or positive ones took it up, and by 2010 they separated all four bases with larger current differences than alpha-hemolysin gave.^[5][7][23]

CsgG came by a different route. It is the secretion channel for curli fibers, and the authors of its structure paper call it "an ungated, non-selective protein secretion channel".^[9] Oxford Nanopore announced in 2016 that the pore of its new R9 flow cells was CsgG, licensed from a laboratory at the VIB research institute in Belgium and from University College London.^[2] The authors of a 2016 history of the field wrote in reply to a critic that the company had held the identity of its pore closely until shortly before.^[24] CsgG's structure had been published two years earlier, and its 0.9 nm constriction is formed by three stacked rings of phenylalanine, asparagine and tyrosine.^[9]

Early experiments formed a lipid bilayer of diphytanoylphosphatidylcholine across a hole about 0.1 mm wide in a Teflon sheet 25 µm thick, with one pore inserted into it.^[1] Oxford Nanopore says that its pores sit in "an electrically-resistant polymer membrane", each on its own microscaffold, with an application-specific integrated circuit measuring each channel individually.^[25] A 2012 report on the announcement of the MinION says that early versions used lipid membranes and found them too fragile for a production sequencer.^[26] A 2016 description of the MinION gives its flow cell 2,048 pores in separate membranes, of which 512 are monitored at a time.^[5]

Biological and solid-state pores have also been combined. In 2010 a collaboration between the group of Cees Dekker at Delft University of Technology and the group of Hagan Bayley inserted a single alpha-hemolysin pore into a solid-state nanopore. A double-stranded DNA attached to the protein was threaded into the solid-state pore by electrophoresis, which directed the protein into place. Insertion succeeded in 30% to 40% of attempts, and the passage of single-stranded DNA showed that the hybrid pore still worked.^[27] The authors offered the hybrid as a route to wafer-scale arrays, a setting in which the protein alone is hard to use because it relies on delicate lipid bilayers for mechanical support.^[27]

Unaided, DNA drawn through a pore by the field moves each base through the sensing region in less than about 1 to 10 µs, and reading a base from picoampere signals needs more than 100 µs.^[5] The solution was an enzyme. A polymerase or helicase bound to the DNA at the mouth of the pore ratchets the strand through, one base per step, and the steps appear as stairs in the current. In October 2011 a phi29 DNA polymerase was combined with an engineered MspA, and reads of a repeating three-base sequence were shown at a conference in February 2012; the two papers that followed described ratcheting "at 5-Å precision" and reading "at single-nucleotide resolution".^[5][28][29] A helicase of the Hel308 family moves DNA through MspA in steps of about 0.33 nm, half a nucleotide.^[8] Oxford Nanopore's R7 chemistry ran at about 70 bases a second per pore, and its CsgG-based R9, introduced in 2016, at about 250.^[2] At 250 bases a second each base spends about 4 ms in the pore, 400 to 4,000 times longer than in unassisted passage.

The current at any instant reflects more than one base. In alpha-hemolysin about 12 nucleobases occupy the 5 nm stem, and all of them are reflected in the current level. In MspA, molecular dynamics simulations suggest that up to five bases still affect each level.^[5] Turning such currents back into letters is the work of basecalling.

Pores were sensors of small molecules before they were read as DNA sequencers. In stochastic sensing, a single engineered pore carries a binding site, and every binding event appears as a step in the current. The frequency of events gives the concentration of the analyte, and its identity is deduced from the magnitude and duration of the steps.^[30][31] Engineered alpha-hemolysin identified and quantified divalent metal ions first. In 1999 a group at Texas A&M held a cyclodextrin inside the channel as an adapter that mediates blocking by organic analytes, and one pore could then analyze a mixture of molecules with different binding characteristics.^[30] A decade later a protein pore with a covalently attached adapter molecule identified nucleoside monophosphates, the products an exonuclease would release from DNA, with an average accuracy of 99.8%.^[32]

Proteins have no uniform charge, so the field does not carry them through a pore the way it carries DNA, and several pores have been engineered to get around that. An electro-osmotic flow engineered into FraC lets polypeptides enter at a fixed voltage whatever their charge. In CytK, sets of charges spaced about 1 nm apart in the lumen create a flow that moves unstructured polypeptides against a strong electrophoretic force, and simulations put the strength of the resulting force at about 20 pN at -100 mV, similar to the electric force on single-stranded DNA.^[18][20]

ClyA is wide enough to take in folded proteins. Human and bovine thrombin, which share 86% of their sequence, gave main current levels that differed by 26 pA, at -50 mV.^[15] The unfoldase ClpX moved proteins through alpha-hemolysin in 2013, and in 2024 it moved them through CsgG on a commercial sensor array in two-residue steps.^[33][10] A helicase pulled a DNA-peptide conjugate through MspA in 2021. A single peptide was read 117 times in one recording, and combining repeated reads brought the error in identifying a single amino acid change below one in a million.^[8]

Natural pores are no longer the only ones. DNA pores that span lipid bilayers were reported in 2013, a large-conductance porin made from DNA origami in 2016, computationally designed transmembrane pores of 12 and 16 helices in 2020, and designed β-barrel pores whose conductances run from 110 pS to 430 pS, for diameters of about 0.5 nm to 1.1 nm, in 2024.^{[34][35][36][37]}

In 2012 the authors of the paper that introduced ClyA as a nanopore wrote that biological nanopores "generally perform better than artificial nanopores as sensors, but they have disadvantages including a fixed diameter."^[15] The sequence of the protein sets the diameter, so an analyte too large for the pore cannot enter, and a small one in a wide pore gives a weak current blockade.

Membranes are the other weak point. Oxford Nanopore instructs that flow cells be stored at 2 to 8 °C from receipt until use, that they never be frozen, and that they be used and returned within 12 weeks.^[38]

Molecular Reality's public guide sets the trade the other way. It describes biological pores as "protein channels that bacteria build themselves" and notes that "the famous MinION sequencer uses those". It describes solid-state pores as "machined chips, usually holes in a thin silicon nitride membrane" that "can survive heat, harsh chemistry, and being tuned to different sizes".^[3] Asked whether it can read DNA yet, the guide answers: "No solid state pore has done it commercially yet." It adds: "That is the plan, and I flag it as a plan."^[3]

• Nanopore sequencing
• Lipid bilayer
• Solid-state nanopore
• History of nanopore sequencing
• Protein sequencing

• Branton, D.; et al. (2008). "The potential and challenges of nanopore sequencing". Nature Biotechnology 26 (10): 1146-1153. doi:10.1038/nbt.1495.