![Figure 1 Protein pores and applications. a, X-ray structures of protein pores αHL, MspA, ClyA, and OmpG. The scale bar of 3 nm length is positioned where the pore spans the lipid bilayer. b, Sequencing with nanopores4. The left panel shows the membraneinserted MpsA pore. A transmembrane potential pulls a negatively charged singlestranded DNA to the trans side and straighten the strand to facilitate base recognition by the reading head. A DNA polymerase motor protein (red) steps DNA and achieves controlled translocation to read the base translocation as current levels (left panel) which reflect not only the translocating base but also neighboring bases. c, Stochastic sensing to detect individual analyte molecules. In this approach, ionic current flowing through a single channel is monitoring as a function of time to detect binding of analyte molecule to an engineered binding site to produce current blockades. In equilibrium, the kinetic rate constant for binding kon = 1/τon [A] where τon is the inter-event interval and [A] the analyte concentration, while the dissociation rate constant koff = 1/τoff whereby τoff is the event duration or dwell time71.](/figures/figure-1-protein-pores-and-applications-a-x-ray-structures-h23592kr.webp)
![Figure 4 Organic synthetic nanopores. a, Crown ethers arranged in stack via a α-helical oligo-leucine scaffold148. b, Rigid-rod β-barrels with four octiphenyl rods that carry eight Lamino acid pentapeptides and interdigitate via H-bonding to assemble the pore103. c,Ppillar[5]arenes where each phenyl group of the five-membered circular arene carrying two short phenylalanine peptides of alternating D-L-D isomer sequence to form interstrand H-boding152. d, Single-wall carbon nanotubes144. e, Helically folded polymeric scaffolds based on oxadiazole-linked tricylic pyrido[3,2-g]quinolines and linear, membraneinteracting alkane extensions which are not shown in the model for reasons of visual clarity155.](/figures/figure-4-organic-synthetic-nanopores-a-crown-ethers-arranged-2stdfvby.webp)
!
1
Molecular Engineering of Membrane Nanopores
Stefan Howorka
University College London, Institute of Structural and Molecular Biology, Department of
Chemistry, London, UK
Membrane nanopores – hollow nanoscale barrels that puncture biological or synthetic
membranes – have become powerful tools in chemical- and bio-sensing, and have
achieved notable success in portable DNA sequencing. The pores can be self-assembled
from a variety of materials, including proteins, peptides, synthetic-organic compounds,
and, more recently, DNA. But which building material is best for which application, and
what is the relationship between pore structure and function? In this Review, I critically
compare the characteristics of the different building materials, and explore the influence of
the building material on pore structure, dynamics and function. I also discuss the future
challenges of developing nanopore technology, and explore what the next-generation of
nanopore structures could be and where further practical applications might emerge.
Membrane nanopores are the most important border crossings in the molecular world.
They form water-filled openings across membrane barriers composed of lipids or semifluid
polymers in order to transport ionic or molecular cargo (Box 1). Given their small
dimensions, nanopores function as size-selective filters that can limit transport to individual
molecules. Nanopores can also select molecules based on charge and other
physicochemical properties, and act as stimulus-responsive molecular valves that open or
close and thereby regulate transport across membranes.
Reflecting these functions, membrane nanopores have been exploited for many
applications. Pores can help sequence individual DNA strands
1-4
, sense a wide range of
analytes of biomedical and environmental relevance
5-11
, and study single-molecule
chemistry and biophysics. They can also regulate transport across cellular bilayers
12,13
or
drug-delivery vesicles
14,15
, or rupture membranes of bacterial pathogens
16
. Sensing and
biophysical studies can also be realized with the related class of solid-state pores
fabricated in thin non-organic films or materials
5,9,10,17-21
. These are, however, outside the
main focus of this review.
In tune with the numerous applications, membrane nanopores can be composed of
several different materials. Historically, protein and peptide pores were the first to be used
given their pre-defined structure and ease of engineering. These biological pores later
inspired the creation of artificial versions built from synthetic organic materials. The most
recent nanopore class is obtained from folded DNA strands. As a characteristic, almost all
pores can be formed in bottom-up fashion by simple self-assembly of the smaller building
units. In this context, building material is used as loose term that encompasses the
smallest unit (e.g. amino acids for proteins), secondary structure elements, and higher-
order architectures.
!
2
With nanopore technology rapidly expanding, there are numerous unanswered
fundamental questions about the construction and function of membrane pores,
specifically about structure-function relationship. For example, what are the characteristics
of the building materials, and how do they differ in terms of chemical tuneability, and
suitability for bottom-up design? Furthermore, how do these characteristics influence the
pores’ structure, dynamics, function and applications? Answering these questions can
contribute to a coherent understanding that underpins the rational design of advanced
membrane nanopores. Knowing the strengths and weaknesses can also help select the
most suitable building material for a given pore application. A survey of existing pores may
finally identify future engineering targets. Excellent existing publications describe a single
or two pore types but very few cover all
7-10,22-30
.
This review compares the four membrane pore classes to obtain a comprehensive picture
of all building materials and their impact on pore design, structure and function (Box 2). An
overview first clarifies what constitutes a membrane nanopore and relates their
advantages and disadvantages to solid-state pores. The subsequent four sections cover
membrane pores composed of protein, peptide, DNA, and synthetic organic molecules, to
identify similarities and differences. For each of these classes, prominent natural or
engineered pores, strengths and weaknesses of the building material, and typical
applications are described. The review concludes by highlighting sophisticated biological
membrane proteins that can inspire the design of advanced nanopores.
Membrane nanopores vs. solid-state pores
A membrane pore is a hollow nanobarrel with a width usually in the range of 1-5 nm that
punctures a biological or synthetic semifluid membrane
31,32
. The detailed barrel structure
and transport properties have to be experimentally verified to establish whether a pore is
present (see Box 3). Relying solely on transport assays without structural confirmation can
be deceiving because synthetic lipids can locally deform the fluid bilayer
33
to yield pore-like
electrical and optical read-out traces. However, high-resolution structural analysis is not
possible for several peptide and synthetic pores that assemble exclusively in the
membrane
24,26,34
. These are nevertheless covered here provided a membrane-spanning
pore is strongly supported by its architecture and structural data. The exciting class of
carrier ionophores is not discussed as they do not form a contiguous nanobarrel
35,36
.
What are the advantages of membrane pores over solid-state pores within inorganic or
polymeric films? By inserting into lipid bilayers, assembled pores are compatible with
applications involving vesicles and cells
12,13
as well as membrane-based analysis
platforms
1-4
. In addition, facile engineering with atomistic precision can tune the pores’
dimensions, dynamics, and interactions with other molecules to a greater extent than
classical solid-state materials such as silicon nitride or silicon dioxide.
The advantage of membrane pores on atomic precision is, however, increasingly rivalled
by recent advances in solid-state pores formed inside atom-layer thin sheets of graphene
18
and MoS
2
20
. Pores in these synthetic materials are so small that they approach the size of
individual nucleotides. The pores could therefore offer the necessary spatial resolution for
nanopore DNA sequencing. As additional advantage, solid-state pores have great
chemical stability and withstand buffer conditions incompatible with fluid membranes. One
striking example is the use of MoS
2
pores with ionic liquids. These liquids are of high
viscosity and slow down the nucleotide translocation through the pore so that the different
isolated nucleotides can be resolved
20
.
!
3
Solid-state nanopores are also of higher mechanical stability than lipid bilayers even
though the latter’s shortcoming can be addressed by several approaches
37
. These include
the reduction of the lateral bilayer size
38
via droplet interface bilayers
39,40
, the use of
hydrogels or inorganic supports
41,42
, the inclusion of polymerizable lipids
43,44
, or the
replacement of lipids with amphiphilic polymers
31,32
. Solid-state materials also make it easy
to control how many pores are fabricated into the thin sheets
18,45-49
. Achieving one channel
per membrane unit is essential for single-molecule DNA sequencing but challenging with
membrane pores that stochastically insert into the semi-fluid membrane. As another
advantage, solid-state pores can easily be read out with advanced physical tunneling
methods not readily compatible with membrane nanopore systems
19
. Nevertheless,
membrane nanopores offer greater chemical, structural and nanomechanical tuneability as
detailed in the following chapters.
Protein pores
Given their multiple favorable features, protein pores are the current workhorses in
nanopore-based DNA sequencing and single-molecule studies
1,5
due to a range of
favorable features (Box 2). The protein pores of biological origin are of varied shape
(examples listed below) and usually have atomistically defined and structurally stable
scaffolds which facilitate rational engineering. Furthermore, proteins are made up of
modular architectural units such as β-sheets and α-helices. Structural fine-tuning can be
achieved with amino acids of different charges, size, functional groups, and
hydrophobicity. As other advantage, proteins are produced in cells or cellular extracts from
genetically engineered DNA templates. Consequently, it is easy to replace, add or delete
amino acids at defined positions, or fuse functional protein domains
50
. Engineering can
also help remove interfering floppy parts
51
or increase pore stability
52
.
While benefiting from biology, engineering of protein pores can tap into the rich repertoire
of synthetic chemistry to tailor the pores’ structure and function (Box 2). Chemical
interventions include the non-covalent placement of molecules into binding pockets, the
covalent modification of selected natural amino acids
53
, and the incorporation of non-
natural amino acids
53
. Moreover, the polypeptide scaffold can be partially replaced with
synthetic stretches via expressed protein ligation
54-56
where a biologically expressed
protein fragment is fused to a synthetic peptide. Alternatively, a complete chemical
synthesis can be achieved by native chemical ligation where two synthetic fragments are
chemically fused
57
.
Protein pores used in nanopore technology cover a broad range of shapes, and are
usually oligomeric and constitutively open. Most are bacterial cytotoxins or conduits for
passive transport. The protein pore used as reference is the cytotoxic heptameric αHL (α-
hemolysin)
58
. Sequencing and stochastic sensing was pioneered with this pore given its
easy insertion into membranes, the hour-long inserted state, the virtual absence of intrinsic
stochastic structural switching, and the stable pore geometry amenable to protein
engineering. αHL features a transmembrane β-barrel of less than 2 nm width and an
extramembrane cap (Fig. 1a). In the context of sequencing, αHL has been replaced by
shuttle pores MspA
59
and CsgG
60
. Cytotoxic ClyA
61
(Fig. 1a) of wider lumen can
accommodate small folded proteins
52
. OmpG
62
(Fig. 1a) and OmpF
63
are unusual given
their composition of solely of single polypeptide chains, while octameric Wza has
membrane-spanning α-helices
64
. Nanopore technology also uses the aerolysin pore
65
, the
potassium channel KscA
66
, mechanosensitive MscL
67
, and an engineered membrane-
inserting version of the bacteriophage phi29 DNA packaging motor
68
; but not cholesterol-
dependent cytolysins or other membrane attack complexes
69
due to their variable number
of subunits and unpredictable pore diameter. In the following, the versatility of pores is
!
4
illustrated by detailing their uses in sequencing, sensing, single-molecule chemistry and
biophysics, and cell biology and nanobiotechnology.
The premier application of protein pores is DNA strand sequencing
1-4
. In nanopore
sequencing, individual single-stranded DNA molecules translocate through a channel with
an internal narrow constriction that serves as reading head (Fig. 1b). Simultaneously
recording ionic current through a single channel reveals a step-wise read-out pattern that
reflects the base sequence (Fig. 1b). Nanopore sequencing is label-free and reads very
long DNA strands not accessible by competing techniques. It directly detects chemically
altered bases of biomedical relevance such as methylated or hydroxymethylated
cytosine
70
. The approach is miniaturizable to the size of a memory stick and hence
portable
1
. Nanopore sequencing also highlights the benefit of a defined pore scaffold and
tunable building blocks. Only a channel with no more than 1.5 nm width can permit the
passage of a single DNA strands as opposed to multiple simultaneously translocating
strands. Pores with suitable dimensions are αHL
58
, MspA
2,3
, and CsgG
60
. Furthermore, the
reading head can be optimized to better distinguish DNA sequences, such as by altering
the constriction’s diameter, height, and hydrophobicity by amino acid replacements
2-4
.
Recently, aerolysin has been shown to resolve individual short oligonucleotides that are 2
to 10 bases long
65
indicating that this pore might also be developed for sequencing.
Pores are also single-molecule sensors for non-DNA analytes
5,71
. They help uncover
scientifically relevant static or dynamic heterogeneities not accessible by conventional
ensemble methods. In one popular sensing mode –stochastic sensing
71
- separate analyte
molecules bind to a defined recognition within the pore to give rise to characteristic current
blockades (Fig. 1c) that differ from the constant current level of the non-bound pore.
Binding sites can be generated via genetic engineering and, optionally, with chemical
modification. One example is a histidine patch that recognizes metal cations within the
αHL pure lumen
72
. Another is a docking site for hollow β-cyclodextrin rings
73,74
that in turn
distinguish small organic drug molecules
73
as well as biomedically relevant
stereoisomers
75
.
In the second strategy of stochastic sensing, molecular receptors are covalently attached
to individual amino acids. Cysteine is a preferred residue as it is easily genetically
introduced, and specifically chemically modified at neutral pH and ambient temperature.
Several molecular receptors were coupled to cysteine such as DNA oligonucleotides
76,77
and protein ligands
78,79
. In these cases, the number and attachment position of molecular
receptors was carefully controlled to achieve a clear and strong current signal upon
analyte binding. For example, a single short DNA oligonucleotide was placed inside the 3
nm-wide cap lumen of αHL to enable hybridization with complementary DNA strands
76
. For
larger protein analytes, multiple recognition agents were positioned close at the channel
entrance
77
. In more extensive genetic engineering, a fusion was generated between a
molecular recognition unit – a G-protein coupled receptor- and a signal generator –a
potassium channel; analyte binding triggered a conformational relay to open the channel
and cause a current signal
50
.
When protein channels are applied to probe single-molecule chemistry, it is similarly
essential to control where and how many receptors are attached to the pore. Several
different chemistries were successfully examined at the single-molecule level. Examples
are the photo-chemical isomerization of diazo-dyes
80
, the kinetic isotopic effects in a
quinone reduction
81
, and organo-arsene reactions that were resolved at unprecedented
kinetic resolution
82
. Reversible bond formation along multiple cysteines created a random
molecular walker
83
.
Protein pores are also excellent tools to study the physico-chemistry of individual organic
and biological polymers. In this type of analysis, single polymers translocate through or
!
5
temporarily reside in a narrow pore without the need for an engineered binding site. The
approach can examine a wide range of biophysical aspects
5
. These studies are not
discussed further as channel engineering usually
84
does not have a key enabling role.
For cell biology and nanobiotechnology, pores have been engineered into stimulus-
responsive nanovalves to control membrane permeability. This is useful when examining
how a cell responds to a locally and temporally defined flux of ions or molecules. For
optical triggering of neuronal current, the Shaker K
+
channel was modified with a channel
blocker via an azobenzene linker. The chromophore’s wavelength-dependent cis-trans
photo-isomerization moved the blocker to and from the pore entrance and altered the
current
12
. A related light gate was installed in the glutamate receptor
13
. To make
membranes permeable for larger cargo, the up 3 nm-wide MscL channel was furnished at
key sites with multiple spiropyran chromophores that undergo reversible light-induced
charge separation and cause pore opening and re-sealing
85
. This nanodevices may be
applied e.g. for drug delivery. Thermal triggering was achieved by genetically placing in
αHL a polypeptide that adopts different temperature-dependent conformations
86
. Similarly,
voltage-sensitivity was programmed by genetically engineering 49 arginine residues into
the pore’s β-barrel; it collapsed at one but re-opened at the other potential
87
. Related
amino acid replacements installed a strong ion filter into the OmpF pore
88
while expressed
protein ligation fine-tuned the selective filter for K
+
in the KscA ion channel
54
. These
engineered channels can be used to build synthetic ionic networks
87
.
Despite many advantages, protein pores have shortcomings (Box 2). Proteins can be
immunogenic which limits their use in therapy. In addition, it is a challenge to create
protein pores with many non-natural amino acids; these can expand the pores’ function.
The experimental hurdle could be overcome with native chemical ligation which fuses
small synthetic peptides to larger biogenic protein fragments
89
. It is furthermore difficult to
make predictable and drastic structural changes such as varying the number of subunits,
with some rare exceptions
90
. Designing de novo pores is even more strenuous. Hence,
there is a shortage of pores wide enough to accommodate proteins or other
marcomolecules for sensing or transport. The following three chapters show that limited
chemical scope can be overcome with peptides and synthetic materials, while DNA offers
considerable freedom in de novo design.
Peptide pores
Peptide pores are smaller than proteins and have usually no more than 50 amino acid.
The short length is an advantage because it is easier to include residues other than the
standard set of 20 proteinogenic L-amino acids (Box 2). For example, peptides with D-
amino acid can be obtained via the nonribosomal biosynthesis
91
. Furthermore, peptides
may be entirely built from synthetic amino acid via solid-phase synthesis
22,92
. As a
consequence, chemically diverse peptide pores feature new scaffolds and functions.
Several biological examples illustrate the chemical and structural characteristics of peptide
pores. Antibiotic gramicidins have a length of only 15 alternating L- and D-amino acids and
assemble into a β-helix in which all residues point outward
93
(Fig. 2a); this is different to
protein β-barrels where sequential residues alternate between inward and outward
orientation. Due to its short length, one gramicidin helix only spans one membrane leaflet
but two half-channels can transiently dimerise to form a membrane-puncturing channel
94
(Fig. 2a). By comparison, the 48 residue-long antibiotic polytheonamide B
95
completely
transverses the membrane as a related β-helix. Another antibiotic peptide, alamethicin, is
composed of standard L- as well as non-traditional 2-aminoisobutyric acid and folds into a
α-helix
96
. Once inserted into a membrane, alameticin forms ion-selective channels of four