Dual-Targeted Extracellular Vesicles to Facilitate Combined Therapies for Neuroendocrine Cancer Treatment

Neuroendocrine (NE) cancers arise from cells within the neuroendocrine system. Chemotherapies and endoradiotherapy have been developed, but their clinical efficacy is limited. The objective of this study was to develop a dual-targeted extracellular vesicles (EV)-delivered combined therapies to treat NE cancer. Specifically, we produced EV in stirred-tank bioreactors and surface tagged both anti-somatostatin receptor 2 (SSTR 2) monoclonal antibody (mAb) and anti-C-X-C motif chemokine receptor 4 (CXCR4) mAb to generate mAbs-EV. Both live-cell confocal microscopy imaging and In Vivo Imaging System (IVIS) imaging confirmed that mAbs-EV specifically targeted and accumulated in NE cancer cells and NE tumor xenografts. Then the highly potent natural cytotoxic marine compound verrucarin A (Ver-A) with IC50 of 2.2–2.8 nM and microtubule polymerization inhibitor mertansine (DM1) with IC50 of 3.1–4.2 nM were packed into mAbs-EV. The in vivo maximum tolerated dose study performed in non-tumor-bearing mice indicated minimal systemic toxicity of mAbs-EV-Ver-A/DM1. Finally, the in vivo anticancer efficacy study demonstrated that the SSTR2/CXCR4 dual-targeted EV-Ver-A/DM1 is more effective to inhibit NE tumor growth than the single targeting and single drug. The results from this study could expand the application of EV to targeting deliver the combined potent chemotherapies for cancer treatment.


Cell Lines, Seed Cultures and Media
The metastatic pancreatic NET cell lines BON-1 (kindly provided by Dr. Mark Hellmich at University of Texas, USA) and QGP-1 (AcceGen Biotechnology, Fairfield, NJ, USA), BON-FLuc generated by stably transfecting BON cells with luciferase-expressing plasmid pGL4. 50 was maintained in DMEM with 10% FBS, 1% nonessential amino acids, and 1% sodium pyruvate at 37 • C and 5% CO 2 in a humidified incubator (Caron, Marietta, OH, USA). The seed train of anti-SSTR2 mAb producing hybridoma was maintained in hybridoma serum free medium with 8 mM L-glutamine in 125-mL shaker flasks at 37 • C, 5% CO 2 and 130 rpm. The human embryonic kidney 293 (HEK 293F) cells (Life Technologies) were maintained in chemically defined HEK 293 medium with 4 mM GlutaMAX in 125 or 250-mL shaker flasks. The viable cell density (VCD) and viability was measured using a Countess II automated cell counter (Fisher Scientific, Waltham, MA, USA). All basal media and nutrient supplements were purchased from Thermo Fisher Scientific (Waltham, MA, USA) or Life Technologies unless otherwise specified.

Construction of mAbs-EV-Ver-A/DM1
First, the EV were produced using our previously published platform [24]. Briefly, 2 L stirred-tank bioreactors (Vineland, NJ, USA) were utilized to conduct EV production. The suspension HEK 293F cells were seeded into the bioreactor vessel with 1.5-L FreeStyle TM 293 expression medium (GlutaMAX) at seeding density of 0.3-0.5 × 10 6 c/mL. A total of 4 g/L Cell Boost 6 was added to basal medium and glucose was supplied to maintain a range of 2-6 g/L during EV production. The production was performed with parameters controlled at Temp 37 • C, pH 7.0, DO 50% and agitation 70 rpm. The EV were purified following our previously reported procedure [24]. Briefly, when HEK293F cell viability dropped to 80-90% in the bioreactor, the culture broth was collected and centrifuged at 4000 g and 4 • C for 1 h. Then the collected spent medium was filtered to remove cell debris and treated with proteinase K to remove secreted host cell proteins. The EV were purified, concentrated and washed using Vivaspin 300 kDa MWCO concentrators (Fisher Scientific, Waltham, MA, USA). Finally, EV were further purified and washed using Vivaspin turbo 100 kDa MWCO concentrators (Fisher Scientific, Waltham, MA, USA). In this study, the fine polishing using Protein A chromatography was not performed. The purified EV were used for surface tagging and drug packing or supplemented with 25 mM trehalose for long-term storage at −80 • C. Second, following the reported procedure [24,36], the anti-SSTR2 mAb was produced by hybridoma cells in 2-L bioreactors and purified using New Generation Chromatography (NGC) system (Bio-Rad, Hercules, CA, USA) equipped with Protein A column. The anti-SSTR2 mAb was titrated using high performance liquid chromatography (HPLC) system (Shimadzu, Columbia, MD, USA) equipped with a UV detector and a viva C-8 column (Phenomenex, Torrance, CA, USA). Third, both anti-SSTR2 mAb and anti-CXCR4 mAb were tagged to EV via 1,2-Distearoyl-sn-Glycero-3-Phosphoethanolamine polyethylene glycol succinimidyl ester (DSPE-PEG-NHS) linker to generate mAbs-EV, with mPEG-DSPE stabilizer (New York, NY, USA) to improve the circulation stability, as we reported previously [24]. The generated mAbs-EV were titrated with NanoSight (Malvern Panalytical, Malvern, UK) with five captures in 5 min per sample using SCMOS camera at room temperature, pump speed of 25 and detection threshold of 5. The morphology of EV was confirmed with a Tecnai Transmission electron microscope (TEM, FEI, Hillsboro, OR, USA).

Packing Ver-A, DM1 and Cyanine 7 (Cy7) in mAbs-EV
Total of 10 × 10 10 EV were incubated with 0.116 mg (214 nanomole) of Ver-A and 0.25 mg of DM1 in 8.6 mL of PBS for overnight at room temperature. The unpacked free drugs were removed using Vivaspin 20 300 kDa column (Fisher Scientific, Waltham, MA, USA) and HPLC (Shimadzu, Kyoto, Japan) equipped with column of 250 × 4.6 mm 2 5 micron, 5 u C18 300A (Phenomenex) was used to titrate Ver-A and DM1 with setup of Buffer A-H 2 O + 0.1% formic acid, Buffer B-acetonitrile + 0.1% formic acid, and flow rate of 1 mL/minute to analyze the packing rate. The Cy7-PE or Liss Rhod-PE fluorescent dye (Alabaster, Alabama, AL, USA) (16.7 nmol) was used to label 1 × 10 12 EV via 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N (Fisher Scientific, Waltham, MA, USA) in the dark at room temperature with horizontal shaking overnight. The fluorescence labelled mAbs-EV or mAbs-EV-drugs can be used to monitor the in vitro uptake and in vivo biodistribution/cancer-specific targeting.

Flow Cytometry
Flow cytometry analysis was performed to quantify the surface expression of SSTR2 and CXCR4 receptors in multiple NET cell lines (BON, QGP, H727, TT and MZ) and a negative control fibroblast cell line (917) using a flow cytometer (LSRII, BD Biosciences, San Jose, CA, USA). The anti-SSTR2 mAb or anti-CXCR4 mAb (RD Systems, Minneapolis, MN, USA) was labelled with an Alexa Fluor™ 647 labelling kit to generate fluorescent AF647-mAb. The staining conditions are 1 µg AF647-mAb per million cells at 4 • C for 30 min as reported before [24,36,51,52].

Confocal Imaging
The in vitro uptake of mAbs-EV by NET cell lines was confirmed with two-or three-color confocal microscopy [24,36,51,52]. The cytoplasm and nucleus were stained with BacMam GFP Transduction Control (Green); the nucleus was stained with DAPI (blue); and the Liss Rhod fluorescent dye (red) was used to monitor the uptake of mAbs-EV. An Olympus 1X-81 confocal microscope with a laser scanning head (Olympus IX81, Center Valley, PA, USA) was used to scan the stained NET cells and collect MitoSox images.

Western Blotting
The NuPAGE™ 4-12% Bis-Tris protein gels were used to run nonreducing SDS-PAGE. The primary mouse antihuman antibody and horseradish peroxidase (HRP)-conjugated secondary anti-mouse antibody (Abcam, Cambridge, UK) were used to detect the intracellular biomarkers and followed with treatment using Luminata Forte Western HRP substrate (Millipore, Boston, MA, USA). The blotted membrane was imaged with MyECL imager (Thermo Scientific, Waltham, MA, USA) and quantified with ImageJ software (National Institutes of Health, Bethesda, MD, USA).

In Vivo Imaging System (IVIS) Imaging
The six-week-old nude (nu/nu) mice (mus) (Jackson Labs, Sacramento, CA) were subcutaneously (s.c.) injected with 3 × 10 6 BON-FLuc cells or QGP cells into the left flank. When tumor volumes reached~75-100 mm 3 , the mAbs-EV-Cy7 or mAbs-EV-Ver-A/DM1 were administered via the tail vein for targeting analysis or anticancer treatment, respectively. The in vivo NET targeting and biodistribution of Cy7 labeled mAbs-EV was evaluated using IVIS. At 24 h after i.v. injections of 30 × 10 10 mAbs-EV-Cy7, IVIS imaging was performed to capture the Fluc bioluminescence (tumor) and Cy7 fluorescence (mAbs-EV). The NET-specific targeting was confirmed by the colocalization of bioluminescence and fluorescence.

In Vitro Anticancer Cytotoxicity
The in vitro anticancer cytotoxicity assay was performed following our published protocol [53]. Briefly, the BON and QGP cells were seeded in 96-well plates with seeding density of 1 × 10 5 cells/mL, Pharmaceutics 2020, 12, 1079 5 of 13 viability of >95%, and volume of 100 µL/well. The Ver-A and DM1 dosages of 0.5, 1, 5, 10, 20, and 40 nM were tested to calculate the IC 50 values. The CellTiter-Glo Luminescent Cell Viability Assay (Promega) was used to measure the cell viability after three days' treatment.

Maximum Tolerated Dose (MTD)
To investigate the tolerated dosages and potential toxicity, five different doses of EV-Ver-A/DM1 (molar ratio of Ver-A:DM1 is 1:1), i.e., 0.31, 0.62, 1.24, 2.48 and 4.96 mg/kg, were administrated into non-tumor-bearing nude mice via intraperitoneal (i.p.) injection (n = 3). Mice were monitored twice daily for a total of eight days and showed no overt changes in general health including water intake, breathing, and locomotion. At the end of the study, mice were sacrificed to collect the major organs including brain, lung, heart, kidney and liver for Hematoxylin and Eosin (H&E) staining to analyze the potential toxicity of Ver-A/DM1.

Hematoxylin and Eosin (H&E) Staining
The tissue samples were embedded in paraffin, sectioned at 5 µm with Leica microtome and mounted on frosted microscope slides at University Pathology Facility. During H&E staining, the embedded slides were dewaxed with xylene and gradient hydrated with 100-0% ETOH, then immersed in hematoxylin solution for 10 min followed by tap water rinse for 5 min and eosin rinse for 12 dips. The slides were dipped in ddH 2 O until the eosin stopped streaking, in 50% ethanol 10 times, and in 70% ethanol 10 times. Then, the slides were equilibrated in 95% ethanol for 30 s and 100% ethanol for 1 min. Finally, the slides were dipped in xylene five times and mounted with cytoseal Xyl.

Statistical Analysis
All the experimental data were presented as mean ± standard error of the mean (SEM). Two-tailed Student's t tests were used to determine the probability of significance between groups. Comparison was performed using a one-way ANOVA followed by post hoc (Dunnett's) analysis. Statistical significance with ** p value of <0.005 was considered for all tests.

SSTR2 and CXCR4 Surface Expression and Dual Targeting
The surface expression of SSTR2 and CXCR4 in NET cell lines was evaluated by staining BON, QGP, MZ and TT cells with SSTR2 and CXCR4 antibodies at 4 • C. As presented in Figure 1, flow cytometry analysis showed that the metastatic pancreatic NET QGP cell had high SSTR2 and CXCR4 expression with binding rate of 65% and 71%, respectively. The metastatic pancreatic NET BON cells had high-level SSTR2 expression (66%) while medium-level CXCR4 expression (47%). The metastatic medullary thyroid NET MZ cells had medium-level surface SSTR2 (40%) and high-level surface CXCR4 (77%). The expression levels of both receptors were relatively low (25%) in medullary thyroid NET TT cells. The pulmonary NET H727 cells had very low SSTR2 (9%) and CXCR4 (10%) expression. The expression level of SSTR2 and CXCR4 in noncancerous foreskin fibroblast cell line 917 was 6% and 3%, respectively. SSTR2 is considered as a good target for low-grade (G1/G2) of NETs. Our previous immunohistochemistry staining of G1 and G2 NET patient tissues (n = 37) showed that 76% samples had high SSTR2 expression [36]. Leijon et al. also reported 74.8% of 151 NET tissues, including 13/14 metastasized samples, overexpressed SSTR2 [33]. We expect that the anti-SSTR2 mAb-EV-drug can target and treat ~75% SSTR2 + NET patients which are comprised of most G1/2 and >20% G3. Furthermore, CXCR4 plays an important role in highly proliferative (G3b) NETs [37,38] and literature reported that CXCR4 is positive in 50% G2 NETs and 80% G3 NETs (n = 137-164) [37][38][39][40]. All these data indicated that SSTR2/CXCR4 dual-targeted EV-drugs could cover ~95% patients with differentiated and metastatic NETs, either SSTR2 + or CXCR4 + . Moreover, dual targeting could also benefit the patients with highly heterogeneous NET. For the SSTR2 -/CXCR4 -NET patients, we can consider targeting an alternative receptor, such as the membrane gastric inhibitory polypeptide receptor (GIPR), which is overexpressed in both primary (n = 103) and metastatic (n = 123) NET tissues [54].
The bispecific antibodies have been developed to achieve dual targeting by combining the two antigens binding fragments of monoclonal antibodies into one antibody. However, the needed antibody engineering is time-consuming and has low flexibility to combine different antibodies. As compared to a bispecific antibody, our dual-targeted mAbs-EV is flexible to target different combination of multiple surface receptors in cancer cells.

NET-Specific Targeting and Uptake of Dual-Targeted mAbs-EV
The homogeneity, quality and size distribution of mAbs-EV were characterized. The transmission electron microscope (TEM) image showed that the targeted EV were physically homogenous ( Figure 2A). The nanoparticle tracking analysis using NanoSight showed that the mAbs-EV had a mean size of 145 ± 39 nm with size distribution of 102-194 nm ( Figure 2B). Moreover, our previous study confirmed the endosomal origin of the generated EV and high purity without contamination of ER, Golgi, mitochondria and nuclear [24]. The live-cell confocal laser scanning microscopy (CLSM) imaging was performed to assess the targeting delivery capability of anti-SSTR2/CXCR4 mAbs-EV. As shown in Figure 2C, the mAbs-EV-Liss Rhod (displayed as red color) targeted and internalized into the cytoplasm of BON and QGP cells. These results indicated that mAbs-EV can bind NET cells and deliver drugs intracellularly. SSTR2 is considered as a good target for low-grade (G1/G2) of NETs.
The bispecific antibodies have been developed to achieve dual targeting by combining the two antigens binding fragments of monoclonal antibodies into one antibody. However, the needed antibody engineering is time-consuming and has low flexibility to combine different antibodies. As compared to a bispecific antibody, our dual-targeted mAbs-EV is flexible to target different combination of multiple surface receptors in cancer cells.

NET-Specific Targeting and Uptake of Dual-Targeted mAbs-EV
The homogeneity, quality and size distribution of mAbs-EV were characterized. The transmission electron microscope (TEM) image showed that the targeted EV were physically homogenous (Figure 2A). The nanoparticle tracking analysis using NanoSight showed that the mAbs-EV had a mean size of 145 ± 39 nm with size distribution of 102-194 nm ( Figure 2B). Moreover, our previous study confirmed the endosomal origin of the generated EV and high purity without contamination of ER, Golgi, mitochondria and nuclear [24]. The live-cell confocal laser scanning microscopy (CLSM) imaging was performed to assess the targeting delivery capability of anti-SSTR2/CXCR4 mAbs-EV. As shown in Figure 2C, the mAbs-EV-Liss Rhod (displayed as red color) targeted and internalized into the cytoplasm of BON and QGP cells. These results indicated that mAbs-EV can bind NET cells and deliver drugs intracellularly. Furthermore, we evaluated the in vivo NET-targeting specificity of Cy7-labeled mAbs-EV, using BON-FLuc and QGP s.c. xenografted nude mouse model. The live-animal IVIS imaging at 24 h after mAbs-EV-Cy7 injection showed that the BON-FLuc xenograft bioluminescence signal or QGP xenograft overlapped with the Cy7 fluorescence signal, suggesting that the anti-SSTR2/CXCR4 mAbs-EV preferentially bound to NE tumors ( Figure 2D). It is also found that the accumulation of mAbs-EV in QGP xenograft was higher than that in BON xenograft, which was probably caused by the higher expression of SSTR2 and CXCR4 receptors in QGP. Moreover, the ex vivo IVIS analysis of tumor and main organs (heart, lung, brain, pancreas, spleen and liver) confirmed the NET xenograftspecific trageting of dual-targeted mAbs-EV and lack of binding to major organs ( Figure 2E). Altogether, these data proved that the constructed SSTR2/CXCR4 mAbs-EV can specifically target NET cells both in vitro and in vivo.

In Vitro Anticancer Cytotoxicity and Synergistic Mechanisms
The in vitro anticancer cytotoxicity of the Ver-A and DM1 was tested in a three-day assay ( Figure  3A). Multiple dosages of these drugs, including 0.5, 1, 5, 10, 20, and 40 nM, were evaluated using both BON and QGP cells. As described in Figure 3A, Ver-A showed the highest toxicity in BON cells, which killed >90% cells at >10 nM within three days. The calculated IC50 values of Ver-A were 2.8 nM for BON cells and 2.2 nM for QGP cells, and the IC50 values of DM1 were 4.5 nM for BON cells and 3.2 nM for QGP cells. These results indicated that the Ver-A and DM1 led to the growth reduction at low, single-digit nanomolar concentrations and reduced NET cells (BON and QGP) proliferation in a dose-dependent manner. In addition, the in vitro assay did not detect any cytotoxicity of the empty EV in BON and QGP cells ( Figure 3B). Furthermore, we evaluated the in vivo NET-targeting specificity of Cy7-labeled mAbs-EV, using BON-FLuc and QGP s.c. xenografted nude mouse model. The live-animal IVIS imaging at 24 h after mAbs-EV-Cy7 injection showed that the BON-FLuc xenograft bioluminescence signal or QGP xenograft overlapped with the Cy7 fluorescence signal, suggesting that the anti-SSTR2/CXCR4 mAbs-EV preferentially bound to NE tumors ( Figure 2D). It is also found that the accumulation of mAbs-EV in QGP xenograft was higher than that in BON xenograft, which was probably caused by the higher expression of SSTR2 and CXCR4 receptors in QGP. Moreover, the ex vivo IVIS analysis of tumor and main organs (heart, lung, brain, pancreas, spleen and liver) confirmed the NET xenograft-specific trageting of dual-targeted mAbs-EV and lack of binding to major organs ( Figure 2E). Altogether, these data proved that the constructed SSTR2/CXCR4 mAbs-EV can specifically target NET cells both in vitro and in vivo.

In Vitro Anticancer Cytotoxicity and Synergistic Mechanisms
The in vitro anticancer cytotoxicity of the Ver-A and DM1 was tested in a three-day assay ( Figure 3A). Multiple dosages of these drugs, including 0.5, 1, 5, 10, 20, and 40 nM, were evaluated using both BON and QGP cells. As described in Figure 3A, Ver-A showed the highest toxicity in BON cells, which killed >90% cells at >10 nM within three days. The calculated IC 50 values of Ver-A were 2.8 nM for BON cells and 2.2 nM for QGP cells, and the IC 50 values of DM1 were 4.5 nM for BON cells and 3.2 nM for QGP cells. These results indicated that the Ver-A and DM1 led to the growth reduction at low, single-digit nanomolar concentrations and reduced NET cells (BON and QGP) proliferation in a dose-dependent manner. In addition, the in vitro assay did not detect any cytotoxicity of the empty EV in BON and QGP cells ( Figure 3B).
The Western blotting analysis of proliferation and apoptosis proteins showed that 5 or 10 nM Ver-A reduced or eradicated the expression of oncogenic antiapoptotic protein cyclin D1 and cyclin dependent kinase inhibitors p21 and p27 in BON cells at 72 h post-treatment. It also decreased the expression of the proliferation signaling protein AKT (protein kinase B) ( Figure 3C). Different from BON cells, the 5 or 10 nM Ver-A did not change the expression of AKT while completely blocked the expression of cyclin D1, p21 and p27 in QGP cells. These results are consistent with literature reporting that Ver-A downregulated the expression of Akt signaling proteins in pancreatic adenocarcinoma and prostate cancer [55,56]. Unlike Ver-A, it is found that 10 or 20 nM DM1 did not change the expression of AKT, Pharmaceutics 2020, 12, 1079 8 of 13 cyclin D1, p21 and p27 in BON and QGP cells. In addition to free drugs, we also tested the EV-delivered combined 5 nM Ver-A and 5 nM DM1. Western blotting results showed that the EV-Ver-A/DM1 significantly reduced AKT expression and completely blocked the expression of cyclin D1, p21 and p27 in both BON and QGP cells. These data indicated that Ver-A and DM1 were involved in the regulation of proliferation and apoptosis of NET cells although the anticancer mechanisms need further investigation. The Western blotting analysis of proliferation and apoptosis proteins showed that 5 or 10 nM Ver-A reduced or eradicated the expression of oncogenic antiapoptotic protein cyclin D1 and cyclin dependent kinase inhibitors p21 and p27 in BON cells at 72 h post-treatment. It also decreased the expression of the proliferation signaling protein AKT (protein kinase B) ( Figure 3C). Different from BON cells, the 5 or 10 nM Ver-A did not change the expression of AKT while completely blocked the expression of cyclin D1, p21 and p27 in QGP cells. These results are consistent with literature reporting that Ver-A downregulated the expression of Akt signaling proteins in pancreatic adenocarcinoma and prostate cancer [55,56]. Unlike Ver-A, it is found that 10 or 20 nM DM1 did not change the expression of AKT, cyclin D1, p21 and p27 in BON and QGP cells. In addition to free drugs, we also tested the EV-delivered combined 5 nM Ver-A and 5 nM DM1. Western blotting results showed that the EV-Ver-A/DM1 significantly reduced AKT expression and completely blocked the expression of cyclin D1, p21 and p27 in both BON and QGP cells. These data indicated that Ver-A and DM1 were involved in the regulation of proliferation and apoptosis of NET cells although the anticancer mechanisms need further investigation.

Maximum Tolerated Dosage (MTD) and Minimal Toxicity
To investigate the MTD and potential systemic toxicity, we treated the nude mice with five doses of anti-SSTR2/CXCR4 EV-Ver-A/DM1: 0.31, 0.62, 1.24, 2.48, and 4.96 mg/kg. The bodyweight and behavior of the mice were monitored twice daily for a total of eight days. There were no overt changes in water intake, breathing and locomotion, and no obvious effects on body weight or overall survival for doses of 0.31-1.24 mg/kg ( Figure 4A). However, the dose of 100 and 200 nM Ver-A and DM1 reduced body weight by ~10% during the first three days post-treatment, then the body weight started increasing from day 4. At the end of this study, mice were sacrificed to collect the important organs, such as brain, lung, heart, kidney and liver, for further toxicity investigation. As shown in H&E staining, none of these organs in the mice treated with 200 nM Ver-A and DM1 had obvious morphology change or necrosis as compared to PBS control group ( Figure 4B). These results indicated that these five dosages of anti-SSTR2/CXCR4 mAbs-EV-Ver-A/DM1 tested in this study had no evident off-target effects in vivo. Considering the change of body weight at 2.48 and 4.96 mg/kg, we used the dose of 1.24 mg/kg in the in vivo anticancer efficacy study.

Maximum Tolerated Dosage (MTD) and Minimal Toxicity
To investigate the MTD and potential systemic toxicity, we treated the nude mice with five doses of anti-SSTR2/CXCR4 EV-Ver-A/DM1: 0.31, 0.62, 1.24, 2.48, and 4.96 mg/kg. The bodyweight and behavior of the mice were monitored twice daily for a total of eight days. There were no overt changes in water intake, breathing and locomotion, and no obvious effects on body weight or overall survival for doses of 0.31-1.24 mg/kg ( Figure 4A). However, the dose of 100 and 200 nM Ver-A and DM1 reduced body weight by~10% during the first three days post-treatment, then the body weight started increasing from day 4. At the end of this study, mice were sacrificed to collect the important organs, such as brain, lung, heart, kidney and liver, for further toxicity investigation. As shown in H&E staining, none of these organs in the mice treated with 200 nM Ver-A and DM1 had obvious morphology change or necrosis as compared to PBS control group ( Figure 4B). These results indicated that these five dosages of anti-SSTR2/CXCR4 mAbs-EV-Ver-A/DM1 tested in this study had no evident off-target effects in vivo. Considering the change of body weight at 2.48 and 4.96 mg/kg, we used the dose of 1.24 mg/kg in the in vivo anticancer efficacy study.

In Vivo Anti-NET Efficacy
The nude mice bearing BON-Luc/QGP xenografts were treated in a dosing interval of five days with PBS, SSTR2 mAb-EV-DM1, SSTR2 mAb-EV-Ver-A/DM1 and SSTR2/CXCR4 mAbs-EV-Ver-A/DM1 at dose of 1.24 mg/kg in 4 groups (n = 6). Figure 5A showed that the NET tumor growth was

In Vivo Anti-NET Efficacy
The nude mice bearing BON-Luc/QGP xenografts were treated in a dosing interval of five days with PBS, SSTR2 mAb-EV-DM1, SSTR2 mAb-EV-Ver-A/DM1 and SSTR2/CXCR4 mAbs-EV-Ver-A/DM1 at dose of 1.24 mg/kg in 4 groups (n = 6). Figure 5A showed that the NET tumor growth was significantly attenuated in the SSTR2/CXCR4 dual-targeted EV-Ver-A/DM1 treatment group compared to PBS control group (p ≤ 0.005). The profiles of bodyweight revealed that there was no difference of bodyweight among all the four groups ( Figure 5B), indicating the overall therapeutic safety and the lack of off-target toxicity during treatment. The wet weight measurement of terminal tumors confirmed the significant anti-NET efficacy of dual-targeted combined therapies ( Figure 5C,D), which reduced tumor growth rate by 71%. The SSTR2 single-targeted combined therapies reduced the tumor growth rate by 39% and the single-targeted DM1 inhibited tumor growth rate only by 27%. These data indicated that our SSTR2/CXCR4 dual-targeted mAbs-EV is an effective drug delivery vehicle for highly potent reagents that are too toxic as conventional chemotherapies (Ver-A and DM1 in this study). Moreover, dual targeting had higher anticancer efficacy than single targeting, and targeted EV enabled the delivery of combined chemotherapies to effectively treat cancers with integrated anticancer mechanisms.

In Vivo Anti-NET Efficacy
The nude mice bearing BON-Luc/QGP xenografts were treated in a dosing interval of five days with PBS, SSTR2 mAb-EV-DM1, SSTR2 mAb-EV-Ver-A/DM1 and SSTR2/CXCR4 mAbs-EV-Ver-A/DM1 at dose of 1.24 mg/kg in 4 groups (n = 6). Figure 5A showed that the NET tumor growth was significantly attenuated in the SSTR2/CXCR4 dual-targeted EV-Ver-A/DM1 treatment group compared to PBS control group (p ≤ 0.005). The profiles of bodyweight revealed that there was no difference of bodyweight among all the four groups ( Figure 5B), indicating the overall therapeutic safety and the lack of off-target toxicity during treatment. The wet weight measurement of terminal tumors confirmed the significant anti-NET efficacy of dual-targeted combined therapies ( Figure  5C,D), which reduced tumor growth rate by 71%. The SSTR2 single-targeted combined therapies reduced the tumor growth rate by 39% and the single-targeted DM1 inhibited tumor growth rate only by 27%. These data indicated that our SSTR2/CXCR4 dual-targeted mAbs-EV is an effective drug delivery vehicle for highly potent reagents that are too toxic as conventional chemotherapies (Ver-A and DM1 in this study). Moreover, dual targeting had higher anticancer efficacy than single targeting, and targeted EV enabled the delivery of combined chemotherapies to effectively treat cancers with integrated anticancer mechanisms.

Conclusions
The EVs have great potential as drug delivery vehicles to treat cancers, but there are some technical challenges to apply EV-based drug delivery in clinics, such as cancer-specific targeting, toxicity or safety, pharmacokinetics or circulation stability, and cost effective biomanufacturing. The developed dual-targeted drug delivery of highly potent anticancer chemotherapies using EV could partially overcome the issues. Moreover, the developed mAbs-EV-drugs can be easily adapted to treat different cancers or tumors by tagging cancer surface receptors-specific antibodies. Despite the achievement so far, the targeting delivery platform of EV requires further development or evaluation before clinical application. For instance, the circulation stability of the constructed mAbs-EV-drugs, such as the stability of linker bridging mAbs and EVs and the stability of packed potent drugs, needs further optimization and full investigation. In addition to inhibiting tumor growth, the loading and delivering capability of combined chemotherapies, gene therapies and biotherapies needs further evaluation, aiming to eliminate tumor cells in vivo. Moreover, a pharmacokinetics study is highly desired to assess the circulation stability of the dual-targeted system-delivered therapies.