Membrane-Suspended Nanopores in Microchip Arrays for Stochastic Transport Recording and Sensing

The transport of nutrients, xenobiotics, and signaling molecules across biological membranes is essential for life. As gatekeepers of cells, membrane proteins and nanopores are key targets in pharmaceutical research and industry. Multiple techniques help in elucidating, utilizing, or mimicking the function of biological membrane-embedded nanodevices. In particular, the use of DNA origami to construct simple nanopores based on the predictable folding of nucleotides provides a promising direction for innovative sensing and sequencing approaches. Knowledge of translocation characteristics is crucial to link structural design with function. Here, we summarize recent developments and compare features of membrane-embedded nanopores with solid-state analogues. We also describe how their translocation properties are characterized by microchip systems. The recently developed silicon chips, comprising solid-state nanopores of 80 nm connecting femtoliter cavities in combination with vesicle spreading and formation of nanopore-suspended membranes, will pave the way to characterize translocation properties of nanopores and membrane proteins in high-throughput and at single-transporter resolution.

They are compatible with cell, organelle, and liposome applications and possess a great potential for biological functions and interactions (Banghart et al., 2004;Volgraf et al., 2006;Howorka, 2017). Nevertheless, the number of channels per square area cannot be controlled precisely, and they lack the chemical and environmental stability of their solidstate counterparts Wang et al., 2011;Yusko et al., 2011;Wei et al., 2012a;Traversi et al., 2013;Heerema and Dekker, 2016;Ding et al., 2020). By reducing the bilayer patches (Urban et al., 2014) and using polymerized lipids (Daly et al., 2006;Heitz et al., 2010) or amphiphilic polymers Nardin et al., 2000), the stability of membrane nanopores can be increased.
Protein pores cannot be easily engineered from scratch as the folding of a polypeptide chain in the quaternary structure is highly complex, and defined pores of >5 nm are very rare or experimentally challenging. For example, the release of proteins from therapeutic vesicles is yet not possible with biological channels (Mahendran et al., 2017). Here, the recently engineered DNA origami-based nanopores constitute a prosperous alternative, featuring the characteristics of simple self-assembly in a bottom-up fashion (Langecker et al., 2012;Burns et al., 2013;Seifert et al., 2015;Burns et al., 2016;Chidchob et al., 2016;Göpfrich et al., 2016a;Göpfrich et al., 2016b;Howorka, 2016;Krishnan et al., 2016;Diederichs et al., 2019;Thomsen et al., 2019;Chen et al., 2020). DNA nanopores can be precisely designed due to the predictable and programmable folding scheme based on simple base-pairing rules of DNA (Douglas et al., 2009a;Douglas et al., 2009b). De novo design generates structures of >20 nm (Rothemund, 2006;Seeman, 2010;Pinheiro et al., 2011;Chen et al., 2015;Jones et al., 2015), which are assembled by simple heating and cooling protocols in the classical scaffold and staple approach, according to which a long scaffold strand is folded with shorter staple nucleotides into a DNA structure (Rothemund, 2006). DNA nanostructures with controlled curvature and twists were generated (Dietz et al., 2009). The first membrane-embedded DNA nanopore was developed by the Simmel group (Langecker et al., 2012) ( Figure 1A), directly followed by a six-helix bundle pore of the Howorka lab (Burns et al., 2013) (Figure 1B), likely indicating an independent parallel development. Both structures share a core of six hexagonally arranged DNA duplexes, differing in wall thickness and height (Langecker et al., 2012;Burns et al., 2013). However, also combinations of DNA origami structures or pores with solidstate nanopores or glass capillaries were developed (Bell et al., 2012;Wei et al., 2012b;Hernández-Ainsa et al., 2013;Hernández-Ainsa et al., 2014). For a detailed description see references (Bell and Keyser, 2014;Göpfrich and Keyser, 2019). Different pore widths were constructed, ranging from 0.5 to 9.6 nm (Krishnan et al., 2016;Göpfrich et al., 2016b;Thomsen et al., 2019), even allowing the translocation of folded proteins (Diederichs et al., 2019) or enabling the interaction with tumor cells (Guo et al., 2017). Nanopores with pore diameters of ∼4.0 and ∼7.5 nm are illustrated in Figures 1C,D (Krishnan et al., 2016;Diederichs et al., 2019). Lipid insertions were achieved by introducing anchors of cholesterol (Langecker et al., 2012;Burns et al., 2013;Burns et al., 2016), porphyrin (Burns et al., 2014;Seifert et al., 2015), or tocopherol (Krishnan et al., 2016). Despite successful incorporation in bilayers, aggregations of multiple pores were observed after attachment of lipophilic groups (Burns and Howorka, 2019;Ohmann et al., 2019). Other modifications such as the removal of negative charges at the backbone (Burns et al., 2013;Burns et al., 2014) or the replacement with neutral peptide nucleic acids were established (Ackermann and Famulok, 2013). By using complementary single strands acting as lock and key at the nanopore entrance, the efflux of small molecules through a DNA nanopore can be controlled, peaking in the discrimination of cargos based on their charges (Burns et al., 2016). The disadvantages of DNA as building blocks are the flexibility and loose packing of the structures so that ions can diffuse through their walls (Maingi et al., 2015;Yoo and Aksimentiev, 2015), which can be partially improved by thickening of the nanopore scaffold (Langecker et al., 2012). It is also reported that small water-filled channels can be formed at the interface between nanopore and membrane, allowing the passage of small molecules (Göpfrich et al., 2016a). The strengths and weaknesses of solid-state and membrane-embedded nanopores are summarized in Figure 2.

Microchip Arrays Suspended By Nanopore-Embedded Biological Membranes
Nanopores are often combined with microchips to ensure parallelization, high-throughput, and single-molecule resolution. Here, micro-or nanocavities are connected via solid-state nanopores with a second aqueous compartment. These nanostructures are fabricated in top-down or bottom-up approaches by photolithography or electron-beam lithography in combination with etching techniques, thermal oxidation, spin coating, and sputtering (Sharma et al., 2006;Buchholz et al., 2008;Herzer et al., 2010;Liebes-Peer et al., 2016;Xu et al., 2019). They provide benefits of miniaturized sizes, high reliability, parallelization, and biocompatibility (Jeon et al., 2010;Ho et al., 2013;Meng et al., 2014;Santos et al., 2015). By spanning lipid bilayers across these solid connections, translocation processes of membrane-embedded nanopores or membrane proteins became observable. These chip architectures are ideal for fluorescent-based high-throughput screenings as exemplified by α-hemolysin and the mechanosensitive channel of large conductance (MscL) (Kleefen et al., 2010;Urban et al., 2014;Diederichs et al., 2018). Membrane coverage across the nanoapertures is achieved by techniques such as lipid painting from organic solvents (Han et al., 2007), vesicle spreading in physiological buffers (Schmidt et al., 2000;Buchholz et al., 2008;Kleefen et al., 2010;Kumar et al., 2011;Wu et al., 2013), or sheardriven membrane formation (Jönsson et al., 2009). The selforganized vesicle spreading approach enabled the use of proteoliposomes resulting in solvent-free membranes harboring complex proteins. Here, the vesicle diameter is critical to successfully ensure coverage by simultaneously avoiding lipid intrusions (Urban et al., 2014;Janshoff and Steinem, 2015). The suspended membrane combines the advantages of increased stability provided by the solid support and high membrane fluidity created by the free-standing patches (Sun et al., 2020). Nevertheless, a tight seal for small ions could not be demonstrated so far. In contrast, painting requires the use of organic solvents, which can harm the function of membrane proteins, however, supports a tight membrane seal impermeable for small ions (Watanabe et al., 2014a).

Femtoliter Cavity-Based Microchips for Optical Recordings
The first cavity-based chips for optical single-channel recordings were developed in the 1990s mainly by the Peters group (Peters et al., 1990;Tschödrich-Rotter et al., 1996;Tschödrich-Rotter and Peters, 1998;Siebrasse and Peters, 2002;Kiskin et al., 2003;Peters, 2003). However, fabrication techniques regrading reliability, accuracy, and parallelism were limited. With the development of advanced fabrication methods, micro-and nanopore arrays with thousands of micro-or nanocavities were constructed and covered by lipid membranes (Buchholz et al., 2008;Sumitomo et al., 2012;Watanabe et al., 2014b). These microchip systems allow high-throughput screening applications for individual translocation processes as exemplified by the bacterial pore α-hemolysin (Kleefen et al., 2010;Watanabe et al., 2014a) and MscL (Urban et al., 2014). Applications are not limited to the analysis of channels and transporters, as cell-penetrating peptides (Kolesinska et al., 2013) or viral capsid membrane fusion and cargo release were investigated with cavity-based chips (Kusters et al., 2014). The Noji lab developed carbon fluorine polymer (CYTOP) coated chips by combining microfluidics and sequential liquid injections (Soga et al., 2015;Watanabe et al., 2016). First, an aqueous solution is infused to fill the cavities with substrates, followed by phospholipids dissolved in chloroform, creating a lipid monolayer at the chamber opening due to the hydrophobicity of the coated material. Finally, a second aqueous solution is injected forming a bilayer at the cavity entrance, whereas a monolayer is formed elsewhere (Watanabe, 2020). Asymmetric membranes were generated by introducing a second chloroform-based solution supplemented with different lipid species (Watanabe et al., 2014b). In addition to the bacterial pore α-hemolysin, the F o F 1 -ATP synthase (Watanabe et al., 2014a) and the phospholipid scramblase TMEM16F were examined by this approach (Watanabe et al., 2018b). Based on the advection-diffusion model, the sequential injection of aqueous solutions with various concentrations of analytes allows to produce concentration gradients of encapsulated molecules across the chip (Watanabe et al., 2018a). Recently, the platform was used to measure single-protein activities in high-throughput by encapsulating soluble enzymes in the microchambers (Sakamoto et al., 2020) or creating hybrid cell reactor systems (Moriizumi et al., 2018).
Despite these multiple applications, the use of organic solvent does not allow the investigation of complex and fragile membrane proteins as required for biomedical screenings. In general, the vesicle-spreading method is superior, as delicate transport complexes can be reconstituted beforehand in proteoliposomes followed by SLB formation as shown in Urban et al. (2014). For vesicle spreading, different chip architectures exist, varying in cavity shape, volume, nanopore diameter, and surface roughness. A comparison of two chip designs published by Urban et al. (2014) and Diederichs et al. (2018) is displayed in Figure 3. The latest silicon chip features truncated inverted pyramidal cavities of ∼50 fL volumes, which are connected by a solid-state nanopore of 80 nm with a buffer reservoir. The transparent Si 3 N 4 top and SiO 2 bottom layers allow readout of both compartments, whereas the second chip architecture displays cylindrical cavities of ∼6-15 fL with nanopores ranging in diameter between 70 and 120 nm. A large difference was observed in the surface roughness which is ∼3.6 nm for the opaque chip ( Figure 3A), whereas the most recent chip design exhibits a surface roughness of ∼0.2 nm ( Figure 3B). These extremely smooth, low strain surfaces in combination with well-defined and tapered pore edges provide optimal conditions for a high coverage upon membrane spreading.

Membrane-Embedded DNA Nanopores Analyzed by Microchips
The recently designed silicon microcavity array was used to analyze translocation processes of two different DNA nanopores at single-channel resolution in high-throughput. The first investigated DNA pore is composed of six interconnected oligonucleotides forming a channel with a height of ∼9.0 nm, a width of ∼5.0 nm, and an inner diameter of ∼1.8 nm (Diederichs et al., 2021). Cholesterol-functionalized DNA strands were attached to drive membrane insertion. This six-helix bundle was used to elucidate how translocation is affected by the analytes' charge and size. Three different organic fluorophores possessing different net charges and sizes were studied. The translocation rate through the negatively charged DNA nanopore decreased with increasing negative net charges of the analytes. The (-1)-charged fluorophore ATTO655 fluxes faster than the (-3)-charged dye FAM, whereas the (-4)charged complexometric calcium indicator Calcein cannot translocate through the DNA nanopore ( Figures 4A,B) (Diederichs et al., 2021). This charge-dependent effect is most likely based on repulsion with the negatively charged pore walls of the DNA backbone. However, only a minor effect of the translocation is assumed for analytes with small numbers of net charges, as the flux of the (-1)-charged ATTO655 is only ∼2.5 times faster through the protein pore α-hemolysin, after correction for the cross-sectional area of the pore lumen (Diederichs et al., 2021). Strikingly, the lipid environment affects the translocation properties of the DNA nanopore, as 30 mol% of the negatively charged 1-palmitoyl-2-oleoyl-snglycero-3-phospho-(1′-rac-glycerol) (POPG) limits the  translocation of the (-3)-charged dye FAM (Diederichs et al., 2021). POPG is expected to cause a constriction of the pore lumen based on electrostatic compensation between lipid and DNA backbone. Thereby, a direct effect of the lipid environment on translocation properties of DNA nanopores was demonstrated (Diederichs et al., 2021). Small pores are well suitable for DNA sequencing due to the tight fitting within the translocation pore (Clarke et al., 2009;Quick et al., 2016). However, bigger pores provide the opportunity for stochastic sensing of proteins and complexes. Currently, biological pores are rather unsuitable for the passage of folded proteins based on their heterogeneities in diameter and due to their complex structures (Knockenhauer and Schwartz, 2016;Kosinski et al., 2016;Lukoyanova et al., 2016;Mahendran et al., 2017). Recently, the new nanopore silicon chip was used to demonstrate translocation of folded EGFP molecules through an artificial DNA nanopore (Diederichs et al., 2019). The latter resembles a funnel with a narrow and wide opening of 7.5 and 22.5 nm and a total height of 46 nm (Diederichs et al., 2019).
Smooth membrane insertion was ensured by 24 cholesterol anchors, and the stability of the pore was guaranteed by multiple layers of duplexes to increase the wall thickness. A remarkable number of ∼66 EGFP molecules can flux through the DNA assembly per second (Diederichs et al., 2019) ( Figure  4C). To mediate transport surpassing the equilibrium of passive diffusion, DNA nanopore protein translocation was coupled to biomolecular antibody recognition, allowing the affinity-based accumulation or trapping of EGFP molecules in the nanoconfinements (Diederichs et al., 2019) (Figure 4D). Current limitations of the recently developed silicon-based cavity array are the comparatively large cavity volumes of 50 fL, permitting the visualization of substrate translocation processes greater than one per second. Additionally, the characterization of transport processes of secondary active transporters is difficult as the membrane does not seal the cavity for small ions, including protons.

Future Prospective
In general, the field of solid-state nanopores and biological counterparts has grown tremendously. Some years ago, it was not imaginable that synthetic pores would ever reach transport behaviors of biological channels (Tagliazucchi and Szleifer, 2015). The multitude of applications and the chemical and physical stimuli to gate conductance equals or exceeds multiple properties of their biological counterparts (Perez Sirkin et al., 2020). Nevertheless, synthetic devices are behind biological pores in ion selectivity, signal amplification, transport of larger analytes, specificity, and many more (Perez Sirkin et al., 2020). Biological analogs inspire research and drive the design of new nanopores also in the field of protein sensing, as it was shown for solid-state nanopores and DNA funnels (Diederichs et al., 2019;Zeng et al., 2019). More and more sensing and sequencing applications with a multitude of analytes are being published. Nevertheless, the possibilities of transferring the knowledge of structural aspects to functional properties are still limited. Especially the research on DNA nanopores has to be intensified regarding controllability, stability, and functionalization to pave the way for diagnostic applications. The detection of cancer biomarkers from blood samples derived from DNA-assisted nanopores, which can also be used for the development of drug-delivery systems, is a good example for this Ding et al., 2020).
By combining solid-state nanopores with microchip systems, single-molecule resolution can be linked with high-throughput and microfluidics. In future, smaller microchips with drastically reduced volumes will be fabricated allowing to characterize medical relevant transporters or to screen lipid membrane conditions affecting the translocation properties of membrane proteins. A promising direction for optical single-molecule detection with nanopores are zero-mode waveguides (ZMWs). ZMWs are small holes in metal films with sub-wavelength apertures for nanophotonic and sensing applications (Assad et al., 2017;Larkin et al., 2017;Han et al., 2019;Spitzberg et al., 2019) as commercialized by Pacific Biosciences for DNA sequencing (Levene et al., 2003;Foquet et al., 2008;Korlach et al., 2008;Lundquist et al., 2008;Eid et al., 2009;Metzker, 2010;Uemura et al., 2010). As the wavelength of light is smaller than the nanoaperture itself, light cannot penetrate in the confined chamber, which results in an exponential decay of the optical field creating a zeptoliter volume (Levene et al., 2003;Foquet et al., 2008). By immobilization of the DNA polymerase at the bottom of these cavities, only photolabeled molecules in the first tens of nanometer can be observed (Korlach et al., 2008;Eid et al., 2009). Sensing applications with ZMWs bear the risk of unspecific binding and biocompatibility, however, protein sensing has recently been demonstrated with palladium ZMWs indicating a promising direction (Klughammer and Dekker, 2021).

AUTHOR CONTRIBUTIONS
TD and RT wrote the manuscript. RT conceived and supervised the work.

FUNDING
This research was supported by the German Research Foundation (SFB 807, TA 157/12-1, and GRK 1986 to RT), the LOEWE program (DynaMem A3 to RT) and ERC Advanced Grant (789121 to RT).