Neoadjuvant chemotherapy affects molecular classification of colorectal tumors

Abstract

The recent discovery of 'molecular subtypes' in human primary colorectal cancer has revealed correlations between subtype, propensity to metastasize and response to therapy. It is currently not known whether the molecular tumor subtype is maintained after distant spread. If this is the case, molecular subtyping of the primary tumor could guide subtype-targeted therapy of metastatic disease. In this study, we classified paired samples of primary colorectal carcinomas and their corresponding liver metastases (n=129) as epithelial-like or mesenchymal-like, using a recently developed immunohistochemistry-based classification tool. We observed considerable discordance (45%) in the classification of primary tumors and their liver metastases. Discordant classification was significantly associated with the use of neoadjuvant chemotherapy. Furthermore, gene expression analysis of chemotherapy-exposed versus chemotherapy naive liver metastases revealed expression of a mesenchymal program in pre-treated tumors. To explore whether chemotherapy could cause gene expression changes influencing molecular subtyping, we exposed patient-derived colonospheres to six short cycles of 5-fluorouracil. Gene expression profiling and signature enrichment analysis subsequently revealed that the expression of signatures identifying mesenchymal-like tumors was strongly increased in chemotherapy-exposed tumor cultures. Unsupervised clustering of large cohorts of human colon tumors with the chemotherapy-induced gene expression program identified a poor prognosis mesenchymal-like subgroup. We conclude that neoadjuvant chemotherapy induces a mesenchymal phenotype in residual tumor cells and that this may influence the molecular classification of colorectal tumors.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Figure 1

Discordant classification of primary colorectal tumors and corresponding liver metastases. (a) The tissue microarray (TMA) was constructed from the resection specimens of primary colorectal tumors and the resection specimens of colorectal liver metastases of 129 patients. Tumor-rich areas were identified via haemotoxylin and eosin stainings and three cores of 0.6 mm were obtained per tumor type. Digital images of TMA immunohistochemically stained slides were obtained via an Aperio Scanscope XT system (Leica Biosystems, Wetzlar, Germany). These were automatically analyzed as described before. Cores with a random forest probability of 60% were scored as 'mesenchymal-like'. Patient subtypes were determined using majority consensus. Here a pie chart shows the distribution of epithelial-like and mesenchymal-like of the primary colorectal tumors in our paired tumor cohort. (b) Patient characteristics of epithelial-like tumors were compared to mesenchymal-like tumors. Age was compared via the Wilcoxon rank sum test, for all other variables the X2-test was used. Minus log10 P-values were calculated and are shown in the graph. Table 1 shows the detailed list of patient characteristics. (c) Kaplan–Meier survival curves of the overall survival after liver resection, calculated with a log-rank test (P=0.276). The blue line represents the patients with epithelial-like colorectal tumors and the green line represents patients with mesenchymal-like colorectal tumors. (d) Pie chart of the classification of liver metastases of our paired tumor cohort. (e) Relationship between the classification of primary colorectal tumors and the corresponding liver metastases. (f) The influence of chemotherapy on the classification of paired tumors. Neoadjuvant chemotherapy for the primary colorectal tumors or liver metastases and adjuvant therapy for the primary CRC were all univariate analyzed via the X2-test. Chemotherapy was considered neoadjuvant if it was given to downsize the tumor, primary or metastases, prior to the surgery. Adjuvant chemotherapy is chemotherapy given after the initial resection of the primary colorectal tumor. Concordant tumor pairs were depicted to discordant tumor pairs, which were separated in switching from epithelial-like to mesenchymal-like and vice versa. Minus log10 P-values were calculated and depicted in this graph.

Figure 2

5-FU-based chemotherapy is associated with a mesenchymal tumor phenotype. (a) In the liver metastases data set two groups were made, chemotherapy naive and chemotherapy exposed, these were compared in distribution of the CMS classification. The bar graph shows chemotherapy before surgery is associated with an increased proportion of mesenchymal-type tumors (CMS4). The CMS subgroups are CMS1: orange, CMS2: blue, CMS3 pink, CMS4: green. (b) The chemotherapy-induced genes were used to cluster the tumors of the CMS cohorts into chemo-induced high and low subgroups (K-means option in R2, using a two group separation) based on single gene P-values. All tumors had previously been classified into molecular subtypes. The graphs show the distribution of the CMS subtypes within the chemo-induced high and low subgroups. The chemo-induced high subgroup is enriched for mesenchymal subtypes (CMS4). (c) Kaplan–Meier curves showing the differences in relapse-free survival between the chemo-induced high and low subgroups in the CMS-3232 cohort.

Figure 3

Chemotherapy induces mesenchymal gene expression in patient-derived colonospheres. (a) Liver metastasis-derived colonospheres were treated with 5-FU for six cycles. RNA was isolated from control (n=5) and 5-FU-treated cells (n=5), and were analyzed by gene expression profiling. The heat map shows all genes that were significantly (P<e−6) upregulated (68) or downregulated (36) in post-treatment tumor cells. See Supplementary Table 2 for a full list of genes. (b) Expression of the 5-FU-induced gene set (68 genes; Supplementary Table 2) was correlated with gene sets reflecting either an epithelial or a mesenchymal tumor phenotype in the data set of the same experiment. Correlations were assessed by using the ‘gene set versus gene sets’ option in the R2 genomics analysis and visualization platform. Gene sets reflecting a mesenchymal tumor cell phenotype positively correlate with the 5-FU-induced gene set (indicated in red), while gene sets reflecting an epithelial phenotype show a negative correlation (in blue). The P-values of the correlations are indicated as minus log10 P-values in green bars. (c) The 68 5-FU-induced genes were used to cluster the tumors of the CMS cohorts into 5-FU-induced high and low subgroups (K-means option in R2, using a two group separation) based on single gene P-values. All tumors had previously been classified into molecular subtypes. The graphs show the distribution of the CMS subtypes within the 5-FU-induced high and low subgroups. The 5-FU-induced high subgroup is enriched for mesenchymal subtypes (CMS4). (d) Kaplan–Meier curves showing the differences in relapse-free survival between the 5-FU-high and 5-FU-low subgroups in both cohorts. (e) Expression of the experimental-derived gene set of 5-FU-induced genes was compared to expression of genes upregulated in chemotherapy-exposed liver metastases by using the ‘relate 2 tracks’ option in the R2 genomics analysis and visualization platform. The XY-plot shows the correlation of the expression of both gene sets (R=0.80, P=2.0e−110) in the CIT subset of the CMS cohort. The CMS subgroups are CMS1: orange, CMS2: blue, CMS3 pink, CMS4: green.