Protein adsorption changes a broad range of fundamental nanoparticle properties. For example, protein adsorption reduced the cytotoxicity of carbon nanotubes and protected PEGylated quantum dots from the fluorescence quenching that normally occurs in physiological buffers [34, 35]. Typically, the formation of the protein corona increases the hydrodynamic diameter of nanoparticles by 3–35 nm and may cause their aggregation into larger clusters (as shown in Fig. 2) [33, 36, 37]. In addition, they often display a slight negative charge (zeta potential between −7 and −25 mV), regardless of their original charge prior to their interaction with serum [30]. Since the size, shape, and charge of nanoparticles are known to affect the in vitro behavior of nanoparticles, the changes to these properties can drastically change observed uptake patterns.

The clearest observed consequence of serum protein adsorption is the decreased uptake of highly charged particles that is observed after exposure to serum or other proteins [32, 33]. Cationic nanoparticles have been observed to have higher uptake than anionic particles in general, and this effect is retained (although less prominent) after their exposure to serum, even if both particle types have the same zeta potentials. This suggests that the difference in uptake is mediated by the proteins present on the surfaces of the nanoparticles [30]. Additionally, nanoparticles with targeting ligands on their surface may lose their binding specificity when exposed to serum, presumably due to competing mechanisms of uptake mediated by these serum proteins [38]. In fact, so many properties are modified by serum protein adsorption that the protein corona has been shown to be more predictive of the observed cell uptake of nanoparticles than any other characteristic of the construct [32].

The presumed mechanism of off-target accumulation of nanoparticles is the coating of their surface with opsonins, which mediate the uptake of these nanoparticles by professional phagocytes. Over 100 different proteins have been identified that adsorb onto nanoparticles following serum exposure, including well-known opsonins such as complement factors and immunoglobulins [39]. The mechanism employed by these proteins to bind to the nanoparticle surface is not well-understood except that charged and hydrophobic particles seem to accumulate more proteins than neutral hydrophilic particles, particularly those coated with dense layers of polyethylene glycol (PEG) [32]. The subtle changes in the nanoparticle surface chemistry (varying types of anionic functional groups) dramatically change the proteins that adsorb on their surface implying that these mechanisms may be quite complex [32]. For example, the protein adsorption on DNA-coated gold nanoparticles depends on the sequence of DNA displayed on the surface and its secondary structure [40]. Given, this apparent selectivity and complexity, we need to understand which proteins are involved in this process and how their binding can be modulated to provide greater control of the interactions of nanoparticles with cells after serum exposure.

Since the mechanisms of serum protein-mediated nanoparticle-cell interactions are not well-understood, the predominant strategy for reducing these undesirable outcomes has been to passivate the surface of nanoparticles with polyethylene glycol (PEG) moieties [41]. Certain zwitterionic surface ligands are also purported to reduce the binding of serum proteins, but more detailed experiments are required to determine how this strategy compares to PEGylation [42]. Surface passivation reduces both macrophage uptake and increases blood circulation half-life for a variety of nanoparticle types [43–45]. In addition, the loss of specificity due to serum protein adsorption and nanoparticle aggregation is also drastically reduced with appropriate passivation of the surface with PEG [37, 46]. However, proteins continue to adsorb on the nanoparticle surface even when the nanoparticle surface is decorated with a very dense layer of PEG [43]. Further, these proteins significantly affect the cell uptake of these particles [32]. Other limitations of PEG include the difficulty of obtaining a high density of PEG on many particle types as well as the fact that PEG increases the hydrodynamic size of nanoparticles which may make the nanoparticle too large for some applications. Alternatives to PEG, including the zwitterionic ligands discussed above, continue to be investigated for this purpose.

Despite the daunting complexity of the interactions of serum proteins with nanoparticles, some efforts have been made to exploit these interactions in order to simplify the design of nanoparticles. Since the adsorption of proteins is essentially inevitable, Prapainop et al. used a small molecule to induce the binding and misfolding of the protein apolipoprotein B to trigger the specific receptor-mediated uptake of these particles in a macrophage cell line [47]. Although nanoparticles are efficiently taken up by macrophages in the absence of specific targeting ligands, this strategy has the potential to be a new direction for creating particles that are easier to synthesize (due to the greater stability of small molecules compared to protein-based targeting ligands) and that display specific targeting in the presence of serum proteins. Hamad-Schifferli’s group has controlled the deposition of serum proteins on gold nanorods to create agglomerate protein corona nanoparticles with high payload capacities [48]. These particles were capable of encapsulating both small molecules (Doxorubicin) and DNA that slowly leach over multiple days; the release of such cargo can also be triggered using the near-IR excitation of the gold nanorods. Unfortunately, both approaches are still preliminary, and it remains to be seen if either strategy can be used for directing the specific uptake of nanoparticles in tumors. This is likely to be the focus of future research.

Despite the increased popularity of research on nanoparticle serum protein adsorption in recent years, it remains unclear how this issue should be addressed. New strategies, in addition to PEGylation, are needed to mitigate the effects of serum protein adsorption on nanoparticle behavior. Nevertheless, the research presented here can be applied in many studies involving nanoparticles. To begin with, researchers should study the effects of nanoparticle-cell interactions in the presence of serum since nanoparticle behavior is known to be dramatically altered by serum protein adsorption [49]. Secondly, nanoparticles that undergo aggregation or large changes in size and shape due to protein adsorption should be avoided in place of more stable particles to improve the reproducibility of observed interactions with cells. Lastly, a greater focus on exploiting the unique behavior of protein adsorption and assembly on nanoparticle surfaces should be pursued since these proteins likely mediate many of the interactions that nanoparticles experience.

Nanoparticle size determines the curvature of the contact surface between the nanoparticle and the cell membrane and thus dictates the nature of the interaction between these two species. Since membrane curvature has been shown to initiate clathrin-mediated endocytosis as well as phagocytosis it is expected that cell uptake would be dramatically affected by nanoparticle geometry [12, 14]. Indeed, for polymer nanoparticles (PEG-acrylate derivatives) Gratton et al. showed that the efficiency of uptake by HeLa cells (cervical adenocarcinoma) decreased as the particle diameters increased from 200 nm to 5 μm among particles with aspect ratios of 1 [50]. Several groups also explored the uptake of spherical nanoparticles smaller than 100 nm in diameter. Chithrani et al. showed that gold nanoparticles ~50 nm in diameter exhibited greater levels of uptake in HeLa cells than other nanoparticles between 15 and 100 nm [51]. Lu et al. and Varela et al. also obtained similar results using mesoporous silica nanoparticles (maximum uptake at 50 nm diameter) and polystyrene nanoparticles (maximum uptake at 40 nm diameter) despite employing varying cell types [52, 53]. Recently, some of these results have been criticized because increasing particle size also favors sedimentation over diffusion thereby increasing the uptake of gold particles that are 50 nm in diameter or larger [54]. Sedimentation also affects polymer particles (PEG-diacrylate) but becomes significant only for particles with diameters greater than 325 nm (disc shaped with height of 100 nm) due to their reduced density compared to gold nanoparticles [55]. Presumably sedimentation applies to many other particles and may complicate the interpretation of current results but this has not been characterized for most particle types. However, the overall trends observed across a variety of particle types still suggest that particles with diameters around 40–50 nm exhibit more rapid endocytosis than particles of other sizes.

The aspect ratios of nanoparticles also affect the contact areas between the particles and the cells and therefore affect endocytic rates across a variety of particle types. This was illustrated by Gratton et al., who observed higher uptake of polymer nanorods with dimensions of 450 × 150 × 150 nm compared to particles with dimensions of 200 × 200 × 200 nm in HeLa cells [50]. However, Agarwal et al. recently reported that polymer nanodiscs (PEG-diacrylate particles) [dimensions of 220 × 220 × 100 nm and 325 × 325 × 100 nm] were taken up faster than nanorods of the same material [dimensions of 400 × 100 × 100 nm and 800 × 100 × 100 nm] by HeLa, HEK293, HUVEC, and BMDC cells [55]. One difference between these studies is that the surfaces of the nanoparticles used by Gratton et al. were positively charged and charge also affects the interactions of nanoparticles with cells. To address these concerns, more studies are needed to decouple the effects of aspect ratio, particle size, core material composition, and surface capping ligands in order to establish design parameters that can be employed to control the endocytic rates of nanoparticles. These results will be promising since aspect ratio allows a second degree of freedom (in addition to particle size) as a means to design nanoparticles and manipulate their behavior in vivo.

Another major challenge to nanoparticle design for in vivo applications is to manipulate the relative endocytic rates between cell types such that nanoparticle accumulation occurs more rapidly in target cell types and more slowly in off-target cell types. By decorating the nanoparticle surface with peptides, proteins, aptamers, or small molecules that specifically bind target cell types, the binding affinity and kinetics of binding of nanoparticles to the cell surface can be increased. Often, the binding affinity of the nanoparticle exceeds that of the free ligand. Wiley et al. modified the surface of gold nanoparticles with varying densities of transferrin and observed that the dissociation constant (K _d ) of the nanoparticle from Neuro2A cells was between 4.9 and 0.014 nM, indicating that the nanoparticles bound these cells much more strongly than free transferrin (K _d of 144 nM) or nanoparticles without transferrin (K _d too large to measure reliably) [56]. Because this enhanced binding is due to the multivalent interaction to multiple cellular ligands, such functionalization may also directly trigger endocytosis by crosslinking surface ligands and initiating the assembly of a clathrin pit [57, 58].