Electrically Switchable Polymer Brushes for Protein Capture and Release in Biological Environments

Abstract Interfaces functionalized with polymers are known for providing excellent resistance towards biomolecular adsorption and for their ability to bind high amounts of protein while preserving their structure. However, making an interface that switches between these two states has proven challenging and concepts to date rely on changes in the physiochemical environment, which is static in biological systems. Here we present the first interface that can be electrically switched between a high‐capacity (>1 μg cm−2) multilayer protein binding state and a completely non‐fouling state (no detectable adsorption). Switching is possible over multiple cycles without any regeneration. Importantly, switching works even when the interface is in direct contact with biological fluids and a buffered environment. The technology offers many applications such as zero fouling on demand, patterning or separation of proteins as well as controlled release of biologics in a physiological environment, showing high potential for future drug delivery in vivo.


Table of Contents
Experimental section Page 3 Figure S1 Page 5 Figure S2 Page 5 Figure S3 Page 6 Figure S4 Page 6 Figure S5 Page 7 Figure S6 Page 8 Figure S7 Page 9 Figure S8 Page 10 Figure S9 Page 11 Figure S10 Page 11 Figure S11 Page 12 Figure S12 Page 13 Table S1 Page 14 Theory Page 15 References Page 16
Diazonium salt synthesis: The synthesis of diazonium salt ( Figure S1) involved a modified literature procedure. [1] Under an inert atmosphere, 4aminophenethyl alcohol (2.94 g, 20 mmol) and tetrafluoroboric acid (48% solution in water, 9.94 g, 113 mmol) were dissolved in acetonitrile (20 mL). In a separate flask, tert-butyl nitrate (2.269 g, 22 mmol) was dissolved in acetonitrile (12 mL). Both solutions were degassed and cooled to −20 ºC alongside 200 mL of diethyl ether. After 20 min the solutions were warmed to 0 ºC, before the tert-butyl nitrate solution was added to the 4aminophenethyl alcohol solution dropwise with stirring. The reaction was then stirred for a further 1 h. The reaction was terminated by dropwise addition of the dark yellow solution to rapidly stirring diethyl ether (200 mL). After additional stirring for 1 h the supernatant was decanted off. The brown colored precipitate was dried and 3.69 g of impure diazonium salt was obtained and carried forward without further purification. To verify the product, 1H NMR spectra were recorded at ambient temperature on a Varian 400 MHz NMR spectrometer. Spectra were analyzed relative to external tetramethylsilane and were referenced to the most downfield residual solvent resonance (CDCl3: δH 7.26 ppm). 1H NMR resonances of the diazonium salt matched those previously reported [1] and analysis revealed a purity of 80%.
For microelectrodes the piranha wash step was omitted to prevent destruction of the surface due to delamination of the gold film with nanoholes.
Surface activation: Gold or platinum surfaces (QCM and SPR sensors) were placed in a glass jar with a septum seal containing diazonium salt 1 (0.301 g, 1.28 mmol) and the jar was purged with N2. In a separate flask, ascorbic acid (0.028 g, 0.16 mmol) was dissolved in water (40 mL) and the solution was degassed for 1 h. Then, the ascorbic acid solution was transferred into the sealed glass jar causing reduction of the diazonium salt. The gold surfaces were stirred in the solution for 1 h by use of a platform shaker (nitrogen bubbles that appear on the surface after 15 min indicate successful diazonium salt monolayer formation), after which they were thoroughly rinsed in water then ethanol and dried. To convert the diazonium monolayer into a polymerization initiator layer, the gold surfaces were exposed to α-bromoisobutyryl bromide (0.222 mL, 1.80 mmol) and triethylamine (0.302 mL, 2.17 mmol) in dichloromethane (20 mL) for 10 min, after which surfaces were rinsed in ethanol and dried under N2.
Surface-initiated polymerization: ATRP was used to prepare PMAA polymer brushes similarly to established protocols. Inhibitor was removed from the monomer tert-butyl methacrylate using an alumina column, after which it was stored at −20 °C, then warmed to room temperature immediately before use. Reactions were carried out using standard Schlenk line techniques under an inert atmosphere of N2. CuBr2 (0.006 g, 0.03 mmol), and PMDTA (0.052 mL, 0.246 mmol) were dissolved in dimethyl sulfoxide (20 mL) and, alongside a separate flask of tert-butyl methacrylate (20 mL, 0.1231 mol), was deoxygenated via vigorous bubbling of N2 for 30 min. The reaction solution and monomer were then transferred via cannula into a screw-top jar (with rubber septa lid) containing initiator-prepared gold surfaces. Quartz crystal microbalance: Sensor crystals coated with gold were used and measurements were performed using a Q-Sense E4 (Biolin Scientific). All frequency and dissipation data shown corresponds to the fundamental resonance at ~5 MHz (no overtones). A flow cell with an electrochemical module (QEM 401) was used to perform in-situ electrochemical experiments. A Gamry Interface 1000E potentiostat (Gamry Instruments) was connected to the electrochemical cell. For every experiment the internal resistance of the circuit was measured (Get Ru) and the open circuit potential was measured to verify an acceptable reference electrode state and correctly connected circuit. The Ag/AgCl reference electrode used was a World Precision Instrument low leakage "Dri-ref" electrode. Potentials were applied either by chronoamperometry (fixed voltage) or by voltammetry sweeps with a rate of 100 mV/s unless otherwise specified. The active electrode area is 0.78 cm 2 for all current data shown. All voltages stated are vs the reference electrode.
Surface plasmon resonance: Measurements were performed on a SPR Navi 220A instrument (BioNavis), both in air and in water. The total internal reflection (TIR) and SPR angle was recorded on three different laser wavelengths and in two different flow channels. The flow rate of buffer used was 20 μL/min. Electrochemical SPR measurements were performed by connecting a potentiostat (same as for QCMD) to a cell designed for this purpose (from the instrument manufacturer). The methodology of analyzing SPR spectra by Fresnel modelling and the quantification has been described in previous work. [2] In brief, the refractive index of the dry polymer brushes was set to 1.522. The deposited diazonium layer was assumed to have refractive index 1.5 (typical for organic coatings). The refractive index of the proteins was assumed to be equal to that of the polymer. To obtain surface coverage, the densities of the dry polymers and proteins were used (1.22 g/cm 3 and 1.35 g/cm 3 respectively).

4
Plasmonic detection with nanohole arrays: Extinction spectroscopy was performed to detect the shift of the resonance peak from nanohole arrays in 30 nm gold films. [3] In brief, the surface was illuminated by a tungsten lamp and a fiber coupled photodiode array spectrometer (B&WTek) analyzed the spectrum. (A small fraction of the light was directly transmitted through the glass in between the electrodes, but this only influences the absolute extinction values, not the resonance shift.) Note that for plasmonic nanohole arrays the signal is a spectral shift in nm, while SPR uses angular shifts in degrees.
Fabrication of microelectrodes: To create microscale stripe electrodes, a laser writer (Heidelberg Instruments DWL 2000) was used. The photoresist (LOR3A) was spin coated at 4000 rpm and baked on a hotplate at 180 °C for 5 min. A second layer of S1813 was spin coated at 4000 rpm and baked on a hotplate at 120 °C for 2 min. The pattern was written by a 60 mW laser beam after which the sample was developed in developer MF-318 for 50 s. The nanohole array was then prepared by colloidal lithography. [3] Finally, lift-off was performed in remover mr-Rem 400.
Desorption of proteins: The surfaces were immersed in a solution of proteins for 30 min to ensure saturated binding, using the kinetics from SPR and QCM measurements as guideline. For protein loading by hydrogen bonds, the surface was rinsed in PBS pH 5.0 and water, then dried with N2.
Desorption was performed by immersing the samples in serum at its native pH. Alternatively, electrochemical release was carried out in a beaker, with a Pt cage as counter electrode and a Ag wire as reference electrode. The wire was preconditioned by depositing chloride ions by applying a +0.5 V (vs the Pt counter) for 10 min in 1 M HCl.
Protein conjugation: Two different fluorophores were used in the conjugation of fluorescent dyes to the amines of BSA: Alexa Fluor 488 and 555 (with tetrafluorophenyl or N-hydroxysuccinimide ester groups). A BSA solution (100 μL, 10 mg/mL, pH 8.5) was mixed with Alexa Fluor dye (100 μg), and the resulting solution was inverted every 10 min for 1 h. The reaction was terminated by the addition of PBS pH 5.0, which reduced the pH to 5.0 and diluted the sample to a protein concentration of 0.2 mg/mL. The fluorescent proteins were immobilized in PMAA brushes in the same manner as the native proteins.
Fluorescence microscopy: All fluorescence measurements were conducted using a Zeiss Axio Observer 7 inverted microscope equipped with an Axiocam506 camera. Microelectrodes in air were imaged using a 10× objective. For Alexa Fluo 488, excitation was at 450-490 nm and emission was collected at 500-550 nm. For Alexa Fluo 555, excitation was at 533-558 nm and emission was collected at 570-640 nm. Faradaic currents are observed in most cases. For DOPA, the redox activity is quite strong initially but quickly decreases, most likely because an organic coating is formed due to electropolymerization. Several species are capable of acidification but hydroquinone is clearly the most efficient.
z / μm pH Figure S4. Calculated local pH as a function of distance z from the surface after different pulse times. The case of pH increase by O2 reduction is modelled. The bulk pH is 7.4 and the diffusivity of protons is set to 9.3×10 −9 m 2 /s. Already after ~1 s, there is a noticeable pH change even z = 10 μm from the surface due to the very high diffusivity of protons. In the brush region (z ≈ 100 nm) the pH goes up to ~12. This explains the efficient switching and that electrostatic repulsion (pH > pI) can be achieved for all proteins. For pH lowering, we found no analytical expression, but the local pH should be on the order of log10 . SPR data at 670 nm on 20 nm Pt, confirming brush synthesis after diazonium grafting and ATRP. The analysis is performed by fitting Fresnel models just like for Au but with different parameters for the metal permittivity. Although no clear resonance dip appears in the spectra, the region next to the total internal reflection shows intensity changes that are sufficient to determine surface coverage. In the initial "background" model, the polymer is excluded. When fitting the spectrum after polymerization, only the thickness is allowed to vary. For the particular spectrum in the plot, the fitted (dry) PMAA thickness was 15 nm.  Figure S11, no H2O2 is added in solution as it is produced by GOX. Table S1. Summary of results from electrochemical QCMD and SPR experiments of hydrogen bond immobilization and release of different proteins. In all cases, immobilization was performed at pH 5 by exposing the surface to a protein concentration of 0.3 mg/mL for ~10 min. The irreversible signals from immobilization (after rinsing) were on the order of ~100 Hz (QCMD) or a few degrees (SPR). Larger proteins tended to immobilize in higher amounts, while no correlation was seen with respect to isoelectric point. However, here the purpose is to look at the pH increase required for full release. The results show that proteins with high pI require a higher pH to leave the surface. The data set also indicates that smaller proteins are more easily released and many are fully desorbed even under conditions where their net charge is actually opposite to that of the brush. Finally, we note that almost no protein with net negative charge remains bound when the brush is negatively charged. Fully released at bulk pH between 7 and 8.
Note that Cs needs to be inserted with unit of M. Given that protons diffuse around 10 times faster than small molecules and n is typically 2 (hydroquinone and H2O2), we get pHsurf ≈ log10(2/Cs). Thus, from Equation 8 it is clear that a few mM concentration of the proton producing species could be sufficient to reach pH as low as ~3.
The main limitation in this model is that the buffering effect from weak acids and bases in the bulk solution is not accounted for. However, this is probably compensated by the fact that the brush extension from the surface is much less (at least an order of magnitude) than the characteristic diffusion distance. Although we present no exact function to describe the pH gradient during acidification, it must have a shape reasonably similar to that in Figure S4, but inverted. Therefore, if buffering species are ignored, the pH inside the brush will be even lower than the average value in the depletion zone. Buffering species will counteract this effect, but mostly in the outer region of the depletion zone (as they will also be depleted close to the surface). Therefore, Equation 8 should be valid for estimating the pH inside the brush.
Convection is very difficult to implement in this kind of analytical modelling, but it can be noted that flow generally reduces the extension of the interfacial region where the pH is altered. [6] This means that the buffering effect from the solution should become more noticeable. However, the flow will also deliver the molecules that cause the pH changes more efficiently to the electrode. These effects of the local pH should roughly cancel each other and thus it is not surprising that no effect of convection was observed in the experiments.