Tuning Electrostatic Gating of Semiconducting Carbon Nanotubes by Controlling Protein Orientation in Biosensing Devices

Abstract The ability to detect proteins through gating conductance by their unique surface electrostatic signature holds great potential for improving biosensing sensitivity and precision. Two challenges are: (1) defining the electrostatic surface of the incoming ligand protein presented to the conductive surface; (2) bridging the Debye gap to generate a measurable response. Herein, we report the construction of nanoscale protein‐based sensing devices designed to present proteins in defined orientations; this allowed us to control the local electrostatic surface presented within the Debye length, and thus modulate the conductance gating effect upon binding incoming protein targets. Using a β‐lactamase binding protein (BLIP2) as the capture protein attached to carbon nanotube field effect transistors in different defined orientations. Device conductance had influence on binding TEM‐1, an important β‐lactamase involved in antimicrobial resistance (AMR). Conductance increased or decreased depending on TEM‐1 presenting either negative or positive local charge patches, demonstrating that local electrostatic properties, as opposed to protein net charge, act as the key driving force for electrostatic gating. This, in turn can, improve our ability to tune the gating of electrical biosensors toward optimized detection, including for AMR as outlined herein.

Scheme SI-1. Engineering of proteins with the noncanonical amino acid (ncAA) azF (p-azido-L-phenylalanine) at defined sites/residues via a reprogrammed genetic code approach Site directed mutagenesis and incorporation was performed as outlined in Scheme S1 above. Details for BLIP 41AzF have been presented elsewhere 2 and are outlined here as part of the general approach for generating the BLIP-AzF mutants.
The BLIP2-azF mutants were constructed using the primers sets given below. PCR was performed with NEB Q5 polymerase according to manufacturer's directions. PCR products were then purified using the QIAgen PCR purification kit. Purified PCR products were recircularised with T4 PNK and Quick ligase (both NEB). DNA sequences were verified by sequencing provided by Eurofins Genomics and are presented below.

G49TAG mutation
CCTGGGGCtagAACAATGACTG CCACCACCGAGGTTGCCATA ATGGCAACCTAGGTGGTGGCCTGGGGCGGTAACAATGACTGGGGTGAAGCTACCGTGCCGGCC  GAAGCGCAGAGTGGTGTGGATGCAATTGCAGGTGGTTATTTTCATGGGCTGGCACTGAAAGGGG  GTAAAGTACTGGGCTGGGGTGCAAATCTGAACGGGCAGCTGACAATGCCGGCGGCGACCCAGA  GCGGCGTTGATGCTATCGCGGCGGGCAATTATCACTCTCTGGCTCTGAAAGATGGGGAAGTGAT  TGCTTGGGGCGGTAACGAGGATGGCCAAACTACGGTGCCGGCCGAGGCCCGTTCCGGTGTAGA  TGCTATTGCGGCAGGCGCTTGGGCGAGCTACGCGCTGAAAGACGGCAAAGTGATCGCCTGGGG  TGATGATTCCGACGGTCAGACCACCGTGCCGGCGGAAGCCCAGTCGGGTGTGACCGCGCTGGA  TGGTGGTGTGTATACCGCGCTGGCAGTAAAAAACGGTGGTGTTATTGCGTGGGGGGATAATTAT  TTTGGCCAGACCACAGTGCCGGCGGAGGCTCAGTCCGGGGTGGATGATGTTGCAGGCGGCATC  TTTCACAGCCTGGCGCTGAAAGATGGTAAAGTTATTGCGTGGGGCGATAATCGCTATAAACAAAC  CACAGTTCCAACCGAGGCGCTGAGTGGCGTGTCGGCCATTGCTTCAGGTGAATGGTATAGCCTG  GCTCTGAAAAATGGTAAAGTAATTGCGTGGGGTAGCAGCCGCACCGCGCCTAGCTCCGTCCAAT  CGGGGGTGAGTTCCATTGAAGCCGGTCCGAACGCCGCTTACGCACTGAAAGGTGGGAGCGGTT  CTGGCCATCATCACCATCATCATTAA   G49TAG gene sequence   ATGGCAACCTCGGTGGTGGCCTGGGGCTAGAACAATGACTGGGGTGAAGCTACCGTGCCGGCC  GAAGCGCAGAGTGGTGTGGATGCAATTGCAGGTGGTTATTTTCATGGGCTGGCACTGAAAGGGG  GTAAAGTACTGGGCTGGGGTGCAAATCTGAACGGGCAGCTGACAATGCCGGCGGCGACCCAGA  GCGGCGTTGATGCTATCGCGGCGGGCAATTATCACTCTCTGGCTCTGAAAGATGGGGAAGTGAT  TGCTTGGGGCGGTAACGAGGATGGCCAAACTACGGTGCCGGCCGAGGCCCGTTCCGGTGTAGA  TGCTATTGCGGCAGGCGCTTGGGCGAGCTACGCGCTGAAAGACGGCAAAGTGATCGCCTGGGG  TGATGATTCCGACGGTCAGACCACCGTGCCGGCGGAAGCCCAGTCGGGTGTGACCGCGCTGGA  TGGTGGTGTGTATACCGCGCTGGCAGTAAAAAACGGTGGTGTTATTGCGTGGGGGGATAATTAT  TTTGGCCAGACCACAGTGCCGGCGGAGGCTCAGTCCGGGGTGGATGATGTTGCAGGCGGCATC  TTTCACAGCCTGGCGCTGAAAGATGGTAAAGTTATTGCGTGGGGCGATAATCGCTATAAACAAAC  CACAGTTCCAACCGAGGCGCTGAGTGGCGTGTCGGCCATTGCTTCAGGTGAATGGTATAGCCTG  GCTCTGAAAAATGGTAAAGTAATTGCGTGGGGTAGCAGCCGCACCGCGCCTAGCTCCGTCCAAT  CGGGGGTGAGTTCCATTGAAGCCGGTCCGAACGCCGCTTACGCACTGAAAGGTGGGAGCGGTT  CTGGCCATCATCACCATCATCATTAA   S213TAG gene sequence   ATGGCAACCTCGGTGGTGGCCTGGGGCGGTAACAATGACTGGGGTGAAGCTACCGTGCCGGCC  GAAGCGCAGAGTGGTGTGGATGCAATTGCAGGTGGTTATTTTCATGGGCTGGCACTGAAAGGGG  GTAAAGTACTGGGCTGGGGTGCAAATCTGAACGGGCAGCTGACAATGCCGGCGGCGACCCAGA  GCGGCGTTGATGCTATCGCGGCGGGCAATTATCACTCTCTGGCTCTGAAAGATGGGGAAGTGAT  TGCTTGGGGCGGTAACGAGGATGGCCAAACTACGGTGCCGGCCGAGGCCCGTTCCGGTGTAGA  TGCTATTGCGGCAGGCGCTTGGGCGAGCTACGCGCTGAAAGACGGCAAAGTGATCGCCTGGGG  TGATGATTCCGACGGTCAGACCACCGTGCCGGCGGAAGCCCAGTCGGGTGTGACCGCGCTGGA  TGGTGGTGTGTATACCGCGCTGGCAGTAAAAAACGGTGGTGTTATTGCGTGGGGGGATAATTAT  TTTGGCCAGACCTAGGTGCCGGCGGAGGCTCAGTCCGGGGTGGATGATGTTGCAGGCGGCATC  TTTCACAGCCTGGCGCTGAAAGATGGTAAAGTTATTGCGTGGGGCGATAATCGCTATAAACAAAC  CACAGTTCCAACCGAGGCGCTGAGTGGCGTGTCGGCCATTGCTTCAGGTGAATGGTATAGCCTG  GCTCTGAAAAATGGTAAAGTAATTGCGTGGGGTAGCAGCCGCACCGCGCCTAGCTCCGTCCAAT  CGGGGGTGAGTTCCATTGAAGCCGGTCCGAACGCCGCTTACGCACTGAAAGGTGGGAGCGGTT  CTGGCCATCATCACCATCATCATTAA To recombinantly produce the BLIP2-AzF variants, E. coli BL21 (DE3) cells (NEB) were transformed with the pET-BLIP2 (kindly provided by Tim Palzkill) and pEVOL-AzF plasmids (provided by Ryan Mehl via Addgene) and grown on LB agar supplemented with 30 µg/mL kanamycin and 35 µg/mL choramphenicol for transformant selection; colonies were picked and grown in 500 mL 2×YT broth with 30 µg/mL kanamycin and 35 µg/mL choramphenicol at 37 °C until reaching an OD600 of 0.4. AzF and IPTG were added, both to 1 mM to induce expression and the cells grown for a further 24 hours at 16 °C with 200rpm shaking. BLIP2 WT was produced in a similar manner but without antibiotic and AzF additions required to cultivate cells transformed with pEVOL-AzF 2 . The cells were pelleted by centrifugation and the cells resuspended in binding buffer (50 mM Tris-HCl, 5 mM imidazole pH8.0). The cells were lysed using a French Press exerting 20,000 psi of pressure. Soluble protein was harvested by centrifugation at 25,000 x g for 10 mins. The supernatent containing histidine-tagged BLIP2 were first purified by cobalt affinity chromatography using a home-made gravity column containing 5mL HisPur cobalt resin (Thermofisher) equilibrated in binding buffer (50 mM Tris-HCl, 5 mM imidazole pH8.0). The column was then washed with 10 column volumes of binding buffer before being eluted in a single step with 20mL 100 mM imidazole, then concentrated and desalted using a PD10 column (GE). Purity of protein was assessed by SDS-PAGE.

Enzyme inhibition by BLIP2 and its AzF variant.
The BL assay measured the initial rate of nitrocefin hydrolysis by TEM WT , by recording the absorbance increase on hydrolysis of the -lactam ring amide bond. Initial rates were recorded using 0.3 nM TEM with increasing concentrations of BLIP2 until full inhibition had been achieved for all BLIP2 variants. The initial rates were plotted and fitted to the Morrison Equation 6-7 using GraphPad Prism software to estimate the Ki app for each interaction.

Additional explanatory methods and text.
To quantitatively determine the effect of the four mutations on BLIP2's ability to bind (and thus inhibit) TEM WT , enzyme inhibition assays were used to measure the apparent inhibition constants (Ki app ) compared to BLIP2 WT . Ki is equivalent to dissociation constant (KD) but in the context of enzyme inhibition. Absolute binding parameters for BLIP2-BL interactions are difficult to determine by normal methods as their dissociation constant is so low (80-540 fM). Working at protein concentrations in this range are impractical for our requirements so a comparative analysis with BLIP2 WT will allow an apparent Ki to be determined. This will allow the direct effect of the mutation on binding affinity to be assessed. The BLIP2 WT -TEM WT dissociation constant (KD) has previously been determined to be 480 fM using separately measured association and dissociation rate constants. However, these experiments require up long term (weeks) incubation of BLIP2 WT -TEM WT complexes in an excess of the inactive TEM 166Ala mutant and was considered unnecessary for our purpose of comparing the variants to BLIP2 WT .
This allowed us to investigate the binding affinities of the BLIP2 variants for the wildtype BL enzyme TEM used in the electrical measurements. BLIP2 WT (wild type) has a high affinity for TEM-1 BL (BL TEM ), which was also observed here (Ki 32 ± 4 pM). Both BLIP2 41AzF and BLIP2 213AzF retain sub-nanomolar affinity for TEM (Ki 78-160 pM). As predicted, BLIP2 49AzF exhibited reduced binding affinity (Ki 2839 ± 406 pM) due to its location at the BLIP2-BL binding interface.

Electrode fabrication and protein attachment
Electrodes were prepared as previously reported. 8 Nanosized electrodes were fabricated on a p-doped Si/SIO2 wafer by a combination of laser and electron beam lithography, followed by evaporating a thin adhesive layer of Cr and a thick layer of Au.
95% semiconducting single walled carbon nanotubes were purchased from Nanointegris. DBCO-amine and pyrene-NHS were purchased from Sigma-Aldrich. All of other chemicals used in this paper were ordered from Sigma-Aldrich. 0.1 mg 95% semiconducting SWNT was dispersed in 500 µL of 1% SDS solution via sonication for 1h. The supernatant of SWNT was collected after being centrifuged for 1h, which was used as stock solution.
To immobilise a small bundle of SWNTs between electrodes, dielectrophoresis (DEP) was performed by applying an alternating current (AC) voltage between electrodes after SDS-dispersed SWNT solution was cast on the electrodes. Typically, the frequency of the generator was switched onto Vp-p = 3V at f=400KHz. The as prepared CNT solution was diluted by 100-fold and cast onto the chip with a pipette (5 µL). After a delay of 10 seconds, the substrate was washed carefully with water to remove SDS and blown gently with Nitrogen gas. Electrical measurement was performed to confirm the immobilisation of SWNTs between electrodes. 0.4mg 1-Pyrenebutyric acid N-hydroxysuccinimide ester (pyrene-NHS) and 0.5mg Dibenzocyclooctyne-amine (DBCO-amine) were dissolved and mixed in 200 µL dimethylformamide (DMF) to give a solution containing 5 mM of pyrene-NHS and 9mM of DBCO-amine. The mixture was placed on a shaker overnight at room temperature to form DBCO-pyrene. Subsequently, 10 µL ethanol-amine was added to the mixture to blocked unreacted NHS groups. The prepared devices were immersed in the mixture solution for 1h. DBCO-pyrene would be immobilised onto the sidewall of SWNT bundles via π-π stacking between pyrene and SWNTs. The devices were rinsed with iso-propanol and DPBS buffer after incubation. Subsequently, 20 µL azide modified BLIP2 variants (200 nM, in DPBS) were cast on the devices. After incubation for 3h, the devices were rinsed with water and blown dried with Nitrogen gas for atomic force microscopy (AFM) imaging. For electrical measurements, pyrene-butanol (PB, ratio of PB to pyrene-NHS is 3:1) was added into the mixture solution after the formation of DBCO-pyrene, serving as spacers to create enough space for the binding of BLIP2 to TEM. Additionally, after incubation for 3h, the devices were rinsed with DPBS buffer and used to performed real time detection of TEM. The biosensing electrical measurements were performed the same day of protein interfacing to minimise the potential degradation/instability of the protein-CNT hybrids over time   We counted the proteins on individual CNTs/small bundles from AFM topographical images ( Figure SI-6). We estimated that there are 2.6 ± 0.7 proteins attached to an individual CNT per 100 nm. This is similar coverage to that observed on using the photo-induced UV-nitrene aziridine functionalisation. 2 To estimate the number of proteins per device, we also calculated the surface area of CNT bundles in each device compared to individual CNTs. Since the height of CNT bundles between electrodes is about 10-20nm, we estimate that there are 10-20 CNTs immobilised in each device. Bundles of CNTs containing 10 or 20 CNTs were modelled on a flat surface using PyMOL and their solvent-accessible surface area measured using the "get_area" command. It was calculated that a 10-CNTs bundle has ~ 5.5 times the available area of an individual CNT on a flat surface and a 20-CNT bundle has ~ 9.3 times the area. The gap between gold electrodes in the FET devices is 300nm. Therefore, we estimate that the number of proteins attached in each device is of 40 to 80 proteins.

Real time detection of TEM in serum
We also performed real time detection of TEM in serum. Steroid-free serum purchased from MP Biomedicals, Inc. was diluted with the DPBS buffer by 10-fold (serum solution). TEM solutions used in this detection was diluted with serum solution. 100 mV bias were applied between source-drain electrodes (gate electrode was grounded). A drop of serum solution (10µL) was drop-cast on the devices: this volume was chosen also due to the miniaturized size of the device (please note that evaporation of liquid can affect the electrical measurements if these are performed in a prolonged time) . Serum solution was added to the devices when the reading of the current was stable. Subsequently, 5µL of TEM solutions (diluted with serum solution) of various concentrations, were cast on the devices. Although transfer characteristics (Isd vs Vg) can provide information about the electronic behavior of the CNTs, we preferred to present the real time biosensing measurements as more relevant -and in our case more accurate -to demonstrate the sensing results of the platform we developed. This because we found that many factors can influence the accuracy of Isd vs Vg data, from the evaporation of liquid affecting the concentration of solution, to issues in the devices due to the deposition of CNT bundles.