Abstract
Purpose: Here, we investigated the clinical relevance of an unprecedented combination of three biomarkers in triple-negative breast cancer (TNBC), both in human samples and in patient-derived xenografts of TNBC (PDX-TNBC): EGFR, its recently identified partner (MT4-MMP), and retinoblastoma protein (RB).
Experimental Design: IHC analyses were conducted on human and PDX-TNBC samples to evaluate the production of the three biomarkers. The sensitivity of cancer cells expressing or not MT4-MMP to anti-EGFR (erlotinib) or anti-CDK4/6 inhibitor (palbociclib) was evaluated in vitro in 2D and 3D proliferation assays and in vivo using xenografts and PDX-TNBC displaying different RB, MT4-MMP, and EGFR status after single (erlotinib or palbociclib) or combined (erlotinib + palbociclib) treatments.
Results: EGFR and MT4-MMP were coexpressed in >70% of TNBC samples and PDX-TNBC, among which approximately 60% maintained RB expression. Notably, approximately 50% of all TNBC and PDX-TNBC expressed the three biomarkers. Single erlotinib and palbociclib treatments drastically reduced the in vitro proliferation of cells expressing EGFR and MT4-MMP when compared with control cells. Both TNBC xenografts and PDX expressing MT4-MMP, EGFR, and RB, but not PDX-TNBC with RB loss, were sensitive to erlotinib and palbociclib with an additive effect of combination therapy. Moreover, this combination was efficient in another PDX-TNBC expressing the three biomarkers and resistant to erlotinib alone.
Conclusions: We defined a new association of three biomarkers (MT4-MMP/EGFR/RB) expressed together in 50% of TNBC and demonstrated its usefulness to predict the TNBC response to anti-EGFR and anti-CDK4/6 drugs used in single or combined therapy.
This article is featured in Highlights of This Issue, p. 1691
Translational Relevance
While triple-negative breast cancers (TNBC) mostly express EGFR, anti-EGFR–targeted therapies are ineffective in unselected patients who show limited, transient, or no response. Furthermore, chemotherapy has limited efficacy. By using xenografts and patient-derived xenografts of TNBC (PDX-TNBC), we demonstrated the efficacy of a combination of anti-EGFR (erlotinib) and anti-CDK4/6 (palbociclib) in MT4-MMP+/EGFR+/RB+ tumors, which represent 50% of patients with TNBC. Our results highlight an interest in using MT4-MMP, RB, and EGFR as predictive biomarkers of tumor response to single or combined therapies. This innovative finding holds promise in improving the identification of patients with TNBC more likely to respond to anti-EGFR and anti-CDK4/6 combination that is not yet used in the clinic for TNBC treatment.
Introduction
Triple-negative breast cancer (TNBC) lacks estrogen receptor (ER) and progesterone receptor (PR) expression and amplification of human EGF receptor 2 (HER2/neu). Although TNBC accounts for less than 20% of invasive breast carcinomas, it represents the most aggressive subtype. Compared with other breast cancer subtypes, TNBC is associated with younger age, worse outcome, and a higher risk of relapse within 5 years. Gene expression profiling studies revealed six distinct molecular subgroups of TNBC including two basal-like (BL1 and BL2) subtypes, a luminal androgen receptor (LAR) subtype, a mesenchymal stem–like subtype, a mesenchymal subtype, and an immunomodulatory subtype (1, 2). The presence of immune-related proteins PD-L1 in 20% of TNBC holds promise for improving the patient outcome with immunotherapeutic approaches (3). As there are no validated efficient targeted therapies against TNBC, current treatment relies only on chemotherapy. Therefore, there is an urgent need for biomarkers and new actionable targets for the design of novel combination therapies. Given the expression of EGFR in 50%–75% of TNBC (4), clinical trials have been conducted with EGFR inhibitors (5, 6). Unfortunately, few patients responded to anti-EGFR antibodies (cetuximab) or tyrosine kinase inhibitors (TKI; erlotinib), highlighting a need for biomarkers for patient stratification and for testing new therapeutic combination regimens (7, 8).
Cell-cycle transition from G1–S phase is operated by interactions between cyclin-dependent kinases 4 and 6 (CDK4/6), cyclin D1, and retinoblastoma protein (RB). These cell-cycle molecular regulators are frequently dysregulated in HER2-negative and ER-positive breast cancers. The combination of endocrine therapy with anti-CDK4/6 drugs (palbociclib, ribociclib, or abemaciclib) is currently used for metastatic ER+ with HER2-negative breast cancer. This new therapeutic guideline was established from the phase III clinical trials (PALOMA, MONALEESA, and MONARCH program) showing increased progression-free survival upon combined treatment when compared with endocrine therapy alone. While TNBCs are enriched in cell-cycle proteins such as cyclin E1, intrinsic resistance to CDK4/6 inhibitor monotherapy has been reported (9, 10). The mechanisms of sensitivity or resistance to CDK4/6 inhibition are insufficiently described and are mainly attributed to the loss of RB in cancers. However, the Cancer Genome Atlas study reported that RB is lost or mutated in only 20% of TNBC, indicating that CDK4/6 inhibitors might have beneficial therapeutic and sustained long-term effects (11).
We have previously identified membrane-type-4 matrix metalloproteinase (MT4-MMP) as a cell surface partner of EGFR in TNBC (12). MT4-MMP enhanced receptor activation upon EGF and TGF binding and stimulated breast cancer cell proliferation in 3D culture. MT4-MMP expression in TNBC xenografts led to increased proliferation, angiogenesis, and lung metastases (13, 14). Interestingly, this result was associated with increased cyclins and CDK activities and RB phosphorylation (12). Through IHC analysis, we previously reported that MT4-MMP is coexpressed with EGFR in approximately 80% of human TNBC samples (15). Here, we investigate the relevance of targeting EGFR and CDK4/6 in TNBCs expressing MT4-MMP, EGFR, and RB. Through a panel of in vitro assays, xenografts, and PDX-TNBC models, we provide evidence that the MT4-MMP/EGFR/RB axis is an important driver of TNBC malignancy and is worth considering to predict the response of TNBC to EGFR and CDK4/6 inhibitor combination therapy.
Materials and Methods
Cohort of patients with TNBC
The retrospective cohort of TNBC samples was provided by the Biobank of the University Hospital of Liège (Liège University, Belgium). For human sample collection for research, written informed consent forms were obtained from the patients and the study was conducted in accordance with the recognized ethical guideline of Declaration of Helsinki and approved by the institutional Ethics Committee of the University Hospital of Liège (Liège, Belgium; file#B707201111974 and file#2015/139). The cohort consists of women aged from 27 to 89 years and diagnosed between February 1999 and September 2013 with a median follow up of 53 months (4.4 years; Supplementary Table S1). Only samples from patients who did not receive neoadjuvant therapy (n = 72) were used in this study. All patients in this cohort were metastasis free (M0). The clinicopathogic features retained in this study are those used in daily basis clinical practice including TNM staging, Ki67, grade, and here we added cytokeratin 5/6 and three biomarkers MT4-MMP/EGFR/RB. For TNM prognostic stage determination, we referred to the 8th edition of the AJCC-UICC (the American Joint Committee on Cancer and Union Internationale Contre le Cancer) that also includes grade.
Cell culture and plasmids
Human breast cancer (MDA-MB-231, HS-578T, and MDA-MB-468) cells and COS-1 cells were purchased from ATCC. Cell line authentication for interspecies contamination was performed by Leibniz-Institute DSMZ. Cell lines were tested by MycoALert Kit (Lonza) to ensure that they were Mycoplasma-free. Cells were grown in DMEM supplemented with 10% FBS, l-glutamine (2 mmol/L), penicillin (100 U/mL), and streptomycin (100 μg/mL) at 37°C, in a 5% CO2 humid atmosphere. All culture reagents were purchased from Invitrogen. Cell transfection and plasmid construction were described previously (12). Transient transfection of HS-578T and COS-1 cells was performed with X-treamGENE according to the manufacturer's instructions (Sigma-Aldrich). MDA-MB-231 HS-578T and MDA-MB-468 cells transfected with MT4-MMP cDNA are referred to as MT4-MMP cells, whereas cells transfected with control vector are labeled CTRL cells.
In vivo tumorigenicity
For subcutaneous (s.c.) injection of breast cancer cells, subconfluent MDA-MB-231 cells were collected in serum-free medium (5 × 106 cells/mL) and mixed with an equal volume of cold Matrigel according to a previous report (16). A cell suspension (106 cells/400 μL) was injected subcutaneously into RAG-1–immunodeficient mice in both flanks (n = 6). All assays were repeated at least three times. PDXs from Institute Curie were established from primary tumors from patients with TNBC (PDX-TNBC) as described previously (17–19), in accordance with institutional guidelines and the rules of the French Ethics Committee (project authorization no. 02163.02) and the Belgian Ethics Committee (project authorization no. 14-1582). For PDX-TNBC transplantation, tumor fragments of 1–2 mm were subcutaneously engrafted into nude mice and tumor volumes were estimated every 3 to 4 days according to the formula V: length × (width)² × 0.4. All animal procedures were performed according to the Federation of European Laboratory Animal Sciences Associations (FELASA) within the accredited GIGA animal facility (University of Liège, Liège, Belgium).
In vivo drug administration
When tumors reached a volume of 100 mm3, the mice were separated into 4 groups for treatment with vehicle, erlotinib (50 mg/kg), palbociclib (75 mg/kg), or a combination (erlotinib + palbociclib). Treatment was administered by gavage 5 days a week for 4 weeks. The mice were weighed every 3 days and monitored for tumor growth until the tumor reached the maximum authorized ethical volume of 1,000 mm3. Then, the mice were sacrificed for tumor and organ collections.
Cell proliferation assays
Cell proliferation was assessed in 2D using a WST kit following the manufacturer's instructions. The cells were incubated in 96-well plates in DMEM with 10% FBS and treated 6 hours later with 0.2% DMSO, erlotinib, or palbociclib at different concentrations and time points. For the 3D culture, cells were suspended in cold Matrigel as described previously (12). After 24 hours, the cells were treated every day for 6–7 days.
siRNA transfection
MDA-MB-231 cells producing MT4-MMP were transfected with siRNA targeting MT4-MMP siRNA-1 (5′-CCCACUUUGACGAUGACGAUU-3′; Eurogentec), siRNA-2 (5′-UGACAGGACUUAUUUCUUU-3′; Dharmacon), or scramble (5′-GUCUCUGUAGGAGUCAUCCUU-3′; Eurogentec). The annealed siRNA (100 μmol/L) was double transfected into cancer cells using Lipofectamine RNAiMAX (Invitrogen) according to the manufacture's protocols. Briefly, siRNA was mixed with RNAiMAX Lipofectamine in DMEM-free serum medium. The mixture was added dropwise to cells (60%–80% confluent). siRNA transfection was repeated after 24 hours. The transfected cells were cultured for 48 hours before being used for the experiments. Successful knockdown was confirmed by Western blot analysis.
Cell proliferation in 3D Matrigel
Cells were suspended in medium (250 μL) with cold Matrigel (250 μL) and seeded on top of a layer of Matrigel mixed with medium as described previously (12). Cells were incubated in 1% FBS-containing media and stimulated with EGF 20 ng/mL every day. After 24 hours, erlotinib and palbociclib (LC Laboratories) were added at the indicated concentrations and untreated cells were incubated with 0.2% DMSO. After 7 days, the cells were recovered from 3D cultures upon treatment with 500 μL of dispase (BD Biosciences) for 2 hours at 37°C. After two washes in PBS, the cell pellet was diluted in PBS (1 mL) and sonicated. Fluorimetric DNA titration was performed by measuring fluorescence in a SpectraMax i3 (Molecular Devices) at 356 nm excitation and 460 nm emission.
Immunoblot analysis
Total cell lysates from cells and tumor fragments were prepared using lysis buffer (Cell Signaling Technology) containing protease inhibitors and phosphatase inhibitors (Roche), and the protein concentration was measured by using the DC Protein Assay Kit (Bio-Rad Laboratories). Total protein extracts from cells (20 μg) and tumors (40 μg) were separated on SDS-PAGE under reducing conditions, and transferred onto polyvinylidene difluoride membranes (NEN). The membranes were saturated with casein (1%, wt/vol) in PBS-Tween-20 (0.1%, vol/vol) and incubated with primary antibodies targeting the following proteins: MT4-MMP (Sigma-Aldrich), p-RB [Ser807-811], total RB (4H1), total EGFR, and p-EGFR [Tyr1068] (D7A5; Cell Signaling Technology). After washes, the membranes were incubated with a secondary horseradish peroxidase (HRP)-conjugated goat anti-rabbit antibody (1:2,000, DakoCytomation) or sheep anti-mouse antibody (1:1,000, DakoCytomation). Immunocomplexes were visualized by chemiluminescence reaction on a luminescent image analyzer (LAS-4000, Fujifilm). For loading control, membranes were stripped and reincubated with anti-β-actin or GAPDH antibodies (Sigma-Aldrich).
Coimmunoprecipitation assay
COS-1 cells expressing EGFR1, HER2, and HER3 were transfected with full-length MT4-MMP cDNA (MT4-MMP) or inactive form (MT4-E249A) containing a FLAG tag inserted in the hinge region. After 24 hours, cells were lysed in lysis buffer (Cell Signaling Technology) containing Protease Inhibitor Cocktail (cOmplete, Roche) and Phosphatase Inhibitor Cocktail (PhosSTOP, Roche). The enzyme was immunoprecipitated from lysates (1 mg) using a mouse monoclonal anti-Flag antibody (2 μg) (cloneM2, Sigma) and 50 μL of 50% slurry protein G coupled to Dynabeads (Invitrogen). The precipitate was subjected to Western blotting using a rabbit monoclonal anti-EGFR antibody (Cell Signaling Technology, D38B1), anti-HER2 (Cell Signaling Technology, 29D8), anti-HER3 (Cell Signaling Technology, D22C5), or anti-HER4 (Cell Signaling Technology, 111B2). The antibodies are revealed with an HRP-conjugated secondary antibody.
IHC
IHC for MT4-MMP and EGFR on tumor sections was performed as described previously in ref. 15 by using a rabbit anti-human MT4-MMP antibody (1/500; AB39028; Abcam) and a rabbit monoclonal anti-EGFR antibody (5B7; 790-4347, Ventana Medical Systems, Roche). Tissue microarrays (TMA) slides containing at least 3 tissue cores of each tumor were used for IHC detection of MT4-MMP, EGFR, and RB.
For HER2, HER3, HER4, and androgen receptor (AR) staining, antigen retrieval was performed with EDTA (Dako, S2367) for HER2 and HER3 or with citrate buffer (Dako, S2031) for HER4. Nonspecific binding was prevented by incubation in blocking solution 1/5 (Cell Signaling Technology, 1519L). Slides were incubated for 1 hour at 1/100 with rabbit anti-human HER2 (Ventana, 790-4493), anti-HER3 (Cell Signaling Technology, 12708), anti-HER4 antibody (Abcam, ab230781), or anti-androgen receptor (Ventana, SPA07). For RB and phosphoRB stainings, antigen retrieval was performed by incubating tumor sections with buffer target (Dako, S1699) at 126°C and 1.4 bar in autoclave for 11 minutes. Endogenous peroxidases were subsequently blocked by 3% H2O2/H2O for 20 minutes, and nonspecific binding was prevented by incubation in Dako Protein Block, Serum-Free (Dako, X0909). Slides were first incubated 60 minutes at room temperature with a rabbit polyclonal targeting human-phosphoRB antibody or RB (1/400; Cell Signaling Technology, # 8516) or a mouse anti-human RB (1/100) diluted in Dako diluent (Dako, S2022). Finally, slides were incubated with HRP-conjugated anti-rabbit or anti-mouse secondary antibodies (Envision System Labeled Polymer-HRP, Dako, K4003 and K4001) for 30 minutes at room temperature. For staining, 3-3′ diaminobenzidine hydrochloride (Dako, K3468) was incubated for 3 minutes. Slides were finally counterstained with hematoxylin and mounted with Eukitt medium for microscope observation. Omission of the first antibody served as negative control.
Proliferating cells in xenografts and PDX were visualized by IHC staining for the human Ki-67. Antigen retrieval of paraffin sections was performed for 11 minutes at 126°C in Target Retrieval Solution by incubation (Dako, S1699). Slides were then subjected to endogenous peroxidase blockade with 3% H2O2/H2O for 20 minutes. Nonspecific binding was prevented by slide incubation in PBS/BSA 10% (Fraction V, Acros Organics) for 1 hour. After 1-hour incubation with anti-Ki-67 antibody (1:100), slides were incubated for 30 minutes with the power vision poly-HRP anti-Mouse/Rabbit/Rat IgG kit (DVPB+110HRP, Immunologic) and revealed with Vector DAB (SK-4100, Vector Laboratories).
Apoptotic cells in tumors were detected by IHC staining for the human activated caspase-3. Antigen retrieval, peroxidase blockade, and nonspecific binding was similar to Ki67 staining. After overnight incubation with anti-caspase-3 antibody (1:300; clone C14, 9661L, Cell Signaling Technology) at 4°C in enhancer buffer, slides were incubated for 30 minutes with En Vision/HRP goat anti-rabbit antibody and revealed with DAB.
IHC staining positivity criteria
Two pathologists analyzed human tumors and TMA sections stained with MT4-MMP, EGFR, and RB in a blinded manner. Sections were considered RB positive, when they presented homogeneous and abundant nuclear staining in the tumor compartment. For MT4-MMP and EGFR staining, we selected a cutoff of 50% of stained tumor cells as described previously (15). Regarding TMA, 3–6 tissues spots were analyzed for each tumor. A sample was considered positive when at least 2 of 3 cores were positive. In case of inconsistent staining between the cores, the sample was excluded from the study. For AR staining, a cutoff of 10% of intense staining of cancer cells was considered positive.
Automatic computer-assisted image analysis of IHC staining
Computerized image analysis and quantification of IHC staining on whole tumor sections were performed using the image analysis toolbox of MATLAB R2014a (8.3.0.532) 64-bit (MathWorks). For quantitative measurement of human Ki67 in xenografts and PDXs treated with erlotinib, palbociclib, or the combination, nuclear Ki67 brown staining was automatically detected and the resulting image was binarized. For activated caspase-3 staining quantification, necrotic areas were delimitated and removed from processed images before automatic measurement of staining density in whole tumor sections. Results are expressed as staining density per tumor area of treated tumor and reported to vehicle-treated tumor (12).
Statistical analysis
For in vitro proliferation assays, differences between experimental groups were assessed by one-way ANOVA test. For tumor growth analysis, differences between experimental groups were assessed by two-way ANOVA test at the endpoint using the Prism 4.0 software (GraphPad). For IHC staining quantification, Wilcoxon–Mann–Whitney tests were performed. Level of significance are: ns, P > 0.05; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.
Results
MT4-MMP, EGFR, and RB are coexpressed in 50% of patients with TNBC
MT4-MMP, EGFR, and RB expressions were examined by IHC analyses in a cohort of 72 human TNBC samples (Fig. 1A). RB expression was maintained in 68% of samples (Fig. 1B). Both MT4-MMP and EGFR were found in 71% of tumors (Fig. 1C), in which RB expression was found in 69% (Fig. 1D). Positivity for the three biomarkers (MT4-MMP/EGFR/RB) was found in 49% of TNBC samples (Fig. 1E). Expression of MT4-MMP and EGFR was not associated with any clinicopathologic feature (ref. 15; Supplementary Table S1), whereas RB expression was higher in stage I tumors (Fig. 1F) and inversely correlated with Ki67 staining (Fig. 1G) with a correlation coefficient (R) of 0.9. Hence, we hypothesized that the expression of MT4-MMP, EGFR, and RB might be used as a predictive biomarker to define TNBC subtypes that could benefit from combined therapies targeting EGFR and the cell-cycle regulators.
Expression of MT4-MMP, EGFR, and RB defines TNBC subgroups. IHC staining for MT4-MMP, EGFR, and RB in serial sections of human TNBC samples (n = 72). A, Images illustrating positive (+) and negative (−) stainings for MT4-MMP, EGFR, and RB (././.). MT4-MMP and EGFR staining was considered positive when the cell membrane and cytoplasm staining was uniformly distributed in the whole tumor section. For RB, only nuclear staining of cancer cells in tumor compartments is considered positive. B and C, Diagrams representing the frequency of TNBC subgroups based on their expression profile of RB, MT4-MMP, and EGFR. D, The frequency of RB in TNBC expressing MT4-MMP and EGFR. E, The frequency of TNBC expressing MT4-MMP, EGFR, and RB together. F, RB expression profile relative to TNM staging. G, Inverse correlation of Ki67 and RB expression percentage in TNBC (top) and graph of correlations (bottom).
MT4-MMP and EGFR expression sensitizes TNBC cells to erlotinib and palbociclib in vitro
We examined the impact of MT4-MMP and EGFR expression on in vitro tumor cell proliferation upon erlotinib or palbociclib treatment. MDA-MB-231, HS-578T, and MDA-MB-468 cells endogenously producing EGFR were transfected with MT4-MMP cDNA (MT4-MMP cells) or control vector (CTRL cells; Fig. 2A; Supplementary Fig. S1A and S1D). Cells were incubated with increasing concentrations of erlotinib or palbociclib for different incubation periods (24, 48, and 72 hours; Fig. 2). In MDA-MB-231, MT4-MMP cells were more sensitive to erlotinib and palbociclib compared with control cells (Fig. 2B and C). Treatment with erlotinib (10 μmol/L) for 72 hours reduced cell proliferation by 60% and 30% in MT4-MMP and control cells, respectively (Fig. 2D). Incubation with palbociclib (5 μmol/L) decreased proliferation of MT4-MMP cells by 90% and by 50% in control cells (Fig. 2E). Similar responses with erlotinib and palbociclib were replicated in HS-578T cells expressing MT4-MMP and EGFR (Supplementary Fig. S1B and S1C), whereas MDA-MB-468 negative for RB (Supplementary Fig. S1G) were sensitive to erlotinib but insensitive to palbociclib (Supplementary Fig. S1E and S1F). Accordingly, two different siRNA directed against MT4-MMP (Fig. 3A) abolished MT4-MMP cell sensitivity to erlotinib (10 μmol/L) and reduced their sensitivity to palbociclib (5 μmol/L) when compared with cells treated with nontargeting siRNA (scrambled; Fig. 3B). Consistent with previous data (12), EGFR phosphorylation upon EGF stimulation for 5, 20, and 60 minutes was increased in MT4-MMP cells (Fig. 3C and D). Similar increase in EGFR phosphorylation was observed in MDA-MB-468 cells incubated with EGF for 60 minutes (Supplementary Fig. S1G). These data highlight a sustained EGFR activation upon ligand stimulation in the presence of MT4-MMP. Importantly, receptor activation was abolished by erlotinib addition alone or in combination with palbociclib. RB phosphorylation was increased after 20 minutes of EGF treatment with a more pronounced effect in MT4-MMP cells (Fig. 3C and D). RB phosphorylation was reduced by palbociclib alone or in combination with erlotinib after 20 and 60 minutes in both cell types (Fig. 3C and D). These data underline that EGFR signaling and RB inactivation were enhanced by MT4-MMP, the pathway that can be inhibited by erlotinib and palbociclib treatments. Sustained MT4-MMP/EGFR/RB signaling in MT4-MMP cells is likely responsible for increased sensitivity to treatment with erlotinib or palbociclib. While the proliferation of MT4-MMP cells was identical to that of CTRL cells in 2D culture, it was considerably higher in 3D culture and in xenografts as we observed previously (12).
MT4-MMP expression sensitizes TNBC cells expressing EGFR and RB to erlotinib and palbociclib in vitro. A, Analysis of MT4-MMP expression levels by Western blot analysis in MDA-MB-231 cells stably transfected with MT4-MMP cDNA or control vector. Cell proliferation of MDA-MB-231 cells expressing MT4-MMP (MT4-MMP cells) and control cells (CTRL cells) after 24, 48, and 72 hours of incubation with erlotinib (B) or palbociclib (C) at indicated concentrations. Rate of cell proliferation compared with control (DMSO) of cells treated for 72 hours with erlotinib (D) or palbociclib (E). *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.
The MT4-MMP/EGFR/RB signaling axis promotes TNBC cell response to erlotinib and palbociclib. A, Western blot analysis for MT4-MMP in MT4-MMP–expressing MDA-MB-231 cells incubated with siRNA (siRNA-1 and siRNA-2) against MT4-MMP or control siRNA (NT-siRNA). Actin is used as a loading control. B, Cell proliferation of MT4-MMP cells incubated with siRNA-1 and siRNA-2 against MT4-MMP (siRNA) or control (scrambled) and treated for 72 hours with erlotinib or palbociclib at 5 μmol/L and 10 μmol/L, respectively. DNA quantification was performed to quantify cell number after treatments. C, Western blot analysis of phosphorylated EGFR and RB (pEGFR and pRB) and total proteins in total cell lysates from control and MT4-MMP cells incubated in DMEM-free FBS for 1 hour. Cells were stimulated with EGF (20 ng/mL) for 5, 20, and 60 minutes and treated with 0.2% DMSO, erlotinib (10 μmol/L), or palbociclib (5 μmol/L). GAPDH is used as a loading control. D, Quantification of EGFR and RB phosphorylation rate after 5, 20, and 60 minutes of cell incubation with EGF in C. E, MT4-MMP cell proliferation in 3D Matrigel was assessed by DNA quantification after treatment with DMSO (0.2%), erlotinib (2.5 μmol/L and 5 μmol/L), or palbocilib (0.1 μmol/L and 0.5 μmol/L) for 7 days in DMEM with 1% FBS. *, P ≤ 0.05.
Combination therapy of erlotinib and palbociclib is effective only in MT4-MMP– and EGFR-expressing xenografts
Reasoning that cancer cell sensitivity to drugs might be affected by the structural organization of cells in the extracellular matrix, we first tested the efficacy of erlotinib and palbociclib in a 3D culture system (on Matrigel) to better recapitulate the in vivo context. Both erlotinib (5 μmol/L) and palbociclib (0.5 μmol/L) reduced the proliferation of MT4-MMP–expressing MDA-MB-231 cells (Fig. 3F). However, the combination of erlotinib (2.5 μmol/L) and palbociclib (0.1 μmol/L) resulted in an additive inhibitory effect, which was more efficient than higher doses of single treatment with erlotinib (5 μmol/L) or palbociclib (0.5 μmol/L; Fig. 3F).
To validate these in vitro findings, we next tested the impact of MT4-MMP expression on in vivo tumor response to erlotinib and palbociclib in mono- and combination therapy. Immunodeficient Rag1−/− mice were subcutaneously injected with MDA-MB-231 cells producing or not MT4-MMP and treated with vehicle, erlotinib (50 mg/kg/day), palbociclib (75 mg/kg/day) or the combination. As previously reported (12–14), MT4-MMP–expressing xenografts grew faster than control xenografts when treated with vehicle (Fig. 4A). While mice bearing control tumors did not respond to single or combined treatments (Fig. 4B), tumor growth of MT4-MMP xenografts was inhibited by erlotinib and palbociclib with the combination eliciting a stronger effect (Fig. 4C). Consistent with the in vitro data, MT4-MMP and EGFR-sensitized tumors to erlotinib and palbociclib single agents with a striking effect using the combination.
Expression of MT4-MMP enhances the xenograft response to erlotinib, palbociclib, and the combination. A, Tumor growth curves of MDA-MB-231 control xenografts (control cells) and xenografts expressing MT4-MMP (MT4-MMP cells) treated with vehicle. Arrows indicate the day when treatment started. Tumor growth curves of control xenografts (B) and MT4-MMP xenografts treated with erlotinib (50 mg/kg/day), palbociclib (75 mg/kg/day), or the combination (C). D, Images illustrating the IHC staining for human Ki67 in xenografts. E, Fold change in Ki67 staining of treated tumors relative to vehicle of control xenografts (top) or MT4-MMP xenografts (bottom). F, Western blot analysis of phosphorylated RB (pRB) and total RB protein of tumor lysates from control and MT4-MMP tumors treated with vehicle, erlotinib, palbociclib, or combination. GAPDH is used as a loading control. G, Quantification of RB phosphorylation (pRB)/RB/GAPDH rate in tumors in C. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.
Combination therapy of erlotinib and palbociclib inhibited the tumor proliferation index and RB phosphorylation in MT4-MMP xenografts
Assessment of Ki67 by IHC (Fig. 4D) and computerized quantification of staining density of control xenografts revealed no significant effect of single treatment with erlotinib on tumor cell proliferation (P = 0.69). In sharp contrast, palbociclib (P = 0.0159) or the combination of palbociclib + erlotinib (P = 0.0159) reduced the proliferation index of control xenografts (Fig. 4E), without significantly affecting the tumor volume at the end of the assay (Fig. 4B). In MT4-MMP tumors, Ki67 density was not affected by erlotinib (P = 0.40), but decreased by approximately 70% with palbociclib (P = 0.0047) and 80% with the combination (P = 0.0001). In contrast, a slight tumor reduction was achieved by palbociclib and erlotinib + palbociclib in control xenografts (20% and 30%, respectively).
Analysis of the RB phosphorylation status of whole tumor extracts revealed a strong phosphorylation of RB in MT4-MMP xenografts when compared with control tumors treated with vehicle. Interestingly, erlotinib + palbociclib combination led to a strong reduction of RB phosphorylation in MT4-MMP tumors (Fig. 4F and G).
MT4-MMP, EGFR, and RB are predictive biomarkers for response to erlotinib–palbociclib combination in PDX-TNBC
To validate our findings in more accurate and robust models of TNBC, we used a panel of 37 PDX of TNBC (Fig. 5A). We performed an IHC study on TMAs of PDX tumors (Supplementary Fig. S2) and found similar expression profiles of MT4-MMP, EGFR, and RB to those of the human samples (Fig. 1B–D). RB expression was maintained in 59% of PDX-TNBC (Fig. 5B). Coexpression of MT4-MMP and EGFR was quite similar to that of human samples with 81% of PDX-TNBC expressing both MT4-MMP and EGFR (Fig. 5C), and 46% being triple positive for MT4-MMP, EGFR, and RB (Fig. 5D). Ki67 density staining was higher in PDX than in human samples and no MT4-MMP–negative cases were found. This is likely due to the higher success of engraftment in mice of the aggressive human tumors with high Ki67 index.
Expression of the MT4-MMP/EGF/RB axis is maintained in PDX-TNBC. A, Images illustrating the histology of a primary TNBC tumor (human) and the PDX-TNBC after 3 passages in the mammary fat pad of NOD/SCID mice (PDX). Diagrams representing the frequency of PDX-TNBC subgroups based on their expression profile of RB (B); MT4-MMP and EGFR (C); and MT4-MMP, EGFR, and RB together (D). E, Image illustrating IHC staining for MT4-MMP, EGFR, and RB in PDX-TNBC. Four groups were defined: MT4-MMP+/EGFR+/RB− (HBCx-4B), MT4-MMP+/EGFR−/RB− (HBCx-14), MT4-MMP+/EGFR−/RB+ (HBCx-16), and MT4-MMP+/EGFR+/RB+ (PDX3).
We next selected PDX-TNBC with 4 different expression profiles of MT4-MMP, EGFR, and RB (Fig. 5E) to further support the concept that the combination of erlotinib + palbociclib is efficient in these models. Nude mice were transplanted subcutaneously with fragments of these selected PDX-TNBC generating five experimental groups: one MT4-MMP+/EGFR+/RB− (HBCx-4B), one MT4-MMP+/EGFR−/RB− (HBCx-14), one MT4-MMP+/EGFR−/RB+ (HBCx-16), and two triple-positive MT4-MMP+/EGFR+/RB+ (PDX3 and HBCx-8). The growth of tumor associated with RB loss was not inhibited by erlotinib, palbociclib, or erlotinib + palbociclib treatments, and this was obtained independently of MT4-MMP and EGFR status (Fig. 6A and B). All RB-positive tumors responded to erlotinib + palbociclib combination therapy. MT4-MMP+/EGFR−/RB+ (HBCx-16) grew slowly, as expected, and was sensitive to palbociclib treatment (Fig. 6C). Intriguingly, this PDX was also sensitive to erlotinib and erlotinib + palbociclib combination. Importantly, a PDX-TNBC producing MT4-MMP, EGFR, and RB (PDX3) was sensitive to erlotinib or palbociclib with a striking improvement with erlotinib + palbociclib treatment (Fig. 6D). Another PDX-TNBC expressing the three biomarkers (HBCx-8) was sensitive to palbociclib but not to erlotinib (ErloR; Fig. 6E). IHC analysis of other HER family members (HER2, HER3, and HER4) in PDX3 and HBCx-8 did not reveal any difference (Supplementary Fig. S3). Moreover, MT4-MMP coimmunoprecipitated specifically with EGFR, but not with HER2 or HER3 in COS1 cells expressing EGFR, HER2, and HER3 and transfected with MT4-MMP cDNA (Supplementary Fig. S4A–S4C). Therefore, the nonresponse of the PDX HBCx-8 to erlotinib is not due to any change in copy number of EGFR gene (data not shown), but most likely due to an intrinsic pathway–driven resistance. Nevertheless, the combination therapy led to a stronger inhibition of tumor growth than palbociclib alone in both PDXs. Finally, we obtained tumor inhibition rates >65% with the erlotinib + palbociclib combination in all the RB-positive PDX-TNBC (Fig. 6F).
PDX-TNBC expressing MT4-MMP and RB were more sensitive to erlotinib and palbociclib combination therapy. Investigation of erlotinib, palbociclib, and combination efficacy in PDX. PDX-TNBC differentially expressing MT4-MMP, EGFR, and RB were grown in nude mice. When tumors reached 100–125 mm3, mice were treated for 5 days a week with erlotinib 50 mg/kg/day, palbociclib 75 mg/kg/day, or the combination for 4 weeks. Tumor growth curves and drug response of the PDX-TNBC: (MT4-MMP+/EGFR+/RB−) HBCx-4B (A), (MT4-MMP+/EGFR−/RB−) HBCx-14 (B), (MT4-MMP+/EGFR−/RB+) HBCx-16 (C), and (MT4-MMP+/EGFR+/RB+) PDX3 (D). E, PDX positive for MT4-MMP+/EGFR+/RB+ (HBCx-8) with intrinsic resistance to erlotinib (ErloR). F, Summary table of tumor growth inhibition of treated tumors compared with vehicle. G–K, Quantification of Ki67-positive cells and Ki67 fold changes in response to erlotinib, palbociclib, and the combination on tumor sections from PDX-TNBC models in A–E. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001.
Quantification of Ki67 staining in intact tumor areas, excluding the necrotic regions, confirmed the tumor growth inhibition in RB-positive PDX-TNBC. The proliferation index rate was not modified by any treatment in PDX-TNBC with RB loss (Fig. 6G and H). As expected, Ki67 density was reduced by erlotinib, palbociclib, and erlotinib + palbociclib in RB-positive PDX-TNBC, except with poorly sensitive PDX-TNBC (HBCx-8) to erlotinib, in which Ki67 was only reduced by the combination (Fig. 6I–K). Similarly, activated caspase-3 staining was markedly increased in tumor compartments of tumors sensitive to erlotinib, palbociclib, and the combination in RB-positive tumors expressing MT4-MMP and EGFR (Supplementary Fig. S5A–S5F).
Discussion
Here, we show that anti-EGFR and anti-CDK4/6 are effective in TNBC tumors expressing MT4-MMP, EGFR, and RB, which represents approximately 50% of human TNBC tumors. By using TNBC xenografts and PDX models established from patients with no prior neoadjuvant therapy (20), we found that the expression of these three biomarkers strongly sensitized tumors to erlotinib–palbociclib combination therapy.
The innovation of this study relies on exploring PDX-TNBC for defining subgroups that can benefit from a combination of anti-EGFR and anti-CDK4/6 drugs. On the basis of the association status of MT4-MMP/EGFR/RB, we identified two subgroups of TNBC with differential response to erlotinib and palbociclib combination: sensitive tumors (RB+/MT4-MMP+/EGFR+) and insensitive tumors with RB loss, independently of MT4-MMP or EGFR expression. Moreover, in TNBCs expressing EGFR and nonresponsive to erlotinib, the combination with palbociclib reversed tumor insensitivity and increased the antitumor effect, indicating that TNBC resistant to EGFR inhibitors could be very sensitive to a combination with CDK4/6 in the clinic. This study demonstrates that the association of the MT4-MMP, EGFR, and RB is worth considering in TNBC to select patients who may benefit from combination-targeted therapy.
TNBC is a molecularly and clinically heterogeneous disease. Paradoxically, TNBC has a higher response to chemotherapy (30%–40%) compared with other breast cancer subtypes, but also has the highest rate of relapse (21). Also, the lack of well-recognized actionable targets in TNBC subclassification hampers the use of available targeted therapies. Although EGFR is frequently overexpressed in TNBC, phase II and phase III clinical trials using anti-EGFR inhibitors as single agents (cetuximab, erlotinib) failed to show clinical benefit in unselected patients with TNBC (4). Consequently, this has drastically decreased the interest of EGFR as a potential target in TNBC (22, 23). Our data show that patient stratification based on the expression of relevant biomarkers for EGFR activation/signaling could help to select subgroups of patients with TNBC that might respond to EGFR inhibitors. MT4-MMP is a specific copartner of EGFR that promotes its activation upon ligand binding resulting in increased cyclin and CDK activities, and enhanced RB phosphorylation (12). The clinical relevance of the MT4-MMP/EGFR axis in TNBC was previously validated in human samples demonstrating that 72.5% of TNBC produce MT4-MMP and EGFR (15). This study conducted on a PDX-TNBC bank further confirms the relevance of MT4-MMP and EGFR double positivity that was found in 81% of PDX-TNBC. Interestingly, MT4-MMP, EGFR, and RB were found present together in 49% of human samples and 46% of PDX-TNBC. While this triple positivity did not correlate with any clinicopathologic features, it defines a TNBC subtype sensitive to EGFR and CDK4/6 inhibitors that notably displayed a higher response to the combination of the two inhibitors. When analyzing the abundance of Ki67 expression in the human TNBC cohort, we found an inverse correlation with intact RB expression (Fig. 1F). Consistent with these data, 30% of TNBC displayed high Ki67 level and low RB expression, which has been suggested as a predictive factor for response to chemotherapy (24, 25). Similarly, we found that approximately 32% of TNBC with RB loss displayed high Ki67 density. Altogether, these data show that 30% of RB loss TNBC would benefit from chemotherapy and 70% of RB positive TNBC would have more profit from targeted therapies. In this context, we are providing a proof of principle of a therapy combination that is worth considering in this TNBC subtype.
While, there is striking evidence of TNBC enrichment for cell-cycle genes (26, 27), Finn and colleagues (9) have reported that TNBC cell lines are clearly less sensitive in vitro to CDK4/6 inhibition when used in monotherapy compared with ER-positive cell lines. Consistent with our in vitro data, MDA-MB-231 and HS-578T cells negative for MT4-MMP were less or not sensitive to palbociclib. The sensitivity to palbociclib was enhanced in these cells after ectopic expression of MT4-MMP. Interestingly, expression of MT4-MMP in MDA-MB-468 cells negative for RB and producing EGFR responded to erlotinib but not to palbociclib. Mechanistically, MT4-MMP enhances EGFR activation and RB phosphorylation (12), which renders cancer cells more sensitive to CDK4/6 and EGFR inhibitors. Accordingly, RB phosphorylation was strongly reduced in MT4-MMP–expressing cells in vitro and in xenografts by palbociclib, in single treatment and more importantly in combination with erlotinib. Studies on other cancer types have reported that sensitivity or resistance mechanisms to CDK4/6 inhibition are orchestrated by RB status (28–30). By linking MT4-MMP–modulated EGFR activation with RB phosphorylation in TNBC, we point to a new signaling axis that can be targeted in TNBC. The EGFR/RB pathway has been explored recently on other cancer types in two studies. Indeed, Zhou and colleagues (29) reported recently an efficient response to the combination of an EGFR inhibitor with MAPK or CDK4/6 blockade in esophageal tumors. In addition, palbociclib reversed non–small cell lung cancer (NSCLC) acquired resistance to EGFR-TKI and the combination of CDK4/6 and EGFR inhibitor has been proposed to be effective in HPV-negative head and neck squamous cell carcinoma (HNSCC; ref. 30). In NSCLC overexpressing EGFR, the response to single anti-EGFR agents was either not efficient or did not show a sustained effect, due to acquired resistance and RB loss (28). Interestingly, we found a drastic inhibition of tumor growth by the combination of CDK4/6 inhibitor with erlotinib in a PDX-TNBC (HBCx-8) expressing EGFR, MT4-MMP, and RB but unresponsive to erlotinib monotherapy (ErloR; Fig. 6E). In our PDX-TNBC study and in other studies conducted on lung and esophageal cancers (28, 29), combination therapies reversed resistance to EGFR inhibitors, supporting our concept that TNBC resistant to EGFR therapy might benefit from a combination of EGFR with CDK4/6 inhibitors.
We cannot rule out the contribution of other factors to the TNBC response to erlotinib and palbociclib. For instance, the expression of androgen receptors (AR) could influence cell cycle in TNBC (31). It has been reported that the combination of enzalutamide, an AR signaling inhibitor, and palbociclib showed higher antiproliferative effect in vitro in AR+ and RB+ cells (32). Evidence of AR and RB signaling in TNBC was proven by the increased sensitivity of TNBC cells expressing AR to therapy combination of palbociclib with PI3K inhibitors (pictilisib and taselisib; ref. 33). These reports have demonstrated a synergistic effect of CDK 4/6 (palbociclib) and PI3K inhibitors (pictilisib and taselisib) in PIK3CA-mutant AR+ TNBC cell lines. In fact, we examined AR expression in our TNBC cohorts by IHC staining with a cutoff of 10% for positivity and found that 17 of 72 (24%) tumors were AR+ tumors, among which 94% were RB+ and 71% coexpressed MT4-MMP and EGFR (Supplementary Fig. S6). In contrast, only 4 of 37 (10%) PDX-TNBC samples were positive for AR. The discrepancy between the prevalence of AR in human samples compared with PDX-TNBC may be due to the enrichment of the most aggressive tumors during the generation of a PDX (34). However, molecular profiling for AR on TNBC fresh tissues in a prospective study would determine the significance of AR found by IHC. Indeed, AR+ tumors are often associated with lower Ki67 and lower aggressiveness (35). Taken together, these data suggest that TNBC can be targeted by combination of CDK4/6 with EGFR, AR or PI3K inhibitors, depending on activated upstream signaling. Of note, we demonstrate that about 50% of TNBC are associated with MT4-MMP/EGFR/RB pathway and could benefit from a combined targeted therapy.
In conclusion, our study demonstrates that anti-EGFR and anti-CDK4/6 drugs are worth considering in TNBC and indicates that the expression of MT4-MMP/EGFR/RB can be used as predictive biomarkers for treatment response of patients with TNBC to a combination therapy. By updating the interest of targeted anti-EGFR therapies in TNBC and identifying three markers predictive of tumor response to a combination of therapy targeting EGFR and CDK4/6, our data pave the way for future clinical trials in TNBC using a personalized medicine.
Disclosure of Potential Conflicts of Interest
J. Collignon is a consultant/advisory board member for Roche, Merck, Amgen, Pfizer, Ipsen, Sanofi, Sirtex, Lilly, Celgene, Bayer, Bristol-Myers Squibb, and Novartis. G. Jerusalem reports receiving commercial research grants from Roche and Novartis, and is a consultant/advisory board member for Novartis, Roche, Pfizer, Lilly, Celgene, Amgen, Bristol-Myers Squibb, Puma Technology, AstraZeneca, Daiichi Sankyo, and AbbVie. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: M. Coibion, A. Noël, N.E. Sounni, G. Jerusalem
Development of methodology: P. Foidart, C. Yip, M. Lienard, L. Montero-Ruiz, G. Jerusalem
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Foidart, C. Yip, J. Radermacher, E. Montaudon, S. Château-Joubert, M. Coibion, V. Jossa, E. Marangoni, G. Jerusalem
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Foidart, J. Radermacher, S. Blacher, E. Maquoi, S. Château-Joubert, A. Noël, N.E. Sounni, G. Jerusalem
Writing, review, and/or revision of the manuscript: P. Foidart, J. Radermacher, J. Collignon, A. Noël, N.E. Sounni, G. Jerusalem
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Montero-Ruiz, A. Noël
Study supervision: J. Collignon, N.E. Sounni, G. Jerusalem
Acknowledgments
The authors thank all the lab members of Sounni and Noël's group for useful technical advice. The authors thank the Biobank of Liège University for providing human samples of TNBC. This project was supported by grants from the Fonds de la Recherche Scientifique - FNRS (F.R.S.-FNRS, Belgium), the Fondation contre le Cancer (foundation of public interest, Belgium), FRS-FNRS, Liege University and Leon Fredericq Foundation, and the Fonds spéciaux de la Recherche (University of Liège, Liège, Belgium).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Footnotes
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
- Received June 13, 2018.
- Revision received October 15, 2018.
- Accepted November 27, 2018.
- Published first November 30, 2018.
- ©2018 American Association for Cancer Research.
References
Volume 25, Issue 6
