GDH-LBD library construction. We used the Saturated Programmable Insertion Engineering (SPINE) algorithm23 to design and generate a vector library that expresses GDH with LBD inserted across the open reading frame (Fig. S1). Python-GDH plasmid (without BsaI or BsmbI recognition sites) were prepared in FASTA format, then submitted to the custom algorithm (https://github.com/schmidt-lab/SPINE). This program generates oligo sequences, their corresponding primers for amplification, and the target backbone primers for inverse PCR. These oligos were microarray-synthesized and amplified as eight oligo pools. Each pool joined its backbone in parallel, then pooled together as the intermediate library. The CFU of the eight sub-libraries ranged from 1200 to 5200, corresponding to > 99.99% coverage. The Sanger sequencing confirmed that nearly 80% of the colonies were perfect variants (Table S1). All eight libraries were pooled together at an equimolar ratio, resulting in the intermediate library with the genetic handle crossing GDH. Lastly, the LBD sequence replaced the genetic handle through BsaI-mediated Golden Gate cloning, resulting in the LBD-inserted GDH library (yielded > 9200 colonies with 81.5% perfect variants).
Deep sequencing of GDH-LBD libraries. The plasmids from the GDH-LBD library were extracted using a Monarch Plasmid Miniprep kit. The product served as the template for specific GDH-LBD amplification with 15 cycles of PCR using the Q5 hot start polymerase. The resulting amplicon was run on 1.2% agarose gels, purified by gel extraction, and quantified by Quant-iT Picogreen. A total of 1.6 µg of DNA was prepared for deep sequencing using the Illumina MiSeq, 2 x 150 bp configuration by GENEWIZ. Insertion sites (Fig. S2) were identified from 140,340 raw sequencing reads. Alignments were processed using the DIP-seq pipeline (http://github.com/SavageLab/dipseq) on both forward and reverse reads.
GDH-LBD whole cell assay. GDH-LBD library DNA was transformed into 50 µL E. coli BL21 chemical component cells (NEB) for protein expression. The transformed cells were diluted to various concentrations and grown on LB agar plates with 50 µg/mL kanamycin overnight at 37 ◦C. The following day, single colonies were picked and transferred to 96 deep-well plates in 600 µL of LB medium, complementing with 10 mM CaCl2, 1 µM PQQ, 50 µM IPTG and 50 µg/mL kanamycin. The cells grew at 25 ◦C with shaking at 60 rpm for 18 hours. The final OD600 were 0.4 ~ 0.6.
Whole-cell GDH-LBD activity was determined spectrophotometrically at room temperature by following the reduction of dichlorophenolindophenol (DCPIP) at 600 nm, using phenazine methosulfate (PMS) as a primary electron mediator (Fig. S3). A reagent solution containing 45 mL MOPS buffer (10 mM, pH 7), 1 mL DCPIP (20 mg dissolved in 5 mL of H2O overnight), 1 mL PMS (45 mg in 5 mL of H2O, freshly made and kept in the dark) and 1 mL glucose (1 M in MOPS, overnight) were prepared in each use. GDH-LBD activity was correlated to the velocity of DCPIP oxidation which was reported as the absorption decreasing at 600 nm over time. The coefficient of variance for whole cells assay was calculated as 11% (Fig. S4). Given the ideal coefficient of variance is 10%,35 our assay provided a reliable screen approach.
GDH-LBD library screening. We first screened the GDH-LBD library for the variants capable of oxidizing glucose with 4-HT. In the colorimetric assay, the GDH-LBD active variants could turn the dark blue reagent solution yellow or colorless. We performed the screen in a Costar 96 flat transparent plate with 10 µL GDH-LBD cells mixed with a 180 µL assay reagent. Positive controls were prepared with cell expression wide-type GDH. Negative control was made with 10 µL cells harboring empty vectors to eliminate the interference of reagent decay over time. By arbitrarily comparing the color decay to the control group, GDH-LBD variants showing activity within 5 minutes were pursued to the second screen round. All the active variants were pooled together, and their plasmids were extracted using the Monarch Plasmid Miniprep kit. The resulting plasmids were sequenced using the Illumina MiSeq, 2 x 150 bp configuration by GENEWIZ, yielding 54430 read counts.
In the second round, 10 µL active GDH-LBD cells were mixed with 180 µL reagent and 10 µL DMSO or 4-HT (1mM dissolved in DMSO). Absorption at 600 nm was recorded every 15 s for 10 min by Tecan Spark plate reader at room temperature with orbital shaking at 500 rpm. Biologically independent experiments (n = 3) were performed in this screening round. The data were collected and processed with two-tailed, independent t–tests for the P value. Only these variants showed statistically different (P < 0.05) activities with 4-HT and DMSO were identified as allosteric variants. LBD-insertion sites on GDH were identified by Sanger sequencing. Their change degree is shown in Fig. S6.
Protein purification and reconstitution. Frozen cells (~ 30 g) were thawed and resuspended in 150 mL buffer (50 mM HEPES, 10 mM imidazole, 300 mM NaCl, 3 mM CaCl2 and 5 mM beta-mercaptoethanol, pH 7) with lysozyme and DNase, and lysed with an AVESTIN EmulsiFlex-C3 homogenizer. The lysate was clarified at 22,000 x g for one hour to precipitate the cell debris. Protein is bound on a 5 mL Histrap FP Ni-NTA column (Cytiva) on an FPLC and washed with a 20% gradient of imidazole buffer (50 mM HEPES, 300 mM imidazole, 300 mM NaCl, 3 mM CaCl2 and 5 mM beta-mercaptoethanol, pH 7). The protein was eluted in 40% imidazole buffer and loaded on a 15 mL HiTrap Desalting column (Cytiva) to remove imidazole. PQQ was added to the protein solution at a 2:1 molar ratio. The mixture was stirred for 30 min. Excess PQQ was removed by loading again on the desalting column. The purity of the protein was confirmed by SDS–PAGE (Fig. S8), and the concentration was measured with Bradford assay.
GDH-LBD protein activity. A reagent solution containing 47 mL MOPS buffer (10 mM, pH 7), 1 mL DCPIP (20 mg dissolved in 5 mL of H2O), 1 mL PMS (45 mg in 5 mL of H2O, freshly prepared and kept in the dark) were prepared for each use. The assay was performed on a flat transparent 96 plate in triplicates. Since GDH has a high turnover rate, each test requires only 5–10 ng of protein. The experiments were prepared by mixing 10 µL protein with a 180 µl reagent solution. Reactions were initialized by adding 10 µL glucose and recorded every 15s for 10 min by Tecan Spark plate reader with orbital shaking at 500 rpm. (The coefficient of variance is 7.6%)
Fc-LPEI synthesis. Fc–LPEI was synthesized as previously reported.30, 31 High molecular weight polyethylenimine (LPEI, 0.100 g, 2.33 mmol) was dissolved in a mixture of acetonitrile (7 mL) and methanol (3 mL) at 80°C. A solution of 3-bromopropyl-dimethylferrocene (0.137 g, 0.45 mmol) in acetonitrile (2 mL) was added to the stirring LPEI solution at 80°C, and the mixture was stirred at 90–100°C for 24 hours. The reaction mixture was cooled to room temperature and the solvent was removed under reduced pressure. Excess ferrocene was removed by soaking the resulting polymer in diethyl ether (10 mL) for 1 hour at room temperature. The diethyl ether was decanted, and excess solvent was removed under reduced pressure. The final product was a brown malleable solid (0.169 g, 71% yield). 1H-NMR in CD3OD: δ (ppm) 1.67 (2H, C-CH2-C), 1.94 (6H, Fc(-CH3)), 2.30 (2H, Fc-CH2-C), 2.50–3.10 (16H, N-CH2-C, polymer backbone), 3.87 (7H, Fc-H).
Protein/Fc–LPEI hydrogel film preparation. Nonwet-proofed AvCarb carbon paper electrodes were cut into strips (3 cm x 0.5 cm). Except for the designed electrode surface area, the rest of the strip was dipped into melting paraffin wax to seal the conductive surface. Before electrochemical testing, the exposed end was coated with protein/Fc-LPEI hydrogel for analysis, while the waxed end was used as an electrochemical connection point.
Six µL GDH-5E+ (16 mg/mL) or bovine serum albumin (BSA, 20 mg/mL) was added to 14 µl of Fc–LPEI solution (12 mg/mL in H2O). The mix was vortexed before adding EGDGE (0.75 µL, 4.4% by volume in H2O). The resulting mixture (~ 20 µL) was then drop-coated onto a glassy carbon electrode (diameter, 3 mm) in 3 µL aliquots or AvCarb carbon electrode (0.5 cm x 0.5 cm) in 10 µL aliquots. The resulting electrodes were cross-linked at 4 ◦C for 6 h.
Cyclic voltammetry (CV). CVs were performed with three electrodes: AvCarb coated with protein/Fc–LPEI as the working electrode, a saturated calomel electrode (SCE) as the reference, and a platinum mesh counter electrode. Experiments were performed at room temperature in 5 mL MOPS buffer (100 mM, pH 7.0). All the protein hydrogel-coated electrodes were electrochemically conditioned with 10 cycles of scanning at 100 mV/s. The electrochemical activities of GDH-5E+ were measured by CV scan at 20 mV/s in a stationary solution for three cycles. The third cycle was used for analysis and reporting. Reported potentials were referenced to the standard hydrogen electrode (SHE) by adding 244 mV to the measured values.
Amperometric i–t curves. A fixed potential of 300 mV vs. SCE was first applied to working electrodes for 100 s. This step could effectively decrease the double–layer charging current in the following tests. Electrodes were soaked in a stirred solution of 100 mM MOPS, pH 7. Glucose was injected into the electrolytes after the current stabilized (usually 180 seconds).
Laccase cathode preparation. Laccase from Trametes versicolor (≥ 0.5 U/mg) was purchased from Sigma–Aldrich. This enzyme deposition solution was prepared by suspending laccase (30 mg) in 75 µl of 200 mM citrate–phosphate buffer (pH 4.5), followed by the addition of 7.5 mg of anthracene–modified multi-walled carbon nanotube (An-MWCNTs)36. The mixture was subjected to successive vortex/sonication steps until the ink was homogenous. TBAB-Nafion (25 µL)32 was added, and a few more vortex/sonication steps were undertaken to promote thorough mixing. 30 µL ink was spread onto the AvCarb electrodes (1.25 cm x 0.8 cm) and dried at 4 ◦C overnight.
Construction of the glucose/O 2 EFC. EFCs were set up in a custom-made electrochemical cell, where a Nafion® 212 proton exchange membrane (PEM) was used to compartmentalize the anodic and cathodic chambers (Fig. S18). In the anodic chamber, we use GDH-5E+/Fc–LPEI coated AvCarb (0.5 cm x 0.5 cm) as the anode, with 4 mL 100 mM MOPS as electrolytes. On the cathodic sides, we used laccase/An-MWCNTs/TBAB-Nafion coated AvCarb electrode (1.25 cm x 0.8 cm). This chamber contained 4 mL 200 mM citrate–phosphate buffer, pH 4.5, having O2 bubbled.
Fabricating OECT device. The OECT devices were fabricated using a low-cost methodology. Glass slides were first cleaned with soap and DI water, then washed in isopropanol and acetone before blow-drying with clean, dry air. Later, thermal evaporation deposited Au (65 nm) on the cleaned glass slides, which were cut into glass strips (1.2 cm x 2.5 cm). Using a new single–edge razor blade, we cut a line along the glass strip, creating a channel with a length of 50 µm. By masking with Kapton tape, we defined a 1 mm width for the channel. Additional areas where Au was deposited were left uncovered to provide electrical contact for alligator clips. The slides were further subject to UV-Ozone treatment for 10 minutes to ensure effective adhesion to the OECT solution. An Ag/AgCl wire was prepared as the gate electrode. The channel of OECT was made of poly (3,4 ethylene dioxythiophene) doped with poly(styrene sulfonate) (PEDOT:PSS). An OECT solution was prepared by ultrasonicating PEDOT:PSS with ethylene glycol (5% v/v) and (3-glycidyloxypropyl) trimethoxysilane (GOPS, 1% w/w). 4-dodecylbenzenesulfonic acid (DBSA, 0.1% v/v) was added to ensure uniform mixing. This OECT solution was spin-coated to the prepared channel at 1000 rpm for 45 seconds and accelerated at 200 rpm/s. The resulting devices were annealed at 1400 ◦C for 30 minutes.
Coupling self-powered sensor with OECT. We connected the source measure meter (SMU) to power the source-drain channel, as shown in Figure S23. The OECT was held at a constant Vsd of – 600 mV while monitoring the Isd. We let the current stabilize for 50 seconds before connecting the cathode to the gate of OECT. After Isd reached a steady state, 40 µL of 2 M glucose comprising 1% (v/v) 4-HT (10 µM, in DMSO) was added to the anodic chamber. The current was constantly monitored, allowing sufficient time to stable before disconnecting from EFC. Meanwhile, EFC’s OCP change was also recorded by a potentiostat (Fig. S24).