De novo design of a nanopore for single-molecule detection that incorporates a β-hairpin peptide

The amino-acid sequence of a protein encodes information on its three-dimensional structure and specific functionality. De novo design has emerged as a method to manipulate the primary structure for the development of artificial proteins and peptides with desired functionality. This paper describes the de novo design of a pore-forming peptide, named SV28, that has a β-hairpin structure and assembles to form a stable nanopore in a bilayer lipid membrane. This large synthetic nanopore is an entirely artificial device for practical applications. The peptide forms multidispersely sized nanopore structures ranging from 1.7 to 6.3 nm in diameter and can detect DNAs. To form a monodispersely sized nanopore, we redesigned the SV28 by introducing a glycine-kink mutation. The resulting redesigned peptide forms a monodisperse pore with a diameter of 1.7 nm leading to detection of a single polypeptide chain. Such de novo design of a β-hairpin peptide has the potential to create artificial nanopores, which can be size adjusted to a target molecule.


MD simulation
The MD simulations of β-barrel structures of SV28 consisting of five and eleven peptides were performed in a DOPC membrane using GROMACS-5.1.4 and -2021.1 1 and CHARMM36 force field 2 . The 3D structures of the SV28 and SVG28 peptides were modeled via a homology modeling technique, which allows prediction of the structure of proteins based on the sequence similarity. 3 HASR protein (PDB ID: 3CSL; chain A; sequence positions, 484-511) 4 was selected as a template structure for the peptides from the Protein Data Bank using BLAST search. 5 The sequence covering and charged amino acid positions were applied as the main criteria for the template selection (Fig. S1). After modeling the β-barrel 3D structures of 5-mer, 7-mer, 11-mer, and 16-mer of SV28 and 7mer SVG28 (Figs. S3 and S21), the simulation systems were prepared by CHARMM-GUI membrane builder. 6 Then, the standard minimization, equilibration, and production procedures of the builder 6 were applied at 300 K temperature. Initially, 5-mer and 11-mer SV28 systems were equilibrated for 100 ns (Fig. S1). Finally, 900 ns MD simulations of 5-mer and 11-mer SV28 in DOPC membrane were performed under NPT conditions. 5mer and 11-mer SV28 system were simulated for a total of 1 µs. The final simulations of the systems were analyzed and discussed. Additionally, MD simulations of 7-mer and 16mer SV28 and 7-mer SVG28 pores were performed for 200 ns. The structures were represented by VMD software. 2 Analyses were performed using GROMACS packages, excluding the pore diameter, which was analyzed using HOLE software. 3 5 Fig. S1 (a) Sequence alignment of SV28 and HASR proteins (PDB ID: 3CSN, Chain A). Blue and red bold letters indicate the positively and negatively charged residues respectively. (b) A ribbon structure of the manually constructed initial model of 11mer SV28. (c) An example of the 11-mer molecular system after a 100 ns long equilibration simulation. Red ribbons show the β-sheet regions of the structures with arrows pointing from the N-terminus to the C-terminus. (d) Secondary structure profiles of half-(upper) and full (lower) length of SV28 peptides during the MD simulations of the monomer. Ribbons show the peptide structures, with the secondary structure indicated by the color of the ribbon (red: β-sheet, cyan: turn, white: random coil structure). Ribbon arrows indicate the direction of the backbone from N-terminal to Cterminal. The pore radius of 5-mer and 11-mer were shown. The pore structures were analyzed by HOLE software and displayed as blue surfaces inside of the barrels. Val10 and Val22 amino acids showing central rim of the pores were displayed as the licorice models. Cyan lines indicate water molecules, and the lipid molecules were omitted for clarity (excluding phosphorus atoms as orange spheres). Green and white spheres indicate the potassium and chloride ions respectively. Structures were displayed by VMD software. The middle and right plots displayed the radius of the final snapshots of 11-mer and 5-mer pore, which were drown by CHARMM-GUI.

Fig. S3
(a) Intrapeptide distance between nitrogen of Val10 and carbon of Val22 during the simulations of 5-mer (black) and 11-mer (red). The distribution of the distance changing in the 5-mer was slightly larger than that in the 11-mer. 5-mer pore structures have shorter turns than 11-mer peptides, which can give structural distortion to the monomer structures. Because of this distortion, the distances in the 5-mer may become longer than that of 11-mer. (b) MD simulation of SV28 with 7-mer and 16-mer after 200 ns running. 1-13 C Labeled Tyrosine (1.15 g, 6.35 mmol) was dissolved in water (40 mL), followed by stepwise addition of Fmoc-OSu (3.71 g, 1.16 mmol) and NaHCO3 (2.96 g, 3.45 mmol), with stirring for 1 day at room temperature. After the reaction mixture was neutralized with 5% hydrochloric acid, EtOAc was added, and the mixture was washed with saturated brine. The organic phase was dried over MgSO4, and removed in vacuo. The residue was applied to a silica gel column and eluted with CHCl3-MeOH (10:0.5) to give 2.35 g (75.7%) of the title compound as a white powder. 1 H NMR (500MHz, CDCl3).

Fmoc-1-13 C Tyr(tBu)-OH
Fmoc-1-13 C-Tyr(tBu)-OCH2CCl3 (1.05 g, 1.78 mmol) was dissolved in tetrahydrofuran (24 mL), and subsequently 50% aqueous acetic acid and zinc powder (5.14 g, 74.7 mmol) were added and stirred for 1 hour at room temperature. After filtration, ethyl acetate was added and the mixture washed sequentially with saturated aqueous sodium chloride and water. The organic phase was dried over MgSO4 and removed in vacuo. The residue was applied to a silica gel column and eluted with a CHCl3-MeOH (10:1) to give 306 mg Fmoc-2-13 C Gly-OH and Fmoc-1-13 C Val-OH 10 Fmoc-2-13 C Gly-OH and Fmoc-1-13 C Val-OH (99% isotopically enriched) were synthesized by the reaction of 9-fluorenylmethyl N-succinimidyl carbonate (Fmoc-OSu) with the 13 C labeled amino acids (Cambridge Isotope Laboratory CIL) based on the following reaction described by A. Paquet: 4 Fmoc-OSu (1 mmol) was reacted with one equivalent of the isotopically labeled amino acid in a water-acetone mixture in the presence of sodium bicarbonate (1 mmol). After stirring overnight at 30 °C, the mixture was acidified to pH 2 with concentrated HCl, and acetone was removed in vacuo. The product was dissolved in chloroform and washed with 0.1 N HCl and water. The organic phases were dried, and the residue was recrystallized by hexane-CH2Cl2. The melting points (m. p.) of Fmoc-2-13 C Gly-OH (174 °C) and Fmoc-1-13 C Val-OH (142 °C) were checked by a capillary method. The m. p. well agreed with the data 5 .

Confirmation of acyl migration by HPLC
The acyl migration of the peptides was confirmed by checking the decrease in the peak area of precursor peptides in HPLC (Fig. S6) because the final products after acyl migration were easily aggregated and were hardly eluted in standard HPLC condition.
The precursors of SV28, SV28-B, and SVG28 were added to 100 mM KOH and The peptides were analyzed by absorbance at 220 nm on an Inertsil ODS-3 column (150 x 4.6 mm, 5 µm, GL Sciences Inc.) using isocratic condition with 100% of MilliQ (containing 0.1% TFA) over 5 minutes, and then a linear gradient from 0% to 100% acetonitrile (containing 0.08% TFA) over 30 min at a flow rate of 1.0 mL/min (Fig. S6).

Measurements of CD spectroscopy
CD spectroscopy was performed at room temperature using the synthesized peptides. In  and a chase between the wells were manufactured on the PMMA plate. Each well had a through-hole in the bottom and Ag/AgCl electrodes set into these holes (Fig. S7c). A polymeric film made of parylene C (polychloro-p-xylylene) with a thickness of 5 μm was patterned with single pores (100 μm diameter.) using conventional photolithography methods, 11 and then fixed between PMMA films (0.2 mm thick) using an adhesive bond (Super X, Cemedine Co., Ltd, Tokyo, Japan). The films, including the parylene film, were inserted into the chase to separate the wells. High throughput measurement (Fig. S7d) was conducted using a JET patch-clamp amplifier (Tecella, Foothill Ranch, CA, USA). 12

Preparation of bilayer lipid membrane and alpha-hemolysin pretreatments
BLMs were prepared in the same method as above using DPhPC (lipids/n-decane, 10 mg/mL) solution. The buffer solution (4.7 μL) with αHL (final concentration 50 nM) was poured into the recording chamber. The buffer solution (4.7 μL) was poured into the ground chamber. In this experiment, a buffer solution (1 M KCl, 10 mM MOPS, pH 7.0) was used.  (Fig. 3g and 3h) were prepared for representing the relationships between the channel conductance and the number of βstrands (Fig. 3g) and between the pore diameter and the number of β-strands (Fig. 3h). 23 Using these calibration curves, we estimated the pore diameter and the number of βstrands of the SV28 and SVG28. We set a threshold conductance form the open level to determine the single-molecule detection in the channel current recoding: SV28-dsDNA 1nS (Fig. 4a); SV28-G4 0.5 nS ( Fig. 4i and 4j); SVG28-PLL 0.4 nS (Fig. 5f); αHL-PLL 0.2 nS (Fig. 5l and 5m). Peaks of conductance histograms (Fig. 3f and Fig. 5c) were founded by 2nd derivative method (threshold: 15% of the counts) and fitted by nonlinear and Gaussian curve fitting method (see the detail on the website of OriginLab:
We considered how SV28 forms a nanopore with a β-barreled structure. There have been many reports on the pore/channel formation of α-helical peptides. These peptides initially bind to the surface of the lipid membrane and form α-helical structures, which subsequently assemble from the monomers to construct the transmembrane nanopore structure. Several pore-forming models have already been proposed for these structures, such as barrel-stave or toroidal models. 18 We have also proposed the assignment of current signals to these models in planar lipid bilayer experiments. 19 Although there are few studies on the pore-formation of β-sheet peptides, it has been reported that the β-sheet peptides also construct barrel-stave and toroidal pores. 20 Our electrophysiological measurements herein also display step and multi-level signals, analogous to those previously assigned to the barrel-stave and the toroidal models.
Although we used experimentally estimation of the pore diameter, the pore diameter of the SV28 nanopore was also calculated using the conductance of the open channel state and the Hille equation, which is a theoretical model that uses the resistance of a cylindrical pore to ion flow. 21 The open channel conductance was determined as the initial step signal 24 from the baseline (≈0 A). The histogram of the pore conductance of SV28 is shown in Fig. 3f. Several peaks are observed in this histogram, with five identified peaks picked by the second deviation method. The peak conductance at 1, 3, 7, 11, and 14 nS are identified, and these give pore diameters of 0.8, 1.7, 2.3, 2.7, and 3.5 nm respectively using the Hille model. The numbers of monomers used in the nanopore assembly were mathematically calculated to be 4, 5, 7, 8, and 10 monomers using the diameters and the size of the β-hairpin molecule.

Fig. S11
The definition of current signal classification.
Step-like signal: the current sharply increases (within 10 ms) and maintains a plateau state (longer than 1 s). Squaretop signal: the current sharply increases (within 10 ms) and proceeds to transit plateau states (shorter than 1 s). Multi-level signal: the current sharply increases (shorter than 10 ms) and proceeds to fluctuate. Fluctuation defined as when the 95% confidence interval of the open level current is larger than that of the baseline. Erratic signal: the current randomly increases with fluctuation.

Detection of poly-L-lysine
In the detection of poly-L-lysine, we used long-poly-L-lysine (L-PLL) which has 132~300 amino acids, and short-poly-L-lysine (S-PLL) which has 50 amino acids. Both L-PLL and S-PLL were added to the buffer solution in a recording chamber. In detection using SVG28, the buffer solution with 24 hours incubated SVG28 was added in a ground chamber. In detection using αHL, the buffer solution with αHL (final concentration 50 nM) was added in a recording chamber with a DPhPC bilayer. The threshold of PLL translocation using SVG28 and αHL was decided as blocking of more than 0.4 and 0.2 27 nS respectively from the open-pore current level. In the exact bootstrap method, the verification of accuracy will be made possible when the sample number is over 200. In this study, our bootstrap procedure took 1000 or 300 samples randomly from the primary common translocation data with 65536 replacements and to calculate the means for these samples. 28 3. Single-molecule detection using SV28 and SVG28 nanopore (Fig. S12 -S19