Morphological characterization of colorectal cancers in The Cancer Genome Atlas reveals distinct morphology-molecular associations: clinical and biological implications

An interpretable deep learning model for detecting BRCA pathogenic variants of breast cancer from hematoxylin and eosin-stained pathological images

Background

Determining the status of breast cancer susceptibility genes (BRCA) is crucial for guiding breast cancer treatment. Nevertheless, the need for BRCA genetic testing among breast cancer patients remains unmet due to high costs and limited resources. This study aimed to develop a Bi-directional Self-Attention Multiple Instance Learning (BiAMIL) algorithm to detect BRCA status from hematoxylin and eosin (H&E) pathological images.

Methods

A total of 319 histopathological slides from 254 breast cancer patients were included, comprising two dependent cohorts. Following image pre-processing, 633,484 tumor tiles from the training dataset were employed to train the self-developed deep-learning model. The performance of the network was evaluated in the internal and external test sets.

Results

BiAMIL achieved AUC values of 0.819 (95% CI [0.673-0.965]) in the internal test set, and 0.817 (95% CI [0.712-0.923]) in the external test set. To explore the relationship between BRCA status and interpretable morphological features in pathological images, we utilized Class Activation Mapping (CAM) technique and cluster analysis to investigate the connections between BRCA gene mutation status and tissue and cell features. Significantly, we observed that tumor-infiltrating lymphocytes and the morphological characteristics of tumor cells appeared to be potential features associated with BRCA status.

Conclusions

An interpretable deep neural network model based on the attention mechanism was developed to predict the BRCA status in breast cancer. Keywords: Breast cancer, BRCA, deep learning, self-attention, interpretability.

Clinicopathological and immune characterization of mismatch repair deficient endocervical adenocarcinoma

Endocervical adenocarcinoma (ECA) is reported increasingly often in young women, and this aggressive disease lacks effective methods of targeted therapy. Since mismatch repair deficiency (dMMR) is an important biomarker for predicting response to immune checkpoint inhibitors, it is important to investigate the clinicopathological features and immune microenvironment of dMMR ECAs. We assessed 617 ECAs from representative tissue microarray sections, gathered clinicopathologic information, reviewed histological characteristics, and performed immunohistochemical staining for MMR, programmed cell death 1 (PD-L1), and other immune markers. Of 617 ECA samples, 20 (3.2%) cases had dMMR. Among them, loss of MMR-related proteins expression was observed in 17/562 (3.0%) human papilloma virus-associated (HPVA) adenocarcinoma and 3/55 (5.5%) non-HPV-associated (NHPVA) adenocarcinoma. In NHPVA cohort, dMMR status was observed in 3 (3/14, 15.0%) patients with clear cells. dMMR ECAs had a higher tendency to have a family history of cancer, larger tumor size, p16 negative, HPV E6/E7 mRNA in situ hybridization (HPV E6/E7 RNAscope) negative, and lower ki-67 index. Among the morphological variables evaluated, poor differentiation, necrosis, stromal tumor-infiltrating lymphocytes, peritumoral lymphocytes, and lymphoid follicles were easily recognized in the dMMR ECAs. In addition, dMMR ECAs had higher CD3+, CD8+, CD38+, CD68+ and PD-1+ immune cells. A relatively high prevalence of PD-L1 expression was observed in dMMR ECAs. dMMR ECAs were significantly more likely to present with a tumor-infiltrating lymphocytes -high/PD-L1-positive status. In conclusion, dMMR ECAs have some specific morphological features and a critical impact on the immune microenvironment, which may provide insights into improving responses to immunotherapy-included comprehensive treatment for ECAs in the future.

Colorectal medullary carcinoma: a pathological subtype with intense immune response and potential to benefit from immune checkpoint inhibitors

INTRODUCTION

Different pathological types of colorectal cancer have distinguished immune landscape, and the efficacy of immunotherapy will be completely different. Colorectal medullary carcinoma, accounting for 2.2-3.2%, is characterized by massive lymphocyte infiltration. However, the attention to the immune characteristics of colorectal medullary carcinoma is insufficient.

AREA COVERED

We searched the literature about colorectal medullary carcinoma on PubMed through November 2023to investigate the hallmarks of colorectal medullary carcinoma's immune landscape, compare medullary carcinoma originating from different organs and provide theoretical evidence for precise treatment, including applying immunotherapy and BRAF inhibitors.

EXPERT OPINION

Colorectal medullary carcinoma is a pathological subtype with intense immune response, with six immune characteristics and has the potential to benefit from immunotherapy. Mismatch repair deficiency, ARID1A missing and BRAF V600E mutation often occurs. IFN-γ pathway is activated and PD-L1 expression is increased. Abundant lymphocyte infiltration performs tumor killing function. In addition, BRAF mutation plays an important role in the occurrence and development, and we can consider the combination of BRAF inhibitors and immunotherapy in patients with BRAF mutant. The exploration of colorectal medullary carcinoma will arouse researchers' attention to the correlation between pathological subtypes and immune response, and promote the process of precise immunotherapy.

Artificial Intelligence Models for the Detection of Microsatellite Instability from Whole-Slide Imaging of Colorectal Cancer

With the advent of whole-slide imaging (WSI), a technology that can digitally scan whole slides in high resolution, pathology is undergoing a digital revolution. Detecting microsatellite instability (MSI) in colorectal cancer is crucial for proper treatment, as it identifies patients responsible for immunotherapy. Even though universal testing for MSI is recommended, particularly in patients affected by colorectal cancer (CRC), many patients remain untested, and they reside mainly in low-income countries. A critical need exists for accessible, low-cost tools to perform MSI pre-screening. Here, the potential predictive role of the most relevant artificial intelligence-driven models in predicting microsatellite instability directly from histology alone is discussed, focusing on CRC. The role of deep learning (DL) models in identifying the MSI status is here analyzed in the most relevant studies reporting the development of algorithms trained to this end. The most important performance and the most relevant deficiencies are discussed for every AI method. The models proposed for algorithm sharing among multiple research and clinical centers, including federal learning (FL) and swarm learning (SL), are reported. According to all the studies reported here, AI models are valuable tools for predicting MSI status on WSI alone in CRC. The use of digitized H&E-stained sections and a trained algorithm allow the extraction of relevant molecular information, such as MSI status, in a short time and at a low cost. The possible advantages related to introducing DL methods in routine surgical pathology are underlined here, and the acceleration of the digital transformation of pathology departments and services is recommended.

SMAD4 depletion contributes to endocrine resistance by integrating ER and ERBB signaling in HR + HER2- breast cancer

Endocrine resistance poses a significant clinical challenge for patients with hormone receptor-positive and human epithelial growth factor receptor 2-negative (HR + HER2-) breast cancer. Dysregulation of estrogen receptor (ER) and ERBB signaling pathways is implicated in resistance development; however, the integration of these pathways remains unclear. While SMAD4 is known to play diverse roles in tumorigenesis, its involvement in endocrine resistance is poorly understood. Here, we investigate the role of SMAD4 in acquired endocrine resistance in HR + HER2- breast cancer. Genome-wide CRISPR screening identifies SMAD4 as a regulator of 4-hydroxytamoxifen (OHT) sensitivity in T47D cells. Clinical data analysis reveals downregulated SMAD4 expression in breast cancer tissues, correlating with poor prognosis. Following endocrine therapy, SMAD4 expression is further suppressed. Functional studies demonstrate that SMAD4 depletion induces endocrine resistance in vitro and in vivo by enhancing ER and ERBB signaling. Concomitant inhibition of ER and ERBB signaling leads to aberrant autophagy activation. Simultaneous inhibition of ER, ERBB, and autophagy pathways synergistically impacts SMAD4-depleted cells. Our findings unveil a mechanism whereby endocrine therapy-induced SMAD4 downregulation drives acquired resistance by integrating ER and ERBB signaling and suggest a rational treatment strategy for endocrine-resistant HR + HER2- breast cancer patients.

Deep learning for dual detection of microsatellite instability and POLE mutations in colorectal cancer histopathology

In the spectrum of colorectal tumors, microsatellite-stable (MSS) tumors with DNA polymerase ε (POLE) mutations exhibit a hypermutated profile, holding the potential to respond to immunotherapy similarly to their microsatellite-instable (MSI) counterparts. Yet, due to their rarity and the associated testing costs, systematic screening for these mutations is not commonly pursued. Notably, the histopathological phenotype resulting from POLE mutations is theorized to resemble that of MSI. This resemblance not only could facilitate their detection by a transformer-based Deep Learning (DL) system trained on MSI pathology slides, but also indicates the possibility for MSS patients with POLE mutations to access enhanced treatment options, which might otherwise be overlooked. To harness this potential, we trained a Deep Learning classifier on a large dataset with the ground truth for microsatellite status and subsequently validated its capabilities for MSI and POLE detection across three external cohorts. Our model accurately identified MSI status in both the internal and external resection cohorts using pathology images alone. Notably, with a classification threshold of 0.5, over 75% of POLE driver mutant patients in the external resection cohorts were flagged as "positive" by a DL system trained on MSI status. In a clinical setting, deploying this DL model as a preliminary screening tool could facilitate the efficient identification of clinically relevant MSI and POLE mutations in colorectal tumors, in one go.

Deep learning framework for comprehensive molecular and prognostic stratifications of triple-negative breast cancer

Triple-negative breast cancer (TNBC) is the most challenging breast cancer subtype. Molecular stratification and target therapy bring clinical benefit for TNBC patients, but it is difficult to implement comprehensive molecular testing in clinical practice. Here, using our multi-omics TNBC cohort (N = 425), a deep learning-based framework was devised and validated for comprehensive predictions of molecular features, subtypes and prognosis from pathological whole slide images. The framework first incorporated a neural network to decompose the tissue on WSIs, followed by a second one which was trained based on certain tissue types for predicting different targets. Multi-omics molecular features were analyzed including somatic mutations, copy number alterations, germline mutations, biological pathway activities, metabolomics features and immunotherapy biomarkers. It was shown that the molecular features with therapeutic implications can be predicted including the somatic PIK3CA mutation, germline BRCA2 mutation and PD-L1 protein expression (area under the curve [AUC]: 0.78, 0.79 and 0.74 respectively). The molecular subtypes of TNBC can be identified (AUC: 0.84, 0.85, 0.93 and 0.73 for the basal-like immune-suppressed, immunomodulatory, luminal androgen receptor, and mesenchymal-like subtypes respectively) and their distinctive morphological patterns were revealed, which provided novel insights into the heterogeneity of TNBC. A neural network integrating image features and clinical covariates stratified patients into groups with different survival outcomes (log-rank P < 0.001). Our prediction framework and neural network models were externally validated on the TNBC cases from TCGA (N = 143) and appeared robust to the changes in patient population. For potential clinical translation, we built a novel online platform, where we modularized and deployed our framework along with the validated models. It can realize real-time one-stop prediction for new cases. In summary, using only pathological WSIs, our proposed framework can enable comprehensive stratifications of TNBC patients and provide valuable information for therapeutic decision-making. It had the potential to be clinically implemented and promote the personalized management of TNBC.

Correlation between NGS panel-based mutation results and clinical information in colorectal cancer patients

Early mutation identification guides patients with colorectal cancer (CRC) toward targeted therapies. In the present study, 414 patients with CRC were enrolled, and amplicon-based targeted next-generation sequencing (NGS) was then performed to detect genomic alterations within the 73 cancer-related genes in the OncoAim panel. The overall mutation rate was 91.5 % (379/414). Gene mutations were detected in 38/73 genes tested. The most frequently mutated genes were TP53 (60.9 %), KRAS (46.6 %), APC (30.4 %), PIK3CA (15.9 %), FBXW7 (8.2 %), SMAD4 (6.8 %), BRAF (6.5 %), and NRAS (3.9 %). Compared with the wild type, TP53 mutations were associated with low microsatellite instability/microsatellite stability (MSI-L/MSS) (P = 0.007), tumor location (P = 0.043), and histological grade (P = 0.0009); KRAS mutations were associated with female gender (P = 0.026), distant metastasis (P = 0.023), TNM stage (P = 0.013), and histological grade (P = 0.004); APC mutations were associated with patients <64 years of age at diagnosis (P = 0.04); PIK3CA mutations were associated with tumor location (P = 4.97e-06) and female gender (P = 0.018); SMAD4 mutations were associated with tumor location (P = 0.033); BRAF mutations were associated with high MSI (MSI-H; P = 6.968e-07), tumor location (P = 1.58e-06), and histological grade (P = 0.04). Mutations in 164 individuals were found to be pathogenic or likely pathogenic. A total of 26 patients harbored MSI-H tumors and they all had at least one detected gene mutation. Mutated genes were enriched in signaling pathways associated with CRC. The present findings have important implications for improving the personalized treatment of patients with CRC in China.

Self-Supervised Learning Reveals Clinically Relevant Histomorphological Patterns for Therapeutic Strategies in Colon Cancer

Self-supervised learning (SSL) automates the extraction and interpretation of histopathology features on unannotated hematoxylin-and-eosin-stained whole-slide images (WSIs). We trained an SSL Barlow Twins-encoder on 435 TCGA colon adenocarcinoma WSIs to extract features from small image patches. Leiden community detection then grouped tiles into histomorphological phenotype clusters (HPCs). HPC reproducibility and predictive ability for overall survival was confirmed in an independent clinical trial cohort (N=1213 WSIs). This unbiased atlas resulted in 47 HPCs displaying unique and sharing clinically significant histomorphological traits, highlighting tissue type, quantity, and architecture, especially in the context of tumor stroma. Through in-depth analysis of these HPCs, including immune landscape and gene set enrichment analysis, and association to clinical outcomes, we shed light on the factors influencing survival and responses to treatments like standard adjuvant chemotherapy and experimental therapies. Further exploration of HPCs may unveil new insights and aid decision-making and personalized treatments for colon cancer patients.

Novel biomarker SARIFA in colorectal cancer: highly prognostic, not genetically driven and histologic indicator of a distinct tumor biology

SARIFA (Stroma AReactive Invasion Front Areas) has recently emerged as a promising histopathological biomarker for colon and gastric cancer. To elucidate the underlying tumor biology, we assessed SARIFA-status in tissue specimens from The-Cancer-Genome-Atlas (TCGA) cohorts COAD (colonic adenocarcinoma) and READ (rectal adenocarcinoma). For the final analysis, 207 CRC patients could be included, consisting of 69 SARIFA-positive and 138 SARIFA-negative cases. In this external validation cohort, H&E-based SARIFA-positivity was strongly correlated with unfavorable overall, disease-specific, and progression-free survival, partly outperforming conventional prognostic factors. SARIFA-positivity was not associated with known high-risk genetic profiles, such as BRAF V600E mutations or microsatellite-stable status. Transcriptionally, SARIFA-positive CRCs exhibited an overlap with CRC consensus molecular subtypes CMS1 and CMS4, along with distinct differential gene expression patterns, linked to lipid metabolism and increased stromal cell infiltration scores (SIIS). Gene-expression-based drug sensitivity prediction revealed a differential treatment response in SARIFA-positive CRCs. In conclusion, SARIFA represents the H&E-based counterpart of an aggressive tumor biology, demonstrating a partial overlap with CMS1/4 and also adding a further biological layer related to lipid metabolism. Our findings underscore SARIFA-status as an ideal biomarker for refined patient stratification and novel drug developments, particularly given its cost-effective assessment based on routinely available H&E slides.

CoNIC Challenge: Pushing the frontiers of nuclear detection, segmentation, classification and counting

Nuclear detection, segmentation and morphometric profiling are essential in helping us further understand the relationship between histology and patient outcome. To drive innovation in this area, we setup a community-wide challenge using the largest available dataset of its kind to assess nuclear segmentation and cellular composition. Our challenge, named CoNIC, stimulated the development of reproducible algorithms for cellular recognition with real-time result inspection on public leaderboards. We conducted an extensive post-challenge analysis based on the top-performing models using 1,658 whole-slide images of colon tissue. With around 700 million detected nuclei per model, associated features were used for dysplasia grading and survival analysis, where we demonstrated that the challenge's improvement over the previous state-of-the-art led to significant boosts in downstream performance. Our findings also suggest that eosinophils and neutrophils play an important role in the tumour microevironment. We release challenge models and WSI-level results to foster the development of further methods for biomarker discovery.

Pathologic Evaluation of Therapeutic Biomarkers in Colorectal Adenocarcinoma

Molecular testing is an essential component of the pathologic evaluation of colorectal carcinoma providing diagnostic, prognostic, and predictive therapeutic information. Mismatch repair status evaluation is required for all tumors. Advanced and metastatic tumors also require determination of tumor mutational burden, KRAS, NRAS, and BRAF mutation status, ERBB2 amplification status, and NTRK and RET gene rearrangement status to guide therapy. Multiple assays, including immunohistochemistry, microsatellite instability testing, MLH1 promoter methylation, and next-generation sequencing, are typically needed. Pathologists must be aware of these requirements to optimally triage tissue. Advances in colorectal cancer molecular diagnostics will continue to drive refinements in colorectal cancer personalized therapy.

The Role of Colonoscopy in the Management of Individuals with Lynch Syndrome: A Narrative Review

The history of Lynch syndrome changed definitively in 2000, when a study published in Gastroenterology demonstrated a significant reduction in mortality among individuals with Lynch syndrome who undergo regular endoscopic surveillance. As a consequence of this clinical evidence, all scientific societies developed guidelines, which highlighted the role of colonoscopy in the management of Lynch syndrome, especially for individuals at high risk of colorectal cancer. Over the years, these guidelines were modified and updated. Specialized networks were developed in order to standardize endoscopic surveillance programs and evaluate all the clinical data retrieved by the results of colonoscopies performed for both the screening and the surveillance of individuals with Lynch syndrome. Recent data show that the impact of colonoscopy (with polypectomy) on the prevention of colorectal cancer in individuals with Lynch syndrome is less significant than previously thought. This narrative review summarizes the current discussion, the hypotheses elaborated and the algorithms depicted for the management of individuals with Lynch Syndrome on the basis of the recent data published in the literature.

Histopathological bladder cancer gene mutation prediction with hierarchical deep multiple-instance learning

Gene mutation detection is usually carried out by molecular biological methods, which is expensive and has a long-time cycle. In contrast, pathological images are ubiquitous. If clinically significant gene mutations can be predicted only through pathological images, it will greatly promote the widespread use of gene mutation detection in clinical practice. However, current gene mutation prediction methods based on pathological images are ineffective because of the inability to identify mutated regions in gigapixel Whole Slide Image (WSI). To address this challenge, hereby we propose a carefully designed framework for WSI-based gene mutation prediction, which consists of three parts. (i) The first part of cancerous area segmentation, based on supervised learning, quickly filters out a large number of non-mutated regions; (ii) the second part of cancerous patch clustering, based on the representations derived from contrastive learning, ensures the comprehensiveness of patch selection; and (iii) the third part of mutation classification, based on the proposed hierarchical deep multi-instance learning method (HDMIL), ensures that sufficient patches are considered and inaccurate selections are ignored. In addition, benefiting from a two-stage attention mechanism in HDMIL, the patches that are highly correlated with gene mutations can be identified. This interpretability can help a pathologist to analyze the correlation between gene mutation and histopathological morphology. Experimental results demonstrate that the proposed gene mutation prediction framework significantly outperforms the state-of-the-art methods. In the TCGA bladder cancer dataset, five clinically relevant gene mutations are well predicted.

Towards targeted colorectal cancer biopsy based on tissue morphology assessment by compression optical coherence elastography

Identifying the precise topography of cancer for targeted biopsy in colonoscopic examination is a challenge in current diagnostic practice. For the first time we demonstrate the use of compression optical coherence elastography (C-OCE) technology as a new functional OCT modality for differentiating between cancerous and non-cancerous tissues in colon and detecting their morphological features on the basis of measurement of tissue elastic properties. The method uses pre-determined stiffness values (Young’s modulus) to distinguish between different morphological structures of normal (mucosa and submucosa), benign tumor (adenoma) and malignant tumor tissue (including cancer cells, gland-like structures, cribriform gland-like structures, stromal fibers, extracellular mucin). After analyzing in excess of fifty tissue samples, a threshold stiffness value of 520 kPa was suggested above which areas of colorectal cancer were detected invariably. A high Pearson correlation (r =0.98; p &lt;0.05), and a negligible bias (0.22) by good agreement of the segmentation results of C-OCE and histological (reference standard) images was demonstrated, indicating the efficiency of C-OCE to identify the precise localization of colorectal cancer and the possibility to perform targeted biopsy. Furthermore, we demonstrated the ability of C-OCE to differentiate morphological subtypes of colorectal cancer – low-grade and high-grade colorectal adenocarcinomas, mucinous adenocarcinoma, and cribriform patterns. The obtained ex vivo results highlight prospects of C-OCE for high-level colon malignancy detection. The future endoscopic use of C-OCE will allow targeted biopsy sampling and simultaneous rapid analysis of the heterogeneous morphology of colon tumors.

Predicting gene mutation status via artificial intelligence technologies based on multimodal integration (MMI) to advance precision oncology

Personalized treatment strategies for cancer frequently rely on the detection of genetic alterations which are determined by molecular biology assays. Historically, these processes typically required single-gene sequencing, next-generation sequencing, or visual inspection of histopathology slides by experienced pathologists in a clinical context. In the past decade, advances in artificial intelligence (AI) technologies have demonstrated remarkable potential in assisting physicians with accurate diagnosis of oncology image-recognition tasks. Meanwhile, AI techniques make it possible to integrate multimodal data such as radiology, histology, and genomics, providing critical guidance for the stratification of patients in the context of precision therapy. Given that the mutation detection is unaffordable and time-consuming for a considerable number of patients, predicting gene mutations based on routine clinical radiological scans or whole-slide images of tissue with AI-based methods has become a hot issue in actual clinical practice. In this review, we synthesized the general framework of multimodal integration (MMI) for molecular intelligent diagnostics beyond standard techniques. Then we summarized the emerging applications of AI in the prediction of mutational and molecular profiles of common cancers (lung, brain, breast, and other tumor types) pertaining to radiology and histology imaging. Furthermore, we concluded that there truly exist multiple challenges of AI techniques in the way of its real-world application in the medical field, including data curation, feature fusion, model interpretability, and practice regulations. Despite these challenges, we still prospect the clinical implementation of AI as a highly potential decision-support tool to aid oncologists in future cancer treatment management.

Clinicopathological characteristics of high microsatellite instability/mismatch repair-deficient colorectal cancer: A narrative review

Colorectal cancers (CRCs) with high microsatellite instability (MSI-H) and deficient mismatch repair (dMMR) show molecular and clinicopathological characteristics that differ from those of proficient mismatch repair/microsatellite stable CRCs. Despite the importance of MSI-H/dMMR status in clinical decision making, the testing rates for MSI and MMR in clinical practice remain low, even in high-risk populations. Additionally, the real-world prevalence of MSI-H/dMMR CRC may be lower than that reported in the literature. Insufficient MSI and MMR testing fails to identify patients with MSI-H/dMMR CRC, who could benefit from immunotherapy. In this article, we describe the current knowledge of the clinicopathological features, molecular landscape, and radiomic characteristics of MSI-H/dMMR CRCs. A better understanding of the importance of MMR/MSI status in the clinical characteristics and prognosis of CRC may help increase the rates of MMR/MSI testing and guide the development of more effective therapies based on the unique features of these tumors.

New thiophene derivative augments the antitumor activity of γ-irradiation against colorectal cancer in mice via anti-inflammatory and pro-apoptotic pathways

BACKGROUND

Colorectal cancer (CRC) is one of the most common types of cancer worldwide and the second cause of cancer-related deaths. It usually starts as an inflammation that progresses to adenocarcinoma. The goal of the present study was to investigate the antitumor efficacy of a new thiophene derivative against CRC in mice and explore the possible associated molecular pathways. The potential of this thiophene derivative to sensitize the CRC tumor tissue to a low dose of gamma irradiation was also investigated.

METHODS

Adult male mice were divided into seven groups; control, group treated with dimethylhydrazine (DMH) for the induction of CRC. The DMH-group was further divided into six groups and treated with either cisplatin, thiophene derivative, γ-irradiation, cisplatin + γ-irradiation, thiophene derivative + γ-irradiation, or left untreated.

RESULTS

DMH induced CRC as evidenced by the macroscopic examination of colon tissues and histopathology, and elevated the activities of cyclooxygenase2 (COX2) and nitric oxide synthase (iNOS). DMH also elevated kirsten rat sarcoma (KRAS) and downregulated the peroxisome proliferator activated receptor (PPARγ) as shown by RT-PCR and Western blotting. DMH exerted anti-apoptotic activity by reducing the expression of phosphorylated p53 and cleaved caspase3 at the gene and protein levels. The flow cytometry analysis showed that DMH elevated the necrosis and reduced the apoptosis compared to the other groups. The colon tissue from DMH-treated mice showed hyperplasia, aberrant crypt foci, loss of cell polarity, typical CRC of grade 4 with lymphocytes and macrophages infiltrating mucosa, muscularis mucosa, and submucosa score 3. Treatment with thiophene derivative or γ-irradiation ameliorated most of these deleterious effects of DMH. The concomitant action of thiophene derivative + γ-irradiation was typified by the better amelioration of tumor incidence and multiplicity, iNOS, PPARγ, p53, caspase 3, and histopathology of colon.

CONCLUSION

Taken together, the new thiophene derivative is a promising therapeutic candidate for treatment of colorectal cancer in mice. It also sensitizes the CRC tumor to the ionizing radiation through anti-inflammatory and pro-apoptotic pathways.

Deep learning model to predict Epstein-Barr virus associated gastric cancer in histology

The detection of Epstein-Barr virus (EBV) in gastric cancer patients is crucial for clinical decision making, as it is related with specific treatment responses and prognoses. Despite its importance, the limited medical resources preclude universal EBV testing. Herein, we propose a deep learning-based EBV prediction method from H&E-stained whole-slide images (WSI). Our model was developed using 319 H&E stained WSI (26 EBV positive; TCGA dataset) from the Cancer Genome Atlas, and 108 WSI (8 EBV positive; ISH dataset) from an independent institution. Our deep learning model, EBVNet consists of two sequential components: a tumor classifier and an EBV classifier. We visualized the learned representation by the classifiers using UMAP. We externally validated the model using 60 additional WSI (7 being EBV positive; HGH dataset). We compared the model's performance with those of four pathologists. EBVNet achieved an AUPRC of 0.65, whereas the four pathologists yielded a mean AUPRC of 0.41. Moreover, EBVNet achieved an negative predictive value, sensitivity, specificity, precision, and F1-score of 0.98, 0.86, 0.92, 0.60, and 0.71, respectively. Our proposed model is expected to contribute to prescreen patients for confirmatory testing, potentially to save test-related cost and labor.

A review on mechanobiology of cell adhesion networks in different stages of sporadic colorectal cancer to explain its tumorigenesis

Sporadic colorectal cancer (CRC) is strongly linked to extraneous factors, like poor diet and lifestyle, but not to inherent factors like familial genetics. The changes at the epigenomics and signalling pathways are known across the sporadic CRC stages. The catch is that temporal information of the onset, the feedback loop, and the crosstalk of signalling and noise are still unclear. This makes it challenging to diagnose and treat colon cancer effectively with no relapse. Various microbial cells and native cells of the colon, contribute to sporadic CRC development. These cells secrete autocrine and paracrine for their bioenergetics and communications with other cell types. Imbalances of the biochemicals affect the epithelial lining of colon. One side of this epithelial lining is interfacing the dense colon tissue, while the other side is exposed to microbiota and excrement from the lumen. Hence, the epithelial lining is prone to tumorigenesis due to the influence of both biochemical and mechanical cues from its complex surrounding. The role of physical transformations in tumorigenesis have been limitedly discussed. In this context, cellular and tissue structures, and force transductions are heavily regulated by cell adhesion networks. These networks include cell anchoring mechanism to the surrounding, cell structural integrity mechanism, and cell effector molecules. This review will focus on the progression of the sporadic CRC stages that are governed by the underlaying cell adhesion networks within the epithelial cells. Additionally, current and potential technologies and therapeutics that target cell adhesion networks for treatments of sporadic CRC will be incorporated.

Clinicopathological and molecular characteristics of RSPO fusion-positive colorectal cancer

BACKGROUND

RSPO fusions that lead to WNT pathway activation are potential therapeutic targets in colorectal cancer (CRC), but their clinicopathological significance remains unclear.

METHODS

We screened 1019 CRCs for RSPO fusions using multiplex reverse transcription-PCR. The RSPO fusion-positive tumours were subjected to whole-exome sequencing (WES).

RESULTS

Our analysis identified 29 CRCs with RSPO fusions (2.8%), consisting of five with an EIF3E-RSPO2 fusion and 24 with PTPRK-RSPO3 fusions. The patients were 17 women and 12 men. Thirteen tumours (45%) were right-sided. Histologically, approximately half of the tumours (13/29, 45%) had a focal or extensive mucinous component that was significantly more frequent than the RSPO fusion-negative tumours (13%; P = 8.1 × 10^-7). Four tumours (14%) were mismatch repair-deficient. WES identified KRAS, BRAF, and NRAS mutations in a total of 27 tumours (93%). In contrast, pathogenic mutations in major WNT pathway genes, such as APC, CTNNB1 and RNF43, were absent. RSPO fusion status did not have a statistically significant influence on the overall or recurrence-free survival. These clinicopathological and genetic features were also confirmed in a pooled analysis of previous studies.

CONCLUSION

RSPO fusion-positive CRCs constitute a rare subgroup of CRCs with several characteristic clinicopathological and genetic features.

A Novel Cartesian Plot Analysis for Fixed Monolayers That Relates Cell Phenotype to Transfer of Contents between Fibroblasts and Cancer Cells by Cell-Projection Pumping

We recently described cell-projection pumping as a mechanism transferring cytoplasm between cells. The uptake of fibroblast cytoplasm by co-cultured SAOS-2 osteosarcoma cells changes SAOS-2 morphology and increases cell migration and proliferation, as seen by single-cell tracking and in FACS separated SAOS-2 from co-cultures. Morphological changes in SAOS-2 seen by single cell tracking are consistent with previous observations in fixed monolayers of SAOS-2 co-cultures. Notably, earlier studies with fixed co-cultures were limited by the absence of a quantitative method for identifying sub-populations of co-cultured cells, or for quantitating transfer relative to control populations of SAOS-2 or fibroblasts cultured alone. We now overcome that limitation by a novel Cartesian plot analysis that identifies individual co-cultured cells as belonging to one of five distinct cell populations, and also gives numerical measure of similarity to control cell populations. We verified the utility of the method by first confirming the previously established relationship between SAOS-2 morphology and uptake of fibroblast contents, and also demonstrated similar effects in other cancer cell lines including from melanomas, and cancers of the ovary and colon. The method was extended to examine global DNA methylation, and while there was no clear effect on SAOS-2 DNA methylation, co-cultured fibroblasts had greatly reduced DNA methylation, similar to cancer associated fibroblasts.

Microsatellite Instable Colorectal Adenocarcinoma Diagnostics: The Advent of Liquid Biopsy Approaches

The introduction of immunotherapy has revolutionized the oncological targeted therapy paradigm. Microsatellite instability (MSI) identifies a subgroup of colorectal cancers (CRCs) which respond to treatment with immune checkpoint inhibitors. Tissue biopsy is currently the gold standard for the assessment of MSI/Mismatch Repair deficiency (MMRd) by means immunohistochemistry or molecular assays. However, the application of liquid biopsy in the clinic may help to overcome several limitations of tissue analysis and may provide great benefit to the diagnostic scenario and therapeutic decision-making process. In the context of MSI/MMRd CRC, the use of liquid biopsy may allow to establish MSI/MMR status if tissue sampling cannot be performed or in case of discordant tissue biopsies. Liquid biopsy may also become a powerful tool to monitor treatment response and the onset resistance to immunotherapy over time and to stratify of MSI/MMRd patients according to their risk of relapse and metastases. The aim of this review is to summarize the main technical aspects and clinical applications, the benefits, and limitations of the use of liquid biopsy in MSI/MMRd colorectal cancer patients.

Swarm learning for decentralized artificial intelligence in cancer histopathology

Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.

A decentralized, privacy-preserving machine learning framework used to train a clinically relevant AI system identifies actionable molecular alterations in patients with colorectal cancer by use of routine histopathology slides collected in real-world settings.

Prognosis and Sensitivity of Adjuvant Chemotherapy in Mucinous Colorectal Adenocarcinoma without Distant Metastasis

In colorectal cancer, whereas mucinous adenocarcinoma (MAC) has several poor clinical prognostic factors compared to adenocarcinoma (AC), the prognosis of MAC remains controversial. We evaluated the prognosis of MAC without distant metastasis and the effects of adjuvant chemotherapy using health insurance registry data managed by South Korea. Patients with colorectal cancer between January 2014 and December 2016 were included (AC, 22,050 [96.8%]; MAC, 729 [3.2%]). We observed no difference in overall survival (OS) between AC and MAC in stages I and II. However, MAC showed a worse OS than AC in stage III disease, especially in patients administered chemotherapy (p < 0.001). These findings persisted after propensity score matching of clinical characteristics between AC and MAC. In addition, transcriptome analysis of The Cancer Genome Atlas (TCGA) data showed increased chemoresistance-associated pathways in MAC compared to AC. In consensus molecular subtypes (CMS) classification, unlike in AC, CMSs 1, 3, and 4 comprised most of MAC and the proportions of CMSs 3 and 4 increased with stage progression. These results suggest clues to overcome resistance to chemotherapy and develop targeted treatments in MAC.

Prognostic Impact of An Integrative Landscape of Clinical, Immune, and Molecular Features in Non-Metastatic Rectal Cancer

Rectal Cancer (RC) is a complex disease that involves highly variable treatment responses. Currently, there is a lack of reliable markers beyond TNM to deliver a personalized treatment in a cancer setting where the goal is a curative treatment. Here, we performed an integrated characterization of the predictive and prognostic role of clinical features, mismatch-repair deficiency markers, HER2, CDX2, PD-L1 expression, and CD3-CD8+ tumor-infiltrating lymphocytes (TILs) coupled with targeted DNA sequencing of 76 non-metastatic RC patients assigned to total mesorectal excision upfront (TME; n = 15) or neoadjuvant chemo-radiotherapy treatment (nCRT; n = 61) followed by TME. Eighty-two percent of RC cases displayed mutations affecting cancer driver genes such as TP53, APC, KRAS, ATM, and PIK3CA. Good response to nCRT treatment was observed in approximately 40% of the RC cases, and poor pathological tumor regression was significantly associated with worse disease-free survival (DFS, HR = 3.45; 95%CI = 1.14-10.4; p = 0.028). High neutrophils-platelets score (NPS) (OR = 10.52; 95%CI=1.34-82.6; p = 0.025) and KRAS mutated cases (OR = 5.49; 95%CI = 1.06-28.4; p = 0.042) were identified as independent predictive factors of poor response to nCRT treatment in a multivariate analysis. Furthermore, a Cox proportional-hazard model showed that the KRAS mutational status was an independent prognostic factor associated with higher risk of local recurrence (HR = 9.68; 95%CI = 1.01-93.2; p <0.05) and shorter DFS (HR = 2.55; 95%CI = 1.05-6.21; p <0.05), while high CEA serum levels were associated with poor DFS (HR = 2.63; 95%CI = 1.01-6.85; p <0.05). Integrated clinical and molecular-based unsupervised analysis allowed us to identify two RC prognostic groups (cluster 1 and cluster 2) associated with disease-specific OS (HR = 20.64; 95%CI = 2.63-162.2; p <0.0001), metastasis-free survival (HR = 3.67; 95%CI = 1.22-11; p = 0.012), local recurrence-free survival (HR = 3.34; 95%CI = 0.96-11.6; p = 0.043) and worse DFS (HR = 2.68; 95%CI = 1.18-6.06; p = 0.012). The worst prognosis cluster 2 was enriched by stage III high-risk clinical tumors, poor responders to nCRT, with low TILs density and high frequency of KRAS and TP53 mutated cases compared with the best prognosis cluster 1 (p <0.05). Overall, this study provides a comprehensive and integrated characterization of non-metastatic RC cases as a new insight to deliver a personalized therapeutic approach.

The seen and the unseen: Molecular classification and image based-analysis of gastrointestinal cancers

Gastrointestinal cancers account for 22.5% of cancer related deaths worldwide and represent circa 20% of all cancers. In the last decades, we have witnessed a shift from histology-based to molecular-based classifications using genomic, epigenomic, and transcriptomic data. The molecular based classification revealed new prognostic markers and may aid the therapy selection. Because of the high-costs to perform a molecular classification, in recent years immunohistochemistry-based surrogate classification were developed which permit the stratification of patients, and in parallel multiple groups developed hematoxylin and eosin whole slide image analysis for sub-classifying these entities. Hence, we are witnessing a return to an image-based classification with the purpose to infer hidden information from routine histology images that would permit to detect the patients that respond to specific therapies and would be able to predict their outcome. In this review paper, we will discuss the current histological, molecular, and immunohistochemical classifications of the most common gastrointestinal cancers, gastric adenocarcinoma, and colorectal adenocarcinoma, and will present key aspects for developing a new artificial intelligence aided image-based classification of these malignancies.

Development and validation of a weakly supervised deep learning framework to predict the status of molecular pathways and key mutations in colorectal cancer from routine histology images: a retrospective study

BACKGROUND

Determining the status of molecular pathways and key mutations in colorectal cancer is crucial for optimal therapeutic decision making. We therefore aimed to develop a novel deep learning pipeline to predict the status of key molecular pathways and mutations from whole-slide images of haematoxylin and eosin-stained colorectal cancer slides as an alternative to current tests.

METHODS

In this retrospective study, we used 502 diagnostic slides of primary colorectal tumours from 499 patients in The Cancer Genome Atlas colon and rectal cancer (TCGA-CRC-DX) cohort and developed a weakly supervised deep learning framework involving three separate convolutional neural network models. Whole-slide images were divided into equally sized tiles and model 1 (ResNet18) extracted tumour tiles from non-tumour tiles. These tumour tiles were inputted into model 2 (adapted ResNet34), trained by iterative draw and rank sampling to calculate a prediction score for each tile that represented the likelihood of a tile belonging to the molecular labels of high mutation density (vs low mutation density), microsatellite instability (vs microsatellite stability), chromosomal instability (vs genomic stability), CpG island methylator phenotype (CIMP)-high (vs CIMP-low), BRAFmut (vs BRAFWT), TP53mut (vs TP53WT), and KRASWT (vs KRASmut). These scores were used to identify the top-ranked titles from each slide, and model 3 (HoVer-Net) segmented and classified the different types of cell nuclei in these tiles. We calculated the area under the convex hull of the receiver operating characteristic curve (AUROC) as a model performance measure and compared our results with those of previously published methods.

FINDINGS

Our iterative draw and rank sampling method yielded mean AUROCs for the prediction of hypermutation (0·81 [SD 0·03] vs 0·71), microsatellite instability (0·86 [0·04] vs 0·74), chromosomal instability (0·83 [0·02] vs 0·73), BRAFmut (0·79 [0·01] vs 0·66), and TP53mut (0·73 [0·02] vs 0·64) in the TCGA-CRC-DX cohort that were higher than those from previously published methods, and an AUROC for KRASmut that was similar to previously reported methods (0·60 [SD 0·04] vs 0·60). Mean AUROC for predicting CIMP-high status was 0·79 (SD 0·05). We found high proportions of tumour-infiltrating lymphocytes and necrotic tumour cells to be associated with microsatellite instability, and high proportions of tumour-infiltrating lymphocytes and a low proportion of necrotic tumour cells to be associated with hypermutation.

INTERPRETATION

After large-scale validation, our proposed algorithm for predicting clinically important mutations and molecular pathways, such as microsatellite instability, in colorectal cancer could be used to stratify patients for targeted therapies with potentially lower costs and quicker turnaround times than sequencing-based or immunohistochemistry-based approaches.

FUNDING

The UK Medical Research Council.

Distinct mutational profile and immune microenvironment in microsatellite-unstable and POLE-mutated tumors

INTRODUCTION

Mismatch repair (MMR)-deficient and DNA polymerase epsilon (POLE)-mutated tumors exhibit a high tumor mutation burden (TMB) and have been proven to be associated with good responses to immune checkpoint inhibitor treatments. However, the relationship between mutational characteristics of MMR-deficient and POLE-mutated tumors and the spatial architecture of tumor-infiltrating lymphocytes (TILs) has not been fully evaluated.

METHODS

We retrieved microsatellite instability-high (MSI-high, N=20) and POLE-mutated (N=47) cases from the clinical next-generation sequencing cohort at Asan Medical Center. Whole-slide immunostaining for CD3, CD4, CD8, FoxP3 and PD-1 were performed with tissue samples of colorectal and gastric cancer (N=24) and the tumor-positive TIL cell densities were correlated with the tumor's mutational features. The findings were compared with the results of similar analyses in The Cancer Genome Atlas-Colorectal Adenocarcinoma (TCGA-COADREAD) cohort (N=592).

RESULTS

The MSI-high group showed significantly higher overall TMBs with a number of insertion/deletion (indel) mutations relative to the POLE-mutated group (median TMB; 83.6 vs 12.5/Mb). Oncogenic/likely-oncogenic POLE mutations were identified with ultrahypermutations (≥100 mutations/Mb) (2/47, 4.3%). Concurrent POLE mutations of unknown significance and MSI-high cases were identified in eight cases (8/67, 11%), and two of these colorectal cancers had multiple POLE mutations, showing an ultramutated phenotype (378.1 and 484.4/Mb) and low indel mutation burdens with complete loss of MSH-6 or PMS-2, which was similar to the mutational profile of the POLE-inactivated tumors. Intratumoral CD3-positive, CD4-positive, CD8-positive, FoxP3-positive and PD-1-positive TIL cell densities were more strongly correlated with the indel mutation burden than with the total TMB (correlation coefficient, 0.61-0.73 vs 0.23-0.38). In addition, PI3K/AKT/mTOR pathway mutations were commonly found in MSI-high tumors (75%) but not in POLE-mutated tumors.

CONCLUSIONS

Indel mutation burden rather than total TMB could serve as a predictor of high TILs in both MSI-high and POLE-mutated tumors. Multiple uncharacterized/non-pathogenic POLE mutations occurring via MMR deficiency within MSI-high tumors may have combined pathogenic roles. A mutated PI3K/AKT/mTOR pathway may be a biomarker that can be used to stratify patients with advanced MSI-high tumors for immune therapy.

In situ functional cell phenotyping reveals microdomain networks in colorectal cancer recurrence

Tumors are dynamic ecosystems comprising localized niches (microdomains), possessing distinct compositions and spatial configurations of cancer and non-cancer cell populations. Microdomain-specific network signaling coevolves with a continuum of cell states and functional plasticity associated with disease progression and therapeutic responses. We present LEAPH, an unsupervised machine learning algorithm for identifying cell phenotypes, which applies recursive steps of probabilistic clustering and spatial regularization to derive functional phenotypes (FPs) along a continuum. Combining LEAPH with pointwise mutual information and network biology analyses enables the discovery of outcome-associated microdomains visualized as distinct spatial configurations of heterogeneous FPs. Utilization of an immunofluorescence-based (51 biomarkers) image dataset of colorectal carcinoma primary tumors (n = 213) revealed microdomain-specific network dysregulation supporting cancer stem cell maintenance and immunosuppression that associated selectively with the recurrence phenotype. LEAPH enables an explainable artificial intelligence platform providing insights into pathophysiological mechanisms and novel drug targets to inform personalized therapeutic strategies.

Current prognostic and predictive biomarkers for gastrointestinal tumors in clinical practice

The pathologist emerged in the personalized medicine era as a central actor in the definition of the most adequate diagnostic and therapeutic algorithms. In the last decade, gastrointestinal oncology has seen a significantly increased clinical request for the integration of novel prognostic and predictive biomarkers in histopathological reports. This request couples with the significant contraction of invasive sampling of the disease, thus conferring to the pathologist the role of governor for both proper pathologic characterization and customized processing of the biospecimens. This overview will focus on the most commonly adopted immunohistochemical and molecular biomarkers in the routine clinical characterization of gastrointestinal neoplasms referring to the most recent published recommendations, guidelines and expert opinions.

Gastrointestinal cancer classification and prognostication from histology using deep learning: Systematic review

BACKGROUND

Gastrointestinal cancers account for approximately 20% of all cancer diagnoses and are responsible for 22.5% of cancer deaths worldwide. Artificial intelligence-based diagnostic support systems, in particular convolutional neural network (CNN)-based image analysis tools, have shown great potential in medical computer vision. In this systematic review, we summarise recent studies reporting CNN-based approaches for digital biomarkers for characterization and prognostication of gastrointestinal cancer pathology.

METHODS

Pubmed and Medline were screened for peer-reviewed papers dealing with CNN-based gastrointestinal cancer analyses from histological slides, published between 2015 and 2020.Seven hundred and ninety titles and abstracts were screened, and 58 full-text articles were assessed for eligibility.

RESULTS

Sixteen publications fulfilled our inclusion criteria dealing with tumor or precursor lesion characterization or prognostic and predictive biomarkers: 14 studies on colorectal or rectal cancer, three studies on gastric cancer and none on esophageal cancer. These studies were categorised according to their end-points: polyp characterization, tumor characterization and patient outcome. Regarding the translation into clinical practice, we identified several studies demonstrating generalization of the classifier with external tests and comparisons with pathologists, but none presenting clinical implementation.

CONCLUSIONS

Results of recent studies on CNN-based image analysis in gastrointestinal cancer pathology are promising, but studies were conducted in observational and retrospective settings. Large-scale trials are needed to assess performance and predict clinical usefulness. Furthermore, large-scale trials are required for approval of CNN-based prediction models as medical devices.

Cell-stiffness and morphological architectural patterns in clinical samples of high grade serous ovarian cancers

High grade serous ovarian carcinoma (HGSOC) is recognized as the most frequent type of ovarian cancer and the main cause of ovarian cancer related deaths worldwide. Although homologous recombination deficiency testing has been adopted in the clinical workflow, morphological analysis remains the main diagnostic tool. In this study Atomic Force Microscopy (AFM) was tested in standard hematoxylin and eosin (H&E) stained sections to investigate the biomechanical properties of different architectural growing patterns of HGSOC. Our results showed that AFM was able to discriminate HGSOC morphological growing patterns as well as patients' stage. Micropapillary pattern, which has been associated to poor outcome, had lower Young's moduli. In addition stage IV HGSOC was significantly softer than stage III cancers. Based on our results, AFM analysis could represent an additional tool in HGSOC morphological diagnosis as the biomechanical proprieties of HGSOC were quantitatively associated to tumor staging and architectural pattern.

Artificial Intelligence for solid tumor diagnosis in digital pathology

Tumor diagnosis relies on the visual examination of histological slides by pathologists through a microscope eyepiece. Digital pathology, the digitalization of histological slides at high magnification with slides scanners, has raised the opportunity to extract quantitative information thanks to image analysis. In the last decade, medical image analysis has made exceptional progress due to the development of artificial intelligence (AI) algorithms. AI has been successfully used in the field of medical imaging and more recently in digital pathology. The feasibility and usefulness of AI assisted pathology tasks have been demonstrated in the very last years and we can expect those developments to be applied on routine histopathology in the future. In this review, we will describe and illustrate this technique and present the most recent applications in the field of tumor histopathology.

Distinct Mutational Profile of Lynch Syndrome Colorectal Cancers Diagnosed under Regular Colonoscopy Surveillance

Regular colonoscopy even with short intervals does not prevent all colorectal cancers (CRC) in Lynch syndrome (LS). In the present study, we asked whether cancers detected under regular colonoscopy surveillance (incident cancers) are phenotypically different from cancers detected at first colonoscopy (prevalent cancers). We analyzed clinical, histological, immunological and mutational characteristics, including panel sequencing and high-throughput coding microsatellite (cMS) analysis, in 28 incident and 67 prevalent LS CRCs (n total = 95). Incident cancers presented with lower UICC and T stage compared to prevalent cancers (p < 0.0005). The majority of incident cancers (21/28) were detected after previous colonoscopy without any pathological findings. On the molecular level, incident cancers presented with a significantly lower KRAS codon 12/13 (1/23, 4.3% vs. 11/21, 52%; p = 0.0005) and pathogenic TP53 mutation frequency (0/17, 0% vs. 7/21, 33.3%; p = 0.0108,) compared to prevalent cancers; 10/17 (58.8%) incident cancers harbored one or more truncating APC mutations, all showing mutational signatures of mismatch repair (MMR) deficiency. The proportion of MMR deficiency-related mutational events was significantly higher in incident compared to prevalent CRC (p = 0.018). In conclusion, our study identifies a set of features indicative of biological differences between incident and prevalent cancers in LS, which should further be monitored in prospective LS screening studies to guide towards optimized prevention protocols.

An Insight Into the Driver Mutations and Molecular Mechanisms Underlying Mucinous Adenocarcinoma of the Rectum

BACKGROUND

Mucinous adenocarcinoma of the rectum accounts for 10% of all rectal cancers and has an impaired response to neoadjuvant chemoradiotherapy and worse overall survival. To date, insufficient genomic research has been performed on this histological subtype.

OBJECTIVE

This study aims to define the mismatch repair deficiency rate and the driver mutations underpinning mucinous adenocarcinoma of the rectum and to compare it with rectal adenocarcinoma not otherwise specified.

DESIGN

Immunohistochemistry and sequencing were performed on tumor samples from our tumor biobank.

SETTINGS

This study was conducted across 2 tertiary referral centers.

PATIENTS

Patients with mucinous adenocarcinoma and rectal adenocarcinoma not otherwise specified who underwent rectal resection between 2008 and 2018 were included.

MAIN OUTCOME MEASURES

Mismatch repair status was performed by immunohistochemical staining. Mutations in the panel of oncogenes and tumor suppressor genes were determined by sequencing on the MiSeq V3 platform.

RESULTS

The study included 33 patients with mucinous adenocarcinoma of the rectum and 100 patients with rectal adenocarcinoma not otherwise specified. Those with mucinous adenocarcinoma had a mismatch repair deficiency rate of 12.1% compared to 2.0% in the adenocarcinoma not otherwise specified cohort (p = 0.04). Mucinous adenocarcinoma and adenocarcinoma not otherwise specified rectal tumors had similar mutation frequencies in most oncogenes and tumor suppressor genes. No difference was found in the KRAS mutation rate (50.0% vs 37.1%, p = 0.29) or BRAF mutation rate (6.7% vs 3.1%, p = 0.34) between the cohorts. No difference was found between the cohorts regarding recurrence-free (p = 0.29) or overall survival (p = 0.14).

LIMITATIONS

The major limitations of this study were the use of formalin-fixed, paraffin-embedded tissue over fresh-frozen tissue and the small number of patients included, in particular, in the mucinous rectal cohort.

CONCLUSIONS

Most mucinous rectal tumors develop and progress along the chromosomal instability pathway. Further research in the form of transcriptomics, proteomics, and analysis of the effects of the mucin barrier may yield valuable insights into the mechanisms of resistance to chemoradiotherapy in this cohort. See Video Abstract at http://links.lww.com/DCR/B464.

UNA PERCEPCIN SOBRE MUTACIONES IMPULSORAS Y MECANISMOS MOLECULARES SUBYACENTES AL ADENOCARCINOMA MUCINOSO DEL RECTO

ANTECEDENTES:El adenocarcinoma mucinoso del recto, representa el 10% de todos los cánceres rectales y tiene una respuesta deficiente a la quimioradioterapia neoadyuvante y una peor supervivencia en general. A la fecha, se han realizado muy pocas investigaciones genómicas sobre este subtipo histológico.OBJETIVO:Definir la tasa de deficiencia en la reparación de desajustes y mutaciones impulsoras, que sustentan el adenocarcinoma mucinoso del recto y compararlo con el adenocarcinoma rectal no especificado de otra manera.DISEÑO:Se realizaron inmunohistoquímica y secuenciación en muestras tumorales de nuestro biobanco de tumores.AJUSTE:El estudio se realizó en dos centros de referencia terciarios.PACIENTES:Se incluyeron pacientes con adenocarcinoma mucinoso y adenocarcinoma no especificado de otra manera, sometidos a resección rectal entre 2008 y 2018.PRINCIPALES MEDIDAS DE RESULTADO:El estado de reparación de desajustes se realizó mediante tinción inmunohistoquímica. Las mutaciones en el panel de oncogenes y genes supresores de tumores, se determinaron mediante secuenciación en la plataforma MiSeq V3.RESULTADOS:El estudio incluyó a 33 pacientes con adenocarcinoma mucinoso del recto y 100 pacientes con adenocarcinoma del recto no especificado de otra manera. Aquellos con adenocarcinoma mucinoso, tenían una tasa de deficiencia de reparación de desajustes del 12,1% en comparación con el 2,0% en la cohorte de adenocarcinoma no especificado de otra manera (p = 0,04). El adenocarcinoma mucinoso y el adenocarcinoma no especificado de otra manera, tuvieron frecuencias de mutación similares en la mayoría de los oncogenes y genes supresores de tumores. No se encontraron diferencias en la tasa de mutación de KRAS (50,0% frente a 37,1%, p = 0,29) o la tasa de mutación de BRAF (6,7% frente a 3,1%, p = 0,34) entre las cohortes. No se encontraron diferencias entre las cohortes con respecto a la supervivencia libre de recurrencia (p = 0,29) o la supervivencia global (p = 0,14).LIMITACIONES:Las mayores limitaciones de este estudio, fueron el uso de tejido embebido en parafina y fijado con formalina, sobre el tejido fresco congelado y el pequeño número de pacientes incluidos, particularmente en la cohorte mucinoso rectal.CONCLUSIONES:La mayoría de los tumores rectales mucinosos se desarrollan y progresan a lo largo de la vía de inestabilidad cromosómica. La investigación adicional en forma transcriptómica, proteómica y análisis de los efectos de la barrera de la mucina, puede proporcionar información valiosa sobre los mecanismos de resistencia a la quimioradioterapia, en esta cohorte. Consulte Video Resumen en http://links.lww.com/DCR/B464.

Histopathological characteristics and artificial intelligence for predicting tumor mutational burden-high colorectal cancer

BACKGROUND

Tumor mutational burden-high (TMB-H), which is detected with gene panel testing, is a promising biomarker for immune checkpoint inhibitors (ICIs) in colorectal cancer (CRC). However, in clinical practice, not every patient is tested for TMB-H using gene panel testing. We aimed to identify the histopathological characteristics of TMB-H CRC for efficient selection of patients who should undergo gene panel testing. Moreover, we attempted to develop a convolutional neural network (CNN)-based algorithm to predict TMB-H CRC directly from hematoxylin and eosin (H&E) slides.

METHODS

We used two CRC cohorts tested for TMB-H, and whole-slide H&E digital images were obtained from the cohorts. The Japanese CRC (JP-CRC) cohort (N = 201) was evaluated to detect the histopathological characteristics of TMB-H using H&E slides. The JP-CRC cohort and The Cancer Genome Atlas (TCGA) CRC cohort (N = 77) were used to develop a CNN-based TMB-H prediction model from the H&E digital images.

RESULTS

Tumor-infiltrating lymphocytes (TILs) were significantly associated with TMB-H CRC (P < 0.001). The area under the curve (AUC) for predicting TMB-H CRC was 0.910. We developed a CNN-based TMB-H prediction model. Validation tests were conducted 10 times using randomly selected slides, and the average AUC for predicting TMB-H slides was 0.934.

CONCLUSIONS

TILs, a histopathological characteristic detected with H&E slides, are associated with TMB-H CRC. Our CNN-based model has the potential to predict TMB-H CRC directly from H&E slides, thereby reducing the burden on pathologists. These approaches will provide clinicians with important information about the applications of ICIs at low cost.

Is medullary carcinoma of the colon underdiagnosed? An audit of poorly differentiated colorectal carcinomas in a large national health service teaching hospital

AIMS

Medullary carcinoma is an uncommon colorectal tumour which appears poorly differentiated histologically. Consequently, it may be confused with poorly differentiated adenocarcinoma not otherwise specified (NOS). The principal aim of this study was to review a large series of poorly differentiated colorectal cancers resected at a large National Health Service (NHS) Teaching Hospital to determine how often medullary carcinomas were misclassified . Secondary aims were to investigate how often neuroendocrine differentiation or metastatic tumours were considered in the differential diagnosis, and compare clinico-pathological features between medullary and poorly differentiated adenocarcinoma NOS.

METHODS AND RESULTS

Histology slides from 302 colorectal cancer resections originally reported as poorly differentiated adenocarcinoma were reviewed and cases fulfilling World Health Organisation (WHO) criteria for medullary carcinoma identified. The original pathology report was examined for any mention of medullary phenotype, consideration of neuroendocrine differentiation or consideration of metastasis from another site. Clinico-pathological features were compared to poorly differentiated adenocarcinoma NOS. Only one-third of medullary carcinomas were correctly identified between 1997 and 2018. The other two-thirds were reported as poorly differentiated adenocarcinoma NOS. The possibility of an extracolonic origin or neuroendocrine carcinoma was considered in 21 and 27% of reports. Most medullary carcinomas exhibited mismatch repair deficiency, were located in ascending colon and caecum and had a lower rate of vascular channel invasion and lymph node metastasis compared to poorly differentiated adenocarcinoma.

CONCLUSIONS

Medullary carcinoma of the colon is often mistaken for poorly differentiated adenocarcinoma NOS and occasionally for neuroendocrine or metastatic carcinoma. Greater familiarity with morphological criteria and use of mismatch repair protein staining should improve diagnosis.

Mucinous Colorectal Cancer is Associated With Expression of the TIM-3 Immune Checkpoint Independently of Microsatellite Instability (MSI) Status

BACKGROUND

Immune checkpoint inhibition has demonstrated success in overcoming tumor-mediated immune suppression in several types of cancer. However, its clinical use is limited to a small subset of colorectal cancer (CRC) patients, and response is highly variable between CRC subtypes. This study aimed to determine the profile of immune checkpoints and factors associated with immune checkpoint inhibitor response in mucinous CRC.

METHODS

Gene expression data from CRC was extracted from the TCGA PanCanAtlas data-freeze release. Gene expression data were reported as batch-corrected and normalized RNA expression derived from RNA-Seq quantification. Clinical, pathologic, and transcriptomic data were compared between mucinous and non-mucinous CRC cohorts.

RESULTS

The 557 cases of CRC eligible for inclusion in this study comprised 486 cases of non-mucinous CRC (87.3 %) and 71 cases of mucinous CRC (12.7 %). High correlation was observed in the expression of the included immune checkpoints. Significantly higher expression of programmed cell death protein 1 ligand (PD-L1) and T cell immunoglobulin and mucin domain 3 (TIM-3) was observed in mucinous CRC than in non-mucinous CRC. In a multiple regression model, significant contributors to the prediction of TIM-3 gene expression were microsatellite instability (MSI) (unstandardized regression coefficient [B] = 1.223; p < 0.001), stage (American Joint Committee on Cancer [AJCC] 2; B = 0.423; p < 0.05), and mucinous status (B = 0.591; p < 0.01).

CONCLUSION

Expression of TIM-3, an emerging immune checkpoint inhibition target, was significantly higher in mucinous CRC, and expression was predicted by mucinous status independently of MSI. These findings should prompt investigation of immune checkpoint signaling in the mucinous tumor microenvironment to clarify the potential for immune checkpoint inhibition in mucinous CRC.

Interpretable survival prediction for colorectal cancer using deep learning

Deriving interpretable prognostic features from deep-learning-based prognostic histopathology models remains a challenge. In this study, we developed a deep learning system (DLS) for predicting disease-specific survival for stage II and III colorectal cancer using 3652 cases (27,300 slides). When evaluated on two validation datasets containing 1239 cases (9340 slides) and 738 cases (7140 slides), respectively, the DLS achieved a 5-year disease-specific survival AUC of 0.70 (95% CI: 0.66–0.73) and 0.69 (95% CI: 0.64–0.72), and added significant predictive value to a set of nine clinicopathologic features. To interpret the DLS, we explored the ability of different human-interpretable features to explain the variance in DLS scores. We observed that clinicopathologic features such as T-category, N-category, and grade explained a small fraction of the variance in DLS scores (R² = 18% in both validation sets). Next, we generated human-interpretable histologic features by clustering embeddings from a deep-learning-based image-similarity model and showed that they explained the majority of the variance (R² of 73–80%). Furthermore, the clustering-derived feature most strongly associated with high DLS scores was also highly prognostic in isolation. With a distinct visual appearance (poorly differentiated tumor cell clusters adjacent to adipose tissue), this feature was identified by annotators with 87.0–95.5% accuracy. Our approach can be used to explain predictions from a prognostic deep learning model and uncover potentially-novel prognostic features that can be reliably identified by people for future validation studies.

Prediction of clinically actionable genetic alterations from colorectal cancer histopathology images using deep learning

BACKGROUND

Identifying genetic mutations in cancer patients have been increasingly important because distinctive mutational patterns can be very informative to determine the optimal therapeutic strategy. Recent studies have shown that deep learning-based molecular cancer subtyping can be performed directly from the standard hematoxylin and eosin (H&E) sections in diverse tumors including colorectal cancers (CRCs). Since H&E-stained tissue slides are ubiquitously available, mutation prediction with the pathology images from cancers can be a time- and cost-effective complementary method for personalized treatment.

AIM

To predict the frequently occurring actionable mutations from the H&E-stained CRC whole-slide images (WSIs) with deep learning-based classifiers.

METHODS

A total of 629 CRC patients from The Cancer Genome Atlas (TCGA-COAD and TCGA-READ) and 142 CRC patients from Seoul St. Mary Hospital (SMH) were included. Based on the mutation frequency in TCGA and SMH datasets, we chose APC, KRAS, PIK3CA, SMAD4, and TP53 genes for the study. The classifiers were trained with 360 × 360 pixel patches of tissue images. The receiver operating characteristic (ROC) curves and area under the curves (AUCs) for all the classifiers were presented.

RESULTS

The AUCs for ROC curves ranged from 0.693 to 0.809 for the TCGA frozen WSIs and from 0.645 to 0.783 for the TCGA formalin-fixed paraffin-embedded WSIs. The prediction performance can be enhanced with the expansion of datasets. When the classifiers were trained with both TCGA and SMH data, the prediction performance was improved.

CONCLUSION

APC, KRAS, PIK3CA, SMAD4, and TP53 mutations can be predicted from H&E pathology images using deep learning-based classifiers, demonstrating the potential for deep learning-based mutation prediction in the CRC tissue slides.

Molecular Insights Into Colorectal Carcinoma

Colorectal carcinoma (CRC) is one of the most common type of cancers and a leading cause of cancer-related deaths worldwide and in the United States. CRC is a heterogeneous disease with a well-characterized stepwise accumulation molecular alteration associated with adenoma formation and progression to carcinoma. We review the genomic and epigenomic pathways, including chromosomal instability, microsatellite instability, and epigenetic instability or CpG island methylator phenotype, their characteristics, and prognosis. We describe the four consensus molecular subtypes of CRC established by the international Colorectal Cancer Subtyping Consortium, their mechanisms to develop cancer, molecular characterization, clinical features, and prognosis. Finally, we review currently used predictive biomarkers.

Histopathological imaging features- versus molecular measurements-based cancer prognosis modeling

For lung and many other cancers, prognosis is essentially important, and extensive modeling has been carried out. Cancer is a genetic disease. In the past 2 decades, diverse molecular data (such as gene expressions and DNA mutations) have been analyzed in prognosis modeling. More recently, histopathological imaging data, which is a "byproduct" of biopsy, has been suggested as informative for prognosis. In this article, with the TCGA LUAD and LUSC data, we examine and directly compare modeling lung cancer overall survival using gene expressions versus histopathological imaging features. High-dimensional penalization methods are adopted for estimation and variable selection. Our findings include that gene expressions have slightly better prognostic performance, and that most of the gene expressions are weakly correlated imaging features. This study may provide additional insight into utilizing the two types of important data in cancer prognosis modeling and into lung cancer overall survival.

Impact of Pre-Analytical Factors on MSI Test Accuracy in Mucinous Colorectal Adenocarcinoma: A Multi-Assay Concordance Study

Immunohistochemistry (IHC) and polymerase chain reaction (PCR) and fragment separation by capillary electrophoresis represent the current clinical laboratory standard for the evaluation of microsatellite instability (MSI) status. The importance of reporting MSI status in colorectal cancer is based on its potential for guiding treatment and as a prognostic indicator. It is also used to identify patients for Lynch syndrome testing. Our aim was to evaluate pre-analytical factors, such as age of formalin-fixed and paraffin-embedded (FFPE) block, neoplastic cell percentage, mucinous component, and DNA integrity, that may influence the accuracy of MSI testing and assess the concordance between three different MSI evaluation approaches. We selected the mucinous colorectal cancer (CRC) histotype for this study as it may possibly represent an intrinsic diagnostic issue due to its low tumor cellularity. Seventy-five cases of mucinous CRC and corresponding normal colon tissue samples were retrospectively selected. MMR proteins were evaluated by IHC. After DNA quality and quantity evaluation, the Idylla™ and TapeStation 4200 platforms were adopted for the evaluation of MSI status. Seventy-three (97.3%) cases were successfully analyzed by the three methodologies. Overall, the Idylla™ platform showed a concordance rate with IHC of 98.0% for microsatellite stable (MSS)/proficient MMR (pMMR) cases and 81.8% for MSI/deficient MMR (dMMR) cases. The TapeStation 4200 system showed a concordance rate with IHC of 96.0% for MSS/pMMR cases and 45.4% for MSI/dMMR cases. The concordance rates of the TapeStation 4200 system with respect to the Idylla™ platform were 98.1% for MSS profile and 57.8% for MSI profile. Discordant cases were analyzed using the Titano MSI kit. Considering pre-analytical factors, no significant variation in concordance rate among IHC analyses and molecular systems was observed by considering the presence of an acellular mucus cut-off >50% of the tumor area, FFPE year preparation, and DNA concentration. Conversely, the Idylla™ platform showed a significant variation in concordance rate with the IHC approach by considering a neoplastic cell percentage >50% (p-value = 0.002), and the TapeStation 4200 system showed a significant variation in concordance rate with the IHC approach by considering a DNA integrity number (DIN) ≥4 as cut-off (p-value = 0.009). Our data pinpoint a central role of the pre-analytical phase in the diagnostic outcome of MSI testing in CRC.

Imaging predictors of BRAF mutation in colorectal cancer

BACKGROUND

Colorectal cancer (CRC) is one of the leading causes of cancer deaths and is associated with various genetic mutations. BRAF mutations, found in approximately 10% of all CRCs, are associated with negative predictive outcomes. The goal of this study was to assess the relationship between the imaging findings and BRAF statuses of CRC patients.

MATERIALS AND METHODS

The study population was colorectal cancer patients who underwent biopsy or surgery in a single institution from September 2004 to October 2018, and in whom the pathologic specimens were tested for BRAF mutation. The exclusion criteria were (1) patients without pre-operative cross-sectional imaging, and (2) patients whose tumors were invisible on imaging. Two hundred and eighty-three patients met the inclusion criteria. Among them, 128 were excluded, and a total of 155 patients were enrolled in the study.

RESULTS

BRAF mutations were significantly more common in female patients (p = 0.007). Patients with mutated BRAF were significantly older than those with wild-type BRAF (p = 0.001). BRAF-mutant tumors were predominant in right-sided colon (p = 0.001) with higher numbers of polypoid- or mass-like morphology (p = 0.019) and heterogeneous enhancement (p = 0.009). Compared to their wild-type counterparts, BRAF-mutated CRCs have a lower occurrence of non-peritoneal, and overall metastases (p = 0.013 and p = 0.004, respectively). Logistic regression analysis showed three significant factors for the prediction of BRAF mutations in CRC patients: right-sided location (p = 0.002), heterogeneous tumor enhancement (p = 0.039), and lack of non-peritoneal metastasis (p = 0.043).

CONCLUSION

By recognizing the specific imaging features of BRAF-mutant CRCs, it would be possible to identify a patient who has a higher risk of carrying BRAF mutation.

Pan-cancer image-based detection of clinically actionable genetic alterations

Molecular alterations in cancer can cause phenotypic changes in tumor cells and their micro-environment. Routine histopathology tissue slides - which are ubiquitously available - can reflect such morphological changes. Here, we show that deep learning can consistently infer a wide range of genetic mutations, molecular tumor subtypes, gene expression signatures and standard pathology biomarkers directly from routine histology. We developed, optimized, validated and publicly released a one-stop-shop workflow and applied it to tissue slides of more than 5000 patients across multiple solid tumors. Our findings show that a single deep learning algorithm can be trained to predict a wide range of molecular alterations from routine, paraffin-embedded histology slides stained with hematoxylin and eosin. These predictions generalize to other populations and are spatially resolved. Our method can be implemented on mobile hardware, potentially enabling point-of-care diagnostics for personalized cancer treatment. More generally, this approach could elucidate and quantify genotype-phenotype links in cancer.

Histology and its prognostic effect on KRAS-mutated colorectal carcinomas in Korea

KRAS mutation is frequently identified in advanced colorectal carcinoma (CRC); however, its prognostic significance and the associated histological features have remained to be clarified. In the present study, the precise histological results and prognostic value of KRAS-mutated CRCs were investigated in patients from South Korea. A retrospective review of the results from KRAS mutation testing, as well as evaluation of the histology of 310 cases of CRC at various stages, were performed. Cross-tabulation and survival analysis were performed according to the KRAS status. Patients with KRAS mutation more frequently exhibited serrated and papillary architectures (P=0.009 and P=0.014, respectively). KRAS mutation was an independent unfavorable prognostic factor for overall survival (OS) according to multivariate analysis (P=0.001), whereas no association was observed with disease-free survival (DFS) (P=0.611). Of note, in the subgroup of KRAS-mutated carcinomas, the presence of a solid component on histology was associated with less favorable OS (P=0.032). Furthermore, among the wild type cases, patients with a micropapillary component had a worse OS than those who did not (P=0.018). However, no subgroup or specific histological features were associated with DFS. In summary, KRAS-mutated CRCs had a moderate association with particular histological features, and according to the KRAS mutational status, there was a certain degree of association between histology and prognosis.

Colorectal carcinomas with mucinous differentiation are associated with high frequent mutation of KRAS or BRAF mutations, irrespective of quantity of mucinous component

Background

Mucinous adenocarcinoma (MAC) is a distinct type of colorectal cancer (CRC) associated with poor response to treatment and poorer prognosis. MAC is diagnosed by WHO definition when the extracellular mucin is more than 50% of the lesion. We aimed at assessing the gene expression profiles of the CRCs with any mucinous features (> 5%) in a retrospective study.

Methods

The data of a 50-gene next generation sequencing (NGS) panel of 166 CRCs was analyzed and the gene mutational profile with morphologic features was correlated.

Results

We identified the different genetic mutation profiles between CRCs with and without mucinous component, but noticed a similar genetic profile between MACs and CRCs with mucinous component, irrespective of the percentage (if mucinous component more than 5%). The different genetic mutation profile related to MSI status was also identified between two groups of tumors. The most frequent mutations in CRCs with mucinous component are KRAS (28/49, 57.1%) and BRAF (19/49, 38.7%), PIK3CA (16/49, 32.6%), followed by APC (12/49, 24.5%) and TP53 (11/49, 22.5%). The combined mutation frequency of the two key factors in the EGFR signaling pathway, KRAS and BRAF, in the CRCs with and without mucinous component is 95.9 and 52.1%, respectively.

Conclusions

The dysregulation of EGFR pathway plays a critical role in the development of CRCs with mucinous component, irrespective of the percentage. The result suggested that the current cut off of 50% mucin component to define mucinous adenocarcinoma might be challengeable.

Cutoff of 25% for Ki67 expression is a good classification tool for prognosis in colorectal cancer in the AJCC‑8 stratification

Ki‑67 expression has been widely used in clinical practice as an index to evaluate the proliferative activity of tumor cells. The cutoff for Ki67 expression in order to increase the prognostic value of Ki67 expression in colorectal cancer varies. The present study assessed the relationship between the 25% cutoff for Ki67 expression and prognosis in colorectal cancer in the AJCC‑8 (American Joint Committee on Cancer 8 edition) stratification. The current trial included 1,090 colorectal cancer patients enrolled from 2006 to 2012 at Huzhou Central Hospital. Ki67 expression was classified according to 25% intervals, dividing the patients into four groups. Measurement data were analyzed by ANOVA, and count data by Crosstabs. Bivariate correlation analysis was performed to assess clinicopathological indicators based on Ki67 expression. Disease‑free survival (DFS) and overall survival (OS) based on Ki67 levels were analyzed by the Kaplan‑Meier method. A total of 1,090 patients of the 2,080 enrolled CRC cases were evaluated (52.4%). Invasive depth, tumor differentiation, tumor size, AJCC‑8, positive number of lymph nodes and chemotherapy status showed significant differences in the various Ki67 expression groups (all P<0.05), with significant correlations (Spearman rho: 0.170, 0.456, 0.22, 0.195, 0.514 and ‑0.201, respectively, all P<0.001). DFS and OS for the different Ki67 level groups based on AJCC‑8 stratification were analyzed, and no significance was found in stage IV (P=0.334). DFS and OS survival rates were assessed at different Ki67 expression levels, and no significant differences were found (all P>0.05). Cox regression analysis showed that invasive depth, lymph node metastasis, tumor differentiation, AJCC‑8 and Ki67 were independent factors affecting colorectal cancer (P=0.030, all others P<0.001). In conclusion, a cutoff of 25% for Ki67 expression is a good classification tool. High Ki67 has a close association with poor prognosis in colorectal cancer and independently predicts prognosis in the AJCC‑8 stratification.