Nanoparticles (NPs) are becoming increasingly important in the chemical sciences, including biomedicine, due to their unique properties, such as high surface area to volume ratios, ability to navigate blood vessels, and enhanced chemical reactivity. [1][2][3] Biomedical applications of NPs include their use as novel drug delivery systems that allow for targeted and controlled drug release, 4 as biosensors for medical diagnostics, 5 and as contrast agents for bioimaging techniques. 6 The growing usage of NPs in these and other fields, such as cosmetics and energy, leads to increased exposure of humans to nanomaterials from the environment representing a growing health concern. 7 Biomolecules, especially proteins, are often readily adsorbed on NP surfaces. [8][9][10] This underscores the importance of understanding the interaction of nanoparticles with proteins at atomistic detail.
Nuclear magnetic resonance spectroscopy (NMR) is a powerful tool to study protein-NP interactions. [11][12][13] Common NMR observables, including chemical shifts and NMR spin relaxation rates, are notably sensitive to protein-NP binding processes. Many sophisticated solution NMR experiments are available to quantitatively report the kinetics and thermodynamics of protein-NP absorption equilibria and provide site-specific information on the structure and dynamics of a protein residing on the NP surface. For instance, the interaction with NPs results in peak amplitude reductions in 1D 1 H spectra or 2D 1 H- 15 N heteronuclear single quantum coherence (HSQC) spectra of proteins and may cause chemical shift perturbations (CSP) reflecting the protein interaction interface or the structural changes of proteins upon NP-binding. The binding kinetics and the protein dynamics in the bound state can be quantified through dark-state exchange saturation transfer (DEST) experiments 14,15 or via transverse R2 NMR spin relaxation.
NPs can be composed of a wide variety of materials, including metals, metal oxides, polymers, and lipids. [16][17][18] The nature of protein-NP interaction is influenced by both the type of material and surface modifications of the NPs. Generally, proteins exhibit tight binding to the noble metal-based NPs, such as gold NPs and silver NPs. For example, proteins GB3 and ubiquitin interact with citrate-coated gold NPs with the adsorption/desorption process taking place on timescales slower than milliseconds. 19,20 The adsorption of proteins onto gold NPs can be abolished by tuning the surface charge of NPs via coating with a mixture of oppositely charged small ligands, for example, lysine and citrate. 21 The interaction with some metal NPs, including uncoated silver NPs and zwitterionic peptide-coated gold NPs, induces protein structural changes or promote aggregation. 22,23 In contrast, interactions between proteins and silica nanoparticles (SNPs) are of a transient nature and fully reversible. 24 Intrinsically disordered proteins characteristically bind to unmodified (i.e. pristine) anionic SNPs via positively charged residues, leading to residue-site specific enhancements in the 15 N transverse R2 relaxation rates, which vary based on the amino acid-residue type and their neighbors. [25][26][27] Globular proteins, on the other hand, interact with SNPs in a predominantly non-specific manner, and their intramolecular dynamics are generally unperturbed. Recently, we developed the nanoparticle-assisted spin relaxation (NASR) method 24,28 that takes advantage of the non-specific, transient binding between globular proteins and SNPs, thereby effectively slowing down the rotational tumbling rate of the proteins. The difference in the 15 N transverse R2 rates in the presence and absence of SNPs ( 15 N-DR2) is directly proportional to the generalized order parameters S 2 of 1 H-15 N-bonds, quantitatively reporting intramolecular protein dynamics all the way up to the nanoparticle tumbling timescale, which depends on the average SNP size according to the Stokes-Einstein-Debye relationship and typically lies in the microsecond (μs) range (see the Supporting Information).
Interactions between proteins and biological liposome NPs also enhances the 15 N-R2 relaxation of globular proteins as demonstrated by Clore and co-workers. [29][30][31] Strikingly, the differential 15 N-DR2 profile of ubiquitin with POPG liposomes 29 shows a very different behavior compared to the profile with unmodified SNPs or Al 3+ -doped SNPs. 28 Unlike for SNPs, where DR2 is proportional to the S 2 , 15 N-DR2 of ubiquitin with liposomes show large variations among residues in a way that is unrelated to the internal dynamics of the protein. 29 Instead, the DR2 profile suggests global diffusive axial rotational motions of ubiquitin relative to the surface of the liposomes. 29 Here, we show that the interaction mode between proteins and organic zwitterion SBScoated SNPs, ZwiSNPs, closely resembles that of proteins interacting with liposomes. Human ubiquitin and the Ras binding domain of human B-Raf (RBD) both display weak yet unique interactions with ZwiSNPs. Based on experimental backbone 15 N transverse R2 relaxation rates, we quantified the motions of both proteins on the surface of ZwiSNPs and identified their different interaction interfaces.
3-(dimethyl-(3-(trimethoxysilyl)propyl)ammonio)propane-1-sulfonate (SBS) coated Ludox TM-40 colloidal silica nanoparticles with an average diameter of 25 nm were synthesized as described previously. 32 The ZwiSNPs were dialyzed into 20 mM sodium phosphate at pH 7.0 and then directly mixed with uniformly 15 N-labeled ubiquitin or RBD of B-Raf in the same buffer to prepare the SNP-doped NMR samples. The final concentrations of the proteins in the samples were 0.6 mM, and the ZwiSNP concentrations were estimated to be around 20 to 30 µM for ubiquitin and around 0.3 µM for RBD. 5% D2O was added to all samples for field lock. RBD samples contained 2 mM reducing agent tris(2-carboxyethyl)phosphine (TCEP).
NMR experiments were performed at 298 K on a Bruker AVANCE III spectrometer operating at 850 MHz 1 H resonance frequency (19.97 T) equipped with a TCI cryoprobe. Ubiquitin and RBD backbone amide 15 N R1 and R2 spin relaxation rates both in the presence and absence of ZwiSNPs were obtained from standard R1 and R1ρ relaxation experiments [33][34][35][36] with recovery delay set to 2 s and the R1ρ spin-lock field strength around 2000 Hz. The enhancements in 15 N-R2 due to the presence of SNPs was then determined for each resonance as DR2 = R2 NP -R2 free . Ubiquitin backbone amide 15 N/ 15 N-1 H CSA/DD transverse cross-correlation rates hxy were measured using the symmetrical reconversion experiment by Pelupessy et al. 37 The recovery delay was set to 1.5 s and the relaxation period T was set to 90 milliseconds (ms) for the free sample and 50 ms for the NP-doped sample. Like the transverse auto-relaxation rate R2, the cross-correlation rates hxy increase upon the addition of nanoparticles with the enhancements determined for each resonance as Dhxy = hxy NPhxy free . The backbone assignments of RBD were taken from BMRB entries 17030 and 30047.
Proteins, when bound to the surface of ZwiSNPs, can exhibit multiple modes of motion. (1) Proteins in the bound state tumble together with the NPs with a rotational correlation time 𝜏 !" . (2) Experimental DR2 profiles suggest global axial rotation about a single axis perpendicular to the NP surface, which gives rise to an order parameter 38
and 𝜃 is the angle between the N-H bond vector and the rotation axis. The correlation time of this rotation is denoted as 𝜏 #$% , and for convenience, we define the global rotation axis to be the z-axis here (Figure 2a). (3) The protein intramolecular motions are assumed to be unperturbed by the interaction with ZwiSNPs, as is the case for unmodified SNPs. 24,28 The site-specific intramolecular motions are described, likewise to the free state, by an order parameter S 2 and an internal correlation time 𝜏 '(% . ( 4) Additionally, global fluctuations on the surface of NPs may exist, such as Gaussian axial fluctuations (GAF) [39][40][41][42] in the tangent plane to the spherical NPs (perpendicular to the rotation z-axis in ( 2)), which is equivalent to 2D GAF motions of the z-axis itself. Here, we assume that these motions are statistically independent with the time correlation function and power spectral density function for calculating relaxation rates given in the Supporting Information.
The zwitterion SBS contains a positively charged quaternary amine (NR4 + ) and a negatively charged sulfonate group (SO3 -) (Figure 1b). SBS coating gives rise to an overall neutral surface charge of the ZwiSNPs in a broad pH range (pH 2-10), reducing protein adsorption effectively. 32 Yet, ZwiSNPs still interact with proteins that are highly positively charged, such as the wellknown "sticky" lysozyme (isoelectric point 43 pI = 11). Lysozyme can be absorbed on the surface of zwitterionic SNPs and leads to SNP agglomeration, although the interaction is weaker than for unmodified SNPs. 32,44 The transverse spin relaxation rates of proteins are exquisitely sensitive to interactions with NPs. Attachment of the protein to NP surface drastically slows down the protein rotational tumbling by 2-3 orders of magnitude, leading to observable enhancements of the transverse R2 relaxation rates even with low bound populations. 24 Here, we monitored the interaction between human ubiquitin and ZwiSNPs by measuring the 15 N transverse R2 relaxation rates of protein backbone amides. We do not observe any chemical shift changes on the 1 H-15 N HSQC spectra between the free and ZwiSNP-doped samples. 15 N-R2 spin relaxation is accelerated in the presence of ZwiSNPs, while 15 N-R1 values remain essentially unaffected. Their interaction with proteins is much weaker than in the case of unmodified SNPs: to reach comparable DR2 values around 200 times higher ZwiSNP concentration is required than for unmodified SNPs.
The differential relaxation rates DR2 obtained for each residue in ubiquitin are shown in Figure 1c. For the unmodified SNP, DR2 is directly proportional to the amplitude of sub-µs protein intramolecular motions S 2 , and the DR2 profile is relatively flat for the majority of ubiquitin residues, except for the flexible b-hairpin loop (Leu8-Thr12) and the C-terminal tail (Leu71-Gly76). 24,28 In contrast, DR2 values obtained with ZwiSNPs show large variations for the rigid parts of ubiquitin, with "spikes" at selected residues including Lys6, Thr22, Gln40, Gln41, Leu55, Ser56, Gln62, Glu64, and Ser65. There is no amino-acid type specificity for the spiking residues.
The 15 N/ 15 N-1 H CSA/DD transverse cross-correlation hxy rates are unaffected by the ms conformational exchange contributions. 45,46 Dhxy values are highly correlated to DR2 (Figure S1)
indicating that the site-specific increments of relaxation rates are not due to residual conformational exchange contributions of protein in the free state induced by ZwiSNPs that are not sufficiently suppressed by the spin-lock during the R1ρ relaxation experiments.
The spiking pattern of the ubiquitin-ZwiSNP DR2 profile closely resembles those observed for POPG liposome NPs by Clore and co-workers, 29 suggesting ubiquitin may undergo rotations around an axis when bound to the surface of ZwiSNPs (Figure 2a) similar to the motions on the liposome NPs. Such global axial rotational motion results in structure-dependent DR2 values, namely DR2 will show a double-dip pattern as the function of the q angle between the N-H bond vector and the rotation axis, reaching minima at the magic angle (q = 54.74° and 125.26°) and a maximum when the bond vector is parallel to the rotation axis (Figure 2d). Fitting of the experimental profile to this global axial rotational motion using the first structure model of the NMR structure of ubiquitin (PDB 1D3Z) shows good agreement (Figure 2). Additional global motions, such as the reorientational fluctuations of the rotation axis, may exist. One possibility is that the rotation axis stochastically reorients in a 2D harmonic potential relative to its equilibrium position, which can be described by the 2D GAF model (see Supporting Information). However, the addition of 2D GAF does not noticeably improve the agreement with experiment (Figure S2).
It shows that if 2D GAF of the axis is present, the fluctuation amplitude must be sufficiently small so that the DR2 profile becomes insensitive to these 2D GAF motions (standard deviation of the fluctuation amplitude sGAF < 7°).
A The best fit gives kex = 4.4×10 5 s -1 , pb = 0.87 %, and trot = 1.6 μs. However, these three parameters are highly correlated, as can be seen from the c 2 surface (Figure S2b), and various combinations would give rise to similar c 2 . However, the range of these parameters can be more confidently defined: kex is between 5×10 3 and 3×10 6 s -1 (comprising the full range that gives rise to a perceptible ΔR2) and trot falls in the range between 0.6 and 3 μs. The orientation of the rotation axis, on the other hand, is well defined (see Figure S2a). The rotation axis together with the electrostatic surface potential of ubiquitin are shown in Figure 2. The axis is orthogonal to the positively and negatively charged faces of ubiquitin.
The Ras-binding domain of B-Raf was chosen because it has a ubiquitin-like structure with five b-strands, one a-helix, and two 310-helices, despite the lack of sequence similarity (Figure 3).
However, the two proteins have rather distinct surface charge distributions. The 15 N-DR2 of RBD also shows a bumpy profile, yet with spikes at different positions than ubiquitin. The DR2 of RBD and ubiquitin are uncorrelated (after accounting for the residue numbering differences based on structure alignment). The profile of RBD can also be well explained by a global axial rotational motion when bound to the ZwiSNP surface, with the DR2 values showing the "double-dip" pattern as a function of the angle between N-H bond and rotation axis using the first structure model of PDB 2L05 (Figure 3e). The best-fit yields kex = 6.0×10 5 s -1 , pb = 0.97 %, and trot = 0.93 μs. Like in the case of ubiquitin, the three parameters are highly correlated. The rotation axis deviates by an angle of 63° from the one of ubiquitin after structure alignment (Figure 3a). Such a large angle is not unexpected based on the distinct surface charge distributions of the two proteins. Similar to ubiquitin, the rotation axis is perpendicular to the most positively charged face of RBD, with the positively charged face likely being the interaction interface with ZwiSNPs.
Zwitterionic coating is a commonly used method to create anti-fouling surfaces. 48,49 The neutral net charge and well-hydrated surface effectively prevent the adhesion of biomolecules and cells, which is a desired property in many biomedical applications. In addition, the zwitterionic coating improves the colloidal stability of NPs under challenging conditions, such as high salt concentrations. For example, SBS-coated SNPs can resist aggregation in 3 M sodium chloride, while unmodified SNPs aggregate within minutes under the same condition. 32,44 SBS-coated SNPs are generally resistant to protein binding. Yet SBS-coated SNPs still interact with proteins that contain large positive surface charge patches (e.g. lysozyme). Here, we observed weak interactions between SBS-coated SNPs and the globular proteins ubiquitin and RBD at low ionic strength (20 mM sodium phosphate).
15 N transverse DR2 rates quantitatively report on the dynamics of proteins in the NP-bound state. In NASR experiments using unmodified SNPs, 24 protein motions remain largely unaffected between the free and SNP-bound states apart from the global tumbling. The resulting 15 N-DR2 values, therefore, directly reflect the intrinsic protein intramolecular motions on the sub-μs timescale. Similar to R2 relaxation, the 15 N CSA/DD transverse cross-correlation hxy is dominated by the spectral density at near zero-frequency J(0) with Dhxy being proportional to the NASR order parameter S 2 (NASR) covering motions up to the μs timescale. It has been known for a long time that hxy is unaffected by chemical exchange on the ms timescale 45,46 that gives rise to exchange broadening increasing the apparent R2 rates. Comparison of hxy and R2 rates of the NP-free sample allows for easy identification of the chemical exchange for each site. 46 In NASR measurements, on the other hand, the chemical exchange contribution to R2 cancels in DR2, since exchange contributes equally irrespective whether the protein is free or NP-bound. Hence, both DR2 and Dhxy are in good approximation unaffected by ms chemical exchange.
In the case of SBS-coated SNPs, 15 N-DR2 revealed additional global axial rotational motions of the proteins in the bound state about axes perpendicular to the NP surface. In fact, for both ubiquitin and RBD, the 15 N-DR2 profiles can be well explained by such axial rotations when using 3D structural models of the free protein taken from the PDB. Since the 15 N-DR2 profiles sensitively depend on the N-H bond vector orientations in the protein relative to the rotation axis, it indicates that for the vast majority of residues the structural models of PDB 1D3Z and 2L05 accurately reflect the average backbone N-H bond orientations with the overall protein backbone structures remaining essentially unchanged in the SNP-bound state. In addition, incorporation of sub-µs intramolecular protein dynamics further improves the agreement with experiment in the case of ubiquitin, especially in the flexible b-hairpin loop and C-terminal tail regions, suggesting that the fast protein motions remain unaltered in the bound state analogous to the situation with unmodified SNPs. The correlation times of the global axial rotational motions trot are determined to be on the low-µs regime between 0.5 to 3 µs, on par with the rotational tumbling time of the ZwiSNPs themselves (τNP = 1.78 μs). The residence times of proteins on the ZwiSNPs fall in the sub-µs to sub-ms range consistent with the kex regime that gives rise to perceptible 15 N-DR2 effects.
The presence of axial rotational protein motions in the bound state observed here has also direct thermodynamic consequences irrespective of timescales. They cause an increase of the entropy in the bound state, which stabilizes protein-ZwiSNP interactions by decreasing the entropy difference between the free and bound states. 50 Furthermore, the backbone relaxation experiments used here can also be extended to the protein side chains 51,52 by measuring NASR 53 of methyl groups.
Similar to the backbone, global axial rotational protein motions will lead to a site-specific, structure dependent increase in transverse relaxation rates of methyl-13 C and 2 H spins.
For both ubiquitin and B-Raf RBD, the global axial rotational motion occurs about one single axis perpendicular to the predominantly positively charged face of the protein. These positively charged regions are likely to be the primary interaction interfaces with ZwiSNPs, facing the negatively charged sulfonate groups in the SBS coatings. While ubiquitin and RBD share very similar 3D structures, the differences in the orientations of their rotation axes and interaction interfaces with the ZwiSNPs are consistent with their distinctive surface charge distributions, highlighting the role of electrostatics in defining protein-SBS coated SNP binding behavior.
Human ubiquitin has been extensively studied using NMR, and the interaction site observed here has also been identified to be driving ubiquitin-NP interactions for various NP types, 29,31,[54][55][56] such as silver NPs and POPG liposomes. The Ras-binding domain RBD in B-Raf selectively interacts with the Ras protein in its active GTP-bound state and triggers downstream signal transduction, playing a key role in the Ras/Raf/MAPK pathway. 57 The interaction interface identified here coincides with the binding site between RBD and K-Ras, where the positively charged face of RBD interacts with the negatively charged Switch I region and the adjacent a1-helix and b2strand. 58 It may allow RBD to be the target of electrostatically driven nanoparticle-based competing agents to disrupt the Ras/Raf/MAPK pathway in dysregulated cells.
Our work demonstrates that proteins can undergo global axial rotations when residing on the surface of zwitterionic nanoparticles. For the two structurally very similar proteins ubiquitin and RBD, the rotational mode of motion on these zwitterionic nanoparticle is found to be the same, except that the rotational axis realigns in response to the different surface charge distributions of these two proteins (Figure 3). This mode of motion observed here mirrors that of ubiquitin when interacting with liposomes made of POPG molecules, which contain a negatively charged phosphate, prepared as large (LUV) and small (SUV) unilamellar vesicles with mean diameters in the tens of nanometers. 29 Liposomes have long served as models to mimic cellular membranes for studying protein-membrane interactions, [59][60][61] and rotational diffusion of proteins is commonly observed for membranes. 62,63 Importantly, the rotation axis in ubiquitin for POPG liposomes is nearly identical to the one observed here for the ZwiSNPs with a similar rotational correlation time in the low-μs regime. By contrast, zwitterionic LUV and SUV consisting of POPC molecules, which contain a positively charged choline group at its solvent exposed end that is preceded by a negatively charged phosphate, did not show significant 15 N-DR2 modulations along the protein backbone. 29 A mixture of neutral and zwitterionic lipids was used for the preparation of bicelles with non-spherical shape with a 2:1 molar ratio showing a very similar 15 N-DR2 profile as the spherical POPG liposomes. 31 The SNPs covered with zwitterionic SBS molecules used in the present work displayed strong 15 N-DR2 modulations reflecting global protein rotation. SBS has a negatively charged sulfonate group at its periphery and a positively charged dimethylammonio group at the center (Figure 1b). It is therefore remarkable how ubiquitin when interacting with fully synthetic, silicabased zwitterionic nanoparticles shows the same motional behavior as when interacting with POPG-based liposomes. Moreover, the repositioning of the charged face, as present in RBD, fully preserves the motional behavior, except for a realignment in the rotation axis so that it remains orthogonal to the nanoparticle surface. This suggests that the motional modes of globular proteins, such as ubiquitin and RBD, with well-defined patches of positively and negatively charged residues is remarkably uniform during encounters with nanoparticles that have a soft, densely charged surface. Importantly, encounters of ubiquitin and other globular proteins with unmodified and Al 3+ -doped silica nanoparticles are fundamentally different as these proteins are transiently immobilized when interacting with the nanoparticle surface. 24,28 The resulting 15 N-DR2 profile is inconsistent with global motion of ubiquitin on the nanoparticle surface. Instead, the 15 N-DR2 profile directly reports the amount of intramolecular protein dynamics along the protein chain at atomic resolution on ps -µs timescales.
In summary, we examined in detail the interaction dynamics between zwitterionic SBS-coated SNPs and ubiquitin and B-Raf RBD protein using 15