Rapid transport of germ-mimetic nanoparticles with dual conformational polyethylene glycol chains in biological tissues

See allHide authors and affiliations

Science Advances  07 Feb 2020:
Vol. 6, no. 6, eaay9937
DOI: 10.1126/sciadv.aay9937


Polyethylene glycols (PEGs) can improve the diffusivity of nanoparticles (NPs) in biological hydrogels, while extended PEG chains severely impede cellular uptake of NPs. Inspired by invasive germs with flagellum-driven mucus-penetrating and fimbriae-mediated epithelium-adhering abilities, we developed germ-mimetic NPs (GMNPs) to overcome multiple barriers in mucosal and tumor tissues. In vitro studies and computational simulations revealed that the tip-specific extended PEG chains on GMNP functioned similarly to flagella, facilitating GMNP diffusion (up to 83.0-fold faster than their counterparts). Meanwhile, the packed PEG chains on the bodies of GMNP mediated strong adhesive interactions with cells similarly to the fimbriae, preserving cellular uptake efficiency. The in vivo results proved the superior tumor permeability and improved oral bioavailability provided by the GMNP (21.9-fold over administration of crystalline drugs). These findings offer useful guidelines for the rational design of NPs by manipulating surface polymer conformation to realize multiple functions and to enhance delivery efficacy.


With the evolution of living organisms, the tunable conformations of molecules play an indispensable role in various biological activities, and a similar principle can be applied in synthetic polymers-decorated systems (1). In the past decades, surface modification by polymers has become a useful and attractive tool in biomedical applications such as drug delivery system engineering (2). Studies have addressed that surface properties, including hydrophilicity and mechanics, are substantially correlated with the grafting density of polymers (3, 4). In particular, tethered polymer chains undergo a conformational transition from the “mushroom” regime at low density to the “brush” conformation at high density. This feature endows even single-component polymers with controllability of the properties of nanoparticles (NPs) (2) and their interactions with biological systems, such as environmentally responsive molecular valves for controlled release (5, 6) and protection for NPs against clearance (7).

Polyethylene glycol (PEG), for instance, has been widely adopted to modify NPs to achieve efficient delivery (8, 9). Brush-like PEGs endow NPs with hydrophilic and “nonfouling” surfaces that can resist protein adsorption, resulting in prolonged circulation time and rapid penetration into biological hydrogels (810). By contrast, mushroom-like chains enable hydrophobic interactions with proteins that are preferable for cellular adhesion (11, 12). Unfortunately, drug vehicles are often required to efficiently pass through human mucosa or tumor tissues that typically contain biological hydrogels (mucus and tumor matrices) and cellular barriers (13). PEGylated NPs encounter the problem that brush-like chains can facilitate the diffusion of NPs in hydrogels while simultaneously causing a pronounced reduction in cellular uptake (14), consequently impairing their drug delivery efficiency. Studies have suggested an optimal chain length for balancing the effects (15); however, this compromise may not fully use the potency of PEGylated NPs, and the actual polymer conformation, even dual conformational chains on a single particle, should be taken into consideration.

Taking inspirations from nature, we notice the interesting occurrence of polymer-like units on germs related to their capability of colonization capability. Some germs have flagellum at their tips with the dynamic properties of “run” and “tumble” that facilitate their rapid transportation in highly viscous medium (16, 17), and in particular, pathogens including Helicobacter pylori penetrate the porous barrier of human mucus layer with the function of these long flagellum (17, 18). On the other hand, short fimbriae play an essential role in mediating the strong nonspecific adhesion of germs to the epithelium during invasion (19). Since these architectures and functions are the result of long-term evolution, they may provide a strong indication on the design of polymer-decorated drug delivery systems. Therefore, we conceived germ-mimetic NPs (GMNPs) with dual conformational PEG chains that might better solve the aforementioned contradictory problems. Although anisotropic NPs such as Janus particles have been investigated to realize multiple functions including targeted delivery, controlled release, and combined therapy (20, 21), NPs with dual polymer conformations for overcoming complex biological barriers have not been proposed.

In this study, taking the advantage of anisotropic mesoporous silica nanorods (NRs) (22), we successfully fabricated anisotropic PEGylated NR, designated GMNP, by choosing an adequate PEG content. Atomic force microscopy (AFM) under fluid conditions revealed that GMNP displayed an anisotropic hydrated PEG corona (HPEGC) with extended PEG chains on the tips and packed mushroom–like chains on the rod body. In vitro results, together with molecular dynamics (MD) simulations, demonstrated that the formation of long PEG chains on the tips help the GMNP escape from the adhesive forces in the biological hydrogels and penetrated the porous meshwork effectively. This was similar to the penetration process of germs driven by the tip-specific long flagellum, and the packed PEG chains on the rod bodies induced strong adhesive interactions with the cell membrane and efficient cellular uptake, mimicking the function of epithelium-adhering fimbriae of germs. Combining the two germ-like characteristics, the GMNP was proved to improve the oral delivery efficiency in the intestinal mucosa and effectively penetrate into the tumor xenografts of the nude mice. These findings experimentally and theoretically advance the knowledge of the rational design of polymer-modified drug vehicles and the ability to manipulate their surface properties for effectively overcoming complex biological barriers.


Characterization of NPs with different surface structures

Three types of mesoporous silica NPs (MSNs) were fabricated: one NR and two nanospheres (NSs) as isotropic counterparts. The size of the NR was approximately 78 nm (short axis) and 220 nm (long axis), while for the two types of NS, the small NSs (SNSs) were about 80 nm, which was similar to the short axis of the NR, and the large NSs (LNSs) were about 140 nm and had a hydrodynamic diameter of about 200 nm, similar to that of the NR (Fig. 1, A and B). Their zeta potentials were similarly negative (Fig. 1C). According to the liquid crystal template–based mechanism, NRs can be inferred to exclusively have porous structures on their tips (22). To verify this anisotropy, we obtained high-resolution electron microscopy images of the NR, and as expected, the NR showed a tip-specific porous structure while the spherical NPs had a uniform distribution of pores on their surfaces (Fig. 1D).

Fig. 1 Characterization of NPs and detection of the anisotropy of NR.

(A) Scanning electron microscopy (SEM) images of NR with a size of about 78 nm by 220 nm, SNS with a size of about 80 nm, and LNS with a size of about 140 nm. (B) Hydrodynamic diameter of NPs. (C) Zeta potentials of NPs. (D) Porous structures of NPs in high-resolution SEM images. Yellow arrows indicate the tip-specific porous structure on the NR. (E) Schematic illustration of detecting hydroxyl densities via the friction between the ─OH groups and COOH-modified AFM probe. (F) Height and adhesion mapping via AFM. Top left, NR body; top right, NR tip; bottom left, SNS; bottom right, LNS. The regions on the rod body showed significantly lower adhesion than the tip regions and the surface of the spherical NPs. (G) Quantification of hydroxyl groups on MSN via thermogravimetric analysis (TGA). The yellow and blue areas indicate the weight loss due to the dehydroxylation and condensation of the hydroxyl groups on the silica surface, respectively. (H) Adhesion force and the inferred hydroxyl density of different regions on MSN. Representative images are presented. Data are means ± standard error of mean. *P < 0.05 and ***P < 0.001, one-way analysis of variance (ANOVA) and Bonferroni’s test.

Since the silica backbone is composed of hydroxyl groups (silanol groups) that are highly associated with surface functionalization (23), we assumed that there could be differences in the density of hydroxyl groups on the rod tip and rod body. To test this assumption, we modified the AFM probes with carboxyl groups to generate frictions with the hydroxyl groups that reflected the local hydroxyl densities (Fig. 1E) (24). The results of AFM adhesion mapping showed that higher friction was detected on the tips of the NRs than on their bodies, implying that the tips had a higher hydroxyl density (Fig. 1F and fig. S1A). This anisotropy might be due to the different surface areas resulting from the tip-specific porous structures or from irregular packing and incomplete condensation during fabrication (23). To measure the average density of hydroxyl groups, we performed a nitrogen adsorption-desorption experiment and thermogravimetric analysis (TGA) (25). The Brunner-Emmett-Teller (BET) surface areas of the NR, SNS, and LNS were 421, 828, and 668 m2/g, respectively, while the weight loss percentages owing to the dehydration and condensation of the hydroxyl groups were 4.0, 10.0, and 8.2%, respectively (Fig. 1G). Therefore, the average density of hydroxyl groups on the NR, SNS, and LNS could be calculated accordingly as 3.2, 4.0, and 4.1 ─OH/nm2, respectively, close to the levels of other silica materials (25, 26). To roughly estimate the heterogeneous densities of hydroxyl groups on the NR, we compared the adhesion forces obtained from AFM, and it was estimated that there were approximately 2.5 and 3.7 ─OH/nm2 on the rod bodies and tips, respectively (Fig. 1H). This anisotropy of the NRs indicated that their fabrication process lead to a heterogeneous PEG grafting density and consequent different polymer chain conformations.

GMNP surrounded with an anisotropic HPEGC

Each type of MSN was then modified with various contents of PEG (percentages denoting the weight ratio of PEG to MSN). The changes in the hydrodynamic diameters and zeta potentials are shown in fig. S1 (B and C). The exact PEG densities grafted onto the NPs were measured via a fluorescamine-based assay (27). As shown in fig. S1D, the PEG densities of the three MSNs with 2% PEG were less than 0.03 chains/nm2, implying that the chains were loosely packed on the substrate and were in a mushroom-like conformation according to theoretical estimations (8). For those MSNs with 4% PEG content, the density of 0.04 to 0.06 chains/nm2 suggested a mushroom-to-brush transition state. While for particles with more than 6% PEG content, the density of higher than 0.06 chains/nm2 indicated that chains were in a brush-like conformation. However, these estimations indicated only isotropic modification, so the anisotropic NR should be carefully examined. Since water molecules exert substantial influence on the swelling and conformation of PEG polymers (28), investigating the HPEGC can be of great importance to more accurately predict the in vivo behavior. Although several approaches and computer simulations for measuring or predicting the thickness and conformation of PEG coatings have been developed (29), conventional analysis may not be able to reflect the anisotropic HPEGC on a single particle. Compared to these conventional methods, electron microscopy techniques are more suitable for showing the corona of a highly dense PEG coating under vacuum (29). Therefore, an AFM test was performed to carefully detect the HPEGC surrounding the NPs under fluid conditions (Fig. 2 and fig. S2) (30).

Fig. 2 Detection of HPEGC via AFM.

(A) Non-PEGylated NPs showed limited adhesion force and apparently lacked a corona. (B) Compared to the non-PEGylated NPs, the SNS and NR with 2% PEGylation showed increased adhesion force resulting from the interactions between the probe and PEG chains. The negligible HPEGC indicated the mushroom structure of PEG. (C) With 4% PEGylation, the NPs showed a clear HPEGC. The SNS showed an isotropic corona, while the NR exhibited anisotropy, in which a thick corona could be detected at the tips and a thin corona was present on the rod body (as indicated by the arrows), structurally mimicking germs. (D) NPs with 6% PEG content showed a uniform, thick HPEGC around them. In particular, PEG brushes were observed under this condition on the rod body (as indicated by the arrow). (E) With a much higher PEG content of 20%, the HPEGC became more rigid. However, the corona could be observed again as the peak force was increased from 1.5 to 5.0 nN. Representative images are presented. Scale bar, 100 nm.

AFM height images, together with adhesion mapping, were adopted to indicate the presence of an HPEGC. As shown in Fig. 2A, no hydrated corona could be detected around the non-PEGylated NPs. By contrast, with the increasing PEG content, the architecture of and the adhesion force induced by the HPEGC became clear. With 2% PEGylation, a negligible corona was found (Fig. 2B), in accordance with the low surface PEG densities (fig. S1D). The adhesion images also suggested that the PEG chains were independently packed on the rod body. When the PEG content increased to 4%, the NR exhibited an anisotropic HPEGC that was thick at the tips but much thinner around the rod body, implying the coexistence of extended brush and packed mushroom chains (Fig. 2C). Therefore, we constructed GMNPs (referred to as NR-4%PEG) that structurally mimicked germs. By contrast, the isotropic SNS and LNS displayed a uniform HPEGC (Fig. 2C and fig. S2D). It could be inferred that with this 4% PEGylation, the PEG density on the SNS and LNS was lower than that on the tip of the NR (as indicated by the thickness of the HPEGC). Furthermore, because of the similar average surface PEG densities (fig. S1D), the density on the SNS and LNS should be higher than that on the body of the GMNP. As mentioned above, the PEG chains presented on SNS and LNS were in a mushroom-to-brush transition state, and this intermediate state could show the trade-off advantages in balancing between diffusion and uptake efficiency but might not be better than the anisotropic state of the GMNP.

As the modification degree increased to 6%, the NPs showed an isotropic, thick HPEGC (Fig. 2D), a strong indication of superior diffusivity and poor uptake efficiency. Moreover, with a 20% PEG content, the whole surfaces of the NPs even had negligible adhesion force, which could be the consequence of the enhanced rigidity of the HPEGC formed by a dense brush–like PEG layer (Fig. 2E). Increasing the forces exerted on the NPs could again reflect the corona.

GMNP function similarly to germs in vitro

We next investigated whether GMNP could function similarly to germs in penetrating biological tissues. After labeling the NPs with rhodamine B isothiocyanate (RITC), a multiple-particle tracking approach was adopted to evaluate the ability of the NPs to move in fresh intestinal mucus. Ensemble-averaged mean square displacement (MSD) values are shown in Fig. 3A, and the logarithmic values of the effective diffusivities of all NPs analyzed are presented as distributions in fig. S3. In accordance with a previous report (31), naked NR displayed a shape advantage over NS, as the diffusivity was 2.8- and 5.1-fold higher than that of naked SNS and LNS, respectively. However, the diffusivity of the GMNP was 16.0-, 40.4-, and 83.0-fold higher than that of the naked NR, SNS, and LNS, respectively. Therefore, the HPEGC played a dominant role in controlling the diffusivity of the NPs and completely overwhelmed the shape effect. The tracking data also showed that compared with NR-2%PEG, the GMNP exhibited a sharp growth in diffusivity (Fig. 3, A and B; as indicated by the arrows), implying that the formation of brush-like PEG chains at tips greatly improved diffusivity, thereby mimicking the flagellum-driven mucus-penetrating ability of germs. By contrast, for the isotropic counterparts, this rapid increase in diffusivity did not appear until uniform PEG brushes formed at a PEG content of 6%. All groups showed a decelerating trend in their growth rate of diffusivity at higher PEG contents, tending to reach a plateau in a similar manner to a previous finding (8). These data suggested that GMNPs had comparable diffusivities with dense brush–like PEG-coated NPs and could rapidly penetrate the mucus barrier.

Fig. 3 In vitro functional evaluation of GMNP.

(A) Ensemble-averaged MSD of NPs diffusing in freshly obtained rat intestinal mucus. (B) Effective diffusivities of NPs. Arrows indicate a rapid growth in diffusivity. (C) The negative impact of PEGylation on uptake by E12 cells with preremoval of mucus (n = 4). (D) Uptake by mucus-containing E12 cells that simulated the multiple barriers of intestinal mucosa (n = 4). (E) Three-dimensional (3D) view of the transport of NPs across multiple E12 barriers at different times. Scale bar, 15 μm. (F) Different cell entry patterns of NPs. The anisotropic GMNPs adhered to the membrane via their body side, while the isotropic NRs were wrapped through the tip first. MEM, membrane. Scale bar, 500 nm. (G) Schematic of the superior efficiency of GMNPs over their isotropic counterparts in traversing across multiple barriers. (H) Transportation of GMNPs and their isotropic counterparts with optimized 4% PEGylation in tumor spheroids. Scale bar, 100 μm. (I) Quantification of fluorescent coverage of NPs at different depths of spheroids. Representative images are presented. Data are means ± standard error of mean. ns, no significant difference; **P < 0.01 and ***P < 0.001, one-way ANOVA and Bonferroni’s test.

To investigate the trends in cellular uptake, we then adopted a mucus-secreting HT29-MTX-E12 (E12) cell monolayer to emulate the intestinal absorption barriers in vivo. The negative impact of PEGylation on cellular uptake was assessed first by removing the mucus layer before incubation. Evidently, PEGylation could substantially impede cellular uptake (Fig. 3C). We noticed that compared with the naked NR, the GMNP with 4% PEGylation showed a nonsignificant 8.7% reduction in the uptake amount, while SNS and LNS with 4% PEGylation suffered a 45.0 and 54.0% decrease in uptake, respectively. Given that the PEG chains surrounding SNS and LNS with 4% PEGylation might be in an intermediate state, the results suggested that such a conformation could, to a certain extent, reduce the cellular uptake. By contrast, despite the PEG brushes on the tip side, the PEG mushrooms on the rod body of the GMNP could maintain the uptake efficiency of these NPs, probably because of the relatively strong interaction of packed polymers with the cell membrane (11, 12). As the PEG content increased to 6%, a marked decrease in uptake of 78.5, 84.4, and 85.8% was detected for the NR, SNS, and LNS, respectively. Moreover, all three groups showed negligible amounts of cellular uptake as the PEG content was continuously increased, which was in accordance with previous findings (32, 33). These data also validated the dominant role of the architecture of the HPEGC over the NP shape in the diffusion and cellular uptake processes.

We then performed an uptake study in the presence of a mucus layer. As shown in Fig. 3D, the mucus layer had a substantial negative effect on the overall uptake amount, while among the NPs assessed, the GMNPs best retained their uptake efficiency, which was 4.3-, 7.0-, and 11.0-fold higher than that of the naked NR, SNS, and LNS, respectively. For SNS and LNS with 4% PEGylation, the results suggested a trade-off effect that compromised between the contradictory requirements of diffusion and uptake. In comparison, the GMNP had a significantly higher uptake amount than the SNS and LNS, indicating that the coexistence of both brush and mushroom chains on NPs could result in better efficacy than that achieved with a trade-off design. The following studies mainly compared GMNP with the two types of NS modified with a PEG content of 4%, along with NR-2%PEG (isotropic mushroom regime) and NR-6%PEG (isotropic brush regime) to avoid bias of the shape effect. A three-dimensional (3D) observation of E12 cells at different times after uptake via confocal laser scanning microscopy (CLSM) was conducted. GMNP penetrated the mucus layer much faster than SNS, LNS, and NR-2%PEG (Fig. 3E). Although the NR-6%PEG could penetrate at a higher rate than the GMNP, negligible fluorescent intensity was found to penetrate below the nuclei. These data certificated the dominant role of the PEG corona in regulating the transport rate. Moreover, since size and zeta potentials can substantially influence the diffusion and uptake efficiency, we fabricated a small NR (SNR) and a large NR (LNR) to fully assure the dominant role of polymer conformation (fig. S4A). After PEGylation, SNR-6%PEG and LNR-2%PEG had similar hydrodynamic diameters and zeta potentials to those of GMNP (fig. S4, B and C). The diffusion and uptake results proved that without dual conformational PEG chains, they could not overcome multiple barriers as efficient as GMNP (fig. S4, D and E). To further address the differences between the GMNP that had anisotropic HPEGC and the NRs with isotropic HPEGC, we detected the orientation of the NPs at the initiation of cell entry (Fig. 3F and fig. S4F). As depicted in superresolution images, GMNPs adhered to the membrane via their body side, while the isotropic NR-2%PEG and NR-6%PEG attached the membrane via their tips, similarly to a previous finding (34). It could be reasonably suggested that the packed mushroom–like PEG chains on the rod body of the GMNP functioned like fimbriae that mediated strong interactions with the cell membrane. On the basis of these data, GMNPs might be superior to their isotropic counterparts for targeting multiple barriers (Fig. 3G).

Tumors are another representative type of tissue with complex biological barriers of a fibrous matrix and cells. In this study, multicellular spheroids were adopted as in vitro models. Apparently, the GMNP deeply penetrated the spheroids and had much higher coverage than the SNS and LNS (Fig. 3, H and I). We also investigated the non-PEGylated NPs and found that they showed high uptake amount in peripheral regions but limited coverage, most likely because of their poor diffusivity (fig. S4G). This result could be explained by the fact that the cellular barrier composed most of the periphery of the spheroids, which was also validated by the finding that NPs with 20% PEGylation all showed limited intensity after incubation (fig. S4H). Therefore, the GMNP with the dual conformational PEG chains could achieve rapid and deep penetration in tumor spheroids in vitro.

Efficient oral and tumor delivery of GMNP in animal models

Given these promising in vitro results, we determined whether the advantageous properties of GMNPs indeed benefited their in vivo delivery efficiency. SNS and LNS with 4% PEGylation, as the optimized isotropic groups, were used for comparison. To avoid the bias of individual diversity, we labeled the isotropic MSNs with a green dye [fluorescein isothiocyanate (FITC)] and mixed them with red RITC-labeled GMNP. After intragastric administration, the NPs were evidently transported along the small intestines at different rates (Fig. 4A). In particular, the GMNPs were better retained in the small intestines than their isotropic counterparts, even after 6 hours (mainly in ileum), while the isotropic MSN were almost completely removed by peristalsis. We also compared GMNP with NR-2%PEG and NR-6%PEG (fig. S5A). The NR-2%PEG were almost removed after 6 hours, similarly to the trends of the NSs, while the NR-6%PEG showed an even better intestinal retention effect than the GMNP 6 hours after administration due to their good mucus-penetrating ability.

Fig. 4 In vivo transportation of NPs in mucosal and tumor tissues.

(A) NP transportation through and retention in rat small intestines. (B) Intestinal uptake examined by CLSM after intragastric (i.g.) administration and quantification of relative fluorescent intensity. Scale bar, 300 μm. (C) Tumor permeation and retention effect of NPs in tumor-bearing nude mice. (D) Quantification of total radiant efficiency. *P < 0.05 and **P < 0.01 for comparison of GMNP and SNS and ##P < 0.01 for comparison of GMNP and LNS, one-way ANOVA and Bonferroni’s test. (E) Tumor slices examined under CLSM at 8 hours after peritumoral injection. Representative images are presented (n = 1 for saline control group and n = 3 for other groups). Scale bar, 200 μm.

However, a good retention effect alone is far from enough, and prompt absorption is the key factor in efficient delivery. The intestine was sliced and carefully examined by CLSM. At 1 hour after administration, a considerable number of GMNP had already reached the villi (Fig. 4B and fig. S5B), especially in the duodenum (fig. S5C), in sharp contrast to the SNS and LNS, which showed limited fluorescent intensity. After 2 hours, a large number of GMNPs were internalized by the villi and clearly transported to the basolateral (BL) side, while comparatively less green signal could be detected in the other groups. After 6 hours, a strong red signal from the GMNP was still observed near the villi, especially those in the ileum (fig. S5D). Notably, despite their good mucus-penetrating ability, negligible NR-6%PEG could be detected on the BL side in all intestinal sections, strongly indicating their poor uptake efficiency (fig. S5, B to D). These data proved that GMNP could be a better option than the other NPs tested for overcoming in vivo mucosal barriers.

A pharmacokinetics (PK) study was also conducted. Saquinavir (SQV), a class IV compound in the Biopharmaceutical Classification System, was loaded into the NPs, and the loading capacity of the SQV@GMNP, SQV@SNS, SQV@LNS, SQV@NR-2%PEG, and SQV@NR-6%PEG was 7.6, 12.2, 6.8, 7.8, and 7.1%, respectively. The release profile of NPs was similar in phosphate-buffered saline (PBS) (fig. S5E). The PK results demonstrated that compared to the other groups, the GMNP group exhibited remarkably improved SQV absorption (fig. S5F). The bioavailability of the SQV@GMNP in 12 hours was 21.9-, 2.6-, 3.9-, 2.6-, and 3.1-fold higher than that of crystalline SQV, the SQV@SNS, the SQV@LNS, the SQV@NR-2%PEG, and the SQV@NR-6%PEG, respectively (fig. S5G). Overall, we suggest that GMNP may be a more promising system for oral delivery and even other mucosal delivery.

In addition to mucosal delivery models, we also performed comparisons among the different NPs in tumor xenograft nude mouse models. After peritumoral injection, the GMNP showed good permeability and retention 4 and 8 hours after administration, while the fluorescent intensity of their isotropic counterparts decreased markedly (Fig. 4, C and D). The confocal images confirmed that numerous GMNPs were widely distributed in the tumor tissues after 8 hours (Fig. 4E). At this point, the NR-6%PEG did not show a good retention effect and accumulated in only the near vascular regions (fig. S6). This finding was in accordance with a previous report, showing that the brush PEG conformation could be beneficial for in vivo retention near blood vessels, although poor cellular uptake could offset this advantage (32). Therefore, dual PEG conformations that mimicked the structures and functions of germs could be more beneficial than a single PEG conformation for tumor delivery.

Different mechanistic regions facilitated the transport of GMNP

To obtain in-depth insight into how the anisotropic HPEGC functions from a microscopic perspective, we first detected the physical moving patterns of the NPs. Under stimulated emission of depletion (STED) microscopy, the GMNP exhibited a hopping effect after being trapped by mucin fibers at a position for an instant (Fig. 5A and movies S1 to S5). It could be inferred that the mushroom- and brush-like PEG chains mediated distinct interactions with the fibers and consequently caused this “adhesion-hopping” diffusion pattern. In contrast, with their nearly isotropic interactions, the SNS and LNS were immobilized in a limited region, and no hopping effect could be detected. In addition, we carefully examined the diffusion patterns of the NR-2%PEG and NR-6%PEG (fig. S7A). The NR-2%PEG showed a limited hopping effect, while the NR-6%PEG diffused much more freely, without characteristic adhesion-hopping pattern observed. Therefore, we demonstrated that although the PEG mushrooms would cause the adherence of the NPs to protein fibers, the brush-like PEG chains on the GMNP would effectively facilitate their diffusion similarly to the motion of germs pushed (or pulled) by the flagellum.

Fig. 5 Mechanistic insights into the superiority of the anisotropic HPEGC.

(A) Diffusion patterns of GMNP, SNS, and LNS in mucus under STED microscopy. As indicated by the yellow arrows, the position of the GMNP could suddenly change after being trapped by mucin for some time, displaying a strong hopping pattern. (B to E) Interactions of the HPEGC on different NPs with the cell membrane detected via AFM. Particles were modified onto the AFM probe, as shown by the SEM images. The average adhesive forces indicated that the body of the GMNP with mushroom-like PEG chains had much stronger interactions with the cell membrane than did the rod tip, SNS, and LNS. Data are means ± standard deviation.

We further measured the adhesion force between the particle and cell membrane. NPs were modified onto the AFM probes, and in particular, the GMNP was modified onto them through different methods to expose their tips (Fig. 5, B to E, and fig. S7B). The force data were recorded, and hundreds of adhesion force measurements were performed. The ensemble-averaged adhesive forces for GMNP body, GMNP tip, SNS, and LNS were 428, 12, 55, and 91 pN, respectively. Compared to the brush-like PEG chains on the rod tip, the PEG chains in the mushroom conformation on the rod body induced a much stronger adhesive interaction with the cell membrane, similar to the functional fimbriae on germs, which could, in part, explain their preserved cellular uptake efficiency. As mentioned in the previous sections, reports have shown that mushroom-structured polymers could promote adhesive interactions with proteins and therefore facilitate cellular adhesion (11, 12). Since “PEG dilemmas,” together with protein corona, are also critical issues for intravenous delivery, we also compared the GMNP with SNS and LNS to further explore the potential applicability of the GMNP (fig. S8). Although the in vitro protein corona levels were similar among the NPs, the GMNP demonstrated longer circulation time and different biodistribution (higher accumulation in lung). We suggested that future efforts might be deserved to study the influence of anisotropic polymer conformation on the biobehaviors of NPs in the circulation.

MD simulations

To gain in-depth insights into the mechanisms underlying our findings, we conducted a series of coarse-grained MD (CGMD) simulations. In the diffusion simulations, a regular polymer network was used to represent biological hydrogel fibers with a mesh size of 14σ, where σ is the unit of length (fig. S9A). Three types of NR with different surface properties but the same size were studied in our simulations (fig. S9B). The rigid NRs were randomly distributed throughout the polymer network, as shown in fig. S9C. The effect of the PEG density in different regions of the NRs was represented by the interaction between the network nodes and NR, reflecting the different affinities between the NR and mucin fibers due to the change in PEG conformation. Then, a series of simulations were performed with different NR-network node interactions (where αA and αB represent the interactions between the chain-tip and chain-body region of the NR, respectively). The results showed that with increasing (αAB), the MSD of the NR decreased (fig. S10A). The diffusion of NRs with low PEG density was impeded in the network due to their high affinity with the network nodes. By contrast, for NRs with anisotropic PEG density, a sharp increase in diffusivity could be found (movie S6). To elucidate the different diffusivities of these different NRs in the polymer network, we analyzed the diffusion processes of three characteristic types of NR, i.e., NRs with high PEG density, NR with anisotropic PEG density, and NR with low PEG density. Figure 6 (A to C) depicts typical 3D centroid trajectories of the diffusion of these three NRs in the polymer network. We found that the trajectory of the NR covered a larger area as the affinity decreased. Notably, compared to the NRs with high and low PEG densities, the NRs with anisotropic PEG density (αA = 0.5, αB = 1.2) diffusing in the network exhibited a remarkable hopping process. The MSD values for different NRs showed that although less obvious, hopping existed within a very short time scale (approximately 3.95 × 105τ), the MSD increased linearly with the same slope when the PEG density was high, suggesting that the diffusion of these NR in the polymer network was Brownian motion (Fig. 6D). The NR with low PEG density would be trapped in the holes in the network due to their high affinity (Fig. 6F). Conversely, for the NR with anisotropic PEG density, the MSD value increased with a small slope within a short time scale but a high slope within a long time scale (Fig. 6E). The hopping process indicated that these NRs would be trapped in a hole in the network within a short time scale (approximately 1.385 × 106τ) but would escape this restriction and diffuse into another hole within a long time scale (movie S6). This adhesion-hopping diffusion mechanism explained how GMNP could rapidly penetrate the network of biological hydrogels with the help of the brush-like PEG chains, similarly to the germs that are pushed or pulled by flagellum to realize their locomotion in the mucus. To verify the robustness of our CG model about the simulation of the diffusivity of the NPs in polymer network, we have tuned the pore size of the network from 10σ to 18σ in the simulations. The results showed that increasing the pore size would augment the diffusivity of the NRs due to the loosened restriction of the network (fig. S10, B and C) and the value of diffusivity drops markedly in the small pore size due to the strong trapping of the network (fig. S10, B and C).

Fig. 6 The diffusivity and internalization processes of different types of NRs from molecular simulation.

Typical 3D centroid trajectories, MSD values, and endocytosis pathway of the NRs with high PEG density (A, D, and E), anisotropic PEG density (B, E, and H), and low PEG density (C, F, and I). In the simulation of the diffusivity of the NPs, αA and αB represent the interactions between the chain nodes and two tip regions and between the chain nodes and body region of the NRs, respectively. For the simulations of the internalization process, the effect of the PEG density was represented by the interaction between the NPs and chain network/membrane bilayer, where εA and εB are the interactions between the receptors and two tip regions and between the receptors and body region of the NRs, respectively. The green and blue dotted lines in (D to F) are the slopes of the plots of the MSD versus time and the time scale of the hopping, respectively. The insets in (G to I) are the stable states of the endocytosis of different NRs.

To further verify the uptake rate of the different types of NPs by cells, we compared the internalization pathways of NPs using other CGMD simulations. The cell membrane was composed of a mixture of receptors/lipids with a ratio of 0.25. Each lipid or receptor molecule is represented by three beads, as developed by Cooke et al. (35, 36). A rigid NP is positioned in close proximity above an equilibrium membrane patch with a size of 160 × 160σ2 at the initial stage. Then, the NP interacted with the membrane via ligand-receptor binding and the kinetic pathway of the membrane-NP interaction. From a series of simulations, we found that the NPs with high PEG density adhered onto only the surface of the membrane and could not be wrapped fully by the membrane, while the NPs with low PEG density would be wrapped fully and internalized by the membrane in a short time regardless of their shape and size (fig. S10D). Figure 6 (G to I) reflects the kinetic pathways of the three different NRs. For the NR with high PEG density, they adhered onto only the membrane surface with a slight rotation due to the weak ligand-receptor binding interaction (Fig. 6G). The NR would be wrapped fully and internalized by the cell membrane with continued rotation in a short time when the PEG density on their surface was low (Fig. 6H). By contrast, the NR with anisotropic PEG density underwent a fast wrapping process at the early stage due to the strong interaction between the body region of the NR and membrane, and then, the wrapping process decreased because of the weak interaction between the tip region of the NR and the membrane (Fig. 6I). Therefore, most of the NRs with anisotropic PEG density would be adhered tightly to the membrane in a short time but took a relatively long time for complete internalization. We suggest that the internalization of some of the NPs by the cell (Fig. 3, C and D) might not represent complete uptake. Nevertheless, the overall efficiency and in vivo results proved that anisotropic GMNP could be a better alternative to isotropic GMNPs.

From the above simulation results, we concluded that although the high PEG density could improve the diffusivity of the NPs in biological hydrogels, the weak membrane-NP interaction led to their poor cellular internalization rate. NPs with low PEG density could be internalized by the cell efficiently, but they would be trapped in the biological hydrogels before contacting the cell. By contrast, the NR with anisotropic PEG density rapidly diffused through the hydrogels with hopping diffusion behavior and could be wrapped by and enter into the cells with considerable efficiency. These computational simulations elucidate the mechanisms lying in the superiorities of GMNP and provide basic principles of designing polymer-decorated NPs for overcoming multiple barriers.


Naturally evolved germs that are equipped with long flagellum and short fimbriae can effectively invade human mucosal tissues. Taking this inspiration, we constructed a GMNP with dual PEG conformations to improve the delivery efficiency of NPs in mucosal and tumor tissues. First, the GMNP structurally imitated germs that were equipped with flagella and fimbriae, as suggested by the AFM data (Fig. 2C). In particular, adhesion mapping indicated that the GMNP had extended chains on the rod tips and packed chains on the rod bodies. Second, GMNP could overcome the diffusion and cellular barriers in a similar manner to those pathogens. The two conformational PEG chains would provide the GMNP with distinct properties on regions of tips and bodies, respectively, which have quite different interactions with biological systems. Therefore, the extended brush–like PEG chains could help the particle to get rid of the adhesive fibers and rapidly penetrate the biological hydrogels (movies S1 and S6). In addition, the mushroom-like PEG chains retained a strong interaction with the cell membrane, functioning like fimbriae that firmly attached to the epithelium, as it is generally known that a majority of biological phenomena are strongly dependent on the structures and functions of polymer-like molecules. Although the dense-brush conformation of PEG polymers has been widely believed to be preferable for in vivo delivery (9), PEG dilemma emerges to be a critical issue that cannot be neglected with regard to efficient delivery. Mucosa and tumors that contain biological hydrogels and cells have contradictory requirements for the surface properties of NPs. The GMNP constructed in this work provide better delivery efficacy, indicating that those evolutionary significance regarding molecular conformation in life sciences may provide natural inspirations for the development of drug delivery systems.

Meanwhile, sufficient control groups were adopted, and deep insights were obtained to fully address the dominant role of the coated PEG chains, rather than the shape, in mediating distinct interactions of the NPs with biological systems. In addition to the superior delivery efficiency of the GMNP, it was interesting to note the two unique mechanisms of the GMNP compared with other isotropic NRs. In the diffusion study, although not fully covered with PEG brushes, the anisotropic GMNP could display remarkable hopping diffusion in biological hydrogels, as proven by STED microscopy and explained in detail by MD simulations. Such a nonconventional moving pattern endowed the GMNP a rapid increase in diffusivity compared to that of the NR-2%PEG and a diffusion rate that was comparable with that of the NR-6%PEG, which had isotropic brush–like PEG chains. The NR-2%PEG might also display a hopping effect due to its intrinsic shape anisotropy, but this effect could be limited because of the strong affinity of these NPs with fibers. On the other hand, during the uptake process, in contrast to NR with an isotropic HPEGC that initially entered cells with their tips (34), the GMNPs adhered to the cell membrane via their bodies and could be wrapped at a rapid rate. Therefore, the conformation of PEG could govern the process of NP cell entry, and the strong interactions of the mushroom-like PEG chains with the cell membrane resulted in high uptake efficiency. Overall, the conformation of the PEG chains around the GMNP played a dominant role in regulating their mucus-penetrating and cellular uptake processes. In the future, studies on the biological behavior of synthetic polymer–based particles can even provide insights into how the macromolecular conformation functions during those currently mysterious biological activities.

It is also noticeable that there are many other strategies for engineering PEGylated NPs, including cleavable PEGylation and ligand modification, which specifically tackle issues under certain environments. Compared to those conventional methods, we focused on the common-shared diffusion and cellular barriers in mucosa and tumors and designed GMNP to play multiple functions on a more general basis. Besides, we suggested that a 4% degree of PEG might not be the best, but what mattered was to decorate PEG chains with different conformation on a single particle. This biomimetic approach could realize more effective delivery.

Moreover, future works may also be focused on the behaviors of GMNP during the circulation. Although, in vitro, the GMNPs demonstrated a similar level of adsorbed protein to that of their isotropic counterparts (fig. S8, A and B); a prolonged circulation time of the GMNP was observed after intravenous administration (fig. S8C). We suggest that the in vivo fluid dynamic effect during circulation, together with the anisotropic HPEGC, probably affected the adsorption of proteins on the GMNP (37). Additional studies on the protein corona aspects of anisotropic PEGylated NPs may need to pay attention to the different dynamic conditions between in vitro and in vivo models. Moreover, compared to the other NPs tested, the GMNP exhibited significantly higher accumulation in the lung 24 hours after administration (fig. S8D), although most particles accumulated in the liver and spleen. It would be meaningful to pursue in-depth insights into these in vivo issues in the future. Novel anisotropic polymer-coated NPs might be developed to balance the requirements of a long circulation time, desirable biodistribution, and effective tissue penetration. Moreover, other issues, such as intracellular and transcellular transport, may also deserve careful investigations on the effect of anisotropic polymer conformations.

In summary, by combining the advantages of diffusivity and cellular uptake, GMNP with dual conformational PEG chains could better overcome complex biological barriers than isotropic NPs. Following a simple but valid principle for the convenience of future scale-up, NPs modified by single-component polymers with multiple conformations and anisotropic properties may be desirable. Future works can also focus on enhancing their potency by selecting more versatile polymers, including polypeptides and polysaccharides, as well as other core NPs. For instance, hybrid lipid systems (38) could be remolded for anisotropic modification of polymers to obtain a larger differential in hydrophilicity/lipophilicity. On the other hand, novel characterization methods for anisotropic properties that are better under biological conditions should be developed to obtain full knowledge and make more accurate predictions about the interactions of NPs with biological systems. This work provides useful guidelines for the future rational engineering of polymer-decorated NPs for improved therapeutic outcomes.



PEG was purchased from Shanghai Peng Sheng Biotechnology (Shanghai, China). RITC, FITC, and (3-aminopropyl) triethoxysilane (APTES) were purchased from Aladdin (Shanghai, China). Hoechst 33342 dye (trihydrochloride) was bought from Yeasen Biotech Co. Ltd. (Shanghai, China). DAPI (4,6-diamidino-2-phenylindole; dihydrochloride) was purchased from Beyotime Biotechnology (Shanghai, China). The bicinchoninic acid (BCA) protein assay kits were purchased from KenGen Biotech (Nanjing, China). Wheat germ agglutinin Alexa Fluor 488 was purchased from Thermo Fisher Scientific (USA). The deionized water was generated by a Millipore Milli-Q system (USA). The Dulbecco’s modified Eagle’s medium (DMEM) and fetal bovine serum (FBS) were purchased from Gibco (USA). SQV was purchased from ChemBest Research Laboratories Ltd. Other chemicals were purchased from Sinopharm Chemical Reagent Company (Shanghai, China).

Cell culture

E12 cell line was cultivated in DMEM supplemented with 1% penicillin-streptomycin, 1% nonessential amino acids, and 10% FBS. The cells were seeded at a density of 1 × 105 cell/cm2 on a 12-well Transwell filter insert (3-μm pore size) and cultivated over 2 weeks. The medium was changed every second day for the first week and then changed every day. The transepithelial electrical resistance value was measured to ensure the formation of monolayer before experiments. Biopsy xenograft of Pancreatic Carcinoma line-3 (BxPC-3), human pancreatic stellate cells (HPSC), and human colon carcinoma cells (Caco-2) were cultured in DMEM, containing 10% FBS and 1% penicillin-streptomycin.


Male Sprague-Dawley (SD) rats and BALB/c nude mice were provided by the Animal Experiments Center of the Shanghai Institute of Materia Medica (Shanghai, China). All the animal experiments were carried out according to the Institutional Animal Care and Use Committee (IACUC) guidelines of the Shanghai Institute of Materia Medica (IACUC code: 2018-03-GY-38).

Fabrication of NPs

Particle shape was controlled by adjusting the reagent’s concentration and temperature. Briefly, for NR, hexadecyl trimethyl ammonium bromide (CTAB; 0.219 g) was dissolved in deionized water. Then, NH4OH (870 μl) was used to alkalify the solution, followed by stirring for 1 hour. After introducing tetraethyl orthosilicate (TEOS; 470 μl) dropwise, the reaction was kept for 4 hours to form the raw NPs. The NPs were collected by centrifugation (13,000 rpm, 10 min) and washed with water and ethanol three times, respectively. After removing the surfactant in ethanol and HCl, the lyophilized (Leica EM CPD300, Leica, Germany) MSN were obtained and stored. For SNS, CTAB (0.4 g) and triethanolamine (0.11 g) were dissolved in water (20 ml) to form the template micelle under 95°C, and then, TEOS was introduced (1.5 ml) to keep the reaction for another 1 hour. The following template-removal processes were the same as that of NR. For LNS, CTAB (0.32 g), KH2PO4 (0.686 g), glycerin (9 ml), and NaOH (0.116 g) were dissolved to form the micelle under 95°C. TEOS (1.8 ml) was then introduced dropwise. Other procedures were similar to those of SNS.

For PEGylation of MSN, particles were suspended in ethanol solution (with a volume ratio of 1:2 for ethanol:water). After the addition of silane-PEG5K-NH2, the reaction was kept for 24 hours at 60°C, and then, the products were washed with ethanol and water and stored in deionized water. Abbreviations of 0, 2, 4, 6, 10, and 20% used in this work represented the initial weight ratio of PEG to MSN.

For fluorescent labeling, RITC powders were dissolved in ethanol, and then, APTES was added to the mixture with stirring in the dark for 24 hours to obtain APTES-RITC conjugates. NPs were then introduced to the solution, and the reaction was further kept for 8 hours. Labeled NPs were washed with ethanol and water. The fluorescence intensity for the same number of particles was adjusted to be similar as previously reported (31) so that the results would be comparable. Similar procedures were performed for labeling NPs with FITC and IR-783.

Characterization of NPs

The hydrodynamic diameters and zeta potentials of the particles were measured in Zetasizer (Nano ZS, Malvern Instruments, UK). The morphology of MSN was obtained under scanning electron microscopy (SEM; S-4800, Hitachi, Japan). BET surface area and Barret-Joyner-Halenda cumulative pore volume were measured via N2 adsorption-desorption via conventional procedures (Autosorb IQ, Quantachrome, USA). The hydroxyl density was quantified via TGA (TG 209 F3, NETZSCH, Germany) according to the previously reported procedures (25). Briefly, lyophilized powders (several micrograms) were placed under atmosphere condition with gradually increased temperature (5°C/min). The samples were balanced for 2 hours at 200°, 400°, and 1000°C, respectively, and weight loss was measured constantly. Although the weight loss was completely in the form of H2O, MSN underwent significant weight loss at 200°C due to the loss of adsorbed water and at 400° and 1000°C due to the dehydration and condensation of the hydroxyl groups. Therefore, the weight loss in the latter two processes reflected the number of hydroxyls. TGA data showed that the percentage of weight loss at the first stage (completely attributed to the loss of adsorbed water molecules, Δw1) was 7.44, 9.52, and 5.96% for NR, SNS, and LNS, respectively. Assume that the original sample weight (i.e., total weight of pure MSN plus adsorbed water molecules, WTotal) was 100%. Therefore, the weight percentage of pure MSN (WPure) could be 92.56, 90.48, and 94.04% for NR, SNS, and LNS, respectively, as calculated byWPure=WTotalΔw1(1)

The percentage of weight loss at the second stage (Δw2) was 3.07, 8.12, and 6.97% for NR, SNS, and LNS, respectively, and the percentage of weight loss at the third stage (Δw3) was 0.66, 0.91, and 0.74% for NR, SNS, and LNS, respectively. Therefore, the weight loss attributed to the surface hydroxyl groups (ΔwH) was the summary of the two stagesΔwH=Δw2+Δw3(2)

Then, the weight percentage of surface hydroxyl groups in purely dried MSN (WH) could be calculated as the following equationWH=ΔwH/WPure × 100%(3)

Here, we got the WH of 4.03, 9.98, and 8.20% for NR, SNS, and LNS, respectively. Last, the exact surface hydroxyl density (DH) was determined by the ratio of number of hydroxyl groups to the BET surface area as the following equationDH=(WH × NA)/(18 × SBET)(4)where the molecular weight of H2O is 18 g/mol, NA is the Avogadro constant, and SBET represents the BET surface area, assuming 1 g of pure MSN for each calculation. Because of the presence of adsorbed water molecules that can significantly interfere the real mass (26), NPs were all dried in muffle furnace (SX2-5-12TP, China) under 200°C before weighing and usage in other experiments.

To estimate the distribution of hydroxyls, gold-coated AFM probe (Bruker, Germany) was modified by mercapto acetic acid to introduce carboxyl groups via the conventional two-step precleaning procedure (39). The AFM test (Dimension Icon and FastScan Bio, Bruker) was performed in the air with controlled cubicle humidity. High concentration of NR solution (100 μg/ml) was dropped on the mica sheet to detect the rod tip, and particles were dried in muffle furnace before test. PeakForce Quantitative Nanomechanical Property Mapping (QNM) mode was adopted, and the adhesion forces resulted from frictions between carboxyls and hydroxyls were collected and analyzed. Three independent tests were carried out, and more than 300 forces from 30 particles were recorded for each group. Data were processed in NanoScope Analysis (version 1.80, Bruker).

Fluorescamine-based assay

To measure the exact PEG density grafted on the NPs, the solution of free silane-PEG5k-NH2 and PEGylated MSN was centrifuged after reaction. The unreacted polymers in the supernatant were collected and conjugated with fluorescamine to form a fluorescent agent. Briefly, a series of PEG solutions with different concentrations in phosphate buffer (3 ml) was prepared. The pH of the solutions was adjusted to 10 by Na2HPO4 and NaH2PO4. Fluorescamine (2 μM) was added into the solution, and the reaction was kept for 15 min. The fluorescence was measured at 390 nm (excitation) and 480 nm (emission), and the calibration curve was obtained. The measurement of free PEG polymers was similar to the procedure used for the calibration curve. Therefore, the PEG density could be calculated by subtracting the unreacted ones from the total amount. The absolute PEG density was presented in the number of PEG molecules packed per square nanometer of apparent surface area (not BET surface area since the critical parameter that affects the conformation should be the PEG density based on the projected area that ignored the influence of pores).

Detection of HPEGC by AFM

NPs were fixed on the mica sheet before the test. Briefly, acetic acid was added to the ethanol solution (95%, 30 ml) to adjust the pH to 3.5, and APTES solution was introduced to reach a final concentration of 0.065% (weight). After hydrolysis under room temperature, the mica sheets were soaked in the solution for 5 min. The prepared NH2-terminated mica sheets were then soaked in chloroform and trimethylamine, followed by introducing citraconic anhydride dropwise. The reaction was kept at room temperature overnight to form carboxyl. NPs were modified with APTES to introduce amino groups before fixation procedures. The COOH-terminated mica sheets were placed into the NP suspensions, followed by introducing 4-dimethylaminopyridine (DMAP, 4 mg). The reaction was kept for 4 hours under room temperature, and NP-fixed mica sheets were soaked in ethanol overnight. AFM test (PeakForce QNM mode in fluid) was performed using silicon probes, and NPs were imaged at a scan rate of 1 Hz (256 samples per line) under 37°C. Feedback gain was 10.00. PeakForce setpoint was 1.5 nN (and additional 5.0 nN for NPs with 20% PEG degree). Height and adhesion images were aligned to illustrate the HPEGC. More than 20 NPs were carefully examined for each group. Data were processed in NanoScope Analysis.

Multiple-particle tracking

The diffusivities of NPs were measured through multiple-particle tracking. Male SD rats were sacrificed, and fresh mucus was carefully isolated from the surface of the intestinal lumen. To minimize the dilution effect that may induce overestimation of diffusivities, particle solution (200 μg/ml, 1 μl) was mixed with mucus (100 μl), followed by a 10-min incubation at 37°C. Movies were captured at a temporal resolution of 26.8 ms using the LAS 4.5 software in fluorescence inverted microscope (DMI 4000B, Leica, Germany). Three independent experiments were performed, and the trajectories of more than 200 NPs were analyzed using ImageJ for each type of NPs. The MSD and effective diffusivities (Deff) were calculated using the following equationsMSDτ=(x(t+τ)xt)2+(y(t+τ)yt)2(5)Deff=MSDτ4τ(6)where x and y represented the coordinates of the NPs and τ is the time scale.

Cellular uptake in E12 cells

To investigate the negative effect of PEGylation on cellular uptake, mucus was preremoved by N-acetylcysteine (10 mM, incubated for 1 hour). The upper medium was replaced by fresh culture (400 μl), and the monolayers were then incubated with RITC-labeled NPs (approximately 4.0 × 1010 particles) suspended in PBS (100 μl) for 1 hour (n = 4). For the studies on multiple barriers, the mucus was removed after the NP incubation, and a 10-min paraformaldehyde fixation was performed to stop the uptake, followed by washing with PBS three times and a 5-min lysis ready for BCA analysis by BCA assay kits (KenGen Biotech, China). The weight of protein was measured to normalize the number of cells in each group. The lysis solution plus the lower medium were collected for the measurement of RITC fluorescent intensity in the microplate reader (Synergy H1, USA) to calculate the particle numbers. The ratio of number of NPs to the total weight of cellular protein represented the uptake efficiency (equal to the ratio of number of particles and weight of cellular protein). For 3D observation by CLSM (FluoView FV10i, Olympus, Japan), the incubation medium at different time points was washed with PBS. Mucin and nuclei were stained by wheat germ agglutinin Alexa Fluor 488 (10 μg/ml) for 15 min and Hoechst 33342 dye (5 μg/ml) for 10 min, respectively. We used the z-stack mode to visualize the monolayer in a 3D pattern. To carefully examine the cell entry pattern of NPs at initiation, images were captured by CLSM (Leica TCS SP8 STED 3X, Leica Microsystems, Germany) at a 100× magnitude, with a resolution of less than 20 nm per pixel. Then, the images underwent a deconvolution process in Huygens Professional to obtain the high-resolution pictures.

Permeation in multicellular spheroids

To construct the in vitro tumor spheroid model, HPSC and BxPC-3 cells (about 5000 cells for each) were coincubated on agarose (2%, w/w) and cultivated for 5 days without physical turbulence to form the sound spheres. Spheroids with a size of 300 to 350 μm were chosen for the following permeation studies. NPs (500 μl, approximately 4.0 × 1010 particles) were added into the spheroid culture and incubated for 3 hours. Then, the spheroids were washed carefully and detected via CLSM. The fluorescent coverage was quantified by ImageJ.

In vivo intestinal retention and uptake

To avoid the bias of individual difference, isotropic NPs were modified with equivalent FITC. After a fast of 18 hours, each male SD rat was intragastrically administered mixed solution of different fluorophore-labeled NPs (approximately 8.0 × 1010 particles for each in a total suspension of 1 ml). Then, the rats were euthanized 1, 2, 3, and 6 hours after administration, and the gastrointestinal tissues were extracted for in vivo imaging (PE IVIS Spectrum, USA). Small intestines were excised into multiple sections and were fixed with 1% paraformaldehyde for 4 hours and dehydrated in 20% sucrose solution overnight at 4°C. Last, the intestines were sliced with a depth of 20 μm using a frizzing microtome (Leica CM1950, Germany), stained with DAPI, and observed under CLSM. Relative fluorescence intensity was quantified in ImageJ.

In vivo tumor permeation

The tumor-bearing model was established by subcutaneous injection of BXPC-3 and HPSC cells (with an identical number of 3.0 × 106 for each) at the right hindlimbs of nude mice. The mice were used to study the penetration and distribution 2 weeks after the formation of tumor xenografts. Because of the interference of normal tissues when performing in vivo imaging, RITC was replaced by IR-783, a near-infrared (NIR) fluorescent dye. At 0.5, 1, 2, 4 and 8 hours after peritumoral injection of IR-783–labeled NPs (approximately 8.0 × 109 particles, 100 μl), NIR images were obtained in the IVIS system. Images were processed in Living Image 3.1 software.

To examine whether the strong fluorescence of GMNP at 8 hours truly resulted from the efficient permeation into tumors, the same experiment was performed again while using RITC-labeled NPs (NIR dyes were not detectable in CLSM). After euthanasia of rats, the tumor tissue was sectioned in the frizzing microtome, and slices were examined under CLSM.

Diffusion patterns of particles in mucus under STED microscopy

Similar experiments were performed as in multiple-particle tracking studies, while movies were captured under STED microscopy (Leica TCS SP8 STED 3X, Leica Microsystems, Germany). Unlike the tracking studies, STED movies focus on a single particle to show its moving pattern. Movies were processed in Imaris.

Measurement of NP cell adhesive forces

AFM silicon probes were first flattened through focused ion beam (FIB; GAIA3 GMU Model, Chech) to expose an area with width of several hundred nanometers (for the convenience of conjugating NPs), and then, the flattened probes were treated via similar procedures as the modification of mica sheet to form COOH-modified probes. NPs with 4% PEGylation (with a particle number of 4.0 × 107 for the studies of SNS, LNS, and GMNP body and 4.0 × 1010 for GMNP tip) were added to immerge the COOH-modified probes (500 μl of solution), and the reactions took 4 hours under room temperature by introducing DMAP. Last, MSN-modified probes were soaked in ethanol overnight and dried for 1 hour. SEM images were obtained to pick out those tips modified with single particle or with several aggregated GMNP that exposed the tip.

Caco-2 cells were cultured in DMEM for 2 days. Before performing force spectroscopy experiments, the cells were rinsed, and the culture was replaced by PBS. Force curves were recorded by AFM in PeakForce QNM mode. The adhesive forces resulted from the interactions between cell membrane and the HPEGC. Data were analyzed in NanoScope Analysis.

Drug loading and PK studies

SQV (60 mg) was dissolved into dimethyl sulfoxide (DMSO) (8 ml), and NPs (80 mg) were added into the solution. The reaction was kept for 36 hours. Then, the solution was centrifuged, and the precipitation was dried in vacuum to remove the DMSO. The products were suspended and centrifuged with methanol three times, and the supernatant was collected for ultraviolet (UV) measurement to gauge the loading capacity. To obtain the release profiles, crystalline SQV (5 mg) and SQV-loaded NPs (5 mg of SQV) were added into water (2 ml) and then added into dialysis bags (2 kDa). The dialysis bags were bathed in 50 ml of PBS. The released SQV was determined by UV at time points of 5 min, 15 min, 30 min, 1 hour, 2 hours, 3 hours, 4 hours, 6 hours, 8 hours, and 12 hours.

For PK study, the powders of SQV-loaded NPs were suspended into deionized water (1 ml). Since SQV was clinically taken up after meal, SD rats were fasted for 18 hours, then given access to food for 2 hours, and administered SQV-loaded NPs intragastrically with a dosage of 10 mg/kg. Crystalline SQV was set as the control group. The plasma concentration at time points of 0.5, 1, 1.5, 2, 3, 4, 6, 8, 12, and 24 hours were measured by high-performance liquid chromatography (HPLC)–UV (239 nm) (although no signal could be detected at 24 hours). Fresh rat blood was collected and centrifuged at 4°C for 10 min (8000 rpm) to separate the serum. For each sample, serum (150 μl) was mixed with ketoconazole methanol solution (30 μl, 8 μg/ml) and NaOH (150 μl, 0.5 M). Then, a mix of methyl tert-butyl ether and dichloromethane (0.5 ml, 9:1, v/v) was added to the above solution to extract the SQV and ketoconazole. The extraction process was performed three times, and the upper organic solution was collected after centrifugation (3000 rpm, 2 min). The collected mix was dried under N2 flow to remove the organic phase and dissolved into the mobile phase for HPLC-UV measurement. Data were analyzed by DAS 2.0 (USA) using noncompartment model.

HPLC condition. Agilent 1200 series HPLC system (Agilent, USA) and HC-C18 (5 μm, 4.6 × 150 mm, Agilent, USA) were used to quantify the blood concentration of SQV. Here, we chose ketoconazole as the internal standard substance. The mobile phase was composed of acetonitrile and water (55%, v/v, pH 7.5 adjusted by phosphate and NaOH). The flow rate was 1 ml/min. The column temperature was set to 25°C. Separate substances were measured by UV at 239 nm.

Ex vivo protein corona evaluation in blood serum

NPs with 4% PEGylation (approximately 4 × 1011 particles) were incubated with freshly obtained serum (0.5 ml) from SD rats for 30 min (37°C). After then, protein-coated NPs were centrifuged, and the precipitation was suspended and treated using radioimmunoprecipitation assay lysis buffer, followed by the BCA analysis. The type of protein was compared via SDS–polyacrylamide gel electrophoresis. Briefly, the voltage was set to 80 V to concentrate the samples and then increased to 120 V to separate the protein.

Half-life time and biodistribution of NPs

SD rats were intravenously administered with NPs of 4% PEGylation (1 ml, approximately 4 × 1012 particles). Blood (300 μl) was collected after 15 min, 30 min, 1 hour, 2 hours, 4 hours, 8 hours, and 24 hours, respectively. The blood serum was then obtained via centrifugation and diluted using acetonitrile (1:1, v/v). The concentration of NPs was measured in the microplate reader. At 24 hours after administration, rats were euthanized, and the heart, liver, spleen, lung, and kidney were obtained for further comparisons. These organs were homogenized, and the fluorescent intensity was quantified.

MD simulations

Similar to the previous works (13, 31), our CG simulation system for investigating the diffusivity of the NPs in mucus was composed of regular polymer network and NPs. The polymer network was used to represent biological hydrogel fibers with a mesh size of 14σ (where σ is the unit of the length in our simulation), as shown in fig. S8. Each fiber was composed of a series of CG beads spanning the entire simulation box with size of 112 × 112 × 112σ3. To simulate entanglement and cross-link of mucin fibers, different fibers were cross-linked by the node beads (fig. S8A). The NPs with the hydrodynamic diameter of 9.0σ was modeled as a rigid particle with different shapes (fig. S8B). In the simulations, the nonbonded interaction between two CG beads was described by Lennard-Jones (LJ) potential as followsV(rij)={4αk[(brij)12(brij)6]rijrc0rij>rc(7)where αk is the depth of the energy well, b is the equilibrium length between two beads, and rc = 2.5σ is the cutoff distance. To describe the effect of the PEG density on the NPs’ surface, the interaction parameters αk between NPs’ different part surface beads and polymer chains node were verified in a wide range from αk = 0.2ε to αk = 1.0ε (ε is the unit of the energy in our simulation), which have captured the diffusivity of different NPs in polymer network (31).

During the simulations, we recorded the center of mass (COM) of each NPs every 10τdd is the unit of the time in our diffusion simulation), and then, the MSD and effective diffusivities were calculated using the following equationMSD(t)=(xtx0)2+(yty0)2+(ztz0)2(8)where x, y, and z represent the COM of the NPs, t is the duration of the time lag, and 〈⋯〉 represents the average of all the NPs. The velocity-Verlet algorithm was used to perform time integration during the simulations. The integration time step was Δt = 0.01τd. To maintain the dynamics of the NPs in the simulation, Langevin thermostat was used to control the system temperature at kBT = 1.0ε, where kB is the Boltzmann constant. We performed constant number-volume-energy integration to update the position and velocity of the beads in each simulation. The total simulation time was 2.05 × 107τd. After about 5.0 × 105τd relaxation, the MSD calculations were started.

The CGMD simulations for the endocytosis of the NPs were based on the solvent-free CG model developed by Cooke et al. (35) and Reynwar et al. (36). Briefly, the membrane was composed with a mixture of receptor/lipid with a ratio of 0.25. Each lipid or receptor molecule was represented by three beads, as shown in fig. S9D. The head of lipid/receptor molecule was formed by one hydrophilic beads (H), while the tails was formed by two hydrophobic beads (T). The neighboring beads i and j in the lipid/receptor are connected together by a finite extensible nonlinear elastic (FENE) potential bond with a constant of kFENE = 30ε and r = 1.5σ. The force constraining the stiffness was described by a simple harmonic spring with a spring constant of kharmonic = 10ε and an equilibrium bond length of r0 = 4.0σ between the head and last bead in each lipid/receptor molecule. The NPs were represented by a rigid particle with different shapes, and the ligands with a density of 0.27/σ3 were distributed on the surface of the NPs uniformly. In the simulation, the interaction between pairs of beads are described by the followingULJ(rij)={4εij[(brij)12(brij)6]rijrcut=2.5b0rij>rcut(9)UWCA(rij)={4εij[(brij)12(brij)6+0.25]rijrcut=2.5b0rij>rcut(10)UCOS(rij)={εij+UWCA(rij)rijrcut=21/6bεijcos2(π(rrcut)2ω)rcut<r<rcut+ω(11)UFENE(rij)=12kFENEr2ln(1r2r2)   rijr=1.5σ(12)Uharmonic(rij)=12kharmonic(rijr0)2   rijr0=4σ(13)

Following the notation from the original paper, the interaction parameters between the different types of beads were listed in fig. S9F. To avoid the distribution of the NPs to the equilibrium state of the membrane, a long-time [about 1 × 105τee is the unit of the time in our endocytosis simulation)] equilibrium simulation was performed before the endocytosis process in the initial stage. Next, an NP was put near the surface of the membrane (fig. S9E), and a long-term simulation was carried out to investigate the kinetics of NP internalization. For the simulations, we used a flat square membrane with a size of 160 × 160σ2 in an NpH ensemble with pressure p = − 10−4kBT3 to model a membrane with very low tension. The velocity-Verlet algorithm was used to perform the time integration. A modified dissipative particle dynamics thermostat was used to control the system temperature at kBT = 1.1ε. The thermostat used a damping parameter of Γ = 1.0ετe2 and a cutoff length of dcut = 3.0σ. In all the above simulations, the periodic boundary condition applied three directions, and all the simulations were performed using the LAMMPS code (40).

Statistical analysis

All data were presented as means ± standard error of mean, except specific notations. Statistical significance was analyzed by one-way analysis of variance (ANOVA) with Bonferroni’s test when multiple groups were compared (ns, P > 0.05; *P < 0.05; **P < 0.01, and ***P < 0.001).


Supplementary material for this article is available at

Fig. S1. Characterization of NPs.

Fig. S2. Schematic illustration and detection of HPEGC of LNS.

Fig. S3. Distribution of logarithmic effective diffusivities for NPs with various PEG degrees at a time scale of 1 s.

Fig. S4. Additional in vitro data.

Fig. S5. Superiority of GMNP for oral delivery.

Fig. S6. Superior tumor permeation of GMNP compared with the NR-2%PEG and the NR-6%PEG.

Fig. S7. Mechanistic studies.

Fig. S8. Protein corona-related and biodistribution studies.

Fig. S9. The molecular models used in the CGMD simulation.

Fig. S10. Diffusion and internalization simulations.

Movie S1. Diffusion pattern of GMNP in the mucus detected by STED.

Movie S2. Diffusion pattern of SNS in the mucus detected by STED.

Movie S3. Diffusion pattern of LNS in the mucus detected by STED.

Movie S4. Diffusion pattern of NR-2%PEG in the mucus detected by STED.

Movie S5. Diffusion pattern of NR-6%PEG in the mucus detected by STED.

Movie S6. Simulation of NR with anisotropic PEG density diffusing in biological hydrogels.

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: The computation experiment was mainly supported by the Supercomputing Center of the Chinese Academy of Sciences. We also appreciate the support in AFM and FIB experiments by H. Li from the Instrumental Analysis Center of Shanghai Jiao Tong University. Funding: We are grateful for the financial support from the National Natural Science Foundation of China (no. 81773651 to Y.G. and nos. 11422215, 11272327, and 11672079 to X.S.), the Strategic Priority Research Program of the Chinese Academy of Sciences (XDA12050307 to Y.G.), the K.C. Wong Education Foundation (to Y.G.), the Opening Fund of the State Key Laboratory of Nonlinear Mechanics, Chinese Academy of Sciences, and the China Postdoctoral Science Foundation (2019M650602). Author contributions: Y.G., X.S., Y.Y., D.N., and F.T. planned and conducted the studies and analyzed the data. Y.G., X.S., Y.Y., D.N., and F.T. prepared the manuscript. Y.Y., D.N., Y.L., and M.Y. prepared and characterized the NPs. Y.Y., D.N., and Y.Z. performed the ex vivo and in vitro experiments. Y.Y., D.N., K.Q., Y.L., and A.W. performed the in vivo studies. Y.Y. and D.N. conducted the mechanistic studies. F.T. performed the MD simulations. All authors carefully reviewed and approved the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article