Giant single molecule chemistry events observed from a tetrachloroaurate(III) embedded Mycobacterium smegmatis porin A nanopore

Biological nanopores are capable of resolving small analytes down to a monoatomic ion. In this research, tetrachloroaurate(III), a polyatomic ion, is discovered to bind to the methionine residue (M113) of a wild-type α-hemolysin by reversible Au(III)-thioether coordination. However, the cylindrical pore geometry of α-hemolysin generates shallow ionic binding events (~5–6 pA) and may have introduced other undesired interactions. Inspired by nanopore sequencing, a Mycobacterium smegmatis porin A (MspA) nanopore, which possesses a conical pore geometry, is mutated to bind tetrachloroaurate(III). Subsequently, further amplified blockage events (up to ~55 pA) are observed, which report the largest single ion binding event from a nanopore measurement. By taking the embedded Au(III) as an atomic bridge, the MspA nanopore is enabled to discriminate between different biothiols from single molecule readouts. These phenomena suggest that MspA is advantageous for single molecule chemistry investigations and has applications as a hybrid biological nanopore with atomic adaptors.

A biological nanopore, which is the core component of a commercial sequencer 1 , is capable of probing the length 2 , sequence 3,4 , and base modifications 5 of DNA and many other biomacromolecules including RNA 6 , Xeno Nucleic Acids 7 , peptides 8 , and proteins 9 . This remarkable sensing performance originates from its biological role as a channel protein 10 . When proper interactions are established, a biological nanopore is able to resolve a monatomic ion 11 , indicating a precision far greater than that of a solid state nanopore. Pioneered since 1997 by Bayley et al. 11 , nanopore-based direct sensing of single ions such as Co 2+ , Ag + or Cd 2+ can be performed by designed ion-amino acid coordination [11][12][13] or by an ion-chelator interaction 14 within engineered α-hemolysin (α-HL) mutants. However, α-HL blockages by single monatomic ions suffer from an extremely low event amplitude (~2-3 pA), due to the cylindrical pore geometry, the small size of the ionic analyte and possible gating behaviors when monitored at a high voltage [11][12][13] . Alternatively, indirect sensing of metal ions can be performed with molecular adapters such as DNA 15 , peptides 16 or cyclodextrins 17 but with a diminished signal specificity and an increased system complexity.
Chloroauric acid (HAuCl 4 ), a well-known Au(III) compound 18 , has been widely used as a precursor for the fabrication of gold nanomaterials 19 . When dissolved in an aqueous solution, it ionizes, producing tetrachloroaurate(III) ([AuCl 4 ] − ), which is a square planar, polyatomic ion with a net charge of −1, in which the Au-Cl bond length is 2.28 Å 20 . Previous reports indicate that tetrachloroaurate(III) is a potent aquaporin inhibitor 20 but investigations at the level of a single molecule have not been reported.
With single molecule evidences, we have found that tetrachloroaurate(III) is an inhibitor of wild-type (WT) α-HL. This results from a coordination interaction established between a Au (III) atom and the thioether residue within the pore restriction. This coordination mechanism was adapted to the conical shaped Mycobacterium smegmatis porin A (MspA) nanopore 21,22 . A further amplified event amplitude, up to~55 pA, was monitored.
To the best of our knowledge, single molecule study of Au(III)thioether coordination chemistry has never been reported, and it brings insights in an aspect of bioinorganic chemistry, such as the design of Au(III) based drugs, which target proteins. The event amplitude, as generated from tetrachloroaurate(III) binding with MspA, is also the largest that has been reported from an inorganic ion when sensed by a nanopore. This suggests that MspA may be a superior template engineered as a nanoreactor to probe chemistry intermediates or kinetics in single molecule. The bound tetrachloroaurate(III) remains in the pore, forming a transient Au (III) embedment as a functional interface for sensing. As a proof of concept, the Au(III) embedded MspA nanopore discriminates between L-cysteine (Cys), L-homocysteine (Hcy) and Lglutathione (GSH) from direct single molecule readouts, which is a great challenge for fluorescence probe based imaging. It thus suggests amino acid or peptide sensing strategies with gold embedded protein nanopores or other embedments as a variety of metalloporins.

Results
Single tetrachloroaurate(III) binding within a WT α-HL. The heptameric WT α-HL is a mushroom-shaped channel protein with a narrow cylindrical stem and an aperture~1.4 nm in diameter at its narrowest spot 23 . Due to the limited acquisition bandwidth (100 kHz) of the patch clamp amplifier (Axon 200B, Molecular Devices), translocations of single inorganic ions through nanopores are not resolvable unless an interaction between the ion and the pore is established. Based on the known sulfur-gold (S-Au) coordination chemistry 24 , methionine (M113) [25][26][27][28] , which is in the proximity of the 1st restriction site of the pore 29 and is the only sulfur-containing amino acid within the inner surface of an α-HL monomer, could form a reversible interaction with freely translocating tetrachloroaurate (III) ions crossing the membrane.
All electrophysiology measurements were performed with a patch clamp amplifier (Axon 200B, Molecular Devices) in an aqueous buffer consisting of 1.5 M KCl and 10 mM Tris-HCl at pH 7.0. All measurements were performed with +100 mV, continuously applied, unless otherwise stated (Methods). Chloroauric acid was added in the cis chamber to reach the desired final concentration. With a single WT α-HL inserted in the membrane, the anionic [AuCl 4 ] − is driven electrophoretically through the pore. Addition of chloroauric acid to the cis compartment with a final concentration of 5 µM results in a reversible current blockage, measuring 5.4 ± 0.7 pA with an average lifetime of 11 ± 1 s (n = 3, Supplementary Table 1, Supplementary Figs. 1 and 2). To verify this sensing mechanism, a Met → Gly mutation (M113G) was introduced by pore engineering ( Supplementary  Fig. 3). No [AuCl 4 ] − ion binding events observed with WT α-HL were detected by α-HL M113G, even when the cumulative chloroauric acid concentration in cis reaches 50 µM (Supplementary Fig. 4). The absence of ion binding events in the M113G mutant is thus evidence for a coordination interaction between Au(III) and the methionine (M113) in the WT α-HL.
Au(III) coordination with amino acids or peptides has been intensively investigated by UV-Vis spectroscopy, NMR spectroscopy and Fourier transform IR spectroscopy. However, these methods are limited by a lack of dynamic information, the requirement of an acidic, low-chlorine environment, a highconsumption of reactants and a lack of single molecule resolution [25][26][27][28] . With nanopores, however, the Au(III)-thioether coordination chemistry was directly monitored from single molecule readouts and with negligible requirements for the measurement environment or the quantity and the purity of the analyte. However, binding of tetrachloroaurate(III) in α-HL shows fluctuations in the trace and these lead to a wide dispersion in the statistics of the event amplitude ( Supplementary Fig. 1). These fluctuations appear as low frequency transitions between different secondary states and may result from non-specific binding of tetrachloroaurate(III) with other amino acids such as lysine (K147 or K131) distributed over its long cylindrical restriction 13 . To exclude non-specific interactions, a combination of multiple mutagenesis would be necessary, and this may result in a high risk of failed pore assembly, complicating further optimizations. However, this mechanism may be adapted to other channel proteins for an enhanced sensing performance and ease of engineering.
Methionine tetrachloroaurate(III) coordination in MspA. In nanopore sequencing, a nanopore with a single, geometrically sharp restriction, as in MspA 21 or CsgG 30 , is advantageous because it has a higher spatial resolution 3 . Though it hasn't been reported previously, direct single ion sensing could also be performed with a geometrically sharp nanopore to acquire an enlarged signal amplitude and to avoid non-specific binding with residues distant from the recognition site. The mutant M2 MspA 21 (D93N/D91N/D90N/D118R/D134R/E139K), which was the first reported nanopore for DNA sequencing, is a funnel shaped, octameric channel protein which measures~1.2 nm in diameter at its narrowest restriction (Methods) 31 . Reported mutations, as in the M2 MspA, are designed to neutralize the original negative charges of WT MspA (PDB ID: 1uun 32 ) for an enhanced capture rate for anions, such as DNA 21 or possibly tetrachloroaurate(III). Based on a visual analysis of the corresponding protein structure, no methionine or cysteine exists within the inner surface of M2 MspA, making it a clean template to which a methionine can be introduced.
Experimentally, as in α-HL M113G ( Supplementary Fig. 4), no [AuCl 4 ] − binding events were detected by M2 MspA even when the cumulative chloroauric acid concentration in cis was raised to 50 µM ( Supplementary Fig. 5). By a single site directed mutagenesis in M2 MspA, a methionine was introduced at residue 91 (Methods), which at 1.2 nm in diameter 22 is the narrowest spot of MspA (Fig. 1a). This MspA mutant namely MspA-M (D93N/D91M/D90N/D118R/D134R/E139K), was prepared in the same way as its predecessor (M2 MspA) (Methods) and showed similar channel properties during its characterization ( Supplementary Fig. 6), indicating an octameric pore assembly, which was unaltered by the mutation.
During a continuous electrophysiology recording as described The event blockage amplitudes, as derived from a representative trace from a 10 min continuous recording, show fully resolved peaks, each with a Gaussian distribution (Fig. 1d) corresponding to different I n blockages. Non-specific events such as statistical counts not subject to the Gaussian distribution, are apparently never observed (Fig. 1d). Presumably, signal contributions from possible non-specific interactions distant from the restriction were weakened as negligible off-focus contributions, and this suggests a significant advantage of using a geometrically sharp channel to probe a variety of single ions or small molecules as a nanoreactor, where pore engineering is simplified to a much reduced amount of amino acids compared with that of α-HL 13 . The reduced complexity of pore functionalization by mutagenesis of MspA thus has a great advantage in the ease of pore engineering. Though tetrachloroaurate(III) events were detectable with a 200 nM HAuCl 4 concentration in the cis, a sharply increased detection frequency was observed (Fig. 1e) when the accumulated HAuCl 4 concentration in cis reached 1 µM. However, the 1/τ off value for I 1 events remains constant with all HAuCl 4 concentrations, indicating that the same type of binding was observed with the various analyte concentrations. When the HAuCl 4 concentration was further increased to 10 µM, I n>1 events become dominant (Fig. 1c), from which the statistics of τ on (time of interevent duration, Supplementary Fig. 8) and the corresponding onrate can be deduced ( Supplementary Fig. 9, Supplementary  Table 3). However, to avoid complications from multi-level binding events, results in Fig

Tetrachloroaurate(III) binding in different confined spaces.
Single tetrachloroaurate(III) binding has so far been demonstrated with two types of channel proteins possessing similar outer dimensions but different geometries (Fig. 2a). However, [AuCl 4 ] − binding within different pores may generate varying single ion behaviors when restricted differently by for example, geometry, charge or electric field in the vicinity of the restriction. These phenomena suggest how further optimization of single ion sensors may be performed, or could inspire the design of goldcontaining compounds as drug molecules which target channel proteins with a high specificity and precision.
Though the experiments were performed identically, [AuCl 4 ] − blockage events in WT α-HL appear as shallow (ΔI = 5.6 ± 0.3 pA) resistive pulses but with a mean dwell time of~9.38 s.
[AuCl 4 ] − binding in MspA-M on the other hand, shows deeper blockage (ΔI = 11.3 ± 0.2 pA) but the event is shorter with a mean dwell time of~0.45 s ( Fig. 2b-d). The dwell time (t off ) follows a single exponential fitting and was derived as described in Supplementary Fig. 8. By analyzing the ΔI of I 1 events from WT α-HL or MspA-M, these differences were systematically observed in independent measurements by different researchers (Supplementary Tables 1 and 2). Though a larger absolute value of ΔI was observed from MspA-M, the percentages of blockage (ΔI/I 0 ) from the two pores are quite similar. We concluded that this is an effect of the pore geometry. The narrowest regions of each pore, where tetrachloroaurate(III) binds, are similar in size ( Fig. 2a) and determine the ΔI/I 0 . However, the wider opening of MspA, which leads to a higher open pore conductance, results in the larger ΔI in the absolute amplitude. The larger ΔI amplitude along with a considerably narrowed dispersion (Fig. 2c) shows that MspA is superior to α-HL for single ion sensing. The conical geometry results simultaneously in a faster voltage drop and a stronger electric field (E z = −dV/dz). The enhanced electrophoretic force and electro-osmotic flow should contribute to the shorter duration time of tetrachloroaurate(III) binding in MspA-M (Fig. 2d).
The local charge distribution within the inner surface of the pore is critical for analyte attraction. It was found that MspA-M captures tetrachloroaurate(III) more efficiently than WT α-HL, where a 200 nM detection limit was observed with MspA-M, which is 5 times lower than that of α-HL. This may result from the positive charges introduced around the larger vestibule (D118R/D134R/E139K) of MspA-M, which was originally designed to attract ssDNA 21 . Similar phenomena have been observed with other biological nanopores, when excessive positive charges in the pore lead to a more efficient DNA capture rate [33][34][35] .
By measuring the voltage dependence for I 1 binding events, acquired with either WT α-HL ( Supplementary Fig. 11a-f) or MspA-M ( Fig. 2e-g, Supplementary Fig. 11g-l), a maximum ΔI =~55 pA was acquired with MspA-M when a +200 mV voltage was applied ( Supplementary Fig. 11l), far beyond any pore blockage signals from a single ion that have previously been reported (Fig. 2h). For [AuCl 4 ] − in particular, binding at +200 mV, the MspA-M signal outweighs that from WT α-HL bỹ 31 pA (Fig. 2h, Supplementary Fig. 11f, l, Supplementary Table 4). On the contrary, [AuCl 4 ] − binding in WT α-HL generates significant baseline fluctuations, which are clearly noticeable for recordings performed with an applied voltage in excess of +100 mV (Supplementary Figs. 11 and 12) and has limited its uses at a high applied voltage for a better amplitude resolution. This phenomenon should result from undesired bindings of [AuCl 4 ] − with other amino acids within the cylindrical stem of α-HL. In the low voltage regime, though barely detectable (0.90 ± 0.08 pA), [AuCl 4 ] − binding within MspA-M was still visible at an applied voltage of +20 mV (Fig. 2e). However, a minimum of +60 mV was needed for WT α-HL to resolve the events ( Supplementary Fig. 11a, b). In summary, by using [AuCl 4 ] − as a model analyte, MspA-M clearly outperforms WT α-HL in many aspects such as a larger event amplitude, a narrower event dispersion, a higher sensing specificity and a lower detection limit. This suggests that MspA is an ideal nano-cavity with which to probe a variety of single molecule chemistry kinetics.
Direct sensing of L-cysteine by Au(III) embedded MspA. When bound to a methionine, the Au(III) atom remains in the proximity of the restriction of MspA for~0.5 s, forming a transient Au (III) embedment as an adaptor for sensing. Besides the demonstrated Au(III)-thioether interaction, a stronger interaction between Au(III)-thiol is expected, as previously reported 25 , which indicates that an Au(III) embedded MspA may sense a variety of thiol-containing molecules. The most abundant biothiols include L-cysteine (Cys), L-homocysteine (Hcy) and L-glutathione (GSH), which are directly involved in crucial physiological processes 36-38 such as protein synthesis 39 , free radical scavenging 36 and normal immune system maintenance 40 . Though presented in the blood plasma with a high abundance, in the~µM range 41 , the structure similarity of these biothiols presents a great challenge for a direct simultaneous discrimination. With distinct physiological roles, discriminative sensing of these biothiols could have great significance in biomedical diagnostics.
Conventionally, sensing of biothiols was performed with high performance liquid chromatography-mass spectroscopy 42 or designed fluorescence probes 43 , but suffers from a time consuming and laborious sample preparation process or the challenge of probe design. Nanopore sensing, which is inexpensive, fast and sensitive, may provide an alternative solution for direct sensing of biothiols. However, a biological nanopore, such as an octameric MspA-M, doesn't directly report signals for all biothiols in general ( Supplementary Fig. 13) without the establishment of an interaction between the analyte and the pore.
On the other hand, the demonstrated Au(III) embedment enables MspA to interact with biothiols via the Au(III)-thiol coordination chemistry. The Au(III)-thiol coordination, which forms a much stronger bond than the established Au(III)thioether coordination, competes with the existing Au(III)thioether bond and consequently speeds up the dissociation of the Au(III) from the pore. Though the described chemical process happens rapidly, it can be monitored by a nanopore sensor, which forms the basis for sensing.
As a proof of concept, nanopore-based biothiol sensing was carried out with MspA-M as described in Methods. Specifically, HAuCl 4 was added to the cis while the biothiols were added to the trans compartment. The two analytes were added to different sides of a nanopore to minimize the spontaneous redox reactions between Au(III) and biothiols before entering the pore restriction (Fig. 3a). With this configuration, the anionic [AuCl 4 ] − was first electrophoretically driven into the pore where it binds to the methionine at the pore restriction. Subsequently, the bound Au(III), which acts as an atomic bridge, captures freely translocating biothiol molecules and is stimulated to dissociate from the thioether group on the pore. Subsequently, another sensing cycle is initiated whenever the next Au(III) binds (Fig. 3b).
L-cysteine (Cys), which is an essential amino acid involved in protein synthesis 39 , is the most well-known biothiol (Fig. 3c). As a test of feasibility, nanopore-based biothiol sensing was performed (Methods) by adding chloroauric acid to the cis and Cys to the trans with 4 µM and 40 µM in concentration, respectively. From electrophysiological recordings, MspA-M reported a different type, 2-step shaped blockage event (Fig. 3d, Supplementary  Video 3), which could be clearly distinguished from binding events when HAuCl 4 was the sole analyte added (Fig. 2). A representative event of such type is composed of three states namely 0, 1, and 1 SH , which corresponds to the open pore state, the [AuCl 4 ] − bound state and the Cys bound state, as described in the molecular model from Fig. 3b. State 1, which represents the [AuCl 4 ] − binding state (Fig. 1b), appear as a flat, less fluctuating and long residing plateau. On the other hand, state 1 SH can be recognized from its characteristic jitter signals (Fig. 3e), which have never been observed from [AuCl 4 ] − binding events. According to reported literatures, the interaction between the methionine residue on the pore and a [AuCl 4 ] − is a highly reversible coordination interaction, with no redox reaction observed. This is also confirmed by the time extended measurements ( Supplementary Fig. 10). However, the characteristic jitter signal might represent a redox reaction between Cys and the bound Au(III), which contains rich chemical intermediates. Nonetheless, the interaction between Au(III) and a thiol group could be indirectly monitored from the significant reduction in the dwell time of state 1 when Cys was added in trans ( Supplementary  Fig. 14). The freshly formed Au(III)-thiol interaction has stimulated the dissociation of Au(III) from the nanopore. With their characteristic event shape, Cys sensing events can be immediately distinguished from other non-specific binding types. To ensure the readability, only Cys sensing events were counted in the statistics which were based on a simple algorithm that an event has to contain all three states as demonstrated in Fig. 3e. Other non-specific event types, including binding and dissociation of one [AuCl 4 ] − without any captured biothiol, sequential binding of two [AuCl 4 ] − ions and intrinsic noises from MspA-M ( Supplementary Fig. 15) were ignored. Nevertheless, the Cys sensing event amounts to 96% of all detectable events from continuously recorded results ( Supplementary  Fig. 16). A scatter plot of the relative blockage amplitude ΔI/I 0 versus the dwell time from the events of [AuCl 4 ] − and Cys sensing is shown in Fig. 3f. Though with a slightly wider dispersion than that of [AuCl 4 ] − (Fig. 3f), the Cys event forms a clear monodispersed distribution. This is clearly demonstrated in the histogram of ΔI/I 0 (Fig. 3g, Supplementary Tables 5 and 6) and the dwell time τ off (Fig. 3h), from which ΔI/I 0 of Cys can be seen to be 0.104 ± 0.008 (Supplementary Table 6, N = 364) and the mean dwell time of state 1 SH was derived as 3.18 ms (Fig. 3g-h).
Similar measurements were also performed with L-asparagine, L-glycine and L-glutamic acid (Supplementary Fig. 17) and no events such as that reported in Fig. 3e were observed. This indicates that this sensing configuration is highly specific to the thiol side chain of an amino acid.
Direct sensing of L-Homocysteine by Au(III) embedded MspA. L-Homocysteine (Hcy), which is a homologue of Cys, is an important intermediate in the metabolism of methionine and cysteine 44 . An elevated Hcy level in the blood serum indicates a high risk of cardiovascular diseases and is a critical parameter in diagnosis 45 . However, Hcy differs from Cys with just one additional methylene group (Fig. 4a), and discrimination between Hcy from Cys is a challenge.
Hcy sensing was performed as described in Fig. 3a (Methods) when 4 µM chloroauric acid was placed in cis and 40 µM Hcy in trans. Systematically deeper blockage events compared with that of Cys (~33 pA, marked by green squares) were observed from continuously recorded traces (Fig. 4b). A representative Hcy event is also composed of three states (Fig. 4c), similar to the behavior of Cys (Fig. 3e). These differ however in the 1 SH state. A scatter plot of ΔI/I 0 vs the dwell time from events of [AuCl 4 ] − and Hcy sensing was presented to show the event dispersion (Fig. 4d). From the corresponding histogram of ΔI/I 0 , the Hcy blockage events measure 0.13 ± 0.01 (N = 330) (Fig. 4e Fig. 18).
To demonstrate simultaneous discrimination between Cys and Hcy from direct single molecule readouts, a nanopore  (Fig. 4f). The scatter plot of ΔI/I 0 vs the dwell time with the corresponding amplitude histogram (Fig. 4g) clearly shows two distinct event types. From the corresponding Gaussian fitting results, the two peaks in the histogram of ΔI/I 0 are 0.105 ± 0.005 for Cys and 0.128 ± 0.013 for Hcy, respectively, which are in agreement with the separately measured values. The above demonstration suggests that a direct discrimination of Cys and Hcy, which differ by only one methylene group, is possible from direct nanopore readouts.
Direct sensing of L-Glutathione by Au(III) embedded MspA. L-Glutathione (GSH), which is a tripeptide (γ-Glu-Cys-Gly) (Fig. 5a), is critical in the maintenance of immune system. It is also the most abundant tripeptide thiol found in human serum 40 .
As demonstrated with Cys and Hcy, the internal thiol in a GSH makes it compatible with the described biothiol-sensing strategy. With 4 μM chloroauric acid in cis and 40 μM GSH in trans, characteristic biothiol-sensing events measuring~36 pA in amplitude were observed (Fig. 5b). From a representative event, the characteristic jitter signal of GSH (state 1 SH ) reports a significantly larger blockage amplitude (Fig. 5c) than that of Cys (Fig. 3e). According to the negative control performed with glutamic acid and glycine separately ( Supplementary Fig. 17), it was confirmed that the internal cysteine of GSH is critical to the generation of the event. The scatter plot of ΔI/I 0 vs the dwell time with the corresponding amplitude histogram for GSH and [AuCl 4 ] − are shown in Fig. 5d. The ΔI/I 0 of GSH measures 0.15 ± 0.02 (Supplementary Tables 9 and 10) with a mean dwell time of 2.82 ms in state 1 SH (Supplementary Fig. 18, Supplementary Table 9), indicating that it can be clearly distinguished from Cys via direct single molecule readouts.
Simultaneous discrimination of Cys and GSH was performed by adding a mixture of Cys and GSH in trans with 10 μM and  Supplementary Fig. 15. c A representative Hcy blockage event. Similar to Fig. 3E, a representative Hcy blockage event is composed of three states as described in Fig. 3b. Characteristic jitter signals can be recognized as state 1 SH . However, the mean blockage amplitude of state 1 SH for Hcy is systematically deeper than that of Cys. With a +100 mV applied voltage, it was found that Cys is much more likely than GSH to be captured by the Au(III) embedded nanopore so that the GSH concentration in the mixture was increased to balance the rate of appearance of both events. From the electrophysiology trace, two types of events were clearly identified according to the difference in their amplitudes (Fig. 5e). The scatter plot of dwell time vs ΔI/I 0 with the corresponding amplitude histogram for GSH and Cys are shown in Fig. 5f, from which two populations of events were clearly visible. This unambiguously confirmed that Cys and GSH can be clearly discriminated from direct nanopore readouts. The time reduction appears to be different and this may result from the differences in steric hindrance when Cys, Hcy, or GSH interact with the Au(III). From the single molecule results, among the three tested biothiols, Cys reacts most strongly with Au(III) ( Supplementary  Fig. 18). The histogram of ΔI/I 0 for Cys, Hcy, and GSH events with corresponding Gaussian fittings are shown in Fig. 6b and an order of I 1,GSH > I 1,Hcy > I 1,Cys can be clearly observed. This was expected because GSH is significantly larger than the other two amino acids and Hcy has an excess methylene when compared with Cys. From the histogram in Fig. 6b, it is clear that Cys, Hcy, and GSH can be clearly distinguished from their distinct ΔI/I 0 values. However, due mainly to the jitter signal of state 1 SH , the ΔI/I 0 value appears with a wide distribution and an inevitable signal overlap. However, statistical data of mean ΔI/I 0 histogram indicates that the distinction between Cys, Hcy and GSH can be achieved from independent measurements (Fig. 6c) Fig. 19), biothiols can also be quantitatively analyzed. All above results demonstrate that the Au(III) embedded MspA could serve as a highly sensitive probe for the differentiation of GSH, Hcy and Cys.

Discussion
Distinct from a recent report of cysteine and homocysteine discrimination using nanopores 46 , in which a time consuming sample preparation is needed and GSH was in principle not detectable, the described method in this paper suggests a strategy that is simple and straightforward using nanopores with atomic adapters. Further engineering with this approach may be done by permanently embedding metal ions using irreversible coordination adaptors 47 . Full discrimination of other amino acids may be achieved with designed atomic adaptors targeting different side groups of the amino acid analytes. The locations of these adaptors within a conically shaped biological nanopore may also be slightly dispersed so that the signal amplitude from different analyte may be tuned to assist full discrimination. Similar pore engineering may be performed by taking the monomeric channel protein OmpG 48-50 as a template for metal embedding. Other ion-amino acid combinations within a variety of biological nanopores such as Cytolysin A 51 , phi29 motor protein 52 or aerolysin 2,53 could also be adapted for different applications.
To the best of our knowledge, single molecule Au(III)methionine coordination chemistry has never been observed in a natural channel protein. It may inspire the design of goldcontaining compounds as drug molecules targeting channel proteins. By single site directed mutagenesis, this mechanism has been adapted to the MspA nanopore. As a consequence of geometric optimization, MspA has magnified the event amplitude of a single [AuCl 4 ] − binding. The observed 55 pA [AuCl 4 ] − event amplitude is the largest known record from a single inorganic ion when sensed by a nanopore. The sharp restriction of MspA along with the simplicity of mutagenesis suggests its role as a superior engineering template for a variety of single molecule chemistry investigations, complementary to its well-known uses in nanopore sequencing.
With its unique physical and chemical properties, gold has been used extensively in a wide range of scientific and industrial applications such as the production of nanoparticles 54 , tunneling electrodes 55 , surface enhanced raman spectroscopy probes 56 and surface plasmonic resonance substrates 57 . However, these technologies lack single molecule control precision comparable to that offered by a biological nanopore, in which the geometry 58 , orientation 59 , polarity 51 , and chemical modifications 11 of both the pore and the analyte can be manipulated and controlled. By taking the embedded Au(III) as an atomic bridge, MspA is enabled with biothiol-sensing capacities, which directly discriminate between L-cysteine, L-homocysteine, and L-glutathione from single molecule readouts. Though demonstrated as a proof of principle, this sensing mechanism is simple, label free, fast and economic and may be engineered into a portable sensor chip. Eventually, the demonstrated result of Au(III) embedment may benefit a wide range of other fundamental scientific research projects in need of a single molecule precision and the properties of gold 60,61 or even other metal elements if properly designed.
Electrophysiology recording and data analysis. All electrophysiology results were acquired by an Axopatch 200B patch clamp amplifier and digitized by a Digidata 1550 A1 digitizer (Molecular Devices, UK). A custom made measurement chamber is separated by a Teflon film (30 µm thick) with an orifice (ø = 100 μm). Before use, the orifice was pretreated with 0.5% (v/v) hexadecane in pentane and then air-dried to evaporate the pentane. 1,2-Diphytanoyl-sn-glycero-3-phosphocholine (DphPC) was used to form a self-assembled lipid bilayer sealing the orifice. This lipid bilayer divides the chamber into cis and trans compartments both filled with 0.5 mL of 1.5 M KCl buffer (1.5 M KCl, 10 mM Tris-HCl, pH 7.0). A pair of Ag/AgCl electrodes were placed in cis and trans side of the chamber, in contact with the aqueous buffer respectively. Conventionally, the cis side is defined as the side which is electrically grounded. Biological nanopores (WT α-HL, α-HL M113G, M2 MspA, or MspA-M) were added to cis for spontaneous pore insertion. Unless otherwise stated ( Supplementary Figs. 11 and 12), all measurements were performed with a +100 mV continuously applied voltage. The acquired single channel data was sampled at 25 kHz and filtered with a corner frequency of 1 kHz. For [AuCl 4 ] − binding events, the recorded current traces were digitally filtered with a 200 Hz low-pass Bessel filter (eight-pole). For sensing of amino acids, the recorded current traces were digitally filtered with a 0.2 kHz low-pass Bessel filter (eight-pole). Event states were detected by the single channel search feature in Clampfit 10.7 and further analyses (histogram, curve fitting and plotting) were carried out in Origin 9.1 (Origin Lab).