Immunohistochemistry

Tumor biopsies were retrieved from the biobank at Karolinska University Hospital. Protein expression was determined by immunohistochemistry on 4-μm-thick formalin-fixed, paraffin-embedded (FFPE) sections. The antibodies used in this study were Notch1 (Cell Signalling Technology, D1E11, Rabbit monoclonal antibody, dilution 1:200), Hes1 (Abcam, 71559, Rabbit polyclonal antibody, dilution 1:800), Ascl1 (Abcam, 206781, Mouse monoclonal antibody, dilution 1:75), and DLL3 (Ventana, 790–7016, Rabbit monoclonal antibody). The Notch1, Hes1, and Ascl1 antibodies were applied manually while the DLL3 staining was performed using an automated immunostaining instrument (VENTANA DLL3 SP347 Assay, Roche Diagnostics). The FFPE sections were de-paraffinized in xylene and rehydrated in alcohol. Antigen retrieval was performed using either citrate buffer pH 6 or EDTA buffer pH 9, depending on the antibody, at 97°C for 20 minutes. For the quenching process, a 30 minute incubation in 0,5% hydrogen peroxidase was performed at room temperature, followed by adding 1% bovine albumin (BSA) for 30 minutes in order to block unspecific antibody binding. The secondary antibodies used were BA-200 anti-mouse for Ascl1 and BA- 1000 anti-rabbit for Notch1 and Hes1 which both were used for 30 minutes at room temperature at a concentration of 1:200.

The next steps included a 30 minute incubation with avidin- biotin enzyme complex followed by a peroxidase substrate DAB, for three minutes.

The sections were counterstained in Mayer´s hematoxylin for one minute followed by dehydration with graded alcohols, xylene and coverslipped with Mountex [18]. For each case, one hematoxylin-eosin staining was performed and a positive/negative control for each protein was carried out according to the manufacturer’s instruction [14]. The evaluation of immunohistochemical stainings was performed by one pathologist (L.K), who was blinded to the clinical data. The number of positive tumor cells was counted under high magnification (x20 and x40) in three random and non-overlapping fields (100 tumor cells per field with a total of 300 tumor cells per specimen) [19].

The scoring of IHC staining in the cases were made into four categories according to the number of positive tumor cells stained; 0: No positive cells, 1; 1–25% positive cells, 2; 26–50% positive cells, 3; 51–75% cells, 4; more than 76% positive cells. The staining intensity was defined as any positivity in the tumor cells of each specimen [19, 20].

Since only DLL3 has a validated cut-off between low vs high expression, this study aimed to establish a cut-off for the other biomarkers based on sensitivity to PDCT using a dichotomizing score with receiver operating characteristic (ROC) curve analysis S5 Fig [21]. The hypothesis was that Notch1 and Hes1 high expression would be more “refractory” to PDCT, while Ascl1 high expression be more “sensitive” to PDCT [16].