Background
Colorectal cancer (CRC) is the fourth most common cancer worldwide and the second leading cause of cancer mortality [
1]. Clinical decision making in CRC is mainly driven by clinical and traditional pathological features including TNM staging. Although these features hold considerable prognostic, and even predictive value, there are profound individual differences in clinical outcome within a single tumor stage, especially for stage II and III [
2]. Also, there is compelling evidence that not all cancers follow the linear-progression model associating with the TNM-stages. For example, in CRC the majority of lymphatic and distant metastases arise from independent subclones, and 40–63% of metachronous metastases develop in patients without lymph node metastasis [
3]. The consensus molecular subtype (CMS) classification is a widely studied transcriptome-based stratification system for CRC defining four disease entities (CMS 1–4) with distinct clinical, biological and molecular features [
4]. Hence, the CMS taxonomy could offer a framework to elucidate whether TNM solely resembles disease progression or also biologically different entities that preferentially present with a specific stage of disease at diagnosis. This study was conducted to investigate the interconnectivity between tumor stage and tumor biology in CRC patients. Subsequently we demonstrate the added value of this knowledge in patients with high-risk stage II colon cancer, a subgroup in which accurate prognostication and selection for adjuvant treatment is still an unmet need.
Methods
Patients and data aggregation
Patients for which information on staging and microsatellite instability (MSI) status was available were selected from the previously reported meta-cohort of Guinney et al. [
4], resulting in 1040 individual patients (accession number GSE39582 [
5] and TCGA). For validation of our findings the chemotherapy naïve stage II CRC patients of the MATCH Cohort [
6] and AMC-AJCCII-90 Cohort (accession number GSE33113) [
7] were used. In the validation cohort high-risk was defined as either T4 or inadequate lymph node assessment (< 10 nodes assessed).
The R2: Genomic Analysis and Visualization Platform was used to extract the aggregated and normalized data (
http://r2.amc.nl).
Differential gene expression analysis
The limma package was used to identify differentially expressed genes (DEG) between the different tumor stages and CMS groups, using the ANOVA test for overall DEG and a limma-test for individual groups. P-values were FDR corrected. For comparing the number of DEG between the overall cohort and CMSs, a random set of 200 patients was sampled 1000 times to correct for the effect of group size on the number of DEG.
Gene signatures
Gene signatures for advanced stage and CMS4 were built using the top 100 DEG (with the lowest FDR corrected p-value) between early (stage I-II) and advanced stage (stage III-IV), and CMS1/2/3 and CMS 4. Gene signature scores were built using the weighted matched z-score of both the up- and downregulated genes of the gene signatures.
Statistical analysis
The Chi-square test was used to assess associations between CMS classification and tumor stage. The Kaplan-Meier method was used to estimate survival. Survival curves were compared using the log-rank test. Disease-free survival (DFS) times of > 60 months were censored at 60 months. We performed a multivariate analysis using a Cox proportional-hazards model with CMS, gender, age, tumor location, T-stage and MSI status as covariables. All statistical tests were 2-sided and considered significant at a P-value lower than 0·05. All analyses were performed using R version 3.6.1.
Discussion
At present we are moving towards a more personalized medicine approach for the treatment of cancer. However, at this stage TNM staging is still the single most important feature guiding treatment decisions for CRC. The CMS classification is a promising classification system for CRC, identifying four subtypes with distinguishing biological features. CMS classification might be a relevant addition to TNM staging in order to provide an optimal treatment strategy for individual patients. Our findings support the hypothesis that tumor stage as defined by TNM, in addition to disease progression, resembles different biological entities. This adds to the argument that the CMS taxonomy is a potential framework to further tailor the prognostication and treatment of patients with CRC.
We have observed a difference in distribution for the CMS within the different TNM stages with mainly a decrease in CMS1 and a profound increase of CMS4 patients with advancing stages of disease. This is in line with the overall good prognosis of the CMS1, which are mainly MSI tumors, and the poor prognosis of the mesenchymal CMS4 subtype [
4]. This may suggest that the poor prognosis for increased stages of disease is (in part) explained by the aggressive tumor biology of CMS4, given the poor disease outcome of CMS4 compared to CMS1–3 cancers. The aggressive nature of the mesenchymal subtype was also demonstrated by a higher progression rate for CMS4 compared to the other subtypes (Fig.
2d).
When stratified for CMSs, we observed a marked decrease in differentially expressed genes between the different tumor stages. Furthermore, a high correlation between the two gene signature scores for stage III/IV and CMS4 was demonstrated. This indicates that at least part of the biological differences between tumor stages are explained by the CMSs. Which in turn supports the hypothesis that different tumor stages are largely driven by tumor biology rather than disease progression.
Furthermore, we showed a possible and valuable clinical implication of the molecular subtypes for the high-risk stage II patients. Current guidelines recommend to consider adjuvant chemotherapy for these patients [
8], which is based on literature showing (limited) prognostic value but no predictive value for the high-risk variables [
9‐
13]. The overt difference in DFS for the CMS4 subtype in the subgroup of high-risk stage II patients suggests that CMS subtyping may be of added value to identify patients that have a high-risk, lymph node negative colon cancer. This effect might partly be explained by stage migration, due to under-staging as a result of low number of assessed lymph nodes; i.e. high-risk stage II tumors contain unrecognized stage III tumors. Another possible explanation for the marked difference in DFS within the CMS4 population is that these tumors behave more like the early-dissemination model [
3,
14], instead of the classical linear-progression model in CRC. In agreement, the existence of early disseminating cancer cells which evolve independently at the metastatic site has been demonstrated in breast cancer [
15]. Therefore CMS4 tumors may benefit from treatment with chemotherapy at an apparently early stage of progression (stage II).
Several clinical studies found that patients with synchronous and metachronous liver metastases had a similar overall survival upon diagnosis of metastatic disease [
16‐
18]. This supports our hypothesis that tumor biology is installed at an early moment in tumor development and that this, rather than the progression over time, is the main determinant for prognosis in these patients. Also, determining the CMS may not only be helpful to identify high-risk stage II patients, but may also be used to select patients for specific treatments. Patients with an MSI tumor (mostly CMS1) are known to have very limited benefit from chemotherapy [
19,
20]. However, these patients may very well benefit from immunotherapy or the addition of Bevacizumab instead of Cetuximab to chemotherapy in metastasized CRC [
21,
22]. For epithelial-like tumors (CMS2/3) there is a predictive value for anti-EGFR therapy [
7,
23]. Patients with a CMS2 tumor were shown to be responsive to Oxaliplatin-containing chemotherapy while mesenchymal tumors (CMS4) seemed refractory to 5FU-based chemotherapy. These results suggest that the CMS taxonomy may also be used to select patients for conventional chemotherapy [
24,
25]. Future prospective studies should be conducted to confirm these hints on CMS-specific drug sensitivity, as these findings originate from retrospective studies.
The current study has several limitations. First, the survival analysis in the subset of stage II colon cancer may be subject to selection bias. Patients with high-risk stage II colon cancer were excluded from the current analysis when they did receive adjuvant chemotherapy. However, on estimate only 10–15% of these patients actually receive adjuvant chemotherapy, and patients with a T4 tumors and inadequate lymph node assessment (both high-risk factors) were present in the aggregated cohorts. Second, the additive value of the CMS for high-risk stage II patients should be validated in larger series given the relatively small number of patients in the high-risk stage II cohort.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.