Insulin-like growth factor 1 receptor activation promotes mammary gland tumor development by increasing glycolysis and promoting biomass production

Abstract

Background: The insulin-like growth factor 1 (IGF1) signaling axis plays a major role in tumorigenesis. In a previous experiment, we chronically treated mice with several agonists of the IGF1 receptor (IGF1R). We found that chronic treatment with insulin analogues with high affinity towards the IGF1R (IGF1 and X10) decreased the mammary gland tumor latency time in a p53^R270H/+WAPCre mouse model. Frequent injections with insulin analogues that only mildly activated the IGF1R in vivo (glargine and insulin) did not significantly decrease the tumor latency time in this mouse model.

Methods: Here, we performed next-generation RNA sequencing (40 million, 100 bp reads) on 50 mammary gland tumors to unravel the underlying mechanisms of IGF1R-promoted tumorigenesis. Mutational profiling of the individual tumors was performed to screen for treatment-specific mutations. The transcriptomic data were used to construct a support vector machine (SVM) classifier so that the phenotypic characteristics of tumors exposed to the different insulin analogue treatments could be predicted. For translational purposes, we ran the same classifiers on transcriptomic (micro-array) data of insulin analogue-exposed human breast cancer cell lines. Genome-scale metabolic modeling was performed with iMAT.

Results: We found that chronic X10 and IGF1 treatment resulted in tumors with an increased and sustained proliferative and invasive transcriptomic profile. Furthermore, a Warburg-like effect with increased glycolysis was observed in tumors of the X10/IGF1 groups and, to a lesser extent, also in glargine-induced tumors. A metabolic flux analysis revealed that this enhanced glycolysis programming in X10/IGF1 tumors was associated with increased biomass production programs. Although none of the treatments induced genetic instability or enhanced mutagenesis, mutations in Ezh2 and Hras were enriched in X10/IGF1 treatment tumors.

Conclusions: Overall, these data suggest that the decreased mammary gland tumor latency time caused by chronic IGF1R activation is related to modulation of tumor progression rather than increased tumor initiation.

Keywords: Hallmarks of cancer; IGF1R; Mammary gland tumor; Next-generation sequencing; Warburg.

Fig. 1

Experimental overview. a Overview of the chronic insulin analogue exposure experiment. b The Kaplan-Mayer MG tumor-free mice plots with the median tumor latency per treatment group; the colored dots indicate the selected tumors for the transcriptomic analysis. c the insulin-like growth factor-1 receptor (igf1r), A isoform of the insulin receptor (ira), and B isoform of the insulin receptor (irb) gene expression levels in MG tissue of young mice (8 weeks), old mice (~50 weeks), and in MG tumors (first graph); the second plot (right) shows the receptor gene expression distribution in the MG tumor tissue of mice chronically exposed to different insulin analogues. d Heat map of the next-generation sequencing data showing hierarchical sample clustering by sample-to-sample distance; the lower table shows the Spearman rank correlation coefficients within the treatment groups (bold) and the coefficient between the different treatment groups (averaged per condition). e The hallmarks of cancer with the features highlighted that we will discuss in view of the chronic insulin analogue exposure experiment. **P < 0.01, ***P < 0.001. ns not significant. (Adapted from Hanahan and Weinberg [9])

Fig. 2

Sustained growth signaling in mammary gland tumor tissue of chronically insulin analogue exposed mice. a The INSR/IGF1R signaling pathway with receptors, downstream targets, and the biological effect. b Hierarchical clustering (Euclidian distance) of the INSR/IGF1R-specific gene expression per MG tumor. The pie diagrams show the distribution of the different treatments in the two clusters. c Bar graph of SVM simulation on the predicted proliferation potential per treatment of MG tumors of the chronically exposed mice (i) and human breast cancer cells (MCF7 IGF1R) exposed for 1 h to the indicated insulin analogue (ii). *P < 0.05. ns not significant

Fig. 3

Tissue invasion and metastasis in mammary gland tumor tissue of chronically insulin analogue exposed mice. a Hierarchical clustering (Pearson correlation) of genes involved in epithelial to mesenchymal transition (EMT) or mesenchymal to epithelial transtition (MET) per MG tumor. The pie diagrams show the distribution of the different treatments in the three clusters. b The E-cadherin/HOECHST immunofluorescent hematoxylin and eosin (H&E) pathology slides of three example tumors showing epithelial or mesenchymal characteristics. c Bar graph of SVM simulation on the predicted migration potential per treatment of MG tumors of the chronically exposed mice (i) and human breast cancer cells (MCF7 IGF1R) exposed for 1 h to the indicated insulin analogue (ii). *P < 0.05. ns not significant

Fig. 4

Warburg effect in mammary gland tumor tissue of chronically insulin analogue treated mice. a Hierarchical clustering (Pearson correlation) of genes involved in glycolysis or oxidative phosphorylation per MG tumor. The pie diagrams show the distribution of the different clusters per treatment group. b Table with the metabolic pathways that were significantly down- or upregulated in the X10/IGF1 treatment groups compared to the vehicle, insulin, and glargine treatment groups. c Table with the metabolic pathways that were significantly down- or upregulated after X10/IGF1 exposure compared to vehicle, insulin, and glargine treatment in the MCF7 IGF1R model

Fig. 5

Predicted biomass production rate is increased in tumors of chronically X10/IGF1-treated mice and could possible explain the decreased tumor latency time in these treatments. a Bar plot of the mean tumor latency time in weeks per chronic treatment. b Bar plot of the normalized biomass production rate per treatment

Fig. 6

Genetic instability in mammary gland tumor tissue of chronically insulin analogue-exposed mice. a The bar plot shows the average number of mutations per MG tumor for all chronic treatments; the dot plot indicates that there is no correlation between the number of mutations of a specific tumor and tumor latency time. b The same as in a, but here we focus on clinically relevant human tumor driver mutations. c Some specific tumor driver mutations are featured in these bar plots. The first bar plot represents the mutations that are enriched in the X10/IGF1 treatment groups; in the second bar plot, the mutations are highlighted that are under-represented in the vehicle treatment group. N shows the number of tumors in which this specific mutation was present