A bioactivated in vivo assembly (BIVA) nanotechnology fabricated NIR probe for small pancreatic tumor intraoperative navigation imaging

Li-Li Li (  lill@nanoctr.cn ) CAS Center for Excellence in Nanoscience, CAS Key Laboratory for Biomedical Effects of Nanomaterials and Nanosafety, National Center for Nanoscience and Technology (NCNST) https://orcid.org/00000002-9793-3995 Han Ren National Center for Nanoscience and Technology (NCNST) Xiao-Xiao Zhao National Center for Nanoscience and Technology (NCNST) Dayong Hou the Fourth Hospital of Harbin Medical University Haodong Yao Institute of High Energy Physics https://orcid.org/0000-0001-5959-8594 Lina Zhao Institute of High Energy Physics, Chinese Acad Sci https://orcid.org/0000-0002-9796-0221 Wanhai Xu Fourth A liated Hospital of Harbin Medical University Hao Wang National Center for Nanoscience and Technology


Introduction
Surgical resection remains the mainstay of treatment for patients with tumor of any grade, including pancreatic cancer, which known as the "king of cancer". [1][2] However, for small lesions or small metastases in pancreatic cancer (tumor diameter < 2.0 cm), which could be hardly identi ed currently, are truly dependent on the experience and expertise of the surgeon. [3] Otherwise, according to the diagnosis of the cutting edge which is made by intraoperative frozen pathological section, the accuracy rate is only about 50%, which greatly increased the risk of recurrence and metastasis. [4] It can be found that the 5-year survival rate of pancreatic cancer without metastasis and whose tumor diameter less than 2.0 cm can be increase to 19-41% after surgical resection. [5] Meantime, the risk of perioperative morbidity can be minimized through the combined use of MIR/CT imaging before operation and uorescence-based intraoperation imaging. [6][7][8] As a real-time tumor imaging molecule, uorescent probe has been widely used in clinical practice to provide information on tumor diagnosis and drug development. [9][10][11][12] The most famous probe is indocyanine green (ICG), which is a Food and Drug Administration (FDA)-approved near-infrared uorescent dye, has been used in clinical intraoperation navigation. [13] These small molecules lack of active targeting and have a poor retention in tumor, which narrowed the detection window for complicated surgery. [11] For a better speci city, researchers developed tumor microenvironment turn-on probe to response ROS, lack of oxygen, pH etc. [14][15][16] However, traditional small molecule uorescent dyes metabolize quickly and are easy to be removed by liver and kidney, which shortens the imaging detection time. [17] Secondly, the aggregation-caused quenching (ACQ) effect of such uorescence dyes limited the molecule accumulation in tumor and the photobleaching behavior also confused the imaging. Other way in recent years, the delivery of uorescent probes in vivo has attracted much attention. [18] Like drug delivery, [19] uorescent probes need to be transmitted through blood to the tumor site for better osmotic enrichment at the tumor site to generate uorescence in a stable and long-term manner. [20][21] Herein, we reported a bioactivated in vivo assembly (BIVA) probe (named M1) for speci c and sensitive pancreatic tumor imaging (Fig.1). The BIVA probe was modular designed with ve modules, including long-term circulation motif (mPEG 2000 ), response tailoring motif (Gly-Pro-Ala), self-assembly motif (Lys-Leu-Val-Phe-Phe-Gly-Cys-Gly), targeting motif (Arg-Gly-Asp) and imaging motif (IR 783 ). Firstly, the probe enabled to long-term blood circulate and had more probability to reach to its target. Next, the targeting motif (RGD) speci c recognized and bonded onto the receptor αvβ3 integrin. The over expressed broblast activation protein-α (FAP-α) on the membrane of cancer-associated broblasts (CAFs) tailored the molecules to induce in situ assembly with typical β-sheet structure. Finally, the assembled nano bers located around the membrane of CAFs in the tumor microenvironment and well labeled the edge of tumor. Based on in vivo self-assembly nanotechnology, the BIVA probe endowed an optimized biodistribution half-life (long blood circulation) and elimination half-life (assembled induced retention effect) with dramatically enhanced area under the curve (AUC) leading to increased drug availability. Otherwise, the targeting and tailoring induced assembly on the membrane of CAFs had a spatial controlled speci city, which also contributed to enhance tumor imaging sensitivity. In this work, we provided an optimized imaging probe delivery system based on BIVA effect, which exhibited a synergetic and enhanced new targeting mechanism: active targeting plus assembly induced retention (AIR) effect. We believed that such a BIVA effect and modular designed BIVA probe will offer us a tool for different imaging molecule delivery.

Results And Discussion
Molecular design, assembled behavior and conformation. In order to study the molecular design and their assembly behavior, we modular designed and synthesized seven molecules (Table 1), which respectively were BIVA probe: M1 (mPEG-GPAKLVFFGC(IR783)GRGD); the non-FAP-α tailoring molecule: M2 (mPEG-GPAKLVFFGC(IR783)GRGD) scrambled response tailoring motif of Gly-Pro-Ala with Ala-Gly-Gly; the nonlabeling molecule of M1: M3 (mPEG-GPAKLVFFGCGRGD) removed IR 783 labeling; FAP-α tailoring residue of M1: R-M1 (AKLVFFGC(IR783)GRGD); the non-targeting molecule: M4 (mPEG-GPAKLVFFGC(IR783)GDTG) scrambled targeting tailoring motif of Arg-Gly-Asp with Asp-Thr-Gly; the nonlabeling molecule of M2: M5 (mPEG-AGGKLVFFGCGRGD) removed IR 783 labeling; FAP-α tailoring residue of M3: R-M3 (AKLVFFGCGRGD). All the synthesized procedure ( Supplementary Fig. S1) and the characterizations of these seven molecules can be found in SI ( Supplementary Fig. S2-Fig. S8). As seen in Table 1, although scrambled the tailoring motif of M2 and targeting motif of M4, the critical assembly concentration (CAC) of these two molecules in aqueous solution was like that of M1, both of which were above 500 μM. In the absence of IR 783 labeling, the CAC values of M3 and M5 molecules in solution were lower than that of M1 and M2, which was due to that fact that the steric hindrance of hydrophilic IR 783 was closed to the self-assembly motif ( Supplementary Fig. S9). When we removed the long-term circulation motif of mPEG 2000 tail of M1 and M3, the CAC of the truncated residues of R-M1 and R-M3 dramatically decreased by more than two orders of magnitude ( Supplementary Fig. S10). The CAC was quantitatively calculated by β-sheet structure sensitive uorescence probe Thio avin T (ThT). All the results indicated that both hydrophilic mPEG tail and IR 783 labeling contributed to the solubility of molecules in aqueous solution. Through the molecular dynamics (MD) simulation calculations, [22] it was clearly observed that both the backbones of M1 and the residual R-M1 of M1 were β-hairpin conformations (Fig. 2). The mPEG tail was close to the self-assembly motif through multiple hydrophobic interactions to stabilize its conformation, including hydrogen bonds VAL4:CYS8, ARG10:ALA1, ARG10: Cy, ARG10:CASP12, ARG10:CASP12, and salt-bridge ARG10:CASP12 on the both side of the hairpin. Interestingly, the labeling of IR 783 was almost perpendicular to the β-hairpin backbone and mPEG tail, which formed a signi cant steric hindrance preventing the further intermolecular assembly. When the mPEG motif was tailored, the backbone of R-M1 remained its β-hairpin structures by hydrophobic interactions, hydrogen bonds GLY9:PHE6, LYS2:CASP12, ARG10:CASP12, LYS2:CASP12, ARG10:CASP12 and salt-bridges LYS2:CASP12, ARG10:CASP12 on the both side of the hairpin, while the IR 783 showed an obvious intramolecular rearrangement, resulting in its alignment parallel to the backbone. The decrease of hydrophilicity of molecule and the exposure of hydrogen bonds on the self-assembled surface were bene cial to the occurrence of intermolecular dynamic assembly.
To further evaluate the assembled structures, the corresponding Circular Dichroism (CD), Fourier Transform Infrared (FTIR) and Wide Angel X-ray Scattering (WAXS) spectra were characterized. As shown in Fig. 3a, CD spectrum of M3 assemblies had a positive band at λ=193 nm and two negative bands at λ=208 nm, λ=225 nm respectively, which implied a β-sheet and α-helix hybrid structure. In contrast, under the same concentration, M1 molecules had a random coil secondary structure in CD spectrum as monomers. Meantime, the FITR spectra of M1 and M3 in Fig. 3b clearly observed the intermolecular interactions. The typical ngerprint peaks of β-sheet bril structure in the amide I region (1700-1620 cm -1 ) included 1694 cm -1 , 1662 cm -1 and 1631 cm -1 . [23] Whereas, the broad peak at 1647 cm -1 indicated that the structure of M1 was disordered. The above evidences indicated that the M3 molecule was easier to assemble than M1 in the absence of IR 783 , and the driving forces of assembly depends on the multiple hydrogen bonds and other weak interactions of the self-assembly motif. After tailoring the hydrophilic balance of mPEG, the R-M1 exhibited a well-ordered β-sheet assembled secondary structures with a typical strong positive band at 196 nm and a wide negative band at 216 nm (Fig. 3c).
The R-M1 molecules had a rapid dynamic assembly process within minutes, and the assemblies in aqueous solution had obvious Dundal phenomenon. As a homologous sequence with amyloid β-protein (Aβ), the self-assembly motif of peptide sequence of KLVFFGCG was similar to that of (Aβ)42 peptide in aggregation kinetics, i.e. dynamic growth from oligomers to amyloid brils. [24] The aggregation starting from the freshly isolated monomers of R-M1, the precipitates were separated in 1 minute and 1 hour, respectively. The FTIR spectra of these two samples ( Fig. 3d) exhibited completely different spectral features. The rapid separated one had a broad peak at 1634 cm -1 , which was identi ed as oligomer; while the one with extended aggregated time had three weak peaks at 1696 cm -1 , 1673 cm -1 and 1633 cm -1 respectively, which indicating as anti-parallel β-sheet brils. [25][26] Characteristic of nucleated growth procedure (Fig. 3e), the aggregation curves with ThT trace had a growth phase for primary process from the initial 17 min, an elongation phase for surface-catalyzed secondary process between 17-30 min, and a nal plain phase after 30 min. The dynamic growth procedure was like the 8-anilino-1naphthalenesulfonic acid (ANS) stained curve ( Supplementary Fig. S11). As known, ANS was sensitive to hydrophobic interaction. [27] When R-M1 were in the initial oligomer, the uorescence intensity of ANS increased due to the enhanced hydrophobicity. When the molecules were elongated and stacked in a higher ordered nano brils, the blue shift of ANS in the β-sheet structures reduced the uorescence. The molecular packing mode of well-ordered brils of R-M1 in Fig. 3f observed a weak re ection at 4.9 Å as laminates space and a strong broad re ection at 10.3 Å as sheet space, which was illustrated in the inserted gure. The bril morphology was characterized by Transmission Electron Microscope (TEM) imaging (Fig. 3g). The statistical calculation of the ber diameter in TEM images was 5.0 ± 0.4 nm ( Supplementary Fig. S12), which was well corresponding to the theoretical calculated two molecule length of R-M1. We assumed that the nano bers were assembled by twisted R-M1 molecules centered on selfassembly motif.
Speci c enzyme tailoring induced nano bril assembly. To further investigate the FAP-α speci c tailoring and BIVA probe assembly in situ simultaneously (Fig. 4a), the High-Performance Liquid Chromatography (HPLC), TEM and Matrix-Assisted Laser Desorption/Ionization Time-of-Flight (MALDI-TOF) mass spectrometry were all used for characterizations. To speci cally cleave the response tailoring motif (Gly-Pro-Ala), the pre-synthesized molecules R-M3, M3 and M5 were set as controls. After incubation with FAP-α for 12 h, the M3 molecules were totally cleaved by the enzyme (Fig. 4b), resulting in truncated residues with different retention times compared to M3 (29.5 min). Compared the residues peaks with R-M3, the primary sharp peak at 27.4 min can be identi ed by the R-M3 control (27.6 min). Meantime, the wide peak at 14.6 min might be the rest PEG residue. In a sharp contrast, after incubation FAP-α with M5, there was no change of the retention peak, which double conformed that the GPA was the FAP-α speci c recognized sequence and the molecule was cut between the amino acid of Pro and Ala. After hydrolyzed by FAP-α, the residues of M1 in situ assembled into nano bers structures under the complex buffer ( Fig. 4c and Supplementary Fig. S13). The tailored residues were identi ed by MALDI-TOF (Fig. 4d), which revealed that M1 were cut into two parts of R-M1 and PEG residue. Additionally, when the molecules of M1 were incubated with Miapaca-2 cells for 2 h, the cell lysis was observed to split into layers. MALDI-TOF con rmed that R-M1 in the precipitate might be the assembly induced precipitate, and the supernatant obviously contained the PEG residues ( Supplementary Fig. S14). In order to further evaluate the speci city of M1 for different enzymes, including FAP-α, pepsin, pancreatin, lipase and BSA, we used ThT (Thio avin T) as a detection probe (Fig. 4e). After co-incubation with M1 for 12 h, only FAP-α group strikingly enhanced the uorescence intensity, which indicated that M1 molecule had speci city for FAP-α induced in situ assembly.
Enhanced targeting and in situ high e ciency nano ber formation located cell outline. In design, the bioactivated in vivo assembly (BIVA) was a triggered and synchronous dynamic assembly system with active targeting cooperative assembly induced retention (AIR) effect. Compared with active targeting mechanism dependent on binding constant Kd, BIVA effect showed an ampli ed mechanism based on primary binding constant Kd and secondary assembly rate constant Ka (Fig. 5a). To further con rm our hypothesis, the molecule M1 and M2 was equally labelled by FITC uorescence probe for cell imaging (Fig. 5b). To simulate the dynamic physiological condition, the cell culture medium was replaced every 15 min. After 1 h incubation with M1 and M2 at the same molecule molar concentration, there were signi cant differences between the two molecules (Fig. 5c). The M1 with BIVA effect had higher uorescence retention rate on the cell membrane, while the M2 with active targeting mechanism signi cantly reduced the uorescence signal during the dynamic incubation. The huge different can be explained by the fact that the secondary assembly rate broke the balance between the targeted ligand and receptor, and thus tended to a stable assembly interaction. The retention e ciency depended on the assembly rate constant Ka. Meanwhile, the rapid dynamic assembly of BIVA probe in situ around the cell membrane, contributed to the e cient formation and retention of nano bers on the cell pro le. When delayed the incubation time up to 2 h, some components could be endocytosed into the cells, but most of them were found on the membranes ( Supplementary Fig. S15). As shown in Fig. 5d, most molecules of M1 were assembled and located on the cell membrane within 1 h of incubation. Then, the isolated cells were collected and lysed, the extracted cell membrane fragments were stained with ThT dye. Not surprisingly, the nano bers on the membrane were all stained by ThT, and the correlation coe cient between ThT and FITC uorescence was high up to 0.83 (Fig. 5e).
In order to further understand the contribution of the active targeting and AIR effect during locating of the cells. The cell transwell experiment were used to quantitively evaluate interference of the cell migration based on different mechanism (Fig. 5f). As expected, untreated cells were easy to migrate to the lower chamber, while the cells treated with active targeting molecule M2 and assembled molecule M3, the migration of cell were reduced. Under the same molar concentration, the M1 treatment group had most inference on cell migration (Fig. 5g). According to the quantitative statistical calculation of the number of cells ( Supplementary Fig. S16), the results in Fig. 5h clearly veri ed that BIVA effect indeed had a high trapping and localization e ciency around cells. Although the targeting group of M2 and BIVA group of M1 had signi cant interaction with cells, there was no obvious cytotoxicity at high concentration of 300 μM (Supplementary Fig. S17).
Metabolic difference and optimized biodistribution enhanced imaging. According to the time-dependent in vivo NIR images, there were signi cant differences in the uorescence distribution among ICG, M2 and M1 mice (Fig. 6a). The representative small molecule probe was ICG, which showed rapid distribution and elimination all over the mice body with no obviously speci c targeting effect on tumor tissue.  Fig. S18).
Otherwise, the molecules of M1 and M2 with a mPEG tail, both had long-time circulation half-life (t 1/2 ), which were 110 ± 3 min and 102 ± 4 min, respectively (Fig. 6b). The blood circulation half-life (t 1/2 ) of ICG was 2.5 ± 0.5 min. The short t 1/2 was related to the rapid distribution and elimination behavior in vivo. In order to understand the contribution of these elements to effective availability of imaging probe. The signi cant parameter of pharmacokinetic: area under the curve (AUC) in tumor tissue were obtained according to the quantitative calculation from uorescence signal. After quantitative calculation of the concentration of probe in tumor area, the time-dependent curve of M1 and M2 were carried out (Fig. 6c).
The area under the curve ranges from 0 h to 120 h (AUC 0-120 h) of M1 was 3.6 times more than that of M2, which mean that with a single dose administration, the average uorescence intensity distribution per unit area of M1 in tumor tissue was 3.6-fold higher than that of M2. The time to peak of M1 was 4 hours later than M2, about at 12 h. The highest signal on tumor of M1 was 1.8 times higher than M2. In addition, the signal elimination of M1 from tumor was quite slow, only 27.6% was reduced between 12 h and 120 h, while the uorescence signal of M2 was disappears completely in the same time interval. Finally, we obtained a stable intraoperative navigation window between 8 h and 96 h for our BIVA probe (M1). As the conclusion, the long-term blood circulation and the dynamic enzyme tailoring both helped the continuous accumulation in tumor area. The in-situ assembly in tumor tissues slowed down the dynamic elimination and prolong the elimination time, which contributed to maintain the imaging signal during surgery operation. The FAP-α speci c tailoring and assembling of M1 differed the tumor from the other tissues, which offered better contrast and biodistribution. To evaluate the imaging property of BIVA probe, the orthotopic pancreatic tumor mice model was built. After intravenous injected M1 and M2 molecules with a dose of 16 mg/kg for 12 h, the mice were sacri ced. When dissected the spleen, the high contrast signal was clearly observed on the orthotopic tumor area (Fig.7a). Then, all the important organs were dissected for ex vivo imaging. The signi cant difference between M1 and M2 on the tissue biodistribution. For BIVA probe M1, the distribution on tumor had obviously selectivity, and the molecules had part retention in the metabolic organs (e.g. liver and kidney). Whereas, the M2 exhibited no signi cant difference in the biodistribution of lung, kidney and tumor, but most of the molecules were stuck in liver. The huge difference between the two molecules can be explained by the high speci c recognition of FAPα to M1, which was conducive to e cient molecular tailoring and assembly in tumor, while the M2 was non-speci c cleavage and accumulation in liver during metabolism. The quantitative analysis results also con rmed the conclusion (Fig.7b). The signal accumulation of FAP-α speci c BIVA probe M1 was twice as much as that of M2. Under the same blood circulation time, organ selectivity depended on the speci city of substrate to target enzyme. Its accumulation amount relied on the cleavage rate of enzyme and aggregation e ciency of molecular residues. The primary nucleation of assembly can induce long lasting growth of the bril in tumor, reduce the metabolic rate, and achieve the retention and accumulation of tumor.
Upon the individual difference, we validated 6 mice under surgery to induce orthotopic tumor in pancreatic head. All the positive results were obtained including the small sized tumor around 2 mm in the diameter (Fig.7c). The ex vivo dissection in Fig.7d provided us a fantastic imaging contrast on tumor and the around spleen tissue, which visualized identi ed the small tumor (~ 2 mm). The statistical results of uorescence signal on tumor were over 9 folds higher than the surround spleen tissue (Fig.7e). The whole tumor histologic section in Fig.7f and Supplementary Fig. S19 stained by Congo Red full viewed the bril distribution inside tumor. As known the FAP-α was a membrane located protein, overexpressed on tumor associate broblast cell and pancreatic cell surface. The FAP-α speci c BIVA probe M1 were well depicted the tumor margin and interstitial space, which concentrated the signal in tumor for better bioimaging, but the M2 has no obvious Congo Red, which mean there was no assembly inside tumor.
Acute toxicity evaluation to organs. The acute toxicity evaluation of M1 to mice were veri ed by blood biochemistry, hemograms and histological analysis. The representative biomarkers of liver function included alanine aminotransferase (ALT), alkaline phosphatase (ALP), aspartate aminotransferase (AST), total protein (TP) and albumin concentration (ALB). Compared to the healthy group (PBS), there had no obvious hepatic toxicity after i.v. injection of 16 mg/kg of M1 for 24 h (Fig. 8a). In addition, the hematological assessment results including creatinine (CREA), white blood cells (WBC), red blood cells (RBC), hematocrit (HCT), mean corpuscular volume (WCV), mean corpuscular hemoglobin concentration (MCH), mean corpuscular hemoglobin concentration (MCHC), red cell volume distribution widthcoe cient of variation (RDW-CV), platelet (PLT)and mean platelet volume (MPV) were carried out (Fig.  8b). All the above indicators in PBS group and M1 group appeared normal, which was basically consistent with the normal range reported in the literature. Then the mice from PBS group and M1 group were sacri ced for the further histological section analysis of the signi cant organs. After Hematoxylin and Eosin (H&E) staining, the slices of heart, liver, spleen, lung, kidney and pancreas were compared and evaluated (Fig. 8c. There was no noticeable organ damage and tissue injury of the two groups. All the evidences revealed that under the imaging dose, the BIVA probe had no acute toxicity performance.

Discussion
Although the optical probes had a great potential in the further clinical trials, the druggability still a bid issue. As a great improvement, the targeting ligand covalent coupling such as antibody, peptide and etc, signi cantly increased the targeting property of the probes. However, the optical probes based on active targeting mechanism were always in a narrow imaging window, which cannot meet the need of stable imaging for intraoperative navigation. Thus, we proposed a bioactivated in vivo assembly (BIVA) nanotechnology to fabricate NIR probe, which offered a new targeting mechanism based on synergy of active targeting and AIR effect. The modular designed BIVA probe were speci c to pancreatic tumor overexpressed FAP-α, resulting a sensitive transformation from monomers to assembles in vitro and in vivo, resulting a high contrast (signal-to-background ratio SBR > 9 folds) imaging window from 8 h to 96 h. Meantime, the mPEG tail provided a long-time blood circulation half-life (t 1/2 ) up to 110 min. The assembling triggered by membrane protein FAP-α induced a speci c pancreatic tumor cells accumulation around cells. The site-speci c assembly contributed to great difference of metabolized time between tumor cells and surround healthy cells, further optimizing the organ distribution. All the efforts realized a highly sensitive small (< 2 mm) orthotopic pancreatic tumor imaging in vivo. We believed that the BIVA nanotechnology would upgrade the imaging probe into clinical drug. The dynamic assembled property and assembled conformation modulation would be optimized according to the surgical imaging needs. Such a delivery system can further be expanded to carry different uorescent molecules, the resulted BIVA probes will bene t future precision medicine. Molecular dynamics (MD) simulation: The GROMACS software package (version 5.1.4) was used to perform energy minimization and molecular dynamics simulations using the AMBER99SB-ILDN force eld. The topology and force eld parameters les for the PEG were generated with Acpype Antechamber.

Methods
All the structures were solvated in a box of TIP3P water model, and then ionized and neutralized with Na + and Cl − ions as 0.15 mol/L. The periodic boundary conditions (PBC) were set for all directions. The NPT ensemble was applied. After energy minimization, each structure underwent NVT equilibration and NPT production. The Nose-Hoover thermostat [29] was used to maintain the temperature of system as 300K.
The pressure was maintained as 1 atm by coupling the semi-isotropic (X+Y, Z) directions of the system using the Parrinello−Rahman algorithm. [30] The van der Waals (vdWs) interactions were computed with a cutoff distance as 12 Å, while long-range electrostatic interactions were handled with the particle mesh Ewald (PME) method. [31] The water molecules were constrained by the SETTLE algorithm, [32] and the hydrogen bonds were constrained to their equilibrium values employing the LINCS algorithm. [33] The timestep was 2 ps in production runs, and the coordinates were saved every 100 ps. Totally time Spectrum range was between 4000-450 cm -1 . All results were averaged among 16 measurements. ATR was used to collect the spectrum.
Transmission electron microscopy (TEM): TEM was used to characterize the morphology of the assemblies of R-M1 and M1 after enzyme cleavage. The instrument used was TECNAI G2 20 s-twin electron microscope. The sample (20 μL, 100 μM, made of deionized water) after standing for 12 h was dropped onto the copper mesh. After standing for 10 min, the upper liquid was removed and the copper mesh was dyed with 2% uranyl acetate for 1 min. Finally, the surface of the copper mesh was washed with distilled water. The copper mesh was placed overnight at room temperature and then observed. and placed in a lateral position (right side down). After the pancreas was exposed under sterile conditions, MiaPaCa-2 cells resuspended in Matrigel were injected into the parenchyma of pancreas of mice.
Subsequently, the body wall and skin incision were successively closed after the pancreas was returned into the peritoneal space. After con rming the tumor growth in pancreas, the tumor speci c recognition of M1 and M2 were investigated with IVIS.
Tumor slices and staining: Tumors were harvest and collected in 4 % paraformaldehyde solution after treated with M1 at a dose of 16 mg/kg for 12 h. The H&E and Congo red staining procedure was performed by Google biotechnology (Wuhan) co. LTD.
Toxicology evaluation: The female Balb/c nude mice (6-8 weeks, n = 3) were sacri ced for blood and major organs collection after treated with M1 at a dose of 16 mg/kg on the 24 h. The histology evaluation of major organs was performed by Google biotechnology (Wuhan) co. LTD. The blood biochemistry and haematology analyses were carried out by Vital River Laboratory Animal Technology Co. Ltd.
Statistical methods: Data are reported as mean ± standard deviation. Statistical analysis of the data was performed with one-way ANOVA followed by post hoc Tukey's test. Statistical signi cance was de ned as *p < 0.05, **p < 0.01 and ***p < 0.001.