Carbon nano-onion-mediated dual targeting of P-selectin and P-glycoprotein to overcome cancer drug r

The transmembrane P-glycoprotein (P-gp) pumps that efflux drugs are a major mechanism of cancer drug resistance. They are also important in protecting normal tissue cells from poisonous xenobiotics and endogenous metabolites. Here, we report a fucoidan-decorated silica-carbon nano-onion (FSCNO) hybrid nanoparticle that targets tumor vasculature to specifically release P-gp inhibitor and anticancer drug into tumor cells. The tumor vasculature targeting capability of the nanoparticle is demonstrated using multiple models. Moreover, we reveal the superior light absorption property of nano-onion in the near infrared region (NIR), which enables triggered drug release from the nanoparticle at a low NIR power. The released inhibitor selectively binds to P-gp pumps and disables their function, which improves the bioavailability of anticancer drug inside the cells. Furthermore, free P-gp inhibitor significantly increases the systemic toxicity of a chemotherapy drug, which can be resolved by delivering them with FSCNO nanoparticles in combination with a short low-power NIR laser irradiation.

Resistance to chemotherapy or molecularly targeted therapies greatly impedes the efficacy of cancer treatments1,2. The drug-resistant cells can survive conventional cancer therapies and may cause tumor recurrence and/or development of distant metastases, which are the major reasons for cancer-related death3,4. Consequently, multidrug-resistant cancer cells have attracted a great deal of attention in the field of oncology for several decades5. Many different mechanisms may enable or promote multidrug resistance, but drug efflux pumps have been considered as one of the key mechanisms, which are located on the cell membrane to efflux anticancer drugs from cancer cells6,7. As a result, various kinds of nanoparticles have been developed to deliver anticancer drugs into multidrug-resistant cancer cells because the drug efflux pumps cannot efflux nanoparticles from cancer cells8,9. Unfortunately, studies have shown that drugs can still be easily pumped out of cells once they are released from nanoparticles, which limits the retention/bioavailability of drugs inside multidrug-resistant cancer cells for effective cancer destruction10,11. Thus, direct inhibition of the drug efflux pumps may be an effective strategy for overcoming the multidrug resistance of cancer.

An important advancement in the understanding of the drug efflux pumps is the identification of the transmembrane P-glycoprotein (P-gp) in the plasma membrane of multidrug-resistant cancer cells12,13. The P-gp multidrug transporter is an ATP-binding cassette (ABC) protein that has been implicated in multiple types of multidrug-resistant cancers7,14. Thus, inhibiting the function of P-gp could be an effective approach to improve the retention/bioavailability of therapeutic agents inside cancer cells. This can be done by using inhibitor to block the P-gp efflux. Indeed, coadministration of the therapeutic agents and P-gp inhibitors in their free forms may reduce the expulsion of the agents by P-gp and increase the intracellular drug concentration15,16. Nevertheless, P-gp pumps also play a crucial physiological role in protecting tissues from poisonous xenobiotics and endogenous metabolites that are widely distributed throughout the body17,18,19. For example, the P-gp pumps are a major mechanism of the blood-brain barrier (BBB) and intestinal wall to minimize the entry of any toxic substances into the brain and blood, respectively (Fig. 1a)20,21. In other words, cellular processes in the body are benefiting from the protection of P-gp and nontargeted systemic administration of P-gp inhibitor in its free form may cause unexpected side effects22,23,24.

To tackle these challenges, here we report a tumor vasculature targeted nanovesicle to codeliver P-gp inhibitor and chemotherapeutic drug. This nanoplatform (FSCNO) is made of carbon nano-onion (CNO) and silica (S) surface-decorated with fucoidan (F), which can specifically bind to P-selectin that is overexpressed on tumor vasculature (Fig. 1b, right)25,26. As a result of this active tumor vasculature targeting and their passive tumor-targeting capacity due to the enhanced permeability and retention (EPR) effects of tumor vasculature compared to normal vasculature27, the FSCON nanoparticles are anticipated to preferentially accumulate in tumor rather than normal tissues after systemic administration in vivo (Fig. 1b, right). Moreover, the release of P-gp inhibitor and chemotherapeutic drug from FSCON nanoparticles is negligible in normal tissue in the absence of an external stimulus (low-power laser irradiation). By minimizing the accumulation of FSCON nanoparticles in normal organs and keeping the inhibitors inside the nanoparticles, the function of P-gp efflux in normal organs to exclude poisonous xenobiotics and endogenous metabolites is minimally affected (Fig. 1b, c, left). In contrast, the release of P-gp inhibitor and chemotherapeutic drugs can be triggered and precisely controlled with low-power near infrared (NIR, 800 nm) laser irradiation in tumor to overcome the multidrug resistance of tumor cells (Fig. 1b, c, right).

Preparation and characterization of nanoparticles

Carbon nano-onions (CNOs), also known as multilayer fullerenes, have been reported for cancer cell imaging28,29. Unfortunately, due to their highly hydrophobic nature, CNOs have poor solubility in water or even commonly used organic solvents (e.g., benzene and toluene) and they tend to form large aggregates in these solvents (Supplementary Fig. 1a). This greatly limits the biomedical applications of CNOs. To address this issue, CNOs were treated with tetraethyl orthosilicate (TEOS) to form silica-CNO (SCNO) nanoparticles (Fig. 2a) with increased solubility and drug loading capacity. In order to decorate them with fucoidan for targeting tumor vasculature, the SCNO nanoparticles were further reacted with (3-Aminopropyl) trimethoxysilane (APTMS) to produce the ASCNO intermediates. Fucoidan was then coated on the surface of the ASCNO intermediates through its electrostatic interaction with APTMS to form the resultant FSCNO nanoparticles (Fig. 2a). Transmission electron microscopy (TEM) images of CNOs are given in Fig. 2b. The nanoparticles readily form aggregates and the nano-onion structure (~50 nm in diameter) can be observed only at the edge of the aggregates at higher magnification (right, Fig. 2b). The onion-like structure is observable in the high-resolution TEM (HRTEM) images (Supplementary Fig. 2). Interestingly, the hybrid SCNO nanoparticles are smaller than the CNOs and their nanostructure can be clearly observed as shown in Supplementary Fig. 3. This is probably due to the reaction between TEOS and CNO according to the Fourier-transform infrared spectroscopy (FTIR) spectrum (i.e., the formation of the Si–O–C bond, Supplementary Fig. 4a), making the CNO structure more compact. After coating the SCNO nanoparticles with fucoidan, the FSCNO nanoparticles of ~20 nm with clear nanostructure and little aggregation can be obtained (Fig. 2c). The successful decoration of fucoidan is confirmed with the surface zeta potential data of the SCNO, ASCNO, and FSCNO nanoparticles. As shown in Supplementary Fig. 5, the zeta potential is negative for the SCNO nanoparticles at room temperature (−23.7 ± 1.8 mV) but becomes positive (38.5 ± 2.4 mV) after modifying them with positively charged APTMS to form the ASCNO intermediates. It returns to negative (−38.0 ± 4.1 mV) after coating the ASCNO nanoparticles with the negatively charged fucoidan, suggesting fucoidan is successfully decorated on the FSCNO nanoparticles. Importantly, unlike the CNOs, the resultant FSCNO nanoparticles have high solubility and stability in both deionized water and cell culture medium (Supplementary Fig. 1a). This is further confirmed with the Tyndall effect that is a result of light scattering by nanoparticles, by shining a red laser beam through various solutions of the CNOs and FSCNO nanoparticles. As shown in Supplementary Fig. 1b, a discernible light track can be observed only in the solutions of FSCNO nanoparticles (in deionized water and cell culture medium).

To inhibit the pumping capability of P-gp, a third-generation inhibitor (HM30181A, HM in short) was used in this study that can selectively and potently inhibit the P-gp function30. However, HM is insoluble in water and was used only through oral delivery in previous studies to inhibit P-gp in the intestinal endothelium. It has never been used to inhibit the P-gp in multidrug-resistant cancer cells. Both the hydrophobic HM and hydrophilic doxorubicin hydrochloride (DOX, a widely used chemotherapy drug) were encapsulated in the FSCNO nanoparticles to form FSCNO-DH nanoparticles, with an encapsulation efficiency of 68.5 ± 1.7% for HM and 89.0 ± 1.5% for DOX at a feeding ratio of 1:20 (HM or DOX: nanoparticles). The corresponding loading content of HM and DOX is 3.3 ± 0.1% and 4.3 ± 0.1%, respectively. To confirm the FSCNO nanoparticles have spaces for encapsulating drugs, nitrogen gas (N2) sorption measurements were carried out to investigate both the surface area and pore size distribution within the nanoparticles. The N2 adsorption isotherm of the FSCNO nanoparticles (Supplementary Fig. 6a) is typical of the type IV isotherm that indicates the micro- or mesoporous feature according to the International Union of Pure and Applied Chemistry (IUPAC) classification31. The surface areas within the FSCNO nanoparticles are 250.3 m2 g−1 and the average pore size is 3.2 nm with two peaks at ~1.4 nm and ~3.1 nm (Supplementary Fig. 6b). This porous structure within the FSCNO nanoparticles renders their capability of drug loading. The loading of drugs inside the nanoparticles rather than on their external surface is further supported by the zeta potential data. As shown in Supplementary Fig. 7a, the zeta potential of the two drugs-laden FSCNO-DH nanoparticles is similar to that of the FSCNO nanoparticles with no drug (Supplementary Fig. 5c), suggesting the drugs are not coated on the surface of the nanoparticles.

The FSCNO-DH nanoparticles is ~28 nm (in diameter) according to the dynamic light scattering (DLS) measurements (Fig. 2d). The successful encapsulation of HM and DOX was further confirmed with UV-Vis absorbance spectrometry. As shown in Fig. 2e, DOX has a strong peak at ~480 nm whereas HM30181A has strong absorbance between 350 and 400 nm. Compared with FSCNO nanoparticles, the DOX-laden nanoparticles (FSCNO-D) have a strong peak at ~480 nm due to DOX in the nanoparticles (Fig. 2f). Similarly, compared to the FSCNO-D nanoparticles, the absorbance between 350 and 400 nm is elevated for the FSCNO-DH nanoparticles because of the strong absorbance of HM30181A in this region (Fig. 2f). The fluorescence spectra show that FSCNO-D nanoparticles and free DOX have the same fluorescence emission peak at ~590 nm after excitation with a 480 nm laser (Fig. 2g), although the fluorescence intensity is decreased compared with free DOX. The latter is probably because of self-quenching of fluorescence and π–π stacking between FSCNO nanoparticles and DOX in the FSCNO-DH nanoparticles32. HM does not have fluorescence and has no effect on the fluorescence of DOX in FSCNO-DH nanoparticles (Fig. 2g). Collectively, these data show that both P-gp inhibitor and chemotherapeutic drug are successfully encapsulated in the FSCNO-DH nanoparticles.

Photothermal effect of FSCNO nanoparticles

Due to its capability of deep tissue penetration, NIR laser has been used as an external stimulus to trigger the release of drugs encapsulated in gold and carbon nanomaterials with strong absorption in the NIR region33. Therefore, we investigated the UV-Vis absorbance spectrum of FSCNO nanoparticles at different concentrations. Interestingly, the absorbance of FSCNO nanoparticles decreases only slightly from ultraviolet to NIR region (Fig. 3a). We also synthesized FSCNT and FSGO nanoparticles with carbon nanotube (CNT) and graphene oxide (GO) that have been widely studied for photothermal applications, for which the same procedure given in Fig. 2a were used except that CNOs was replaced with CNTs and GOs, respectively. As shown in Supplementary Fig. 8, scanning electron microscopy (SEM) image of CNTs shows their tube structure is changed after reacted with silica and coated with fucoidan to form the FSCNT nanoparticles. Although the sheet structure of GO is difficult to identify in the SEM image because they are very thin, the thickness of the sheet structure is significantly increased and clearly visible after reacted with silica and coated with fucoidan to form the FSGO nanoparticles (Supplementary Fig. 8). The FTIR spectra of the SCNT and SGO exhibit absorption bands corresponding to Si–O–Si (νs at 800 and νas at 1080–1200 cm−1, where the subscripts s and as represent symmetric and asymmetric, respectively) and Si–O–C (νs at 954 and νas at 1070 cm−1), suggesting the formation of silica structure and the reaction between TEOS with CNT or GO (Supplementary Fig. 4b–c). The absorbance of FSCNT and FSGO nanoparticles decreases significantly from the ultraviolet to NIR region for all the concentrations studied (Fig. 3a). In particular, the absorbance of the FSCNO nanoparticles in the NIR region is 2–3 times higher than that of the FSCNT and FSGO nanoparticles when the concentration is more than 0.25 mg ml−1, implicating the superior photothermal property of the FSCNO nanoparticles. Indeed, the temperature of FSCNO nanoparticle solution increases faster than that of the FSCNT or FSGO nanoparticle solution at the same concentrations under NIR laser irradiation (1 W cm−2, Fig. 3b). The photothermal effects of the FSCNO nanoparticles is further demonstrated in centrifuge tube with near infrared thermography (Supplementary Fig. 9). Interestingly, the photothermal effect of the FSCNO nanoparticles can also be observed in their dry state. For this study, the FSCNO nanoparticle solutions at various concentrations (1.8–280 ng mm−2) was dropped on paper, completely dried, and irradiated with NIR laser (1 W cm−2). As shown in Fig. 3c, the paper was observed to ignite in ~1–3 s in the area with 7 ng mm−2 or more FSCNO nanoparticles, while burning was seen at ~10 s in the area with 3.5 ng mm−2 nanoparticles and it was not seen in the area with 1.8 ng mm−2 nanoparticles in 20 s. The stability of the photothermal effects was examined with three cycles of NIR irradiation (1 W cm−2, on-off: 2–10 min) of the FSCNO nanoparticle solutions continuously. As shown in Fig. 3d, the temperature change in the solutions is similar for all the three cycles. This is not surprising as the laser does not affect the integrity of the nanoparticles (Supplementary Fig. 10). Importantly, the release of P-gp inhibitor and chemotherapeutic drug from the nanoparticles is minimal at least for 10 h and can be precisely controlled with NIR irradiation (Fig. 3e), to minimize their uncontrolled release in normal tissue.

P-gp inhibition and anticancer capacity of FSCNO-DH nanoparticles in vitro

Two types of multidrug-resistant human cancer (NCI/ADR-RES and A2780ADR) cells were used in this study together with a nondrug resistant type of human cancer (OVCAR-8) cells for control. We first confirmed the expression of P-gp in NCI/ADR-RES and A2780ADR cells but not the OVCAR-8 cells with confocal microscopy. As shown in Supplementary Fig. 11, immunostaining of the P-gp can be clearly observed in the cell membrane of both NCI/ADR-RES and A2780ADR cells but not the OVCAR-8 cells. Presumably due to the efflux function of the P-gp pumps, the fluorescence of DOX is not observable in both NCI/ADR-RES and A2078ADR cells after incubated them with free DOX for 6 h. In contrast, free DOX (10 μg ml−1) can be observed predominantly in the cell nuclei when it is co-incubated with various concentrations of free HM (1 and 5 μg ml−1 for Supplementary Fig. 12 and 10 μg ml−1 for Fig. 4a), and the fluorescence in nuclei is stronger for cells treated with 5 or 10 μg ml−1 HM than 1 μg ml−1 HM. These data indicate that the HM indeed can inhibit the function of P-gp pumps in a dose-dependent manner. Next, we treated NCI/ADR-RES and A2078ADR cells with DOX-laden FSCNO (FSCNO-D) nanoparticles and irradiated with laser (FSCNO + L, 0.5 W cm−2). Although the overall DOX fluorescence is increased after laser irradiation, the DOX fluorescence in nuclei is still weak (Fig. 4b). This is presumably because the P-gp could still pump out the released drugs. In contrast, the DOX fluorescence in nuclei is highly increased after NIR irradiation of cells treated with the FSCNO-DH nanoparticles (Fig. 4c). As a control, free DOX can enter the OVCAR-8 cells regardless of treatment with or without HM (Supplementary Fig. 13) and there is no obvious difference in the distribution of DOX in the OVCAR-8 cells treated with FSCNO-D and FSCNO-DH (Supplementary Fig. 13), suggesting the HM has minimal impact on the drug uptake of the nondrug resistant cancer cells. These results are further confirmed with flow cytometry. As shown in Supplementary Fig. 14a–b, the DOX in NCI/ADR-RES and A2780ADR cells is minimal if they were treated with free DOX and greatly increased when free DOX is combined with HM (DOX+HM) for treating the cells. The DOX fluorescence intensity is stronger in FSCNO-DH+L than FSCNO-DH treated cells, probably due to the self-quenching effect of DOX fluorescence in the FSCNO nanoparticles (Fig. 2g). Similar to the microscopy data, HM has minimal effect on the DOX uptake of OVCAR-8 cells (Supplementary Fig. 14c).

The in vitro anticancer capacity of FSCNO-DH nanoparticles was conducted with NCI/ADR-RES, A2780ADR, and OVCAR-8 cells. Compared with OVCAR-8 cells, free DOX is inefficient to kill NCI/ADR-RES and A2780ADR cells (Fig. 4d), due to the drug-resistant nature of the NCI/ADR-RES and A2780ADR cells. The efficacy of DOX in killing NCI/ADR-RES and A2780ADR cells is significantly enhanced when coadministrated with 1–10 μg ml−1 HM (dissolved with DMSO at 1 mg ml−1 in stock). Of note, the efficacy of DOX in killing OVCAR-8 cells also increased when coadministrated with HM (5 and 10 μg ml−1), which is probably due to the DMSO used to dissolve HM. As shown in Supplementary Fig. 15, DMSO at the concentration of 5 or 10 μl ml−1 could kill some of all the three types of cells. Overall, HM of 5 ug ml−1 is enough to block the P-gp function and used for the following studies. Although using FSCNO-D nanoparticles can enhance the anticancer effect as a result of improved delivery of DOX into the multidrug-resistant cells, the enhancement is limited for the two drug-resistant cells (Fig. 4d). To check the effect of NIR irradiation on the in vitro anticancer capacity of the nanoparticles, the laser powers at 0.05, 0.2, 0.5, and 1 W cm−2 were studied. The results indicate that only the highest power (1 W cm−2) combined with FSCNO-D nanoparticles can effectively kill all the three different types of cancer cells. However, this is not because of the anticancer effect of DOX as the empty FSCNO nanoparticles could effectively kill all the three different types of cells under the laser irradiation at 1 W cm−2(Supplementary Fig. 16a). Since the NIR irradiation alone does not compromise the viability of cells at the different laser powers (Supplementary Fig. 16b), the observed anticancer effect is probably due to the aforementioned superior photothermal effect of the FSCNO nanoparticles (Fig. 3 and Supplementary Fig. 9). In contrast, FSCNO-DH nanoparticles (HM: 5 μg ml−1) could efficiently destroy all the three types of cells at a laser power as low as 0.2 W cm−2 (Fig. 4d). This is probably due to the superior photothermal effect of FSCNO nanoparticles, which could effectively trigger the release of HM to inhibit the P-gp and enhance the retention of DOX in the drug-resistant cells for the observed effective destruction of the cells. Therefore, the laser power of 0.2 W cm−2 is used for the following studies and the photothermal effect of the FSCNO nanoparticles at various concentrations (0.06–0.50 mg ml−1) under this laser power is given in Supplementary Fig. 17.

Tumor vasculature targeting

To deliver P-gp inhibitor and DOX into tumor rather than normal tissue, the FSCNO nanoparticles were designed to target the tumor vasculature through P-selectin as aforementioned (Fig. 1b). To confirm this, we activated the human umbilical vein endothelial cells (HUVECs) with human tumor necrosis factor-alpha (TNF-α) to mimic the tumor vasculature34. As shown in Fig. 5a, the expression of P-selectin is evident in activated HUVECs (aHUVECs) while it is negligible in the HUVECs without activation. The targeting capability of SCNO-DH or FSCNO-DH nanoparticles was then investigated by incubating them with 2D-cultured aHUVECs for 3 h at 4 °C. Indeed, the FSCNO-DH nanoparticles could efficiently bind with P-selectin on the aHUVECs while the binding between SCNO-DH nanoparticles and aHUVECs is not observable (Fig. 5b), showing the crucial role of the fucoidan on the surface of the FSCNO-DH nanoparticles in targeting the aHUVECs. The latter is probably via the binding between fucoidan and P-selectin overexpressed on the aHUVECs. This is further supported by the data showing less P-selectin antibody could bind with the FSCNOs nanoparticle-treated aHUVECs than the SCNO-DH nanoparticle-treated aHUVECs (Fig. 5b), probably because some of the P-selectins have been occupied by the FSCNO-DH nanoparticles and hard for the antibody to access.

To further confirm the targeting capability of FSCNO nanoparticles, we fabricated a polydimethylsiloxane (PDMS) microfluidic device for culturing the aHUVECs and HUVECs under continuous dynamic perfusion (Fig. 5c and Supplementary Fig. 18a) to mimic the in vivo blood perfusion in blood vessels. The HUVECs and aHUVECs were cultured in two different and parallel chambers in this device with independent inlets and outlets (inlet2–outlet2 or inlet3–outlet3) for sample injection and the two chambers are separated with a fluid-passable partition wall (Fig. 5c and Supplementary Fig. 18b). Computational modeling data show that fluid flow is almost independent in each chamber during cell injection from inlets 2 or 3 (Supplementary Fig. 18c), making this device ideally suitable for culturing two kinds of cells at same time. The targeting capability was studied by injecting cell culture medium with FSCNO-DH nanoparticles into the device through inlet1 for exiting via outlet1. As a result, the velocity of injected nanoparticle solution is homogeneous and nearly the same in the two chambers (Fig. 5c, ~0.6 mm s−1) which is close to the blood flow speed in capillaries35. With this device, green fluorescence protein (GFP)-expressing aHUVECs and HUVECs were cultured in the two chambers for 12 and 24 h, respectively. As shown in Fig. 5d(where the fluid passes aHUVECs first) or Fig. 5e (where the fluid passes HUVECs first), the aHUVECs and HUVECs can form tube-like structure in the device after culturing for 12 h and particularly 24 h. When perfusing cell culture medium dissolved with FSCNO-DH nanoparticles for 3 h in the device, the FSCNO-DH nanoparticles can target only the aHUVECs regardless of the position of the two chambers with HUVECs and aHUVECs (Fig. 5d, e), confirming the aHUVECs-targeting capability of FSCNO-DH nanoparticles under dynamic flow conditions.

The stability of the FSCNO-DH nanoparticles was examined by incubating the FSCNO-DH nanoparticles in medium for 2 days and then investigating their capability of targeting the aHUVECs versus HUVECs under both static culture in Petric dish and dynamic culture in the aforementioned microfluidic device. As shown in Supplementary Fig. 19a for static culture or 19 b for dynamic culture of aHUVECs, the FSCNO-DH nanoparticles after incubating with cell culture medium can still target the aHUVECs regardless of the culturing conditions. This is probably due to the strong interaction between the sulfonyl group (in fucoidan) and amino group (on ASCNO nanoparticles) by reducing the nucleophilicity and basicity of the amino group to form sulfonamide-like stable interaction36,37, which ensures the stability of fucoidan on the nanoparticle surface. In order to show the fucoidan density is important for the targeting effect, FSCNO nanoparticles with reduced/low fucoidan (FlowSCNO nanoparticles) were prepared. As shown in Supplementary Fig. 20a, the zeta potential of the FlowSCNO-DH nanoparticles is increased to −11.2 ± 3.8 mV if the amount of fucoidan during the nanoparticle preparation is reduced from 5 to 2 mg. Consequently, the targeting capability of the FlowSCNO-DH nanoparticles is weakened compared with the FSCNO-DH nanoparticles (Supplementary Fig. 20b versus Fig. 5b or Supplementary Fig. 19a). It is worth noting that P-selectins are expressed in all the tumor cells used in this study (Supplementary Fig. 21), suggesting the FSCNO-DH nanoparticles can also target tumor cells.

Next, we investigated the in vivo tumor-targeting capability of the FSCNO nanoparticles by encapsulating indocyanine green (ICG, an NIR dye) in them to form the ICG-laden FSCNO (FSCNO-I) nanoparticles for in vivo imaging. The encapsulation of ICG does not change the zeta potential of the FSCNO nanoparticles (Supplementary Fig. 7b versus Supplementary Fig. 5c). NCI/ADR-RES, A2780ADR, and OVCAR-8 cells (1 × 106 cells per mouse) were injected subcutaneously at the left dorsal side of the upper hindlimb of mice. Comparatively, NCI/ADR-RES, A2780ADR and OVCAR-8 cells were co-injected with aHUVECs at the right side of the mice to study the targeting effects (N + EC, A + EC, and O + EC in short, 1 × 106cancer cells and 2 × 106 aHUVECs per mouse, Fig. 6a). The mice were used for imaging when the tumor volume is ~100 mm3. At 3 or 6 h after intravenous injection, free ICG (50 µg per mouse in 100 µl of saline) is observable throughout the mice body without specific tumor accumulation (Fig. 6b). In contrast, enhanced fluorescence of ICG is observable in the tumor areas at both sides (labeled with L for left and R for right, respectively) for mice treated with the FSCNO-I nanoparticles (ICG: 50 µg per mouse in 100 µl of saline) at both 3 and 6 h after i.v. injection (Fig. 6b). To confirm the targeting capability of FSCNO nanoparticles, various organs and tumors from both sides were harvested for ex vivo imaging after sacrificing the mice at 6 h. The ICG fluorescence in tumors of ICG-treated mice is minimal (Fig. 6c). In contrast, strong fluorescence of ICG can be observed in the tumors of FSCNO-I nanoparticle-treated mice formed with all the three different types of cancer cells with/without aHUVECs (Fig. 6c). Importantly, the intensity of ICG fluorescence in tumors of the N+EC, A+EC, and O+EC groups of mice treated with the FSCNO-I nanoparticles are significantly stronger than that in tumors grown without aHUVECs or free ICG treatment (Fig. 6c, d), supporting the tumor vasculature-targeting capability of the FSCNO nanoparticles observed in vitro. The tumors were then cryo-sectioned and stained with CD31 antibody to examine the distribution of FSCNO-I nanoparticles via the ICG fluorescence. As shown in Fig. 6e–g and Supplementary Fig. 22, the FSCNO-I nanoparticles can target the tumor blood vessels formed from both the host endothelial cells and injected aHUVECs.

In vivo antitumor capacity

To check the antitumor capacity of FSCNO-DH nanoparticles, two drug-resistant (NCI/ADR-RES and A2780ADR) and one nondrug resistant (OVCAR-8) tumor models were used for in vivo studies. Similar to the in vivo imaging studies, co-injection of aHUVECs were included in each tumor model to understand how the tumor vasculature-targeting capability of the FSCNO-DH nanoparticles may affect their antitumor capacity in vivo. As shown in Fig. 7a, for NCI/ADR-RES tumors, the co-injection of aHUVECs (With EC, the Saline group) does not significantly influence the tumor growth compared with injection of the NCI/ADR-RES cells alone (W/O EC, the Saline group) during the 7 weeks of experiments. Unlike the in vitro data that shows the combination of free DOX and HM (DOX+HM) may kill the cancer cells better than free DOX alone in Petri dish where bioavailability of the two agents is high (Fig. 4d), the DOX + HM treatment has minimal effects on the growth of tumors in vivo. This is not surprising because the NCI/ADR-RES tumor cells are multidrug-resistant and the bioavailability of the two agents in tumor is low when they applied systemically in their free form, which may be overcome by encapsulating them in the tumor-targeting FSCNO nanoparticles for delivery. Indeed, a decrease in tumor volume is observable for the treatment of FSCNO-DH nanoparticles and the treatment of FSCNO-D nanoparticles with laser irradiation (FSCNO-D+L), especially for the With EC condition (Fig. 7a). This is probably because more FSCNO nanoparticles could accumulate in the tumors grown from both cancer cells and aHUVECs for the condition as shown in Fig. 6c. Importantly, the treatment of FSCNO-DH nanoparticles combined with NIR irradiation (FSCNO-DH+L) gives the best antitumor efficacy. Since the mild photothermal effect has minimal impact on the viability of tumor cells (Supplementary Fig. 16a), this result further confirms that controlled release of P-gp inhibitor is crucial to overcome the multidrug-resistant capability of NCI/ADR-RES cells for killing them in vivo. Similarly, the FSCNO-DH+L treatment is more effective in killing tumors in the With EC group (Fig. 7a), confirming the importance of tumor vasculature targeting for enhancing the efficacy of nanoparticle-mediated cancer therapy.

Unlike NCI/ADR-RES tumor cells, co-injection with aHUVECs (With EC, the Saline group) could significantly enhance the tumor growth of the drug-resistant A2780ADR cells compared with injection of the A2780ADR cells alone (W/O EC, the Saline group, Fig. 7a and Supplementary Fig. 23a). These tumors grew much faster than the NCI/ADR-RES tumors and reached the maximum size allowed in our animal protocol after only 16 days. Similar to that observed for the NCI/ADR-RES tumors though, the DOX+HM treatment has minimal impact on the growth of the A2780ADR tumors with EC or W/O EC, both the FSCNO-DH and FSCNO-D+L treatments could inhibit the tumor growth, and the FSCNO-DH + L treatment gives the best therapeutic outcome among all the treatments (Fig. 7a). This further confirms the importance of controlled release of P-gp inhibitor in overcoming cancer resistance in vivo. Of note, there is no significant difference between the With EC and W/O EC groups for the FSCNO-DH+L treatment. This is mainly because the tumors in With EC group grow faster than that in the W/O EC group. In other words, the decrease in tumor volume for the With EC group after the FSCNO-DH+L treatment is much more than that for the W/O EC group (Supplementary Fig. 23b), demonstrating the importance of tumor vasculature targeting for enhancing the efficacy of nanoparticle-mediated cancer therapy.

Interestingly, co-injection of aHUVECs (With EC, the Saline group) does not significantly influence the nondrug resistant OVCAR-8 tumor growth compared with injection of the OVCAR-8 cells alone (W/O EC, the Saline group). As expected, for the nondrug-resistant OVCAR-8 tumors, there is no significant difference between the FSCNO-D+L and FSCNO-DH+L treatments (Fig. 7a), because the P-gp expression is minimal in these cells. It is also not surprising that the DOX+HM treatment has minimal impact on the growth of the A2780ADR tumors with EC or W/O EC. However, the FSCNO-DH treatment is more effective than the DOX+HM treatment in inhibiting the tumor growth, presumably because of the improved bioavailability of DOX delivered with the FSCNO-DH nanoparticles. Importantly, all the nanoparticle formulations are more effective in killing the tumors in the With EC than the W/O EC group, further demonstrating the importance of tumor vasculature targeting for enhancing the efficacy of nanoparticle-mediated cancer therapy. The aforementioned information was further confirmed with both the size and weight of the tumors from the various conditions and groups (Fig. 7b, c). Hematoxylin and eosin (H&E) stain shows extensive necrosis or shrinking in the tumors from the FSCNO-DH+L group while tumors from all the other groups are more viable (Fig. 7d).

Lastly, the body weights of mice with the FSCNO-DH+L treatment were stable during the whole observation period (Supplementary Fig. 24a). This may be due to the controlled drug release inside tumors and minimal drug release outside tumors (Fig. 3e) as the body weight is stable for all mice treated with all the nanoparticle formulations (Supplementary Fig. 24a). Moreover, there is no evident damage to major organs collected from mice in the FSCNO-DH+L group as compared with the saline group, according to the H&E staining data (Supplementary Fig. 24b). In contrast, significantly decreased body weights of mice treated with DOX+HM were observed during the experiments, suggesting the DOX+HM treatment could cause systemic toxicity. This is confirmed by the H&E staining data showing cellular damage (vacuolization) in the heart and apoptotic hepatocytes in the liver of mice treated with DOX+HM. Interestingly, our previous studies show that free DOX alone at the same dose as that used in this study does not induce evident decrease in body weight of mice11,38, suggesting that the free HM co-injected with free DOX significantly increase the systemic toxicity of the free DOX. This might be because the injection of free HM could cause the dysfunction of the P-gp pumps in normal tissues. It is worth noting that the NIR laser irradiation alone could increase the temperature of tissues. Due to the superior photothermal effect of FCNO nanoparticles, an NIR laser power as low as 0.2 W cm−2 was used for the in vivo studies in this work. This could avoid the potential damage to normal tissue (with no nanoparticles) surrounding the tumor at a higher power (e.g., 1 W cm−2, Supplementary Fig. 25). Collectively, these data show the FSCNO-DH nanoparticles can improve not only the efficacy but also the safety of cancer chemotherapy via targeting and inhibiting the P-gp pumps.

Mesoporous silica nanoparticles have been used to encapsulate various hydrophilic and hydrophobic drugs39,40,41. In this study, the sequence of encapsulating DOX and HM in the FSCNO nanoparticles was optimized. The encapsulation efficiency of HM is only ~24% if mixing HM with the FSCNO nanoparticles first, which is much lower than that (~69%) of the optimized method of mixing DOX with the nanoparticle first. This suggests that DOX might have some interactions with HM to enhance the encapsulation of HM inside the FSCNO nanoparticles. The encapsulation of DOX is probably because of the electrostatic interaction between DOX (positively charged) and silica (negatively charged) and the π–π stacking interaction between DOX and CNO. Although the exact mechanisms for the encapsulation of hydrophobic agents like HM in silica are still not well understood42, one possible mechanism is the electrostatic interaction between the negatively charged silica and the local positive charge (e.g., the amino group) in the HM molecule. Another possible mechanism is the π-π stacking between the naphthalene-like structure in HM and the tetracene structure in DOX and CNO. This is supported with the FTIR data shown in Supplementary Fig. 26a. The aromatic bonds at 1582 cm−1 can be observed only in the DOX, FSCNO-D, and FSCNO-DH groups, suggesting successful encapsulation of DOX in the nanoparticles. Similarly, methyl and carboxyl peaks at 1438 and 1407 cm−1 are observed mainly in the HM, FSCNO-H, and FSCNO-DH groups, confirming the existence of HM in the nanoparticles. Interestingly, multiple peaks associated with DOX and HM disappear or reduce after they are encapsulated inside the nanoparticles (i.e., FSCNO-D, FSCNO-H, and FSCNO-DH). This might be due to the π–π stacking between FSCNO nanoparticles and DOX/HM that decreases the molecule vibrations. Because the disappeared or reduced peaks associated with DOX or HM are not evidently affected if they are simply mixed with the FSCNO nanoparticles for the measurements (i.e., FSCNO+D, FSCNO+H, or FSCNO+D+H, Supplementary Fig. 26b). Moreover, the π–π stacking between FSCNO nanoparticles and DOX is supported by the fluorescence spectra shown in Fig. 2g. This is consistent with the literature43,44, showing the DOX fluorescence intensity decreases after it is encapsulated in nanoparticles that have π-stacking interactions with DOX, compared with free DOX.

For controlled drug release using NIR laser, our data show that the triggered drug releases from silica-fullerene (note: fullerene is a carbon nanomaterial as with nano-onion) hybrid nanoparticles is mainly because of the NIR laser-induced temperature increase in a previous study38. Similarly, the temperature in the FSCNO-DH nanoparticles should increase during the laser irradiation. As a result, the Brownian motion of all molecules is increased, which may destabilize the electrostatic and π–π stacking interactions between HM/DOX and the surface within the nanoparticles. This may free the HM and DOX for them to diffuse out of the nanoparticles under concentration gradient.

When controlling the drug release, HM and DOX are triggered to release simultaneously from the FSCNO nanoparticles during the NIR irradiation. Nonetheless, HM can quickly, selectively, and potently inhibit the P-gp function. As a result, although some DOX may be pumped out of cells before the function of all P-gps is inhibited, it could re-enter the cells after the HM binds with all the P-gps quickly. To support this, the multidrug resistance cells (NCI/ADR-RES and A2780ADR) were either pre-treated with HM for 30 min (HM-DOX) or without any HM pre-treatment (HM+DOX) before incubating them with both HM and DOX for 3 h to examine the DOX uptake by the cells. As shown in Supplementary Fig. 27a, b for NCI/ADR-RES and A2780ADR cells, respectively, the intracellular fluorescence intensity of DOX for the HM-DOX treatment is similar to that for the HM+DOX treatment. Similarly, the cytotoxicity of the HM-DOX treatment is not significantly different from that of the HM+DOX treatment (Supplementary Fig. 28a,b). These data confirm that releasing HM and DOX simultaneously from the FSCNO nanoparticles is of significance for overcoming cancer drug resistance.

Lastly, the purpose of co-injecting aHUVECs with tumors cells was to mimic the condition of tumors with various degree of P-selectin expression in this study. This is also helpful to confirm the importance of the P-selectin targeting capability of the FSCNO nanoparticles for enhancing cancer therapy. Importantly, the FSCNO-DH+L treatment can still significantly inhibit the growth of the NCI/ADR-RES, A2780ADR, and OVCAR-8 tumors without co-injection of the aHUVECs, compared with the treatments of free drug, FSCNO-D+L, and FSCNO-DH. In other words, the FSCNO-DH nanoparticles with laser irradiation can efficiently inhibit the growth of multidrug-resistant tumors in general and is even more effective for tumors with high P-selectin expression or ample blood vessels.

In summary, we synthesized a fucoidan-decorated silica and CNO hybrid (FSCNO) nanoparticle for targeting the tumor vasculature and P-gp overexpressed on drug-resistant cancer cells. We further revealed the superior light absorption at the NIR region of CNO (in comparison to the widely studied CNT and GO). As a result, controlled release of P-gp inhibitors and chemotherapeutic drugs can be achieved for the FSCNO nanoparticles under NIR irradiation at a very low power. The FSCNO-DH nanoparticles can precisely compromise the function of P-gp pumps and overcome the drug resistance of multidrug-resistant NCI/ADR-RES and A2780ADR cancer cells both in vitro and in vivo. The tumor vasculature-targeting capability of the nanoparticles via the binding between fucoidan and P-selectin overexpressed in tumor vasculature is confirmed by three different models: static culture of activated human endothelial cells in Petri dish, dynamic culture of activated human endothelial cells in microfluidic device, and human tumors in mice. In particular, the microfluidic model developed in this study may be valuable for mimicking the in vivo perfusion in blood vessels to examine the targeting capability of nanoparticles. Ultimately, the chemotherapeutic agent (DOX) and P-gp inhibitor (HM)-laden FSCNO nanoparticles exhibit superior safety and efficacy for therapy of two types of multidrug-resistant tumors and one type of nondrug-resistant tumors without evident side effects. This study may lay the foundation of developing a promising nanotechnology-based strategy for combating cancer multidrug resistance.