Combined epithelial marker analysis of tumour budding in stage II colorectal cancer

Abstract

Tumour budding predicts survival of stage II colorectal cancer (CRC) and has been suggested to be associated with epithelial-to-mesenchymal transition (EMT). However, the underlying molecular changes of tumour budding remain poorly understood. Here, we performed multiplex immunohistochemistry (mIHC) to phenotypically profile tumours using known EMT-associated markers: E-cadherin (adherence junctions), integrin β4 (ITGB4; basement membrane), ZO-1 (tight junctions), and pan-cytokeratin. A subpopulation of patients showed high ITGB4 expression in tumour buds, and this coincided with a switch of ITGB4 localisation from the basal membrane of intact epithelium to the cytoplasm of budding cells. Digital image analysis demonstrated that tumour budding with high ITGB4 expression in tissue microarray (TMA) cores correlated with tumour budding assessed from haematoxylin and eosin (H&E) whole sections and independently predicted poor disease-specific survival in two independent stage II CRC cohorts (hazard ratio [HR] = 4.50 (95% confidence interval [CI] = 1.50-13.5), n = 232; HR = 3.52 (95% CI = 1.30-9.53), n = 72). Furthermore, digitally obtained ITGB4-high bud count in random TMA cores was better associated with survival outcome than visual tumour bud count in corresponding H&E-stained samples. In summary, the mIHC-based phenotypic profiling of human tumour tissue shows strong potential for the molecular characterisation of tumour biology and for the discovery of novel prognostic biomarkers.

Keywords: colorectal cancer; digital pathology; epithelial-to-mesenchymal transition; integrin β4; multiplex immunohistochemistry; prognostics; tumour budding.

Figure 1

Survival analysis of tumour budding from H&E‐stained whole sections assessed by visual scoring according to ITBCC (2016) guidelines. (A) Three‐tier Kaplan–Meier (log‐rank) plots for DSS, where Bd 1 = 0–4 buds (84.8%), Bd 2 = 5–9 buds (7.8%), and Bd 3 ≥ 10 buds (7.4% of patients). The values in brackets indicate the number of events/patients. (B) Dichotomisation of patients for low and high budding groups using optimal separation (high group ≥7 buds; 11% of patients). Cox univariate regression analysis; **p < 0.01; ***p < 0.001. Ref, reference.

Figure 2

mIHC staining for epithelial markers using a multi‐focal TMA demonstrates heterogeneity of ITGB4 expression. (A) Representative areas of tumour‐adjacent normal epithelium (benign), tumour centre, and tumour front were visually determined using H&E‐stained whole‐tissue sections, and 1.2‐mm punch cores were used to construct a TMA cohort of 232 patients. The TMA cohort sections were stained for simultaneous detection of PanCk (cytokeratins), ZO‐1 (apical tight junctions), ITGB4 (basal membrane), E‐cadherin (ECADH, adherens junctions), and DAPI (nuclei). High‐resolution five‐channel scanning (0.22 μm/pixel) was performed prior to visual observation and automated digital image analysis of the multiplex images. Scale bar = 2 mm. (B) Example images of tissue cores of normal (left), centre (middle), and front (right) areas of CRC samples, respectively. Scale bar = 200 μm. (C,D) Heterogeneity of ITGB4 expression: strong cytoplasmic ITGB4 (red) expression in invasive cell buds (C) and weak ITGB4 expression in cell buds (D). White arrowed lines and their corresponding intensity plots for E‐cadherin (ECADH, green) and ITGB4 (red) as measured along the line pixels to the direction of the arrow (intensity scale = 0–255) are shown as insets. Scale bar = 40 μm.

Figure 3

Automated digital image analysis for the segmentation of epithelial areas (clusters), buds, and ITGB4‐high buds. (A) A scheme of the epithelial segmentation using combined channels of PanCk, E‐cadherin (ECADH), and ITGB4. CellProfiler software was used for the automated detection of epithelial clusters and buds (clusters of 1–5 cells). ITGB4‐high buds were defined as buds with high ITGB4 expression (cut‐off = median ITGB4 across all the clusters; 26 intensity units; Scale = 0–255). See image analysis details in Section 2. The rectangles indicate blow‐up regions shown in the lower panels. Scale bar = 200 μm. (B) Channel intensities across all the epithelial clusters within normal epithelium (Normal Epi) (B) (n = 6774), tumour centre (C) (n = 17 272), and tumour front (F) cores (n = 19 369). Scale bar = 400 μm. (C) Median channel intensities within buds (n = 29 227) and non‐buds (n = 7985) of tumour cores. (D) Mean counts of buds and ITGB4‐high buds within different cores of the TMA cohort. (E) Median tight junction (TJ) perimeter in pixels within normal and tumour cores. (F) Rank‐weighted co‐localisation (RWC) analysis of epithelial markers within epithelial non‐buds and buds. The markers' pixel‐based localisations and their intensities were compared pairwise using CellProfiler. The error bars represent standard deviation except in D, where standard error of the mean was used. *p < 0.05; **p < 0.01; ***p < 0.001 (non‐parametric Mann–Whitney U test).