Single‐Molecule Observation of Intermediates in Bioorthogonal 2‐Cyanobenzothiazole Chemistry

Abstract We report a single‐molecule mechanistic investigation into 2‐cyanobenzothiazole (CBT) chemistry within a protein nanoreactor. When simple thiols reacted reversibly with CBT, the thioimidate monoadduct was approximately 80‐fold longer‐lived than the tetrahedral bisadduct, with important implications for the design of molecular walkers. Irreversible condensation between CBT derivatives and N‐terminal cysteine residues has been established as a biocompatible reaction for site‐selective biomolecular labeling and imaging. During the reaction between CBT and aminothiols, we resolved two transient intermediates, the thioimidate and the cyclic precursor of the thiazoline product, and determined the rate constants associated with the stepwise condensation, thereby providing critical information for a variety of applications, including the covalent inhibition of protein targets and dynamic combinatorial chemistry.


General
Reagents were purchased from commercial sources (as indicated) and used as received. NMR spectra were recorded on a Bruker AVIII HD 400 spectrometer. Chemical shifts for 1 H are reported in ppm on the δ scale, and referenced to the residual solvent peak. Coupling constants (J) are reported in Hertz (Hz). The following abbreviations are used to describe signal multiplicity for 1 H spectra: s: singlet, d: doublet, t: triplet, m: multiplet, dd: doublet of doublets. Low and high resolution mass spectra were recorded on a Micromass Platform 1 spectrometer and a MicroTOF mass spectrometer using electrospray ionization.

Protein preparation
αHL monomers were prepared by in vitro transcription-translation (IVTT) as previously reported [1] .
Specifically, engineered αHL polypeptides were expressed by using a commercial IVTT kit: E. coli T7 S30 Extract System for Circular DNA (Promega). To suppress transcription by E. coli RNA polymerase, the T7 S30 extract provided in the kit was treated with rifampicin prior to use (1 µg mL -1 final concentration). A standard reaction comprised: DNA template (3.2 µg), amino acid mix minus methionine (supplied with the kit, 5 µL), S30 premix without amino acids (supplied with the kit, 20 µL), [ 35 S]methionine (2 µL, 1,200 Ci mmol -1 , 15 mCi mL -1 , MP Biomedicals), T7 S30 extract (supplied with the kit, 15 µL), and nuclease-free water to a final volume of 50 µL. To make heteroheptamers, plasmids encoding the WT αHL and the mutant αHL with a D8 tag were mixed in a ratio 6:1 (WT: mutant). The IVTT mixture was incubated at 37°C for 1 h.
Heptamerization was carried out by the addition of rabbit erythrocyte membranes (3 µL, ~1 mg protein mL -1 ) and incubation at 37°C for 1 h. The mixture was then centrifuged for 10 min at 25,000 × g. The supernatant was removed, and the pellet resuspended in MBSA buffer (200 µL, 10 mM 3morpholinopropane-1-sulfonic acid (MOPS), 150 mM NaCl, 1 mg mL -1 bovine serum albumin, pH 4 7.4). The wash with MBSA was repeated before the pellet was resuspended in 2X Laemmli sample buffer (50 µL) and electrophoresed in a 5% SDS/PAGE gel at 70 V for 15 h. The αHL heptamers containing different numbers of mutant subunits were separated in the gel based on their different electrophoretic mobilities, which were determined by the number of octa-aspartate (D8) tails. The top and bottom bands corresponded to WT7 and (mutant-D8)7, respectively. The second band from the top was the desired heteroheptamer containing a single engineered subunit. To extract heptameric pores, the gel was first dried without fixation on Whatman 3M filter paper under vacuum for 5 h at room temperature. After visualization by autoradiography with Kodak BioMax MR film, the desired bands were cut out from the gel with a scalpel. Each excised band was rehydrated in TE buffer (300 µL, 10 mM Tris· HCl, 1 mM ethylenediaminetetraacetic acid (EDTA), pH 8.0) for 1 h at room temperature. The filter paper was then removed, and the rehydrated gel was macerated with a pestle.
The resulting slurry was filtered through a 0.2 µm hydrophilic membrane filter (Proteus Mini Clarification Spin Column, Generon). The filtrate was stored in 10 µL aliquots at -80 °C.
Unless otherwise stated, the other chemicals were purchased from Sigma-Aldrich.
Planar bilayer recordings were performed following the method established by Montal and Mueller [2] .
Two Delrin chambers were separated by a 25 µm-thick Teflon film containing an aperture (60 µm in diameter), which was pre-treated with 1% (v/v) hexadecane in pentane. Each chamber was then filled Ionic currents were recorded by using a patch clamp amplifier (Axopatch 200B, Axon Instruments) with a 4-pole low-pass Bessel filter (80 dB/decade) at room temperature (20 °C ± 1 °C). Signals were digitized with a Digidata 1320A digitizer (Molecular Devices), connected to a computer running the pCLAMP 9.2 software suite (Molecular Devices). Unless stated otherwise, the signal was filtered with a corner frequency of 10 kHz and sampled at 50 kHz.
To study CBT chemistry, gel-purified αHL heptamers containing a single cysteine were added to the grounded cis chamber. To facilitate pore insertion, a potential of +200 mV was applied with constant stirring in the cis compartment. After a single pore had inserted, the potential was changed to -50 mV and the stirring stopped. To build a CBT nanoreactor, CBT-Mal was added to the cis side as a freshly by stirring for 30 s. For the reversible reactions between CBT and monothiols that generated two current blockade levels, individual events were identified by using a threshold based single-channel search in Clampfit to generate idealized traces. The mean dwell times (<τ>) of the CBT nanoreactor (i), the thioimidate (ii), and the tetrahedral adduct (iii) were determined using QuB software (QUB 2.0, Buffalo University). A maximum likelihood (MIL) algorithm was used to calculate the rate constants based on the idealized trace using a three-state model [3] . The bimolecular reactions between CBT and monothiols, and between thioimidates and monothiols were first order in monothiol concentration (k = 1/(<τ>[monothiols]), M -1 s -1 ), and the rate constants were extracted from the slopes of the plots (1/<τ> vs [monothiols]) determined by linear regression analysis. The dissociation of the 6 monothiols from either the tetrahedral adducts or the thioimidates was independent of the concentrations of monothiols, and the rate constants were obtained by averaging the dissociation rates (k = 1/<τ>, s -1 ) obtained at different concentrations of monothiols.
For the irreversible condensations between CBT and aminothiols, current traces were idealized and imported into QuB as described above. A four-state model was used to derive the mean dwell times (<τ>) of the CBT nanoreactor (i), the thioimidate (ii), and the tetrahedral intermediate (iii) using MIL algorithm. The first step was bimolecular and hence followed k = 1/<τ>[aminothiol], whereas the later two steps were unimolecular and followed k = 1/<τ>. Once the thiazoline or dihydrothiazine was formed, a new experiment with a separate nanopore was started.
Transient spikes of varying amplitudes in the baseline of single-channel recordings were observed with the wild-type pore in the absence or presence of CBT-Mal (10 µM), with the cysteine mutant nanopore (e.g. 117C) before the reaction with CBT-Mal, and more frequently after the CBT nanoreactor was formed. Perfusion of the cis compartment three times after the functionalization to remove excess CBT-Mal did not alter the spiky character of the current. The transient spikes were therefore attributed to intrinsic fluctuations within the nanopore structure, which were more pronounced when the interior wall was modified. The spikes were neglected during analysis.
The mixture was shaken at room temperature for 2 h and then centrifuged.

Products of condensations between aminothiols and CBT nanoreactors
Rapid interconversions between two discrete current levels were observed as the final product signature for all reactions between aminothiols and CBT nanoreactors. Different interconversion frequencies were recorded with different individual CBT nanoreactors upon reaction with the same aminothiol (e.g. Cys, Figure S3). With a given nanoreactor, the interconversion rate was highly voltage-dependent for charged molecules (e.g. the thiazoline formed with Cys or the dihydrothiazine formed with hCys) and less so with neutral species (e.g. the thiazoline formed with CysOMe) ( Figure   S4). We moved the tethering site of the reactive CBT group along the length of the β barrel to examine the change in product current pattern. At sites 117 and 121, interconversions between two states were seen after the reaction with Cys, whereas at position 123, interconversions were recorded between 4 states ( Figure S5). We speculated that the appendant product molecule was able to interact with various sites on the interior wall.
In additions to the interactions with the inward-facing amino acid side chains, the tethered molecules might undergo internal conformational changes that contribute to the generation of current signatures of the products. For example, there might be cis-trans isomerization in the urea group through rotation about the C−N bond, for which the activation barrier has been reported to be 11.3 kcal mol -1 in DMF/DMSO solutions [4] (>15000 s -1 at 20 °C ). There might also be cis-trans isomerization in the luciferin structure through rotation about the C-C bond connecting the thiazoline and the benzothiazole rings. While experimental measurements of the rotation rates are lacking, computational studies predicted energy barriers for rotations from the cis or trans conformations to be 1.7 and 7.2 kcal mol -1 , respectively [5] .

Direct attachment of the product molecule to a cysteine nanopore
To confirm that we formed the thiazoline product with CysOMe within the nanoreactor, we synthetized the corresponding amino-L-luciferin methyl ester (LucOMe) and derivatized it with pmaleimidophenyl isocyanate. The resultant LucOMe-Mal was used to modify the cysteine nanopore in situ to generate the proposed product directly ( Figure S6). Given that the isocyanate was reactive towards both amine and carboxylate groups, derivatization of amino-L-luciferin-the product with Cys-with p-maleimidophenyl isocyanate was not straightforward and therefore not completed.
We added LucOMe-Mal to the trans compartment for direct attachment because the reaction with LucOMe-Mal was retarded when the molecule was added to the cis side, although it was less of a problem with CBT-Mal. It is therefore worth highlighting the dimensions of the product molecules ( Figure S7). In the most extended form, LucOMe-Mal can elongate to ~2 nm ( Figure S7), which is greater than the constriction of the β barrel (~1.4 nm).