Screening of an annexin‐A2‐targeted heptapeptide for pancreatic adenocarcinoma localization

Annexin A2 (ANXA2) encodes an oncoprotein whose expression has been found to correlate with poorer overall survival (OS) of pancreatic adenocarcinoma (PAAD) patients. Although peptides are available for targeting ANXA2, none of these were initially selected to target this protein specifically. Here, we took ANXA2 as a molecular target for PAAD and employed the phage display technique to screen for a new ANXA2‐targeted peptide. The resultant heptapeptide, YW7, was firstly labeled with fluorescein isothiocyanate (FITC) to evaluate its selectivity in cellular uptake, and further with the near‐infrared fluorescent (NIRF) dye Cy7 to assess in vivo distribution in a mouse model bearing PANC‐1 human pancreatic cancer xenografic tumors. We found that both FITC‐YW7 and Cy7‐YW7 probes showed significantly higher uptake in PANC‐1 cells compared to the HPDE6‐C7 pancreatic epithelium cells. Mice intravenously injected with Cy7‐YW7 showed higher tumor‐to‐background ratios (TBRs) (~ 2.7‐fold) in tumor tissues compared to those injected with Cy7 alone. Our study suggested that YW7 is a novel peptide targeting ANXA2 and Cy7‐YW7 is an NIRF probe potentially useful for the early detection of PAAD.


Introduction
Pancreatic adenocarcinoma (PAAD) is a malignant digestive tract tumor with an extremely poor prognosis. The 5-year survival rate of PAAD is only 10% [1]. Thus far, surgery is still the best option for PAAD treatment, but the lack of strategies to diagnose PAAD early unqualifies 80-85% of PAAD patients, who have already developed metastasis at the time of diagnosis, for surgical resection [2]. Furthermore, the inability to locate tumors in preoperative and intraoperative examinations imposes a challenge for complete tumor resection. There is an urgent need for new strategies for early tumor detection [3] and localization [4] to improve the treatment outcomes of PAAD patients. Molecular imaging is a promising tool for non-invasively characterizing biological events and thus facilitating both preoperative and intraoperative decision-making [5]. Currently, molecular probes for magnetic resonance imaging (MRI), positron emission tomography (PET), and single-photon emission tomography (SPECT) have been developed [6] and demonstrated the potential to facilitate early detection of PAAD in preclinical models. For example, a radiolabeled peptide specific to the DNA damage signal, 111 Inanti-cH2AX-TAT, has been used for the sensitive detection of pancreatic cancer at early stages in mice [7]. In another study, IGF1-targeted gadolinium-labeled nanoparticles TBR have been developed as a theranostic platform for both MRI detection and treatment of murine pancreatic cancer models [8]. Therefore, molecular imaging is very important for early detection to achieve timely, accurate, personalized treatment. Based on this, the ability of molecular targeted probes to accurately target tumor biomarkers specific to tumor aggressiveness is of critical importance.
Peptide-based imaging probes have emerged as an amenable tool to achieve higher tumor permeability and specificity. Many researchers have selected peptides as potential tumor ligands, due to their smaller size and versatility for chemical modification and costeffectiveness [9]. Tumor-targeted peptides can be derived from natural products, bio-panning libraries, or synthetic peptide arrays. Phage display technology [10] is a classic and efficient way to screen for short peptides against a specific target using the biopanning process. Molecular targeting peptides that screened from phage display, such as VEGFR, EGFR, and HER2-targeted peptides, have specifically shown high tumor-specific affinities [11]. Recently, annexin A2, a 36KD calcium-dependent phospholipid-binding protein of the annexins family encoded by the ANXA2 gene, has been identified as a novel biomarker in PAAD as elevated ANXA2 expression was found to significantly correlate with poorer overall survival (OS) and disease-free survival (DFS) of PAAD patients [12]. Besides, ANXA2 has been found to be intracellularly and extracellularly overexpressed in various types of tumors, in which ANXA2 translocates to the tumor cell membrane and facilitates the generation of plasmin in the extracellular matrix, resulting in enhanced cancer cell invasion and migration [13,14]. ANXA2 has been used as a valuable biomarker for both cancer diagnosis and prognosis [15]. Thus far, peptides have been discovered for developing cancer-specific imaging probes via targeting ANXA2. For example, TM601 (a 36-amino acids peptide) [16], which is a synthetic peptide with the same sequence as chlorotoxin (CTX) derived from the venom of the scorpion Leiurus quinquestriatus, has been found to target both matrix metalloproteinase (MMP2) and phosphorylated ANXA2 and used to develop an optical imaging probe named BLZ100 for detecting multiple types of tumors. Other CTX-based bioconjugates, such as 131 I-TM601, are also being investigated for tumor detection [17]. In a recent study, a cyclic octapeptide, LS301, was also found to target both integrin receptors and phosphorylated ANXA2 in the presence of calcium, and the ability of LS301 to emit near-infrared fluorescence makes it a desirable imaging probe to illuminate tumor margin [18]. However, those peptides were not initially selected to target ANXA2. We postulated that peptides that target ANXA2 would presumably allow selective homing to membrane-bound ANXA2 in the PAAD mouse model, because of the low levels of ANXA2 in non-tumor tissues. The discovery of such an ANXA2specific peptide could greatly facilitate further clinical development of diagnostic and therapeutic agents using ANXA2 as the molecular target.
Using the phage display technique, we identified a heptapeptide YW7 to specifically target the recombinant protein of ANXA2. The FITC-YW7 probe demonstrated a high selectivity in binding PANC-1 pancreatic cancer cells in vitro. We then conjugated YW7 with the NIRF dye Cy7 as the long wavelength of NIRF (700-900 nm) is desirable to achieve higher TBRs in deep tissue, meanwhile reducing autofluorescence [19]. We showed that the molecular fluorescent imaging probe Cy7-YW7 demonstrated strong PAAD targeting capabilities in vivo, making it a promising probe for enhancing the diagnostic efficiency of PAAD.

Cell culture and patient tissue samples
A total of four human pancreatic cancer cell lines (BxPC-3, SW1990, MIA PaCa-2, and PANC-1) were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). MIA PaCa-2 and PANC-1 cells were cultured in DMEM medium supplemented with 100 UÁmL À1 penicillin, 100 lgÁmL À1 streptomycin, and 10% FBS (all from Gibco, Carlsbad, CA, USA). Additionally, 2.5% horse serum was added to the medium of MIA PaCa-2 cells. BxPC-3 cells were cultured in RPMI-1640 medium (Gibco) with 100 U•mL À1 penicillin, 100 lg•mL À1 streptomycin, and 10% FBS. SW1990 cells were cultured in L-15 medium (Gibco) with 100 U•mL À1 penicillin, 100 lg•mL À1 streptomycin, and 10% FBS. The normal human pancreatic ductal cell line HPDE6-C7 was obtained from Jennio Biotech Co., Ltd (Guangzhou, China) and cultured in RPMI-1640 medium containing 10% FBS, 100 U•mL À1 penicillin, and 100 lg•mL À1 streptomycin. All cells were grown at 37°C with 5% CO 2 . Six pairs of human pancreatic cancer tissue samples were obtained from six patients who underwent surgery at the First Affiliated Hospital of Chinese Medical University from 2020. All patients who provided clinical specimens signed the written informed consent form. All the procedures were performed in accordance with the 1964 Declaration of Helsinki principles and its later amendments or comparable ethical standards. All protocols were approved by the Medical Ethics Committee of Chinese Medical University (Approval No. [2020]102).

Peptide synthesis
Solid-phase synthesis was performed to synthesize peptides and fluorescent dye-labeled peptides. FITC was labeled to the N-terminal of YW7 peptide via a spacer of 6-aminocaproic acid by Sangon Biotech Co., Ltd. (Shanghai, China). Cy7 was linked to the N terminal of YW7 peptide via spacer of PEG 3 by Chinese Peptide Co., Ltd. (Shanghai, China). YW7 peptide and fluorescent probes were characterized by MALDI-TOF mass spectrometry and reverse-phase high-performance liquid chromatography (RP-HPLC) and all products had a purity of ≥ 95%.

Bioinformatics analysis
The RNA-seq data expressed as FPKM and the related clinical information of PAAD were downloaded from The Cancer Genome Atlas (TCGA, http:// cancergenome.nih.gov/), and the FPKM data were transformed to transcripts per million (TPM) values for all the samples by using R version 3.6.3 software. Genotype-Tissue Expression (GTEx) database was obtained from UCSC Xena browser (https:// xenabrowser.net/datapages/). Mann-Whitney test was used to detect the differences in expression between tumor and normal tissues. The overall survival (OS) and progression-free survival (PFS) result was expressed as Kaplan-Meier (K-M) curve by "SURVMINER" and "SUR-VIVAL" R package. Univariate Cox regression analysis was implemented to evaluate prognostic analysis using R version 3.6.3 software. The nomogram was further established to predict cancer prognosis individually. The criterion for significance was P < 0.05 for all comparisons.

Animals
The 4-week-old male BALB/c nude mice were purchased from Beijing Hfk Bioscience Co., Ltd. (Beijing, China) and raised at the animal facility of China Medical University. Each mouse was subcutaneously injected with 1 9 107 PANC-1 cells into the right flanks. Mice were housed at 22 AE 1°C and kept on 12-h light/dark cycles, with 50% humidity. Distilled water and standard mouse chow diet were provided ad libitum. All animal studies were carried out in accordance with the guidelines published by the Institutional Animal Care and Use Committee of China Medical University (Shenyang, China). All protocols were approved by the Medical Ethics Committee of Chinese Medical University (Approval No.CMU2020320).

Western blotting analysis
ANXA2 expression in human cells and patients' tissues was determined by western blotting. Cells were lysed on ice in RIPA Lysis Buffer (Beyotime, Shanghai, China) with 1% PMSF for 30 min. Fresh tissues were suspended in RIPA Lysis Buffer with 1% PMSF, homogenized for 2 min, and then lysed on ice for another 30 min. The supernatant was collected and the protein concentration was determined after centrifuging for 10 min at 8000 g. The protein concentration was detected by BCA Protein Assay Kit (Beyotime). Proteins extracts were separated by 12% SDS/PAGE gel and then transferred to the PVDF membranes. Next, the membranes were blocked with 5% skimmed milk powder at room temperature for 1 h. Rabbit primary antibodies specific to ANXA2 (1 : 1000; Cell Signaling Technology, Billerica, CA, USA; #8235S) were used to incubate the membranes at 4°C overnight. Anti-b-tubulin mouse primary antibodies (1 : 5000; EarthOx Life Science, Millbrae, CA, USA; E021040-01) were applied as a loading control. On the following day, the membranes were incubated with the secondary antibodies conjugated with horseradish peroxidases at room temperature for 1 h. ECL reagents (Millipore, Billerica, CA, USA; WBULS0500) were applied to the membranes in the dark and protein bands were visualized by the chemiluminescence imaging system. The band density was quantified by IMAGEJ software (National Institutes of Health, Bethesda, MD, USA).

Recombination and purification of ANXA2
ANXA2 protein expression plasmid was constructed by inserting the coding sequence of GST-fused ANXA2 into pGEX-6P-1 plasmid by OBiO Technology Corp. Ltd., Shanghai, China. using the BamHI and XhoI enzymic sites. The recombined plasmid, pGEX-6P-ANXA2 or pGEX-6P-1, was then transfected into BL21 (DE3) E. coli. Isopropyl b-d-1-thiogalactopyranoside (IPTG) was added to the culture medium when OD600 of the bacteria culture reached 0.6 to induce ANXA2 expression. After induction, and the culture flask kept shaking for another 5 h, the bacteria were collected and lysed with lysozyme. Finally, the resulting supernatant was loaded to GST tag purification resin (Beyotime), and the resin was washed with PBST buffer six times before adding elution buffer and collecting elutions. Purity and product size were evaluated by SDS/PAGE and Coomassie blue staining.

Phage display screening toward ANXA2
The Ph.D.-7 phage display library (New England Biolabs, Beverly, MA, USA) was used to screen candidate peptides with high affinity toward ANXA2 but not GST. A quantity of 0.5 lg of GST or ANXA2 recombinant protein was immobilized on a 96-well plate overnight. 1% BSA was used to block nonspecific binding. In the first round of panning, phage solution was applied to wells coated with GST and incubated for 1 h with gentle shaking. Subsequently, unbound phages were collected by aspirating the remaining solution and then added to GST-fused ANXA2 recombinant protein by gentle shaking. The bound phages were eluted with 0.2 M glycine-HCl buffer (pH 2.2) of 100 lL, followed by neutralizing with Tris-HCl (pH 9.1). The collected phages were titrated on LBampicillin/IPTG/X-gal plates and amplified by culturing with E. coli (ER2758) and the amplified phages were used for another round of panning following the aforementioned procedures. In the third round of panning, eluted phages were cultured on an LB-ampicillin/ IPTG/X-gal plate and eventually, phages were randomly selected for DNA extraction and sequencing. DNA sequencing services were provided by Sangon Biotech Co., Ltd.

Microscale thermophoresis (MST) assay
The binding affinity of YW7 peptide to ANXA2 was measured by MST through Monolith NT.115 instrument (NanoTemper Technologies, Munich, Germany). In MST assay, YW7 was labeled with FITC and the concentration was kept unchanged. The unlabeled ANXA2 protein was diluted according to the gradient. The reaction solution was MST buffer containing 0.05% Tween-20. After a short time of binding reaction, the sample was loaded into NT.115 standard capillary. The resulting dissociation constant (K D ) and binding curve were calculated by the MO. Control Analysis software (NanoTemper).

Molecular operating environment (MOE) assay
Docking analysis of YW7 to ANXA2 protein was performed by MOE Version 2020.09 (Chemical Computing Group Inc., Montreal, Canada) according to compute energy scoring functions. The structure of ANXA2 (PDB:4HRE) was downloaded from the PDB database (https://www.rcsb.org/) and YW7 was generated from the Protein Builder Module of MOE. The spatial structure of YW7 was optimized by the Energy Minimize Function of MOE. Docking was conducted automatically by default parameters. The structure with the lowest free energy of binding was selected.

Evaluating YW7 binding to ANXA2 using affinity column and LC/MS detection
To prepare the ANXA2-immobilized affinity column, CNBr-activated Sepharose (GE Healthcare Bio-Sciences AB, Uppsala, Sweden) was swelled by 1 mM HCl and coupling buffer. Purified recombinant ANXA2 of 0.7 mg was loaded onto the resin and incubated at 4°C overnight. After blocking with ethanolamine, the resin was washed with 50 mM Tris-HCl (pH 8.0) and 50 mM Glycine-0.5 M NaCl (pH 4.0) buffers alternatively for three rounds. Then, the resin was washed with PBS and 1 mg YW7 dissolved in DEPC was added to the ANXA2-loaded resin. After binding for 1 h at 4°C, the resin was washed again with 50 mM Tris-HCl (pH 8.0) and 50 mM Glycine-0.5 M NaCl (pH 4.0) buffers. Eluents were eventually mixed with 50% acetonitrile and analyzed by UPLC-G2XS-QTOF/MS (Waters, Milford, MA, USA).

Characterizing the binding of FITC-YW7 to cells
We suspended cells in a concentration gradient of FITC-YW7 (0-1.5 lM) at 37°C for 30 min. Then, cells were washed with PBS three times and resuspended in PBS and finally transferred into a black 96well plate (1 9 10 5 per well). Fluorescence intensity of the cells was measured by a fluorescence multimode microplate reader (Tecan, Mnnedorf, Switzerland) at an excitation wavelength 490 nm and emission wavelength 518 nm. The apparent dissociation constant K d was calculated by fitting the data to nonlinear binding curves using the embedded function of GRAPHPAD PRISM 7.0 (GraphPad Software, Inc., San Diego, CA, USA).

SiRNA knockdown of ANXA2
Small interfering RNA (primer sequences: forward 5 0 -AGACCAAAGGUGUGGAUGAUU-3 0 and reverse 5 0 -UCAUCCACACCUUUGGUCUUU-3 0 ) was used to downregulate the expression of ANXA2 (si-ANXA2). A scrambled sequence siRNA was used as a negative control and a blank group was treated without siRNA. PANC-1 cells were seeded in a six-well plate with serum-free medium and treated with siR-NAs and lipofectamine 3000 reagent (Invitrogen, Carlsbad, CA, USA). At 48 h after transfection, cells were harvested, and total RNA and proteins were extracted. Knockdown efficiency was evaluated by western blotting and quantitative real-time PCR. Primer sequences were constructed by Sangon Biotech Co., Ltd.

Measurement of cellular uptake by flow cytometry and fluorescence imaging
PANC-1 and HPDE6-C7 cells were seeded in six-well plates at a density of 3 9 10 5 cellsÁmL À1 and incubated overnight. The cells were incubated with 25 lM FITC-YW7 for 30 min and 1 h for flow cytometry analysis of bound and internalized peptide, and cells were incubated with 1, 2, 4 lM FITC-YW7 or Cy7-YW7 for 1 h for fluorescence imaging of the live cells. Cells were washed with PBS three times and centrifuged for 5 min after 37°C incubation. Si-ANXA2or si-NC-transfected PANC-1 cells were detached with nonenzymatic dissociation solution from six-well plates. 20 lM FITC-YW7 was added to the suspended cells and incubated at 37°C for 15 min. Cells were washed with PBS and centrifuged for 5 min repeating three times. The supernatant was then discarded and the cells were resuspended in PBS. Percent of total cells with fluorescent signals was determined by flow cytometry using a FACSCelesta flow cytometer (BD Biosciences, San Jose, CA, USA). Fluorescence intensity was quantified by an imaging system (Berthold LB983, Bad Wildbad, Sttugart, Germany).

Immunofluorescence
Patients' pancreatic tumor tissues were embedded in OCT embedding medium, cut into 4 lM slices and frozen at À80°C. Cells were seeded in 24-well plates with glass coverslips and cultured overnight. The following day, frozen tissue sections were fixed with cold acetone, and cells were fixed with 4% paraformaldehyde. FITC-YW7, anti-ANXA2 primary antibody, TRITClabeled goat anti-rabbit IgG (1 : 200; EarthOx Life Science, E031320-02), and DAPI (Beyotime, C1002) were added to tissues or cells followed by blocking with 3% BSA and goat serum separately. Cells or tissue cells were visualized under a confocal laser scanning microscope (C2 plus, Nikon, Japan) at 408 nm (blue), 488 nm (green), and 543 nm (red).

In vivo and ex vivo imaging
Mice-bearing tumors about 6-8 mm in diameter (about 6 weeks after injection) were divided into two groups and Cy7 or Cy7-YW7 (8 nmol diluted in 100 lL) was injected intravenously into mice, respectively. Fluorescence signals of mice were acquired in a live imaging system (Berthold LB983 Night OWL II) using a red filter set (700-780 nm wavelength). During imaging, mice were anesthetized with 2% isoflurane. For ex vivo imaging, mice were sacrificed 3 h after probe injection, and the hearts, livers, spleens, lungs, kidneys, and tumors were collected and then imaged. Tumor tissues were fixed in 4% paraformaldehyde for further histological analysis. Fluorescence signals of different organs were measured by INDIGO software (Bad Wildbad, Stuttgart, Germany) and the binding specificity was gauged by TBRs.

Hematoxylin-eosin (H&E) and immunohistochemical (IHC) staining
Patients' tissues and tumor tissues from PANC-1-derived xenografts were fixed in 4% paraformaldehyde overnight at 4°C. The fixed tissues were dehydrated, embedded in paraffin, and cut into 4 lm slices. The slices were deparaffinized by heating at 65°C for 2 h. Heat-induced epitope retrieval was conducted using citrate buffer at 100°C for 20 min and 3% H 2 O 2 was used to remove the endogenous peroxidase. After blocking with goat serum for 30 min, tissue slices were incubated with anti-ANXA2 antibody overnight at 4°C. The next day, the tissue slices were incubated with the secondary antibody conjugated to horseradish peroxidase (HRP) at 37°C for 60 min and diaminobenzidine (DAB) for 3 min. Finally, hematoxylin and eosin (H&E) staining was done and images were obtained under a microscope (NIKON ECLIPSE E200, Tokyo, Japan).

Statistical analysis
In the present study, statistical analyzes and graphics were evaluated by GRAPHPAD PRISM 7.0 (GraphPad Software, Inc.), SPSS 19.0 software package (SPSS Inc., Chicago, IL, USA), and R software version 3.6.0. Comparation of the differences between groups was processed by Student's t-test and Mann-Whitney test. Survival analysis was estimated using the Kaplan-Meier method, and any differences in survival were evaluated with the log-rank test. The hazard ratio (HR) and 95% CI associated with the expressions of h-prune were estimated through a multivariable Cox regression model to assess the association between ANXA2 expression and OS or PFS. P < 0.05 was considered statistically significant. Quantitative data were expressed as mean AE SD.

Result
3.1. ANXA2 acts as a potential independent survival and prognostic factor in pancreatic adenocarcinoma We compared the expression levels of ANXA2 in 178 PAAD tissues of The Cancer Genome Atlas (TCGA) and 171 normal pancreatic tissues of GTEx database of TCGA; ANXA2 was expressed at higher levels in PAAD tissues than in normal pancreatic tissues (P < 0.0001, Fig. S1A). The survival analysis showed that high levels of ANXA2 expression were associated with poor OS (n = 88, P < 0.001) and PFS (n = 88, P < 0.001) in PAAD (Fig. S1B). We further applied the multivariable Cox regression model to investigate the independent prognostic force of the signature. Multivariable analysis results indicated that ANXA2 (HR = 1.001, 95% CI = 1.000-1.002; P = 0.007) have prognostic value for the OS, ANXA2 (HR = 1.001, 95% CI = 1.000-1.002; P = 0.001), residual tumor (HR = 2.311, 95% CI = 1.565-3.413; P < 0.001) and tumor position (HR = 2.613, 95% CI = 1.146-5.956; P < 0.022) had prognostic value for the PFS. These results demonstrated that ANXA2 is an independent prognostic factor in PAAD. Age, T, N, grade, stage, and tumor position and residual tumor are also potential-independent prognostic indicators of PAAD (Fig. S1C). We further constructed the nomograms based on ANXA2 and other independent prognosis factors (age, gender, grade, stage, and anatomic neoplasm subdivision), in order to demonstrate ANXA2 is important to predict the survival probabilities of PAAD patients. The expression of ANXA2 makes the greatest contribution to the prediction of 1-, 2-, and 3year OS, and calibration curves demonstrated that this nomogram was close to the actual survival duration indicating the superior predictive capacity (Fig. S1D).
To further investigate ANXA2 expressions in PAAD, patients' tissues from the First Affiliated Hospital of Chinese Medical University were collected.
Proteins in PAAD patients' tissues and PAAD cell lines (BXPC-3, SW1990, MIAPaCa-2, PANC-1) were detected by western blotting. As shown in Fig. 1A-C, PAAD tissues showed a markedly higher expression compared with their counterparts. The highest level of ANXA2 was observed in PANC-1 cells and BXPC-3 cells, and the normal pancreatic cells HPDE6-C7 exhibited a relatively lower level of ANXA2 expression (Fig. 1D,E). Localization and expression were confirmed by detecting fluorescent-labeled ANXA2 antibody in PAAD cells (Fig. 1F,G). Collectively, these data suggested that abnormally expressed ANXA2 may be a potential prognostic predictor in the highly malignant PAAD.

Identification of ANXA2-targeted heptapeptide YW7 by phage display
To screen for an ANXA2-targeted peptide, we first established a GST-ANXA2-inserted E. coli vector (pGEX-6P-1) as shown in Fig. 2A. The vector was transformed into BL21(DE3) bacteria, followed by IPTG induction at 22°C. We purified the GST-ANXA2 and GST protein using GST-tag affinity purification, and 12% SDS/ PAGE was performed to confirm the size (62 kD) and purity of the products (Fig. 2B). Then, phage display was carried out following steps in the flow chart as shown in Fig. 2C. In phage display, we used GST-ANXA2 as the target protein and GST as a negative binding protein, on which high-throughput screening of an M13 phage display library was performed. We achieved a high enrichment of ANXA2-binding phages after three rounds of screening (Fig. S2A). After DNA sequencing and amino acid translation, we obtained a peptide with the sequence of YWRGVYN among three randomly sequenced phage colonies, and we named the peptide YW7 (Fig. S2B). Peptide YW7 and FITC-labeled YW7 (purity > 95%) were synthesized by solid-phase synthesis for further confirmatory experiments (Fig. 2D).

Confirmation of the binding affinity of YW7 to ANXA2
To confirm ANXA2 is the binding target of YW7, MOE was utilized to perform preliminary characterization of the affinity of YW7 toward ANXA2. The structure of ANXA2 (PDB:4HRE) was downloaded from PDB database, and the structure of YW7 was optimized by the MOE-protein builder module. After the molecular docking, the binding mode of ANXA2 in complex with YW7 peptide with the lowest free energy was selected and further analyzed as shown in Fig. 3A. It can be seen that Ser64 and Gln138 of ANXA2 formed H-bond with Tyr1 and Trp2 residues of YW7; in addition, both Asp110 and Glu139 of ANXA2 formed an ionic bond and a H-bond with the Arg3 residues of YW7. These interactions proved that YW7 has a good affinity (SVM classifier = À48.59) to ANXA2. To further confirm the binding affinity of YW7, resin fixed with ANXA2 overnight was loaded to a column. Then, YW7 was added to the column and eluted after 1 h binding from the column and LC/ MS was utilized for the characterization of bound YW7 (Fig. 3B). We also tested the binding affinity between FITC-YW7 and ANXA2 by MST. MST data suggested FITC-YW7 binds to ANXA2 with a dissociation constant (K D , app) of 12 AE 0.69 lM (Fig. 3C).
To further verify that the binding between YW7 and ANXA2 on tumor cells, we further knocked down of ANXA2 on PANC-1 cells using small interfering RNAs (siRNAs). Downregulation of ANXA2 after knockdown was verified by both western blotting and quantitative real-time PCR analysis after knockdown (Fig. 3D-F). The fluorescence signal of FITC-YW7 was detected by flow cytometry and fluorescent microscopy after binding with PANC-1 cells. As expected, we found that PANC-1 cells with lower expression of ANXA2 were demonstrated a lower fluorescence signal intensity toward FITC-YW7 (Fig. 3G-I) and to verify that ANXA2 overexpression is essential for the binding of FITC-YW7 on PANC-1 cells. . P values were determined using paired Student's t-test. (C) Representative image of immunohistochemical staining of ANXA2 in PAAD patient tissue (tumor) and normal adjacent pancreas (normal) tissue (n = 2). Scale bar, 100 lm. (D, E) The relative protein level of ANXA2 was detected by western blotting in HPDE6-C7 and indicated PAAD cell lines (n = 3 for each cell). Scale bar, 20 lm. P values were determined using paired Student's t-test. (F, G) Localization of ANXA2 using confocal images (red, collected at 543 nm). All the data are represented as mean AE SD. *P < 0.05, ***P < 0.001.

Selectivity of FITC-YW7 on PANC-1 cells in vitro
In the above analysis of YW7, ANXA2 was verified to be the binding target of YW7 and to mediate the binding ability to PANC-1 cells. Thus, we further explored the binding selectivity of YW7, based on the abnormal expression of ANXA2 in tumor cells. We characterized the binding of FITC-YW7 to PANC-1 cells as shown in Fig. 4A. An apparent binding constant (K d ) of 86.62 nM was obtained for PANC-1 cells, but a lower K d of 502.2 nM for HPDE6-C7 cells. Next, the cellular uptake of FITC-YW7 was assessed by flow cytometry. After incubation with FITC-YW7 (25 lM) for 30 min and 1 h, approximately 90% of PANC-1 cells demonstrated a strong FITC signal, while only about 10% of HPDE6-C7 cells showed a strong FITC signal. Moreover, FITC-YW7 uptake in PANC-1 cells was shown to be in a time-dependent manner (Fig. 4B). Thereafter, we incubated PANC-1 cells with FITC-YW7 and anti-ANXA2 antibodies for confocal microscopy detection. Our data demonstrated that PANC-1 cells exhibited significantly stronger signals from FITC-YW7 and anti-ANXA2 antibodies, compared to HPDE6-C7 cells. Moreover, FITC-YW7 colocalized with anti-ANXA2 antibodies on PANC-1 cells (Fig. 4C,D). Similarly, FITC-YW7 showed greater binding on ANXA2 high expressed PAAD tumor than paired normal adjacent pancreas tissues (Fig. 4E,F). Scale bar, 20 lm. Data are represented as mean AE SD. P values were determined using paired Student's t-test. *P < 0.05, **P < 0.01, ***P < 0.001 versus blank; ## P < 0.01, ### P < 0.001 versus si-NC.

Specifically constructed Cy7-YW7 probe targeting PANC-1 in vitro and in vivo
In our study, we also constructed the Cy7-YW7 to assess the binding selectivity in vivo. First, constructing an NIRF probe target to ANXA2, Cy7 fluorescent dye was conjugated to the N terminal of YW7 using a short PEG 3 linker in between (Fig. 5A). The binding of FITC-YW7 or Cy7-YW7 to PANC-1 cells was observed using a fluorescent imaging system. We found that Cy7-YW7 (1, 2, 4 lM) were preferred to be internalized by PANC-1 cells in a dose-dependent manner. As a control, PANC-1 cells treated with Cy7 showed a significantly lower fluorescence signal compared to those treated with Cy7-YW7. However, in the FITC-YW7 binding assay, PANC-1 presented lower sensitivity compared to Cy7-YW7 (Fig. S3A,B). Cy7-YW7 diluted in PBS at final concentration of 4 lM was used to incubate with PANC-1 and HPDE6-C7. Our findings revealed that a stronger fluorescence signal of Cy7-YW7 was detected after internalization into PANC-1 cells compared to HPDE6-C7. Cy7 (4 lM) treated cells have little change in the fluorescence signal (Fig. S3C,D). The in vivo pancreatic  tumor targeting specificity of Cy7-YW7 was assessed on mice-bearing PANC-1 cell xenografts. Mice with tumors of 6-10 mm in diameter were intravenously injected Cy7-YW7 or Cy7 (8 nmol, n = 4 for each group), and images were acquired at 0, 1, 2, 2.5, 3 h after injection (Fig. 5B). Using those injected with Cy7 as controls, our data showed that stronger fluorescence signals in tumor tissues were observed in mice injected with Cy7-YW7 (Fig. 5C). The highest TBRs of mice injected with Cy7-YW7 occurred at 2.5 h after injection (TBR Cy7-YW7 = 2.7 AE 0.6, vs. TBR Cy7 = 1.5 AE 0.2, P < 0.05), and we examined the ex vivo tissue distribution of Cy7-YW7 at 2.5 h after injection (Fig. 5D). As shown in Fig. 5G, substantial retention of Cy7-YW7 was detected on tumor xenograft, a small amount of fluorescence signal on lung, kidney, liver, and little was on other tissues. Cy7 showed no significant difference in ex vivo distribution in major organs except that its tumor tissue accumulation was lower (Fig. 5E,F). ANXA2 immunostaining also suggested that ANXA2 is highly expressed in PANC-1 pancreatic tumor tissue (Fig. 5G).

Discussion
With the capability of noninvasively recognizing tumor biomarkers with high specificity, molecular imaging is a promising approach to improve the accuracy of tumor detection [20]. Early detection of PAAD is still a clinical challenge due to the lack of imaging biomarkers and the highly fibrotic and dense tumor stroma, which limits its accessibility by imaging agents [21]. Peptides, which are a class of therapeutic and targeting tumor ligands commonly used for constructing molecular imaging probes, have a high biocompatibility and cost-effectiveness, with their smaller size highly amenable for deep tissue penetration [22]. In this study, we employed the phage display technique to screen for a new heptapeptide specific to ANXA2, an oncoprotein overexpressed on the surface of tumor cells during tumor development, and consequently constructing a small-molecular NIRF probe with a high affinity and selectivity to pancreatic cancer. The choice of ANXA2 as a biomarker was based on previous reports that high ANXA2 levels are strongly correlated to the survival of pancreatic cancer patients, which was also verified by our bioinformatical study [12,23]. We showed that the heptapeptide YW7 has a high binding affinity toward ANXA2 (K D = 12 AE 0.69 lM). Based on YW7, we further developed an ANXA2-specific NIRF probe, Cy7-YW7, for PAAD detection. In our proof-of-concept study, injecting 8 nmol Cy7-YW7 resulted in rapid tumor accumulation in the mouse-bearing PANC-1 tumor xenografts, that is, significant tumor-to-background contrast could be seen as early as 1 h after injection and the highest TBRs were observed after 2.5 h. In contrast, Cy7 alone failed to induce enhanced tumor contrast. Thus, Cy7-YW7, a stable and sensitive NIRF probe, may be further recognized in clinical applications for the imaging of PAAD [24].
To our best knowledge, Cy7-YW7 is the first ANXA2-specific optical imaging probe that was discovered by phage display. Phage display technology is a traditional biopanning process but the most widely used method to biologically screen specific peptides. Therefore, the screening of peptides by phage display technology mimics the immune selection process by repeating and amplifying of specific binding phages [25]. Despite that phosphorylated ANXA2 expression is more specific to tumor compared to the unphosphorylated form, our findings show that both FITC-and Cy7-YW7 also possess a high specificity to PANC-1 cells. Importantly, Cy7-YW7 was proven to accumulate in tumor with a low background signal (TBRs~2.7). These data support our hypothesis that ANXA2-targeted peptides would selectively target tumor cells as ANXA2 is commonly overexpressed on tumor cell membranes. Compared to previously identified ANXA2-targeted peptide probes, BLZ-100 [9], which is based on a 36-amino acid peptide, and LS301, is based on an octapeptide [18], YW7 has a shorter amino acid sequence of only seven amino acids. Although YW7 and others share the same target of ANXA2, the fast clearance and high penetration of Cy7-YW7 due to the small size are likely other contributing factors of the high TBRs. In addition, as demonstrated in our study and previous studies, ANXA2 is also abundantly expressed in other types of cancers, making Cy7-YW7 a potentially valuable probe to detect other ANXA2-positive cancers.
Due to the delayed diagnosis of pancreatic cancer, surgery is still the most effective treatment for pancreatic cancer. However, the recurrence rate of PAAD still remains high (~80%) [2]. Our study paves the way for utilizing Cy7-YW7 to facilitate PAAD localization and delineating margin pre-, intra-, or postoperatively. Indeed, the most recently reported ANAX2targeted LS301 was shown to be capable of illuminating tumor edges due to its selective binding to the invasive front [18]. We expect that Cy7-YW7 has the potential to address the need to facilitate complete PAAD tumor resection by targeting ANXA2overexpressing tumor cells. Furthermore, our data also supported the YW7 internalization that followed binding to ANXA2 in pancreatic cancer cells, suggesting that YW7 has the potential to guide drugs to recognize tumor cells or even cross the cell membrane.
Compared with antibody counterparts, YW7 is of smaller size for more efficient tumor penetration. Given that endocytotic and non-endocytotic pathways are the main mechanism that peptides use to be internalized into cells, YW7 is likely to conduct natural ligand-receptor interactions and induce internalization through endocytosis. However, further data are needed to support this hypothesis. Meanwhile, the YW7 peptide could also serve as a valuable building block for synthesizing new MRI and PET probes, and enable early noninvasive detection of PAAD, which has not been feasible [26,27]. YW7 may also benefit the construction of novel PAAD-targeted theranostic agents, such as targeted nanoparticles (NPs), peptide-drug conjugates (PDC), targeted photodynamic agents for photodynamic therapy (PDT), where specific detection of PAAD through targeting ANXA2 would greatly enhance their translational potential.

Conclusions
Annexin A2 (ANXA2) is a cancer biomarker and is involved in the occurrence and development of a variety of tumors, particularly PAAD. We have screened an ANXA2-targeting heptapeptide, termed YW7, based on the phage display technique and found that FITC-YW7 selectively binds to pancreatic cancer cell PANC-1. Furthermore, we developed a novel NIRF probe Cy7-YW7. Compared to Cy7-treated group, Cy7-YW7 showed enhanced tumor sensitivity and specificity from 1 h after injection in mice-bearing PANC-1 cell xenografts. Our data suggested that YW7 is a new ANXA2-and potential PAAD-specific peptide. Cy7-YW7 is a novel NIRF probe that can facilitate PAAD tumor detection by specifically accumulating in tumor tissues in vivo. Cy7-YW7 is a promising molecular imaging probe for early detection and image-guided surgical resection for PAAD.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Survival prognostic value of ANXA2 in pancreatic cancer. (A) The differential expression level of ANXA2 expression between tumor and normal tissues on TCGA and GTEx cohorts, which were estimated by the Mann-Whitney test (P < 0.001). (B) Kaplan-Meier survival curves showed that patients with lower ANXA2 expression (n = 88) had longer OS (P < 0.001) and PFS (P < 0.001) than those with higher expression of ANXA2 (n = 88). P values were determined using a log-rank test. (C) Multivariate Cox regression of ANXA2 as a prognosis factor for OS and PFS of PAAD patients in TCGA. (D) Construction of a prognostic nomogram of PAAD patients by integrating the ANXA2 expression and other independent prognostic indicators (age, gender, grade, stage, and anatomic neoplasm subdivision). Calibration curves estimated the deviation of predicated 1-, 2-, and 3-year OS and actual survival duration.  Fig. S3. Fluorescent-labeled YW7 probes imaging in vitro. (A, B) Cell imaging after incubation with fluorescent-labeled YW7 using a live imaging system by a green filter set (490-525 nm) for FITC-YW7 or a red filter set (700-780 nm) for Cy7-YW7. PANC-1 cells were incubated with the same concentration gradient