M2 macrophage microvesicle-inspired nanovehicles improve accessibility to cancer cells and cancer stem cells in tumors

Graphical Abstract Cancer cells and cancer stem cells (CSCs) are the major players of cancer malignancy and metastasis, but they are extremely difficult to access. Inspired by the vital role of macrophages and microvesicle-mediated cell–cell communication in tumors, we herein designed M2 macrophage microvesicle-inspired nanovehicle of cabazitaxel (M-CFN) to promote accessibility to cancer cells and CSCs in tumors. In the 4T1 tumor model, M-CFN flexibly permeated the tumor mass, accessed cancer cells and CD90-positive cells, and significantly promoted their entry into CSC fractions in tumors. Moreover, M-CFN treatment profoundly eliminated aldehyde dehydrogenase (ALDH)-expressing CSCs in 4T1 and MCF-7 tumors, produced notable depression of tumor growth and caused 93.86% suppression of lung metastasis in 4T1 models. Therefore, the M2 macrophage microvesicle-inspired nanovehicle provides an encouraging strategy to penetrate the tumor tissues and access these insult cells in tumors for effective cancer therapy. Supplementary Information The online version contains supplementary material available at 10.1186/s12951-021-01143-5.


Introduction
Cancer cells and cancer stem cells (CSCs) are the major players that orchestrate cancer malignancy and metastasis in many types of solid tumors [1,2]. Cancer cells are a subgroup of cells that are heterogeneously distributed in tumor tissues, and CSCs are a small subpopulation of cancer cells with principal features of self-renewal, proliferation and differentiation, which is only 0.05-3% of total cancer cells in tumors [3][4][5]. Cancer cells and CSCs usually undergo a dynamic reciprocal transformation process [6,7]. CSCs can transform into rapidly dividing cancer cells to foster tumor progression, and cancer cells can evolve into CSCs to protect themselves from therapeutic intervention. Some commonly used anticancer drugs, such as paclitaxel, cisplatin, and sunitinib, efficiently damage rapidly dividing cancer cells [1,8,9], but they are ineffective in eliminating CSCs; worse still, they may unexpectedly enrich the ratio of CSCs in tumors [9][10][11]. To date, only a few therapeutic agents have been used alone to directly eliminate CSCs or in combination with modulatory agents to sensitize them to standard therapies [1,[12][13][14][15][16][17][18][19]. However, cancer cells and CSCs are frequently embedded in the dense tumor stroma comprising versatile stromal cells and extracellular matrix networks, making them extremely difficult access [5,20,21]. Rationally, therapeutic agents should be delivered across the tumor stroma barriers and access cancer cells and CSCs to exert anticancer effects [22,23]. Although nanovehicles hold great potential for accumulating at tumor sites, they are often trapped around the tumor vasculature and ineffective in penetrating the tumor tissues, drastically restricting their access to cancer cells and CSCs in tumors [24]. Therefore, it is impressively desired to develop novel delivery strategies to improve their accessibility to these insult cells for cancer therapy.
Cancer cells are usually embedded in the tumor stroma and surrounded by a variety of accessory cells, such as cancer-associated fibroblasts (CAFs), tumor-associated macrophages (TAMs) and endothelial cells etc. [25][26][27]. CSCs are believed to reside in CSC niches comprised of versatile accessory cells, extracellular matrix (ECM) components, and networks of cytokines and growth factors [6,28,29]. Of note, TAMs, usually M2-phenotype macrophages in tumors, are the major supportive cells in tumor microenvironments and CSC niches, that play a vital role in promoting cancer cell proliferation, metastasis and maintaining the CSC state [2,[30][31][32]. Compelling evidence has revealed that macrophages can interact with breast cancer CSCs via a variety of mechanisms to support CSC niches, such as the binding of CD11b to CD90 markers [33][34][35]. Interestingly, TAMs can communicate with cancer cells or CSCs in tumors via extracellular vesicles such as microvesicles or exosomes [29,[36][37][38][39][40][41], which provides a substantial opportunity to diffuse across the tumor stroma and access cancer cells and CSCs in tumors. Moreover, due to the possible immunogenicity of the macrophages from a different donor, autologous macrophages have been used in clinical trials to treat a variety of patients with tumors or other diseases, and have been validated to be safe in Phase I clinical trials for liver cirrhosis [42][43][44][45]. To date, few if any reports of M2-macrophage microvesicles have been explored as delivery vehicles to improve drug accessibility to these insult cells in tumors.
Based on this rationale, we herein designed a M2-macrophage microvesicle-inspired nanovehicles of cabazitaxel (CTX) (M-CFN) to promote their accessibility to cancer cells and CSCs in tumors to promote effective cancer therapy (Scheme 1). CTX, a chemotherapeutic drug with poor affinity to the efflux transporter P-glycoprotein, was selected as the therapeutic agent to kill cancer cells and CSCs [46,47]. M-CFN was developed by camouflaging cabazitaxel-loaded polyfluorocarbon nanoparticles (CFN) with M2-macrophage membranes to endow them with macrophage microvesicle-mimic features, thereby promoting their intratumoral permeation and improving accessibility to the cancer cell and CSC fractions in tumors. We investigated the capacity of M-CFN to promote tumor penetration and access cancer cells and CSCs in 4T1 tumors. Moreover, we measured the therapeutic benefits of M-CFN and its effects on eradicating CSCs in 4T1-and MCF-7-tumor models.
Murine 4T1 breast cancer cells, RAW 264.7 cells and human MCF-7 breast cancer cells were provided by the Cell Bank of Shanghai, Chinese Academy of Sciences (CAS, Shanghai, China). 4T1 cancer cells with stable expression of green fluorescence proteins (4T1-GFP) were supplied by Sciencelight Biology Science and Technology Co., Ltd (Shanghai, China). 4T1 and 4T1-GFP cells were incubated with RPMI-1640 media containing 10% FBS (Gibco) at 37 °C and 5% CO 2 in a humidified incubator, while RAW 264.7 and MCF-7 cells were cultured using Dulbecco's modified Eagle medium (DMEM) with 10% FBS (Gibco) under the same conditions. Female BALB/c mice (18-22 g) and BALB/C-nu nude mice (18-22 g) were obtained from the Shanghai Experimental Animal Center, CAS (Shanghai, China) to develop tumor models for further measurements. To develop the 4T1-or 4T1-GFP-induced tumor model, approximately 1,000,000 cells were injected into the second mammary pad of BALB/C mice. To induce the MCF-7 breast cancer model, approximately 1 × 10 7 cells were subcutaneously injected into BALB/C-nu mice. When the tumor size reached approximately 200 mm 3 , the tumor mass was collected, cut into small pieces and implanted into the second mammary pads of BALB/C-nu mice for further measurements. These animal experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee of Shanghai Institute of Materia Medica (SIMM), CAS.

Preparation of M-CFN
M-CFN was prepared in two steps: (i) preparation of CFN, and (ii) isolation of M2 macrophage membranes and camouflaging them onto CFN to form M-CFN.
CFN was prepared with PFP, PCL-PEG and CTX (10:10:1, w/w) using an emulsion-evaporation technique. In brief, these ingredients were dissolved in trichloromethane, dropped into double-distilled water at a 1:5 volume ratio and then emulsified using a sonication probe (JYD-650L, Shanghai, China). Then, the emulsion was evaporated under reduced pressure (B-480, Buchi, Switzerland) to remove trichloromethane to obtain CFN. The CTX concentration in CFN solution was approximately 1.0 mg/mL. M2 macrophages were induced from RAW 264.7 cells by incubation with 40 ng/mL interleukin 4 (IL-4) (Pep-roTech, USA) in DMEM for 48 h. Then, the phenotype of the induced macrophages was examined by flow cytometry analysis (FACSCalibur, BD Biosciences) after incubation with anti-CD86-APC and anti-CD206-PE antibodies according to the manufacturer's protocols. M2 macrophages were denoted as CD86 − CD206 + cells. Thereafter, the induced M2 macrophages were collected to isolate the macrophage membranes according to our previously described method [48].
M-CFN was prepared by camouflaging CFN with a macrophage membrane using an extrusion method. In brief, 1 mL of CFN (CTX, 1.0 mg/mL) was mixed with

Characterization of CFN and M-CFN
The morphologies of CFN and M-CFN were determined using transmission electron microscopy (TEM, 120 kV, Talos, FEI, USA) after negative staining with saturated uranyl acetate solution. The size distribution and zeta potential values were measured using dynamic light scattering (DLS) analysis on a Malvern Nano ZS90 analyzer (Malvern, UK). The encapsulation efficiency (EE) values of CTX in CFN and M-CFM were determined using high performance liquid chromatography (HPLC) analysis. The unentrapped CTX was separated from the CFN or M-CFN by centrifugation using an ultrafiltration tube (10 KDa) at 8000g for 20 min. The drug levels in the total solution and filtration were determined by HPLC analysis (Waters, USA) under the following conditions: column, XBridge C18 Column (250 mm × 4.6 mm, 5 μm i.d., Waters); mobile phase, acetonitrile-water (70:30, v/v); flow rate, 1.0 mL/min; detection wavelength, 254 nm. The EE values (%) were defined as the ratio of the amount of entrapped drug compared to the total drug amount in the nanovehicles.
To evaluate their stability in the mimicked physiological fluids, CFN and M-CFN were incubated with phosphate buffered saline (PBS, pH 7.4) at room temperature for 24 h and the particle size was determined using the Nano ZS90 analyzer at predetermined time points. Moreover, CFN and M-CFN were incubated in PBS (pH 7.4) and complete fetal bovine serum (FBS) at 37 °C for 30 h. At certain time intervals, samples were collected, and EE values were monitored by HPLC analysis as described above to characterize the percentage of drug remaining in the nanovehicles.

In vitro cellular uptake in parent 4T1 cancer cells and 3D tumorspheres
The cellular uptake of CFN and M-CFN was measured in parental 4T1 cells and 4T1-induced 3D tumorspheres using confocal laser scanning microscopy (CLSM) (TCS-SP8 STED, Leica, Germany) and fluorescence spectrum analysis (EnSpire, PerkinElmer, Singapore). CFN and M-CFN were labeled with a hydrophobic probe of DiIC18(3) (DiI) by physical encapsulation for the measurements.
To visualize their uptake into parental 4T1 cells under CSLM, cells were seeded into a 24-well plate using a round coverslip at 50, 000 cells/well and cultured overnight. DiI-labeled CFN and M-CFN were added to each well at 1 µg/mL DiI, and incubated for 4 h. Then, the cells were stained with Hoechst 33342 for visualization under CSLM. To determine their uptake in CSC-enriched 3D tumorspheres, the tumorspheres were incubated with DiI-labeled CFN and M-CFN at 1 µg/mL DiI for 4 h, and counterstained with Hoechst 33342 for CLSM detection. For quantification, the parent 4T1 cells and tumorspheres were incubated with DiI-labeled CFN and M-CFN at 1 µg/mL DiI for 4 h and then harvested for quantification by fluorescence spectrum analysis (EnSpire, PerkinElmer, Singapore).

Cytotoxicity in parent 4T1 cancer cells and CSC-enriched 3D tumorsphere cells
The cytotoxicity of CFN and M-CFN was measured in the CSC-enriched 4T1 tumorspheres and parental 4T1 cancer cells. To evaluate cytotoxicity in CSC-enriched tumorsphere cells, tumorspheres that were developed from 5000 cells in 24-well ultralow attachment plates were treated with free CTX, CFN and M-CFN at CTX concentrations ranging from 5 ng/mL to 50 μg/mL or comparable doses. After 5 days of incubation, the cell viability from each group was measured using a cell counting kit-8 (CCK8). In contrast, to evaluate the cytotoxicity in parental 4T1 cells, cells were seeded into 96-well plates at 3,000 cells/well and incubated overnight. Then, free CTX, CFN and M-CFN were added to each well with CTX ranging from 5 ng/mL to 50 μg/mL or comparable doses and incubated for an additional 48 h. Cells without any treatment were used as a negative control. The cell viability from each treatment was measured using the CCK8 assay kit on a microreader analyzer (EnSpire, PerkinElmer, Singapore).

Effects on inhibiting tumorsphere formation and eliminating ALDH-positive CSCs
The effects of M-CFN on inhibiting tumorsphere formation and eliminating ALDH-positive CSCs were measured in 4T1 and MCF-7 cancer cell models. In brief, 4T1 cells or MCF-7 cells were pretreated with free CTX, CFN or M-CFN at 0.25 μg/mL CTX for 12 h. Cells without any treatment were used as controls. Then, these cells were collected and seeded into 6-well ultralow attached plates at 40,000 cells/well, and incubated with the aforementioned serum free DMEM/F12 culture media for 7 days to form the tumorspheres. The tumorspheres from each treatment were imaged under an inverted microscope (IX81, Olympus, Japan). Thereafter, the tumorspheres from each group were collected, dissociated into single cells and stained using an Aldefluor ™ kit (#01700, STEM-CELL Technologies). The proportion of ALDH-positive CSC fractions in tumorspheres was monitored using flow cytometry analysis (FACSCalibur, BD, USA).

In vivo tumor accumulation, permeation and cancer cell access
The in vivo tumor accumulation of M-CFN and CFN was measured in 4T1-induced tumor models. For in vivo imaging, CFN and M-CFN were labeled with a hydrophobic dye of DiIC18(7) (DiR) by physical encapsulation. When the tumor size reached approximately 100 mm 3 , DiR-labeled CFN and M-CFN were injected into tumor-bearing mice via the tail vein at 1.0 mg/kg DiR. At predetermined time points, mice were anesthetized and the fluorescence signals of DiR from these two groups were monitored using an in vivo imaging system (IVIS Spectrum, PerkinElmer, USA). At 8.0 h postinjection, mice were necropsied, and the major organs including the heart, liver, spleen, lung, kidney and tumor from each group were collected and imaged using the in vivo imaging system.
To quantify the drug distribution in the major organs, tumor-bearing mice were intravenously administered free CTX, CFN and M-CFN at 10 mg/kg CTX (n = 3). Eight hours after injection, mice were necropsied, and the major organs from each treatment were collected, homogenized with 0.5 mL of dimethyl sulfoxide, and centrifuged at 10,000g for 10 min. The drug concentration in the supernatant was determined using HPLC analysis as described above.
To investigate the intratumoral permeation behaviors, DiI-labeled CFN and M-CFN were intravenously administered to tumor-bearing mice at 1.0 mg/kg DiI. Eight hours after injection, mice were necropsied and the tumor tissues were collected for cryostat sectioning at 10 µm (CM1950, Leica). Afterward, the sections were counterstained with DAPI for visualization under CLSM (TCS-SP8 STED, Leica, Germany). The fluorescence signals of DiI in the entire tumor sections were documented to outline their intratumoral distribution. The captured images were analyzed using ImageJ software (National Institutes of Health, Bethesda, MD, USA) for better clarification.
To evaluate their accessibility to the cancer cell fraction in tumors, the DiI-loaded CFN and M-CFN were injected into 4T1-GFP-induced tumor models via the tail vein at 1.0 mg/kg DiI. At 8.0 h postinjection, the tumor tissues were removed, embedded in frozen media and sectioned at 10 µm on a cryotome (Leica 1950, Germany). The sections were stained with DAPI for measurement. The green signals of cancer cell clusters, red signals of these two nanovehicles and the blue signals of nuclei were recorded to evaluate their accessibility to the cancer cell regions in tumors. The images were further analyzed using the ImageJ software.
To measure their access to CD90-expressing cancer cells and CSCs in 4T1 tumors, the tumor sections were treated with a primary antibody against CD90 (Affinity, DF6849, 1:200) followed by Alexa Fluor 488-labeled goat anti-rabbit IgG (H + L) (Yeasen, 1:200) for CLSM detection. The access of nanovehicles to the CD90-positive regions in tumors was denoted as the colocalization of these two fluorescence signals, which was further analyzed using the ImageJ software. Moreover, the access of CFN and M-CFN to CD90-expressing cells in tumors was determined using flow cytometric analysis. In brief, tumor tissues from DiI-labeled CFN-and M-CFN-treated groups were collected, cut into small pieces and treated with digestion buffer containing 0.2 mg/mL collagenase type IV (Yeasen, 40510ES60), 0.2 mg/mL hyaluronidase (Yeasen, 20426ES60) and 0.1 mg/mL dexoyridonuclease I (Yeasen, 10608ES25) at 37 °C for 2 h. The cell suspensions were collected by centrifugation, and then incubated with PE-anti-CD90 (Biolegend, 202,523, 1:50) at 4 °C for 1 h. The uptake of CFN or M-CFN into CD90-expressing cells was determined using flow cytometry.

In vivo access to CSC fractions in tumor
The in vivo access of CFN and M-CFN to the CSC fractions in tumors was measured in 4T1-induced tumor models. DiI-labeled CFN and M-CFN were intravenously administered to tumor-bearing mice at 1.0 mg/kg DiI. At 8.0 h postinjection, the tumor tissues were removed, frozen and sectioned at 10 µm (Leica 1950, Germany). The tumor sections were stained with primary antibodies against ALDH1A1 (Abcam, ab52492, 1:100), SOX2 (Abcam, ab97959, 1:1000) and Oct4 (Abcam, ab200834, 1:50), followed by Alexa Fluor 488-labeled IgG (H + L) secondary antibody (Beyotime, Jiangsu, China). In contrast, these sections were stained with DAPI for CLSM detection. The access of nanovehicles to these CSC fractions were denoted as the overlap of the red fluorescence signals of nanovehicles and the green signals of typical markers of CSCs in tumor sections. These images were analyzed using Image-Pro Analyzer 3D 7.0 (Media Cybernetics, MD, USA) to monitor the colocalization of nanovehicles with CSCs in tumors and were analyzed using ImageJ software to evaluate the access of nanovehicles to CSC fractions in tumors.

In vivo therapeutic efficacy
The in vivo therapeutic efficacy of M-CFN was measured in a 4T1-induced metastatic breast cancer model, which was developed as described above. When the tumor volume reached approximately 100 mm 3 , the tumor-bearing mice were intravenously administered PBS control, MM, free CTX, CFN and M-CFN at 10 mg/kg CTX or comparable doses every three days for a total of six injections (n = 5). The tumor size and body weights from each group were monitored. After six cycles of injection, mice were necropsied and the tumor tissues from each treatment were collected, imaged and accurately weighed to calculate the tumor inhibition effects. To assess apoptosis in tumors, tumor tissues were subjected to TUNEL assays. Concurrently, the lungs from each treatment were collected and the visualized metastatic nodules in each lung were counted to calculate the inhibitory effects on lung metastasis. Furthermore, histological examinations of the lung tissues from each treatment were performed using hematoxylin and eosin (H&E) staining to detect metastatic foci in the lungs.
To confirm the therapeutic benefits of M-CFN, we evaluated antitumor efficacy in an MCF-7-induced tumor model. When the tumor size reached 100 mm 3 , tumor-bearing mice were injected with PBS control, free CTX, CFN or M-CFN at 5 mg/kg CTX on days 1 and 6 for a total of two injections. During the treatments, the tumor volume and body weight changes were recorded. When the tumor volume in the PBS group was approximately 1000 mm 3 , the experiments were terminated, and the tumor tissues from each treatment were collected, imaged and weighed to evaluate the inhibitory effects on tumor growth.

In vivo effects on eradicating ALDH-expressing CSC fractions in tumors
To detect the CSC-eradicating ability of M-CFN, the tumor tissues from each group in 4T1-and MCF-7 tumor models were embedded in paraffin and sectioned at 5 µm. The sections were successively incubated with anti-ALDH1A1 (Abcam, ab52492, 1:100) and Alexa Fluor 488 labeled IgG (H + L) (Beyotime, Jiangsu, China). In contrast, the nuclei were stained with DAPI and monitored using a pannoramic scan (Pannoramic MIDI, 3D Histech, Hungary). The ALDH-positive (ALDH high ) CSC fractions in tumors are denoted as green fluorescence signals. The captured images were analyzed using ImageJ software for quantification.

Statistical analysis
Data are expressed as the mean standard deviation (SD) and were analyzed using Student's t-test. Statistical differences were significant at *p < 0.05 and very significant at **p < 0.01.

Preparation and characterizations of M-CFN
Initially, M-CFN was fabricated in two steps: (1) preparing CFN with CTX and amphiphilic polyfluorocarbon polymeric materials (PFP) and poly(ε-caprolactone)block-poly(ethylene glycol) (PCL-PEG); (2) inducing M2 macrophages from murine RAW264.7 cells and isolating M2 macrophage membranes to camouflage CFN via an extrusion method to form M-CFN. Details of the synthesis and characterization of PFP are provided in the Additional file 1: Figs. S1, S2. M2 macrophages were induced by incubating RAW264.7 cells with 40 ng/mL IL-4 for 48 h and characterized by flow cytometry analysis, which were denoted as CD86 − CD206 + cells (Fig. 1A). The TEM images indicated that both CFN and M-CFN were nanosized spherical particles with a size less than 50 nm (Fig. 1B, C). In contrast to CFN, M-CFN displayed a typical core-shell nanostructure (Fig. 1C). When CFN was decorated with the isolated macrophage membrane, the average hydrodynamic diameter increased from 54.79 ± 1.50 nm for CFN to 73.59 ± 1.92 nm for M-CFN (Fig. 1D). Meanwhile, the zeta potential values were altered from 0.74 ± 0.25 mV for CFN to -9.24 ± 1.93 mV for M-CFN. Moreover, the protein profiles in macrophages, macrophage membranes (MM), CFN and M-CFN were determined by electrophoresis assays. As shown in Fig. 1E, no protein signals were detected in CFN, but the major protein compositions in macrophage membrane were largely retained in M-CFN. Meanwhile, western blotting analysis indicated that typical markers of macrophages, including CD11b, colony stimulating factor 1 receptor (CSF1R), β1-integrins, α4-integrins and C-C motif chemokine receptor 2 (CCR2), were readily detected in the macrophage derived microvesicles, MM and M-CFN, but were undetectable in their counterpart CFN (Fig. 1F). These data indicated that MM preserves the typical markers of macrophage microvesicles, and MM was effectively camouflaged onto CFN in the M-CFN system, which may endow them with macrophage microvesicle-mimicking features. In addition, the expression of CD11b, a surrogate marker of macrophages that recognizes CSCs in tumors, was also confirmed by flow cytometry analysis (Additional file 1: Fig.  S3). Increasing data have shown that extracellular vesicles play a key role in tumor progression and could be used as effective delivery vehicles for targeted drug delivery [50,51]. Concurrently, TAMs can effectively communicate with CSCs or cancer cells in tumors with extracellular vesicles such as microvesicles or exosomes [2,6,36]. Accordingly, the combination of nanosized vehicles and macrophage microvesicle-mimicking MM in the M-CFN system provides a substantial opportunity to improve accessibility to CSCs or cancer cells in tumors. Although the CD11b marker is constitutively expressed in both M1 and M2 macrophages, the highly expressed markers CSF1R, CCR2 and α4β1 integrins on the M2 macrophage surface facilitate their accumulation at tumor sites and promote their accessibility to cancer cells and CSCs in tumors [52].
Then, we investigated the EE of CTX in CFN and M-CFN using HPLC analysis. The measured results indicated that the EE value of CTX was 99.79 ± 0.01% for CFN and 99.64 ± 0.11% for M-CFN, indicating the high encapsulation of CTX in these two nanoparticles. Meanwhile, upon incubation in phosphate buffered saline (PBS, pH 7.4) or fetal bovine serum (FBS), the EE values of CTX in CFN and M-CFN barely changed with incubation time (Fig. 1E, F). After 30 h of incubation, over 95% of CTX remained in CFN and M-CFN. In addition, when they were incubated in PBS (pH 7.4) for 24 h, no changes in the mean diameter of CFN and M-CFN were detected (Additional file 1: Fig. S4). These data confirmed their good stability in mimicked physiological fluids, suggesting their feasibility for in vivo delivery.

Cellular uptake and CSCs-eradicating efficacy
The cellular uptake of M-CFN was determined in comparison to CFN in CSC-enriched 3D tumorspheres and parental 4T1 cancer cells using confocal laser scanning microscopy (CLSM) and fluorescence spectra analysis. The results indicated that M-CFN was effectively internalized into both parent 4T1 cells and CSC-enriched tumorspheres with a fluorescence intensity stronger than that of CFN ( Fig. 2A, B). Compared with CFN, the cellular uptake of M-CFN increased 1.43-fold in parental 4T1 cancer cells and 1.54-fold in CSC-enriched tumorspheres (Fig. 2C, D). CD11b expressed on macrophages is crucial for their binding to the CD90 markers of breast cancer CSC [2]. Expression of CD90 in CSC-enriched 4T1 tumorsphere cells and parental 4T1 cancer cells was verified by immunofluorescence imaging, which could interact with CD11b in macrophages (Additional file 1: Fig. S5). To investigate the role of CD11b in M-CFN preferential uptake, we blocked CD11b on the M-CFN particles with anti-CD11b antibodies and measured their uptake in the CSC tumorsphere model. Compared to M-CFN, the uptake of anti-CD11b treated M-CFN was significantly reduced by 32.50% (Fig. 2E), indicating the important role of CD11b in the preferential cellular uptake of M-CFN in CSC-enriched tumorsphere cells.
In the CSC-enriched 4T1 tumorsphere cells and parental 4T1 cancer cells, free CTX, CFN and M-CFN exhibited considerable cytotoxicity in a concentrationdependent manner (Fig. 2F, Additional file 1: Fig. S6). Then, we evaluated the effect of M-CFN on inhibiting the self-renewal ability of 4T1 and MCF-7 breast cancer cells using a tumorsphere-forming assay. The residual 4T1 or MCF-7 cancer cells from the CTX-, CFN-and M-CFN-(0.25 µg/mL CTX or comparable concentrations)-treated groups were cultured in ultralow attachment plates for tumorsphere-formation (Fig. 2G). After 7 days, typical large tumorspheres were clearly observed in the control group, but only a few small cell spheres were detected in the CTX and CFN groups. Importantly, small cell clusters or single cells were visualized in the M-CFN treated group, suggesting their considerable inhibition of the self-renewal ability of residual cancer cells. Moreover, to validate the CSC-eradicating effects, we measured the proportion of ALDH-expressing CSCs in tumorspheres from these treatments using flow cytometry analysis (Fig. 3). In 4T1-induced tumorsphere models, the percentage of ALDH high CSCs dramatically decreased from 27.84 ± 3.31% in the control group to 4.61 ± 1.09% in the M-CFN group, resulting in an 83.45% eradication of ALDH high CSCs. Similarly, in MCF-7 tumorsphere models, the proportion of ALDH high CSCs was obviously reduced from 27.0 ± 5.37% in the control group to 7.94 ± 0.40% in the M-CFN group, resulting in a 70.58% eradication of ALDH high CSCs. These data confirmed the more effective CSC-eradicating effects of M-CFN than free CTX and CFN (**p < 0.01), which could be primarily attributed to the potent cytotoxicity of CTX and its effective incorporation into the M-CFN system.

In vivo tumor accumulation, permeation and CSCs access
The efficient accumulation and deep penetration of nanoparticles in tumor tissues represent a prerequisite for accessing cancer cell and CSC fractions that are frequently located in the deep interior regions of the tumor mass [21,22,24,53]. The in vivo imaging results indicated that the red fluorescence signals of CFN and M-CFN were clearly detected at the tumor sites with high fluorescence intensity (Fig. 4A). The preferential tumor accumulation of CFN and M-CFN was also validated by ex vivo imaging of the major organs (Fig. 4B). At 8.0 h postinjection, CTX levels in tumors from the M-CFN group increased 4.17-fold compared with free CTX but was comparable to that of the CFN group (Fig. 4C). Then, we measured the intratumor permeation of CFN and M-CFN in 4T1 tumors. The red fluorescence signals of M-CFN were extensively visualized in the entire profile of tumor tissue, but those of CFN were only detected in the exterior regions (Fig. 4D). The flexible intratumor permeation of M-CFN was also confirmed by image analysis (Fig. 4E). These data confirmed the notable efficacy of M-CFN in promoting tumor accumulation and intratumor permeation, which was conducive to facilitating their access to the cancer cell and CSC fractions in tumors.
Subsequently, we measured the accessibility of these two nanovehicles to the cancer cell fractions in tumors developed from 4T1 cancer cells with stable expression of green fluorescence proteins (4T1-GFP) (Fig. 5). The red fluorescence signals of M-CFN largely overlapped with the green signal-clustered 4T1-GFP regions with high intensity. In contrast, the red fluorescence signals of CFN were primarily localized in regions near the green fluorescence-labeled cancer cell clusters (Fig. 5A). Meanwhile, imaging analysis indicated that the signal profiles of M-CFN and 4T1-GFP cancer cells largely coincided, while those of CFN and 4T1-GFP were two separate peaks with negligible overlaps (Fig. 5B). Meanwhile, the enhanced effects of M-CFN over CFN in accessing the CD90-positive cell fractions in 4T1 tumors were also validated by CLSM imaging and image analysis (Fig. 5C, D). Flow cytometry examinations indicated that uptake of M-CFN in the CD90-expressing cells of 4T1 tumors was considerably enhanced by 2.94-fold in comparison to that in CFN-treated tumors (Additional file 1: Fig. S7). These data effectively confirmed the superiority of M-CFN over CFN in accessing cancer cells and CD90-expressing cells in tumors. By analyzing the formulation of M-CFN and CFN, the advantage of M-CFN in accessing cancer cells and CD90-positive cells in tumors was primarily due to macrophage camouflaging in M-CFN. In addition, the data from the in vitro cellular uptake and in vivo access to CD90-expressing cells effectively confirmed the significance of the CD11b-CD90 interactions in the preferential targeting abilities of the M-CFN particles.
Importantly, CSCs are a rare subpopulation of cancer cells in heterogeneous tumors that account for the tumor progression and metastasis [3,7]. The accessibility of M-CFN to the CSC fractions in 4T1 tumors was evaluated using CLSM imaging. ALDH1A1, SOX2 and Oct4 are typical characteristic markers of breast cancer CSCs [7,53]. To evaluate their CSC-accessing ability, the CSCs in 4T1-induced tumors were labeled with these specific markers by immunofluorescence staining for CLSM detection. As shown in Fig. 6A and C, the red fluorescence signals of M-CFN were considerably colocalized with cells labelled using the specific CSC markers ALDH1A1, SOX2 and Oct4, but the signals of CFN barely merged with these CSC markers. Furthermore, these images were analyzed to confirm the CSC-accessing capacity of M-CFN. The colocalization of M-CFN with ALDHA1-, SOX2-or Oct4-expressing CSC fractions in tumors was significantly higher than that of CFN ( Fig. 6A and C). Compared with CFN, the colocalization index of M-CFN with CSCs was significantly increased by 60.19-, 7.39-and 5.82-fold, respectively. Meanwhile, image analysis indicated that the signals of M-CFN largely overlapped with those of the CSCs with specific markers, but those of CFN and CSCs barely colocalized, indicating the preferential access of M-CFN to the CSC fractions in tumors ( Fig. 6B and D). Therefore, M-CFN exhibited profound efficiency in promoting intratumor transport and improving accessibility to the cancer cell and CSC fractions in tumors, which was much more efficient than unmodified CFN. Due to the robust physical barriers in tumor tissues, commonly used nanomedicines are usually trapped around the tumor vessels and ineffective penetrating the tumor tissues. The bioinspired design of the M-CFN system could circumvent these obstacles and promote their intratumor delivery. Considering the flexible intratumor permeating ability of macrophage microvesicles and the CD11b-CD90 interactions between TAM and CSC fractions, the macrophage microvesicle-inspired features of the M-CFN system could be the major contributor to the profound CSCaccessing capacity.

In vivo therapeutic efficacy on tumor growth and metastasis
Encouraged by the prevailing CSC-accessing ability of M-CFN, we finally measured their antitumor efficacy in 4T1-induced and MCF-7 induced breast cancer models (Fig. 7). In the tumor growth profiles of 4T1 metastatic tumor models, tumor progression was inhibited by free CTX, CFN and M-CFN but was mostly unaffected by MM treatment (Fig. 7A).  Fig. S9). In addition, the body weight was mostly unchanged in response to these treatments (Additional file 1: Fig. S10). More importantly, we evaluated the effect of M-CFN on eradicating CSC fractions in tumors by immunofluorescence assays, which were denoted as green fluorescence signals. As shown in Fig. 7C, the green fluorescence signals in the M-CFN-treated group were barely detected, and the fluorescence intensity was much weaker than that of the other treatments. The quantified results revealed that the fractions of ALDH high CSCs in tumors in the M-CFN-treated group were significantly reduced by 93.24% compared to the PBS control (Fig. 7D). Moreover, in the M-CFN-treated group, the proportion of ALDH high CSCs in tumors was only 2.91% of that in the free CTX group and 6.09% of that in the CFN group, confirming the superior efficacy of M-CFN in eradicating ALDH high CSCs in tumors. The better CSC-eradicating efficacy of M-CFN over CFN was primarily attributed to its prominent CSC-accessing capacity owing to its macrophage microvesicle-mimicking properties.
Concurrently, we measured the efficacy of M-CFN in suppressing the incidence of lung metastasis. The lungs are one of the most frequent organs of breast cancer metastasis in both preclinical models and clinical patients [54,55]. Visualized metastatic nodules in the lungs from each group were recorded to evaluate treatment efficacy in inhibiting lung metastasis (Fig. 7E, Additional file 1: Table S1). M-CFN treatment caused 93.86% inhibition of lung metastasis compared to the PBS control. Moreover, the average number of metastatic nodules in the lung from the M-CFN-treated group was only 11.11% that of the free CTX group and 21.88% that of the CFN group. The antimetastatic effects of M-CFN were also confirmed by histological examinations of the metastatic foci in lungs from each treatment (Fig. 7F). These results suggested the prevailing efficacy of M-CFN in suppressing the incidence of lung metastasis, which could be related to the profound eradication of CSC fractions in the primary tumor sites.
To further confirm the efficacy of M-CFN, we measured the therapeutic effects on inhibiting tumor growth and eliminating CSCs in human MCF-7 induced tumor models ( Fig. 7G-J). In the tumor growth profiles, tumor size was moderately inhibited by free CTX and CFN but significantly retarded by M-CFN treatment (Fig. 7G). The body weight of each group rarely changed during these treatments (Additional file 1: Fig. S11). At the final time point, tumor size in the M-CFN-treated group decreased by 73.91% compared to the PBS control and was much lower than that in the free CTX and CFN groups. The tumor inhibitory effects of M-CFN were also validated by measuring the tumor weights from each treatment (Fig. 7H, Additional file 1: Fig. S12). Meanwhile, M-CFN treatment produced the best efficacy in reducing the ALDH high CSC fractions in tumors (Fig. 7I). Compared with the PBS control, M-CFN treatment caused a 92.52% reduction in ALDH high CSC fractions in MCF-7 tumors (Fig. 7J). Moreover, the proportion of ALDH high CSCs in tumors from the M-CFN group was only 14.11% of free CTX and 23.20% of CFN. As a result, M-CFN treatment produced notable efficacy in suppressing tumor growth and eliminating the ALDH high CSC fractions in MCF-7 tumors. From these data, it can be deduced that the bioinspired design of the M-CFN system could effectively combine the advantages of the nanotool delivery system and the tumor-tropic features of M2 macrophages and demonstrated encouraging potential for targeting the cancer cell and CSC fractions in tumors for effective cancer treatments. However, due to the possible immunogenicity of macrophages from a different donor or species, this design of macrophage microvesicle-inspired nanovehicles may be limited to autologous macrophages. In addition, in view of the high expression of some immune checkpoints on the M2 macrophage surface [44], their impacts on cancer immunotherapy will be considered in our future studies.

Conclusions
In summary, we successfully designed a M2 macrophage microvesicle-inspired nanovehicle of cabazitaxel (M-CFN) by camouflaging macrophage membranes onto CFN to preferentially access cancer cells and CSCs in tumors for cancer therapy. In the 4T1 tumor model, M-CFN effectively permeated across the tumor mass and accessed both the cancer cell and CSC fractions. Moreover, M-CFN effectively promoted accessibility to the ALDHA1-, SOX2-or Oct4-expressing CSC fractions in tumor. Importantly, M-CFN treatment profoundly eliminated the ALDH high CSC fractions in 4T1-and MCF-7-induced tumors Fig. 7 In vivo therapeutic efficacy and eradication of ALDH high CSC fractions in tumors. A Tumor growth profiles from each group in 4T1 tumor models (n = 5). B Tumor weights from each group in 4T1 tumor models. C Effects on eradicating the ALDH high CSCs in 4T1 tumors, wherein the ALDH high CSCs were denoted as green fluorescence signals, scale bar, 50 μm. D Quantified effects on eradiating ALDH high CSCs in 4T1 tumors, **p < 0.01. E Average number of lung metastatic nodules from each group in 4T1 tumor models, **p < 0.01. F Histological examinations of lungs from each treatment in 4T1 tumor models, scale bar, 200 μm. The metastatic foci were denoted as cell clusters with darkly stained nuclei (red arrows). G Tumor growth profiles from each group in MCF-7 tumor models (n = 5). H Tumor weights from each group in MCF-7 tumor models, *p < 0.05, **p < 0.01. I Effects on eradicating the ALDH high CSCs in MCF-7 tumors, wherein the ALDH high CSCs are denoted as green fluorescence signals, scale bar, 50 μm. J Quantified effects on eradiating ALDH high CSCs in MCF-7 tumors, **p < 0.01 and produced notable therapeutic benefits in these two breast cancer models. Therefore, M2 macrophage microvesicle-inspired nanovehicles represent an encouraging bioinspired drug delivery nanoplatform for permeating tumor tissues and accessing cancer cells and CSCs in tumors for effective cancer therapy.