Programmable nano-reactors for stochastic sensing

The study addresses the long-standing challenge of directly monitoring single-molecule chemical reactions, which typically occur too rapidly and are traditionally inferred from ensemble measurements. While several advanced techniques (scanning probe microscopy, enhanced Raman spectroscopy, molecular junctions, single-molecule force spectroscopy, and biological nanopores) interrogate different molecular properties, they each have limitations for resolving discrete reaction steps. Prior nanopore approaches, especially engineered α-hemolysin pores, suffered from small signal amplitudes, distributed unwanted reactive sites, complex hetero-oligomer assembly requirements, and instability upon lumen modification. The authors propose PNRSS (programmable nano-reactors for stochastic sensing) as a generalizable approach that relocates reactive site engineering from the protein pore to a programmable synthetic polymer strand (e.g., DNA), positioned precisely within a structurally advantageous nanopore (MspA). This aims to democratize single-molecule chemistry measurements with higher resolution, flexibility in reactive site placement/number, and compatibility with AI-driven event classification.

The paper reviews single-molecule interrogation methods and prior nanopore engineering: engineered α-hemolysin with fixed reactive sites enabled ion binding detection but yielded low amplitudes (~1–5 pA) and complex engineering due to oligomeric symmetry and distributed reactivity. Semi-synthetic α-HL via native chemical ligation and adaptor-based approaches (cyclodextrins, protein adaptors) expanded capabilities but required complex chemistry or suitable adaptors and often limited resolution. MspA’s conical geometry has shown superior event amplitudes (~55 pA) in recent chemistry applications. DNA is highlighted as an ideal programmable polymer due to ease of synthesis and modification. The authors position PNRSS to overcome earlier constraints by precisely positioning designed reactive chemistries on a tethered strand within a nanopore with high field focusing.

Concept and components: PNRSS uses a synthetic polymer strand (primarily DNA) composed of modules: tether site (for streptavidin binding), extension section (length tuned with ~3.5 Å precision to place the reactive section in the pore constriction), reaction section (fixed reactants: e.g., nucleobases, azide-click conjugates like phenylboronic acid), and traction section (to maintain electrophoretic force). The strand is docked and stretched in a nanopore (mainly M2 MspA octamer; α-HL also tested), forming a fixed-reactant nano-reactor. Mobile reactants are introduced in solution.

Nanopore setup: A DPhPC bilayer separates cis/trans chambers (500 μL each) with 1.5 M KCl, 10 mM HEPES (pH 7.0 or 8.0). Ag/AgCl electrodes connected to an Axopatch 200B amplifier; sampling at 25 kHz, analog low-pass 1 kHz. Positive potentials of +160 or +180 mV typically applied. Single MspA pores inserted from cis; buffer exchange to prevent multiple insertions.

Strand capture and current levels: A biotin-TEG-modified DNA strand is precomplexed with streptavidin and added to cis (10–20 nM). Upon capture, a static residual current level (Is) is established. PNRSS events are further transient blockades (Ib) atop the Is level due to mobile reactant binding to the fixed reactant. Event parameters: dwell time τoff, inter-event interval τon, and amplitude ΔI = I1 − I0.

Chemistries implemented:

• Metal coordination: A DNA strand with adjacent guanines (13G/14G) flanked by abasic sites coordinates Ni2+ (mobile reactant added to trans, 0–1 mM). DFT calculations (Gaussian16, M06/6-31+G(d), LANL2DZ for Ni) explored coordination geometries and spin states.
• Click-functionalization: An alkyne DNA template (14TAK) was modified by CuAAC with functional azides (e.g., 4-(azidomethyl)benzene boronic acid) to generate 14PBA (PBA installed at site 14). Products validated by MS and single-channel behavior.
• Polyol binding to PBA: Mobile reactants included catechol, ethylene glycol, glycerol, L-lactic acid, vitamin C, vitamin B6, resorcinol (as negative control). Tris buffer at pH 7–8 used to probe protonation-state-dependent intermediates with PBA.
• Irreversible oxidation: H2O2 used as mobile reactant to convert PBA to phenol. After loss of boronic functionality (no further polyol/H2O2 events), a voltage protocol ejected and reloaded a fresh strand to repeat cycles.

Event analysis and kinetics: Events extracted via Clampfit 10.7; further analysis in Origin and ggplot2 for scatter plots. Kinetics assumed simple schemes: 1/τon = kon[analyte], 1/τoff = koff for reversible bimolecular association/unimolecular dissociation. Voltage (80–160 mV), salt concentration, and temperature dependencies characterized; temperature-controlled experiments on an Orbit Mini reader provided Arrhenius-like behavior.

Machine learning-assisted classification: Event traces split by Butterworth filtering (100 Hz cutoff) into low-pass and high-pass components. Standard deviations of both components during bound states formed 2D features. A support vector classification (DarwinML 2.0) was trained on 1455 labeled events (norepinephrine, epinephrine, isoprenaline) to classify analytes with high accuracy. The approach was also applied to distinguish remdesivir vs its triphosphate metabolite.

Controls and generality: A nonreactive strand (14X) showed no Ni2+ events; resorcinol showed no PBA binding; MspA vs α-HL comparison demonstrated generality with differing amplitudes. A strand lacking the traction section failed to maintain trapping, underscoring module necessity.

• High-amplitude single-molecule events: Ni2+ binding to dual guanine on a DNA strand within MspA produced a single event population with ΔI ≈ −60 pA, much larger than typical α-HL ion events (~2 pA). Kinetics followed: 1/τon ∝ [Ni2+] and τoff independent of [Ni2+], consistent with bimolecular association and unimolecular dissociation.
• Broad PBA polyol sensing: A PBA-functionalized strand (14PBA) detected multiple polyols (catechol, ethylene glycol, glycerol, L-lactic acid, vitamin C, vitamin B6) with distinct kinetic signatures; resorcinol (no 1,2/1,3-diol) showed no events. Many PBA-bound events were positive-going, attributed to increased conductance from anionic boronate esters.
• Observation of intermediates: With tris as mobile reactant, two bound levels appeared (protonated vs deprotonated forms) at pH 8.0, directly revealing transient intermediates and charge-dependent conductance effects.
• Irreversible reactions monitored repetitively: H2O2 caused reversible spiky events then irreversible oxidation of PBA to phenol (loss of PBA reactivity). Voltage ejection/reloading enabled multiple cycles of observing the irreversible chemistry within one experiment.
• Catecholamine discrimination with AI: Norepinephrine, epinephrine, isoprenaline all bound PBA but produced distinct event amplitudes/noise. After frequency splitting, SVC classification achieved accuracies: isoprenaline 99.9%, norepinephrine 98.6%, epinephrine 95.5%. Event amplitudes were ~21–32 pA in MspA, significantly larger than prior α-HL work (~1 pA), enabling clear separation.
• Antiviral drug distinction: Remdesivir vs remdesivir triphosphate both yielded positive-going PBA-binding events but were clearly separated by dwell time and bound-state noise; 2D SD features (high-/low-pass) showed two distinct populations, enabling direct discrimination.
• Generality across pores: PNRSS also worked in α-HL, though with smaller event amplitude (−4.6 pA for isoprenaline) than MspA (−32.2 pA), highlighting the advantage of MspA’s conical geometry and field focusing.
• Operating conditions: Higher voltage and salt increased event amplitude; dwell times largely voltage-independent. Charged analytes (norepinephrine) showed higher event rates at higher potentials; neutral analytes (catechol) less affected, indicating minor electroosmotic contribution. Temperature modulated on/off rates and affinity following Arrhenius behavior.
• Detection limits and sample compatibility: Defined LOD as ≥5 events in 10 min. Catecholamines had ~1 µM LODs (physiological levels are nM), whereas vitamin B6 had ~400 nM LOD and was detectable in spiked human urine without matrix interference, enabling quantification via calibration.

PNRSS relocates reactive-site engineering from the nanopore protein to a programmable tethered polymer, overcoming challenges of hetero-oligomer assembly, unintended reactivity, and low amplitudes. By placing designed chemistries (e.g., dual guanine ligands, phenylboronic acid) precisely in the MspA constriction, PNRSS achieves large, well-resolved signals that support kinetic modeling and recognition of subtle structural differences among small molecules. The method captures transient intermediates (e.g., protonation states of tris) and uniquely allows repetitive monitoring of irreversible chemistries by ejecting/reloading strands. The combination of high-resolution electrophysiology, frequency-domain feature extraction, and machine learning greatly enhances analyte discrimination, demonstrated for catecholamines and nucleoside analogues (remdesivir vs its active triphosphate). The approach shows generality with different pores (MspA, α-HL) and tunability across voltage, salt, and temperature, indicating broad applicability for single-molecule chemistry sensing, pharmacokinetics, and drug screening.

The study introduces PNRSS as a flexible, high-resolution platform for nanopore-based single-molecule chemistry. By programming reactive groups onto a tethered polymer positioned in a conical nanopore (MspA), the authors report 20 distinct reactions across ions, oxidants, polyols, vitamins, catecholamines, and nucleoside analogues, often revealing single-molecule kinetics and intermediates with large signal amplitudes. Repetitive observation of irreversible reactions and AI-assisted event classification expand the utility for chemical identification. Future directions include: integrating multiple fixed reactants on a single strand for multiplexed sensing; covalently attaching strands to pores for improved stability and resolution; adapting other nanopores (CsgG, aerolysin, ClyA, FraC, PlyA/B, phi29) for larger or more complex analytes; optimizing parallelization and flow-cell design to improve detection limits and throughput; exploring additional therapeutic nucleoside analogues for pharmacokinetic monitoring.

• Detection sensitivity: With single-pore measurements in relatively large chambers, LODs for some targets (e.g., catecholamines ~1 µM) are above physiological concentrations (nM). Parallel nanopore arrays and optimized flow cells are needed to boost efficiency.
• Strand–pore configuration: Demonstrations used non-covalently docked strands; permanent conjugation could further improve stability and resolution but was not implemented.
• Chemical side reactions: CuAAC triazole linkers can chelate transient metal ions, potentially interfering with measurements; mitigations include chelators (e.g., EDTA) or alternative bioorthogonal chemistries.
• Preparative limitations: PNRSS is a sensing approach with a nano-reactor scale not suited for synthesis/preparative chemistry.
• Generalizability metrics: While generality was shown with α-HL, event amplitudes are pore-dependent; broader validation across diverse pores/reactants remains for future work.