Cellular journey of nanomaterials: Theories, trafficking, and kinetics

Engineered nanomaterials (NMs) are increasingly fabricated in various fields involving consumer goods, waste management, and biomedical applications such as drug delivery, diagnosis, and treatment of pathological conditions. While these NMs are intentionally or unexpectedly in contact with the human body, there are growing concerns about their intracellular journey, especially considering the therapeutic or deleterious effects after they cross the cell membrane. In this review, the cellular journey of NMs including internalization, intracellular trafficking, and deposition/exocytosis is systematically discussed. This work highlights the accumulation of NMs in cells not only depends on the moment of NMs crossing the cell membrane but also at the following trafficking and exocytosis process. A deeper understanding of the cellular journey of NMs implies that an alternative strategy to fabricate specific targeting NMs is to bypass a few pathways of intracellular trafficking to achieve potent therapeutic effects with minimal toxicity. After comprehensively reviewing the cellular journey of NMs, current progress and application scenarios of kinetic models are discussed. Finally, this review focuses on the bottleneck problems and the corresponding solution technologies for studying the cellular journey of NMs. Recent progresses on the cellular journey of NMs provide new insights into the fabrication of biomedical NMs and facilitate technology development for probing the nano‐cell interaction with high temporal‐spatial resolution.


INTRODUCTION
Extensive efforts have been devoted to the development of nanotechnology over the past decades.Numerous engineered nanomaterials (NMs) are fabricated in various fields including medical treatment, electronic industry, energy materials, waste management, and mechanical engineering. [1,2]Specifically, NMs usually exhibit unique biological effects compared with their ionic counterpart.For instance, overuse of antibiotics leads to the developed resistance of bacteria that finally results in the emergence of "superbugs", and nanotechnology is considered the next-generation technique for solving this problem. [3]NMs may execute multiple bactericidal pathways, making it difficult for bacteria to adapt to these therapeutics. [4]In addition, NMs could replace traditional microelement fertilizers in the agriculture field to accelerate plant growth. [5]In recent years, nanocarriers have This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.© 2023 The Authors.Aggregate published by SCUT, AIEI, and John Wiley & Sons Australia, Ltd. been designed to achieve controlled delivery of pesticides. [6]oreover, various NMs have been fabricated for the delivery of anticancer drugs or chemodynamic therapy.[9] Beyond these applications of NMs in the antibacterial, agricultural, and biomedical fields, the adverse effects of NMs on the ecological system and human health have attracted growing attention.In aquatic systems, it has been estimated that the concentration of engineered nanoparticles (NPs) was in the range of ng/L-μg/L, [10,11] while the concentration of natural NPs could even reach mg/L level. [12]15] F I G U R E 1 Schematic illustration for the cell-killing mechanism of a few widely reported biomedical nanomaterials.These nanomaterials are usually designed to target lysosomes, mitochondria, or nuclei to achieve therapeutic effects.
Since most NMs exhibited biological effects after being internalized, a full understanding of the cellular uptake, trafficking, and exocytosis of NMs could maximize the therapeutic effect and minimize their adverse effect, especially in their biomedical application.Shao et al. [16] prepared several aggregation-induced emission-active organic materials with well-controlled morphology, and found that only the spherical shape of nanoaggregates exhibited excellent tumortargeted ability.Wu et al. [17] found that spherical nucleic acid-grafted Au NPs could bypass the lysosomal degradation.Ding et al. [18] reported that blocking the exocytosis of mesoporous silica NPs could efficiently enhance the therapeutic efficacy of delivered drugs.[21] Except for a few specific NMs, most NMs would be internalized, entrapped, and trafficked through vesicles, which is completely different from the fate of micromolecules and ions.24] Even so, the difficulty for NMs to cross the barrier composed of the plasma membrane is always underestimated.Through etching to remove adsorbed NPs on the cell membrane, Malysheva et al. [25] found that only 9%-38% of citrate-capped AgNPs associated with cells were internalized.After the internalization, subsequent intracellular trafficking would affect the subcellular distribution and exocytosis of NMs, while bypassing the acidic organelle (e.g., lysosome) in the cellular journey of NMs may lead to minor cytotoxicity. [26]Overall, the therapeutic effect of NMs largely depends on their accumulation in the cell, which is a combined effect of the internalization and exocytosis process.Therefore, it is crucial to understand the underlying mechanisms of NMs accumulation, which is also important for the kinetic study of NMs.
A few reviews have systematically discussed the cellular journey of NMs including internalization, intracellular trafficking, and exocytosis. [20,27,28]However, these reviews mainly focused on the influence of the physicochemical properties of NPs on their cellular uptake, while factors influencing the trafficking and exocytosis of NPs were rarely discussed.In addition, the kinetic study of NMs is critical for assessing the biological effect of NMs, especially for uncovering the control step of NMs accumulation and distribution.It is also helpful for comparing the potential risks of NMs with different compositions, coatings, sizes, or shapes, which could guide the fabrication of biomedical NMs.However, few reviews discussed the relationship between the kinetic model and the cellular journey of NMs.Therefore, it is uncertain to establish the appropriate kinetic model.This review focuses on the underlying mechanisms of cellular internalization of NPs including theories, intracellular trafficking, and exocytosis, which would benefit the kinetic study of the bio-uptake of NMs.Finally, we discussed the current problems in studying the cellular journey of NMs and the relevant cutting-edge technologies.

THE THEORY FOR CELLULAR INTERNALIZATION OF NMs
Although the cellular internalization of NMs has been uncovered for decades, there is still a great research gap in the internalization mechanisms.Unlike small molecules which could directly penetrate the cell plasma membrane through passive diffusion, NMs could hardly achieve penetration due to their relatively large size and polarity.Usually, the internalized NMs were trapped in a small vesicle composed of lipids. [29,30]Therefore, other factors beyond concentrationdependent diffusion drove the cellular internalization of NMs.[33] The question becomes what drives the NMs adsorption on the cell plasma membrane and which factor leads to the subsequent internalization of adsorbed NMs.
Usually, the cell plasma membrane is composed of lipids and proteins with asymmetrical distributions. [34]The phosphate group in the phospholipids leads to the negative surface potential of the cell membrane, which is favored by the adsorption of positively charged NMs.Apart from the electrostatic force, other forces including the van der Waals force, repulsive hydration force, hydrophobic force, steric force, and bridging force may be encountered in the adsorption of NMs. [35]Bailey et al. [36] reported that polymer like polymethacrylic acid increased the adhesion of gold NPs on the supported lipid bilayer.Melby et al. [37] found that coating of serum protein (i.e., protein corona) greatly reduced the attachment of cationic Au NPs through steric repulsion.More interestingly, they also found that protein corona increased the adsorption of anionic Au NPs.These implied that the nano-membrane interaction always depended on the multiple competition of several forces.Even so, most current studies are based on the artificial supported phospholipid membranes, while the in vivo characterization of nano-membrane interaction is still a great challenge.In the artificial phospholipid bilayer, proteins are hardly embedded.Usually, these proteins embedded in the plasma membrane could serve as specific receptors for binding NMs, or form specific domains which strengthen the rigidity of plasma membrane. [38]Wang et al. reported that artificial phospholipid bilayer might capture the electrostatic interaction with similar responses as plasma membranes, but the integrity of artificial phospholipid bilayer is much more vulnerable than the plasma membrane. [39]fter the adsorption of NMs on the plasma membrane, the adhesion strength needs to go over the energy barrier composed of binding energy and stretching energy to drive the internalization of NMs.Generally, adhesion strength included the nonspecific forces as mentioned above, as well as the specific forces between ligands and receptors.The latter forces are critical for designing nanomedicine, which could deliver drugs to targeted cells/organs.Larger NMs always displayed higher contact area along with higher adhesion strength, but the bending, stretching, and deformation of the membrane also induced the dramatical increase of the energy barrier.[42] For most engineered NMs discharged into the environment, nonspecific adsorption should be the primary driving force for cellular internalization.It was estimated that the low limit of the radius of endocytosed NPs was 5 nm. [21]It is possible that small NMs may agglomerate into large clusters to be internalized.Even so, the low environmental concentration (below ng/L) of engineered NMs [43] implied that the uptake of small NMs should be limited unless these NMs undergo heteroaggregation with other natural particles.Apart from that, the presence of exogenous substances may reshape the nanosurface, giving specific ligand-receptor reactions to NMs.Hayashi et al. [44] reported that coelomic proteins secreted by Eisenia fetida gave recognizable biological identity to Ag NPs, which accelerated the accumulation of NPs.
Despite the advances in the theoretical derivations for NMs internalization, various endocytosis pathways have been uncovered in recent years. [28,45,46]These endocytosis pathways include clathrin-mediated endocytosis, fast endophilin-mediated endocytosis, CLIC/GEEC endocytosis, micropinocytosis, phagocytosis, and caveolae-mediated endocytosis, among others.Only micropinocytosis and phagocytosis endocytosis could internalize the NMs with a diameter over 200 nm, while the limit size for other endocytosis pathways was usually below 100 nm.Even so, NMs with large aspect ratios (e.g., nanowires with hundreds of nm long) could still be internalized through clathrinmediated endocytosis. [47]Until now, it is still difficult to conclude which factor determines the endocytosis pathway of NMs due to the complex nano-membrane interaction, including the NMs type, extracellular medium, cell type, and physicochemical properties of NMs (Table 1).
Table 1 shows that micelle NMs were generally endocytosed by caveolae-mediated pathways, regardless of the cell type and composition, size, and coating of NMs.In addition, clathrin and caveolae-mediated pathways were likely to be the most frequent endocytosis pathways for various NMs.However, it is supposed that clathrin-dependent uptake may be an artifact induced by the protein corona.Many studies reported the abundance of albumin in the protein corona, [48][49][50] and albumin is known to be endocytosed through a clathrin-mediated pathway. [51,52]Cheng et al. [53] reported that protein corona increased the clathrin-dependent uptake of 50 nm Au NPs, thus it is crucial to consider the exposure medium before investigating the endocytosis pathways of NMs.For caveolae-mediated pathway, all the known inhibitors would inevitably change the membrane fluidity, which therefore may interfere with other uptake mechanisms.Caco2 cell was lack of caveolae-mediated pathway, [45] but conflict results were reported (Table 1).
Inhibitors were always employed to study the endocytosis pathways of NMs (Table 1).However, these commercial inhibitors exhibited an 'off-target' effect on several endocytosis pathways.Ho et al. [30] used several inhibitors to study the endocytosis pathway of Au NPs, and found that only m-βCD and dynasore largely inhibited the uptake of NPs, indicating the caveolae-mediated uptake and dynamindependent uptake.However, genetic knockdown of relative genes failed to support these endocytosis pathways.There were possibly other undiscovered routes for NPs internalization, which could be influenced by commercial inhibitors.Recently.Sommi et al. [54] reported that NPs were internalized by a new uptake route through microvilli adhesion, which was influenced by caveolae-related inhibitors.The potential 'off-target' effect of commercial inhibitors was proposed in a recent review. [45]Overall, bio-TEM images and gene knockout/RNA silence are recommended for verifying the endocytosis pathway of NMs.The employment of inhibitors for specific pathways should be carefully checked to minimize the 'off-target' effect.Multiple inhibitors for the same endocytosis pathway may strengthen the solidity of conclusion.In addition, the exposure time for inhibition testing should be minimized as long-term inhibition may interfere with the exocytosis of NMs or induce cytotoxicity, giving the illusion of inhibiting endocytosis.
Apart from the above endocytosis pathways involving the vesicle generation for NMs transport, several specific NMs could directly penetrate the cell plasma membrane by passive diffusion.These NMs always shared tiny sizes (∼5 nm). [77,78]in et al. [79] developed a model to simulate the interactions of Au NPs with model lipid membranes, showing that an increase of positive charge led to the direct penetration of NPs into the cytoplasm while negative NPs could not be internalized.Further study indicated that adsorption of cationic Au NPs generated a nanoscale hole on the model membrane to assist the translocation of small cationic Au NPs. [80]Simi- Caveolae/dynamindependent Inhibitors [55] 75 nm PVP Near sphere RPMI with FBS Earthworm coelomocytes

Micropinocytosis/ clathrin/caveolae
Inhibitors [76] *The frequently-used nanomaterials for toxicity study (e.g., Ag, Cu, Fe, CNT, graphene, and nano/microplastics) and biomedical application (MOF, silica, and micelle) were chosen to summarize their endocytosis pathways into cells (Data collected from wos database).'-' represents that related information could be not found or confirmed in the literature.

F I G U R E 2
Overview of the primary intracellular trafficking and exocytosis pathway of nanomaterials (NMs) in cells.Most NMs would be transferred into the early endosome once internalized, while a few NMs could directly penetrate into the cytoplasm.Then, the endosome sorting would determine whether NMs would be recycled back into the cell membrane, and transported into the late endosome/lysosome, or the Golgi apparatus.The endosomal/lysosomal escape may lead to the leakage of NMs into the cytoplasm, which further targets organelles such as mitochondria and nuclei.The main exocytosis pathways of NMs include endosome recycling, exosome-mediated pathway, secretory pathway through Golgi apparatus, and lysosome exocytosis.The role of lysosome exocytosis in the depuration of NMs is doubtful as discussed in the main text.
larly, Panja et al. [78] found that small Au NPs (<10 nm) with positive charge could directly penetrate the plasma membrane.Interestingly, Verma et al. [77] found that alternating the distribution of anionic and hydrophobic groups on the nano-surface also helped Au NPs to penetrate the plasma membrane directly, but the mechanism is unknown.

THE INTRACELLULAR TRAFFICKING OF NMs
For most internalized NMs, they are trapped in a small vesicle after crossing the plasma membrane (Figure 2).Although NMs could be endocytosed by different pathways, numerous studies uncovered that NMs were first transferred to early endosomes. [45]The early endosome is weakly acidic (pH 6.5), [81] therefore most NMs would remain stable in the early endosome, but a few NMs could be degraded.Cu 2 O and CuO NPs would release ions in the endosomal compartment as they were soluble in the culture medium. [82,83]In the following trafficking, NMs in the endosome were released into the extracellular environment through endosome recycling, or accumulated in multivesicular endosomes followed by transport to lysosomes. [30,84]The latter trafficking is always along with progressive acidification and degradation.In most studies, lysosomes acted as the final storage for NMs. [85,86]The low pH (<5.5) in the lysosome and abundance of hydrolytic enzymes led to the degradation of NMs, even for the inert materials.Kolosnjaj-Tabi et al. [87] studied the 1-year fate of iron oxide-coated Au NPs in mice and demonstrated the degradation of Au NPs within hepatic and splenic lysosomes.Balfourier et al. [88] suggested that ROS production in the lysosome activated the degradation of Au NPs.The lysosomal degradation of other NMs including plastic NPs, [89] carbon nanotubes, [90] and Ag NWs [91] were also reported, but the digestion ability of lysosome appeared to be limited as few studies reported the complete degradation of NMs in the lysosome.
Another key vesicular trafficking pathway in the cells is the bi-directional transport between the trans-Golgi network and endosomes. [92]Deposition of NMs in the Golgi apparatus was accordingly found. [93]The fate of NMs in the Golgi apparatus remained obscure as NMs in the Golgi apparatus were much less than those deposited in the lysosome.Apart from the above vesicular trafficking pathway including the endosome-lysosome and endosome-Golgi apparatus route, some studies reported that NMs were located in the cytoplasm, mitochondria, or nucleus.Many studies reported that single-walled carbon nanotubes were located in the cytoplasm using bio-TEM and fluorescence microscopy, but the mechanism remains unknown. [94,95]Wang et al. [96] recently reported that AgNPs could penetrate mitochondria in HepG2 cells.Zheng et al. [97] investigated the subcellular distribution of quantum dot (QD) with four different coating compositions and found that only GSH-coated QD could selectively accumulate in the mitochondria.They suggested that GSH carriers might facilitate the sequestration of GSH-coated QD in mitochondria.Malatesta et al. [98] found that chitosanbased NPs occurred in both the cytoplasm and nucleus, which was confirmed by fluorescence microscopy and bio-TEM.
However, studies on the subcellular distribution of NMs beyond the deposition in endosome/lysosome are limited, except for specifically designed nanomedicines for targeting organelle.These nanomedicines are always grafted with specific ligands such as TPP for targeting mitochondria [99] and peptides for targeting nuclei. [100]There is still no convincing mechanism to explain how NMs directly target the organelles such as mitochondria or nuclei.The generally accepted mechanism suggests that these randomly distributed NMs should first escape from the endo-lysosomal vesicle, and are then accumulated in the organelles.
Many factors governed the endosomal escape of NMs, but the direct inducement included membrane fusion, osmotic pressure, mechanical strain, and membrane destabilization. [101]Membrane fusion mainly occurs for some viruses and lipid NPs.The proton sponge effect is often described as a classic theory explaining the increase of osmotic pressure.The buffering capacity of NMs inhibited the drop of pH in the endosome, leading to the continued pumping of protons into the endosome.This resulted in the influx of chloride counterions and water molecules that eventually lyse the endosome.The proton sponge effect was proposed much earlier, but there has not yet been any direct proof.Roy et al. [102] first confirmed the theory by designing fluorescent pH-sensor microcapsules.Some specific NMs might swell or aggregate to increase the mechanical strain of the endosomal membrane, which eventually lyses the membrane.This mechanism has been employed in nanomedicine to selectively deliver drugs into the cytoplasm, while NMs are designed to be responsive to the pH or enzyme in the endosome. [103,104]In comparison, membrane destabilization is unusually induced by specific interactions between NMs (e.g.cationic NPs and nanotubes) and endosomal membranes [105,106] or released small molecules from NMs. [107] The lysosomal escape of NMs almost followed the same mechanisms as an endosomal escape.Borkowska et al. [84] reported that mixed-charge AuNPs selectively aggregated in the lysosome of cancerous cells, further leading to lysosome swelling and disruption of the lysosomal membrane.Moreover, the specific environment in the lysosome may accelerate the ROS generation that disrupted the lysosomal membrane, which was reported for Fe-based NPs due to the Fenton reaction. [108,109]As maintaining lysosomal integrity and function is crucial for cellular homeostasis, [110] the lysosomal escape of NMs may lead to cell death, which is different from the consequence of endosomal escape.
Overall, the intracellular trafficking pathways of NMs indicated that endosomes played the most critical role in the intracellular trafficking of NMs (Figure 2).Limited transfer from endosome to lysosome resulted in the deposition of NMs in the endosome or Golgi apparatus.The sorting of cargo in the endosome determined whether their following fate was degraded, transferred, or exported, that is, endosome sorting. [111]From the perspective of NMs trafficking, there is no solid mechanism for critical factors determining the sorting of NMs.Numerous studies pointed out that the size, [112] coating, [113] and charge [114] of NMs could influence the intracellular trafficking of NMs.The shape of NMs may not influence the trafficking of NMs. [115]Although Hinde et al. [116] reported that polymeric NPs with different shapes entered the nucleus with various efficiencies, they also found that all NPs entered into the endosome and showed the same efficiency for endosomal escape.Therefore, ligand and size may be the most critical factors as coating and charge instead of shape changed the ligand on the nanosurface.Ligand decided the ligand-receptor interaction that triggered endosome sorting, while size may influence the endocytosis pathway to affect the following trafficking.Li et al. [117] found that chondroitin sulfate (CS) was accumulated in the Golgi apparatus of hepatic stellate cells, and the coating of CS successfully led to the accumulation of NPs in the Golgi apparatus.However, Tekle et al. [118] found that ligand-coupled QD was arrested within endosomes although ligands themselves targeted the Golgi apparatus.A few studies suggested that the binding strength of ligand-receptor interactions may be also critical. [24,119]A loose association of ligand and receptor would result in the recycling of NMs and receptors back to the cell membrane, while a stable ligand-receptor complex would lead the vesicle to reach the lysosome for degradation.This may explain why most of NPs were located in the lysosome as the high surface area of NPs led to the high binding strength between NPs and receptors.In addition, cell type would also critically determine the intracellular trafficking of NMs, [120] which may depend on the intrinsic difference between endosomal functions and receptors on the plasma membrane.

THE EXOCYTOSIS OF NMs
Besides these intracellular depositions of NMs, NMs could be exported through the fusion of the vesicle with the outer plasma membrane or the non-vesicle-related secretion (Figure 2).The latter exocytosis pathway suggested a direct translocation of NMs from the cytoplasm to the extracellular medium, which has been rarely studied.It was likely that the transport rate of non-vesicle-related secretion was low. [121]Chu et al. [122] reported that silica NPs located in the cytoplasm were hardly exported compared with NPs in the lysosome.In comparison, the vesicle-dependent exocytosis dominated the export procedure, which included early/late endosome recycling, lysosome exocytosis, and secretory pathway from Golgi.Endosomal recycling existed extensively in the cellular system, which was crucial for maintaining the plasma membrane polarity and regulating the residence time of cargo. [123]It could be divided into two kinetic routes: a fast-recycling pathway where cargo was directly back to the plasma membrane, and a slow-recycling pathway where cargo transited through the endocytic recycling compartment before being exported.Specific GTPases were associated with certain recycling pathways.It has been reported that Rab4 was related to the fast-recycling pathway, in which Rab11 and Rab35 mediated the slow recycling directly from early endosomes, and Rab15, 17, 20 and Rab25 controlled trafficking through the apical recycling endosomes to the apical plasma membrane. [124,125]There are only a few studies on endosome recycling using colocalization analysis.Sandin et al. [86] reported that relatively few polystyrene NPs were able to access the endocytic recycling pathways according to the low colocalization between NPs and Rab11.One possible reason was the transient endosomal recycling which was hardly recorded.Previous studies suggested that the recycling of transferrin receptors usually finished in several minutes. [126]Therefore, fast endosomal recycling may lead to the apparent low dependence of NMs on the endocytic recycling pathway.Another possibility was that endosomes mostly played as a transfer stop for NPs deposition, which resulted in the low colocalization between NPs and endosomes with the prolonged exposure time, much less between NPs and recycled endosomes.
Lysosome is always considered a slow turnover compartment. [127]Bourquin et al. [128] reported that mitosis rather than lysosome exocytosis led to the reduction of NPs bioaccumulation.A few studies reported the participation of lysosome exocytosis in the export of NMs.Yanes et al. [129] reported the dominant role of lysosome exocytosis in the export of mesoporous silica NPs, while the regulation of NPs exocytosis could be achieved by inhibiting or accelerating lysosomal exocytosis.Gravely et al. [130] also found that the expulsion of single-walled carbon nanotubes from cells depended on lysosome exocytosis.These inconsistent results may suggest that the lysosome exocytosis pathway was NMs-dependent or cell-dependent.Another possible reason was the 'off-target' effect of the inhibitor.Inhibitors for lysosome exocytosis, like Bafilomycin A1 and nocodazole, could also affect endosome recycling. [131]Generally, strong evidence is still lacking to confirm the role of lysosome exocytosis in the export of NMs.
Trafficking vesicles could deliver NMs into the Golgi body, where NMs could be further exported through some specific secretory pathways.Ding et al. [18] reported that MSNs-PDA NPs were transported out of cancer cells via Rab8/10-and Rab3/26-mediated exocytosis pathways, while the former represented the GLUT4 translocated vesicles, and the latter was related to the classic secretory vesicles.2D nanosheets such as MoS 2 also exhibited the same exocytosis mechanism mediated by secretory pathways from the Golgi apparatus. [132]Apart from these classic exocytosis pathways, there were some newfound pathways such as exosome-mediated pathways.Ho et al. [127] recently reported that some specific NPs with dodecyl loading could upregulate the exocytosis-and vesicle-related genes, then NPs were mainly exported in the sub-100 nm, CD81-enriched exosomes.It should be noted that the exocytosis pathway of NMs was time-dependent, largely different from the endocytosis pathway.It was mainly due to the change in the subcellular distribution of NMs with exposure time, where they usually followed the route from the endosome to the lysosome.NMs would first concentrate in the fast turnover compartment such as the endosome, and after long-term exposure, they should be deposited in the slow turnover compartment like the lysosome.Park et al. found that the amount of exocytosed chitosan NPs from HeLa cells decreased with preexposed time. [133,134]This should be crucially considered in the kinetic study of the depuration of NMs in cells.
As turnover rates of different compartments varied greatly, either the amount or kinetics of exocytosis of NMs largely depended on their subcellular location.Therefore, factors influencing the exocytosis of NMs would partially overlap with factors influencing the subcellular distribution of NMs.Since these factors have been discussed above, only the possible mechanisms determining the exocytosis of NMs in a given compartment are highlighted here.Especially, particle size critically determined the exocytosis of NMs.Chithrani et al. [29] reported that Au NPs with small sizes would be more quickly depurated compared with the larger NPs, and a higher fraction of small AuNPs would be eventually exocytosed.This phenomenon was supported by a few studies. [121,135]One generally accepted explanation was that larger NPs exhibited higher surface area and ligand density leading to stronger binding constant with receptors, which therefore would be hardly released.By adjusting the binding strength between receptors and NPs coated with transferrin, Wu et al. [136] showed that high binding strength increased the retention time of NPs.Apart from the binding strength, the weight or amount of NMs in the vesicles may also influence the exocytosis rate as the transport of vesicles in the cells was dynein-dependent requiring consumption of ATP. [137]The calculated transport velocities of NMs in the vesicle [137,138] were always lower than the reported velocities (0.7 μm/s) for dynein-mediated transport in vitro, which implied that the weight of cargo influenced the transfer rate of the vesicle.It could also explain the low exocytosis rate of large NPs.

KINETIC MODELS DESCRIBING THE INTERNALIZATION, TRAFFICKING, AND EXOCYTOSIS OF NMs
Yi et al. [40] used a mathematical model to simulate that NPs needed dozens of minutes to be fully wrapped.Gu et al. [139] used rod-shaped Au NPs to track their rotational dynamics on the cell membrane under a differential interference contrast microscope.With high-speed photography, the rotational motions of Au nanorods could be recorded in real-time mode, which indicated the wrapping time for a single nanorod.These authors showed that TAT-modified nanorods could be internalized after 4 min on the cell membrane, while PEGmodified nanorods were not internalized.They also found that specific binding could greatly facilitate the wrapping of nanorods, while the wrapping time was reduced to ∼25 s.Due to technological limitations, most works preferred to simulate the wrapping process of NMs using molecular dynamics simulation.Shen et al. [140] reported that the elasticity of NPs could influence their wrapping efficiency, which also depended on the size and shape of NPs.Soft NPs needed more time to be fully internalized.[143] Although great progress in these simulations has been achieved in recent years, they could only simulate the internalization process of NMs.The bioaccumulation of NMs depended on internalization as well as intracellular trafficking and exocytosis.Therefore, kinetic models are accordingly developed to clarify the key factors controlling how fast and how many NMs could be accumulated in cells (Figure 3).
Until now, many kinetic models have been developed to describe the internalization process of NMs (Table 2).Although the introduction of more parameters in the model may come up with better simulation results, most models preferred to simplify some cellular processes of NMs due to the limited dimensions of the dataset.Otherwise, the kinetic data may be overfitted and the fitted parameter would have a much larger uncertainty.The application scenarios of different kinetic models are suggested in Table 2.
The simplest kinetic model is the first-order uptake model, which assumes that the uptake rate of NMs is linearly dependent on the exposure concentration (v u = k u *c e ).Although

F I G U R E 3
The diagram for explaining the foundations of kinetic models.First-order kinetic is always used to establish the relationship between uptake rate and concentration.The extracellular concentration (C 0 ) of nanomaterials (NMs) could be influenced by many factors including aggregation, sedimentation, reduction due to uptake, dissolution, and surface change due to ambient substances (e.g., protein corona).C m , C le , C e , C g , and C l represent amounts of NMs on the cell membrane and in the late endosome, early endosome, Golgi apparatus, and lysosome, respectively.C in represents the intracellular amounts of NMs, which is the sum of C le , C e , C g , and C l .The cellular journey of NMs includes several kinetic processes, which are adsorption, internalization, trafficking, and exocytosis.Several reported kinetic models based on the specific cellular journeys of NMs are listed.It is recommended to simplify some cellular processes to choose the most appropriate kinetic model for different purposes.

Involved cellular process Application scenario Reference
First-order uptake model Internalization low exposure concentration and short-term exposure duration [144]   One-compartment model Internalization and exocytosis low exposure concentration [145]   Two-compartment model; Physiologically based pharmacokinetic model Internalization, transfer, and exocytosis low exposure concentration [146, 147]   Membrane adsorption-internalization model

Membrane adsorption, internalization, and exocytosis
short-term exposure duration [152]   this model does not meet the actual entry pathway of NMs, it has been proven to be a useful empirical model for describing the NMs uptake. [153]It could be inferred that the foundation of the first-order uptake model is based on that the internalization of NMs is independent of each other and no factors (e.g., receptors) limit the internalization of NMs.Rees et al. [31] hypothesized that the internalization of QD depended on the endosome formation on the plasma membrane and the timely interaction between QD and the newly formed endosome.They eventually inferred that the accumulated NPs per cell was a product of concentration and the duration of expo-sure, which well matched with their experimental data and the classic first-order kinetic model.Their model was established based on two principles, including no competition for receptors on the plasma membrane and no exocytosis of NMs.Therefore, the first-order uptake model is suitable for low exposure concentration and short-term exposure duration.A few models are established based on the first-order uptake model, and further consider the subsequent trafficking or exocytosis of NMs.For a one-compartment model, the depuration rate of intracellular NMs is assumed to be linear with the intracellular concentration of NMs (v d = k d * c i ).Two-compartment model and physiologically based pharmacokinetic model (PBPK) are the modified versions of the one-compartment model, while they divide the intracellular NMs into several compartments (c i = c 1 +c 2 +c 3 +…) and NMs in different compartments could be unidirectional or bidirectional transferred. [154,155]These segmented compartments could represent conceptual definitions (e.g., exchangeable portion and unexchangeable portion) and also specific organs or organelles.Therefore, these models are highly programmable.Meanwhile, they could be also applicable to study the exocytosis of NMs, [156] while the exposure concentration of NMs is set as 0. Although these models largely consider the cellular process of NMs, it is suggested that they are only suitable for low exposure concentration as the influences of particle aggregation and consumption of receptors on the internalization of NMs are ignored.
Obviously, these discussed models could not explain the gradual decrease of uptake rate for NMs at high exposure concentration as they assumed that uptake rate was always linearly related to the exposure concentration.Wilhelm et al. established a kinetic model that first considered the membrane adsorption of NMs and consumption of receptors during endocytosis. [148]Therefore, the maximum bioaccumulation of NMs would be limited by receptor density and cell area.They introduced a parameter named 'reactive surface', which represented a part of the plasma membrane that could be internalized by cells to create an intracellular endosome.The 'reactive surface' reflected the receptor density for NMs on the plasma membrane.This model also included the adsorption of NMs on the plasma membrane which was modelized by a Langmuir adsorption, thus the adsorbed and internalized NMs could be distinguished. [32]ilhelm et al. [150] employed the model to investigate the kinetic uptake of NMs in different cell types.They found that Raw macrophages exhibited a much higher ratio of 'reactive surface' compared with the other 13 cell types, which was in accordance with the high internalization ability of the phagocytic cells.Liang et al. [32] used this model to explain the dual role of natural organic matter in the internalization process of NMs.Overall, this model could be useful to compare the membrane adsorption and internalization ability of NMs.However, their model assumed that receptors would be exhausted regardless of external exposure concentration.The intracellular transfer and exocytosis of NMs were also not considered.Therefore, this model is suitable for high exposure concentration with negligible exocytosis for NMs.
Ohta et al. [152] developed a similar model as the Wilhelm et al. work.They also assumed that the membrane adsorption of NMs followed the Langmuir adsorption, and the exocytosis of Si QD was taken into consideration based on the dissociation between Si QD and receptors in the endosome.Their model well predicted the exocytosis of NMs and bioaccumulation of NMs at different exposure concentrations.Jin et al. [151] developed a single-particle tracking technique to calculate the endocytosis rate and exocytosis rate of SWC-NTs.Based on that, they established a model by considering the aggregation of SWCNTs on the cell membrane, which explained the size-dependent internalization of NMs.Overall, although it is a charming prospect to develop a general kinetic model including the membrane adsorption, internalization, intracellular transfer, and exocytosis of NMs, few studies could achieve this goal as the transfer rate and subcellular location of NMs are hard to be directly determined.Therefore, it is suggested to choose the most appropriate kinetic model based on their requirement, although some studies preferred to use previously reported values in the literature for assignments to parameters in the model.

TECHNIQUES FOR PROBING NANO-CELL INTERACTIONS
Although great progress on the cellular internalization, trafficking, and exocytosis of NMs has been achieved in recent years, many questions still baffle.To minimize the adverse effect and maximize the therapeutic effect of biomedical NMs, it is critical to fully understand the nano-cell interaction at the sub-cellular level.However, it is still largely unknown which factor determines the endosomal sorting of NMs and which pattern could explain the kinetic trafficking and exocytosis of NMs in cells.Much effort has been devoted to the internalization mechanisms of NMs, while the biological effects of NMs mostly depend on their targeted organelle and accumulation (i.e., trafficking and exocytosis).A deeper understanding of the cellular journey of NMs requires techniques for probing the nano-cell interactions at the sub-cellular level, which have been well-reviewed in previous work. [20]In general, techniques for studying the cellular journey of NMs could be classified into three groups according to their purposes: 1) tracking the kinetic internalization of NMs; 2) locating the subcellular distribution of NMs with high resolution; and 3) quantifying the internalized NMs accurately.Thus, this review focuses on the corresponding solution technology (Figure 4) and discusses the application scenario for different techniques (Table 3).

Real-time tracking
Until now, multiple live-cell imaging techniques have been developed for tracking the nano-cell interaction in real-time, which include fluorescence microscopy, Raman microscopy, hyperspectral darkfield microscopy, phase contrast microscopy, atomic force microscopy (AFM), and scattered light imaging (Table 3).Fluorescence microscopy may be the most widely used technique for tracking the movement of NMs in the living cell.This technique always suffers disadvantages like photobleaching and cytotoxicity induced by high laser power, thus NMs with high photostability and high quantum yield are usually recommended for long-term tracking of the cellular journey of NMs.Compared with other live-cell imaging techniques, fluorescence microscopy exhibits relatively high resolution, especially for the latest super-resolution microscopy techniques, [166] therefore tracking the movement of single particles becomes possible.Jin et al. [167] calculated the endocytosis and exocytosis rate of SWCNT in NIH-3T3 cells based on single-particle tracking.
Although fluorescence microscopy is a popular method for tracking NMs in vivo, it is a labeling technique that limits its application for various NMs.Even though some NMs could be coated with dyes to emit fluorescence, it is noticed that such coating may reshape the nano-surface, which actually changes the bioaccumulation and toxicity behavior of NMs.

TA B L E 3
Techniques for probing nano-cell interactions at different application scenarios.

NMs
Real-time tracking Fluorescence Microscopy High temporal and spatial resolution; single-particle tracking Labeling techniques that may change the nano-surface to affect the biological effect; fluorescence quenching SWCNT [167] Raman Microscopy Steady signal for long-term tracking; a non-destructive technique for tracking specific organic materials

Low temporal resolution
CNTs, plastic NPs [168,169]   Hyperspectral Darkfield Microscopy Suitable for monitoring the coating/dissolution/aggregation of intracellular NMs; steady signal for long-term tracking; Low temporal resolution Au NPs, Ag NPs [84,170,171] Phase Contrast Microscopy Label-free technique with high temporal and spatial resolution; single-particle tracking and Fe

O
3 NPs [159] Atomic Force Microscopy Super sensitivity for tracking the endocytosis process of single NMs Only suitable for specific NMs; complicated operation Au NPs [158] Scattered  [160] Biodistribution with high resolution (<100 nm) Nano-SIMS

Element-mapping analysis to colocalize
NMs with other species

Tedious sample processing
Ag NPs [161,172] FIB-SEM Superior z-axis resolution to construct a 3D image

Expensive and time-consuming
Au NPs [173] Bio-TEM Super-resolution for NMs and sub-cellular organelles Artifact from sample processing Au NPs [30,127] Super-resolution fluorescence Microscopy Compatible with other probes to determine the distribution and ambient microenvironment of NMs Labeling techniques that may change the nano-surface to affect the biological effect SiO 2 NPs and QD [174,175] Quantification of internalized

Simple operation
The etching/quenching protocol should be carefully verified for accurate determination Au NPs, Ag NPs, SiO 2 NPs [25,163,176] Smart-designed nanomaterials

Simple operation
The internalized NMs may be underestimated polyelectrolyte capsules [165]  F I G U R E 4 Techniques for studying the nano-cell interactions includes i) real-time tracking; ii) biodistribution with high resolution (<100 nm); iii) quantification of the internalized NMs.Reproduced with permission from (A), [157] (B), [158] (C), [159] (D), [160] (E), [161] (F), [162] (G), [127] (H), [163] (I), [164] and (J). [165]man microscopy is an alternative non-destructive technique for tracking some specific organic materials, for example, CNTs.Kang et al. [168] utilized the Raman microscopy to track the dynamic of multiple pristine CNT species in the RAW 264.7 macrophages.Raman microscopy has proved its reliability for long-term tracking NMs, but its low resolution (∼μm) limits their application in the biological system.In addition, the Raman signal for most NMs is weak, thus a few studies specifically labeled NMs using the surface-enhanced Raman scattering technique.Through encapsulating specific small molecules and metal NPs (usually Au NPs) in the targeted NMs, the Raman signal of small molecules could be increased by hundreds of times, which therefore could be used for tracking NMs. [169]yperspectral darkfield microscopy is an ideal technique for tracking noble metal NPs due to their intrinsic high scattering, which effectively differentiates the NPs from the background signal (Figure 4).Moreover, the shifting of the surface plasmon resonance (SPR) peak could indicate the surrounding microenvironment of NPs.This mechanism has been usually used for studying the intracellular aggregation of Au NPs as the aggregation would induce a remarkable red shift to the SPR peak of Au NPs. [84,170]The scatter light imaging technique is also based on the intrinsic scattering of NMs (Figure 4), which was recently developed by Wang et al. [160] Compared with darkfield microscopy, scatter light imaging exhibits higher spatial resolution.This technique could be applied to various NMs and is capable of distinguishing AgNPs with a 10 nm size difference.In addition, Wang et al. [158] developed a promising technique based on AFM to track the internalization process of Au nanorods in the living cell.Au nanorod was connected with the AFM tip through a PEG linker, therefore the rotation and internalization of the nanorod would result in the change of force captured by the AFM tip.Their results indicated that nanorods with a high aspect ratio would undergo intermittent rotation before being internalized.

Biodistribution with high resolution
Numerous techniques have been developed to locate the subcellular distribution of NMs. [20]As the low resolution of imaging techniques may overestimate the overlap between NMs and organelles, there is an increasing need to develop high-resolution techniques.Here, the advanced techniques with high resolution (<100 nm) are mainly discussed.The resolution of traditional fluorescence microscopy is always limited by the diffraction limit.In recent decades, several super-resolution microscopy techniques that could break the diffraction limit, for example, stimulated emission depletion microscopy (STED), and stochastic optical reconstruction microscopy (STORM), have been developed and applied in various fields. [177]Schubbe et al. [174] calculated the agglomeration of silica particles in A549 cells based on the super-resolution provided by STED micrographs.Zwaag et al. reported that the mean size of 80 nm PS NPs observed by STORM well matched its actual size, while the detected value by conventional wide-field microscopy was overestimated (Figure 4). [162]Another advantage of super-resolution fluorescence microscopy is adaptable with various fluorescence probes, which could further determine the surrounding microenvironment (e.g., pH, viscosity, and protein) of NMs.
As an alternative technique, electron microscopy has been widely employed due to its high resolution (∼nm) and adaptability.After fixation, staining, and dehydration of samples, the subcellular distribution of NMs could be easily observed by transmission electron microscopy (TEM).Choi et al. [30,127] investigated the endocytosis and exocytosis mechanism of alkyl-coated Au NPs through bio-TEM.Even so, the sample for bio-TEM is easily contaminated by artifacts during the sample preparation. [178]Some studies would combine EDS to verify the existence of NMs. [179]In addition, it is difficult for bio-TEM to locate NMs with a low electronic density (e.g., plastic NPs and carbon NMs) from the background signal.Compared with TEM, focused ion beam scanning electron microscopy (FIB-SEM) could construct 3D imaging of organelle and NMs in the cell with superior z-axis resolution.This technique utilizes the ion beam to remove the surface of the sample while the SEM carries out high-resolution imaging.Therefore, FIB-SEM is an ideal technique to study nano-organelle interaction, organelle-organelle interaction, and cell-cell interaction.
Apart from the electron microscopy technique, mass spectrometry imaging has also exhibited potent applications.Compared with other mass spectrometry imaging techniques (e.g., laser ablation inductively coupled plasma mass spectrometry), the resolution of Nano-SIMS is greatly enhanced (50 nm).The Nano-SIMS uses an ion source to produce a primary beam of ions that further erode the sample surface and result in the release of secondary ions.These ions could be determined through a mass spectrometer, where tested elements could be identified and measured.In addition, it is feasible to differentiate isotopic elements for Nano-SIMS, therefore NMs could be specifically labeled with an isotope to be distinguished with a background signal or corresponding ion.Shao et al. [161] used labeled Ag 109 NPs and Ag 107 ions to investigate the respective uptake and transport mechanisms, and the Nano-SIMS technique provided visualized evidence of Ag NP-induced autophagy in the oyster gill cells (Figure 4).Through the combination of Nano-SIMS and electron microscopy, their other work integrally described the subcellular distribution and transformation of Ag NPs in the oyster larvae. [172]

Quantification of internalized NMs
For some NPs, less than 10% of NPs associated with cells were internalized. [25]It is also assumed that adsorbed NMs may account for most of the bioaccumulated NMs for algae and bacteria with protective cell walls.Therefore, it is important to quantify the internalized NMs to assess their biomedical or toxicity effect.However, conventional methods such as inductively coupled plasma mass spectrometry and flow cytometry could not distinguish the adsorbed NMs on the plasma membrane and internalized NMs across the membrane.Until now, it is still a technical challenge.Few instruments could automatically distinguish the adsorbed NMs and internalized NMs. [180]or some specific NMs (e.g., nano-Au, nano-Ag, and nano-SiO 2 ), etching could effectively eliminate the adsorbed NMs for quantifying the internalized NMs. [25,163,176]Cell membrane would act as a barrier to protect intracellular particles from dissolution.Similarly, extracellular and intracellular fluorescence NMs could also be distinguished by quenching (Figure 4).Rejman et al. [181] used a high concentration of Trypan Blue solution to quench the fluorescence of extracellular microspheres, thus enabling the determination of internalized fraction.Another alternative is to design smartresponsive NMs.Generally, signals of these NMs would change in the intracellular microenvironment due to their specific pH. [165]However, this method may only detect NMs trapped in the acidic lysosome as pH (∼6.5) in the early endosome is much closer to that in the culture medium.

CONCLUSIONS
The development of cutting-edge techniques with high spatial and temporal resolution has greatly advanced the progress in nano-bio interaction.Also, smart-responsive nano-structures with precisely defined features are continued fabricated to elucidate the nano-bio interaction mechanism and exploit new phenomena.Especially, the development of real-time tracking techniques, high-resolution microscopies, and molecular dynamics simulation has greatly accelerated the knowledge of the internalization mechanism of NMs.
It is now generally accepted that specific ligands on the nano-surface could give NMs targeting ability to specific organelles or cells.For cancer therapy, most of the current studies would prefer the strategy of increasing NMs accumulation in the cancerous cells or inducing cytotoxicity in the specific cancerous microenvironment (e.g., low pH and high H 2 O 2 ).Compared with these strategies, bypassing the trafficking of nanomedicine in normal cells could be an alternative and potent strategy to minimize their adverse effects. [84]However, for the intracellular trafficking and exocytosis of NMs, it is still difficult to develop a universal theory to assign the intracellular pathway of NMs.The composition of surface coatings and binding strength between receptors and ligands may be critical for determining intracellular trafficking and exocytosis.Combining methods including analysis techniques, precisely controlled nano-structure, machine learning, and computer simulation will likely shed light on this area in future work.

A C K N O W L E D G M E N T S
We thank anonymous reviewers for their comments.This study was supported by the Hong Kong Research Grants Council (CityU 11102321 and C6014-20 W) and the Shenzhen Municipal Science and Technology Innovation Commission (JCYJ20210324134000001).

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.

R E F E R E N C E S
(A) Darkfield images of cells incubated with CTAB-rods, NH 2 -PEG-rods, COOH-PEG-rods, CTAB-spheres, NH 2 -PEGspheres, and COOH-PEG-spheres.The spot color displayed the plasmon frequency depending on the particle shape and particle aggregation.(B) Au nanorods were connected with AFM to detect the endocytosis process based on the force tracing technique.(C) Scatter-enhanced phase contrast could simultaneously enable the visualization of cells and ENMs with high temporal and spatial resolution.(D) The label-free NPs could be tracked at a single-particle level using scattered light imaging techniques.(E) Correlative SEM images and NanoSIMS mapping of subcellular localization of the 109 Ag/ 107 Ag ratio, 107 Ag -, and 109 Ag − in different cells.(F) Analysis of polystyrene nanoparticles of different sizes internalized by HeLa cells.Cells were imaged with conventional widefield and STORM microscopy.The quantification through size histogram and particle averaging indicated STORM could accurately determine the actual size of nanoparticles (NPs).(G) TEM images showed the extracellular localization of alkyl 4% -PEG-AuNPs from Kera-308 cells.(H) Etching significantly eliminated the adsorbed Au NPs on the cell membrane according to the fluorescence images and quantification analysis.(I) QSY 9 was used for quenching the fluorescence of DPA-QDs adsorbed onto the chamber surface and the cells.(J) The fluorescence of polyelectrolyte microcapsules responded with pH change, which could be used for distinguishing adsorbed NMs and internalized NMs through flow cytometry.