Stromal fibronectin expression in patients with resected pancreatic ductal adenocarcinoma

Patients from this study were all diagnosed with PDAC and underwent pancreatectomy at the Department of Surgery, Skåne University Hospital, Lund and Malmö, Sweden. A total of 138 formalin-fixed, paraffin-embedded tissue samples were included. The study period spanned from 1996 to 2017. Hematoxylin and eosin stained tissues from all patients were re-evaluated by our pathologist (A.S.) in accordance with the WHO 2010 classification. As controls, disease-free pancreatic tissues from patients with serous (n = 3) or mucinous (n = 1) cystadenoma were included. The baseline characteristics of patients with PDAC are presented in Table 1. Ethical approval for this study was granted by the local human ethics committee at Lund University (Ref 2017/320). The study follows the REMARK guidelines where possible [20].

To minimize experimental variability and gain reproducibility, tissue microarray (TMA) technology was applied to the formalin-fixed paraffin-embedded specimens [21]. From each specimen, four sites of cancerous tissues with a diameter of 2 mm were obtained, which were marked by our pathologist (A.S.) and stabilized into paraffin blocks by an automated tissue array device (Minicore® 3, Alphelys, Plaisir, France). The established blocks based on TMA were then sliced into sections with a thickness of 3 μm for further immunohistochemical assessment. Each TMA slide contains around 120 cores, corresponding to samples from 30 patients with 4 replicates. Duplicated TMA slides underwent immunohistochemical staining (Fig. 1).

Immunohistochemistry was performed as previously described [22]. Briefly, TMA slides were firstly pre-warmed for 1 h at 60 °C. Secondly, slides were added to EnVision FLEX Target Retrieval Solution low pH (K800521–2, Dako, Glostrup, Denmark) heated to 96 °C for 20 min in an automated PT Link (Dako, Glostrup, Denmark). Then, slides were immersed in phosphate-buffered saline for 5 min, which was repeated twice. Subsequently, slides were immersed into phosphate-buffered saline containing hydrogen peroxide (0.3%) and methanol (1%) for 10 min. The sections were then incubated with 5% goat serum for 1 h at room temperature (RT). After careful removal of the liquid on the slides, successive incubation with avidin and biotin blocking kit (SP-2001, Vector Laboratories, Burlingame, CA, USA) was conducted for 15 min at RT, respectively. Next, the primary antibody, a rabbit recombinant monoclonal FN1 antibody (Abcam, Cambridge, UK; cat no ab2413; dilution 1:4000), was added on the slides. One slide was added with solvent without primary antibody for quality control. After making sure that all tissues were covered by the diluted antibody, samples were preserved in a refrigerator at 4 °C overnight. The next day, biotinylated secondary goat anti-rabbit antibody (BA-1000, dilution 1:200, Vector Laboratories) was applied on the slides at RT for 1 h. To amplify the target antigen signal, an avidin-biotin-peroxidase complex (Vectastain Elite ABC-HRP Kit, PK-6100, Vector Laboratories) was prepared according to the instructions of the manufacturer and used to immerse slides for 30 min at RT. Then, the specimens were covered by chromogen diaminobenzidine (SK-4100, Vector Laboratories) for 5 min, which was followed by deionized water immersion for 5 min. The slides were then immersed in Mayer’s hematoxylin (Histolab, Gothenburg, Sweden) for 30 s and quickly replaced in running tap water for 5 min. Lastly, the slides underwent routine dehydration in alcohol and xylen before mounting by Pertex (Histolab).

The reactivity of the FN1 antibody in samples was evaluated by our pathologist (A.S.), who was blinded to the survival information. The scoring algorithm was modified from Norihiro et al. [23] and takes the proportion of stained cells into consideration, as well as the intensity of the staining. The reactivity was scored in a semi-quantitative manner, which was categorized as negative if less than 10% staining was observed in the stroma; and mild, moderate, or strong based on the intensity if the percentage was > 10%. Low expression was defined as negative and mild reactivity, whereas high expression represented moderate or strong reactivity.

FN1 expression was evaluated in the tumor stroma component, localized to non-malignant fibroblasts and extracellular matrix. The epithelial tumor component was negative. Stromal FN1 expression was negative in 21 (15.3%) of tumors, while 66 (47.8%) tumors had mild FN1 expression, 44 (31.9%) had moderate expression, and 7 (5.1%) had strong FN1 expression. Figure 2 shows representative immunohistochemical images of FN1 expression in PDAC. FN1 was not stained in acinar cells and islets of Langerhans of disease-free control pancreatic tissues (data not shown). The public Human Protein Atlas database also shows absent or minimal expression of FN1 in normal pancreas (https://​www.​proteinatlas.​org/​ENSG00000115414-FN1/​tissue/​pancreas) [24].

When compared to the low FN1 expression group, the high FN1 expression group had significantly larger tumor size (P = 0.002), more advanced T stage (P = 0.039) and N stage (P = 0.009), and worse AJCC stage (54.0% vs 28.7% with stage III, P = 0.003) (Table 2). Furthermore, adjuvant chemotherapy was more common in patients with high FN1 expression as compared to the low expression group (92.0% vs 78.3%, P = 0.040). No association was observed between FN1 expression and age, gender, tumor location, tumor differentiation, and resection margin status (all P > 0.05).

Kaplan-Meier analysis showed that there were no differences in either disease-free survival (DFS) or overall survival (OS) when comparing high FN1 expression and low FN1 expression (median DFS 17. 9 vs 12.3 months; median OS 23.8 vs 24.5 months; both P > 0.05) (Fig. 3). By using Cox analyses, FN1 expression was not found to be associated with OS or DFS (P > 0.05, Tables 3 and 4). On multivariable analysis, only histological grade and resection margin status significantly correlated with DFS or OS. Notably, in our study, there were six patients without complete clinical data (Table 1). Re-analysis with exclusion of these six patients resulted in similar results and the same conclusion.