Active Targeting Nanotheranostic System for Dual-Modality Imaging-Guided Chemo-/Photodynamic Therapy of Pancreatic Cancer

Background: Pancreatic cancer (PC) is one of the most devastating types of cancers worldwide and has a remarkably poor survival rate, emphasizing the need for more effective strategies for the diagnosis and therapy of PC. Upconversion nanoparticles (UCNPs) have gained a privileged place in the biomedical eld due to their outstanding properties. Besides, epithelial cell adhesion molecule (EpCAM) as one of the key biomarkers of pancreatic cancer stem cells, is a vital target for theranostic, diagnostic, and/or therapeutic intervention in nanomedicine. In this study, the theranostic nanosystem (EpCAM-UCMSNs-MX) was formed from the mesoporous silica-coated UCNPs functionalized with anti-EpCAM monoclonal antibody, and then one anticancer drug and photosensitizer, mitoxantrone (MX), was loaded into the mesoporous silica. The nanotheranostic system was used to target caner stem cells for realizing simultaneous dual-modality MR/UCL imaging and synergetic chemotherapy and NIR-triggered PDT. Results: After conducting series of characterizations, the nanotheranostic systems own superior uniform sphericity and long-time stability. In vitro and vivo experiments show the nanocomposites have good biocompatibility and can target caner stem cells to realize simultaneous dual-modality MR/UCL imaging. Furthermore, in comparison with UCMSNs-MX and free MX, MX-loaded UCMSNs conjugated with anti-EpCAM monoclonal antibody (EpCAM-UCMSNs-MX) are eciently endocytosed by cancerous cells and show synergetic effect with PDT in vitro. In vivo experiments reconrm the synergistic effects observed with the combination of EpCAM-UCMSNs-MX and PDT, which results in better treatment outcomes as compared to chemotherapy or NIR irradiation alone that fail to show any noticeable systemic toxicity. Conclusions: The resulting nanotheranostics were shown to target caner stem cells to confer simultaneous dual-modality MR/UCL imaging and induced intracellular reactive oxygen species exposed to 980 nm excitation, leading to synergetic chemotherapy and NIR-triggered PDT. These results offer a promising strategy for designing a multifunctional nanotheranostic system for dual-modality imaging-guided synergistic oncotherapy.


Introduction
Pancreatic cancer (PC) is one of the most intractable cancers and presents extremely limited treatment options, and the 5-year survival rate is below 5% [1]. Although surgical procedure is the preferred treatment for patients with PC, approximately 80% patients suffer from complications that are unmanageable with curative surgery [2]. Moreover, other forms of therapies, including chemotherapy, radiotherapy and immunotherapy, exhibit limited e cacies [3,4], highlighting the need for the development of other therapeutic strategies.
Photodynamic therapy (PDT) has recently drawn extensive attention as a potential cancer treatment, owing to its noninvasive properties and negligible drug resistance [5]. Upon proper light irradiation, photosensitizers (PSs) may convert light energy into oxygen molecules, generating local cytotoxic reactive oxygen species (ROS) that ablate malignant cells[6]. Despite making important success for PDT, signi cant challenges still exist such as poor cancerous endocytosis and insu cient therapeutic effect [7,8]. In addition, the applications of PDT is restricted by its limited penetration depth, as many PSs are activated by visible or UV light [9]. Mitoxantrone (MX), a type of broad-spectrum antitumor drug, is known to have less systemic toxicity than other anthracycline antibiotics. MX has been studied in nanomedicine research [10][11][12][13][14]. Of note, in view of two major absorption peaks at 610 and 660 nm, MX may act as an e cient PS to kill cancer cells following irradiation at 660 nm wavelength [15]. In particular, MX has been veri ed as a remarkably e cient PS that mediates cell apoptosis and damage under light excitation.
In comparison with chemotherapy or PDT alone, the integration of chemotherapy and PDT has been shown to offer better therapeutic outcomes [16][17][18][19][20]. However, previous studies failed to show any in vivo results about introducing one reagent as anticancer drug and PS simultaneously or only studied the therapeutic effects of intratumorally injection or used a higher power density of laser, most importantly rare literature have been reported on active chemo-/PDT targeting pancreatic cancer stem cells [21][22][23]. It is common knowledge that epithelial cell adhesion molecule (EpCAM), as one of the key biomarkers of cancer stem cells (CSCs), involves the proliferation, differentiation, migration, and invasion of cancer. EpCAM is becoming a promising target for theranostic, diagnostic, and/or therapeutic intervention in nanomedicine. The use of MX coupled with EpCAM to allow MX release in CSCs may facilitate the release of adequate ROS, thereby resulting in cell apoptosis and death. However, the current nanosystems exhibit common drawbacks such as the laborious synthetic strategy, poor PS-loading ability, and lack of active CSCs-targeting superiority [24]. The construction of an active EpCAM-targeting drug delivery system that allows e cient loading and targeted delivery of MX is desirable.
Integrating diagnostic and therapeutic function into one nanoplatform, theranostic-oriented nanosystem may realize imaging and therapy simultaneously, thereby facilitating personalized medicine possible.
Lanthanide ion-doped upconversion nanoparticles are deemed as dual-modality magnetic resonance/upconversion luminescent (MR/UCL) imaging probes, owing to their superior physicochemical properties [25,26]. UCNPs-based combination chemotherapy and PDT were demonstrated to show superior treatment outcomes relative to single therapy modality.
Herein, we synthesized innovative active targeting nanoparticles (EpCAM-UCMSNs) by loading MX and used these for MR/UCL dual-modality imaging to investigate their active targeting abilities and therapeutic e cacies in vivo (Scheme 1). Our studies have demonstrated that the integration of PDT and chemotherapy may synergistically improve the e ciency of cancer therapy, which may be a good inception for the application of chemo-/PDT for the clinical treatment of PC.

Synthesis and characterizations
We synthesized NaYF 4 :Yb,Er nanoparticles using a typical thermal decomposition according to our previous report [27]. As displayed in Fig. 1a, NaYF 4 :Yb,Er nanoparticles were monodispersed with an average size of 22.7 nm. Afterwards, a homogenous layer of NaGdF 4 was grown onto NaYF 4 :Yb,Er nanoparticles. As showcased in Fig. 1b, the synthesized NaYF 4 :Yb/Er@NaGdF 4 nanoparticles (UCNPs) displayed a uniform spherical morphology with an average size of 30.3 nm. TEM characterization (Fig. 1c) revealed the good dispersity of the synthesized UCMSNs. UCNPs exhibited a hexagonal phase ( Fig. S1a), which is favorable for achieving highly luminescent and MR imaging probes. Furthermore, elements such as Na, F, Y, Yb, Er, and Gd in UCNPs were determined with energy-dispersive X-ray (EDX) spectroscopy (Fig. S1b). The photoluminescence intensity of UCNPs standardized based on the concentration of Y element increased by 3.5-fold relative to that of NaYF 4 :Yb,Er nanoparticles; the emission intensity mostly contributed to the in vivo detection sensitivity (Fig. 2a). The surface modi cation of nanoparticles with mesoporous silica is known to enhance their biocompatibility and stability under physiological conditions. Furthermore, mesoporous silica may serve as an e cient drug reservoir and vehicle for encapsulating drugs [28][29][30][31]. With these advantages in mind, we converted the oil-phase OA-UCNPs in cyclohexane to an aqueous phase following treatment with a weak acid. CTAC was introduced as for the modi cation of mesoporous silica on ligand-free UCNPs, followed by the coating of mesoporous silica shell onto UCNPs as previously described [32].
EpCAM, as one important biomarker of CSCs in PC, has been established as the target for cancer therapy.
Here, anti-EpCAM monoclonal antibody was covalently grafted onto the surface of UCMSNs based on esteri cation reaction, as con rmed by FTIR absorption spectra characterization. As indicated in Fig. 2b, the characteristic bands at 800-1,200 cm − 1 for UCMSNs, UCMSNs-NH 2 , and UCMSNs-EpCAM may be indexed to Si-O-Si stretching, while the broad band at 3,463.6 cm − 1 corresponded to the stretching vibration of the amino group on UCMSN-NH 2 . The C-H bands at approximately 2,900 cm − 1 and C = O stretching bond at 1,458.0 and 1,377.2 cm − 1 were correlated with the covalent bonding of anti-EpCAM monoclonal antibody on UCMSNs, thereby con rming the successful attachment of anti-EpCAM monoclonal antibody. Furthermore, both UCMSNs and UCMSNs-EpCAM showed a narrow size distribution in PBS; the hydrodynamic diameters of UCMSNs and UCMSNs-EpCAMs were 82.5 and 93.5 nm (Fig. 2c), respectively. Three kinds of nanoparticles were well dispersed in PBS, FBS, and DMEM cell medium for 28 days (Fig. S2). Thus, these particles were markedly stable under physiological conditions. The size range and high stability of these nanoparticles were appropriate for theranostic applications.
Drug release from UCMSNs-MX An effective drug delivery system is the one that shows no reaction with the encapsulated drug while at the same time prevents the premature release of the drug. The sustainable release of the encapsulated compound will result in better therapeutic effects on tumors. As shown in Fig. S3, UCMSNs exhibited high BET surface area (290.1 m 2 /g) and large pores (4.5 nm), which serve as useful properties for drug delivery. We investigated if MX, a widely used antitumor drug and PS, may be loaded into the mesopores of UCMSNs and observed a loading capacity of 23.6%. The in vitro release of MX from UCMSNs-MX was estimated with the dialysis method. As observed from the release pro le in Fig. S4, MX was released from UCMSNs at a relatively faster rate, which may speed up the drug accumulation in cancerous cells and increase the therapeutic effect. The most prominent feature is that MX may play the role of an antitumor drug and PS in antitumor therapy. The encapsulation of MX into the active EpCAM-targeting drug delivery system (UCMSNs-EpCAM) may allow its direct delivery into the cytoplasm, thereby resulting in enhanced

Toxicity evaluation
Before conducting biological experiments, the potential cytotoxicity of UCMSNs/UCMSNs-EpCAM was examined against BxPc-3 cells using CCK-8 assay. As displayed in Fig. 3b&c, no remarkable change was seen in the viability of cells following their incubation with higher concentration (1,000 µg/mL) of these particles for 24/48 h. The negligible cytotoxicity of UCMSNs/UCMSNs-EpCAM showcased their excellent in vitro biocompatibility, which is critical for their in vivo applications. These observations may be associated with the mesoporous silica modi cation on UCNPs in the form of a stabilizing layer, which decreases the leakage of possible toxic ions and the consequent systemic toxicity of nanoparticles.
Biosafety is one key consideration for biomaterials to translate into biomedical applications. Therefore, healthy mice were administrated with 150 µL physiological saline containing UCMSNs/UCMSNs-EpCAM (50 mg/mL) to study their in vivo short/long-term toxicity. The results of the complete blood chemistry assay ( Fig. S5-6) showed that all serum biochemistry parameters were within the physiological range even at 30 days post-injection. Meanwhile, no remarkable weight loss was seen in mice treated with UCMSNs/UCMSNs-EpCAM or control (Fig. S7). Furthermore, no obvious injuries and pathological changes were reported in H&E-stained samples of major organs (Fig. S8), thereby demonstrating that the synthesized nanoparticles are safe for further applications.

Cellular uptake study
EpCAM is overexpressed on the cell membrane of many solid tumors, especially highly upregulated in virtually all epithelial carcinomas. On the other hand, its expression is almost absent in most normal cells, indicative of its potential role as a promising target for theranostic applications. The cellular internalization of UCMSNs/UCMSNs-EpCAM was analyzed using CLSM equipped with an external CW 980 nm laser source. The upconversion luminescence emission of UCNPs was used for in vitro uptake study. BxPc-3 cells were treated with UCMSNs (non-targeted), UCMSNs-EpCAM (targeted), or UCMSNs-EpCAM coupled with free anti-EpCAM monoclonal antibody (the blocking group). As displayed in Fig. 4, a much higher green signal was seen in cells treated with UCMSNs-EpCAM compared to those treated with UCMSNs, demonstrating that anti-EpCAM may remarkably improve the cellular internalization of nanoparticles through receptor-mediated endocytosis. These results con rm that the speci c binding between EpCAM-UCMSNs and EpCAM on the membrane of cancerous cells is essential for the e cient targeting of UCMSNs-EpCAM.

Synergetic effect of Chemo-/PDT in vitro
Aside from its application as a dual-modality imaging probe, UCMSNs may be used for loading anticancer drug, owing to their unique mesoporous structure. We loaded MX into UCMSNs and evaluated the production of cytotoxic 1 O 2 from UCMSN-EpCAM-MX in response to the synergetic effect of chemotherapy and PDT in vitro. As showcased in Fig. 5a, no cell death was observed exposed to NIR irradiation alone. Without laser illumination, cells treated with free MX and MX-loaded UCMSNs (UCMSNs-MX) at 2.5 µg/mL concentration for 24 h showed a survival rate of 12% and 15%, respectively. On the contrary, the percentage of cell death was higher in the group treated with EpCAM-UCMSNs-MX (22.5%) as compared with that treated with UCMSNs-MX (17.4%); the increased treatment e cacy of EpCAM-UCMSNs-MX over MX-UCMSNs was related to more cellular uptake of targeted UCMSNs via antigen-antibody-mediated endocytosis. However, neither MX-UCMSNs nor EpCAM-UCMSNs-MX showed any PDT effect due to no 1 O 2 production. The viability of cells treated with pure MX with or without laser irradiation were 83.7% and 76.8%, respectively. Thus, no signi cant cell death was observed when treated with MX and NIR excitation. Upon NIR irradiation, cell death dramatically increased to 68.8% with MX-UCMSNs and 95.7% with EpCAM-UCMSNs-MX. This drastic difference may be associated with the gradual uptake of EpCAM-UCMSNs-MX in the cytoplasm of BxPc-3 cells through endocytosis as compared to the enhanced permeability and retention (EPR) effect observed for UCMSNs-MX. These results suggest that EpCAM-UCMSNs-MX exhibited superior therapeutic effects as compared to UCMSN-MX and free MX. EpCAM-UCMSNs-MX could accumulate in cancer cells through speci c antigenantibody-mediated endocytosis and release MX to show synergetic chemo-/PDT effect. The high-Z ions (such as Yb 3+ and Gd 3+ ) may help to the process of photosensitization. Furthermore, EpCAM-UCMSNs-MX may be used for EpCAM-targeted dual-modality imaging and MX delivery to achieve synergetic chemot-/PDT effect.
It is known that PDT produces ROS and induces programmed cell death via the activation of apoptotic pathway, wherein protease caspase-3 plays a key role in apoptosis [33]. Caspase-3 is derived from the proteolytic cleavage of the dormant procaspase-3 in the cytosol during the process of apoptosis.
Caspase-9 is rst activated, followed by the subsequent activation of the downstream caspase-3 [34]. In addition, mitochondrial injuries may trigger the release of cytochrome c (cyt c) in the cytoplasm of cancer cells after a series of different treatments. The Bcl-2/Bax protein complex plays a critical role by maintaining the mitochondrial membrane permeability and may induce the expression of downstream caspase-3, which in turn mediates cell survival or death [35,36]. To further explore if ROS can inhibit tumor cell growth and increase their sensitivity to PDT, BxPc-3 cells received eight different treatments as mentioned above. The expression level of proteins involved in the mitochondrial apoptosis pathway was measured by western blot analysis to evaluate the potential mechanism underlying chemo-/PDT-induced cell death. As shown in Fig. 5b, the expression levels of Bax, caspase-3, and caspase-9 substantially increased in the cells subjected to chemo-/PDT treatment as compared with the cells from control, UCMSNs-MX plus NIR irradiation, and EpCAM-UCMSNs-MX without NIR irradiation (P < 0.05) group, thereby indirectly verifying the best antitumor effects in vitro. Caspase-3 in the group treated with EpCAM-UCMSNs-MX with NIR irradiation was approximately 2.4-fold higher than that observed for the control group, further demonstrating the activation of caspase-3 during PDT process (Fig. 5c). The level of Bcl-2 was greatly downregulated in the group with chemo-/PDT as compared with MX group (P < 0.05, Fig. 5d). The chemo-/PDT group exhibited the best anticancer effect, suggesting that EpCAM-UCMSNs-MX plus NIR irradiation may remarkably modulate the expression of survival-related proteins and consequently induce cell death. Thus, MX combined with PDT may suppress cell growth and increase cell apoptosis through the upregulation of Bax, caspase-9, and caspase-3 and downregulation of Bcl-2 expression (Fig. 5e&f). Taken together, mitochondrial apoptosis induced by the synergetic effect of chemo-/PDT was thought to be the key mechanism underlying BxPc-3 cell death.

Dual-modality MR/UCL imaging in vivo
As mentioned before, UCNPs emit luminescence at 660 nm wavelength upon excitation at 980 nm wavelength and may serve as outstanding imaging probes in vivo. The in vivo tumor accumulation of UCMSNs was studied in BxPc-3 tumor-bearing mice. Mice were administrated with UCMSNs/EpCAM-UCMSNs intravenously, respectively. As shown in Fig. 6a, green luminescence was notably different for both non-targeted and targeted group. The signal intensity was much higher in mice administrated with EpCAM-UCMSNs than those with UCMSNs, emphasizing the superior in vivo targeting ability (Fig. 6b). In comparison to UCMSNs that owned passive tumor-targeting ability alone, anti-EpCAM-conjugated UCMSNs owned both active and passive targeting abilities, thus followed by reducing more accumulation even at 48 h. To evaluate the biodistribution of the nanoparticles, mice were euthanized after 48 h of administration, and the tumor and various organs were harvested to image ex vivo. As shown in Fig. S9, much accumulation of nanoparticles was seen in the liver and spleen; moreover, the excised tumors showed more uptake of targeted nanoparticles as compared with non-targeted ones. Although the active targeting ability of EpCAM-UCMSNs may be in uenced by unspeci c proteins in the tumor microenvironment, the e cient in vivo imaging results suggests that the nonspeci c adsorption of plasma proteins, if any, would fall within a very narrow range. Taken together, our results demonstrate that EpCAM-UCMSNs may effectively target PC cells, which warrants further studies in the future.
To explore the active targeting ability of EpCAM-UCMSNs, the in vivo MR scans were conducted before and after the intravenous injection with UCMSNs/EpCAM-UCMSN. Anti-EpCAM monoclonal antibody was grafted onto UCMSNs to allow binding to EpCAM overexpressed on CSCs of PC. As shown in Fig. 6c&d, MR imaging signal intensity increased by about 20.1% and 40.8% in BxPc-3 tumors treated with nontargeted and targeted nanoparticles, respectively, after 4 h of injection. This observation con rms that UCMSNs/EpCAM-UCMSNs were largely delivered to the tumor sites via active/passive targeting capabilities, which increased the signal intensity in MR imaging. Therefore, UCMSNs-EpCAM may be involved into one promising MR contrast speci c to tumor sites.
In vivo synergetic chemo-/PDT The above in vitro results of satisfactory dual-modality imaging and synergetic therapeutic e cacy encouraged us to investigate the therapeutic e cacies of UCMSNs-MX in vivo. When tumors of mice were palpable, they received the above treatments as mentioned above. Changes in relative tumor volume are shown in Fig. 7a. Mice treated with EpCAM-UCMSNs-MX showed better anticancer response than those treated with UCMSNs-MX (P < 0.05), demonstrating the superiority of active targeted drug delivery system. Moreover, the group treated with EpCAMs-UCMSNs-MX + NIR exhibited better results than that treated with EpCAM-UCMSNs-MX (P < 0.01), owing to the synergetic effect of MX-based chemotherapy and PDT triggered by NIR irradiation. The mice exposed to NIR illumination alone failed to show any therapeutic effect as compared with the control group. Furthermore, EpCAM-UCMSNs-MX + NIR group displayed remarkably signi cant delay in the tumor growth as compared with UCMSNs-MXs + NIR group (P < 0.05), as more EpCAM-UCMSNs-MX accumulated at the tumor sites and thus more MX was leaked in the tumor cells that contributed to the synergetic effect of chemo-/PDT. Final tumor volumes are summarized in Fig. S10. Tumors treated with EpCAM-UCMSNs-MX + NIR demonstrated maximum necrosis, as observed by images of H&E-stained tumor slices, which highlight the enhanced anticancer effect of the combination of chemotherapy and PDT as compared with other therapy types (Fig. S11). Meanwhile, no signi cant weight losses were observed in mice treated with EpCAM-UCMSNs-MX + NIR and EpCAM-UCMSNs-MX; a decrease in the body weight of mice treated with NS or NIR alone was observed over time, possibly attributed to the tumor progression and dyscrasia (Fig. 7b). Especially, mice from the chemo-/PDT group showed a remarkable increase in the survival period (P < 0.05, Fig. 7c). In brief, these results suggest that EpCAM-UCMSNs-MX based chemo-/PDT is an effective therapeutic strategy for the treatment of PC.

Biodistribution and targeting e ciency
To further evaluate active targeting e ciency of the EpCAM-UCMSNs/UCMSNs, the concentrations of Y 3+ in the collected tumor and major organs were investigated based on ICP-MS analysis. As shown in Fig. 8, the nanoparticles primarily accumulated in the liver and spleen 12 h post-injection, con rming that the UCNPs-based nanocomposites could accumulate in the organs of the reticuloendothelial system. EpCAM-UCMSNs gradually increased 24 and 48 h post-injection and reached a peak higher than that of UCMSNs.
This remarkably difference was attributed to active targeting ability due to anti-EpCAM-mediated endocytosis. However, the contents of Y 3+ stepped dowm in the tumors and major organs at 48 h postinjection, and this is due to slow clearance of the nanoparticles through a hepatobiliary route to a large extent. In addition, the targeting e ciencies of the above nanocomposites are statistically different at 24/48 h post-injection, respectively (P < 0.05). These results recon rm the excellent active targeting ability of the EpCAM-UCMSNs.

Conclusions
We report the development of a potent theranostic nanosystem based on the combination of UCNPs and MX, which may be used for the dual-modality MR/UCL imaging-guided synergetic chemo-/PDT. China. Tetraethyl orthosilicate (TEOS) sodium chloride (NaCl), triethanolamine (TEA), methanol, and sodium hydroxide (NaOH) were purchased from Lingfeng Chemical Reagent Company (Shanghai, China). Anti-EpCAM monoclonal antibody was supplied by eBioscience (Austria). MX was procured from Toronto Research Chemicals Inc (Canada). Cell counting kit-8 (CCK-8) was purchased from Dojindo (Tokyo, Japan). All reagents were used as received unless otherwise stated.

Synthesis of UCNPs@mSiO 2 nanoparticles (UCMSNs) and UCMSNs-EpCAM
Cyclohexane solution of OA-UCNPs (50 mM, 2.5 mL) was added into 10 mL diluted hydrochloric acid (pH 4.0) and then stirred for 2 h. The products were under centrifugation, rinsed thrice with deionized water, and nally dispersed in 20 mL deionized water. TEA (0.01 g) and CTAC (0.5 g) were added and vigorously stirred for 2 h. A total of 10 mL ligand-free UCNPs solution was then added slowly, and the mixture was sonicated for 1 h. A total of 200 µL TEOS was added dropwise into the system and stirred vigorously at 80°C for 1 h. The obtained UCNPs@mSiO 2 nanoparticles (denoted as UCMSNs) were precipitated, washed thrice with ethanol, and extracted with 30 mL methanol solution of NaCl (1 wt %) at 25°C for 3 h to clear off the excess CTAC. After several cycles of extraction, UCMSNs were dispersed in deionized water for further use.
Anti-EpCAM monoclonal antibody was conjugated onto the surface of UCMSNs via an EDC/NHS coupling chemistry. Brie y, 50 µg anti-EpCAM monoclonal antibody in 10 mL MES buffer (pH: 6.0, 0.1 M) was reacted with equimolar ratio of EDC and NHS at 25°C for 15 min. Following incubation, UCMSNs-NH 2 nanoparticles dispersed in 100 mL phosphate-buffered saline (PBS, pH 8.5) were added to the above mixture and the system was performed for 2 h at 25°C. The resulting UCMSNs-EpCAM nanoparticles were centrifuged thrice with PBS to remove the unbound antibody and the nal product was dispersed in 0.5 mL deionized water for use in subsequent experiments.

Characterization instruments
The morphologies of the above nanoparticles were studied using JEOL JEM-1200EX transmission electron microscope (TEM, Japan), while the size distribution study was performed using Image J

NIR-induced ROS generation in vitro
The intracellular ROS generation was determined with an oxidation sensitive uorescent probe 1,3diphenylisobenzofuran (DPBF) [28]. Brie y, 3000 BxPc-3 cells were cultured into a 96-well plate overnight, then were rinsed with sterile PBS and incubated with DMEM containing EpCAM-UCMSNs-MX/UCMSNs-MX (300 µg/mL) for 4 h. After incubation, the medium was removed and the cells were incubated with 100 µL DPBF for 1 h, then the cells were rinsed with PBS three times and illuminated with 980 nm light (320 mW/cm 2 ) for 10 min. The production of ROS was uorometrically determined by examining the amount of DPBF and comparing it with the predetermined DPBF standard curve.
Cellular uptake study and targeting e ciency of UCMSNs/UCMSNs-EpCAM BxPc-3 cells were cultured in confocal laser scanning microscopy (CLSM) special cell culture dish under a humidi ed 5% CO 2 atmosphere at 37°C. After reaching about 80% con uence, the cells were rinsed thrice with PBS. Following washing, the cells were treated with DMEM solution containing UCMSNs (200 µg/mL) and UCMSNs-EpCAM (200 µg/mL), respectively. After 4 h co-incubation, cells were gently rinsed with PBS three times and DAPI (1:1000) were used to stain the cell nuclei for 15 min. The excitation wavelength used was 980 nm.
In vivo toxicology study of UCMSNs/UCMSNs-EpCAM Healthy female athymic nude mice (weight: 20 g, 5 weeks old) were purchased from the Model Animal Research Center of Nanjing University and raised at Laboratory Animal Center of Southeast University. All animal experimental procedures were approved by the Institutional Animal Care Committee at Southeast University. Mice were randomly divided into three groups as follows (six mice per cohort): two groups were administrated with a single dose of UCMSNs/UCMSNs-EpCAM in physiological saline (30 mg/mL, 150 µL) intravenously via tail vein. The third group was administrated with only 150 µL physiological saline and used as the control. Before blood collection, mice were anesthetized after 7 and 30 days of treatment and blood samples (approximately 500 µL) were acquired through cardiac puncture for biochemistry assays. All mice were euthanized and the major organs (heart, liver, spleen, lung, and kidney) were excised and xed in a 10% formalin solution, followed by hematoxylin and eosin (H&E) staining.

Synergetic effect of chemo-/PDT in vitro
A total of 10 5 BxPc-3 cells were planted in a six-well plate and incubated overnight. After 80% con uence, the cells were rinsed with sterile PBS three times, then divided into eight groups and treated as follows:  To further evaluate the treatment e cacy, the tumors and major organs were collected for H&E staining. The representative photos of tumors were recorded, respectively. Seven mice from each group were used for analyzing survival time.
Biodistribution and targeting e ciency As mentioned above, BxPc-3 tumor-bearing mice were administrated with EpCAM-UCMSNs/UCMSNs (15 mg/mL, 150 µL, 5 mice per group) via tail vein, and euthanized at 12, 24, and 48 h post-injection. The contents of Y 3+ in the tumor and major organs were measured based on ICP-MS. In addition, the targeting e ciency was evaluated as follows: targeting e ciency (%) = (total content of Y 3+ in the sample/total content Y 3+ in the injected nanoparticles) × 100%.

Statistics analysis
All the obtained data were analyzed using GraphPad Prism software (version 5.0). Differences between two groups was regarded as statistically signi cance for *P < 0.05 and very signi cant for **P < 0.01 and ***P < 0.001. Figure 1 TEM images and size histograms of NaYF4:Yb,Er (a), NaYF4: Yb,Er@NaGdF4 (UCNPs, b) and