Data-driven identification and semi-automated quantification of molecular targets for tumour-imaging of colorectal liver metastases and primary colorectal tumours

Abstract

Background: Accurate pre- and intraoperative assessment of disease extent is crucial for curative treatment of colorectal liver metastases (CRLM) and primary colorectal cancer (pCRC). Tumour-targeted nuclear imaging can enhance pre- and postoperative tumour staging, while tumour-specific fluorescence-guided surgery can improve intraoperative visualization. However, validated targets remain limited, particularly for CRLM. This study aimed to identify and validate novel molecular targets for CRLM using a data-driven approach. Additional objectives included pCRC target expression and evaluating target expression in neoadjuvant-treated patients.

Results: Using a data-driven RNA-based discovery approach (Euretos), candidate targets for colorectal liver metastases were identified. Of these, the six highest-ranking targets-CEACAM5, EPCAM, CEACAM6, MUC13, FXYD3, and CDH17-were selected for validation through immunohistochemistry (IHC). Semi-automated image analysis quantified IHC staining intensity (0-100) for tumour epithelium and background on a per-pixel basis. A patient's staining pattern was regarded as positive if mean tumour epithelium scored positive (>25), background negative (<25) and tumour epithelium >25 points higher than background, referred to as relative positive expression. The proportion of CRLM samples showing relative positive expression was 79% for CEACAM5, 45% for EPCAM, 80% for CEACAM6, 24% for MUC13, 58% for FXYD3, and 22% for CDH17. CEACAM5/CEACAM6 combined positivity reached 90% (either one positive), showing that targeting both markers enables molecular imaging in nearly the entire population. Staining intensities were similar in pCRC epithelium (P >0.05). Neoadjuvant-treated CRLM exhibited higher expression for all targets.

Conclusion: Using a novel data-driven approach, six potential imaging targets were successfully identified and validated. CEACAM5 and CEACAM6 emerged as strong targets that, regardless of neoadjuvant therapy, covered nearly the entire CRLM population-supporting their further probe development and clinical translation.

Keywords: Colorectal cancer; Colorectal liver metastases; Fluorescence-guided surgery; Molecular imaging; Neoadjuvant treatment; Nuclear imaging.

Fig. 1

Workflow and target selection for tumour-imaging in colorectal liver metastases and primary colorectal cancer. The Euretos ‘single target for cancer algorithm’ was used to identify genes with high TPM expression in colorectal cancer and low expression in liver parenchyma, with thresholds set at >50 TPM and <5 TPM, respectively. The Surfaceome gene set (4673 genes predicted to encode membranous proteins) served as input. The nine highest-ranked genes underwent further evaluation based on three additional target criteria. Of the remaining qualifying genes, an initial IHC analysis was performed, and those showing promising results proceeded to final IHC validation in the entire cohort. Results: Of the 136 genes meeting the TPM criteria, the nine highest ranked genes were selected. Of these, two (CLDN4, FAM3D) were excluded based on the additional selection criteria, and 1 (TSPAN1) was excluded after initial IHC screening. Six targets advanced to final IHC validation in a cohort of 62 CRLM and pCRC patients. Euretos = Data driven bio-informatic research platform. Single target for cancer algorithm = Tool within Euretos platform that allows for data-driven target selection. GTEx = Genotype-tissue expression database, TCGA = The cancer genome atlas, TPM = Transcripts per million (RNA), CRC = colorectal cancer. *Liver protein expression as assessed by Human Protein Atlas

Fig. 2

Representative immunohistochemical double staining of all six identified targets in colorectal liver metastases and primary colorectal carcinoma. Representative image of different staining intensity scores (0, 25, 50, 75 and 100) for all six targets, including a representative image of liver parenchyma (LP). Brown chromogen depicts extracellular matrix (targeting collagen I, collagen VI, elastin), blue (Vector Blue Substrate) depicts tumour epithelium. LP = liver parenchyma

Fig. 3

Target expression in colorectal liver metastases. (A) Staining intensity scores (0–100) of each target for CRLM tumour epithelium vs. liver parenchyma (LP) in boxplots. The X represents the mean. The dots represent a patient’s score (B) Percentage of patients with absolute positive staining for CRLM tumour epithelium and liver parenchyma separately. Absolute positive staining was defined as a staining intensity score >25, (C) Percentage of patients with relative positive staining. Relative positive staining was defined by three criteria: (1) positive staining in CRLM epithelium (>25), (2) negative/weak staining in liver parenchyma (<25), and (3) at least 25-point higher staining in CRLM epithelium compared to liver parenchyma. CRLM=colorectal liver metastases, LP=liver parenchyma

Fig. 4

Target expression in primary colorectal cancer. (A) Staining intensity scores (0–100) of each target for primary colorectal cancer (pCRC) tumour epithelium vs. mucosa, in boxplots. The X represents the mean. The dots represent a patient’s score (B) Percentage of patients with absolute positive staining for pCRC tumour epithelium and mucosa separately. Absolute positive staining was defined as a staining intensity score >25, (C) Percentage of patients with relative positive staining. Relative positivity was defined by three criteria: (1) positive staining in pCRC epithelium (>25), (2) negative/weak staining in mucosa (<25), and (3) at least 25-point higher staining in pCRC epithelium compared to mucosa. pCRC=Primary colorectal cancer

Fig. 5

Comparison of CRLM target expression in patients with and without neoadjuvant chemotherapy. (A) Staining intensity scores, (0–100) of each target for primary colorectal cancer, (pCRC) tumour epithelium vs. mucosa, in boxplots. The X represents the mean. The dots represent a patient’s score, (B) Percentage of patients with absolute positive staining for pCRC tumour epithelium and mucosa separately. Absolute positive staining was defined as a staining intensity score >25, (C) Percentage of patients with relative positive staining. Relative positivity was defined by three criteria:, (1) positive staining in pCRC epithelium, (>25), (2) negative/weak staining in mucosa, (<25), and, (3) at least 25-point higher staining in pCRC epithelium compared to mucosa. pCRC = Primary colorectal cancer

Fig. 6

Complementarity of targets in colorectal liver metastases. Per-patient relative staining outcome (positive or negative) for all six targets. As CEACAM6 had the highest percentage of CRLM with relative positive staining, it was taken as the reference. Relative positive staining was defined by three criteria: (1) positive staining in CRLM epithelium, (>25), (2) negative/weak staining in liver parenchyma, (<25), and, (3) at least 25-point higher staining in CRLM epithelium compared to liver parenchyma. NAT = neoadjuvant treated, CRLM = colorectal liver metastases