Chlorin e6-loaded goat milk-derived extracellular vesicles for Cerenkov luminescence-induced photodynamic therapy

Photodynamic therapy (PDT) is a promising cancer treatment strategy with rapid progress in preclinical and clinical settings. However, the limitations in penetration of external light and precise delivery of photosensitizers hamper its clinical translation. As such, the internal light source such as Cerenkov luminescence (CL) from decaying radioisotopes offers new opportunities. Herein, we show that goat milk-derived extracellular vesicles (GEV) can act as a carrier to deliver photosensitizer Chlorin e6 (Ce6) and tumor-avid 18F-FDG can activate CL-induced PDT for precision cancer theranostics. GEV was isolated via differential ultracentrifugation of commercial goat milk and photosensitizer Ce6 was loaded by co-incubation to obtain Ce6@GEV. Tumor uptake of Ce6@GEV was examined using confocal microscopy and flow cytometry. To demonstrate the ability of 18F-FDG to activate photodynamic effects against cancer cells, apoptosis rates were measured using flow cytometry, and the production of 1O2 was measured by reactive oxygen species (ROS) monitoring kit. Moreover, we used the IVIS device to detect Cherenkov radiation and Cerenkov radiation energy transfer (CRET). For animal experiments, a small-animal IVIS imaging system was used to visualize the accumulation of the GEV drug delivery system in tumors. PET/CT and CL images of the tumor site were performed at 0.5, 1, and 2 h. For in vivo antitumor therapy, changes of tumor volume, survival time, and body weight in six groups of tumor-bearing mice were monitored. Furthermore, the blood sample and organs of interest (heart, liver, spleen, lungs, kidneys, and tumor) were collected for hematological analysis, immunohistochemistry, and H&E staining. Confocal microscopy of 4T1 cells incubated with Ce6@GEV for 4 h revealed strong red fluorescence signals in the cytoplasm, which demonstrated that Ce6 loaded in GEV could be efficiently delivered into tumor cells. When Ce6@GEV and 18F-FDG co-existed incubated with 4T1 cells, the cell viability plummeted from more than 88.02 ± 1.30% to 23.79 ± 1.59%, indicating excellent CL-induced PDT effects. In vivo fluorescence images showed a peak tumor/liver ratio of 1.36 ± 0.09 at 24 h after Ce6@GEV injection. For in vivo antitumor therapy, Ce6@GEV + 18F-FDG group had the best tumor inhibition rate (58.02%) compared with the other groups, with the longest survival rate (35 days, 40%). During the whole treatment process, neither blood biochemical analysis nor histological observation revealed vital organ damage, suggesting the biosafety of this treatment strategy. The simultaneous accumulation of 18F-FDG and Ce6 in tumor tissues is expected to overcome the deficiency of traditional PDT. This strategy has the potential to extend PDT to a variety of tumors, including metastases, using targeted radiotracers to provide internal excitation of light-responsive therapeutics. We expect that our method will play a critical role in precision treatment of deep solid tumors.

the three critical components of PDT [2]. Under the excitation of light at an appropriate wavelength, the excited PS photochemically reacts with cellular substrates or molecular oxygen to produce ROS, eventually leading to cancer cell death and inhibition of tumor growth [3]. Singlet oxygen ( 1 O 2 ) is the most damaging ROS because it reacts with primary components of cells [4]. Although great strides have been made over the past few decades, oncological PDT remains in its infancy. Key challenges such as limited external light penetration and precise delivery of photosensitizers have restricted the clinical use of PDT to specific dermatological indications and superficial tumors [5].
Internal excitation light sources based on chemiluminescence, bioluminescence, and Cerenkov luminescence (CL) promise to expand the number of indications amenable to PDT [6,7]. CL is produced by decaying radioisotopes, such as 198 Au, 124 I, 90 Y, 89 Zr, 68 Ga, 64 Cu, and 18 F, which emit β particles (positrons or electrons) at speeds faster than the phase velocity of light in the dielectric medium [8,9]. CL occurs in the 250-600 nm range, matching well the absorption spectra of most PS [10,11]. In addition, CL has the potential to generate light more homogeneously throughout tumor tissues and helps negate the effects of poor PS specificity [12,13]. Despite these advantages, CL remains a low-intensity light source because of the low CL photon flux from the radioisotopes, which leaves clinicians uncertain about the benefits of CL-induced PDT [9]. Thus, it is important to develop a CL-induced PDT strategy that ensures efficient energy transfer and tumor accumulation but minimizes potential side effects. The common diagnostic radiotracer 18 F-fluorodeoxyglucose ( 18 F-FDG) is an ideal light source for CL-induced PDT because of its pure positron decay ( 18 F half-life, 109.8 min; β + decay, 97%, 0.635 MeV) [14]. Additionally, the increased uptake and retention of 18 F-FDG by various tumors makes it an ideal PDT light source. 18 F-FDG enables efficient generation of CL at the tumor site without the need for additional techniques to direct the radionuclide to the tumor.
The choice of PS crucially affects the efficacy of PDT. In the photochemical reaction, the PS is activated by light and transfers the absorbed energy to the reactant. However, owing to their hydrophobicity, weak bioavailability, nonspecific phototoxicity, and tendency to self-aggregate, commonly used PS are far from ideal [15]. Therefore, we need an efficient drug delivery system (DDS) to optimize the biodistribution of PS molecules and guide them to tumor sites. Extracellular vesicles (EVs), including apoptotic bodies (500-5000 nm), microvesicles (50-10000 nm), and exosomes (30-150 nm), are promising biocompatible nanocarriers for drug delivery and tumor targeting [16,17]. An advantage of using EVs as therapeutics is the ability of their lipid membrane to fuse with target cells where they can deliver therapeutic cargos while exhibiting low levels of immunogenicity [18].
Almost all types of cells secrete EVs, which are naturally present in body fluids, including blood, saliva, urine, cerebrospinal fluid, and breast milk. Among these, milkderived EVs are known for their easy production and high yields compared with those derived from cell culture media or blood plasma. Milk-derived EVs have displayed minimum toxicity and non-immunogenicity in healthy animals [19]. Given their accessibility and reproducibility, milk-derived EVs are a focus of current research [20,21]. Goat milkderived EVs (GEV) are viable natural nanocarriers previously used to deliver siRNA for cancer theranostics [20].
The widely studied amphiphilic chlorin e6 (Ce6) is promising PS for PDT because it has a high 1 O 2 quantum yield when irradiated [22][23][24]. Ce6 has absorption peaks at approximately 400 and 660 nm. On the one hand, the broad 400-nm absorption line matches the CL spectrum of 18 F, and it has a high 1 O 2 generation rate in response to CL excitation; this spectral coupling of Ce6 and CL is termed Cerenkov radiation energy transfer (CRET). CRET refers to transferring CL energy from a radionuclide to another fluorescent receptor. CL can induce the receptor molecules to emit fluorescence, and the fluorescence intensity of CL itself decays [25]. On the other hand, the 660 nm absorption of Ce6 provides a more tissue-penetrating wavelength for real-time fluorescence imaging in vivo compared with traditional PS such as porphyrin. Therefore, Ce6 is suitable for CL-induced PDT and visualization of the DDS [26].
In this study, GEV was used as a DDS to increase and extend PS retention at tumor sites and minimize off-target accumulation of PS, which are toxic in their free form. Specifically, we designed a Ce6-loaded GEV nanoplatform to achieve efficient Ce6 tumor delivery and then used it for realtime fluorescence imaging and CL-induced PDT. When triggered by 18 F-FDG located at tumor sites, CL-induced PDT killed tumor cells by activating Ce6-mediated ROS generation. Our method offers a novel way of integrating foodderived natural biomaterials and nuclear imaging probes in daily clinical practice to overcome current limitations of PDT (Scheme 1). We believe that this interesting combination may pave the way for more efficient cancer treatment regimens.

Materials and methods
Cell culture 4T1 murine breast cancer cells were preserved in our laboratory (Hubei Province Key Laboratory of Molecular Imaging) and propagated in Roswell Park Memorial Institute (RPMI)-1640 medium (Gibco, Gaithersburg MD, USA) supplemented with 10% v/v fetal bovine serum (Gibco, USA) and 1% v/v penicillin-streptomycin solution (Solarbio, Beijing, China). Cells were cultured at 37°C in a humidified atmosphere containing 5% CO 2 .

Isolation of GEV
GEV were isolated using sequential ultracentrifugation of fresh commercial goat milk. First, goat milk was ultracentrifuged at 8000 × g for 30 min at 4°C to remove lipid and cell debris. Residual lipid and dead cells in the skim milk supernatant were removed using ultracentrifugation at 13,000 × g for 60 min. Next, the casein was removed from the supernatant by adding 0.025 g/L chymosin and incubating at 37°C for 4-6 h. The resulting whey was filtered through a 0.45-μm membrane to remove excess cell debris. Finally, the preliminary GEV products were centrifuged at 120,000 × g for 90 min, and the resulting GEV in the pellet were resuspended in phosphate buffered saline (PBS; Gibco, USA), passed through a 0.22-μm membrane, and quantified according to their surface protein content using a BCA Protein Assay Kit (Beyotime, Shanghai, China).

Synthesis of Ce6@GEV
When used as a carrier of the hydrophobic drug curcumin, EVs improved its solubility, stability, and bioavailability [27]. In a similar way, the Ce6 photosensitizer was loaded into GEV by co-incubation to obtain Ce6@GEV. In brief, Ce6 (5 mg/mL in DMSO) was mixed with GEV (2.12 mg/ mL final concentration) in a 1:1 volume ratio. Then, the mixture was stirred at 2500 rpm overnight in the dark at 37°C. Next, free Ce6 molecules were removed using ultrafiltration (50 mL, 100 kDa, Millipore, USA). The Ce6@GEV, obtained as a green pellet, were dispersed in PBS.

Characterization of GEV and Ce6@GEV
The morphologies of GEV and Ce6@GEV were characterized using transmission electron microscopy (JEM-1400Plus, Japan). Hydrodynamic diameters and zeta potentials of GEV and Ce6@GEV were measured using dynamic light scattering (Brookhaven Instruments, USA) at room Scheme 1 The schematic of chlorin e6-loaded goat milk-derived extracellular vesicles for multimodal imaging and Cerenkov luminescenceinduced photodynamic therapy temperature and continuously monitored for 7 days to assess their stability in vitro. Representative cell surface markers (Tsg101, CD63, and CD9) of GEV and Ce6@GEV were analyzed using Western blotting. To measure Ce6 encapsulation efficiency, a standard curve was constructed from the UV-vis-NIR spectra collected using a microplate reader (Bio-Rad, USA). Cell Counting Kit-8 (CCK-8) (HY-K0301, MCE, USA) was used to assess the cytotoxicity of different GEV and Ce6@GEV concentrations to 4T1 cells.

Intracellular ROS detection
Briefly, confocal dishes were each seeded with 2 × 10 5 4T1 cells and incubated for 36 h. Next, 1000 μL serumfree medium was added to each dish, followed by Ce6@ GEV (20 μL) or Ce6 (6 μL) to yield a final concentration of 1 μg/mL Ce6. After 6 h incubation, the treated medium was aspirated and the cells were washed twice with PBS (pH 7.4), then 2,7-dichlorodihydrofluorescein diacetate (DCFH-DA; Yeasen, Shanghai, China) was added to each dish and the cells were incubated for 30 min at 37°C. To stimulate intracellular accumulation of Ce6, 18 F-FDG (3.7 MBq) was added to each dish. After two half-lives, the cell nuclei were stained with DAPI. Finally, antifade mounting medium (Beyotime, Shanghai, China) was added to prevent fluorescence quenching, and a fluorescence confocal microscope was used to detect the fluorescence of dichlorofluorescein (DCF, Ex/Em = 488/525 nm), Ce6 (Ex/Em = 405/680 nm), and DAPI (Ex/Em = 364/454 nm). The above experiments were carried out under dark conditions.

Cell apoptosis assay
The effects of each treatment were examined using an apoptosis kit (Calcein-AM/PI; Meilunbio, Dalian, China). 4T1 cells were inoculated into 6-well plates (2 × l0 4 cells/well), incubated overnight, then washed twice with PBS. The wells were then supplemented with untreated culture medium or medium containing 2 μg/mL Ce6@GEV or Ce6. After 6 h incubation, 2.96 MBq 18 F-FDG was added to each well, and after two half-lives, the cells were digested and stained with 200 μL PBS containing 2 μM calcein-AM and 8 μM propidium iodide for flow cytometry analysis (FACSCalibur, BD, USA). The apoptosis rate was calculated as the percentage of cells undergoing early or late apoptosis.

Imaging Cerenkov radiation and CRET In Vitro
To detect CRET, we used the small-animal In Vivo Imaging System (IVIS; PerkinElmer Inc., USA) to capture 18 F emission in response to different treatments (PBS, GEV, Ce6, 18 F-FDG, 18 F-FDG + Ce6, or 18 F-FDG + Ce6@GEV). To image the radioactivity-dependence of 18 F-FDG emissions, CL was measured in response to 0, 0.037, 0.37, 1.85, 3.7, and 7.4 MBq 18 F-FDG, using open, 520-, 620-, 670-, and 710-nm filters. To image the concentration-dependence of Ce6@GEV emissions, 1 mg/mL Ce6@GEV stock solution was diluted in 100 μL PBS in a black 96-well plate to achieve final Ce6 concentrations of 0, 10, 25, 50, 100, and 200 μg/mL. 18 F-FDG was added (1.48 MBq/well) at the time of CL imaging, yielding a final volume of 150 μL. Images were captured using open, 520-, 570-, 620-, 670-, and 710-nm emission filters, a binning parameter of 8, and a 60-s exposure. CRET ratios were calculated by manually delineating the average radiance (photons/s/cm 2 /sr) of the regions of interest as follows [28]: where Cerenkov + Fluorophore x and Cerenkov + Fluorophore y are the radiances detected within spectral windows x and y centered on the fluorophore and Cerenkov radiation emission wavelengths, respectively, and Cerenkov x and Cerenkov y are the corresponding radiances in the absence of a fluorophore. In the case of 18 F-FDG and Ce6@GEV, Eq. 1 becomes: where ave denotes the average value.

Tumor-bearing mouse models
All animal experiment protocols were reviewed and approved by the Animal Care and Use Committee of Tongji Medical College, Huazhong University of Science and Technology. 4T1 cells (5 × 10 6 ) suspended in 100 μL PBS were subcutaneously injected into the right front limb of BALB/c mice (female, 4-5 weeks old, Beijing Vital River Laboratory Animal Technology Co., Ltd, China) fed in a pathogen-free environment. Imaging studies were performed when the tumors reached approximately 5 mm in diameter, and treatment experiments started when tumor volumes reached 60-80 mm 3 .

Fluorescent imaging in vivo
When the tumor diameters reached 5 mm, the tumorbearing mice were injected with 25 mg/kg Ce6@GEV or Ce6 in a volume of 100 μL via the tail vein. In vivo fluorescence imaging was conducted 1, 2, 4, 6, 8, 16, 24, and 48 h after injection, using the IVIS equipped with a fluorescence filter set (Ex/Em = 680/790 nm). Tumors and organs of interest (muscle, large intestine, small intestine, kidneys, spleen, liver, and stomach) were resected and imaged 24 h later. As a control, Lipo-DiD (26.3 mg/kg) was injected into tumor-bearing mice. In vivo fluorescence imaging was conducted using the IVIS equipped with a fluorescence filter set (Ex/Em = 620/670 nm).

Positron emission tomography/computed tomography and CL imaging in vivo
Mice were anesthetized using 2% isoflurane 0.5, 1, and 2 h after injection of 18 F-FDG, and images were acquired using a micro-PET/CT scanner (Novel Medical, InliView-3000B, China). For CL imaging, 6 h after Ce6@GEV administration, 18 F-FDG (5.55 MBq) was injected and CL imaging (binning parameter, 16; exposure, 300 s) was performed 0.5, 1, and 2 h later using the IVIS without a fluorescence filter set. The control group was not injected with Ce6@GEV before 18 F-FDG injection (n = 3 per group).

In vivo toxicity evaluation
Experiment groups and treatment methods were as described above (n = 5 per group). The general state of the mice was monitored daily, and they were weighed every other day. On days 7 and 20 of treatment, blood samples were collected for hematological analysis. A blood biochemical analyzer (AU-480, Beckman, USA) was used to evaluate markers of liver and kidney function, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), blood urea nitrogen (BUN), and creatinine (CRE). On day 20 of treatment, organs of interest (heart, liver, spleen, lungs, and kidneys) were collected for H&E staining and then examined using an optical microscope (IX73, Olympus, Japan).

Statistical analysis
All data are reported as mean ± standard deviation (SD).
Comparisons between different groups were evaluated with the unpaired Student's t-test. Values of P < 0.05 are considered statistically significant. Statistical significance was performed with GraphPad Prism 8.0 software.

Synthesis and characterization of GEV and Ce6@GEV
GEV was successfully isolated from fresh commercial goat milk using sequential ultracentrifugation. The surface protein content of the extracted GEV was 2.12 mg/mL, as calculated using the bicinchoninic acid method (Fig. S1A). Ce6 was loaded into GEV by co-incubation (Fig. 1A). After several rounds of ultrafiltration, the filtrate gradually became colorless, indicating that free Ce6 molecules had been removed. The Ce6@GEV was obtained as a green pellet and dispersed in PBS (Fig. 1B). Figure 1C shows representative transmission electron microscopy images of GEV and Ce6@GEV. As expected, typical protein markers, including Tsg101, CD63, and CD9, were identified in GEV and Ce6@GEV (Fig. 1D). The Ce6 encapsulation efficiency of GEV was 11.06 ± 1.07% (Fig. S1B−C). The GEV and Ce6@GEV were well dispersed in PBS, with hydrodynamic diameters of 173.88 ± 5.33 and 182.19 ± 7.5 nm (Fig. 1E), respectively, and zeta potentials of −7.66 ± 1.3 and −9.64 ± 1.94 mV, respectively. The above differences in hydrodynamic diameter and zeta potential indicated that Ce6 was successfully loaded into GEV. UV-vis-NIR spectra indicated that GEV encapsulation did not affect the optical properties of Ce6 (Fig. 1F). The mean hydrodynamic diameters and zeta potentials of GEV and Ce6@GEV did not change significantly over 7 days, which indicated their excellent stability (Fig. 1G−H). The polydispersity indexes of GEV and Ce6@GEV at 4°C on day 7 were 0.14 ± 0.02 and 0.11 ± 0.01, respectively (Fig. 1I). GEV and Ce6@GEV also had good stability in serum over 7 days (Fig. S2A−C). The survival rate of 4T1 tumor cells treated with high GEV concentrations (50 μg/mL) was 87.99 ± 2.9%, indicating that GEV had low toxicity. Cell viability was reduced to 77.94 ± 6.96% at the highest Ce6@GEV concentration (50 μg/mL) because the high concentration of Ce6 had a cytotoxic effect after entering the 4T1 cells ( Fig. S3A−B).

In vitro cell uptake and CL-induced PDT
As shown in Fig. 2A, after 4 h incubation with Ce6@GEV or free Ce6, confocal microscopy showed that both Ce6@ GEV and Ce6 were taken up by 4T1 tumor cells. It can be found that the Ce6 was in the nucleus of the cells. Quantification of fluorescence intensity revealed that the cytoplasmic concentration of Ce6@GEV (29.14 ± 2.54 RFU) was significantly greater than that of free Ce6 (22.53 ± 3.78 RFU) ( Fig. 2B, P < 0.05). Flow cytometry analysis also indicated that Ce6 in Ce6@GEV could be efficiently delivered into tumor cells (Fig. S4). As illustrated in Fig. 2C, D, high concentrations of free Ce6 molecules were clearly cytotoxic. No cytotoxicity was observed when 0-3.7 MBq of 18 F-FDG, 0-50 μg/mL GEV, or 0-4 μg/mL Ce6@GEV were incubated with 4T1 cells. However, as shown in Fig. 2E, cell viability after coincubation of Ce6@GEV and 1.48 MBq 18 F-FDG (23.79 ± 1.59%) was markedly lower than that resulting from incubation in medium only (108.09 ± 1.85%), indicating an excellent PDT effect. Moreover, cell viability after treatment with 50 μg/mL Ce6@GEV + 18 F-FDG (34.04 ± 3.11%) was lower than that after treatment with 10 μg/ mL Ce6@GEV + 18 F-FDG (59.79 ± 3.98%, n = 7), which demonstrated the Ce6 concentration-dependent antitumor efficacy of Ce6@GEV (Fig. 2F). The IC 50 (half maximal inhibitory concentration) value of 1.48 MBq 18 F-FDG to activate Ce6@GEV was 0.177 μg/mL (Fig. S5). Next, we assessed the ROS production of Ce6@GEV coincubated with 18 F-FDG using confocal imaging of green fluorescent intracellular DCF (Fig. 3A). Semi-quantitative analysis showed that significantly greater ROS were generated by treatment with Ce6@GEV + 18 F-FDG (24.38 ± 4.07 RFU) than with Ce6@GEV (15.37 ± 2.24 RFU) (Fig. 3B, P < 0.01). The above results indicated that when triggered by 18 F-FDG, CL-induced PDT can generate a large number of ROS in 4T1 cells. Flow cytometry analysis showed that although there was no significant difference between the rates of apoptosis after Ce6 + 18 F-FDG and Ce6@GEV + 18 F-FDG treatments, the former produced more late apoptotic cells (Q2, 20.03 ± 4.28%). This may be explained by the cytotoxic effects of free Ce6 within the first 6 h of drug incubation (i.e., before 18 F-FDG was added). The apoptosis rates in response to Ce6 + 18 F-FDG and Ce6@GEV + 18 F-FDG treatments were 39.73 ± 2.50% and 39.27 ± 3.07%, respectively, which further confirmed the effectiveness of CL-induced PDT in 4T1 cells (Fig. 3C, D).

Imaging Cerenkov radiation and CRET in vitro
To detect Cerenkov radiation and CRET, we used the IVIS to capture 18 F emissions in the absence of excitation light, with and without emission filters. As shown in Fig. 4A, the PBS solution containing GEV and Ce6 was nonfluorescent, whereas intense fluorescence was emitted from solutions containing 18 F-FDG, 18 F-FDG + Ce6, and 18 F-FDG + Ce6@GEV (1.85 MBq 18 F-FDG). Quantitative analysis of the average radiance (Fig. 4B) showed that the 18 F-FDG + Ce6@GEV solution had the highest fluorescence intensity (249,250 ± 17,964 photons/s/cm 2 /sr; p < 0.05), which was consistent with the visual inspection of the images in Fig. 4A. Moreover, the intensity of CL emitted by 18 F-FDG increased in a radioactivity-dependent manner (Fig. S6A−C). CL emissions in the absence of excitation light with 520, 620, 670, and 710 nm filters are shown in Fig. 4C. Using the 520-nm filter, the 18 F-FDG solution had the highest fluorescence intensity, which was consistent with 250-600 nm CL resulting from 18 F decay [29,30]. Using the 670-nm filter, solutions of 18 F-FDG + Ce6 and 18 F-FDG + Ce6@GEV had the highest fluorescence intensities, which resulted from the red fluorescence generated by 18 F-FDG-excited Ce6 (Fig. 4D). Using black 96-well plates, we then imaged Ce6@GEV at different concentrations (0, 10, 25, 50, 100, and 200 μg/mL) in PBS with 1.48 MBq/well of 18 F-FDG. Images were captured using open, 520-, 570-, 620-, 670-, and 710-nm emission filters to distinguish the bulk CL (<520 nm) from excited Ce6 fluorescence induced by CRET (>620 nm) (Fig. 4E). The relative radiance in the red-filtered images (>620 nm) increased commensurately with the concentration of Ce6@ GEV, while that of the blue-filtered images (<520 nm) decreased, consistent with absorbance of Cerenkov radiation and emission by Ce6@GEV (Fig. 4F−G). Using Eq. 2, the CRET ratio was calculated to be 1.93 ± 0.01 at 100 μg/mL Ce6@GEV (Fig. 4H).

Multimodal imaging of 4T1 tumor-bearing mice
Longitudinal fluorescence imaging was performed 1, 2, 4, 6, 8, 16, 24, and 48 h after administration of Ce6@GEV or Ce6. Fluorescence imaging showed that Ce6@GEV could effectively accumulate and remain in tumor tissue for more than 24 h, but accumulation of free Ce6 in tumor tissue was negligible (Fig. 5A). A stronger fluorescence signal within the tumors and fewer background signals were measured 6 h after injection of Ce6@GEV (Fig. 5B). The tumor-to-liver fluorescence ratios 6, 8, 16, and 24 h after injection of Ce6@ GEV were 0.40 ± 0.08, 0.64 ± 0.06, 1.21 ± 0.01, and 1.36 ± 0.09, respectively, indicating that the fluorescence intensity of tumor tissue was higher than that of liver tissue 16 and 24 h after injection (Fig. 5C). Ex vivo images obtained 24 h after administration of Ce6@GEV were consistent with in vivo images (Fig. 5D). From the fluorescence imaging, the second highest fluorescence intensity was found in the liver, and almost no fluorescence was observed in the gastrointestinal tract and kidney. This suggested that the nanodrug may be mainly metabolized and degraded by the liver, which may attribute to the properties of nanoparticles. In contrast, most of Lipo-DiD accumulate in the liver and spleen at 48 h after intravenous injection, and only a small amount of Lipo-DiD target the tumor (Fig. S7).
Having confirmed excellent Ce6 localization in 4T1 tumors, we further evaluated the tumor-targeting ability of 18 F-FDG and its CRET efficiency at tumor sites. As shown in the PET/CT and CL images (Fig. S8−9), tumor-avid 18 F-FDG was able to concentrate at 4T1 tumor sites and generate sufficient CL for in vivo Cerenkov-induced PDT. Hence, we showed that GEV could deliver the Ce6 photosensitizer to tumor tissues and that 18 F-FDG might be suitable for precision cancer theranostics using CL-induced PDT.

In vivo antitumor effect of CL-induced PDT
After demonstrating that 18 F-FDG and Ce6@GEV could effectively locate at tumor sites, we further explored their combined antitumor effect using CL-induced PDT in 4T1 tumor-bearing mice. A schematic diagram of the experimental process is shown in Fig. 6A. The 35-day survival rate of mice treated with Ce6@GEV + 18 F-FDG was 40% at the end of treatment. However, mice in all other groups reached their endpoints (tumor volume > 1500 mm 3 or mortality, Fig. 6B). The tumor inhibitory rate 20 days after treatment with Ce6@GEV + 18 F-FDG (58.02%) was significantly higher than the corresponding rates after treatment with Ce6 + 18 F-FDG (11.01%), 18 F-FDG (7.29%), Ce6@ GEV (14.68%), and Ce6 (3.4%; n = 5, p < 0.001), which showed that CL-induced PDT had excellent antitumor efficacy (Fig. 6C−D). As expected, H&E staining and Ki67 staining of tumors 7 and 20 days after different treatments further confirmed that Ce6@GEV + 18 F-FDG resulted in optimal tumor suppression (Fig. 6E).  In vivo toxicity evaluation BALB/c mice (n = 5) received intravenous injections of Ce6@GEV + 18 F-FDG, Ce6 + 18 F-FDG, 18 F-FDG, Ce6@ GEV, Ce6, or PBS at therapeutic doses to evaluate their potential toxicity. The body weight of mice treated with Ce6@GEV + 18 F-FDG decreased by approximately 9% in the first 2 days after treatment, which may have been a side effect of tumor cell death caused by ROS produced in the tumor area. There was no obvious weight loss 20 days after any treatment (Fig. 7A). The blood biochemistry results showed no significant hepatic or renal toxicity, with liver and kidney function markers being normal 7 days (Fig. S10A−C) and 20 days (Fig. 7B−D) post-treatment. Also, H&E staining provided no significant evidence of major damage to the heart, liver, spleen, lungs, or kidneys (Fig. 7E). Taken together, our results demonstrate that combining CL of 18 F-FDG with Ce6@GEV not only inhibited tumor growth but prolonged survival time. Moreover, this treatment method did not elicit obvious systemic toxicity.

Discussion
Our study validated a Ce6-loaded GEV nanosystem, namely Ce6@GEV, for 18 F-FDG CL-triggered cancer PDT. Our results demonstrated the possibility of harnessing CL as an internal excitation light source to achieve depth-independent CL-mediated therapy. This strategy could extend PDT to a variety of tumors, including metastases, using targeted radiotracers to provide internal excitation of light-responsive therapeutics.
The choice of DDS critically influences the final PS concentration in target cells or tissues, and PDT effects depend on the uptake of PS molecules by tumor cells [31]. A challenge for artificial DDSs is their rapid clearance by the reticuloendothelial system and fast blood clearance [32]. Liposomes are the most widely used delivery vehicles, but have shortcomings associated with allergic reactions, facile oxidative degradation, and suboptimal reproducibility [33,34]. EVs are an emerging platform for diagnosis and therapy, owing to their natural origin and liposome-like structure [35]. The prospect of developing EVs as a DDS has piqued great interest because of their excellent biocompatibility and tumor-targeting [36]. Among the possible sources of EVs, milk enables their facile production in high yields compared with cell culture fluids or blood plasma [37,38].
Although widely used as a PS, Ce6 is far from ideal for PDT because of its hydrophobicity, weak bioavailability, and non-specific phototoxicity [15]. To circumvent the above shortcomings, Ce6 was encapsulated in GEV, an advanced DDS capable of enhancing drug penetration of biological membranes and carrying both hydrophobic and lipophilic agents. Moreover, the unique fusogenic properties and enhanced permeability and retention of GEV improve the efficiency of drug delivery to tumors [39]. In this study, GEV were successfully isolated from commercial goat milk in high yields (2.12 mg/mL) and with hydrodynamic diameters (173.88 ± 5.33 nm) (Fig. 1E) that enabled them to passively target solid tumors via the enhanced permeability and retention effect. The average hydrodynamic diameter and zeta potential of the GEV did not change significantly over 7 d (Fig. 1G−H). GEV added to 4T1 cells showed no significant toxicity and good tumor uptake, making them an ideal DDS. Moreover, Ce6@GEV showed a peak tumorto-liver ratio of 1.36 ± 0.09 24 h after intravenous injection (Fig. 5A), which was consistent with our previous study of tumor cell-derived exosome metabolism in vivo [40,41]. This result not only highlighted the good tumor targeting of GEV but also indicated that the co-incubation method of loading Ce6 into GEV does not disrupt the normal structure and thus in vivo biodistribution of GEV.
The rapid development of nanotechnology has seen biomedical applications of CL in imaging and therapy attract significant attention [10]. There are several issues to consider. The CL intensity mainly depends on the choice of radionuclide. 82 Rb is the brightest Cerenkov emitter, producing approximately 80 photons/decay. Radionuclides such as 90 Y or 68 Ga, yielding >30 photons/decay, are preferable to 18 F, 89 Zr, and 64 Cu if high-intensity CL is desired [42]. Considering the availability and practical application of these radionuclides in nuclear medicine and molecular imaging, the popular clinical radiotracer 18 F-FDG was used in our study. In addition to functioning as a PET tracer, 18 F-FDG as a PDT light source is favored by its higher uptake and retention by various tumors compared with healthy tissue. After demonstrating that Ce6@GEV could effectively target 4T1 tumor cells, we verified the therapeutic effect of CL-induced PDT in vitro and in vivo. The luminescent signal from the high-energy positrons emitted from 18 F-FDG was detected using the IVIS without an external excitation source in vitro (Fig. S4A) and in vivo (Fig. S5A). As proof of principle, we quantified the CRET ratio by imaging the energy transfer of 18 F-FDG-generated CL to Ce6@GEV in vitro (Fig. 4E). The highest CRET ratio was 1.93 ± 0.01 at 100 μg/mL Ce6@ GEV and 1.48 MBq 18 F-FDG (Fig. 4H).
Nanoparticles provide a useful platform for therapeutic agents such as radioisotopes, chemicals, and PS. Radioisotopes can be encapsulated within or linked to nanoparticles, or the two can be delivered separately. Radioisotopes are generally bound to nanoparticles owing to their high degree of co-localization and short interaction distance. For instance, Ni et al. synthesized 89 Zr-radiolabeled, porphyrin-decorated magnetic nanoparticles for CL-induced PDT. While a radionuclide and PS co-loaded in nanoparticles could achieve satisfactory therapeutic effects, liver damage was evident 7 days post-injection because most of the nanoparticles accumulated there [12]. Indeed, 30-99% of exogenous nanoparticles are known to accumulate and be metabolized in the liver [43]. Our therapeutic strategy to initiate PDT using CL from 18 F-FDG could reduce side effects on normal tissues while suppressing tumor growth. We pre-injected Ce6@GEV to passively target the tumor as a CL energy receiver and then injected 18 F-FDG after a Fig. 7 In vivo side effect assessment. A Body weight change curves at 20 days after treatment. B-D Blood biochemistry data including liver function makers (ALT, AST, and ALP) and kidney function markers (CRE and BUN) in different treatment groups at 20 days after treatment. There were no statistically significant differences among groups regarding different indicators. E Representative histochemistry analysis of major organs (heart, liver, spleen, lungs, and kidneys) stained with hematoxylin-eosin. Scale bar = 200 μm high tumor-to-liver concentration ratio of Ce6@GEV was attained. Efficient CRET was achieved by maximizing the co-localization of Ce6@GEV and 18 F-FDG at the tumor site, thereby optimizing tumor-targeted PDT efficacy and minimizing side effects on the liver and other healthy tissues. Cell viability and ROS assays suggested that 18 F-FDG CL can stimulate Ce6@GEV to generate sufficient ROS to kill 4T1 tumor cells. H&E and Ki67 staining showed that the highest apoptosis and lowest cell proliferation rates were achieved in vivo using Ce6@GEV + 18 F-FDG treatment (Fig. 6E). Based on the above results, our work has created a path for clinical translation of depth-independent PDT.
More refinements are needed to develop this CL-induced PDT system. In terms of a DDS, there is still no standardized method for EV isolation. Ultracentrifugation is a more common separation strategy, although nanovesicles and protein aggregates may be partially damaged by high centrifugal forces, which may complicate the application of EVs [44]. As a PDT light source, 68 Ga has higher CL intensity than 18 F, 89 Zr, and 64 Cu and better biosafety than 90 Y. Compared with other radionuclides, 68 Ga therefore has more potential for clinical development, but an efficient method is required to deliver it to different tumors. The use of external excitation light for superficial or endoscope-accessible lesions is well established, but relative efficacy of radionuclides (lower photon flux) and external beams (higher photon flux) for PDT of deep tumors requires further exploration.

Conclusions
This work devoted to deliver Ce6 to tumor sites using GEV, then induces photodynamic effects using 18 F-FDG generated CL. We verified that Ce6@GEV could effectively target 4T1 tumors, and tumor-avid 18 F-FDG was able to accumulate at tumor sites to generate sufficient CL. When triggered by 18 F-FDG, CL-induced PDT could significantly inhibit tumor growth and prolong the survival time of 4T1 tumor-bearing mice. Our study provides a new approach to CL-induced PDT that can overcome the deficiency of traditional PDT methods that rely on an external light source. We expect our method will play a critical role in precision treatment of deep solid tumors.