Programmed Death Ligand-1 expression in stage II colon cancer - experiences from a nationwide populationbased cohort

The population and sources of data has previously been descreibed in detail [17]. In brief all patients surgically treated for stage II CC in 2002 in Denmark were identified by a search in the nationwide registry administrated by the Danish Colorectal Cancer Group (DCCG) (N = 746). Exclusion criteria were as follows: missing tumour block (N = 11), insufficient tissue for analyses (N = 2), incorrectly staged patients (N = 25), treatment with adjuvant chemotherapy (N = 26)/radiotherapy (N = 1) and death within 90 days after the operation (N = 75). Furthermore patients with loco-advanced disease (N = 8) and patients diagnosed with another malignancy prior to CC were excluded from the study (N = 26), and the final study population comprised 572 patients.

Histopathological data were obtained by microscopic examination and from the national Patobank containing all pathology reports in Denmark. The term “not assessed” was used if the pathological feature was not described. Clinical data were obtained from The National Patient Registry.

Recurrence primarily occurred within the first five years of follow-up and to encompass the majority of recurrences a follow-up period of seven years was selected.

Formalin-fixed, paraffin-embedded tissue blocks were collected from the departments of pathology in Denmark. The tissue blocks were stored and transported at room temperature. One tumour block representing the deepest invasive margin, was selected from each patient. Prior to inclusion, all histological slides from each tumour were evaluated by first a trainee and afterwards an experienced pathologist.

From the selected tumour blocks serial 4 μm thick tissue sections were cut and mounted on FLEX IHC Microscope Slides (K8020, DAKO, Glostrup, Denmark). One whole tumour section per patient was used for the evaluation of PD-L1 expression. Staining was performed using a Ventana BenchMark ULTRA (Ventana Medical Systems, Tucson, Arizona, USA) automated immunohistochemistry (IHC) slide staining system. Tissue sections were heated and deparaffinised in EZprep (no.950–102, Ventana). Pre-treatment and demasking were carried out using ULTRA CC1 (no. 950–224) and ULTRA CC2 (no. 950–223), and endogenous peroxidase activity was blocked by Optiview Peroxidase Inhibitor (no. 760–700 Ventana). The slides were incubated with a rabbit monoclonal anti-PD-L1 (clone SP263A, no. 790–4905/741–4905 Ventana) for 16 min. This clone was chosen based on a pilot study. For amplification Optiview HQ Universal Linker (no. 760–700, Ventana) and Optiview HRP Multimer (no. 760–700, Ventana) was each used for 8 min.

The primary antibody was visualized using Optiview H₂O₂ and DAB (no. 760–700, Ventana), followed by Optiview Copper (no. 760–700, Ventana). Counterstain was done using Hematoxylin II (no. 790–2208, Ventana) and bluing Reagent (no. 760–2037, Ventana). Finally the histological slides were cover slipped with Tissue-Tek PERTEX (Histolab Products AB, Göteborg, Sweden).

Evaluation of microsattellite instability (MSI) was performed using IHC. Tumours displaying loss of one or more of the 4 mismatch repair proteins (MLH1, MSH2, MSH6 and PMS2) were considered as MSI, whereas tumours with intact mismatch repair proteins were considered as microsatellite stable (MSS). Staining of mismatch repair proteins was performed using a DAKO Autostainer Link 48 (DAKO) with monoclonal mouse antibody against MLH1 (Novocastra, Leica, Germany, clone ES05, dilution 1:100, product code NCL-L-MLH1), MSH2 (Novocastra, Leica, clone 25D12, dilution 1:100, product code NCL-L-MSH2), MSH6 (BD Transduction Laboratories, clone 44/MSH6, dilution 1:200, material number 610919), and PMS2 (BD Pharmingen, clone A16–4, dilution 1:500, material number 556415). Tissue sections were incubated for 30 min at room temperature with the primary antibodies diluted in Envision Flex antibody diluent (code S2022 DAKO). The antibody signal was amplified using Envision Flex+ Mouse(Linker) (DAKO) for 20 min. Bound antibodies were detected using Envision FLEX/HRP (DAKO) and visualized by Envision FLEX DAB (DAKO) and chromogene diluted in Envision Flex Substrate Buffer (DAKO). The sections were incubated in 0.5% CuSO₄ in TBS buffer pH 7.6 for 10 min to enhance the immunohistochemical staining. Sections were counterstained with Meyer’s hematoxylin (Merck, Damstadt, Germany).

Tumour PD-L1 expression was evaluated based on immunostaining of the cell membrane of the epithelial tumour cells. The immunostaining of the stromal cells were not evaluated. Tumour cells were considered positive when any cell membrane staining (partial of complete) was present. Staining intensity was not evaluated and cytoplasmatic immunoreaction was not considered.

The percentage of positive tumour cells were scored semi-quantitatively as 0 (no positive tumour cells), 1 (≤1% positive), 2 (> 1 and ≤ 5% positive), 3 (> 5 and ≤ 20% positive), 4 (> 20 and ≤ 50% positive) and 5 (> 50% positive) (Fig. 1). A subset of 50 randomly selected tumours was examined by a second pathologist in order to assess inter-observer variation. For prognostic evaluation data were dichotomized, using 5% PD-L1 expression as cut-off. In absence of a standardized scoring system the cut-off was based on previously studies [9, 12, 14].

The inter-observer reproducibility of PD-L1 scoring was evaluated by kappa statistics. Simple and weighted kappa (κ) values were calculated, and agreement was described according to Landis et al [18] as moderate, substantial, and almost perfect for κ values of 0.41–0.60, 0.61–0.80, and 0.81–1, respectively.

Chi²-statistics were used to test associations between clinicopathological variables. A p-value < 0.05 was considered significant. The statistical analyses were performed using STATA software version 14.0 (StataCorp, Texas, USA), and all statistical tests were two-sided.

Patient characteristics are summarized in Table 1. In the follow-up period of seven years, 266 (46.5%) patients died; 110 (19.2%) patients experienced disease recurrence and 78 (13.6%) patients were diagnosed with another cancer. The median age at time of surgery was 73 years (range 29–95), and the mean follow-up time was 6.9 years (range 3–84 months).

The IHC staining of PD-L1 often had a highly heterogeneous expression both between the central part of the tumour and the invasive margin and along the invasive tumour front (Fig. 2).

Results regarding PD-L1 expression are displayed in Table 2. Nearly half of the population (46%) had no PD-L1 expression in tumour cells. After dichotomization, using 5% PD-L1 expression as cut-off, 35 (6%) of the tumours were classified as high PD-L1. In the group of MSI 31 (18%) of the tumours were classified as high PD-L1 and in the subgroup of MSS 4 (1%) were classified as high PD-L1. High PD-L1 was related to female gender (p = 0.028), high malignancy grade (< 0.001), right side localization (p < 0.001) and MSI (p < 0.001).

The inter-observer agreement for the semi-quantitative evaluation of PD-L1 expression was moderate with κ = 0.418 and weighted κ = 0.573. The agreement improved to substantial, when categorizing the data into high PD-L1 expression (> 5%) or low PD-L1 expression (≤5%), κ = 0.691.

The 5-year RFS for the population with low PD-L1 was 69.2% versus 67.7% in the group with high PD-L1 expression, and OS was 74.7% versus 70.5%, respectively. When considering patients with MSI tumours the 5 year RFS for low PD-L1 was 77.4% versus 67.5% in the group of high PD-L1, and OS was 79.4% versus 70.7%. No significant differences in survival rates were observed, considering the entire cohort (Fig. 3). In the group of patients with MSI tumours the Kaplan Meier curves were separated for RFS, but results were insignificant, p = 0.256 (Fig. 4). This also accounted the group of patients with MSI T3 tumours (N = 155), p = 0.149.

Outcomes from the corresponding univariable Cox regression analyses are shown in Table 3. Patients with MSI tumours and high PD-L1 expression did not have a significant worse OS or RFS, HR = 1.104 (0.604–2.016), p = 0.748 and HR = 1.429 (0.769–2.653), p = 0.258, respectively. Age ≥ 73 years, T4 tumour and perforation were significantly related to an adverse outcome of both OS and RFS. Patients with MSI T3 tumours and high PD-L1 expression did neither have a significant worse OS or RFS, HR = 1.531 (0.757–3.100), p = 0.236 and HR = 1.637 (0.833–3.216), p = 0.153, respectively.