Insulin-like growth factor receptor and sphingosine kinase are prognostic and therapeutic targets in breast cancer

Cell culture reagents were purchased from Trace Biosciences (North Ryde, New South Wales, Australia) and Nunc (Roskilde, Denmark). Bovine insulin, methanol, calcium chloride, magnesium chloride, crystal violet powder and 1-(4,5-Dimethylthiazol-2-yl)-3,5-diphenylformazan were purchased from Sigma-Aldrich. Enhanced chemiluminescence (ECL) reagent was SuperSignal West Pico Chemiluminescent Substrate (Pierce Biotechnology). The dual IGF1R/insulin receptor tyrosine kinase inhibitor (OSI-906; also referred to as linsitinib) was purchased from MedChem Express (Princeton, NJ) and the SphK inhibitor 2-(p-hydroxyanilino)-4-(p-chlorophenyl)thiazole (SKI-II) from Calbiochem [21]. Antibodies raised against phospho-Y1135/1136 IGFR1, IGFR1 beta chain, phospho-Ser473 AKT and total AKT, 4E-BP1 and eIF4E were purchased from Cell Signaling Technology (Beverley, MA). The antibody to detect SphK1 (ab16491) for western blots was purchased from Abcam and for SphK1 immunohistochemistry, from Abgent (AP7237c).

Breast cancer tissues were obtained from the Australian Breast Cancer Tissue Bank (ABCTB), Westmead, NSW, Australia for the purposes of this study. This study was approved by the Human Research Ethics Committee of the Northern Sydney Local Health District (Reference Numbers: RESP/15/125 and LNR/15/HAWKE/182) for the analysis of human breast cancer tissues samples obtained from the Australian Breast Cancer Tissue Bank. All samples obtained from this bank were de-identified and were from donors who had given written informed consent for their banked tumor tissue to be used in future research projects.

A total of 236 FFPE breast tissue samples were approved for use, comprised of five tissue micro-arrays (TMA) in duplicate or triplicate cores (0.6 mm³) (187 patients in total) and 49 whole face tissue sections. All patient samples had molecular subtyping from the ABCTB for ER, PR and HER2 expression by IHC and/or FISH analysis (for HER2) (Table 1). Patient information provided by the ABCTB included gender, disease status (all reported as invasive), pathology notes where applicable, primary histologic diagnosis and histopathological grade (Table 1). Patient follow-up data provided by the ABCTB consisted of diagnosis age, year of first breast event, time of follow-up since diagnosis and follow-up status (median follow up; 61 months (Table 1). In addition the ABCTB provided information relating to the therapy the patients received included the following: 1) anti-endocrine therapy; 2) HER2-therapy and 3) chemotherapy (Table 2). All studies were performed with approval from the Northern Sydney Local Health District (NSLHD) Human Research Ethics Committee (HREC), which assessed it as a low-negligible risk study.

IHC was performed on 4 μm FFPE sections using an automated tissue stainer (Autostainer, DAKO, Glostrup, Denmark) according to standard manufacturer’s operating procedures. Antigen retrieval was performed using a water bath heated to 99.2 °C for 20 min in freshly made 10 mM citric acid monohydrate adjusted to pH 6.0. The sections were quenched in 0.3% hydrogen peroxide for 5 min, blocked with 5% goat serum for 30 min and incubated in primary antibodies: IGF1Rβ antibody (no cross-reaction with the insulin receptor (InsR)) (#3027, Cell Signaling, Danvers, MA, USA 1:100), p-IGF1R (#ab39398, Abcam, Melbourne, VIC, Australia, 1:200) and SphK1 (#AP7237c, Abgent, San Diego, CA, USA, 1:200) for one hour at room temperature. Protein detection was subsequently performed using the DAKO-Envision Dual Link Labelled Polymer (Anti-Rabbit) (#K5007, Dako, Botany, NSW, Australia) for 30 min and the ImmPACT NovaRed Peroxidase Substrate Kit (#SK-4805, Vector Laboratories, Burlingame, CA, USA) for 10 min at room temperature. All antibodies were optimized using a series of dilutions on a TMA comprised of ten ER-positive and negative breast cancer patient tissues in duplicate to determine an optimal dilution for IHC staining. Final dilutions were closely assessed for specific membranous, cytoplasmic and/or nuclear staining in line with the literature. Negative controls were included for all IHC using a rabbit immunoglobulin fraction (#X0936, DAKO) at the final concentrations of the primary antibodies and a tissue sample incubated with the anti-rabbit antibody in the absence of a primary antibody.

Manual scoring was assessed on all samples for subsequent statistical analysis, with examples shown in Fig. 1. IGF1R and p-IGF1R staining expression levels were manually assessed using the HER2 scoring system described in the Hercep Test manual (DAKO) as follows: no staining = 0, faint staining = 1, weak to moderate staining = 2 and strong staining = 3, in line with published studies [27, 28]. Positive staining was defined as membrane/cytoplasmic and/or nuclear staining detectable in ≥10% of tumor epithelial cells. SphK1 staining levels were manually assessed as: no staining (<10% of tumor epithelial cells with cytoplasmic staining) = 0, weak = 1, moderate = 2 and strong = 3, in accordance with previously published data [29‐31].

Kaplan-Meier (KM) survival analysis was performed using SPSS v.22 (IBM, Armonk, NY, USA) to determine prognostic significance of high protein expression of phospho- and total IGF1R, and SphK1, for overall survival (OS) and disease-free survival (DFS) in non-stratified patient cohorts, and after stratification for high or low ER, PR, and HER2. KM analysis was also performed to determine the relationship between IGF1R and SphK1 protein expression in patients stratified for anti-endocrine therapy treatment. For KM analysis of IGF1R, p-IGF1R, and SphK1, staining scores described above, on a scale of 0 to 3, were converted to a new binary scale where 0 or 1 = low, and 2 or 3 = high. For analysis of IGF1R-SphK1 and p-IGF1R-SphK1 co-expression, a binary co-expression score was used: either or both analytes low = low co-expression, both analytes high = high co-expression. Significance (P < 0.05) was determined by the log-rank test. Multivariate survival analysis was performed in SPSS by the Cox proportional hazards method, using forward stepwise (likelihood ratio) regression, with significance set at P < 0.05. To determine the relationship between IGF1R and SphK1 in non-stratified and stratified patient groups including: ER+/−, PR+/− and HER2+/−, age, tumor grade correlation analysis was performed in SPSS using Spearman correlation analysis with statistical significance set at P < 0.05. All data used in these analyses are included as Additional file 1: Table S1.

The ER-positive MCF7 and T47D and ER-negative HCC1806 and HCC70 breast cancer cell-lines were purchased from the American Tissue Culture Collection (Manassas, VA, USA) and cultured and maintained in phenol-red RPMI-1640 medium supplemented with 5% FBS, 15 mM Hepes and 10 mg/mL bovine insulin at 37 °C in humidified 5% CO₂ atmosphere.

The MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay and clonogenic survival assay were performed by plating 2 × 10³ MCF7, T47D, HCC1806 and HCC70 cells per well of a 96-well plate (MTT-assay) or a 6-well plate (clonogenic assay) in phenol-red RPMI culture media supplemented with 5% FBS and culturing for 24 h. The cells were treated with culture media containing 5% FBS in addition to OSI-906 (0, 0.1, 0.4, 1.6 or 6.4 μM) and/or SKI-II (0, 0.16, 0.8, 4, 10 or 20 μM) in combination for 96 h for MTT-assay and 10-14 days for clonogenic assay. For the MTT-assay 30 μl of 1-(4,5-dimethylthiazol-2-yl)-3,5-diphenylformazan (thiazolyl blue formazan) solution prepared at 2.5 mg/ml, in PBS (containing 0.9 mM calcium chloride plus 0.5 mM magnesium chloride) was added per well and incubated at 37 °C for 4 h. The cells were solubilized in 150 μl of DMSO for 15 min at room temperature (RT). Absorbance was read using a plate reader at 460 nm. For the clonogenic assay, the cells were washed twice in PBS (0.9 mM calcium chloride plus 0.5 mM magnesium chloride) solution at RT, fixed in 0.1% crystal violet prepared in final 20% methanol solution and destained in tap water. The cells were left to air-dry overnight and images obtained using FujiFilm Luminescent Image Analyzer LAS-300 and single colonies were counted using open colony forming unit (CFU) software (http://​opencfu.​sourceforge.​net/​).

3 × 10⁵ MCF7 and HCC-1806 breast cancer cell-lines were plated per well of 6-well plates, cultured for 24 h and subsequently treated with OSI-906 (0.1, 0.4 or 1.6 μM) and/or SKI-II (4 μM) for 24 h. Protein lysates were prepared and 20 μg of protein was further run on SDS-polyacrylamide gels, transferred to Hybond C nitrocellulose and probed with antibodies to detect phospho and/or IGF1R, AKT, 4E-BP1, eIF4E (Cell Signaling) and SphK1 (Abgent) steady-state protein levels as previously described [32]. Β-actin antibody was used as a loading control. Subsequent to the addition of the ECL-reagent, a FujiFilm Luminescent Image Analyzer LAS-300 (Stamford, CT) was used for band detection and image production.

The effect of drug treatments on MTT and clonogenic assays was first analysed across the full OSI-906 dose-range using 2-way ANOVA for repeated measures (SPSS) with experiment and SKI-II dose as factors, and OSI-906 dose as the repeated measure. Comparison of individual dose combinations was performed using 1-way ANOVA with Tukey’s post-hoc test. (GraphPad Prism v.7, La Jolla, CA, USA). Statistical significance was defined as a p-value <0.05*, p < 0.01**, p < 0.001*** and p < 0.0001****. Drug synergism was determined using the Chou Talalay method to calculate the combination index (CI) <1 [33].

In FFPE tissue sections, high total IGF1R protein expression, in isolation or combined with SphK1, was associated with a better overall survival (OS) rate when analyzed across all patients suggesting IGF1R, alone or together with SphK1, acts as a positive prognostic indicator (Table 3, p = 0.018, Fig. 2a and p = 0.028,; Fig. 2b, respectively). These highly significant effects on OS were not seen for DFS (data not shown). No significant relationship between SphK1 expression alone and OS was identified (Table 3). In contrast, no significant improvement in OS was detected in patients with high p-IGF1R protein expression alone or combined with SphK1 in our patient analysis (Table 3, p = 0.303, Fig. 2c and p = 0.118, Fig. 2d, respectively). However, pIGF1R and SphK1 high co-expression did lead to lower p-value compared to either protein in isolation (Table 3). After stratification for ER status we found, in line with previous studies, an improved OS in ER-positive breast cancer patients who express high IGF1R protein in isolation (Table 3, p = 0.048, Fig. 3a) or combined with SphK1 (Table 3, p = 0.051, Fig. 3b), compared to ER-negative breast cancers (Table 3, p = 0.445, Fig. 3c and p = 0.582, Fig. 3d, respectively). This effect was further supported by a strong positive correlation between IGF1R and SphK1 in ER-positive (p = 0.001), but not ER-negative (p = 0.936) breast cancer patients (Table 4). In support of the literature reporting that pIGF1R can be detected in the nucleus of the cell, bind DNA and act as a transcription factor, we also found that some patient samples showed positive nuclear IHC staining for pIGF1R (Additional file 2) [34].

After stratification of tumors as either HER2 +/− or PR +/−, there was no significant prognostic effect of IGF1R (Additional file 3A-D) or p-IGF1R (Additional file 4A-D) in isolation, or combined with SphK1 expression (data not shown), in our patient cohort. Similarly, no significance was observed for high vs. low SphK1 protein expression alone following stratification for HER2 or PR (Table 3). However, as shown in Table 4, a significant association was observed between both p-IGF1R and IGF1R expression and ER-positivity (p = 0.007 and p < 0.001, respectively), and between IGF1R expression and PR-positivity (p < 0.001). Furthermore, p-IGF1R expression was negatively correlated to HER2-positive expression (Table 4, p = 0.002), yet this was not evident for IGF1R expression. Moreover, SphK1 high expression was inversely associated with both ER-positivity (p < 0.001) and PR-positivity (p = 0.003) in breast tumors (Table 4) which supports literature findings [35]. We were not able to identify any association between IGF1R or SphK1 and tumor grade; however, p-IGF1R was inversely correlated to tumor grade in our analysis (Table 4, p = 0.004). Lastly, we identified an inverse correlation of IGF1R expression to following stratification for age (i.e. ≥51y; average age of menopause) in our analysis (Table 4, p = 0.025).

To determine the combined effect of the IHC measurements and hormone receptor status on patient survival we undertook multivariate Cox regression (proportional hazards) analysis, using the forward stepwise (likelihood ratio) method. Analyzing the influence of IGF1R, p-IGF1R and SphK1 alone on overall patient survival, only IGF1R was significant (Table 5; p = 0.008). None of these variables was significant for DFS (Table 5). When ER and PR status were added to the model, IGF1R remained the only significant variable (Table 5: p = 0.010 for OS, not significant for DFS). However, adding ER, PR and HER status to the model, HER2 became the dominant variable when OS was the survival endpoint (Table 5; p < 0.001), with IGF1R expression approaching significance (Table 5; p = 0.053). For DFS, only HER2 was significant among the six variables (Table 5; p = 0.007).

Given that we identified a relationship between high IGF1R protein expression alone and combined with high SphK1 expression, and overall patient survival, that was most evident for ER-positive breast cancers, we undertook further analysis to determine whether a similar relationship existed when samples were stratified for anti-endocrine therapy. KM analysis identified that anti-endocrine therapy is associated with an improved OS (Table 3, p = 0.006, Fig. 4a) in the unstratified cohort. Examining only samples from anti-endocrine therapy treated women, high IGF1R expression (Table 3, p = 0.028, Fig. 4b), and high IGF1R/SphK1 co-expression (Table 3, p = 0.034, Fig. 4c) were both prognostic for improved OS, whereas high SphK1 expression was not (Table 3, p = 0.101, Fig. 4d). This supports our previous findings that high IGF1R/SphK1 co-expression is associated with an improved disease outcome specifically in ER-positive breast cancer patients in relation to anti-endocrine therapy.

The observation that high IGF1R and SphK1 expression are prognostic for improved overall survival is paradoxical, given that the literature suggests IGF1R and SphK1 are oncogenic mediators in breast cancer. Since the positive prognostic effect was only significant for ER-positive cancers, we undertook in vitro studies to compare whether therapeutically co-targeting IGF1R and SphK1 using the dual IGF1R/InR inhibitor OSI-906 and the SphK inhibitor SKI-II might be a viable clinical approach for ER-negative vs. ER-positive cancers. To determine optimal dose ranges to identify a combination effect of the two drugs, initial MTT-assay experiments were performed on each drug alone, i.e. OSI-906 at 0.1, 0.4, 1.6 and 6.4 μM (Fig. 5) and SKI-II at 0.16, 0.8, 4, 10 and 20 μM (Additional file 5), as well as all combinations of these doses (data not shown). Using immunoblot analysis we confirmed inhibition of p-IGF1R phosphorylation and downstream IGF1R signaling (i.e. p-AKT and 4E-BP1 hyper-phosphorylation) following treatment with OSI-906 and SphK1 steady-state protein levels by SphK1 in the MCF7 ER-positive and HCC1806 ER-negative cell-line to demonstrate effective drug-target inhibition by the therapies (Additional file 6). Based on the initial MTT-experiments we identified that a fixed concentration of SKI-II at 4 μM in combination with OSI-906 in the range of 0.1-6.4 μM was the most effective in reducing cell viability (Fig. 5).

OSI-906 showed dose-dependent inhibition of the ER-positive MCF7 breast cancer cell line by both cell-viability and colony-formation assays, with a near-maximal effect seen at 1.6 μM (Fig. 5a and e). The addition of 4 μM SKI-II significantly increased sensitivity to OSI-906 by 2- to 4-fold in both assays (p < 0.001 by repeated measures ANOVA for both assays), with the greatest effect of the drug combination, compared to individual treatments, seen at 4 μM SKI-II and 0.4 μM OSI-906 (Fig. 5a; p < 0.01 and Fig. 5e; p < 0.001, and Additional file 7A). The ER-positive T47D cell-line was somewhat less sensitive to OSI-906, but a similar sensitising effect of 4 μM SKI-II was seen (Fig. 5b and f; p < 0.001 by repeated measures ANOVA for both assays). Greatest combination effects, compared to individual treatments, were seen at 4 μM SKI-II and 1.6 μM OSI-906 co-treatment in the cell-viability assay (Fig. 5f; p = 0.052), and at 4 μM SKI-II and 0.4 μM OSI-906 co-treatment in the colony-formation assay (Fig. 5f; p < 0.01 and Additional file 7B).

The ER-negative HCC1806 cell line showed similar OSI-responsiveness to T47D, and similarly, the addition of 4 μM SKI-II significantly increased responsiveness (Fig. 5c and g; P < 0.001 by repeated measures ANOVA for both assays). The greatest combination effect of SKI-II and OSI-906 co-treatment, compared to individual treatments, was observed on colony formation at the highest dose of 6.4 μM OSI-906 (Fig. 5g; p < 0.05 and Additional file 7C). ER-negative HCC70 cells were highly resistant to inhibition by OSI-906 (Fig. 5d), and even in the presence of 4 μM SKI-II failed to achieve 50% inhibition of cell viability at the highest tested OSI-906 dose of 6.4 μM. HCC70 cells expressed low levels of p-IGF1R and SphK1 protein (data not shown) which is likely a contributing factor to the lack of co-treatment effectiveness.

Calculation of the Chou-Talalay combination index confirmed our findings that co-treatment with OSI-906 and SKI-II led to synergistic effects using OSI-906 at 0.4 μM combined with SKI-II at 4 μM on cell viability and/or colony formation in the MCF7 and T47D ER-positive cell-lines, and at the higher dose of OSI-906 at 6.4 μM in the HCC1806 ER-negative cell-line (i.e. CI < 1; synergism; Table 6). Drug synergy could not be calculated for HCC70 cells owing to their strong resistance to OSI-906.