Cell membrane coating integrity affects the internalization mechanism of biomimetic nanoparticles

Quantification of ratio of full cell membrane coating

The acquisition of cell membrane material used for coating usually involves two steps: extracting cell membranes from harvested cells and extrusion through porous membranes to create cell membrane-derived vesicles⁶. We hypothesized that this resulting mixture of cell membrane fragments and vesicles (Supplementary Fig. 1) fusing with NPs by extrusion or sonication could result in three classes of products characterized by the degree of membrane integrity: uncoated, partially coated and fully coated NPs (Fig.1a). In order to probe the integrity of the cell membrane coating, we developed a fluorescence quenching assay in which the NPs labeled with fluorescent nitro-2,1,3-benzoxadiazol-4-yl (NBD) by sequential chemical covalent coupling (Supplementary Fig. 2) were treated with dithionite (DT), a negatively charged reducing regent that cannot cross membranes^22,23. DT reduces NBD to 7-amino-2,1,3-benzoxadiazol (ABD), which irreversibly quenches the fluorescence of exposed NBD-labeled NPs (Fig. 1b). We hypothesized that if the NPs were fully coated, then the fluorescence signal would remain after the addition of the DT quencher into the solution. In contrast, if the NPs were only partially coated or totally uncoated, the fluorescence intensity would progressively disappear due to the reduction of the NBD dye with DT (Fig. 1c). Therefore, by measuring the remaining fluorescence, we would be able to calculate the proportion of fully coated NPs.

a Schematic illustration of the preparation procedure of the cell membrane-coated mesoporous SiO₂ nanoparticles (CM-SiO₂ NPs). Membrane materials including fragments and vesicles were derived from source cells through extraction and extrusion processes. Then, the membrane materials were further fused with SiO₂ NPs to obtain CM-SiO₂ NPs by extrusion or sonication. The resulting mixture contained uncoated SiO₂ NPs, partially coated SiO₂ NPs and fully coated SiO₂ NPs. b Reduction of the fluorescent nitro-2,1,3-benzoxadiazol-4-yl (NBD) to the nonfluorescent 7-amino-2,1,3-benzoxadiazol (ABD) with dithionite (DT). c Schematic representation of probing the integrity of the cell membrane coating by the fluorescence quenching assay. In the presence of DT, which is a membrane-impermeant fluorescence quencher, the fluorescence of uncoated SiO₂ NPs (labeled with NBD) and partly coated SiO₂ NPs disappeared while only that of fully coated SiO₂ NPs was retained. d, e TEM images of bare SiO₂ NPs (d) and CM-SiO₂ NPs (e). Insets are magnified images of the areas highlighted with the respective yellow dashed box. Scale bars, 200 nm (top); 50 nm (bottom). f Schematic representation of symmetrically distributed NBD-labeled giant unilamellar vesicles (GUVs). g Typical confocal laser scanning microscopy (CLSM) image of symmetrical GUVs. Scale bar, 20 μm. h Schematic showing that SiO₂ NPs were endocytosed into CT26 cells after incubation 24 h, then localized within endosomes. i TEM images of CT26 cell endocytosed SiO₂ (E-SiO₂) NPs. Inset shows a magnified image of the area highlighted with a yellow dashed box. Scale bars, 2 μm (top); 200 nm (bottom). j Representative fluorescence traces corresponding to DT treatment of SiO₂ NPs, CM-SiO₂ NPs, GUVs and E-SiO₂ NPs. DT was first added at t = 120 s to bleach the fluorescence of NPs with a defective cell membrane coating. After a stable baseline was obtained, 1% Triton X-100 (TX-100) was added at 420 s to disrupt the integrated cell membrane and bleach the remaining fluorescence of fully coated NPs. k CLSM images of CM-SiO₂ NPs, GUVs and E-SiO₂ NPs before (top) and after (bottom) addition of DT. Scale bars, 20 μm. l Quantification of the relative fluorescence intensity of CM-SiO₂ NPs as well as the blank control (bare SiO₂ NPs) and positive control (GUVs and E-SiO₂ NPs). Data represents mean ± SD (n = 3). One-way ANOVA followed by post hoc Tukey test was used to determine the significance of data. ns: not significant.

To investigate the feasibility of this approach, mouse colon carcinoma (CT26) cells were used as a model cancer cell line from which cell membranes were extracted, while ~70 nm mesoporous SiO₂ NPs (Supplementary Fig. 3) were selected as a model core nanomaterial due to their excellent drug delivery and imaging capabilities²⁴. The purified cancer cell membranes were first obtained by emptying cells of their intracellular components by applying a combination of hypotonic lysing, mechanical membrane disruption, and differential centrifugation²⁵. The cell membrane material used for coating was then formed by physical extrusion through a 400 nm porous polycarbonate membrane. Finally, the resulting cell membrane materials were coextruded with SiO₂ NPs through a 200 nm porous polycarbonate membrane to obtain cell membrane-coated SiO₂ (CM-SiO₂) NPs. The successful fusion of cell membrane materials with SiO₂ NPs was directly visualized by transmission electron microscopy (TEM), which clearly showed the spherical SiO₂ core surrounded by an outer lipid bilayer shell of around 9 nm in thickness (Fig. 1d, e). These results were comparable to the reported thickness of natural cellular membrane phospholipid bilayers (5–10 nm)^26,27. Furthermore, dynamic light scattering (DLS) measurements revealed that the hydrodynamic diameters of SiO₂ NPs slightly increased from 118.6 ± 0.4 nm to 139.8 ± 5.4 nm after cell membrane coating (Supplementary Fig. 4a), confirming the presence of the coating cell membrane around the NPs. Meanwhile, the surface zeta potential changed from −41.2 ± 0.9 mV to −36.0 ± 0.3 mV (Supplementary Fig. 4b), due to charge screening by the cell membrane. To further verify the successful cell membrane coating on SiO₂ NPs, CT26 cells were incubated with CM-SiO₂ NPs for 4 h, where SiO₂ NPs and cell membrane were labeled with Cyanine 5 (Cy5) and 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI), respectively. Confocal laser scanning microscopy (CLSM) images demonstrated that the red fluorescence derived from SiO₂ NPs matched well with the green fluorescence derived from CT26 cell membrane (Supplementary Fig. 5), indicating the successful formation of CM-SiO₂ NPs. Furthermore, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis demonstrated that the protein composition of CM-SiO₂ NPs was similar to that of cell membranes (Supplementary Fig. 4c), indicative of a good retention of the characteristic proteins presents in the membrane of these cancer cells. Importantly, the obtained CM-SiO₂ NPs displayed good colloidal stability after 6 days of storage in phosphate-buffered saline (PBS) at a concentration of 1 mg/mL (Supplementary Fig. 4d). Collectively, these results confirmed the presence of cancer cell membranes on SiO₂ NPs cores, while no general conclusion could be drawn on the extent of the membrane coating.

In addition to the CM-SiO₂ NPs, we prepared NBD-labeled giant unilamellar vesicles (GUVs; Fig. 1f, g) via a natural swelling method²⁸ and endocytosed SiO₂ (E-SiO₂) NPs (Fig. 1h) as the control groups to test whether DT is able to cross the cell membrane. In order to produce E-SiO₂ NPs that were encapsulated by the membrane structures of living cells, CT26 cells were incubated with SiO₂ NPs (200 μg/mL) for 24 h. TEM images revealed that E-SiO₂ NPs were mainly located in the lysosomes (Fig. 1i), indicating that the SiO₂ NPs were fully surrounded by native cellular membranes. With the CM-SiO₂ NPs, GUVs and E-SiO₂ NPs in hand, we next tested the proposed fluorescence quenching assay. After addition of DT to the sample (Fig. 1j), the fluorescence of bare SiO₂ NPs decreased immediately and was completely lost after 5 min, consistent with a reduction of the free NBD dye. With respect to the NBD-labeled GUVs, we observed a rapid decrease in fluorescence (approximately 45%), consistent with the assumption that the only outer leaflet NBD-phospholipids were reduced, while the inner-leaflet phospholipids were protected. Following the addition of Triton X-100 (TX-100), the fluorescence intensity was further reduced by ~50%, as TX-100 can disrupt the GUVs and allow DT to react with the NBD-labeled phospholipids in the inner leaflet. Notably, a rapid decrease (~90%) in the fluorescence of the CM-SiO₂ NPs was observed after the addition of DT, indicating that the partial coating was dominant. To rule out the possibility that this greater fluorescence reduction could be due to protein-mediated permeation of DT across the membrane rather than partial coating, we measured the fluorescence of E-SiO₂ NPs, which were protected by integrated living cellular membranes. As expected, only ~20% of the fluorescence was lost upon addition of the DT, which could be attributed to the free SiO₂ NPs adsorbed onto the surface of cellular membranes. Similar to GUVs, a complete reduction was seen after permeabilizing the cells with TX-100, indicating that the remaining fluorescence of E-SiO₂ NPs had resulted from the protection of cellular membranes. This fluorescence quenching was further visualized in the CLSM images, where the fluorescence signal of CM-SiO₂ NPs completely disappeared, while that of GUVs and E-SiO₂ NPs remained after the addition of DT (Fig. 1k). We attributed the remaining fluorescence of GUVs to the NBD inserted in the inner leaflet. These results confirmed that DT was not able to cross the cell membrane, unless the cell membranes were not fully integrated, making the NBD-labeled SiO₂ NPs available for reaction. In addition, we observed that the adsorption of bovine serum albumin (BSA) on NPs didn’t reduce the quenching (Supplementary Fig. 6), indicating that the protein adsorption has little effect on our proposed fluorescence quenching assay. The quantification analysis revealed that the ratio of full coating of CM-SiO₂ NPs was only ~6% (Fig. 1l), which was due to the existence of a few fully coated SiO₂ NPs (Supplementary Fig. 7). Altogether, these results demonstrated that most of SiO₂ NPs were partially coated in contradiction to the common assumption that the extrusion process results in fully coated NPs^20,29,30.

Validation of partial cell membrane coating

Next, we applied the proposed quantification method to compare the cell membrane integrity of different coating approaches. To date, ultrasonic fusion (Fig. 2a) and membrane extrusion (Fig. 2b) have been the two most commonly used methods in the fabrication of cell membrane-coated NPs. Among them, the mechanical force imposed by ultrasonic energy or mechanical extrusion leads to a disruption of the membrane’s structure, followed by the spontaneous formation of the cell membrane coating¹. To explore the effect of coating methods on membrane integrity, we compared the ratio of full cell membrane coating of three methods: sonication, extrusion and a combination of these two approaches (Fig. 2c–f). Although we found the ratio of full coating was low with all the three methods (sonication: 1.8 ± 0.1%; extrusion: 6.2 ± 0.3%; combined sonication-extrusion: 6.5 ± 0.3%), it seemed that the extrusion process was more efficacious than sonication in the formation of a full cell membrane coating. Given that these top-down biomimicry approaches require that cell membranes and NPs come together under conditions of strong mechanical stress, we further investigated whether using milder coating approaches (e.g., simple incubation with cells) could retain the membrane integrity of biomimetic NPs. To do so, the exosome membrane-coated NPs that were recovered after exocytosis of the NPs previously endocytosed by cancer cells³¹ were chosen as an example of a milder coating method. After the CT26 cells were incubated with positively charged mesoporous SiO₂ NPs, we first collected the exocytosed SiO₂ (Ex-SiO₂) NPs by centrifugation of the supernatants (Supplementary Fig. 8a). Upon membrane coating, an increase in the hydrodynamic diameter from 132.8 ± 5.8 nm to 162.8 ± 4.4 nm was detected (Supplementary Fig. 8b) and the zeta potential of SiO₂ NPs changed from +26.2 mV to −23.7 mV (Supplementary Fig. 8c), validating the successful preparation of Ex-SiO₂ NPs. TEM imaging further confirmed the presence of membrane structures on the SiO₂ NPs surfaces (Supplementary Fig. 8d-f). Consistent with the top-down approach, the ratio of full coating of Ex-SiO₂ NPs was approximately 6% which was comparable with the value of the CM-SiO₂ NPs (Supplementary Fig. 8g), suggesting that this natural coating method was also unable to generate a complete membrane integrity for the resultant biomimetic NPs.

a, b Schematic illustration of the preparation of cell membrane-coated NPs with a sonication method (a) and a physical co-extrusion method (b). c Quantification of the ratio of full cell membrane coating with different coating methods (sonication, extrusion, and combined sonication-extrusion). d−f TEM images of CM-SiO₂ NPs fabricated using sonication (d), extrusion (e), and combined sonication-extrusion (f). Scale bars, 100 nm. g Quantification of the ratio of full cell membrane coating for SiO₂ NPs coated with different source cell membrane materials (HeLa, macrophage, platelet, and RBC). h TEM images of different sizes of nonporous Stöber SiO₂ NPs before and after coating with cell membranes. Scale bars, 100 nm. i Quantification of the ratio of full cell membrane coating for cell membrane-coated nonporous Stöber SiO₂ NPs of different sizes. j TEM images of Fe₃O₄ NPs, ZIF-8 NPs, Au NPs, PLGA NPs, and porous silicon (PSi) NPs before and after coating with cell membranes. Scale bars, 100 nm. k Quantification of the ratio of full cell membrane coating for different core materials (Fe₃O₄ NPs, ZIF-8 NPs, Au NPs, PLGA NPs, and PSi NPs). Data represents mean ± SD (n = 3). One-way ANOVA followed by post hoc Tukey test was used to determine the significance in c and g. p = 1.7E-6 (c: sonication+ and extrusion– vs. sonication– and extrusion+), p = 1.2E-6 (c: sonication+ and extrusion– vs. sonication+ and extrusion+), p = 6.1E-4 (g: RBC vs. Hela), p = 7.0E-4 (g: RBC vs. macrophage), p = 3.3E-4 (g: RBC vs. platelet). **p < 0.01. ns: not significant.

Since the cell membranes extracted from different cells contain a unique set of surface proteins, a variety of cell types were used as a source of membrane coating material to confer different functionalities onto the NPs. To examine whether the source of cell membrane materials could affect the cell membrane integrity of the resultant biomimetic NPs, membranes derived from four different cell types including cancer cells (HeLa), immune cells (RAW264.7 macrophages), platelets and RBCs, were used to construct biomimetic NPs. A partial coating was identified also when mesoporous SiO₂ NPs were coated with these different cell membranes (Fig. 2g), consistently with what observed with the CT26 cell membranes. We noted that RBC-based NPs did exhibit a higher ratio of full coating (~10.5%) than membranes obtained from other cells, probably because of their well-preserved cell membrane structure after hypotonic lysis and extrusion (Supplementary Fig. 9). It has been recently reported that the surface charge of NPs plays a key role in the formation of supported lipid bilayers on NPs³². Inspired by this, we further analyzed the cell membrane integrity of positively and negatively charged NPs (Supplementary Fig. 10a–d), in which positively charged NPs were prepared without using succinic anhydride modification (Supplementary Fig. 2). Compared with the negatively charged NPs that can be extruded easily, the mixture of the positively charged NPs and negatively charged membranes resulted in extensive aggregation (Supplementary Fig. 10a) through the strong electrostatic interactions that impede the extrusion process. This aggregation caused a significant decrease in ratio of full coating (Supplementary Fig. 10e), indicating that the presence of a negative surface charge favored the formation of a full cell membrane coating, consistent with its reported roles in coating³³. In addition, the strong affinity between positively charged NPs and negatively charged cell membranes could cause the collapse of the fluidic lipid bilayer, resulting in the arrangement failure of the local lipid upon the full coverage process³⁴. Furthermore, to ascertain whether higher amounts of cell membrane materials could result in higher rates of fully coated NPs, we measured the ratio of full coating by increasing the weight ratio of the cell membrane materials to NPs from 0 to 5 (Supplementary Fig. 10f). As expected, increasing the cell membrane content, gradually improved the ratio of full coating, which reached a maximum value (~16.6%) at a weight ratio of 5. However, further increases in the cell membrane content caused a significant blocking of the porous membrane that impeded the extrusion process.

On the basis of the knowledge that the formation of a supported lipid bilayer on NPs can be influenced by the curvature of NPs³⁵, we hypothesized that the size of core materials would also be an important parameter determining the integrity of the cell membrane of our biomimetic NPs. To test this hypothesis, we first prepared a series of nonporous Stöber SiO₂ NPs with different diameters, which were achieved by tuning the amounts of ammonium hydroxide in the synthetic procedure (Supplementary Table 1). The TEM images and size distribution histograms in Supplementary Fig. 11a–f show that the nonporous Stöber SiO₂ NPs had average sizes of 30 ± 2.9, 62 ± 8.1, 100 ± 7.0, 149 ± 7.8, and 190 ± 8.6 nm. To obtain cell membrane-coated NPs, each of these nonporous Stöber SiO₂ NPs was extruded with an equal mass of CT26 cell membrane vesicles through polycarbonate membranes with appropriate pore sizes (200 nm for 30 and 63 nm NPs, and 400 nm for 100, 149, and 190 nm NPs). DLS measurements revealed that after cell membrane coating, the diameter of these differently sized NPs was increased by 10–20 nm, whereas the zeta potential was decreased (Supplementary Fig. 11g, h). Consistent with the results discussed above, a partial coating was mostly observed in these differently sized NPs, even in the case of the smallest NPs with a diameter of 30 nm (Fig. 2h), which was further confirmed by the ratio of full coating results (Fig. 2i). Furthermore, we noted that the ratio of full coating of the smallest NPs (~15%) was much higher than that of the other NPs, which could be attributed to its favorable bending energy requiring a lower interaction force to pull the membrane to the surface^35,36,37.

According to what needs to be ultimately delivered to the targeted cells or tissues, the synthetic core materials are also important in the design of an effective cell membrane‐coated nano-construct. Hence, we next focused on understanding the effect of different core materials on the integrity of the membrane coating. To this end, we selected NPs made of magnetite (Fe₃O₄), zeolitic imidazole framework-8 (ZIF-8), gold (Au), PLGA, and porous silicon (PSi). Coating of CT26 cell membranes with these different NPs led to a consistent increase of the hydrodynamic size (Supplementary Fig. 12a) and of a change in the zeta potential (Supplementary Fig. 12b), demonstrating the presence of membrane coatings on the surfaces of these NPs. The partial coating of these NPs was clearly observed in the TEM images (Fig. 2j), and this was further supported by the ratio of full coating results (Fig. 2k). We attributed the highest ratio of full coating observed in Au NPs (8.3%) to the strong affinity of the free sulfhydryl group on the membrane protein for the Au NPs, which provides an extra force for cell membrane coating. Collectively, the experiments described above indicate that the partial coating is a general phenomenon existing in cell membrane‐based biomimetic NPs.

Cellular uptake of partially coated NPs

To explore whether the partially coated CM-SiO₂ NPs displayed self-recognition of the corresponding homologous cell line and enhanced cellular internalization, we checked the NP uptake capability of three cell lines including CT26, HeLa and MCF-7 (Fig. 3). The biocompatibility of CM-SiO₂ NPs was tested on CT26 cells and showed negligible deleterious effects (Supplementary Fig. 13), in line with previous reports of membrane-coated NPs³⁸. The CLSM images showed that CT26 cells exhibited a higher red fluorescence (Cy5-labeled NPs-based formulations) as compared to the other cell lines after incubation with CM-SiO₂ NPs, whereas all of the cell lines treated with SiO₂ NPs and DOPC lipid bilayer coated SiO₂ (LB-SiO₂) NPs displayed a similar weak red fluorescence (Fig. 3a), verifying the targeting ability of the cell membrane to its source tumor cell. We measured quantitatively the intracellular uptake of the different cell types by using flow cytometry (Fig. 3b–e). It demonstrated that after 4 h incubation with NPs, the fluorescence intensity of homologous CT26 cells was nearly 2.2- and 1.6-fold higher than that of HeLa cells and MCF-7 cells, respectively, indicating the specific binding ability of CM-SiO₂ NPs to the homologous CT26 cells. This source cell-specific targeting property of CM-SiO₂ NPs was further evidenced by TEM in CT26, HeLa and MCF-7 cells (Fig. 3f–h and Supplementary Fig. 14), in which a large number of CM-SiO₂ NPs localized in the endocytic vesicles could be observed in CT26 cells while CM-SiO₂ NPs were rarely present in other cell lines. Taken together, our data suggested that even a partial coating of SiO₂ NPs with CT26 cells membrane was sufficient to increase the affinity of the particles to the source cancer cells, a functionality that can be attributed to the transference of cell adhesion molecules with homotypic binding properties³⁸.

a Representative CLSM images of three cell lines (CT26, HeLa and MCF-7) after 4 h incubation with SiO₂ NPs, LB-SiO₂ NPs and CM-SiO₂ NPs. The SiO₂ cores were labeled with Cy5 (red), and the cell nuclei were stained with 4’,6-diamidino-2-phenylindole (DAPI; blue). Scale bars, 20 μm. b–d Flow cytometric analysis of CT26 cells (b), HeLa cells (c) and MCF-7 cells (d) incubated with blank solution, SiO₂ NPs, LB-SiO₂ NPs and CM-SiO₂ NPs. e Quantification of the mean fluorescence intensities for the three cell lines (CT26, HeLa, and MCF-7). Data represents mean ± SD (n = 3). One-way ANOVA followed by post hoc Tukey test was used to determine the significance of data. ***p < 0.001. f–h TEM images of CT26 cells (f), HeLa cells (g), and MCF-7 cells (h) after 4 h incubation with SiO₂ NPs and CM-SiO₂ NPs. Insets below are magnified of each image in the area highlighted with the respective yellow dashed box. Scale bars, 2 μm (top); 200 nm (bottom).

We next probed whether the coating of the cancer cell membrane also affected the internalization of the NPs by macrophages. In contrast to the case of tumor cells, the results of CLSM (Supplementary Fig. 15a) and flow cytometry (Supplementary Fig. 15b, c) both indicated that the macrophage cell line RAW264.7 showed significantly reduced binding and/or internalization of CM-SiO₂ NPs as compared to that of bare SiO₂ NPs and LB-SiO₂ NPs. This was in a good agreement with the results from a previous study showing that the presence of a cell membrane coating reduced the clearance of NPs by the reticular endothelial system^39,40, suggesting that even partial cell membrane decoration seemed to suffice in improving the short half-life of traditional NPs.

Endocytic entry mechanism study of partially coated NPs

Our in vitro targeting results raised a fundamental question regarding how partially coated NPs enter the cells. To systematically address this issue, we first sought to find out the endocytic pathway of partially coated NPs by CT26 cells. To do so, we pre-treated cells with a low temperature (4 °C) and various pharmacological endocytosis inhibitors before the addition of CM-SiO₂ NPs (Supplementary Fig. 16). The uptake of CM-SiO₂ NPs was notably inhibited at 4 °C, suggesting they were being internalized via an energy-dependent endocytosis process. The cellular uptake of CM-SiO₂ NPs was reduced to ~14% in the presence of chlorpromazine (CPZ; inhibitor of clathrin-dependent endocytosis), whereas exposure to genistein (GEN; inhibitor of caveolin-dependent endocytosis) and cytochalasin D (CytD; inhibitor of macropinocytosis) had negligible effects, indicating that partially coated NPs were internalized primarily by clathrin-dependent endocytosis, generally referred to as “receptor-mediated endocytosis”⁴¹.

In receptor-mediated endocytosis, the energy obtained upon NPs’ ligands binding to cell surface receptors requires an overcoming of the deformation energy of the membrane⁴². Given that the surface of CM-SiO₂ NPs was partially covered with ligands, we next wondered whether such receptor–ligand binding strength would be strong enough to drive NPs over the energy barrier during the uptake. We did this by employing dissipative particle dynamics (DPD) simulations to simulate the receptor-mediated endocytosis of partially coated NPs (Supplementary Fig. 17; see more details in the section of Model and Simulation). Here, we defined the degree of cell membrane coating as the ratio of the cell membrane coating surface area to the total NP surface area (Supplementary Fig. 18). We first analyzed the effect of different coating degrees (0%, 15%, 30%, 50%, 80%, and 100%) on a single NP’s wrapping (Fig. 4a). Consistently, the previously reported TEM images of cell membrane-coated NPs supported the existence of partial coating^26,43,44. This phenomenon could be attributed to the support effect that induced constraints on the lipid mobility⁴⁵, as a non-constrained patch would adopt the thermodynamically favorable circular shape. The results revealed that a larger coating degree could induce more cell membrane bending around the CM-SiO₂ NP, resulting in a lower z-position below the membrane and a higher wrapping ratio (Supplementary Fig. 19). It is noteworthy that during the clathrin-mediated endocytosis process, after entrapment by clathrin-coated pits, the NP could be completely internalized into the cell via active membrane shrinkage, a process that can be triggered by clathrin and actin filaments polymerization^46,47. Based on this phenomenon, in our following simulations we chose a coating degree of 50% as the criteria to determine whether a single NP could be internalized by cells because its final position was close to the center of the phospholipid bilayer. However, the proportion of CM-SiO₂ NPs with a coating degree larger than the critical coating degree was only 7.4% (Fig. 4b), suggesting that most of the NPs were unlikely to be individually internalized by the cells. This observation was further supported by examining the TEM images of the NPs internalization at different incubation time points (Supplementary Fig. 20). These observations showed that the CM-SiO₂ NPs located in the early endosomes were mainly aggregated, while 9.5% NPs were individual in an early endosome. This value (9.5%) was comparable to the proportion (7.4%) of NPs with a coating degree greater than the critical coating degree, further supporting that the threshold value of the coating degree we selected was appropriate. Remarkably, the CT26 cells incubated with NPs at a lower concentration (1 μg/mL; Supplementary Fig. 21) still exhibited a similar ratio of individual NPs aggregated in a single early endosome to high concentration (50 μg/mL; Supplementary Fig. 20), suggesting that these NPs had to aggregate for the uptake.

a Top: TEM images of SiO₂ NPs with different cell membrane coating degrees (0%, 15%, 30%, 50%, 80%, and 100%). Scale bar, 50 nm. Bottom: the corresponding final dissipative particle dynamics (DPD) simulation snapshots of the wrapping of the CM-SiO₂ NPs by a modeled cell membrane. It shows that the wrapping of a single CM-SiO₂ NP with a higher coating degree is easier than that of an NP with a lower coating degree. b Cell membrane coating degree distribution of as-prepared CM-SiO₂ NPs, which is calculated from TEM images (n = 325). The inset shows the proportion of SiO₂ NPs with a low cell membrane coating degree (<50%). c–e Typical DPD simulation snapshots of multiple CM-SiO₂ NPs: aggregation number (n) = 2 (c), 4 (d), and 9 (e). The coating degree of each NP is 33%. The top panel (t = 0 τ) shows the setup of the simulation system and the other panels (t = 160,000 τ) displays the final equilibrated NP-membrane structure at the top view and the profile view. f Time evolution of CM-SiO₂ NPs positions along the membrane normal direction. Z represents the distance between the center of the NP and the cellular phospholipid bilayer (inset). g Comparison of positional change of CM-SiO₂ NPs with different aggregated numbers (2, 4, and 9) and coating degrees (16%, 25%, 33%, and 40%). Data represents mean ± SD (n = 20). h TEM images showing the different states of CM-SiO₂ NPs during receptor-mediated interactions with CT26 cells. Scale bars, 100 nm. i Schematic illustration of a possible endocytic entry mechanism for partially coated NPs.

Based on the above results of single NP simulation and TEM analysis, we hypothesized that the NPs with low coating degrees (<50%) aggregated together and provided more ligands to enter the cells. To test this hypothesis, we performed simulations examining the interaction between the cell membrane and the multiple CM-SiO₂ NPs with a coating degree of 33% (aggregation number: 2, 4, and 9; Fig.4c–e). Interestingly, the NP aggregates were more likely to enter the cell membranes as the aggregation number increased (Fig. 4f). The final structure of the NP-membrane showed that the aggregated NPs rotate spontaneously after entering into the membranes to promote more ligands binding with the cell receptor (Supplementary Fig. 22a-c and Supplementary Movies 1–3). Regarding the aggregated NPs as an individual particle, such rotation provided more ligands on the side of the NP aggregates, which subsequently promoted the wrapping of the NPs (Supplementary Fig. 22c). Inspired by the simulation results, we further proposed a two-dimension geometric model to predict the required coating degree for half wrapping the NPs aggregates (Supplementary Fig. 23). Under these experimental conditions, the required coating degree for two NPs was about 29%−32% and that for three NPs was about 23%−26%, which was much smaller than the one required for internalizing an individual NP (50%). It should be clear that this predicted value would be slightly smaller than the actual three-dimension (n = 4, 9) required coating degree (possible reasons for this discrepancy are given in Supplementary Note 1).

Given that the coating degree of the CM-SiO₂ NPs mostly varied from 5% to 50% (Fig. 4b), we explored the interactions between the membrane and the CM-SiO₂ NPs with different coating degrees (16%, 25%, 33%, and 40%; Fig. 4g and Supplementary Fig. 24). With an increasing aggregation number, the NPs could more easily enter into the cells and a lesser coating degree was needed to achieve 50% wrapping by the membrane. However, it was found that when the coating degree decreased to 16%, the cooperation within the aggregated NPs was significantly weakened, resulting in a failure of the internalization by the cell membrane. This is because at such a low coating degree, the rotation cannot promote binding of any extra ligands to the receptors, which only occurs at much higher coating degrees (Supplementary Fig. 22d, e). Therefore, in the case of the NPs with a low coating degree (20%), increase in the aggregation number (n = 16) cannot solely promote their internalization (Supplementary Fig. 25). Remarkably, the aggregation number of NPs that enter the cells is limited by the size of clathrin-coated vesicles (approximately 100–300 nm)⁴⁸. Taking into account the nonuniformity of the coating degrees within the NPs aggregates, the NPs with a very low coating degree (20%) may be able to enter the cell with the help of NPs with a higher coating degree (40%; Supplementary Fig. 26). Taken together, these results strongly indicate that the CM-SiO₂ NPs can also enter into the cells by a process called aggregated cooperation when the coating degree is lower than 50%. To validate the prediction emerging from our computational model, we used TEM to capture the membrane-NPs structures at the different interaction stages (Fig. 4h), which demonstrated that the dispersed NPs would aggregate after adhering onto the membranes and subsequently first induce membrane wrapping and then internalization of the NP aggregates. Compared with the individually internalized NPs (Supplementary Fig. 27), multiple NPs are wrapped within a much larger endosome (200–400 nm) and thus need more proteins to actively bend the membrane.

Finally, putting the theoretical modeling and experimental results together, we proposed a possible endocytic entry mechanism for the partially coated NPs (Fig. 4i). The NPs with a high coating degree (≥50%) were able to undergo individual cell entry. By contrast, those NPs with a low coating degree (20%−50%) needed to first aggregate with each other on the cell’s surface before they could synergistically enter into cells. The aggregation number of NPs with low coating degrees highly depends on their degree of cell membrane coating. The NPs with a very low coating degree (<20%) may be hardly engulfed by the membranes even with cooperation with each other, but they are able to be engulfed when they can aggregate with NPs with higher coating degrees (e.g., ≥40%).