Introduction
Members of the receptor tyrosine kinase (RTK) superfamily are often aberrantly expressed and/or activated in human tumors and many have been successfully targeted using antibody-based therapies or tyrosine kinase inhibitors (TKI) [
1]. In breast cancer, ErbB2 has proven to be an excellent target; however, only 25% of cancer patients are eligible for an ErbB2-directed therapy [
2,
3]. Currently much effort is going into uncovering other RTKs that when inhibited could impact disease. The fibroblast growth factor receptors (FGFRs) and their ligands have been implicated in many different types of tumor, including breast cancer. Indeed, amplification of
FGFR1 or
FGF3 has been detected in approximately 10% or 15% of primary tumors respectively, while patients with
FRFR1 amplification are more likely to develop distant metastasis [
4], as such FGFRs are considered to be highly relevant therapeutic targets [
5,
6].
The 4T1 and 67NR mammary cancer cell lines are widely studied models for basal-like breast cancer that have similar genetic backgrounds but different metastatic potential. When implanted in Balb/c mice the 67NR cells form mammary tumors that do not metastasize, while the 4T1 mammary tumors are able to spread to and grow in distant organs [
7]. We have previously shown that both tumor cell lines display autocrine FGFR activity due to co-expression of FGFRs and ligands. Using the FGFR selective inhibitor, dovitinib (TKI258) [
8], we showed that the 4T1 and 67NR cancer cell lines are dependent upon FGFR signaling for proliferation and survival, and that mammary tumor outgrowth is significantly slower in dovitinib-treated mice [
9]. While tumors from dovitinib-treated animals displayed a strong reduction in FRS2/Erk pathway signaling, the phosphatidyl inositol 3'kinase (PI3K)/Akt pathway showed little or no downregulation [
9]. In the results presented here we further explored the role of the PI3K/Akt/mammalian target of rapamycin (mTOR) pathway and RTKs that regulate this pathway in the 4T1 and 67NR models.
We show that the combination of dovitinib with the PI3K/mTOR inhibitor, NVP-BEZ235 [
10], strongly downregulates the FRS2/extracellular signal-regulated kinase (Erk) and PI3K/Akt/mTOR signaling pathways, resulting in high levels of apoptosis and tumor stasis. Using an unbiased approach to screen for active receptors, anti-phosphotyrosine receptor antibody arrays (P-Tyr RTK), we identified high levels of P-epidermal growth factor receptor (P-EGFR) and P-ErbB2 in the tumors. Testing the pan-ErbB inhibitor AEE788 [
11] in the 4T1 and 67NR models revealed that only the combination of AEE788 and dovitinib resulted in blockade of the FRS2/Erk and PI3K/Akt/mTOR pathways, high levels of apoptosis with prolonged tumor stasis, and in the 4T1 model a highly significant decrease in lung metastasis. Our results suggest that
in vivo, but not
ex vivo, both breast cancer models become dependent upon co-activation of FGFR and ErbB receptors for PI3K/Akt/mTOR pathway activity, demonstrating the importance of the tumor environment in influencing receptor activity and response to targeted inhibitors. In the models we studied, optimal blockade of tumor growth and metastatic spread was only achieved by combining an FGFR inhibitor with the PI3K/mTOR inhibitor or with the pan-ErbB inhibitor. Considering that breast tumors co-express multiple RTKs including ErbB and FGFRs [
12,
13], these results have important implications for targeted therapy.
Materials and methods
Kinase inhibitors
The inhibitors dovitinib [
8], NVP-BEZ235 [
10] and AEE788 [
11] were provided by Drs. D Graus-Porta, S-M Maira and G Caravatti (Novartis Institutes for Biomedical Research, Basel, Switzerland). All inhibitors were prepared as 10 mmol/L dimethyl sulfoxide (DMSO) stocks for
in vitro use or diluted in the corresponding carrier for
in vivo experiments.
Cell lines, in vivotreatments and analysis
The 4T1 and 67NR cell lines [
7] were maintained as described [
9]. We examined the 4T1 cell line for mutations in
PI3KA,
K-Ras and
FGFR3. We sequenced exons 9 and 20 of
PI3KCA, exons 1 and 2 of
K-Ras and exons 7, 10 and 15 of
FGFR3; none of these exons were mutated. Animal experiments were performed according to the Swiss guideline governing animal experimentation and approved by the Swiss veterinary authorities. The 4T1 and 67NR cells (5 × 10
5 cells) were injected into the fourth mammary fat pad of 10-week-old BALB/c mice (Harlan Laboratories, Netherlands). Once palpable, tumors were measured daily and volume was calculated using the following formula: Volume = Height × ((Diameter/2)
2 ×
π)
Mice were randomly distributed into groups when tumors reached 50 to 100 mm3. Different groups were treated for the indicated times with different doses depending upon the experiment: vehicle (water or polyethylene glycol 300), dovitinib (per oral (p.o.), once daily) AEE788 (p.o., thrice weekly), NVP-BEZ235 (p.o., once daily), the combination of dovitinib (20 mg/kg) and AEE788 (40 mg/kg), or dovitinib (20 mg/kg) and NVP-BEZ235 (10 mg/kg). For experimental metastasis, 2.0 to 2.5 × 105 4T1 cells were injected into tail veins; 24 hrs later, mice were treated with PEG300 or NVP-BEZ235 (10 days at 20 mg/kg); alternatively, 7 days after injection, treatment was started for 11 days; dovitinib (20 mg/kg), NVP-BEZ235 (10 mg/kg), AEE788 (40 mg/kg), dovitinib/AEE788 or dovitinib/NVP-BEZ235. At the end, lungs were isolated and placed in Bouin's solution to visualize and count metastases (Leica MacroFluo Z6, Leica Microsystems, Heerbrugg, Switzerland). Results are reported as average number of nodules per group.
Tumor serial transfer
Inhibitor-treated mice were sacrificed and tumors were digested for 1 hr at 37°C in Collagenase (1 mg/ml), Dispase (1 mg/ml) and DNAse (50 KU/ml) to a single cell suspension. Hematopoietic cells labeled with CD45-biotin (Biolegend, San Diego, USA) were removed from samples using anti-biotin magnetic bead depletion (EasySep, StemCell Technologies, Grenoble, France) and tumor cells were enriched via discontinuous percoll density gradient separation (GE Healthcare, Glattbrugg, Switzerland). Equal numbers of tumor cells were injected into recipient Balb/c mice. Tumors were visible by 7 days; tumor-take was 100%.
Analysis of drug effect on circulating tumor cells
Circulating tumor cells in 4T1 tumor-bearing mice were quantified as described in [
14] and collected cells were cultured in media supplemented with 60 μM 6-thioguanine to select for 4T1 cells [
7]. After 14 days colonies were stained and counted.
Immunofluorescence and image measurements
For immunohistological analysis, tumors were dissected and frozen in optimum cutting temperature compound (OCT) on a 2-methylbutane, dry ice bath. Cryosections (10 μm) were fixed in 1:1 methanol/acetone, blocked with 1% rat, donkey and goat serum and stained using antibodies for CD31-FITC (Biolegend, San Diego, USA. Clone 390), phosphorylated histone-H3-Alexa-Fluor647 (Biolegend, Clone HTA28), and cleaved caspase-3 (Cell Signaling Technology, Inc. Danvers, MA, USA 9661) with Alexa-Fluor594 conjugated anti-Rabbit IgG (Invitrogen, Lucerne, Switzerland) secondary. Sections were mounted with Prolong gold containing 4',6-diamidino-2-phenylindole (DAPI) (Invitrogen) and images acquired using an Axio Imager Z2 LSM700 confocal microscope (Zeiss, Feldbach, Switzerland). Images were analyzed using Image J (NIH, MD, USA). Image measurements were taken in 15 to 20 digital images (10 × objective lens) from three to four separate tumor specimens, with four images taken in each quadrant of the tumor perimeter and one in the center region. The area of CD31 or cleaved caspase-3 immunoreactivity was measured as the number of pixels above the fluorescence threshold (typically 15 to 30) as a proportion of total pixels within defined tumor boundaries; values were not influenced by tumor size. Phosphorylated histone-3-positive cells were manually counted, and the final values presented as an average number of cells per field.
Gene expression analysis and pathway enrichment
The 4T1 tumors were taken from mice treated with vehicle, or with 15 or 40 mg/kg dovitinib for 2 and 8 hrs or for 1, 3 and 10 days. Total RNA was isolated from three tumors/treatment time points using RNeasy Mini Kits (Qiagen, Hilden, Germany) following the manufacturer's instructions. Sentrix MouseWG-6 V2 arrays (Illumina, San Diego, CA, USA) were used for expression profiling. Quality control of the RNA (Agilent Bioanalyzer), as well as labeling and array hybridization was performed at the DKFZ microarray Core Facility. Data were normalized with the variance stabilization transformation algorithm (Bioconductor vsn package) and genes with significant change (
P < 0.05, log-fold-change > 0.5) were recorded. The microarray data have been submitted to Array-Express [E-MTAB-1260: EBI]. The Bioconductor limma package was used to identify differentially expressed genes and two-step regression (Bioconductor maSigPro package) was applied to identify genes with temporal expression changes. STRING [
15] and DAVID Bionformatics Resources 6.7 [
16] were used to map protein interactions and for functional gene enrichment, respectively. R-script was used to generate the plots for EGFR and its ligands (see Figure S5 in Additional file
1).
Western blot analysis and RTK phosphorylation detection
Protein lysates were prepared and western analyses were performed as described [
9]. The following antibodies were used: P-mTOR, P-ErbB2, P-FRS2, P-Akt, PERK1/2, P-S6, Akt, ERK1/2, S6 all from Cell Signaling, FRS2 (Santa Cruz Biotechnology, Inc. Dallas, Texas, USA) and ErbB2 [
17]. Blots were probed using an appropriate horseradish peroxidase-conjugated secondary antibody (Amersham, GE Healthcare, Glattbrugg, Switzerland) and developed using Western Pico ECL substrate kit (GE Healthcare). Detection of phosphorylated RTKs in tumor lysates was performed using a Proteome Profiler Array kit (R&D systems, Abingdon, UK. cat. No. ARY014) as per the manufacturers protocol. Quantification of signal was determined using Image J software.
Statistics
For determining statistical significance in all quantifications, non-parametric Mann-Whitney U-tests were used; all data are presented as mean ± SD. Data were considered significant for P-values < 0.05 and are denoted as follows: *P < 0.05, **P < 0.01.
Conclusions
Targeting RTKs with antibodies or kinase inhibitors is a clinically validated anti-cancer approach; however, the effectiveness of individual inhibitors is often short-lived and resistance emerges. Experimental approaches have revealed numerous feedback loops in tumor cells and have shown that blocking one signaling pathway, be it the receptor [
22,
23] or downstream targets [
24,
25], is not sufficient to cause tumor regression, thereby allowing resistant cells to emerge. Moreover, inhibition of Akt or PI3K has been shown to increase the activity of multiple RTKs [
24]. Taken together, it appears that the utility of single pathway inhibitors might be limited and that resistance to RTK inhibitors may often be due to activation of other RTKs that restore signaling [
23]. Indeed, we and others have shown experimentally that ligand activation of EGFR or ErbB2/ErbB3 heterodimers overcomes the inhibitory effects of trastuzumab by stimulating downstream signaling pathways [
26‐
28].
The 4T1 and 67NR models have been useful for examining the impact of FGFR inhibition on tumor growth and metastatic spread [
9]. We previously showed that blocking FGFR
in vitro was sufficient to inhibit Erk and PI3K signaling and to induce cell death via blockade of the latter pathway [
9].
In vivo targeting of FGFR significantly slows tumor growth, but neither tumor stasis nor strong inhibition of PI3K/Akt signaling was observed [
9]. As we show here, the PI3K/mTOR inhibitor NVP-BEZ235 robustly blocks this pathway and the combination of dovitinib + NVP-BEZ235 had significantly better anti-tumor and anti-metastatic activity than treatment with single inhibitors. It is becoming clear that
in vivo responses to kinase inhibitors is optimal only when tumors show high levels of apoptosis [
29]. Indeed, we show here that the most durable tumor responses and the highest levels of apoptosis were observed in mice treated with the FGFR inhibitor in combination with either the PI3K/mTOR inhibitor or the pan ErbB inhibitor. For both treatment modalities strong inhibition of the FGFR/FRS2/Erk pathway and the PI3K/Akt/mTOR pathway was observed.
An important goal of this work was to uncover a tyrosine kinase receptor that when inhibited would block PI3K/Akt/mTOR pathway activity. The ErbB RTKs were interesting to target for different reasons: EGFR and ErbB2 are both active in the tumors; ErbB2 signals strongly to the PI3K pathway through ErbB3 [
19], and pan-ErbB inhibitors that block all three receptors are in clinical use [
20]. Thus, we were surprised to find that the pan-ErbB inhibitor AEE788 [
11,
21] when given alone blocked ErbB receptor activity, but had no impact on PI3K/Akt signaling. Only the combination of AEE788 with dovitinib caused a significant block in the PI3K/Akt/mTOR pathway and resulted in strong anti-tumor activity. When AEE788 is dosed as a single agent it is possible that one of the other active RTKs maintains PI3K/Akt signaling. Indeed, the fact that addition of dovitinib to AEE788 did block the pathway, suggests that this is the case. We detected other RTKs with moderate levels of P-Tyr in 4T1 tumors and specific inhibitors for two of them, PDGFR and VEGFR, are available. We have already tested
in vivo activity of the VEGFR inhibitor PTK787 [
30], but as a single agent we found no change in 4T1 tumor outgrowth [
9]. It will be interesting to examine the potential of blocking PDGFR or VEGFR in combination with AEE788 or dovitinib in future work.
It is interesting to compare the effects of the different inhibitors. First, the combination of dovitinib + AEE788 appears to be somewhat more effective than that of dovitinib + NVP-BEZ288. An analysis of the phosphorylation status of signaling proteins did not reveal major differences in tumors from these treatment groups; both combinations efficiently block the PI3K/Akt/mTOR pathway. However, the decrease in mitosis and increase in apoptosis was consistently stronger in the dovitinib + AEE788 treatment group, suggesting that targeting ErbB receptors has broader downstream effects compared to targeting only PI3K/mTOR. In experiments aimed at testing the durability of treatment response, this combination was also more effective. Indeed, 67NR tumors from mice treated with dovitinib + AEE788 remained static in the timeframe of our studies. Second, in tumors from dovitinib-treated mice, alone or with both combinations, we observed a significant decrease in CD31 staining, which was accompanied by changes in vessel morphology. NVP-BEZ235 has been shown to impact on tumor vessel permeability [
31], but as a single agent in our studies, we did not observe significant changes in the vessels. Since dovitinib also targets VEGFRs [
8,
11] we have previously tested the effects of another more selective VEGFR inhibitor, PTK787 [
32] and found that this inhibitor had no effect on 4T1 tumor outgrowth [
9]. However, in the work we present here we consider it possible that the decreased CD31 staining in tumors from dovitinib-treated mice might be due to the combined effects of blocking VEGFR and FGFR activity. Third, looking at the effects of the inhibitors on 4T1 metastasis, a few conclusions can be made. After tail vein injection of 4T1 cells, the combination of dovitinib + AEE788 is clearly the best treatment of those tested, suggesting that in the lung environment tumor cells continue to be dependent on ErbB and FGF RTKs. In contrast, we were surprised to see that once 4T1 cells colonized the lungs, NVP-BEZ235 treatment had no effect on metastatic growth. In primary tumors, treatment with NVP-BEZ235 or dovitinib had similar significant effects on proliferation and apoptosis, albeit lower than the combination. These results suggest that in the lung environment the PI3K/mTOR pathway is not so important for tumor cell growth. Finally, the dovitinib + NVP-BEZ235 combination strongly inhibited intravasation from the primary tumor into the bloodstream and/or tumor cell survival, which very likely contributes to the low number of spontaneous lung metastases in these animals.
Multiple RTKs are often active in cancer cells and combinations of RTK inhibitors have been shown to be better at blocking PI3K/Akt/mTOR signaling than individual inhibitors (for example, by Stommel
et al. [
33]). Considering FGFR and ErbB receptors as targets, non-small cell lung cancer cell lines have been shown to respond better to combinations targeting both, compared to individual treatments [
34]. Our previous work with human breast cancer cell lines showed that the combined inhibition of FGFR and ErbB receptors caused a complete loss of PI3K/Akt/mTOR pathway activity and a robust block in
in vitro proliferation [
35]. Using a panel of tumor-derived cell lines with defined sensitivity to ErbB kinase inhibitors, it was shown that many of these tumor cell lines are rescued by FGF addition [
23]. These
in vitro results clearly show that FGFR activation can, in many cases, circumvent ErbB receptor inhibition. Does this occur in cancer patients? In a small group of ErbB2-positive breast cancer patients treated with lapatinib, those whose tumors had elevated levels of FGFR2 had a shorter time to progression than the low FGFR2 group [
36]. The amount of genome-wide information available for breast tumors is increasing at a rapid pace and should help in choosing patients for whom simultaneous inhibition of ErbB and FGF receptors might be appropriate.
FGFR amplification has been found in some basal-like breast cancers, a group that also has
EGFR amplification [
12].
FGFR1 is preferentially amplified in estrogen receptor-positive tumors and in our experience these often co-express ErbB family members. Indeed, some breast tumors with copy number changes in both
ERBB2 and
FGFR1 were recently described [
13]. In addition to genomic alterations including copy number changes or mutations, ligand-mediated receptor activation might also play an important role. It has been known for many years that FGF8 and FGF10, both ligands for FGFR2, are overexpressed in human breast tumors [
37,
38], suggesting that antibodies to screen for active FGFR2 would be very useful. There are many useful diagnostic tools for identifying ErbB receptor alterations in human tumors. In the future the development of additional reagents that can be used to predict FGFR activation in tumors would be an important area to pursue.
In the work presented here we show that combinatorial inhibition of FGFR and ErbB receptors has a very significant impact on the
in vivo tumor growth and metastatic spread of breast cancer models. Considering the emerging evidence that breast tumors co-express ErbB and FGFRs, our results have important implications for targeted therapy. There are multiple ErbB family inhibitors available for clinical use, and additional, more selective FGFR inhibitors, such as NYP-BGJ398, are now starting clinical development [
39]. In the future it should be possible to select breast cancer patients for whom combination therapy would be appropriate.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
AI and JG designed and carried out the in vivo work and the analyses of the tumors and metastases described in this manuscript. JD helped with some of the initial in vivo experiments and discussions. AI performed western analyses and PCR analyses and phosphor-RTK analyses on tumors. RH performed the phospho RTK analysis on cell lines. SO and SW performed the microarray studies. NH conceived the study, participated in the experimental design and helped to draft the manuscript. All authors have read and approved the manuscript.