Versatile ginsenoside Rg3 liposomes inhibit tumor metastasis by capturing circulating tumor cells and destroying metastatic niches.

Limited circulating tumor cells (CTCs) capturing efficiency and lack of regulation capability on CTC-supportive metastatic niches (MNs) are two main obstacles hampering the clinical translation of conventional liposomes for the treatment of metastatic breast cancers. Traditional delivery strategies, such as ligand modification and immune modulator co-encapsulation for nanocarriers, are inefficient and laborious. Here, a multifunctional Rg3 liposome loading with docetaxel (Rg3-Lp/DTX) was developed, in which Rg3 was proved to intersperse in the phospholipid bilayer and exposed its glycosyl on the liposome surface. Therefore, it exhibited much higher CTC-capturing efficiency via interaction with glucose transporter 1 (Glut1) overexpressed on CTCs. After reaching the lungs with CTCs, Rg3 inhibited the formation of MNs by reversing the immunosuppressive microenvironment. Together, Rg3-Lp/DTX exhibited excellent metastasis inhibition capacity by CTC ("seeds") neutralization and MN ("soil") inhibition. The strategy has great clinical translation prospects for antimetastasis treatment with enhanced therapeutic efficacy and simple preparation process.

INTRODUCTION

Triple-negative breast cancer (TNBC) is the most aggressive breast cancer subtype and is 2.5-fold more likely to metastasize than other breast cancer subtypes within 5 years of diagnosis (). Metastasis accounts for more than 90% of TNBC-caused mortality in women (). In clinic, metastatic TNBC is almost incurable and the 5-year survival rate is only about 20% (). The dissemination, seeding, and colonization of circulating tumor cells (CTCs) are the decisive steps of distant metastasis. Once CTCs colonize and metastasize successfully, current clinical treatments usually fail to carry out solid effects (). Because CTCs are relatively more vulnerable before successful colonization (), we should strive to target and destroy metastatic tumor cells in their cradle. Therefore, CTCs, the “seeds” of metastasis, have emerged as valuable therapeutic targets for metastasis inhibition. In addition, following dissemination from primary sites, CTCs only survive and proliferate under favorable microenvironment (“soil”). The prepared soil for the “seeding” of CTCs, termed as a “metastatic niche (MN),” helps CTCs colonize and proliferate with the engagement of inflammatory cells, chemokines, and cytokines (). Therefore, apart from CTCs, the MNs, where the seeding and colonization process of CTCs happen, would be the other desirable target for the management of early cancer metastasis ().

Meta-analysis of randomized clinical trials has shown that taxane-based chemotherapy regimen represents the mainstay of therapeutic strategies for TNBC metastasis patients, and docetaxel (DTX) is a representative first-line antimetastatic TNBC drug (). However, metastatic TNBC has a poor response to conventional chemotherapy and is prone to being resistant to chemotherapy drugs (). Ginsenoside Rg3, the main component of “Shenyi capsule,” was approved as an anticancer drug by the State Food and Drug Administration of China in 2000 and has been synergistically applied with chemotherapeutic drugs to improve the treatment of breast cancer in clinic (, ). It has been proved to be an effective agent to sensitize tumor cells to taxane drugs (). Apart from that, Rg3 can promote antitumor immunity by regulating signal transducers and activators of transcription 3 (STAT3) signal transduction pathway (), indicating its potential in regulating and destroying MNs. Therefore, Rg3 is expected as an adjuvant agent with DTX to improve the cytotoxicity of DTX on CTCs and realize the regulation of MN.

However, because Rg3 degrades rapidly in the gastrointestinal tract and the blood and cannot reach the tumor site with anticancer drug simultaneously, its antitumor effect is greatly restricted. Fortunately, nanocarriers, especially liposomes, have emerged as a platform for the codelivery of combined drugs. Even so, nanotechnological chemotherapy for cancer metastasis still presents a unique challenge and has, so far, shown limited success in clinical translation. For example, Doxil and Abraxane, the marketed nanoformulations of doxorubicin and paclitaxel, were proved inefficient against breast cancer metastasis (, ). Increasing evidence suggests that the regulation of metastatic growth differs from that of primary tumor growth and questions the clinical validity of traditional nanodrug systems (). Most current nanotherapeutic strategies focusing on eliminating primary tumors are based on the enhanced permeability and retention (EPR) effect in well-vascularized primary tumors. However, early metastases are usually poorly vascularized and cannot be accessed by liposomes via EPR effect, not to mention the capture of CTCs, thus hindering the function of conventional liposomal formulations. Generally speaking, to realize CTC capturing and early metastatic lung targeting, the passive targeting strategy of nanodrugs is not workable yet, which means that active targeting is a critical factor that determines whether drugs can be effectively delivered to tumor cells. However, only a handful of active targeting nanodrugs have entered clinical trials, and none have passed early-phase testing in humans and have been approved to the market to date (). The key challenge is that currently researched active-targeting nanocarriers usually use the surface modification of the active targeting ligands or antibodies, which involve complex synthesis and formulation processes. It brings difficulties for large-scale production and quality consistency. Moreover, the stability and delivery process of synthetic active-targeting nanocarriers in human body are unpredictable and uncontrollable (), and there are safety-related concerns as well. The previous results in our laboratory demonstrated that ginsenosides have the potential to replace cholesterol as a liposome membrane material, and, unexpectedly, the ginsenoside-based liposomes were found with strong active tumor cell–targeting ability (). By analyzing the structure of ginsenoside Rg3, we found that its steroid ring structure, side chain of C17, and hydroxyl of C3 site satisfy all conditions proposed by researchers that required for liposome membrane regulators (fig. S1) (), which means that Rg3 has the potential as a stabilizer by inserting into the liposome membrane and interacting with phospholipids. The glucose moieties in the hydrophilic part of Rg3 will theoretically extend out of the liposome surface, which is the perfect ligand for glucose transporter 1 (Glut1) overexpressed in TNBC (), implying that Rg3 may have the ability to target and capture CTCs after inserted into liposomes. It is attractive that Rg3 can not only work as an adjuvant drug but also be concurrently used as an active targeting ligand and a membrane-stabilizing material. Thus, the Rg3-based liposomes can simultaneously realize the active targeting of CTCs while avoiding the complicated preparation process. Synchronous delivery of Rg3 with chemotherapeutics, such as DTX, can maximize the cytotoxic effect of DTX on CTCs and maintain the regulatory ability of Rg3 on MNs, yielding twice the result with half the effort.

On the basis of the above assumption, an Rg3-based liposome loading with DTX (Rg3-Lp/DTX) was prepared. We hypothesized that the substitution of cholesterol with Rg3 in the liposomes would endow them with Glut1-targeting capability to continuously capture CTCs. Rg3 would also exert its synergistic antitumor effects with DTX and enhance the response of CTCs to DTX, preventing the de novo metastasis. Furthermore, as Rg3-Lp/DTX reached the lungs via its active targeting ability to the disseminated CTCs, the formation of MNs would be prevented and reversed by the liposomes. We speculated Rg3-Lp/DTX could successfully inhibit metastasis by targeting and depleting CTCs (seeds) and preventing MN (soil) formation. The greatest advantage of this liposomal system is that it can be manufactured with simple components and processes, which has been validated in our pilot study. The system, therefore, has immense potential for clinical translation.

RESULTS

Preparation and characterization of Rg3-Lp/DTX

Except for replacing cholesterol with Rg3, Rg3-Lp/DTX was prepared by thin-film hydration method with the same process as that for conventional cholesterol liposomes (C-Lp) (Fig. 1A). The average sizes of C-Lp/DTX and Rg3-Lp/DTX were 106.8 ± 2.9 and 80.3 ± 3.7 nm, respectively (Fig. 2A and table S1), measured by dynamic light scattering (DLS). Representative transmission electron microscopy (TEM) images showed that Rg3-Lp/DTX had a spherical structure similar to that of C-Lp/DTX (Fig. 2B). Similar to those of C-Lp/DTX, the drug-loading capacity and encapsulation efficiency of Rg3-Lp/DTX were determined to be 6.5 ± 0.2 and 91.0 ± 2.6%, respectively.

Fig. 1.

Schematic diagram of Rg3-Lp/DTX preparation and its inhibiting mechanism on lung metastasis of TNBC.

(A) Preparation of Rg3-Lp/DTX by thin-film hydration method. (B) Because Rg3 extends its glucose moieties out of the surface of the liposome, Rg3-Lp/DTX can accurately capture CTCs through Glut1-Rg3 interaction. After reaching metastatic lesions with the disseminated CTCs, Rg3 can inhibit C─C chemokine ligand 2 (CCL2) secretion of tumor cells and thus prevent the recruitment of MDSCs and TAMs, destroy the formation of MNs, and promote the immune surveillance of tumor cells by cytotoxic T lymphocytes (CTLs).

Fig. 2.

Preparation and characterization of C-Lp/DTX and Rg3-Lp/DTX.

(A) Size distribution of C-Lp/DTX and Rg3-Lp/DTX. (B) Morphology of C-Lp/DTX and Rg3-Lp/DTX. Scale bars, 100 nm. (C) The vivid interaction between Rg3 and DSPC. Rg3 and DSPC were represented by stick and surface map as purple and green color, respectively; the water phase was omitted for the simplicity of display. (D) Typical coordinations of Rg3 with DSPC lipids and water molecules (H2O) in 3D (left) (Rg3: purple; PC: green; H2O: O in red and H in white sticks; hydrogen bond interactions: red dotted lines) and 2D (right) (Rg3: blue; PC: brown; H2O: single red dot; hydrogen bond interactions: green dotted lines) models. The rough coordinates were defined by setting the z axis as the membrane’s normal line, and Z = 0 means the center of the membrane bilayer. (E) Pyrene micropolarity I1/I3 (378/383) in pure liposomes (Lp), C-Lp, and Rg3-Lp. (F) Fluorescence anisotropy of DPH obtained from Lp, C-Lp, and Rg3-Lp. (G) Release stability of Rg3-Lp/DTX and C-Lp/DTX in 10% fetal bovine serum (FBS). All data are represented as means ± SD; n = 3; *P < 0.05, **P < 0.01, and ***P < 0.001.

To verify whether Rg3 can intersperse in the phospholipid bilayer and locate its position in the liposome bilayer, molecular dynamics (MD) simulation of the 1,2-dioctadecanoyl-sn-glycero-3-phosphocholine (DSPC)–Rg3 system was performed. The initial messy system consisted of 6 Rg3 molecules, 128 DSPC lipid molecules, and 10,484 simple point charge (SPC) water that was run for 700 ns of MD with a time step of 2 fs. During the dynamics, DSPC/Rg3 molecules were observed to rapidly aggregate in the first 200 ns and maintained a steady state in the following 500 ns (fig. S2A). Furthermore, the parameter “area per lipid (AL)” was applied to assess the validity of the simulation, and the results are show in fig. S2B. AL value rapidly dropped from 1.031 nm2 with the initial irregular state in the first 250 ns and then fluctuated smoothly around 0.641 nm2, indicating that Rg3 can stably intersperse in the DSPC bilayer. A larger mixed bilayer system (DSPC:Rg3 = 300:90, consistent with Rg3-Lp formulation) was established to locate the exact position of Rg3 in the bilayer. To improve computational efficiency, the main conformation of Rg3 was extracted (fig. S2C) as the initial structure for building pre-equilibrium larger mixed bilayers. MD simulation was performed for a total of 300 ns for the system, and AL value was also calculated to evaluate the stability (fig. S2D). To determine the relative position of the components in the bilayer, the density distributions of major headgroups of Rg3 and DSPC were analyzed from the entire simulated trajectory, and the results are displayed in fig. S2F. The density profiles along the bilayer’s normal line showed the symmetrical distribution of both Rg3/phosphorylcholine (PC) headgroups; the peaks’ positions of Rg3 headgroups with ±2.037 nm were almost the same as PCs’ position with ±2.055 nm. However, the peak area of Rg3 headgroups was larger than that of PC, indicating that the Rg3 headgroups had some conformations beyond the scope of DSPC and directly into the water molecular layer. To visually display the position of Rg3 headgroups in the MD simulation, the main cluster conformation was extracted and displayed in Fig. 2C. Rg3 were randomly dispersed in the bilayer. There are numbers of Rg3’s headgroups penetrating deeply into the water. To explore the mechanism of interaction between Rg3 and DSPC, the typical conformations were selected for analysis. As illustrated in Fig. 2D with two-dimensional (2D) and 3D views, the terminal pyran ring of the Rg3 hydrophilic group extended out of the phospholipid membrane plane with multiple water molecules surrounding its hydroxyl substituents. The internal pyran ring with ─OHs can also interact with the solvent water. Meanwhile, the pyran ring can form hydrogen bonds with PCs, between the hydroxyl group (O12) of Rg3 and the carbonyl oxygen (O32) of PC (Fig. 2D).

As illustrated in MD simulation, Rg3 can spontaneously form a stable bilayer with PC (Fig. 2, C and D). In addition, it was proved that almost all of Rg3 was inserted into the phospholipid bilayer in Rg3-Lp (table S5). As a liposome bilayer regulator, membrane micropolarity (Fig. 2E) and fluidity (Fig. 2F) of Rg3-Lp were investigated compared with those of pure phospholipid liposomes (Lp) and conventional C-Lp. The ratio of I1/I3 in Lp solution was 1.12 ± 0.03, while the ratio of I1/I3 in Rg3-Lp and C-Lp solution was 0.93 ± 0.02 and 1.02 ± 0.03, respectively. As shown in Fig. 2E, the micropolarity of the membrane was substantially reduced by the addition of Rg3 compared with that of Lp and C-Lp, which indicated the improvement in the hydrophobicity inside the bilayer of Rg3-Lp. It was mainly due to the fact that the regulator molecules were filled into the gaps of the phospholipid molecules and formed strong interactions with phospholipid molecules when they were embedded in the phospholipid bilayer, so that polar molecules outside the bilayer did not affect the micropolarity inside the layer (). Compared with Lp, the polarization decreased after adding Rg3 or cholesterol, which means that the fluidity increased at 25°C. However, the degree of polarization of Lp at 37°C was lower than that at 25°C, while the value remained stable as the temperature got higher in the Rg3-Lp group (Fig. 2F). It was highly possible that Rg3 fixed the amine groups of PCs on the surface of the liposome via the formation of hydrogen bonds with the polar headgroup of PC as illustrated in Fig. 2D and restricted the movement of acyl chains of PC molecules at a higher temperature. As a result, the ordering of the membrane and the stability of liposome were enhanced. The improved release stability of Rg3-Lp/DTX in 10% fetal bovine serum (FBS) solution was shown in Fig. 2G. Obvious aggregation and sedimentation were observed in C-Lp/DTX after 12 hours, while the particle size and character of Rg3-Lp/DTX remained stable (fig. S2G), which may be attributed to the strong interaction between Rg3 and phospholipid molecules, and the long hydrophilic chains of Rg3 could probably form a hydrated layer on the surface of the liposome to reduce agglomeration among liposomes as well.

Precise CTC-active targeting effect of Rg3-Lp

To simulate the microenvironment of CTCs, the efficiency of liposomes to capture CTCs in vitro was studied on 4T1 cells spiked into red blood cells (RBCs) at a mixture ratio of 1:103. The 4T1/RBC mixture was incubated with the liposomes for 4 hours at 37°C, shaking at 120 rpm. As shown in Fig. 3 (A and B) and fig. S5A, the fluorescence intensity of 4T1 cells treated with Rg3-Lp was 2.17 times of that treated with C-Lp, indicating that Rg3-Lp could capture more circulating 4T1 cells than C-Lp in the blood. However, no significant difference was observed between the uptake of the two liposomes by RBCs.

Fig. 3.

Precise CTC-targeting ability of Rg3-Lp.

(A) Quantitative analysis of cellular uptake of 1,19-dioctadecyl-3,3,39,39-tetramethylindodicarbocyanine perchlorate (DiD)–loaded C-Lp (C-Lp/DiD) or Rg3-Lp (Rg3-Lp/DiD) by circulating 4T1 cells and RBCs. (B) Representative confocal laser scanning microscope (CLSM) images of cellular uptake of C-Lp/DiD or Rg3-Lp/DiD by circulating 4T1 cells and RBCs. Scale bars, 50 μm. Inset scale bars, 10 μm. (C) Representative 10 min of IVFC results after intravenous injection of CFSE-4T1 cells and the following Rg3-Lp/DiD or C-Lp/DiD. (D) Representative confocal intravital microscopy (IVM) images of the mice ear blood vessels after intravenous injection of CFSE-4T1 cells and the following C-Lp/DiD or Rg3-Lp/DiD. (E) In vivo fluorescent and BLI images of the mice and ex vivo fluorescent and BLI images of various organs at 4 hours after the injection of Luc-4T1 cells and DiD-loaded liposomes. (F) Semiquantification of the fluorescent signals in lung tissues excised from mice with early metastatic lesions. n = 3; ***P < 0.001. ns, not significant.

In vivo CTC-targeting ability of Rg3-Lp was investigated with in vivo flow cytometry (IVFC) (). The mice were fixed on the table after the sequential injection of carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled 4T1 cells and 1,19-dioctadecyl-3,3,39,39-tetramethylindodicarbocyanine perchlorate (DiD)–labeled liposomes. Ear microcirculation of the mice was then immobilized and visualized under illumination with a 488-nm light-emitting diode (LED) and a 633-nm LED for the collection of CFSE and DiD signaling, respectively. After selecting the artery of interest, fluorescent signals were recorded. Different from traditional in vitro methods with low CTC isolation efficiency from the blood sample, this method could quantitatively monitor every individual CFSE-4T1 cell that passed through the laser slit across the artery over time. The green and red signals represented CFSE-labeled 4T1 cells and DiD-loaded liposomes flowing across the artery of the mouse ear, respectively. Thus, the double-positive signals represented circulating CFSE-4T1 cells recognized and captured by DiD-loaded liposomes (fig. S5B). As shown in Fig. 3C, 15 ± 2 double-positive signal peaks (indicated with dark arrows) were detected in 10 min after the administration of Rg3-Lp, accounting for approximately 50% of all circulating 4T1 cells passing by, whereas only 1 ± 1 double-positive signal peaks were found in the C-Lp group. Confocal intravital microscopy (IVM) imaging was also performed under Nikon two-photon confocal laser scanning microscope (CLSM) to dynamically monitor the real-time CTCs captured by the liposomes in vivo (movie S1 and Fig. 3D). Consistent with IVFC findings, notably increased CFSE-4T1 cells were detected exhibiting DiD and CFSE signals after Rg3-Lp/DiD injection.

To further explore the active targeting ability of Rg3-Lp, Luc-4T1 cells and DiD-loaded liposomes were injected into mice sequentially. Dual-mode in vivo imaging was conducted to analyze the colocalization of fluorescent signal and bioluminescence (BLI) signal in the same mouse after 4 hours. As can be seen in Fig. 3E, Luc-4T1 cells were trapped and disseminated in the lungs by tiny pulmonary capillaries and revealed BLI signals in the lungs 4 hours after injection. If the liposomes could actively target to tumor cells, then the liposomes would be intercepted in the lung along with the captured CTCs in the blood; on the other hand, they could also potentially be attracted to the lung when uncaptured CTCs in the blood disseminated in the lung. In general, the DiD-loaded active targeting liposomes should have enhanced signals in the lungs after the injection of 4T1 cells. Accordingly, the fluorescence signal of Rg3-Lp in the lungs was almost twice that of C-Lp (Fig. 3, E and F). Because it was not enough for CTCs to construct the EPR environment in only 4 hours, the increased accumulation of Rg3-Lp was mainly due to its strong active targeting capacity to 4T1 cells.

Targeting ability of Rg3-Lp to already formed lung metastasis

The targeting capability of Rg3-Lp in the already formed lung metastatic mouse model was studied. The mice were injected with Luci-4T1 cells. Rg3-Lp/DiD and C-Lp/DiD were administered after 14 days. The biodistribution of the liposomes in mice was imaged at different time points (Fig. 4A). As depicted in Fig. 4 (A and B), improved accumulation of Rg3-Lp was found in the metastatic site compared with that of C-Lp. Quantitative analysis showed that the accumulation of Rg3-Lp in the lung was 1.30-fold higher than that of C-Lp (Fig. 4C). As shown in Fig. 4D, the fluorescent images of frozen slices from the mice showed a much stronger purple signal (DiD) in the Rg3-Lp group than that in the C-Lp group, and Rg3-Lp merged better with the green fluorescent protein (GFP) signal of tumor cells (GFP-4T1). Under a static condition at 37°C in vitro, after incubation with C6-loaded liposomes for 4 hours, the fluorescent signal of Rg3-Lp in the 4T1 cell monolayer was 1.71-fold higher compared with that of C-Lp. After treated with various Glut1 inhibitors, the uptake of Rg3-Lp was significantly decreased, and no effect was observed on the uptake of C-Lp (Fig. 4E), indicating that the active targeting ability of Rg3-Lp could be attributed to the interaction with Glut1.

Fig. 4.

Targeting of Rg3-Lp and C-Lp to already formed lung metastasis and cellular uptake of Rg3-Lp and C-Lp in static 4T1 cells.

(A) In vivo fluorescence imaging of the mice with TNBC lung metastasis at 1, 2, 4, 8, 12, and 24 hours after administration with C-Lp/DiD (top) and Rg3-Lp/DiD (bottom) and in vivo BLI image of the same mice in each group at 24 hours after administration of DiD-labeled liposomes (the sites marked by the white dashed box are the corresponding lung metastasis lesions according to the BLI images). (B) Ex vivo fluorescent and BLI images of isolated mice organs. (C) Semiquantification of fluorescent signals in major organs excised from lung metastasis–bearing mice. (D) Representative immunofluorescence (IF) images of frozen lung slices from mice with TNBC lung metastasis. Blue: DAPI for staining cell nucleus; green: GFP-4T1 cells; red: Glut1; and purple: DiD-loaded liposomes. Scale bars, 20 μm. (E) Quantitative analysis of cellular uptake of C-Lp/C6 and Rg3-Lp/C6 in static 4T1 cells with or without pretreatment of different Glut1 inhibitors. n = 3; *P < 0.05, **P < 0.01, and ***P < 0.001.

Targeting mechanisms of Rg3-Lp

To explore the targeting mechanisms of Rg3-Lp, the molecular docking of Rg3 and cholesterol with Glut1 was performed (Fig. 5, A and B). The docking score of cholesterol and Rg3 was −7.1 and −11.7, respectively. The lower score of Rg3 implied that it was much easier to bind with Glut1. It could be found that Rg3 interacted with Glut1 through the hydrogen bond between the glycosyl chain of Rg3 and the amino acid residue of Glut1 (Fig. 5A). Gln 282, Gln 283, Trp 388, and Asn 411 that formed hydrogen bonds with the Rg3 outer glycosyl group were typical amino acid residues within the Glut1 transmembrane region, of which Gln 282 and Gln 283 were key residues for glucose binding (). In contrast, cholesterol lacked the ability to interact with Glut1 because of the absence of glycosyl groups (Fig. 5, A and B). To verify this calculation, the binding affinity of Rg3 for Glut1 was first confirmed using surface plasmon resonance (SPR). As shown in Fig. 5C and fig. S7A, Rg3 exhibited reproducible and concentration-dependent binding responses with Glut1. The kinetic responses indicated that a population of Rg3 have formed very stable complex with Glut1 immobilized on CM5 chip surface because its response units (RU) did not return back to baseline during the dissociation phase (Fig. 5C). In contrast, the binding responses of cholesterol returned to baseline rapidly (Fig. 5D), indicating an overall weak interaction. The value of Rg3 equilibrium dissociation constant (Kd) was 1.97 × 10−7 M (table S6), indicating that Rg3 had a relatively strong affinity for Glut1. In addition, on the basis of the results in Fig. 5D, Rg3-Lp showed a much higher response than C-Lp when flowing through the Glut1-immobilized sensor surface, implying that Rg3 retains its interaction with Glut1 after prepared into Rg3-Lp. Combined with the results of MD simulation in Fig. 2C, it can be deduced that Rg3 had the potential to extend its glycosyl from the surface of liposomes. Fluorescently labeled wheat germ agglutinin (WGA) was used as a probe to verify the deduction. WGA is a protein that can selectively bind to glycosyl residues on the cell membrane surface. Besides, in solution, WGA exists as a heterodimer with a molecular weight of approximately 38,000 Da, which prevents WGA from penetrating into the membrane (). Only glycosyl groups on the outer surface of the liposome membrane could be detected. Therefore, Texas Red-X–labeled WGA was used here as a probe for the surface glycosyl characterization of Rg3-Lp. The results showed that both Lp and C-Lp could not be labeled by WGA, while the fluorescence signal of Rg3-Lp was greatly enhanced (Fig. 5E), indicating the existence of glycosyls exposed on the outer surface of Rg3-Lp membrane. To further verify the conjecture, the in vitro cellular uptake of C-Lp and Rg3-Lp by Glut1-knockdown 4T1 cells (4T1Glut1−) and in vivo distribution of C-Lp and Rg3-Lp in mice injected with Luci-4T1Glut1− cells were investigated. On the basis of the Western blot (WB) results in fig. S7C, the expression of Glut1 in 4T1Glut1− cells was much lower than that in normal 4T1 cells after Glut1 knockdown. The uptake of Rg3-Lp in 4T1Glut1− cells was significantly reduced compared to that in normal 4T1 cells, and there was no significant difference with that of C-Lp group (Fig. 5F and fig. S7D). The targeting capability of liposomes to lung-retained CTCs in vivo was assessed with 4T1Glut1− cells. The results in Fig. 5G indicated that the fluorescence signal of Rg3-Lp in the lungs of the mice injected with 4T1Glut1− cells was significantly decreased to 0.54-fold of the mice administered with normal 4T1 cells, while there was no difference in C-Lp signals in the lungs between the two groups (Fig. 5H). In immunofluorescence (IF) staining images of the corresponding lung sections (Fig. 5I), the bright yellow signals represented the successfully merged signals of GFP-4T1 cells (green), Glut1 (red), and DiD-loaded liposomes (purple). Rg3-Lp accumulated more at the sites revealing green signals than at other sites and showed much better merged signals with Glut1 than those in C-Lp. In addition, the accumulation of Rg3-Lp at GFP-positive sites decreased notably with the knockdown of Glut1.

Fig. 5.

Active targeting mechanisms of Rg3-Lp to tumor cells.

(A) Maps of Rg3-Glut1 and cholesterol-Glut1 interaction, the arrows represent H-bond interactions. (B) 3D overlay showing the interaction of Glut1 with Rg3 (bright yellow) and cholesterol (purple). The bright yellow dotted lines represent H-bond interactions between Rg3 and Glut1. (C) Binding kinetics of Rg3-Glut1. (D) Binding curves of Rg3-Lp and C-Lp (1.52 μM Rg3 or Chol) with immobilized Glut1. (E) WGA binding signal of Lp, C-Lp, and Rg3-Lp to identify the exposed glycosyl on the liposome membrane. (F) Flow cytometry analysis of cellular uptake of C-Lp/C6 and Rg3-Lp/C6 in normal and 4T1Glut1− cells. (G) In vivo and ex vivo fluorescent and BLI images of the mice at 4 hours after the sequential injection of Glut1-knockdown/normal Luc-4T1 cells and DiD-loaded liposomes. (H) Semiquantification of DiD signal in the lungs from the mice at 4 hours after the injection of Glut1-knockdown/normal Luc-4T1 cells and DiD-loaded liposomes. (I) Representative IF images of the lung frozen slices from the mice at 4 hours after the injection of Glut1-knockdown/normal Luc-4T1 cells and DiD-loaded liposomes. Blue: DAPI for staining cell nucleus; green: GFP-labeled 4T1 cells; red: Glut1; and purple: DiD-loaded liposomes. Scale bars, 100 μm. n = 3; ***P < 0.001. ns, not significant.

Enhanced cytotoxicity and selective elimination of CTCs in the blood by Rg3-Lp/DTX

Thiazolyl blue tetrazolium bromide (MTT) assay was used to evaluate the cytotoxicity of the DTX-loaded nanoformulations on 4T1 cells. The calculated median inhibitory concentration (IC50) was 6.82 ng/ml for DTX, 1.40 ng/ml for C-Lp/DTX, 1.34 ng/ml for Nanoxel-PM, 0.04 ng/ml for Rg3/DTX, and 0.12 ng/ml for Rg3-Lp/DTX after incubation with 4T1 cells for 48 hours (Fig. 6A and table S7). The results demonstrated that the cytotoxicity of DTX was substantially enhanced when coadministered with Rg3. It was worthy to address whether the activity of Rg3-Lp/DTX in selective elimination of CTCs remained under biological environment. We then determined the apoptotic rate of 4T1 cells and leukocytes in the blood after different treatments (Fig. 6B and fig. S9, A and B). Briefly, 4T1 cells were spiked in the prepared fresh blood and shaken with free DTX, C-Lp/DTX, Nanoxel-PM, Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX at the DTX concentration of 100 ng/ml for 4 hours at 37°C at 120 rpm. Erythrocytes were lysed after platelets were removed by gradient density centrifugation. The remaining cells were then stained with peridinin-chlorophyll-protein–cyanine 5.5 (PerCP-Cy5.5)–labeled anti-CD45 to distinguish leukocytes from 4T1 cells and 4′,6-diamidino-2-phenylindole (DAPI) and fluorescein isothiocyanate (FITC)–annexin V to determine the variability of the cells. The populations of early apoptotic 4T1 cells after treatment with Rg3-Lp/DTX (18.60 ± 2.65%) were significantly higher than those of the cells treated with C-Lp/DTX, Nanoxel-PM, and free DTX (5.51 ± 1.04, 4.43 ± 0.81, and 3.38 ± 0.57%, respectively) (Fig. 6B). In addition, there was no significant difference in the apoptotic rates of leukocytes between the groups (Fig. 6B and fig. S9B). It may be attributed to the selective targeting ability of Rg3-Lp/DTX as mentioned above. Rg3-Lp/DTX exhibited greater binding capacity to CTCs than C-Lp/DTX and Nanoxel-PM because of the overexpression of Glut1 on CTCs. However, there was no significant increase in the binding and apoptotic capacity of Rg3-Lp/DTX to the leukocytes in the blood because leukocytes did not overexpress Glut1 (). All of these data indicated that Rg3-Lp/DTX still exerted its cytotoxicity to CTCs in the blood.

Fig. 6.

Cellular cytotoxicity and apoptosis effect of the liposomes.

(A) In vitro cytotoxicity assay of different formulations on 4T1 cells after 48 hours of co-incubation (n = 6). (B) Percentages of early, late, and total apoptosis of 4T1 cells and leukocytes in blood/4T1 mixture after different treatment. (C) WB analysis of NF-κB p65, Bax, and Bcl2 expression in 4T1 cells after different treatments. (D) Quantification of WB protein levels by group. n = 3; ***P < 0.001.

Notably, although Rg3 had no obvious cytotoxic effect when administered alone, it could obviously enhance the cytotoxicity of DTX to tumor cells when simultaneously used (Fig. 6A). As illustrated in Fig. 6 (A and B), the cytotoxic and apoptotic effects in the free Rg3/DTX group were obviously stronger than those in the free DTX group, which demonstrated that, in addition to the targeting effect of Rg3-Lp, Rg3 itself could greatly improve the antitumor effect of DTX. It has been proved that nuclear factor κB (NF-κB) is a key regulator that promotes cell proliferation, suppresses apoptosis, and stimulates metastasis and angiogenesis (). Activation of NF-κB has been found in TNBC, which can lead to overexpression of downstream signaling target antiapoptotic Bcl2 and decrease proapoptotic Bax to confer aggressive growth and chemoresistance (). Ginsenoside Rg3 was reported to promote cytotoxicity and apoptotic effect of chemotherapy drugs such as paclitaxel on TNBC through inhibiting NF-κB signaling and regulating Bax/Bcl2 expression (). Therefore, to elucidate the mechanisms of the combined effect of DTX and Rg3, the expression levels of NF-κB p65, Bcl2, and Bax were detected by WB analysis in 4T1 cells and lung tissues after treated with various drugs (fig. S11A and Fig. 6C). The results showed that the levels of NF-κB p65 and Bcl2 expression were significantly decreased, while the expression of Bax increased in vitro in 4T1 cells after treated with Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX (Fig. 6, C and D). For in vivo assays, the expression of NF-κB p65, Bax, and Bcl2 in the lung after different treatments were quantified by WB analysis. As shown in fig. S11B, Rg3-Lp and Rg3-Lp/DTX groups had the most obvious effect on NF-κB p65 decrease and Bax/Bcl2 ratio increase for the enhanced delivery efficacy of Rg3-Lp. The level of NF-κB p65 in the lung tissue of Rg3-Lp/DTX group decreased to 0.06- and 0.08-fold of that in phosphate-buffered saline (PBS) and free DTX group, respectively, and the corresponding Bax/Bcl2 level in the Rg3-Lp/DTX group was 2.80- and 3.9-fold higher than that in the PBS group and free DTX group, respectively. The results showed that Rg3 promoted cytotoxicity and apoptotic effect of DTX by decreasing the expression of NF-κB and antiapoptotic Bcl2 and by promoting the expression of proapoptotic Bax.

Inhibition of Rg3-Lp/DTX on lung metastasis of TNBC

TNBC lung metastases were developed by tail vein administration of Luc-4T1 cells to mice. The mice were randomly divided into eight groups and BLI-imaged after the injection. The treatment was initiated right after BLI imaging with a DTX dosage of 5 mg/kg (on days 0, 7, 14, and 21) (Fig. 7A). Lung metastasis in each group was closely monitored by weekly BLI imaging for up to 21 days. As shown in Fig. 7 (B to D), compared with other groups, Rg3-Lp/DTX could completely inhibit the progression of TNBC lung metastasis. According to the BLI signals recorded in the mice of each group (Fig. 7B), almost no increase was found in the Rg3-Lp/DTX group after the 21-day treatment, and the lung signal value in the group was only 0.02-fold of that in the Nanoxel-PM group at the end point (Fig. 7C). From the ex vivo BLI imaging of the lungs isolated from the mice in each group, almost none of the mice treated with Rg3-Lp/DTX developed lung metastases, whereas metastases could be observed in all other groups, including Rg3, Rg3-Lp, C-Lp/DTX, and Nanoxel-PM groups (Fig. 7D). We further proved that the metastasis inhibitory activity of Rg3-Lp/DTX could notably benefit the survival of mice (Fig. 7E and fig. S12B). The median survival time of the mice treated with PBS, DTX, C-Lp/DTX, Nanoxel-PM, Rg3, Rg3/DTX, Rg3-Lp, and Rg3-Lp/DTX was 27, 32, 37, 42, 32, 36, 35, and more than 63 days, respectively (Fig. 7E). The lungs of the mice were excised at the end of the experiment and photographed to observe the changes in the size and morphology. As shown in fig. S13 (A and C), the size and weight of the lungs in PBS group were obviously increased, which proved the successful construction of metastatic lesions and the rapid growth of metastatic tumors in the group. However, the lungs in the Rg3-Lp/DTX group were substantially reduced, due to the lowest lung metastases in the Rg3-Lp/DTX group. Moreover, the number of metastatic nodules in Rg3-Lp/DTX was greatly decreased compared to that in other groups, which was only one-third of that in the Nanoxel-PM group (Fig. 7F).

Fig. 7.

Effect of Rg3-Lp/DTX on lung metastasis and survival time of mice.

(A) Schematic design of TNBC lung metastasis therapy. (B) Metastasis progression curves depicted from in vivo BLI signal intensity (n = 4). (C) Quantitative analysis of total BLI signals detected in isolated lungs at the end of the treatment (n = 4). (D) BLI images of lung metastasis at different time points in the mice treated with various drugs and ex vivo lung BLI images at the end point (n = 4). (E) Survival time of the mice in various groups displayed as Kaplan-Meier curves (n = 5). (F) Quantification of metastasis node numbers of excised lungs from the mice in different groups (n = 5). **P < 0.01 and ***P < 0.001.

Effect of Rg3-Lp/DTX on metastatic microenvironment and its mechanism

The arrest of CTCs in the lungs does not definitely cause metastases. The formation of lung metastases depends on whether CTCs can successfully colonize and proliferate in the lungs (). Newly disseminated cancer cells were particularly vulnerable to immune surveillance (), which might lead to weaker BLI signals of the mice at the seventh day after Luci-4T1 injection than those at the first day in all groups, even in the PBS group (Fig. 7, B and D, and fig. S12B). The BLI signal value in the PBS group at day 7 was 6.96 ± 2.34 × 105 photo/cm2 per second, which was significantly lower than 19.04 ± 8.33 × 105 photo/cm2 per second at day 0. This phenomenon was in line with the data from experimental mouse models and clinical evidence (). However, the signal values in the PBS group increased rapidly to 10.5-fold after 14 days compared with those at day 7 (Fig. 7, B and D), indicating that cancer cells broke out of latency since 14th day and reinitiated overt outgrowth. It was reported that, as the seeds for tumor metastasis, CTCs need a suitable soil formed in the lung MNs, which is beneficial to the colonization and proliferation of the seeds (). Tumor-derived factors are the key factors for the formation of MNs (), and immunosuppression is one of its main characteristics (). It is highly possible that immune cells as the body’s defense “guard” lowered the BLI signal value in the first 7 days after tumor injection, and immunosuppressive MNs were formed after 14 days. The lungs were therefore more hospitable for CTCs to colonize and grow, resulting in the growth outbreak. As can be seen in Fig. 7 (B and D), a sharp increase in BLI signals occurred in all groups except Rg3-Lp/DTX after 14 days. One reason for this phenomenon might be the CTC-capturing ability and enhanced cytotoxicity effect of Rg3-Lp/DTX as mentioned above, and the other reason might be the MN modulation effect of Rg3-Lp/DTX. We found that the increasing rate of BLI signal in the Rg3-Lp group without DTX was also slower than that in PBS, free Rg3, and DTX groups after 14 days (Fig. 7, B and D). Rg3 was reported to inhibit the secretion of C─C chemokine ligand 2 (CCL2), a tumor-derived chemokine with a crucial role in MN formation, by suppressing the activation of STAT3 (). Combined with the low cytotoxicity effect of Rg3 alone as shown above (Fig. 6), we speculated that Rg3-Lp might exert an MN-destroying effect after targeting to the lung along with the CTCs that contributed to the lower BLI signal than that in free Rg3 and free DTX groups.

To verify the mechanisms of Rg3-Lp on MNs regulation, the effect of Rg3-Lp on STAT3-CCL2 pathway in 4T1 cells (Fig. 8, A to D) and lung tissues (Fig. 8E and fig. S14, A and B) were explored. The activation of STAT3 signal is directly related to the increase in CCL2 secretion (). As shown in fig. S15A, the expression of phosphorylated STAT3 (p-STAT3) in 4T1 cells was significantly decreased after incubation with stattic, a well-known STAT3 phosphorylation inhibitor (). As a result, the concentration of CCL2 decreased significantly after treatment with stattic (fig. S15C), indicating that CCL2 secretion can be interrupted by inhibition of STAT3 phosphorylation. As illustrated in Fig. 8 (A and B), the p-STAT3 expression in 4T1 cells in the groups treated with Rg3, including Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX groups, were obviously lower than those in PBS, DTX, C-Lp/DTX, and Nanoxel-PM groups. However, the expression of STAT3 in different groups with Rg3 was almost the same, indicating that Rg3 effectively suppressed the phosphorylation of STAT3. As a downstream effector of STAT3 activation, the gene expression of CCL2 in 4T1 cells was measured. As shown in Fig. 8C, the gene expression of CCL2 in 4T1 cells treated with Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX were significantly lower than those treated with free DTX and other DTX-containing formulations. Subsequently, the quantitative analysis of CCL2 in 4T1-cultured medium (CM) with different treatments was conducted by enzyme-linked immunosorbent assay (ELISA) (Fig. 8, D and E). The concentration of CCL2 in the Rg3-Lp group was 0.85-fold of that in the DTX group, and the CCL2 concentration in the Rg3-Lp/DTX group was 0.71-fold of that in the Nanoxel-PM group. There was no significant difference in CCL2 concentration between Rg3 and Rg3-Lp groups, which revealed the CCL2 secretion inhibition ability of free Rg3 and the formulated Rg3-Lp. The change in p-STAT3 level in 4T1 cells after the intervention of each group was consistent with that in CCL2 expression in quantitative polymerase chain reaction (qPCR) and ELISA analysis. Moreover, the activation of STAT3 is an acknowledged factor that induces CCL2 gene expression and promotes the occurrence and metastasis of tumors through the p-STAT3–CCL2 pathway (). Therefore, Rg3 was proved to have the potential to inhibit CCL2 expression in 4T1 cells via suppressing the activation of STAT3. To verify the STAT3-CCL2 signaling inhibition effect of Rg3 observed at the cellular level, the levels of p-STAT3 and CCL2 in the lungs of mice after different treatments were investigated (fig. S14, A and B, and Fig. 8E). Different from the equivalent suppressive effect on p-STAT3–CCL2 signaling between Rg3 and Rg3-Lp in vitro, the p-STAT3 level in lung tissue of Rg3-Lp and Rg3-Lp/DTX groups was a quarter of that of the C-Lp/DTX group, while there was no significant difference in the p-STAT3 level between free Rg3 and C-Lp/DTX groups (fig. S14, A and B). Accordingly, the CCL2 expression level in lung tissues of the Rg3-Lp group was significantly lower than that of free Rg3 and C-Lp/DTX groups (Fig. 8E). It was probably because more Rg3 could reach the lung MNs after prepared into Rg3-Lp for its ability to capture CTCs in the blood and actively target to the CTCs disseminated in lung MNs, revealing that the preparation of Rg3 into liposomes played an essential role in the efficient delivery of Rg3 to the lung and p-STAT3–CCL2 inhibition.

Fig. 8.

Differential regulation of lung metastasis niches treated with various drugs.

(A) WB identification and comparison of p-STAT3 and STAT3 expression in 4T1 cells after different treatment. (B) Semiquantitative results of the relative level of p-STAT3 and STAT3 in 4T1 cells obtained in WB assays. (C) Gene expression of CCL2 in 4T1 cells treated with different drugs determined by qPCR. Concentration of CCL2 in 4T1-CM (D) and metastatic lung tissues (E) after different treatment measured by ELISA. (F) The migration ability of TAMs when incubated with 4T1-CM pretreated with different formulations. (G) Flow cytometry analysis of lung-infiltrating Gr1 high CD11b + granulocytic (G-MDSC) and Gr1int CD11b + monocytic (M-MDSC). (H) TAM populations (CD45+/F4/80+/CD206+) in lung tissues detected by flow cytometry. (I) Histogram analysis of CD4+ and CD8+ lymphocytes in mice treated with Rg3-Lp/DTX, control groups, and PBS group. (J) Full scanning images of hematoxylin and eosin (H&E), p-STAT3, CCL2, MDSC (Gr1-red and CD11b-green) and TAM (F4/80-red and CD206-green) staining of lung tissues of the mice treated with different drugs and the zoomed-in images of p-STAT3 and CCL2 staining of lung tissues from the full-scan images in each group (the black box represents the field of view selected for magnification). Scale bars, 5 mm for full scanning images and 0.2 mm for magnified images. n = 3; *P < 0.05, **P < 0.01, and ***P < 0.001.

CCL2 is a potent chemoattractant for myeloid-derived suppressor cells (MDSCs) and tumor-associated macrophages (TAMs), which are key immune cells that exert immunosuppressive effects in the metastatic microenvironment (). Because Rg3-Lp significantly decreased the CCL2 expression of cancer cells as proved above, it had the potential to reduce the levels of MDSCs and TAMs, which were recruited by CCL2, in the lungs of the mice. As depicted in fig. S14 (C and D), the optical density (OD) value in Fig. 8F represented the level of TAMs migrating to the lower chamber containing 4T1-CM after different treatments. The OD value in the Rg3-Lp/DTX group was significantly lower than that in the C-Lp/DTX group (Fig. 8F and fig. S14D). It was probably due to the lowered CCL2 concentration in 4T1-CM after Rg3 treatment reduced the migration capability of TAMs. The results obtained in Fig. 8F showed that Rg3 was capable of inhibiting the recruitment of immune-suppressive cells such as TAMs by cancer cells via inhibiting the secretion of CCL2 in vitro, and the hypothesis was then investigated in vivo. As depicted in Fig. 8 (G and H), relative abundances of MDSCs and TAMs in the lungs treated with Rg3-Lp/DTX and Rg3-Lp were significantly lower than those in other groups. The levels of MDSCs and TAMs in the Rg3-Lp/DTX group were also significantly reduced compared with those in the Rg3-Lp group. It was not only Rg3 itself that can reduce the secretion of CCL2 by inhibiting p-STAT3–CCL2 pathway but also DTX and Rg3 delivered by Rg3-Lp/DTX can also kill tumor cells and further inhibit the secretion of CCL2, thus reducing the number of recruited cells. Because MDSCs have multiple pathways that suppress the effects of cytotoxic T lymphocyte (CTL) and induce imbalance in T cell differentiation (), the reduction of MDSCs in Rg3-Lp and Rg3-Lp/DTX groups contributed to the significantly increased abundance of CD4+ T cells compared with other groups (Fig. 8I). A more direct effect of Rg3-Lp/DTX on the MN regulation in vivo could be observed in Fig. 8J, in which the metastatic nodules in lung tissues were stained with hematoxylin and eosin (H&E). The high-expression areas of p-STAT3 and CCL2 signals in lung tissue coincided with the position of the metastatic nodules. The levels of MDSCs (Gr1+ and CD11b+) and TAMs (F4/80+ and CD206+) were represented by yellow signals in lung tissues that were merged by the red signals (Gr1 and F4/80) and green signals (CD11b and CD206). The p-STAT3 and CCL2 signals of the lung tissues in the Rg3-Lp/DTX group were much weaker than those in other groups, and the yellow signals in Rg3-Lp/DTX were correspondingly lower than those in other groups, demonstrating that fewer MDSCs and TAMs were recruited to the lungs to form the MNs in the Rg3-Lp/DTX group (Fig. 8J).

DISCUSSION

To improve the therapeutic efficacy of chemotherapeutics against TNBCs with high molecular heterogeneity, the combination of drugs with different antitumor mechanisms has been extensively studied and applied in clinic (, ). This study successfully developed a versatile Rg3-Lp/DTX that could effectively inhibit TNBC metastasis by capturing the seeds in the blood and simultaneously destroying the soil in the lung. The metastatic 4T1 model was chosen because it robustly recapitulates many features of human TNBC (), which is refractory to most therapeutic agents, including chemotherapeutic drugs used clinically (, ). Shenyi capsule, whose main component is ginsenoside Rg3, has been approved in China since 2000 and applied clinically in combination with chemotherapeutics for the treatment of breast cancer and lung cancer to maximize the clinical efficacy of chemical anticancer drugs (, ). In the present study, we found that Rg3 itself did not exhibit obvious 4T1 cytotoxicity and proapoptotic activity. However, when DTX was administered with Rg3, its effect on tumor cell cytotoxicity and apoptosis was greatly improved by Rg3 in vitro. The IC50 values of Rg3/DTX and Rg3-Lp/DTX were only about 1/10 of those of DTX alone. On the basis of the WB analysis results, it can be explained that the promoting effect of Rg3 on DTX cytotoxicity of TNBC was attributed to the inhibition of NF-κB activation. Activated NF-κB increases the expression of antiapoptotic protein Bcl2 and inhibits the expression of proapoptotic protein Bax, which has been found in TNBC frequently (). The inhibition of NF-κB signaling by Rg3 is responsible for the sensitization of cancer cells. Therefore, Rg3 has potential as an anticancer adjuvant agent in the combined chemotherapy of TNBC.

Although the combination of Rg3 and DTX has achieved enhanced CTC-neutralizing effect in vitro, when administered in vivo, the two drugs cannot reach the tumor site synchronously and act synergistically due to the different systemic distribution and rapid clearance. Thereby, the combination effect will be greatly reduced. Nowadays, liposomes are universally acknowledged as a platform for efficient codelivery of combined drugs. However, different from primary breast tumor, the treatment of metastasis by conventional liposomes is hindered by the difficulty in capturing CTCs and delivering the drugs to the early metastatic site at the same time. Although liposomes co-loading of DTX and Rg3 can achieve the targeting delivery to primary solid tumors through EPR effect, the strategy is no longer suitable for the capture of CTCs and the targeting of metastatic lesions in early metastatic stages due to the lack of EPR effect. Therefore, “next-generation” liposomes used tumor-targeting ligand modification to improve their targeting capacity. Unfortunately, no active targeting liposomal product has been approved until now due to its complicated formulation components, multistep preparation processes, and failure to meet therapeutic expectations in clinical trials ().

By analyzing the structure of Rg3, we found that Rg3 not only can exert its efficacy as an adjuvant drug but also has the potential as a membrane material to substitute cholesterol and as an active ligand for tumor targeting (, ). Gallay et al. found that three conditions are required for liposome membrane regulators to act on the phospholipid bilayer, i.e., (i) hydroxyl of C3 site, binding to the hydrophilic groups of the phospholipid; (ii) a planar ring structure; and (iii) the side chain of C17, aligning with the hydrophobic fatty acid chain of the phospholipid (). Rg3, as an analog of cholesterol, satisfies all the above conditions, implying that it has the ability to interact with phospholipid molecules to stabilize the bilayer of liposomes. The results of liposome membrane micropolarity and fluidity assays of Rg3-Lp suggested that Rg3 was interspersed in the phospholipid bilayer and regulated the micropolarity and fluidity of the bilayer by interacting with phospholipid molecules and thus formed a much more stable bilayer with PCs than cholesterol. Moreover, from the results obtained in Rg3-Glut1 docking and SPR experiments, Rg3 was proved having a strong affinity with Glut1 via its glycosyls, indicating that it is a promising TNBC active targeting ligand of Glut1 overexpressed in 4T1 cells (). In contrast, because of the lack of glycosyl groups, the affinity of cholesterol with Glut1 was minimal and negligible. To further locate the position of Rg3 in the phospholipid bilayer, we conducted an MD simulation of Rg3-PC system and found that Rg3 could spontaneously form a stable bilayer with PCs. It could interact well with PCs and fill the gaps between them, functioning as a liposome membrane stabilizer and regulator. Therefore, Rg3-Lp remained stable in 10% FBS for 96 hours. In addition, the glycosyl moieties of Rg3 can extend out the liposome to potentially bind with Glut1. Among the two glycosyl groups of the hydrophilic part of Rg3, the pyran ring of the inner glycosyl with ─OHs can form hydrogen bonds with PCs and can also extend a part of the inner glycosyl group out of the membrane. The terminal pyran ring of the outer glycosyl moiety fully extends out of the phospholipid membrane plane of the liposome to specifically bind to Glut1 overexpressed on 4T1 cells. The simulation result was further confirmed by WGA assays. A stronger fluorescence signal of Texas Red-X–labeled WGA was observed in Rg3-Lp than that in C-Lp. As a result, it can be concluded that glycosyl groups of Rg3 could extend out of the liposome membrane to realize the interaction with Glut1 because only the glycosyl groups extending out of the Rg3-Lp membrane could bind to fluorescence-labeled WGA. This finding inspired us to construct Rg3-Lp/DTX to realize the active targeting delivery of the combined drugs without excess chemical modification of other ligands. To verify this conjecture, IVFC and metastatic lung targeting experiments were carried out. Rg3-Lp exhibited much higher CTC capturing and CTC colonies targeting capacity than C-Lp. When the Glut1 gene in 4T1 cells was knocked down, the targeting effect of Rg3-Lp in vitro and in vivo disappeared correspondingly. It strongly proved that Rg3-Lp mainly achieved CTC capturing and lung targeting by Rg3’s glycosyls extending outside. Subsequently, DTX encapsulated in Rg3-Lp can exert an optimal chemotherapeutic effect on captured CTCs under the joint action of Rg3. After Rg3-Lp reaches the lungs via targeting to the disseminated CTCs, Rg3 can reduce the level of tumor-secreted CCL2 by suppressing the activation of STAT3, resulting in lower abundance of MDSCs and TAMs recruited to MNs, which could help the immune system recognize CTCs arrested in the lung and then prevent CTC colonization and proliferation.

In summary, a multifunctional liposomal system was developed on the basis of the versatile capabilities of ginsenoside Rg3 in the present study. The liposome was composed of only three components: phospholipids, Rg3, and the loaded DTX. In the liposomes, Rg3 acts as a membrane material replacing cholesterol to keep the stability and fluidity of liposomes, as an active targeting ligand to bind with the Glut1 on tumor cells, as a MN modulator to destroy MNs, and as a chemosensitizer of DTX to maximize its chemotherapeutic effect. By actively targeting to the CTCs (the seeds) and neutralizing the metastatic microenvironment (the soil) in the lungs simultaneously, Rg3-Lp/DTX could successfully inhibit metastases of TNBC, with a dramatically improved efficacy compared to the solution, conventional liposomes, and marketed micelles of DTX. Rg3-Lp/DTX is easy to be prepared based on its simple composition, both in laboratory and in factory. The formulation is currently under preclinical studies and demonstrates immense potential to provide an effective and safe chemotherapeutic regimen for the treatment of TNBC and associated lung metastasis.

MATERIALS AND METHODS

Study design

The objective of this study was to determine whether Rg3 can be used as an active targeting membrane material and adjuvant drug with DTX to inhibit TNBC lung metastasis. Mice bearing metastatic model were used for these controlled laboratory experiments. It was confirmed in vitro and in vivo that the Rg3-Lp could help realize the active targeting delivery of DTX and thus realize metastasis inhibition. A variety of anticancer modalities was compared, including C-Lp and marketed DTX productions. BLI was used to inspect for metastatic progression. Survival was selected prospectively as a primary end point. Safety and tolerability were evaluated by measuring the weight and H&E staining of major organs. All the animal experiments were performed in accordance with the guidelines evaluated and approved by the Institutional Animal Care and Use Committee (IACUC), School of Pharmacy, Fudan University (Shanghai, China).

In all experiments, animals were randomly allocated to control or experimental groups. No statistical method was used to predetermine the sample size, which was estimated only on previous experience with assay sensitivity and the different animal models. Unless otherwise specified, three independent experimental replicates were performed. The investigator was blinded to the group allocation before surgery. The surgeries were performed independently for at least three times, and the surgeon was blinded when the experiment was conducted.

Materials

DTX, ginsenoside 20 (S)–Rg3, and Nanoxel-PM (Samyang Biopharm) were gifted by Bensu Medicine Technology Co. Ltd. (Shanghai, China). Cholesterol was purchased from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). PL-100M was purchased from AVT Medicine Technology Co. Ltd. (Shanghai, China). Coumarin-6 (C6), Hoechst 3334w, DAPI, 1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide (DiR), DiD, MTT, and 1% crystal violet solution were purchased from Meilun Biotechnology Co., China. Annexin V–FITC/propidium iodide (PI) apoptosis detection kit and phosphatase inhibitor cocktail were provided by Yeasen Biotechnology Co., China. Penicillin, streptomycin, and 0.25% trypsin-EDTA were purchased from Invitrogen Co., USA. Protein GLUT1 was obtained from Wuhao Bio-Tech Co., China. CM5 sensor chips were purchased from GE Healthcare Bio-Sciences AB (Sweden). Anti-Glut1 small interfering RNA (siRNA) (CCAACUGGACCUCAAACUUTT, AAGUUUGAGGUCCAGU-UGGTT) and siRNA-mate were provided by GenePharma Co., China. All other chemicals and solvents used in this study were of reagent grade or high-performance liquid chromatography (HPLC) grade.

Cell culture

(GFP- or Luc-) 4T1 cells and RAW cells were provided by the Shanghai University of Traditional Chinese Medicine (Shanghai, China). Raw and 4T1 cells were maintained in Dulbecco’s modified Eagle’s medium and RPMI 1640 supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 μg/ml) at 37°C in a humidified atmosphere containing 5% CO2, respectively.

Animals

Female BALB/c mice (20 ± 2 g) were provided by Shanghai Slac Laboratory Animal Co. Ltd. (China) and maintained under standard housing conditions of the Department of Experimental Animals, Fudan University (Shanghai, China). All the animal experiments were performed in accordance with the guidelines evaluated and approved by the IACUC, School of Pharmacy, Fudan University (Shanghai, China).

Preparation and characterization of the liposomes

PL-100M, Rg3 (or cholesterol), and DTX (10:3:1) (w/w) were dissolved in the mixed solvent of chloroform and ethanol (1:1) (v/v). The solvent was dried under vacuum at 48°C and then the film was hydrated with 5% glucose solution at 48°C. The C-Lp/DTX and Rg3-Lp/DTX liposomes were formed after ultrasound with a probe ultrasonic instrument (JYD-650, Zhixin Instrument Co. Ltd., Shanghai, China) at 270 W for 30 times. Size and zeta potential of the freshly prepared liposomes were measured by DLS (Malvern Instruments Ltd., UK), and morphology of the micelles was observed under TEM (Tecnai G2 F20 S-Twin; FEI, Hillsboro, OR, USA).

Preparation of fluorescent dye-loaded Rg3-Lp

The preparation of fluorescent dye–loaded liposomes was similar to that of DTX-loaded liposomes. PL-100M and Rg3 (or cholesterol) (10:3) (w/w) were dissolved in the mixed solvent of chloroform and ethanol (1:1) (v/v). Then, the fluorescent dye dissolved in dimethyl sulfoxide (DMSO) solution (C6, DiD, 10 mg/ml, 10 μl) was added. The following steps are the same as those described in the “Preparation and characterization of the liposomes” section.

Analysis of Rg3 amount in each part of liposome

PL-100M, Rg3 (or cholesterol), and DTX (10:3:1) (w/w) were dissolved in the mixed solvent of chloroform and ethanol (1:1) (v/v). The solvent was dried under vacuum at 48°C, and then the film was hydrated with 5% glucose solution at 48°C. The Rg3-Lp were formed after ultrasound with a probe ultrasonic instrument (JYD-650, Zhixin Instrument Co. Ltd., Shanghai, China) at 270 W for 30 times. A total of 500 μl of liposome suspension was taken, and free Rg3 was removed by ultrafiltration. The filtrate was diluted to 2 ml with methanol. Another 500 μl of liposome suspension was diluted to 50 ml with methanol. The concentration of the free Rg3 (Cfree Rg3) and total Rg3 (Ctotal Rg3) was measured by HPLC. The encapsulation efficacy (EE) of Rg3 was calculated according to the following equation

To obtain the concentration of Rg3 in external aqueous phase, 500 μl of liposome suspension was ultrafiltered, and the concentration of the filtrate was measured by HPLC. Because drugs are distributed homogeneously in the aqueous phase (both inside and outside the liposomes) in the process of liposome-passive drug loading (thin-film hydration ultrasonic method) (, ), the concentration of Rg3 in the inner aqueous phase (CRg3 in inner aqueous phase) is the same as that in the outer aqueous phase.

Calcein-entrapped Rg3-Lps were prepared according to the following procedure: PL-100M, Rg3 (or cholesterol), and DTX (10:3:1) (w/w) were dissolved in the mixed solvent of chloroform and ethanol (1:1) (v/v). The solvent was dried under vacuum at 48°C, and then the film was hydrated with 10 mM calcein at 48°C. The liposomes were formed after ultrasound with a probe ultrasonic instrument at 270 W for 30 times. A total of 500 μl of liposome solution was separated from free (unentrapped) calcein by gel filtration chromatography using a Sephadex G-50 column. Liposomes were disrupted by addition of 10% (v/v) Triton X-100 to release the loaded calcein. Entrapment efficiency was calculated by measurement of fluorescence emitted from entrapped calcein. Captured volume (Vinner aqueous phase) in the 500-μl liposome solution was obtained from the calculated values of calcein entrapment efficiency (, ). Therefore, the amount of Rg3 in the inner aqueous phase was obtained

Stability of the liposomes in 10% FBS

Size of C-Lp/DTX and Rg3-Lp/DTX were evaluated for 96 hours at 37°C in 10% FBS. At 0.5, 1, 2, 4, 8, 12, 24, 48, 72, and 96 hours, size distributions of the liposomes were measured by DLS. In vitro DTX release behavior was measured by using HPLC. One milliliter of liposome solution with 10% FBS was sealed into a dialysis bag [molecular weight cutoff (MWCO) = 3500 Da; 18 mm] and immediately placed into glass bottles containing 100 ml of releasing medium [PBS (pH 7.4) containing 0.5% SDS]. The bottles were shaken at 100 rpm, 37°C. Then, 0.5 ml of the medium was withdrawn, and an equal volume of fresh release medium was refilled at various time points.

Measurement of liposomal membrane micropolarity

A total of 0.1 ml of the tritium solution (4 × 10−7 M) was placed in a test tube, and 5 ml of different liposome solutions was added after volatilization. After sonication for 10 min, the mixture was left at room temperature for 12 hours. The fluorescence intensity was measured at 373 and 384 nm, respectively, at an excitation wavelength of 338 nm, which is represented by I1 and I3. The value of I1/I3 correlates with the microenvironmental polarity of liposomes.

Measurement of the liposomal membrane fluidity

A total of 1 ml of DPH solution (2 × 10−6 M) was taken and added to 5 ml of liposome solution. At λEx = 360 nm and λEm = 430 nm, the fluorescence intensity was measured after equilibrating for 12 hours. The degree of polarization was calculated according to the following formulawhere ‖ represents the excitation and emission polarizers are in parallel, ⊥ represents the excitation and emission polarizers are vertical, and G is the grating correction factor.

It is known that the greater the degree of polarization, the smaller the fluidity. Lp, C-Lp, and Rg3-Lp were prepared and the degree of polarization was measured at 25° and 37°C.

MD simulation of Rg3-Lp

MD simulation was performed to study the interaction mechanism between DSPC and Rg3. Self-assembly method was applied to visualize the studied properties of the studied system. The initial system consisted of six Rg3 molecules, 128 DSPC lipid molecules, and 10,484 SPC water. The initial structure of the system was generated by using the gmx insert-molecules tool that was provided in the GROMACS package. All above molecules were randomly placed in the simulation box. The initial messy state was run for 700 ns of MD with a time step of 2 fs. On the basis of the self-assembly result to improve computational efficiency and save computing time, the main principles of Rg3 conformation were extracted to construct the larger mixed bilayers (DSPC:Rg3 = 300:90) by the use of MemGen tool (http://memgen.uni-goettingen.de/) for further MD simulation.

During MD simulations, the CHARMM36 force field was used and the periodic boundary condition was applied in all xyz directions with the neighbor list updated for every time step. The simulations were performed in an isothermal-isobaric (NPT: constant composition, T = 300K and P = 1.0 atm) ensemble. At the same time, the time constants were equal to 0.5 and 10 ps using the Parrinello-Rahman for the temperature and pressure couplings, respectively. The compressibility was set to 4.5 × 10−5 bar−1. A cutoff distance of 1.2 nm was used for the long-range neighbor list of electrostatic and van der Waals interactions. The particle-mesh Ewald (PME) method was applied to calculate long-rang electrostatic interactions with the PME grid of 0.12 nm in the reciprocal-apace interactions and cubic interpolation. In addition, the H-bonds lengths were constrained using the LINear Constraint Solver (LINCS) algorithm. The visual molecular dynamics software was used for molecular visualization.

The AL was often considered to be a common physical property to determine whether the phospholipid bilayer has reached equilibrium. Usually, the phospholipid bilayer structure is along the z axis during the simulation, and the area of a single phospholipid molecule can be calculated by the following formulawhere the Lx and Ly were the length of the box in the x and y direction, respectively, while the Nm is the total number of molecules in the bilayer.

Furthermore, the order parameters of phospholipids indicate the degree of order, regularity, and symmetry of the system. It is very useful to study the order parameters of phospholipid tails. Phospholipids are embedded in the bilayer in different ways, which has a great influence on the nonpolar tails of phospholipidswhere θ is the angle between the z axis (the membrane normal) of the bilayer and the vector C to C. The angular bracket is the average value of the ensemble. When Scd = 1, 0.5, and 0, it means that the orientation and ordering of the phospholipid tails in the bilayer with respect to the bilayer normal are the same, opposite, and randomly arranged, respectively.

WGA assay

Appropriate amount of Lp, C-Lp, and Rg3-Lp was incubated with Texas Red-X–conjugated WGA (1 μg/ml) in the dark at room temperature for 10 min. After that, the mixture was dialyzed overnight at 4°C with a 100-kDa MW dialysis to remove free WGA. The fluorescence intensity was measured using a fluorescence spectrophotometer [excitation wavelength/emission wavelength (Ex/Em): 595/615 nm].

Modeling of Rg3 and Chol interaction with Glut1

The structures of Rg3, cholesterol (Chol), and Glut1 (Protein Data Bank ID 4PYP) were imported into Schrödinger maestro version 11.8. LigPrep was used for energy minimization of the Rg3 and Chol structure. The 3D coordinate of Glut1 was subjected to energy minimization after assigning proper bond orders and ionization states along with charge fixing. The best docked conformation of the test compounds was identified on the basis of glide energy, docking score, the hydrogen bonds, and hydrophobic interactions. The generated 2D scheme illustrated protein-ligand interactions briefly, and ribbon/surface view representations of docked complexes were generated using PyMOL.

SPR experiments of Glut1-Rg3

SPR measurements were performed with a BiaCore T200 instrument (Pharmacia Biosensor). The experiment was performed on the research-grade CM5 sensor chip (carboxylated dextran surface) with the running buffer including HBS-EP [10 mM Hepes (pH 7.4), 150 mM NaCl, 3 mM EDTA, and 0.005% polyoxyethylenesorbitan surfactant]. The Glut1 protein was immobilized on the chip surface via amide linkages at pH 4.0. Rg3 was diluted in HBS-EP buffer to a serial of concentrations from 0.78 to 5 μM and flowed over the Glut1-immobilized CM5 chip to record resonance changes to assess binding affinity. Data transformation, overlay plots, and interaction analysis were prepared with BIA evaluation software. The binding response of Rg3-Lp and C-Lp (1.52 μM Rg3 or Chol) was also tested to investigate whether Rg3 can still interact with Glut1 after being developed into Rg3-Lp.

Circulating 4T1 cells capture in RBCs

4T1 cells were collected at the density of 1 × 106 cells per tube. To simulate the physiological state of circulating tumors, 4T1 cells in each tube were mixed with 1 × 109 RBCs. They were incubated at 120 rpm, 37°C. After treated with C-Lp/C6 and Rg3-Lp/C6 at the C6 concentration of 100 ng/ml for 4 hours, nuclei were then stained with Hoechst 33342 (5 μg/ml) before analyzed by flow cytometry (BD Biosciences, USA). To better observe the tumor cells, we chose GFP-4T1 cells that express green fluorescence and mixed it with RBCs. The mixture was incubated with DiD-loaded C-Lp or Rg3-Lp at 120 rpm, 37°C [DiD (100 ng/ml)], and then centrifuged to wash away free liposomes. The colocalization of GFP-4T1 cells with DiD-loaded C-Lp and Rg3-Lp was observed under the laser confocal live cell imaging system (LSM710, Zeiss), respectively.

Uptake of liposomes by static 4T1 cells

4T1 cells were seeded at a density of 2 × 105 cells per well into 12-well plates. For Rg3-Lp/C6 + glucose, Rg3-Lp/C6 + phlorizin, and Rg3-Lp/C6 + quercetin groups, the medium was removed and replaced by 20 mM glucose, phlorizin, and quercetin solutions, respectively, after 12 hours. These three solutes should be dissolved in glucose-free medium. When incubated for 1 hour, cells were treated with C-Lp/C6 and Rg3-Lp/C6 at the C6 concentration of 100 ng/ml. Incubated for 4 hours, trypsinized, collected, and washed three times with fresh PBS (pH 7.4), the cells were analyzed by flow cytometry (CytoFlex S, Beckman Coulter Inc., USA). For quantitative study, nuclei were then stained with Hoechst 33342 (5 μg/ml) before subjected to fluorescence microscopic observation (Zeiss LSM 710, Oberkochen, Germany). All the fluorescence images were taken under identical conditions with the same exposure time.

Uptake of liposomes with Glut1-knockdown 4T1 cell

4T1 cells were transinfected with anti-Glut1 siRNA. After 72 hours of incubation, the glut1 protein expression level was examined by WB using primary anti-Glut1 antibodies (ab115730, Abcam). GAPDH (glyceraldehyde phosphate dehydrogenase) was used as a loading control. These Glut1-knockdown cells were treated with C-Lp/C6 and Rg3-Lp/C6 for 4 hours, and the cells were analyzed by flow cytometry (CytoFlex S, Beckman Coulter Inc., USA). For quantitative study, nuclei were then stained with DAPI (5 μg/ml), and Glut1 was stained according to the protocol for the “IF staining of Glut1” section before subjected to fluorescence microscopic observation (Zeiss LSM 710, Oberkochen, Germany).

In vitro cytotoxicity study

4T1 cells were seeded in 96-well plates at a density of 5 × 103 cells per well and cultured in complete medium for 24 hours. Then, the cells were incubated with a series of concentrations of DTX, C-Lp/DTX, Nanoxel-PM, Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX. After 48 hours, we add 100 μl of sterilized MTT solution (0.5 mg/ml in Hanks’ balanced salt solution) at 37°C for 4 hours. The MTT solution was removed and 150 μl of DMSO was added to each well and shaken at a speed of 100 rpm for 15 min to dissolve the formed formazan crystal. Absorbance at 570 nm was detected by Multiskan MK3 microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). Cell viability was calculated using untreated cells as control.

Apoptosis detection of CTC in blood

A total of 1 × 106 4T1 cells were spiked in the 300 μl of fresh blood and sheared with an equal volume of free DTX, C-Lp/DTX, Nanoxel-PM, Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX at the DTX concentration of 100 ng/ml for 4 hours at 37°C, 120 rpm. RBCs were lysed after platelets were removed by gradient density centrifugation. The remaining cells were then stained with PerCP-Cy5.5–labeled anti-CD45 to distinguish leukocytes from 4T1 cells and DAPI and FITC–annexin V to determine the variability of the cells. Briefly, cells were incubated in 300 μl of binding buffer containing 5 μl of annexin V–FITC and 5 μl of DAPI for 15 min at room temperature away from light. The extent of apoptosis was detected via a flow cytometer (FACSCalibur, BD, USA). For confocal observation, GFP-4T1 cells were applied instead; the following steps were the same as mentioned above. Annexin V–PE and DAPI were stained before visualization under a CLSM (Zeiss LSM 710, Oberkochen, Germany).

Migration assay of TAMs

TAMs were obtained by incubating RAW264.7 cells with interleukin-4 (IL-4) (20 ng/ml; PeproTech) overnight. TAM chemotaxis in response to 4T1-CM was determined by measuring the number of migrated cells through a polycarbonate filter with 5-mm pore size in 24-well Transwell chambers. The upper chamber contained 1 × 105 TAMs in 100 μl of RPMI 1640 containing 0.1% bovine serum albumin. Lower chambers contained 600 ml of 4T1-CM pretreated with PBS, DTX, C-Lp/DTX, Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX. After incubation (12 hours at 37°C), the upper TAMs of the chamber were wiped off with a cotton swab. TAM cells below the chamber were fixed with methanol for 20 min and stained by 0.1% crystal violet for 15 min, washed with double-distilled water for three times, and photographed; then, the crystal violet was dissolved by adding 0.1 ml of 33% acetic acid solution to each well and measured at 570 nm after shaking by Multiskan MK3 microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). The OD value of each group represents the number of TAM cells that migrated through the chamber membrane.

IF staining of Glut1

GFP-4T1 cells were transinfected with anti-Glut1siRNA by siRNA-mate. After 72 hours, cells were incubated with the primary anti-Glut1 antibody (ab115730, Abcam) overnight at 4°C; Cy3-labeled fluorescent secondary antibody (33108ES60, Yeasen) was then used to visualize Glut1. Nuclei were then stained with Hoechst 33342 (5 μg/ml) before subjected to fluorescence microscopy (Zeiss LSM 710, Oberkochen, Germany). All the fluorescence images were taken under identical conditions with the same exposure time.

IVFC assay

Mice were injected with 1 × 106 CFSE-labeled 4T1 cells via tail vein, followed by intravenous injection with 100 μl of Rg3-Lp/DiD or C-Lp/DiD at an equivalent DiD concentration of 1 μg/ml. The mice were then anesthetized and fixed on the working stage. An LED configured with a charge-coupled device camera was used to visualize major veins and arteries of the left ear. Lasers of 488 and 633 nm were focused onto the selected favorable artery with the emission wavelength 510 ± 10 and 670 ± 20 nm for the collection of CFSE-labeled cells and DiD-loaded liposomes, respectively.

Exploration of the interaction between STAT3 and CCL2

4T1 cells were seeded in six-well plates at a density of 5 × 105 cells per well and cultured in complete medium for 24 hours. Cells were then incubated with 2 μM stattic [MedChemExpress (MCE), NJ, USA] for 24 hours. IL-6 (20 ng/ml) was added to the medium 1 hour before cell and CM collection. The harvested cells and CMs were used for WB and ELISA analysis, respectively. The detailed protocol of WB and ELISA can be found in the “Western blot” and “ELISA of CCL2 levels” sections, respectively.

Targeting of liposomes to CTC colonies trapped in the lung

Rg3-Lp/DiD and C-Lp/DiD were injected into mice immediately after injection of Luc-4T1 cells. After 4 hours, the mice were intraperitoneally administered with 200 μl of d-luciferin (5 mg/ml) 5 min before their near-infrared (NIR) and BLI imaging under an in vivo imaging spectrum (IVIS) system (PerkinElmer, USA). Major organs were then collected and imaged immediately for NIR imaging and BLI imaging. The images were analyzed by Living Imaging software. Frozen slices of lung tissues were stained with anti-CD31 antibody (ab28364, 1/50 dilution) and anti-Glut1 antibody (ab115730, Abcam) at 4°C overnight, washed three times with PBS, then incubated with a secondary antibody Cy3-goat anti-rabbit immunoglobulin G (IgG) H&L (33108ES60, Yeasen) at 37°C for 2 hours, washed three times with PBS, and lastly stained with DAPI for 10 min before confocal observation (Zeiss LSM 710, Oberkochen, Germany).

Targeting of liposomes to already formed lung metastasis

NIR imaging and BLI were simultaneously performed by real-time IVIS. Fourteen days after model establishment, mice were divided in two groups and imaged at different time points after injected with Rg3-Lp/DiD and C-Lp/DiD (n = 3), respectively. After 24 hours, the mice were intraperitoneally administered with 200 μl of d-luciferin (5 mg/ml) 5 min before their imaging under the IVIS imaging system. Major organs were then collected and imaged immediately under the same system. Lung section and Glut1 IF staining was performed for CLSM observation.

Targeting of liposomes to Glut1-knockdown CTC colonies trapped in the lung

Except for adding the Glut1-knockdown groups treated with same volume of Glut1-knockdown Luc-4T1 cells, the experimental procedure was the same as described in the “Targeting of liposomes to CTC colonies trapped in the lung” subsection.

Metastasis inhibition assay

Mice were injected with 2 × 105 Luc-4 T1 cells through the tail vein (n = 4). Subsequently, DTX, C-Lp/DTX, Nanoxel-PM, Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX (at the DTX concentration of 5 mg/kg) were intravenously injected on days 0, 5, 10, 15, and 20 after Luc-4T1 cell injection, respectively. Mice treated with saline were taken as a control. The BLI images were conducted every 7 days to monitor the progression of metastasis. Twenty-one days after treatment, lungs of animals were excised and BLI-imaged. The obtained lung tissue of each tissue was used for quantification of lung-infiltrating lymphocytes, WB, PCR, and ELISA described in the following text.

Survival time investigation

A total of 2 × 105 Luc-4T1 cells were injected to the lateral tail vein of female nude mice to allow the formation of lung metastasis. At 4 hours after injection, in vivo BLI imaging was performed to confirm the localization of 4T1-Luc cells in mouse lungs using an IVIS system. Then, mice were randomized into six groups (n = 5 per group) and treated with PBS, DTX, C-Lp/DTX, Nanoxel-PM, Rg3, Rg3-Lp, Rg3/DTX, and Rg3-Lp/DTX (5 mg/kg per dosage, every 5 days) for 30 days. Lung metastasis of Luc-4T1 was monitored by weekly in vivo BLI imaging for up to 63 days. Mice were euthanized, and organs were excised for weighing. Lung metastatic nodules were then counted and morphologically examined. Lungs and other organs were embedded in paraffin and sectioned at 10 μm, after which the H&E, p-STAT3, CCL2, TAM, and MDSC staining assays were performed, respectively.

Quantification of lung-infiltrating lymphocytes

The excised lungs (n = 3) in the “Metastasis inhibition assay” section were ground and passed through a 200-mesh sieve. The obtained cell suspensions were costained for T cells (CD45, CD4, and CD8), MDSCs (CD45, Gr1, CD11b, and CD45), and TAMs (CD45, F4/80, and CD206) for FACS (fluorescence-activated cell sorting) analysis (BD Biosciences, USA). The obtained lung slices in the “Survival time investigation” section were stained to analysis the abundance of MDSCs and TAMs, respectively.

Western blot

4T1 cells after different treatments for 24 hours were washed with PBS and scraped into a lysis buffer containing the proteinase and phosphatase inhibitor cocktail. Protein concentrations were measured with the bicinchoninic acid (BCA) protein assay Kit (Beyotime Biotechnology, China). A total of 50 μg of protein lane was loaded and run on the polyacrylamide gel with a tris/glycine running buffer system and then transferred onto a polyvinylidene difluoride membrane. Anti-STAT3 (ab119352, Abcam), anti–p-STAT3 (ab75314, Abcam), anti–phospho-NF-κB p65 (GB11142, Servicebio), anti-actin (30102ES40, Yeasen), anti-Bax (ab32503, Abcam), and anti-Bcl2 (ab182858, Abcam) were incubated overnight at 4°C. The peroxidase-conjugated secondary antibody (Yeasen, China) was used, and the signals were detected by adding the enhanced chemiluminescence Western blotting detection reagents (Amersham Biosciences, Piscataway, NJ, USA). Lung tissues were obtained following the protocol from the “Survival time investigation” section. The tissue protein was obtained by grinding in a lysis buffer containing the proteinase and phosphatase inhibitor cocktail, and the following steps were the same as described before.

ELISA of CCL2 levels

ELISA was conducted to evaluate the level of CCL2 in vitro and in vivo. Concentrations of CCL2 in each 4T1 culture supernatant after different treatment for 24 hours were determined according to the manufacturer’s instructions (Lianke, Shanghai, China). The lung tissues were obtained according to the protocol from the “Metastasis inhibition assay” section; the lungs were homogenized with precooled PBS (5 ml of PBS per 1 g of tissue). The prepared homogenate was centrifuged at 5000g for 5 min, and the CCL2 concentration of CCL2 in the supernatant was detected with a microplate reader (450 nm) (Thermo Multiskan MK3, USA).

Quantitative PCR

Total RNAs were extracted from 4T1 cells with different treatments for 24 hours using TRIzol (Servicebio, China). Reverse transcription was performed. Quantitative real-time PCR analyses using the comparative cycle threshold (CT) method were performed on fluorescence qPCR instruments (7300, Applied Biosystems ABI). After an initial incubation at 95°C for 10 min, amplification was performed for 40 cycles at 95°C for 15 s and 60°C for 60 s. The mouse CCL2 primer pair was 5′-CCAGCAAGATGATCCCAATGAGT-3′ and 5′-CATTAGCTTCAGATTTACGG-GTC-3′. The mouse GAPDH primer pair was 5′-CCTCGTCCCGTAGACAAAATG-3′ and 5′-TGAGGTCAATGAAGGGGTCGT-3′.

Statistical analysis

Statistical analysis was performed using GraphPad Prism software version 7.01. Differences between two experimental groups were determined by two-tailed Student’s t test and among multiple groups by one-way analysis of variance (ANOVA) with Tukey’s post hoc test. For paired observations and repeated measurements over time, we used two-way ANOVA with Bonferroni’s multiple comparison posttest. All bar graphs show means ± SD. *P < 0.05, **P < 0.01, and ***P < 0.001.