Over the past few decades, the interaction of nanomaterials (NMs) with biomolecules has become one of the most exciting areas of applied and basic research in bio- and material science. The rapid use of nanobiomaterials in different areas of industries, technologies, and medicine is raising obvious questions about the safe implementation of its prescribed applications [1]. The conjugation of biomolecules with NMs not only introduces biocompatible functionalities into these nanoclusters but also leads to stabilization of the complex (Figure 1). Again, proteins are functionally relevant biomolecules and attachment of NM to protein has found vital applications in catalysis, imaging, and understanding the structural modification of proteins and control of protein catalytic activity with external fields, e.g., magnetic [2]. NM-protein conjugates are not limited to biological applications but rather have outreached fields like material sciences and physical sciences with device design [2] such as biofuel cells and nanosensors.

Compared to bulk materials, the surfaces of NMs have higher free energy owing to their unique electronic distributions, which enable NM surfaces to adsorb proteins preferentially [3]. Evaluation of this change in free energy that occurs during protein adsorption provides the fundamental basis for analyzing the nature of the driving force for the reaction. Free energy change (∆G) can either be entropic or enthalpic or both. While entropic change (∆S) associated with protein adsorption is due to either the dehydration of uncharged surfaces or the removal of the electric double layer from charged surfaces, enthalpic changes (∆H) are mostly due to hydrogen bonding or the formation of donor–acceptor coordination bonds between the protein and the adsorbent surface [3]. The layers consisting of bound or adsorbed proteins around NPs are allegorically called “protein corona” (PC) [4, 5] for their structural resemblance to the original cosmological term, corona, meaning the rarefied gaseous envelope of the sun and other stars. This PC-bilayer then grows rapidly to introduce the bound NM into biological fluid and participate in the exchange of proteins based on their abundance and affinity towards particular NM [5]. One of the main purposes of forming this “layered delivery system” is to prevent any undue cell damage by some NMs (carbon nanotubes, graphene, etc.) [6]. The formation of PC is closely associated with inducing protein conformational changes which have both useful and deleterious effects. Given all these, one of the primary objectives of this review is to recapitulate the physical chemistry of the process as well as the biological impact of PC formation associated with the protein conformational changes, to discuss the current state of the art, and to point out the limitations and highlight the future scope.

There are several experimental techniques to capture changes in size and shape of NMs upon protein adsorption (Figure 2). They include Transmission and Scanning Electron Microscopy (TEM, SEM), Atomic Force Microscopy (AFM), Nanoparticle Tracking Analysis (NTA), Dynamic Light Scattering (DLS), Fluorescence Correlation Spectroscopy (FCS), and Differential Centrifugal Sedimentation (DCS). DLS and FCS determine the thickness of the protein shell by comparison of the hydrodynamic diameter of NPs, which can be measured from their diffusion coefficients (Supplementary) by the Stokes-Einstein equation, before and after the adsorption. Again, since the adsorption of proteins on NPs usually leads to quenching, the changes in NP-fluorescence are also indicative of the formation of a PC [9]. Sometimes, DLS is particularly difficult to apply when NPs are in the size range of proteins (say, <10 nm), as unbound proteins may meddle with. In such troubleshoots, the community has come up with Dielectrophoretic DLS (DDLS) [10] as the alternative where the unwanted signals from unbound proteins are completely suppressed due to the incubation of the PC with complex matrices [9], while on the other end the dissociation coefficients of the NP-protein interactions can be determined using FCS.

The changes in particle size are accompanied by a concomitant change in electrokinetic potential (ζ) during the corona formation (Supplementary). As a matter of fact, ζ-potential is the only accessible resource to characterize the double layer properties of colloidal systems [11]. The colloidal dispersion exhibits acceptable physical stability when the ζ-potential of the particles falls outside the range of −15 to +15 mV. The increased magnitude in ζ-potential, in turn, increases the repulsive forces between particles preventing their aggregation. Again, the protein concentration has a direct effect on the hydrodynamic radius of the interacting NPs independent of the ζ-potential [12], as revealed by Monopoli et al., in SiO₂ NPs [2, 9].

One of the major conclusions from virtually all experimental findings was that the PC formation is associated with structural (or conformational) changes in the adsorbed proteins. However, all such studies suffer in their (atomistic) details due to the lack of enough high-resolution experimental structural techniques to be implemented to probe the induced conformational changes. FT-IR is a very sensitive technique to capture the conformational changes as reflected in the vibrational bands corresponding to the amide (or peptide) bonds, also used to probe the covalent adduct formation of Cysteines in the PCs. Again, CD can indirectly capture the structural changes, as revealed in the case of the human iron-protein transferrin upon interacting with SPIONs [13]. Another sensitive indicator is the change in intrinsic fluorescence of a buried tryptophan upon getting exposed to the polar solvent. This can be probed by bis-ANS binding, a common marker for protein unfolding [14]. An example of an effective use of Trp-fluorescence has been demonstrated in probing the CuNP-BSA interaction [15].

Charged molecular surfaces lead to the formation of an electric double layer when placed in aqueous solutions. The 1st layer of ions gets directly adsorbed on the surface through electrostatic forces leading to the formation of salt-bridges [15], hydrogen bonds [16], charge-dipole, and dipole-dipole interactions while the 2nd layer remains composed of free ions adsorbed to the surface by weaker electrostatic forces (Table 1). Principal forces stabilizing protein corona originate from two fundamental and distinct sources: (i) repulsive forces between interacting proteins help keep nanoparticles in solution (e.g., BSA is used to stabilize gold nanoparticles). However, attractive or depletion forces may dominate at times between proteins so as to drive aggregation of the particles, which further influence change of the size and density of the nanoprotein complexes. (ii) Owing to their unique electronic distributions, the high free energy at the surfaces of nanoparticles (compared to bulk materials) results in absorption of various molecules, most notably proteins. The surface charge on the proteins also triggers electrostatic interactions (hydrogen bonds, salt-bridges, charge-dipole, and dipole-dipole interactions) influencing protein absorption on nanosurfaces. This layer is called the diffused layer as it generally moves in the fluid. The formation of this charged double layer on the NM surface induces local VDW interactions [17] [best explained by quantum electrodynamics [18]], yet another component is known to affect the adsorbed protein’s conformations. Sometimes induced charges affect the adsorption of ions onto gold surfaces in the gas phase at a strength similar to chemical bonds while ions and charged peptides in solution are influenced at a strength similar to intermolecular bonds [19]. Hydrogen bonding may further be mediated between polar amino acid residues (e.g., serine, threonine, asparagine, and glutamine) of the adsorbed protein and exposed polar groups (e.g., hydroxyl groups on oxidized metals) on the NM surface [20]. Another major component influencing PC formation and the related conformational changes is the hydrophobic interaction [21] which originates due to the strong antipathy of the nonpolar groups when exposed to a polar environment. Hydrophobic interactions bring about the largest structural changes to the adsorbed proteins as the inner hydrophobic regions of the protein get unfavourably exposed to the solvent. Such an unfavourable exchange was encountered in α-lactalbumin, a largely hydrophobic protein, upon interacting with ZnO-NP leading to the deformation of its secondary structures [22]. A similar event has also been reported in case of CNT which when incorporated into the hydrophobic core leads to the destruction of the protein functionality by blocking the active site [23, 24].

PC formation is thought to involve an intrinsic “memory function” [12] which appears to depend on the particular NP interacting with the particular protein sequence. One view addressing the modus operandii of this “memory function” hypothesizes that the first protein to get adsorbed on a given NP surface holds the key to develop the “memory” as it is the first protein which has the largest abundance to PC [12, 25]. The “memory” is often developed by means of a conformational change induced on this first adsorbed protein, which, in turn, influences the subsequent (specific) protein-protein interactions constrained by the modified growing surface of the NP-PC assembly [12, 25]. Similarly mechanisms invoke an NP-induced conformational change on the first adsorbed protein; an NP-PC assembly can also potentially get access to otherwise inaccessible endogenous and exogenous substances by biomimicry [12, 26]. This feature has in fact been used as a strategy for selective targeting of drugs, say, by cloaking NPs on the membranes of certain immune cells, thereby providing a shielding effect against an undesired antigen rejection [12, 27].

One of the key attributes of NMs per se is their “unique” surface properties, arising due to the relative distribution of core and surface electrons, leading to a high surface area-to-volume ratio in comparison with bulk material [28]. In fact, this serves as one of the most fundamental points of discrimination between NMs and bulk matters and forms the basis of the widespread biomedical applications of NMs [29]. Again, one way to look at the protein-NM interface is to treat the interaction between two surfaces of distinctly different origin and scale. To that end, the protein surface has been traditionally well defined and categorized as solvent accessible surfaces [30], VDW surfaces [31, 32], and molecular Conolly surfaces [33] with commonality as well as uniqueness among the different surfaces. So to say, the solvent accessible surface in proteins has found a wide array of applications over the years in the structure dependent functional characterization while the VDW and molecular Connolly surfaces have been particularly useful to study internal packing within folded globules, extending the exercise to facilitate protein design [34]. Again, by definition the molecular surface being smoother than a corresponding VDW surface appropriately tunes the numerical ranges of the related shape descriptors (Sc: shape correlation) as applicable to the packing studies [31, 32, 34]. On the other hand, the NM surfaces of diverse origins can be adequately described by morphological and physicochemical descriptors such as surface curvature, surface energy, charge, and topography [35]. In such a surface view of the protein-NM interaction, however, one has to consider the difference between the scales at which the two surfaces (i.e., protein and NM surfaces) operate (Figure 3), as is evident in the PC model. In other words, NM surface provides a relatively flat large anchoring particulate mesh (receptor) wherein protein surfaces (ligand) can potentially adhere depending on the compatibility of surface charge, shape, and possible solvation terms. This scale-difference in the two interacting partners makes such interaction analogous to, say, the docking of elongated helical peptides [36] (ligand) onto large protein surfaces (receptor), wherein complementarity in shape and surface electrostatic potential may apply as critical constraints, however, intuitively to a less degree than what has been found in both protein-protein interactions [37, 38] as well as in the folding of globular proteins to be envisaged as the docking of their interior components [39]. Generally, in a biomolecular interaction, if large molecular surfaces of proportionately equivalent size need to associate specifically, the constraints in shape and electrostatics are found to be higher as the surfaces need to be carefully tailored over large patches (say ~1600 Ų for protein-protein interfaces [40] buried upon association. An equivalent comprehensive characterization of the NM-protein interface in terms of shape and electrostatic complementarities demands more structural as well as computational studies of the atomistic scale.

One of the significant challenges pertaining to the study of NM-protein interactions is the thermodynamic characterization of the induced protein conformational changes, whether reversible or irreversible (Figure 5). This is particularly challenging given the lack of experimental structural data and appealing as protein function is directly linked to its conformation. Not all NPs bring about the same changes on the same protein and factors controlling the surrounding chemical environment like pH, further influencing the changes. The conformational changes also vary as a function of the binding site. For example, BSA upon adsorption to gold NPs displays more stable and rigid conformational changes (irreversible) at both secondary and tertiary structural level while more flexible (reversible) changes are encountered for the same protein adsorbed to the boundary surface of gold NPs [20]. Again, virtually no conformational change was detected on BSA upon adsorption to C60 fullerene while only marginal changes were obtained upon binding to ZnO-NPs [51]. Furthermore, a dose dependence on the induced changes was also found in the case of BSA [52]. On the other hand, irreversible changes were detected in albumin and lacto peroxidase upon adsorption to silica NP while RNAse and lysozyme retained their native structure when adsorbed to the same NP [53]. Again, in systems like CuO-BSA, Fe₂O₃-Hb, and SWNT-Lyz no remarkable changes were detected [23, 24, 54].

The binding of proteins onto NMs can result in NM aggregation along with protein aggregation. It has been shown that protein self-assembly in the presence or absence of NMs is highly controlled not only by proteins’ properties, like aggregation rate and inner stability, but also by the physiochemical properties of NMs, like shape, size, and NMs/protein ratio. Upon interaction with NMs, some proteins predominantly consist of ß-sheet structure and tend to self-assemble inside, whereas others get aggregated randomly [8]. For example, lysozyme-gold NPs formed assembly when they were mixed and underwent protein conformational change on gold NPs. The surface bound lysozyme interacted with other proteins in the solution followed by gold NP aggregation. Positively charged part of protein gets exposed in solution when it adsorbed onto a NM surface. This phenomenon depends on the 3D structure and the position of lysine and arginine residues [60]. Lowering the pH below the pI value of a protein can encourage NMs aggregation. 6-mercaptopurin, citrate, ω-mercaptoundecanoic acid, surface stabilizing molecules, etc. are known to influence the NMs aggregation as they have different pKa values from the NMs surface. Their dissociation from the surface affects the aggregation of NMs after partial adsorption of protein onto NMs surface. Protein mixed with NPs usually plays crucial role in the cluster formation [8, 61]. Amyloidogenic peptides influence protein-NP aggregation by interacting with other peptides and formation of nuclei for protein-NP aggregation triggered through the combination of long range electrostatic repulsion and short-range attraction between proteins bound to NPs [8, 61].

The NP-induced reversible conformational changes also affect the downstream protein-protein interaction cascade, which, in turn, influences cellular signaling and eventually transcription. Also, such changes can increase the accessibility of the active sites of certain enzyme(s) towards NPs [8]. To be more specific, NP-protein binding has two major consequences. Generally, irreversible changes to the protein lead to a reducing bioactivity of the concerned protein while changes that level up to scale the protein energy in a controlled way favour the overall functionality of the system. For instance, Si-NP was able to induce a molten globule like structural change in carbonic anhydrase, while upon removal of the NPs three intermediate native-like conformations were obtained and the catalytic activity of the enzyme was retained [60]. It is known that NP-protein corona (PC) rather than bare NP determines the NP-Cell interactions, including endocytic pathway and biological responses. The experiments for endocytosis specify the fact that different endocytic pathways might be responsible for the alternative roles of PC in the interaction of sized NPs with different cell lines. The therapeutic and toxicological profiles of a NP might be altered by altering cell uptake as well as cellular response through NP-PC incorporation. It was observed that, in the presence or absence of a PC, silica NP showed different uptake efficiencies. AuNP particle size and surface ligand grafting density can also affect the adsorption of serum proteins and cell uptake in macrophages [62]. Covalently bound horse radish peroxidase, substilisin, Carlsberg, etc. were also shown to retain their catalytic activity and native structure after removal of SWNT from the bound form [60]. One of rare but important factors on the composition of PC is the changes in incubation temperature of the NPs. For instance, plasmonic heat induction changes composition of hard corona adsorbed on gold nanorods which was further supported by computational molecular modeling studies [61].

Till now the structural modification of proteins upon NMs interaction have been shown by most of the literature in in-vitro studies. Still there is not much evidence of protein conformational changes due to NMs interaction after nanobiosystem is taken up by living organisms. So, the obvious question is what would be the fate of protein as well as NMs in an in-vivo study?

Most of the studies were based on the adsorption of protein onto NMs to discuss the effect. But apart from adsorption, there are other modes of NM-protein interaction, like cross linking, entrapment, encapsulation, self-assembly, etc. So, the effect on protein structures for these alternative modus operandii of nanoprotein interactions is another essential query that needs to be addressed adequately.

There are very few instances where NM induced enzymatic activity due to interaction. If nanomaterials can stabilize a protein structure, they may also potentially hyperactivate the enzyme by changing its secondary structure/modulating its folding to expose more catalytic sites. So, what are the other possibilities of functional gain of a protein due to NM interaction? Apart from aggregation, what other structural and functional modulations may be experienced in NMs due to protein interaction? And lastly, is it possible to use engineered protein or surface stabilizing agents for NMs to minimise the structural damage and aggregation?

Characterization and analysis of proteins bound to NM surface are an important step to understand the effect of NM-PC for biological and medical applications. Research has thus characterized the physical chemistry of PC formation and associated protein conformation. One of the insightful ideas was the ‘memory of PC formation.’ Again, the potential consequences of NM-protein interaction are the alteration of biomolecular structure. Conformational change may affect protein function, e.g., enzymatic activity. However, the changes can also be utilized to good effects as exemplified in the case of serving as a thermostable protective layer for certain mesophilic enzymes. Aggregation of NP can potentially change its interaction pattern affecting the normal cell homeostasis. Interestingly, most studies about NM-protein interactions are done in-vitro. So, understanding the behavior of NM-protein complexes in-vivo remains yet to be explored, which is a challenging task. Another current lacuna in the field is the lack of enough high-resolution structural techniques which throws an open challenge to the computational structural biology community to model the interaction and dynamics such large and complex macromolecular assemblies. But in some current simulation methods, advances in the understanding of chemical bonding, in the development of force fields, and in the development of chemically realistic models are described [69]. For example, an introduction of the silica models or hydroxyapatite model provides full validation for interfaces [70, 71]. A successful endeavour towards this direction will also enable the study of the surface architecture of NM-PC complexes in atomistic details. This is of utmost importance to decipher the scope of the most popular concept of ‘complementarity in biomolecular recognition’ in such complexes. Also, from a computational end, it cannot be left overlooked that there is a significant scope of the ever-so-successful machine learning approach [72, 73] to be implemented to predict the yet unexplored NM-protein interactions (Figure 6) once enough high-resolution structural and binding data are available by either experimental or computational means.

The implementation of the “artificial intelligence” approach to the nanoprotein system is currently only hypothetical, however, once grown enough matured, it appears perfectly rational to imagine instinctive prediction of nanosurfaces, potentially facilitating beneficial interactions, with a concomitant estimation of the corresponding risk profiles. The machine learning approach has been extremely successful across various branches of informatics and there is no reason to believe that it won’t be successful here as well. Overall, emergence of the computational structural field in probing NM-protein complexes should definitely foster future translational applications of nontechnology in the human body as the subject is heavily dependent on NM-protein interactions.