Flat Solenoidal Ice‐Binding Proteins as Scaffolds for Solid‐Binders

Solid interfacing biomaterials is a crucial aspect of bionanotechnology and important for applications such as biosensing. Because of their potentially large contact area, flat solenoidal proteins are ideal scaffolds for designing proteins binding to surfaces of man‐made solids such as minerals, metals, and plastics. To explore this opportunity, a naturally occurring flat solenoidal protein: the Rhagium inquistor Antifreeze Protein from the insect Rhagium inquisitor is re‐designed. By mutating 4, 6, and 10 out of its 4 × 5 arrays of threonines into arginines, it have arrived at the silica‐binding proteins RiSiBP‐4, RiSiBP‐6, and RiSiBP‐10. Variants with increasing numbers of arginines bind stronger to silica, but are also less stable and increasingly difficult to produce. It is found that the RiSiBP‐6 variant binds strongly to silica yet still has good stability and easy production. It is shown that sfGFP‐RiSiBP‐6 fusions allow for the functional display of a monolayer of sfGFP cargo on silica surfaces, suggesting the general usefulness of flat solenoidal proteins as scaffolds for designing solid‐binding proteins.

flat beta-solenoid structure, with one side containing the icebinding residues that consist of an highly ordered matrix of 20 threonine residues that are all pointing in exactly the same direction, and the other side being hydrophilic, containing mostly polar and charged amino acids. [17] In the absence of stable de novo designs for β-solenoid scaffolds, we argue that RiAFP is an ideal scaffold for re-engineering as a flat protein to also bind to other solid surfaces by engineering the threonine residues on the ice-binding site. While for binding to crystalline surfaces (such as ice, or minerals) both precise placement and orientation of binding groups, and an extended flat binding face are crucial, in many other cases (for example amorphous solids, alloys, or plastics), it is especially an extended flat binding interface which will be crucial, and constraints on the placement of binding residues may be less stringent.
As proof of principle, we here consider (amorphous) silica as a model surface, and re-engineer RiAFP to strongly bind to silica (and no longer to ice) by substitution of the ice-binding residues for silica-binding ones. We show that these substitutions can be done based on no rational design because it is quite well-known which amino acid preferentially binds to specific solid surfaces from analysis of the ample SBP sequences. In the case of silica, it is especially arginine that binds strongly, mainly through electrostatic interactions [18,19] which is confirmed by the many silica-binding SBPs that are rich in arginines [3,[16][17][18][19] . We, therefore, engineer a series RiAFP variants with 4, 6, and 10 arginines replacing ice-binding threonines on the ice-binding face of the RiAFP proteins, generating a series of flat silica-binding proteins RiSiBP-4, RiSiBP-6, and RiSiBP-10. We demonstrate the RiSiBP proteins adsorb as a thin monomolecular layer on silica surfaces, and have an affinity for silica surfaces that increases with their number of designed arginines.
Finally, we demonstrate that cargo fused to the RiSiBP proteins can be functionally displayed on silica surfaces. More generally, we conclude that the rational design of surface binding motifs via residue engineering of proteins with well-defined solid binding surfaces is a successful strategy for rational design of folded and flat surface binding proteins.

Results and Discussion
Antifreeze proteins (AFPs), also known as ice-binding proteins (IBPs), have evolved in cold-adapted organisms to bind nucleating ice crystal surfaces to, among others, inhibit formation of damaging ice crystals. [15] Although the precise mechanism of the ice-binding is still opaque, the protein surface residues responsible for ice-binding on the so-called ice-binding site (IBS), have been identified for most AFPs. For Rhagium inquisitor AFP (RiAFP) this is a highly ordered flat matrix of 4 × 5 threonines on one side of its beta-solenoid fold (Figure 1). The IBS of RiAFP spans a large atomically flat surface area of ≈450 nm 2 where the threonines all protrude in the same plane. Therefore the IBS is ideally suited for rational design of other surface interacting residues.
As proof of principle, we here engineer Rhagium inquisitor Silica Binding Proteins (RiSiBPs). To this end, we mutated a varying number of threonine residues on the IBS into positively charged arginine residues, a residue type well known to bind strongly to silica, mainly via electrostatic interactions. [18,19] We thus hypothesize that RiSiBPs with increasing numbers of arginines will have a likewise increasing affinity for silica, and potentially other negatively charged materials.
Based on the RiAFP crystal structure (PDB ID: 4DT5) we design 3 silica-binding variants by respectively selecting 4, 6, and 10 threonines in the 4 × 5 residue matrix, to be mutated into arginines. In selecting the threonines to be mutated, we aimed at homogeneously distributing arginines over the 4 × 5 matrix. The resulting proteins are referred to as RiSiBP-4, RiSiBP-6, and RiSiBP-10. For clarity we will refer to the RiAFP wildtype protein as RiSiBP-0, having zero arginines on the ice-binding site. Figure 1 shows cartoon model structures of RiSiBP-0, RiSiBP-4, RiSiBP-6, and RiSiBP-10 generated with Rosetta Design. The ice-binding site is highlighted on RiSiBP-0. The figure also illustrates that the arginine mutations are indeed equally distributed over the IBS.
Next, to increase the solubility and expression of the designed proteins, we included a highly soluble super folder green fluorescent protein (sfGFP) at the N-terminal to make sfGFP-RiSiBP fusion proteins. The sfGFP can be cleaved off www.advmatinterfaces.de using a tobacco etch virus (TEV) protease cleavage site which also acts as a spacer between sfGFP and RiSiBP (Figure 2A). All designs include a C-terminal hexahistidine tag such that TEV-cleaved RiSiBP can be purified on a Ni-NTA column. Complete sequences for sfGFP-RiSiBP-0, sfGFP-RiSiBP-4, sfGFP-RiSiBP-6, and sfGFP-RiSiBP-10 can be found in Table S1, Supporting Information.
We recombinantly expressed sfGFP-RiSiBP in E.coli and purified the proteins using a Ni-NTA column followed by a polishing size-exclusion chromatography (SEC) purification (Figures S1 and S2, Supporting Information). The purified sfGFP-RiSiBP proteins were cleaved with TEV protease, and RiSiBP was isolated using a Ni-NTA column followed by another SEC purification to produce high-purity monomeric RiSiBP ( Figure S3, Supporting Information). Figure 2B shows the SDS-PAGE of the final purified proteins used in this study. Lanes 1-3 show final purified sfGFP-RiSiBP-0, sfGFP-RiSiBP-4, and sfGFP-RiSiBP-6. Lanes 4-6 show final purity of RiSiBP-0, RiSiBP-4, and RiSiBP-6. The sfGFP-RiSiBP-10 is not shown, because it expressed poorly and the majority of the expressed protein was found to be insoluble ( Figure S1, Supporting Information). Similarly, enforcing 12 arginines on one side of a flat de novo designed helical repeat protein (DHR10), [20] the same scaffold used for the de novo design of mica binders, [12] also led to poor expression and insoluble protein ( Figure S4, Supporting Information). This together demonstrates that independent of the scaffold type, the design of high densities of positive charged can be problematic for soluble expression and folding.
To verify that secondary structure remains the same despite the arginine mutations we obtained circular dichroism (CD) spectra and compared them to the RiSiBP-0 spectrum, for which a crystal structure is available (PDB ID: 4DT5). Figure 3A shows the normalized ellipticity for all RiSiBPs overlap with a single minimum at 220 nm, indicating high beta-sheet content. These results are expected considering that RiSiBP-0 is a betaroll solenoid containing predominantly beta-sheets [17] (Figure 1). To test the influence of the arginine mutations on the scaffold structure stability, a thermal denaturation ramp was performed which shows RiSiBP proteins gradually unfold at high temperature with melting temperatures of T m ≈ 50°C for RiSiBP-0 and RiSiBP-4, and T m ≈ 38 °C for RiSiBP-6 ( Figure 3B). This demonstrates that incorporation of arginines on the IBS of the RiAFP somewhat reduces the thermal stability of the scaffold, which is not unexpected given the increasingly high density of arginines on the sheets resulting in increased electrostatic repulsion.
Having established that the RiSiBP designs have the correct secondary structure elements and are thermostable at room temperature, we set out to test whether the threonine to arginine substitutions affect ice-binding activity by performing an ice recrystallization inhibition (IRI) assay. [21] AFPs bind to ice owing to the well-ordered presentation of the threonine array, and it has been demonstrated previously that the incorporation of just a few steric mutations with high entropy side chains (such as arginines) on IBS of native AFPs can greatly reduce their IRI activity. [22] In the IRI assay, small ice crystals are nucleated and kept at −8 °C for 45 min in the presence of  The molar ellipticity is normalized at 220 nm (θ norm ). All RiSiBP show a highly similar spectrum typical for proteins rich in beta-sheet content, indicating the protein is folded to the native state. B) thermal melting curves, with on y-axis fraction of native protein (P native ) at λ = 220 nm, indicating thermal melting points of 49, 48, and 38 °C for RiSiBP-0, RiSiBP-4, and RiSiBP-6 respectively. www.advmatinterfaces.de 0.2 mg mL −1 RiSiBP protein. Figure 4 shows that in the presence of RiSiBP-0, the native RiAFP, ice crystals have not grown in volume over the course of 45 minutes. This crystal growth inhibition is the hallmark of IRI activity. Therefore we confirm RiSiBP-0 is indeed folded to its active native conformation, with its threonine array on the IBS are presented. For the arginine mutants RiSiBP-4 and RiSiBP-6, much faster growth of large ice crystals is observed, confirming that indeed only a few threonine to arginine substitutions on the IBS are enough to reduce the IRI properties of RiSiBP-0.
Having established the native-like folding of RiSiBP proteins, we next test the effect of arginine mutations on RiSiBP scaffold on the binding strength of protein to silica (SiO 2 ) surface using Quartz Crystal Microbalance Dissipation (QCM-D). In QCM-D, the oscillation frequency of the SiO 2 -coated quartz sensor is monitored and changes with adsorbed mass to the surface. Provided dissipation is limited, the adsorbed mass scales linearly with the decrease in oscillation frequency (ΔM∝ − Δf) according to the Sauerbrey equation. [23] Figure 5A shows QCM-D results for the adsorption of 1 µM of the RiSiBPs, followed by washing with PBS. We find RiSiBP-0 adsorbs to SiO 2 but after washing with PBS, RiSiBP-0 desorbs almost completely, indicating binding is likely due to weak nonspecific interactions. At 1 µM, the kinetics of adsorption of RiSiBP-4 and RiSiBP-6 are similar, and both have a significantly higher plateau adsorption than RiSiBP-0. They differ however in their desorption behavior, with significant desorption during PBS washing for RiSiBP-4, but no desorption during PBS washing for RiSiBP-6. Figure 5B shows the limiting adsorbed masses per unit area, as estimated from the Sauerbrey equation, for RiSiBP layers versus the concentration at which the layers were formed. A similar mass per unit area is observed for RiSiBP-6 as compared to RiSiBP-4, possibly indicating full monolayer coverage for both cases. The decrease in coverage after prolonged washing with PBS is shown in Figure 5C, again demonstrating that only for RiSiBP-6 there is almost no desorption after prolonged washing with PBS. During QCM-D measurements of RiSiBP-0, RiSiBP-4, and RiSiBP-6, the measured energy dissipation (ΔD) remains well below 1 ppm during the measurements, indicating a only a thin layer of proteins is adsorbed, with no dissipation, thus allowing the use of the Sauerbrey equation for estimating coverages ( Figure S5, Supporting Information). The fact that the adsorbed layers are thin and homogeneous is further confirmed by atomic force microscopy ( Figure S6, Supporting Information).
Finally, to demonstrate the usefulness of RiSiBP to display functional proteins on silica surfaces, we coat silica nanoparticles (d = 163 nm) with RiSiBP-sfGFP, where the sfGFP is a placeholder for functional proteins one might want to display, such as capture proteins for certain ligands in biosensing. Assuming monolayer coating of the RiSiBP-sfGFP on the silica nanoparticles, we expect a layer height with h ∼ 6 nm based on the cross-sectional diameter of crystal structures of sfGFP (PDB ID: 2B3P) and RiSiBP (PDB ID: 4DT5) ( Figure 6A). Using Dynamic Light Scattering (DLS), we measured hydrodynamic diameters of the silica nanoparticles incubated at various concentrations of sfGFP-RiSiBP-0 and sfGFP-RiSiBP-6. From the slight increase in the hydrodynamic diameters of the particles, we are able to estimate layer thicknesses. Indeed, for

www.advmatinterfaces.de
both sfGFP-RiSiBP-0 and sfGFP-RiSiBP-6, at protein concentrations of ≈500 nM, we find a saturation layer thickness h ∼ 6 nm ( Figure 6B-C). For sfGFP-RiSiBP-6 at concentrations >500 nM we observe layer thicknesses h ∼ 13 nm ( Figure 6B). Possibly, multilayer formation at high concentrations can be caused by electrostatic attraction between the RiSiBP-6 and sfGFP domains which have opposite net charges of, respectively, +4 and −8. Overall though, the DLS results confirm monolayer coating of RiSiBP-sfGFP on silica surfaces, with the RiSiBP being bound to the surface and sfGFP pointing into the bulk, as illustrated in Figure 6A.
To test the mobility of sfGFP immobilized on silica surfaces through RiSiBP tagging, we used fluorescence recovery after photobleaching (FRAP). Figure 7 shows results of FRAP experi-ments for 1 µM of sfGFP-RiSiBP-0 and sfGFP-RiSiBP-6. For sfGFP-RiSiBP-0 we observe that bleached proteins are gradually being replaced by unbleached ones via exchange with the bulk phase. Exchange via the bulk and hence fluorescence recovery is significantly less for sfGFP-RiSiBP-6, consistent with our earlier experiments that determined that this protein is not easily rinsed away by PBS. For both sfGFP-RiSiBP-0 and sfGFP-RiSiBP-6 the boundary of the bleached area remains sharp, which would not be the case if there was significant in-plane mobility.

Conclusion
To demonstrate the concept of engineering flat solenoid proteins for solid binding, we have designed a series of silicabinding proteins by introducing arginine mutations in a flat solenoid antifreeze protein. The RiSiBP-6, with 6 arginines, binds strongly to silica surfaces, cannot be easily washed away, and has low in-plane mobility when adsorbed. It can be fused with a cargo protein such as sfGFP, and on silica surfaces sfGFP-RiSiBP-6 fusions form a monolayer coating with thickness h ∼ 6 nm, with the sfGFP being oriented towards the bulk.
Introducing more and more arginines leads to both production problems and lower protein stability. Possibly the introduction of self-assembly, such as fibril formation, could further increase their surface binding strength. [24,25] In the long run, the de novo computational design of flat beta-solenoids could lead to highly stable scaffolds that may tolerate many more surface mutations as compared to native beta-solenoids such as the RiAFP that we investigated here. Such scaffolds would also be suitable for decoration with other surface-binding amino acids, to target other surface materials such as plastics, metals, and other oxides.
By illustrating the concept of the surface engineering of flat solenoid proteins for solid binding using a natural ice-binding protein, we hope to stimulate further work in this area to eventually arrive at highly stable de-novo designed beta-solenoid binders for a wide range of solid materials that can be used for bio-material interfacing.

Experimental Section
Construct Cloning: Genes encoding sfGFP-RiAFP proteins were codon optimized using codon harmony (version 1.0.0) and obtained as synthetic DNA fragments from Twist Bioscience. The gene fragments were cloned into pET-24(+) vector using restriction digestion with BamHI and XhoI restriction endonucleases (New England Biolabs) and the insert was validated by sequencing.
Protein Expression and Purification: A shaker flask with 1L luberia-broth (LB) medium + 50 mg L −1 kanamycin was inoculated from an overnight starter culture of T7-Express E.coli containing the expression plasmid. The culture was incubated shaking (125 rpm) until 0.6 < OD 600 < 0.8 at 37 °C. Protein expression was induced by 1 mM isopropyl B-Dthiogalactoside (IPTG) and expression was continued at 18 °C overnight. The broth was centrifuged at 6000 x g and the cell pellet was resuspended in cold 30 mL lysis-wash buffer (50 mM Tris-HCl pH 8.0, 400 mM NaCl, 30 mM imidazole) + 1 mM phenylmethylsulfyl fluoride (PMSF) serine protease inhibitor. Cells were lysed by sonication on ice for 7 min. with a 2s on-off duty cycle at 85% amplitude using a Q-sonica Q125 with a CL-18 probe. The lysate was centrifuged at 30000 x g for 30 min. at 4 °C and the clarified supernatant was applied two times on lysis-wash buffer equilibrated Ni-NTA resin with a column volume (CV) of 2 mL. The resin was washed with 25 CVs lysis-wash buffer and eluted with 3 CVs elution buffer (50 mM Tris-HCl pH 8.0, 400 mM NaCl, 300 mM imidazole). The eluate was further purified and buffer-exchanged via size exclusion chromatography (SEC) on an Agilent Infinity II GPC with a Superdex 200 Increase 10/300 GL (GE Healthcare) column in PBS (10 mM phosphate pH 7.4, 137 mM NaCl, 3.7 mM KCl). Fractions containing monomeric sfGFP-RiSiBP proteins were pooled, and protein concentration was determined using a molar extinction coefficient of 23630 M −1 cm −1 at 280 nm. Purified sfGFP-RiAFP variants were stored at 4 °C.
To purify cleaved RiSiBP protein, the sfGFP-RiSiBP eluate was diluted 20X in cleavage buffer (25 mM Tris-HCl pH 8.0, 100 mM NaCl and 10% glycerol) and ≈3 mg of home-made tobacco etch virus (TEV) protease [26] was added and incubated for 48 hr. at room temperature. The cleaved protein solution was applied to a 2 mL Ni-NTA resin washed with 5 CVs of lysis-wash buffer, and eluted in 3 CVs elution buffer. The eluate was spin concentrated using Amicon 3.5 kDa spin filters (Merck Millipore) and further SEC purified on a Superdex 75 10/300 GL (GE Healthcare) equilibrated in PBS. The monomeric fractions containing RiSiBP protein were pooled and spin concentrated to 1 mg mL −1 using a molar extinction coefficient of 1615 M −1 cm −1 at 280 nm, flash frozen and stored at −20 °C.
Circular Dichroism: Samples were prepared by thawing RiSiBP aliquots on ice and diluting to 0.2 mg mL −1 in PBS. A spectropolarimeter J-715 (JASCO Corporation, Japan) to record CD profiles in a quartz cuvette with a 1 mm path length. Spectral scans were averaged over 20 measurements and obtained at 10 °C with a scan rate of 50 nm min −1 using a response time of 2s. For thermal denaturation experiments, circularly polarized light was recorded at a fixed wavelength λ = 220 nm and a response time of 2s at a heating rate of 1°C min −1 .
Ice Recrystallization Inhibition: RiSiBP protein aliquots were thawed and diluted to 0.2 mg mL −1 in PBS + 20% w/w sucrose. Samples of 2 µL were sandwiched between two precleaned cover slides, which were rapidly frozen at 20 °C min −1 to −40 °C to generate tiny polycrystalline crystals using a Linkam LTS420 cold stage attached to a Nikon ECLIPSE Ci-Pol Optical Microscope. The samples were annealed for 90 min at a sample annealing temperature of −8 ± 1 °C. Microphotographs were taken every 2 min using a Lumera3 CCD camera to follow the ice growth over time. Image analysis was done in ImageJ.
Quartz Crystal Microbalance Dissipation: A Q-Sense E4 (Biolin Scientific) QCM-D instrument was used to quantify protein binding to silica. QCM-D quartz sensors coated with SiO 2 (QSX 303, Biolin Scientific) are cleaned according to the manufacturer's instructions. Samples for QCM-D experiments were prepared by diluting freshly thawed RiSiBP protein aliquots in PBS buffer. Prior to the measurements, the instrument was equilibrated with PBS buffer for at least 20 min. at a flow rate of 50 µL min −1 . The experiment was started by flushing the sensors with RiSiBP protein at 50 µL min −1 for ≈70 min. Next, the flow was briefly stopped, and inlets were switched to PBS buffer to wash the sensors for ≈55 min. Analysis of QCM-D data was done using Q-Sense Dfind software (Biolin Scientific). Experimental observables are the changes in resonance frequency (Δf) and energy dissipation (D). If energy dissipation is small enough, the adsorbed mass per unit area can be (ΔM) estimated from the change in resonance frequency through the Sauerbrey equation: Where n is the number of the odd harmonic (n = 1,3,5, …) and the numerical prefactor, for a 5 MHz crystal was, per unit area of crystal surface, C = 17.7 ng (cm 2 •Hz) −1 . The active area of the crystal was 0.2 cm 2 .
Dynamic Light Scattering: A ZS-Nano (Malvern, UK) DLS instrument was used to quantify the hydrodynamic diameter of silica nanoparticles coated with sfGFP-RiAFP protein. Light scattering was measured at a fixed angle of 173°, in a quartz cuvette at 20 °C. Samples were prepared by mixing silica nanoparticles of 163 nm diameter (Bang Laboratories), PB (10 mM phosphate pH 7.4), PBS, and sfGFP-RiAFP purified proteins at various concentrations. Prior to measurement, samples were incubated >5 min. at 20 °C. In all samples, the final NaCl concentration was kept constant at 28.2 mM to reduce silica nanoparticle aggregation due to charge screening. [27] Hydrodynamic radii are calculated based on the average of 15 measurements, obtained using the Zetasizer software version 7.13 (Malvern, UK).
Fluorescence Recovery After Photo-Bleaching: Microscopy slides and cover glasses were prepared by plasma cleaning for 2 minutes, and resting for >5 min. Two strips of double-sided tape were stuck on the microscope slide, with the distance just shy of the length of the cover glass ( Figure S7, Supporting Information). Between the two strips, a 20 µL droplet containing 1 µM sfGFP-RiSiBP in PBS pH 7.0 was pipetted. The drop was incubated for 2 min, after which the cover glass was placed on top of the tape strips to create a capillary. Through the capillary 150 µL of 1 µM sfGFP-RiSiBP protein was flushed and the ends were sealed with grease to prevent evaporation. The slide was imaged using an oil immersion Plan Apochromat 60x objective (Apo 60x Oil λS DIC N2) with a numerical aperture (NA) of 1.42 on a Nikon Ti confocal microscope with a Nikon C2 laser module. During time-lapse imaging, the focal plane (1.5 AU) of the cover glass was maintained using perfect focus system in the NIS-Element software. Images were recorded at 1 frame per second using a 488 nm laser. Each measurement started with 11 frames obtained at 0.3% laser power to collect a prebleach baseline. Bleaching was done by increasing laser power to 20% and 50% using 488 and 405 nm lasers for 6 seconds. Fluorescence recovery was measured by scanning at 0.3% 488 nm laser for 30 min.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.

Data Availability Statement
The data that support the findings of this study are available in the supplementary material of this article.