Introduction
For two decades there has been an intensive debate about the human risk for increased cancer incidence when using insulin analogues [
1]. Insulin and insulin analogues act via the insulin receptor (IR), of which there are two isoforms, IRA and IRB, and to a lesser extent via the insulin-like growth factor 1 receptor (IGF1R). An increased residence time as well as an increased binding affinity of synthetic insulin-like molecules towards IRA and especially IGF1R might affect carcinogenesis [
2,
3]. Downstream signaling via these insulin receptor family members occurs via distinct pathways. Upon stimulation with insulin analogues with high affinity for IRB the PI3K/Akt pathway is activated, which is related to the metabolic role of insulin [
4]. In contrast, activation of the IGF1 pathway via IGF1R results in an upregulation of both the MAPK/Erk signaling cascade and an asymmetrical activation of the Akt pathway [
5], which is directly related to the limited mitogenic effect of insulin. Stimulation of the IRA results in a downstream signaling cascade similar to that observed after IGF1R activation [
5]. Epidemiological studies indicate a strong association between expression levels of both IGF1 and its receptor (IGF1R) and cancer initiation/progression [
6,
7]. This creates a situation whereby insulin analogues with increased affinity for IRA and/or IGF1R may increase the cancer hazard.
Current approaches to assess the intrinsic carcinogenic potential of insulin analogues are limited. The binding affinity of insulin analogues to both IR and IGF1R and their subsequent activation and overall mitogenic capacity are currently used as part of the risk assessment in terms of carcinogenic potential of newly developed insulin analogues. This
in vitro assessment is limited by the variability in the cell lines, culture conditions and proliferation assays that are used [
8,
9]. Furthermore, kinetic parameters, such as administration, distribution, metabolism and excretion, cannot always be captured in these
in vitro models. Some insulin analogues that are currently on the market have also been tested in two-year carcinogenicity studies in rats and/or mice [
10,
11]. So far only insulin X10 (also called AspB10) has been indicated to increase the tumor incidence in the chronic rodent studies and, consequently, it never reached the market [
1,
10,
11]. While most insulin analogues, including glargine [
12-
14], were negative in these chronic bioassays, several epidemiological studies showed an increased breast cancer risk [
15-
18], which could not be observed by others [
19-
26]. Altogether these data stirred concerns that growth factor-like insulin analogues are potential non-genotoxic carcinogens [
27].
Genetically engineered mouse (GEM) cancer models constitute powerful, alternative methods to assess the carcinogenetic potential of non-genotoxic compounds [
28]. This, in particular, involves GEMs with constitutive or conditional tissue specific deletion of tumor suppressor genes [
29]. We have described a mammary gland specific dominant negative mutated p53 mouse model, p53
R270H/+WAPCre [
30]. The model is based on a point mutation corresponding to the p53 mutated hotspot p53.R273H in the human
Li Fraumeni cancer syndrome. Mutant p53 is only expressed when Cre recombinase is induced by the whey acidic protein promoter (WAP) in the mammary gland. This leads to spontaneous mammary gland tumor formation initiated within a year.
Here we used the p53R270H/+WAPCre model to evaluate the carcinogenic potential of several insulin (like) molecules: insulin NPH, insulin glargine, insulin X10 and IGF1. We demonstrated that chronic exposure to insulin X10 and IGF1 significantly promotes mammary gland tumor development, while glargine and insulin do not. Yet, glargine-related tumors do have a different pro-mitogenic signaling that is distinct from control and insulin treated mice, and more reminiscent of insulin X10 and IGF1.
Methods
The p53R270H/+WAPCre mouse model
Heterozygous p53.R270H as well as the WAPCre mice were backcrossed with FVB mice over 15 times to yield a >99.99% FVB genetic background. Heterozygous conditional p53.R270H mice, >8-week-old, were crossed to transgenic WAPCre mice of the same age to generate p53
R270H/+WAPCre mice [
30,
31]. The mammary gland specific Cre recombinase splices out the intronic floxed stop cassette of the p53.R270H allele that eventually would lead to spontaneous mammary gland tumor formation of one-year-old p53
R270H/+WAPCre female mice. A high expression WAP promoter was used [
32], which is already active in the mammary gland of non-pregnant, non-lactating virgin mice. Therefore, we used only nulliparous mice in this experiment. A PCR/digestion based assay was used for genotyping using the same primers as previously described [
30] and subsequent restriction analysis using Hsp92II. The presence of the R270H mutation leads to digestion of the 486 bp p53 PCR amplicon into a 269 and 217 bp product. The presence of Cre recombinase was verified by a 676 bp amplicon [
30]. All mice were fed
ad libitum with RM1 diet (SDS, technilab-BMI, Someren, Holland).
Preparation of insulin, insulin analogues and IGF1 injection solutions
The treatments included: insulin NPH (Insuman Basal, Sanofi Aventis, Gouda, The Netherlands), insulin glargine (Lantus, Sanofi Aventis), insulin X10 (AspB10, Novo Nordisk, Alphen aan den Rijn, The Netherlands) and IGF1 (Increlex, Ipsen, Hoofddorp, The Netherlands). All compounds were dissolved in their original vehicle solutions: glargine (glycerol 0.2 mol/L, m-cresol 0.025 mol/L, ZnCl2 0.0002 mol/L adjusted to pH 4.0), insulin (glycerol 0.2 mol/L, NaH2PO4 0.00135 mol/L, phenol 0.0063 mol/L, m-cresol 0.0138 mol/L, ZnCl2 0.0001 mol/L adjusted to pH 7.4), X10 (glycerol 0.2 mol/L, phenol 0.0063 mol/L, m-cresol 0.0138 mol/L, ZnCl2 0.0001 mol/L adjusted to pH 7.4) and IGF1 (benzyl alcohol 0.083 mol/L, sodium chloride 0.1 mol/L, polysorbate 20 0.0016 mol/L, acetic acid 0.0072 mol/L, sodium acetate 0.05 mol/L adjusted to pH 5.4).
Experimental set-up
The experimental setup of the studies was examined and approved by the institute’s Ethical Committee on Animal Experimentation, in accordance with national and European legislation.
In a first panel of short term experiments, the maximal pharmalogical dose (MPD) was determined for each compound in our mouse model. In a 10 hour experiment (n = 54) the glucose drop was measured (Freestyle light, 70812–70, Abbott, Olst, The Netherlands) sequentially every hour after a single subcutaneous injection with insulin NPH, glargine, regular insulin, X10 or IGF1 (Additional file
1: Figure S1). A wide concentration range based on the literature was evaluated: insulin NPH and insulin glargine (25 nmol/kg to 125 nmol/kg); insulin X10 (480 nmol/kg to 1,800 nmol/kg); and IGF1 (5 mg/kg to 15 mg/kg) [
12]. In a subsequent experiment (n = 52) the effects of frequent injections were determined (Additional file
1: Figure S2). Besides blood glucose levels, the weight and overall well-being were determined during one month of daily injections.
Based on the above data, the long-term exposure experiment was designed in which the mice were injected with 50% and 80% of the MPD according to Additional file
2: Table S1. Once the female p53
R270H/+WAPCre mice were about eight-weeks old, they were randomly distributed in the dose groups using the program ‘randomice’. Mice were weighed every week, and a standard injection solution per compound/concentration was made to ensure an injection volume in the range of 60 to 130 uL. To avoid adverse reactions at the injection site due to frequent injection, three subcutaneous injection sites were used: neck, upper back and lower back. Mice were injected every other day for up to 67 weeks and palpated for tumors twice a week. A typical mammary gland tumor could be detected once it had a volume of about 8 mm
3. Dimensions were noted to monitor tumor growth. Once the tumor reached a volume of about 1 cm
3, the mouse was sacrificed and dissected. Tumors from other origins were generally difficult to palpate, so other fitness markers (weight loss, skin condition, motility, and so on) were used to decide when to sacrifice an animal.
Histopathology and immunofluoresence
When mice were destined for sacrifice based on tumor size or other markers, they were euthanized one day after the last injection, blood was collected and serum was extracted (mini collect, Greiner Bio-one B.V., Alphen aan de Rijn, The Netherlands) according to the manufacturer’s protocol. A quarter of the tumor, liver, lung, pancreas, kidney and spleen were fixed in a neutral aqueous phosphate buffered 4% solution of formaldehyde (Klinipath/VWR, Duiven, The Netherlands) for 24 hours and stored in 70% EtOH. These tissues were embedded in paraffin wax, sectioned at 5 μm, and stained with hematoxylin and eosin (H & E) for histopathologic evaluation. Immunofluorescence (IF) was performed on twenty representative tumors (based on the H & E characterization) from the different tumor types. Citrate buffer was used for antigen retrieval, blocking was performed in 10% normal goat serum (NGS). Antibodies were diluted (1:10 to 1:800) in 1% NGS buffer. Antibodies used were smooth muscle actin (A2547, Sigma-Aldrich, Zwijndrecht, The Netherlands), cytokeratin 5 (PRB-160P, Covance, BioLegend, London, UK), cytokeratin 8 (rdi-pro61038, Fitzgerald/Bioconnect, Huissen, The Netherlands), E-cadherin (610181, BD Transduction, Breda, The Netherlands). Fluorescently labelled secondary antibodies were from Jackson Laboratories, Suffolk, UK. Images were made on a Nikon Eclipse TS100 microscope. The histopathology was based on analysis of the macroscopic tumors. The epithelial and mesenchymal cell distributions were estimated by a pathologist on morphological characteristics of these cell types.
Western blotting
A quarter of the tumor was snap frozen in liquid nitrogen, stored at −80°C and used for subsequent protein expression profiling. A tumor piece was ground and lysed with 300 μL tumor lysis buffer (20 mM Tris–HCl, pH 7.4, 137 mM NaCl, 2 mM ethylenediaminetetraacetic acid (EDTA), 1% Triton, 10% glycerol, 1:100 protease inhibitor cocktail) for 30 minutes at 4°C. After centrifugation (10 minutes, 10,000 rpm), supernatant was collected and the protein concentration was determined (BCA TM Protein Assay Kit; Thermo Scientific, Rockford, IL, USA). The Western blotting procedures and all antibodies were identical to [
8]. In addition, antibodies targeting Her2 (Cell Signaling Technology, Danvers, MA, USA), β-actin, GAPDH, ER-α and EGFR, (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were used. The same endogenous control (EC) sample consisting of a mix of 30 random mammary gland tumor samples was used on each blot.
Statistical analysis
Statistical analyses of (mammary gland) tumor-free survival curves included calculation of censored Kaplan-Meijer distribution of survival of two different treatment groups and comparison by a two-sided log-rank test using Graphpad Prism version 4.00 software. The same software was used to determine P-values over the mean weight, percent of epithelial cells and average number of mammary gland tumors using the two-sided t-test. For the correlation plots linear regression was applied.
Discussion
In the present study, we used the p53R270H/+WAPCre mouse model to assess the carcinogenic potential of two insulin-like molecules, X10 (AspB10) and glargine in direct comparison to insulin NPH and IGF1. Our data indicate that the highly mitogenic compounds, insulin X10 and IGF1, which both stimulate the IGF1-receptor, significantly decreased the latency time for tumor development. This was not observed for glargine and insulin NPH. Moreover, we demonstrated that tumors derived after mitogenic insulin analogue treatment induce mammary gland tumors with enhanced intracellular signaling through either the Erk or Akt pathway, which was not observed in control animals. Our data indicate that the p53R270H/+WAPCre mouse model is sensitive to evaluate the intrinsic higher mitogenic potential of insulin-like compounds and the associated contribution to cancer development. Interestingly, while this mouse tumor model largely gives rise to EMT mammary gland tumors irrespective of insulin treatment conditions, the mitogenic insulin-like molecules drive the formation of tumors with enhanced key mitogenic signaling activity.
X10 as well as IGF1 significantly decreased tumor latency time. These two compounds also induced a significant weight increase. However, there is no overall correlation between tumor latency time and mouse weight at tumor detection date (Additional file
1: Figure S5), indicating that chronic treatment of X10 and IGF1 affects tumor latency directly by mitogenic signaling rather than indirectly by obesity.
Similar to insulin NPH, chronic exposure to insulin glargine did not significantly affect tumor latency time compared to the vehicle treatment, although tumors developed slightly earlier in the glargine treated animals. These glargine data are in line with a previous report by Stammberger
et al. in which a lifelong exposure of glargine did not show any difference in the incidence of mammary tumors reported in both mice and rats when comparing with the NaCl, vehicle-control, or the NPH insulin treated groups in wild type mice and rats [
13]. When conducting a pathology study on the different tumors, we found no difference between insulin NPH versus glargine induced tumors. These histopathology results are in agreement with a study performed by Besic
et al., in which a clinical and histopathological screening was performed on breast carcinomas of diabetic patients who were either on a glargine or other insulin (analogue) therapy [
37].
We previously showed in
in vitro models that insulin IGF1, X10 and glargine can strongly activate the IGF1R-mediated signaling, but that only IGF1 and X10 have an increased mitogenic potential [
8]. Insulin glargine did not significantly enhance cell proliferation, which was largely explained by the rapid metabolism of glargine, preventing a sustained activation of IGF1R-mediated mitogenic signaling [
8]. These insulin-like compounds have been tested for carcinogenic side-effects in wild type mice and rats [
10,
11,
38], but to our knowledge they have never been tested in sensitive humanized
in vivo models. Our current
in vivo findings are in agreement with these
in vitro observations. Yet, despite the fact that insulin glargine did not significantly enhance MG tumor development, many insulin glargine tumors also demonstrated enhanced Erk and Akt activity, which was hardly observed under control conditions. This suggests that insulin glargine is not inert and may affect the intracellular signaling in the developing tumor.
The majority of the obtained tumors in our p53
R270H/+WAPCre mouse model were classified as EMT tumors; these tumors are thought to be in the epithelial to mesenchymal transition state. This type of MG tumor is not a commonly found human breast cancer subtype. Nevertheless, as in human breast cancer, the intracellular signaling varied considerably between the various tumors. In particular, a strong enhancement of Erk and Akt activity was evident in the mitogenic insulin-like molecules groups, which was hardly observed for untreated mice. This variation induced by our insulin treatment conditions probably provides a better representation of human breast cancer in general and EMT tumors in particular. Hence, a mitogenic treatment setting might provide enhanced information on human breast cancer development. Yet, given the biased formation of EMT tumors in the current study setup, further carcinogenic studies with insulin-like molecules are required with other humanized mouse MG tumor models. This could involve genetically engineered mouse models (GEMMs) with a human specific mutation in, for example, the PI3K signaling pathway [
39].
Long-term administration of IGF1, X10, glargine and insulin led to tumors with significantly higher p-Akt levels compared to vehicle treated animals. This indicates that the PI3K signaling cascade is up regulated upon stimulation with the insulin-like molecules. Similarly the MAPK signaling pathway was up regulated after IGF1, X10 and glargine treatment. Interestingly, long-term stimulation with insulin did not affect the p-Erk1/2 levels in the obtained tumors.
At this moment we do not know the exact mechanism by which the more mitogenic insulin like molecules promotes MG tumor development. Since Erk and Akt activity did not per se coincide in the various tumors and/or relate to enhanced IGF1R levels, direct ligand-mediated activation of the Erk and Akt pathways seems unlikely. Possibly, mutations in either Erk and/or Akt pathway components are underlying the enhanced activation of these signaling molecules. In such a scenario, activation of the IGF1R and/or IR may promote the selection of the initiated cells that have incorporated mutations in Erk/Akt pathway components (for example, Ras or PI3K). This could provide a suggestion that treatment with human insulin analogues may initiate the development of mammary tumors with an altered mutational and/or signaling spectrum. More in-depth molecular analysis of the mouse tumors at the genome and proteome level is needed to further understand the underlying mechanism for the enhanced tumor formation by the IGF1 and X10 treatment conditions and their potential role in either enhanced tumor initation and/or progression. While we studied the effect of insulin analogues on MG tumor development, we cannot exclude that insulin analogues with high affinity for the IGF1R may also promote breast cancer progression, either locally or at distant sites, or modulate sensitivity to anticancer drugs. This needs further investigation and our different mouse MG tumor banks that show different expression levels of IGF1R and IR could contribute to this.
Conclusions
The p53R270H/+WAPCre mouse model is a sensitive and human relevant model to test the carcinogenic properties of insulin-like molecules, as is apparent with insulin X10 and IGF1. Insulin glargine was tested in this study and did not show a significantly decreased tumor latency time compared to insulin NPH, although the MAPK-signaling pathway was upregulated as found for X10 and IGF1. As is the case in humans, rapid conversion of glargine into metabolically active metabolites M1 (and to a lesser extent M2) is likely to be the reason for the low carcinogenic potential of subcutaneous injected glargine. All in all, based on the current tumor model, the data suggest that glargine users are not facing an increased carcinogenic hazard compared to insulin NPH users. Yet, future studies in mouse models that lead to more human relevant tumors remain important to fully exclude a role for current clinically relevant insulin analoques in the development and/or progression of human breast cancer.
Acknowledgements
We thank Dr. Virginie Boulifard (Ipsen, France) for providing the Increlex, IGF1. We are grateful to Dr. Norbert Tennagels and Dr. Ulrich Werner (Sanofi-Aventis, Germany) for providing us with insulin glargine, insulin basal and vehicle solutions for the exposure experiments. We thank Dr. Bo Falck Hansen (Novo Nordisk, Denmark) for providing the X10. We would like to thank Conny van Oostrom, Silvia Neggers and Ilma Rietbroek for excellent technical assistance during the in vivo experiments and Saskia van der Wal-Maas for her technical help with the pathology. Ben ter Braak is acknowledged for critically reviewing our manuscript. This study was funded by the National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands in the Strategic Research Program SOR 2010 (S/360003).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
Conceived and designed the experiments: BTB, CS, HVS, BVDW, JWVDL. Performed the experiments: BTB, CS, ENS, EK, DS. Analyzed the data: BTB, EK, DS. Wrote the paper: BTB. Reviewed and corrected paper: CS, HVS, DS, EK, ENS, BVDW, JWVDL. All authors read and approved the final manuscript.