Human PHCAA cells, QBC-939 and FRH-0201 were purchased from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China) and maintained in Dulbecco’s Modified Eagle medium supplemented with 10% exosome-depleted fetal bovine serum in a humidified atmosphere containing 5% CO₂.

Four-week-old BALB/c nude mice were purchased from GemPharmatech (Nanjing, China). A total of 3×10⁶ QBC939 cells transfected with an empty vector or the Cripto-1-OE vector, were subcutaneously injected into the flank of nude mice (24 mice for each group). Tumor diameters were measured every 2 days, and tumor volume was calculated as follows: volume= (length×width²)/2. After 2, 4 and 6 weeks, the mice were sacrificed, and their blood and solid tumors were collected. All procedures were approved by the Animal Management Committee of Shandong University.

We obtained the approval of the local ethical committees, and all subjects gave written informed consent to participate. Three cohorts of subjects from Qilu Hospital of Shandong University and Shandong Provincial Third Hospital were enrolled in this study. In the first cohort, pre-operation serum samples were collected from a group of 217 cases with HC (n=65), cholangitis (n=47) and PHCCA (n=115). The PHCCA cases were newly diagnosed and previously untreated; their tumors were confirmed by postoperative histopathological analyses. In the second cohort, sera and matched tissue was collected from 34 PHCCA patients. Moreover, postoperative sera were collected from 15 of the PHCCA cases in the second cohort. The third cohort was comprised of 105 PHCCA patients with available formalin-fixed tumor tissues. These patients underwent radical resection and were followed up regularly. PHCCA was staged according to the 7th AJCC/UICC tumor-node-metastasis (TNM) classification system.

Serum was separated by 2-step centrifugation (1,600g for 10 minutes followed by 16,000g for 10 minutes as previously described (22). Tissues were embedded in paraffin and cut into 4μm sections. For tissue microarray (TMA) construction, core tissues of 1.5 mm in diameter from each tumor block, as confirmed by hematoxylin and eosin staining, were re-embedded into the recipient TMA block as we described previously (2).

Exosomes were isolated from conditioned medium (CM) and serum using ultracentrifugation. CM was collected with centrifugation of 800g for 10 min and 2,000g for 10 min. Next, the CM supernatant, or cell-free serum diluted with D-PBS, was filtered through a 0.22μm filter and then ultracentrifuged at 120,000g for 70 min at 4°C. The pellets were washed with D-PBS and purified by further ultracentrifugation at 120,000 g for 70 min at 4°C. Exosomal markers CD81, TSG101 and Calnexin were detected by western blotting assay.

Cellular and exosomal lysates were prepared using RIPA buffer (Life Technologies) and quantified using a Pierce™ BCA Protein Assay Kit (Thermo Scientific). Proteins were separated by SDS-PAGE and transferred to PVDF membranes (Millipore). After blocking in 5% skimmed milk for 30 min, the membrane was incubated with primary antibodies overnight at 4°C. The following primary antibodies were used: CD81 (Abcam, ab79559, 1:1000 dilution), TSG101 (Abcam, ab83, 1:1000 dilution), Calnexin(Abcam,ab133615, 1:1000 dilution) and Cripto1 (Abcam, ab108391, 1:1000 dilution). Afterwards, an HRP-conjugated secondary antibody (Beyotime, Shanghai, China) was added at 1:5000 before incubation for 1h at room temperature. Chemiluminescence signal was detected using enhanced ECL Reagent (Vazyme, Nanjing, China) on the FluorChem E Chemiluminescent System (Cell Biosciences, Santa Clara, CA, USA).

The levels of exosomal Cripto-1 were detected using a Human Cripto-1 DuoSet ELISA kit (R&D) with some methodological alterations. In brief, 96 well-plates were coated with 100 μl of anti-CD81antibody (Abcam, ab79559, 1:250 dilution) per well, instead of the Capture Antibody in the kit, before incubation overnight at room temperature. After blocking plates with 300 μl/well Reagent Diluent for at least 1 h, 100 μl exosomes were added and incubated for 2 h at room temperature. Subsequently, we added 100 μl of the Detection Antibody before incubation for 2h, 100μl Streptavidin-HRP for 20 min, and then 100 μl Substrate Solution for 20 min. After blocking with 50 μl Stop Solution, the optical density (OD) of each well was determine immediately at 450 nm on an ELISA plate reader (Thermo Scientific, United States).

Serum CEA and CA19-9 levels were detected using an electrochemiluminescence method on the Cobas E601 Analyzer (Roche Diagnostics GmbH, Germany); the upper limits were defined as 5 ng/ml and 37 U/ml according to the manufacturer’s directions.

Slides were immersed in 3% hydrogen peroxide to inactivate endogenous peroxidase and incubated in EDTA buffer (pH = 9.0) for antigen retrieval. After blocking by 1% bovine serum albumin, sections were incubated with primary antibodies overnight at 4°C. Subsequently, the slides were incubated in secondary antibody prior to 3,3’-diaminobenzidine solution (Zsbio, Beijing, China), which enable visualization. The following primary antibodies were used: Cripto-1 Antibody (R&D, MAB2772, 1:100 dilution), Ki-67 (CST, #9027, 1:400 dilution), E-Cadherin (CST, #3195, 1:400 dilution). The IHC results were evaluated using the Quant Center software and calculated by a histochemistry score (H-score) system (23). The formula for the H-score was: H-score=(percentage of cells of weak intensity×1)+(percentage of cells of moderate intensity×2)+percentage of cells of strong intensity×3).

A Kruskal-Wallis test was employed for global comparison of Cripto-1, CEA or CA19-9 level among multiple groups, and further post hoc multiple comparisons were examined using a Mann–Whitney U test. A chi-square test was used for comparison of categorical variables. Logistic regression modelling was performed to combine biomarkers and generated predicted probability value. The correlation between Cripto-1 and E-cadherin expression was assessed by spearman correlation analysis. Survival curves were generated using the Kaplan-Meier method, and compared by a log-rank test. A Cox regression model was applied for identifying the independent prognostic factors. All above statistical analyses were performed by SPSS software, 22.0 and figures were prepared using GraphPad Prism, 9.1. The area under the curve (AUC) was calculated on the receiver operating characteristic (ROC) curve and compared using MedCalc 9.3.9.0. The cutoff point was determined based on the Youden index (sensitivity+specificity-1). Statistical significance was defined as two-sided P<0.05.

We isolated exosomes from the CM of two sets of PHCAA cells, QBC-939 and FRH-0201. Exosomes purified from PHCAA cells showed enrichment of exosomal markers, CD81 and TSG101, compared to total cell lysis, and had no Calnexin expression, an endoplasmic reticulum marker known to be absent in exosomes and present in cells lysis ( Figure 1A ). Western blotting analysis also showed Cripto-1 was expressed both in PHCAA cells and their exosomes ( Figure 1A ).

To observe whether Cripto-1 was released with exosomes in vivo, we subcutaneously injected QBC-939 cells into nude mice to establish a PHCCA mouse model. After 2 weeks, 4 weeks and 6 weeks post-injection the mice were bled, and tumors were harvested. As shown in Figure 1B , the concentration of serum exosomal Cripto-1 was gradually increased in mice xenografted QBC-939 cells (treated group), but not detected in normal mice (control group). Levels of exosomal Cripto-1 at 4 weeks and 6 weeks in the treated group were significantly higher than those in the control group and at 0 weeks in the treated group. IHC analysis showed the xenograft tumors had high Cripto-1 and ki-67 expression ( Figure 1C ), confirming exosomal Cripto-1 come from injected QBC-939 cells.

An exoELISA method was developed for detection of exosomal Cripto-1 in serum as shown in Figure 2A . Repeat measurements of 3 serum samples, from the same batch or different batches, resulted in an average intra-assay coefficient of variation (CV) and inter-assay CV of 4.27% and 5.52% ( Figure 2B ). Then, the serum samples were evaluated following repeated freezing and thawing, with different lengths of room temperature incubation; the results of this experiment showed no significant changes to the levels of exosomal Cripto-1 ( Figures 2C, D ).

By analyzing a large number of serum samples, measuring levels of exosomal Cripto-1 in serum showed a significant difference among HC, cholangitis and PHCCA groups ( Table 1 ). Moreover, Cripto-1 levels were independent of age and sex ( Table 1 ). Further pairwise comparison demonstrated exosomal Cripto-1 was significantly increased in the sera of the PHCCA group compared with the HC and cholangitis groups (both at P<0.001), whilst there was no significant difference between the HC and cholangitis groups ( Figure 3A ). In PHCCA patients, the levels of exosomal Cripto-1 increased as the TMN stage increased ( Figure 3B ). Further analysis showed exosomal Cripto-1 levels were significantly elevated in patients with lymph nodes metastasis ( Figure 3C ) or distant metastasis ( Figure 3D ), but not an advanced T stage ( Figure 3E ). Meanwhile, there were no significant differences observed when PHCCA cases were stratified by differentiation ( Figure 3F ), tumor size ( Figure 3G ) and neural invasion ( Figure 3H ). Cripto-1 showed no obvious association with age and sex ( Table S1 ). Correlations between levels of exosomal Cripto-1 and clinicopathological characteristics are shown in Table S1 .

ROC curve analyses realized that exosomal Cripto-1 was robust in discerning PHCCA patients (n=115) from cholangitis (n=47) and HC subjects (n=65), with an AUC of 0.87 4 ( Figure 4A ). When the Youden index reached a maximum from the ROC curve, the corresponding cutoff value of exosomal Cripto-1 for diagnosing PHCCA was 0.82, achieving a sensitivity of 79.1% and a specificity of 87.5%. Moreover, it was highly capable in discerning different TNM stage of PHCCA patients from cholangitis and HC subjects ( Figure S1 ). In contrast, exosomal Cripto-1 levels could not distinguish cholangitis cases from healthy controls ( Figure S2 ). And, exosomal Cripto-1 had limited differential diagnosis value for PHCCA with or without lymph nodes metastasis ( Figure S3 ).

To better understand the diagnostic performance of exosomal Cripto-1 for PHCCA, the levels of traditional serum tumor markers, CA19-9 and CEA, were assessed in all subjects. As shown in Table 1 , there were significant differences between the HC, cholangitis and PHCCA groups. Further, CA19-9 in the PHCCA group was increased compared to the cholangitis and HC groups (both at P< 0.001), while CEA was only significantly higher than patients in the cholangitis group (P< 0.01) ( Figure S4 ). The AUC values of CA19-9, CEA and the combined CA19-9 and CEA for PHCCA diagnosis were 0.773 ( Figure 4B ) and 0.596 ( Figure 4C ) and 0.773 ( Figure 4D ), much lower than that of exosomal Cripto-1. When the combined exosomal Cripto-1, CA19-9 and CEA levels were used, the AUC was 0.888( Figure 4E ), not significantly enhanced compared to exosomal Cripto-1 alone. Since exosomal Cripto-1 had no significant association with CA19-9 and CEA, and no significant difference was observed between CA19-9 or CEA positive and negative groups ( Figure S5 ), its diagnosis performance was assessed for CA19-9 or CEA negative individuals. As shown in Figure 4F , the exosomal Cripto-1 also had a high diagnostic capacity in subjects where diagnosis was missed in clinics. The diagnostic performance characteristics of exosomal Cripto-1, CA19-9 and CEA for PHCCA, such as sensitivity, specificity, positive likelihood ratio and negative likelihood ratio, are shown in Table 2 .

To explore the relationship between Cripto-1 in the exosomes of sera and tissues, 34 pairs of tumor tissues and sera were collected from PHCCA patients who underwent radical resection. Western blot analysis showed Cripto-1 was increased in PHCCA tissues compared to the corresponding normal bile duct tissues ( Figure 5A ). Then, the expression of Cripto-1 in 34 PHCCA tissue samples was detected by immunohistochemical staining and classified into low and high Cripto-1 expression groups, according to median of H-score ( Figure 5B ). The corresponding sera detected by exoEILSA showed that exosomal Cripto-1 levels in immunohistochemically Cripto-1-high cases, were significantly higher than in Cripto-1-low cases ( Figure 5C ). Among the aforementioned cases, 15 further cases were detected 1-week post surgery. The quantitative assay showed that after surgery, the median levels of exosomal Cripto-1 decreased from 1.25 (0.96-3.26) to 0.85(0.62-1.82) ( Figure 5D ).

Anti-Cripto-1 was used for immunohistochemical staining, in a retrospective cohort of 105 cases with PHCCA who had underwent radical resection. By evaluating the H-scores, the patients were divided into subgroups with low or high expression of Cripto-1( Figure 6A ). Data from Table S2 showed Cripto-1 expression significantly associated with lymph nodes metastasis, whilst there was no relationship with other clinicopathological factors, such as gender, age, differentiation, tumor size, T stage and neural invasion. As patients with high exosomal Cripto-1 expression in tissues or serum tended to have metastasis and EMT was a critical event involved in tumor metastasis (10, 11), the expression of E-cadherin, a known EMT marker, was assessed ( Figure 6B ). As shown in Figure 6C , the Cripto-1 expression negatively correlated with E-cadherin expression and patients with high Cripto-1 expression lowly expressed E-cadherin.

To investigate the prognostic properties of PHCCA, univariate analysis was performed using a Kaplan–Meier survival curve. As shown in Figure 6D , patients with high Cripto-1 expression had significantly lower survival rates than patients with low Cripto-1 expression. Moreover, patients with high Cripto-1 and low E-cadherin expression had the lowest probability of survival ( Figure 6E ). However, there were no significant survival differences between patients with low and high E-cadherin expression ( Figure S6 ). In addition, patients with advanced T stage or positive lymph nodes metastasis had poor overall survival ( Figure S7 ). Cripto-1 expression, along with the clinicopathological factors, was subjected to Cox model for multivariate analysis; except for T stage, Cripto-1 was identified as an independent prognostic biomarker ( Table 3 ).