This research was carried out at the Department of Hepatobiliary and Pancreatic Surgery, People’s Hospital of Zhengzhou University (Zhengzhou, China). From August 2018 to September 2020, 60 PDAC patients receiving radical surgery were enrolled. The inclusion criteria were as follows: (a) radical resection confirmed by postoperative pathology; (b) PoV blood samples collecting was prior to specimen separation intraoperation; (c) definite pathological diagnosis of PDAC. The exclusion criteria were as follows: (a) previously treated with anticancer therapy before surgery, including radiotherapy, immunotherapy and chemotherapy; (b) patients with incomplete clinicopathological data; (c) patients with distant metastasis or others primary tumor. Ultimately, sixty patients met the inclusion criteria and were enrolled in this study. This study is consistent with the guidelines of the Declaration of Helsinki and approved by ethics committee of Henan Province People’s hospital. Informed consent was obtained from each participant.

Portal venous (PoV) blood sample (5ml) collection was performed prior to specimen separation during surgery ( Figure 1A ). Peripheral blood (PV) samples (5ml) were obtained from the cubital vein before surgery ( Figure 1B ). Then, these blood samples were placed in ethylenediaminetetraacetic acid (EDTA) vacutainers for the CTC phenotype detection.

The Canpatrol™ CTCs detection platform (Sur Exam, Guangzhou, China) had been introduced by previous reports (5, 17, 18). Firstly, CTCs were isolated by a filtration system containing a membrane with 8-μm diameter pores (Sur Exam, Guangzhou, China). The nuclei were stained by4',6-diamidino-2-phenylindole (DAPI). RNA-in situ hybridization technology was used to identify the CTC phenotypes targeting different mRNA sequences that encode epithelial biomarkers (EpCAM, CK8/18/19), mesenchymal biomarkers (Vimentin and Twist), and the leukocyte biomarker CD45. CTCs phenotypes were analyzed with a fluorescent microscope, the epithelial CTCs (E-CTCs) were stained with red fluorescence, mesenchymal CTCs (M-CTCs) were stained with green fluorescence and hybrid CTCs (H-CTCs) were stained with both red and green fluorescence. Besides, leukocytes were stained with white fluorescence ( Figure 1C ).

Clinicopathological data from 60 PDAC patients were collected, including age, sex, hepatitis, diabetes, preoperative serum CA19-9 level and postoperative serum CA19-9 level (referring to the first postoperative review result), tumor location (head and neck vs body and tail), degree of tumor differentiation (medium and poorly differentiated vs well differentiated), TNM stage (referring AJCC cancer staging manual, 8th edition) (19), nerve invasion, tumor size, surgical method, PoV and PV CTCs subtype counts.

The postoperative follow-up mainly focuses on the disease progression, including recurrence, metastasis, death until to January 2021. According to the guidelines of PCCA, patients are recommended to undergo a comprehensive examination for status assessment every 3 months in the first year, every 3 to 6 months during the second to third years, and then every 6 months (20). The length of the recurrence free survival (RFS)/metastasis free survival (MFS)/overall survival (OS) was measured from the date of surgery until the date of recurrence/metastasis/death occurring.

Postoperative tumor progression was stratified into two mutually exclusive categories: local recurrence and distant metastasis and they were judged mainly by imaging examination. If a patient had diagnosed with both local recurrence and distant metastasis at a follow-up, then he was classified into the metastasis group for metastatic lesions can represent a more lethal progression.

SPSS 21.0 and GraphPad Prism 8.0 software were used for data analysis. E-CTCs, M-CTCs, H-CTCs and T-CTC counts were used to investigate the prognosis assessment value and the correlation with clinicopathological parameters. Receiver operating characteristic curve (ROC) was used to determine the cut-off value of each CTCs subtype counts on different clinical outcomes. Then, we divided these patients with different CTC subtype count into different groups.

Continuous variables are presented as the median with inter-quartile range (IQR). CTCs count differences were compared by the nonparametric Mann–Whitney U test and/or the Kruskal-Wallis test. The Kaplan–Meier method with the log-rank test was used to assess the differences between the different groups. Univariate and multivariate factor analyses of prognosis-related factors were conducted using a Cox regression model to identify independent predictors. P<0.05 means statistically significant.

Sixty patients were enrolled in this study, including 34 men and 26 women with a median age of 56.5 (range: 30–83 years old). The number of patients in I stage, II stage, and III stage of TNM stage were 16, 35, 9 respectively. In addition, PV blood samples were collected from 32 patients among them. Although it was an invasive operation to obtain portal venous blood, we had not observed immediate or delayed complications from portal vein draws including hematoma formation or gastrointestinal bleeding. With a median follow‐up duration of 15 months (range: 3–29 months), 11 (18.3%) patients experienced local recurrence, and metastasis occurred in 30.0% (18/60), including 9 liver only metastases, 4 lung only metastases, 3 cases of multiple sites metastases, beside 3 patients had been diagnosed with both recurrence and metastatic. By the last follow-up, 17 patients had died due to tumor progression.

Thirty-two patients were implemented both PoV and PV CTCs phenotype tests, and we found that the T-CTC detectable rate in PV was 87.5% ( Figure 2A ), while the PoV T-CTC detectable positive ratio was high to 96.9% ( Figure 2B ). Paired comparison of CTCs phenotype counts in different vessels of 32 patients, we found the count of T-CTC and three kinds of subtypes of CTC in PoV were all significantly higher than in PV ( Figure 2C ) (P<0.05).

To further investigate the correlation between the PoV and PV CTC counts, we performed a linear regression analysis and found that the counts of T-CTCs, H-CTCs, E-CTCs per 5ml of PoV were significantly correlated with the count of PV (R² = 0.349, P < 0.001; R² = 0.335, P < 0.001; R² = 0.135, P = 0.039; respectively) ( Supplementary Figures 1A–C ), while no significant association was observed in the count of M-CTCs per 5 mL of PoV and PV sample (R² < 0.001, P = 0.973) ( Supplementary Figure 1D ).

During the follow-up (median months=15, range 3 to 29), five patients (15.6%) suffered recurrence, eleven patients (34.5%) presented with tumor metastasis, and eleven patients (34.5%) died from tumor-related causes. The median counts of T-CTCs, E-CTCs, H-CTCs and M-CTCs were 2, 0.5, 1, and 0, respectively.

To evaluate the prognostic value of the PV CTC phenotype count in PDAC, we compared the count differences of T-CTC and CTCs subtype among patients with different outcomes including recurrence, metastasis, death. We had not noticed a correlation between PV CTCs subtype counts with survival and metastasis events ( Figures 3B, C ). However, we found that recurrent patients had higher E-CTC counts than recurrence-free patients (P<0.05) ( Figure 3A ). ROC curve analysis exhibited that E-CTC≥2/5 ml means a higher risk of tumor recurrence (P=0.02) ( Figure 3D ), and Kaplan–Meier curve analysis shown that patients with high PV E-CTC (E-CTC ≥ 2/5 ml) counts had significantly shorter RFS than patients with low E-CTC counts (P=0.002) ( Figure 3E ).

Peripheral blood CTCs of 18 unresectable advanced PDAC patients were also collected before and after the first cycle chemotherapy and chemotherapy mainly adopts gemcitabine - based monotherapy or combination regimens. Comparing the PV CTCs subtype count difference in resectable and unresectable PDAC patients before chemotherapy, we noticed that unresectable patients had significantly higher H-CTCs than resectable patients (P=0.035) ( Supplementary Figure 2A ). Besides, Kaplan-Meier analysis shown that there was no significantly difference on overall survival in patients with increased or non-increased T-CTCs/E-CTCs/H-CTCs/M-CTCs ( Supplementary Figures 2B–E ) (P=0.915, P=0.149, P=0.505, P=0.164, respectively).

We obtained PoV blood samples for CTC phenotype test from 60 patients. Fifty-eight patients (96.7%) had detectable T-CTC in the PoV, and the median counts of T-CTCs, E-CTCs, H-CTCs and M-CTCs were 7, 2, 4, and 0, respectively. During the follow-up (median months=15, IQR 10.25 to 18), eleven patients (18.3%) presented with disease recurrence, eighteen patients (30%) presented with tumor metastasis, and seventeen patients (28.3%) died due to tumor-related causes.

The correlations between the clinicopathologic variables and the PoV CTC phenotype count of the 60 PDAC patients are listed in Table 1 . The Mann-Whitney U test was used to compare the correlation between CTCs subtype counts and the clinicopathologic variables and outcomes. In this research, we noticed a correlation of the T-CTC count with an abnormal postoperative CA19-9 level and tumor recurrence; a correlation of the E-CTC count with a postoperative abnormal CA19-9 level, T stage and tumor recurrence; a correlation of the H-CTC count with a postoperative abnormal CA19-9; a correlation of the M-CTC count with lymph node invasion, tumor differentiation degree and tumor metastasis (P<0.05) ( Table 1 ).

Based on the correlation between the clinical outcome and the CTC phenotype counts, ROC curves were used to determine the cut-off value of the CTC phenotype counts Figures 4A–D . The sensitivity, specificity and cut-off value are depicted in Table 2 . According to tumor recurrence or recurrence-free, the cut-off values of the T-CTC and E-CTC counts were 14/5 ml and 5/5 ml, respectively. The cut-off value of M-CTC to describe postoperative metastasis was 1/5 ml, and M-CTC≥1/5 ml was also an unfavourable factor of OS ( Figure 5 ).

Based on the cut-off value, PoV CTC subtype counts were divided into high or low groups. Then, Kaplan–Meier analysis was used to compare the prognostic differences among the CTC counts in different groups. The Kaplan–Meier curves showed that PoV T-CTC≥14/5 ml predicts a shorter RFS (P<0.05), PoV E-CTC≥5/5 ml also predicts a shorter RFS (P<0.05), while PoV M-CTC≥1/5 ml predicts a shorter MFS (P<0.05), and a shorter OS (P<0.05)

A scholar had also adapted the Canpatrol™ CTC platform, then divided the CTCs into five phenotype with different plastic and reversible phenotypes (21), referring this division criteria, the H-CTCs were further classified into three subtypes as epithelial predominant (E>M), intermediate (E≈M), or mesenchymal predominant (E<M) according to the signal numbers of epithelial or mesenchymal biomarker ( Figure 6A ). H-CTCs are the main component of CTCs in both PV and PoV; interestingly, E ≈ M is also the most abundant subtype of H-CTCs ( Figure 6B ). We noticed a correlation between PoV E>M H-CTCs with postoperative recurrence, besides we had not found correlations between the other CTCs phenotypes with postoperative metastasis and survival ( Figures 6C–E ). The ROC curve results showed that the cut-off value of E> M H-CTC counts for recurrence assessment was 3/5 ml ( Figure 6F ). The Kaplan–Meier curves demonstrated that E>M H-CTC≥3/5 ml means a shorter RFS ( Figure 6G ).

To identify the influence of baseline characteristics and particular PoV CTC phenotype count on the resectable PDAC postoperative prognosis. We noticed that postoperative CA19-9>37 U/ml, high T stage, PoV E>M H-CTC≥3/5 ml, PoV T-CTC≥14/5 ml, and PoV E-CTC≥5/5 ml were significantly associated with RFS (all P<0.05), while lymph node invasion and PoV M-CTC≥1/5 ml were significantly associated with MFS (all P<0.05) by univariate Cox regression. However, M-CTC≥1/5 ml and lymph node invasion were both associated with significantly shorter OS (all P<0.05) ( Table 3 ). Next, the multivariable analysis revealed that postoperative CA19-9>37 U/ml was a significant independent predictor of RFS (95CI 1.237–40.908; P=0.028); lymph node invasion (95%CI 1.286–19.546; P=0.020) and PoV M-CTC≥1/5 ml (95%CI 1.893–24.396; P=0.003) were significantly associated with shorter MFS; and PoV M-CTC≥ 1/5 ml (95%CI 1.03–9.27; P=0.043) was significantly associated with shorter OS ( Table 4 ).