Background
The human epidermal growth factor receptor 2 (HER2), known as ErbB2 or Neu, is a tyrosine kinase receptor protein encoded by the
ERBB2 (
HER2) gene [
1]. HER2 is a member of the epidermal growth factor (EGF) receptor family along with EGFR/HER1, ERBB3/HER3, and ERBB4/HER4. The four receptors are transmembrane proteins with an intracellular tyrosine kinase domain (although ERBB3/HER3 is considered kinase impaired). While HER2 is the only family member that does not bind to a ligand, it forms heterodimers with the other EGF receptor protein members and shows strong catalytic kinase activity, efficiently triggering downstream signaling through phosphatidylinositol-3 kinase (PI3K) and mitogen-activated protein kinase (MAPK) [
1]. Approximately 15–20% of primary breast cancers show HER2 protein overexpression and/or
HER2 gene amplification [
2], which is associated with poor prognosis. The development of humanized monoclonal antibodies binding the extracellular domain of HER2 (e.g., trastuzumab, pertuzumab), EGFR-HER2 small molecule kinase inhibitors (e.g., lapatinib, neratinib, or tucatinib), and antibody-drug conjugates (e.g., T-DM1 or DS-8201) has revolutionized HER2-positive breast cancer treatment [
3]. Still, most patients with metastatic disease eventually progress on anti-HER2 therapy due to de novo or acquired resistance, and 20–30% of patients with early HER2+ breast cancer relapse [
4‐
6]. Therefore, elucidating the mechanisms of resistance to anti-HER2 drugs is pivotal to further improve patients’ survival outcomes.
Receptor activator of nuclear factor kappa-Β ligand (RANKL) and its receptor (RANK) belong to the TNF superfamily. The fundamental role of RANK signaling in osteoporosis and bone metastasis inspired the development of denosumab, a monoclonal antibody against RANKL, for the treatment of skeletal-related events (SREs) linked to osteoporosis and cancer [
7]. RANK signaling activation in the breast epithelium promotes tumor initiation, progression, and metastatic spread. Thus, RANK and RANKL have emerged as promising targets for breast cancer prevention and treatment [
8]. RANKL is expressed in progesterone receptor-positive cells and acts as a paracrine mediator of progesterone in the mammary epithelia [
9,
10]. Increased RANK receptor expression is more frequent in hormone receptor-negative tumors and high-grade breast cancer, but it is also found in a subset of luminal tumors [
11,
12]. RANK signaling controls proliferation and stemness in BRCA1-mutant and oncogene-driven mammary tumors [
13,
14]. Interestingly, RANK signaling inhibition has been shown to reduce HER2 tumorigenesis in preclinical studies [
9,
15]. In human tumors, RANKL and HER2 levels predict metastasis to the bone in breast cancer better than RANKL alone [
16].
Some of the common (intrinsic or acquired) resistance mechanisms to trastuzumab and/or lapatinib treatment are impaired HER2 binding, parallel/downstream pathway activation, ER signaling, cell cycle de-regulation, or escape from antibody-dependent cellular cytotoxicity (ADCC) [
17]. Personalized treatment of HER2-positive breast cancer and better predictive biomarkers to anticipate therapy resistance will contribute to the identification of patients that will benefit from new combinatorial therapies, paving the way for HER2-positive breast cancer precision medicine [
18].
In this study, we unveiled a functional relationship between RANK and HER2 signaling using HER2-positive breast cancer patient samples and cell lines. Upon analyses of HER2-positive breast cancer samples from treatment-naive patients and residual disease at surgery after neoadjuvant anti-HER2 therapy, including paired samples from the phase II SOLTI-1114 PAMELA trial, we observed that anti-HER2 treatment or resistance to anti-HER2 therapy both resulted in increased RANK expression. Additionally, when we analyzed the effects of RANK modulation on anti-HER2 treatment in HER2-positive breast cancer cell lines, we observed that enhanced RANK signaling led to increased lapatinib resistance.
Methods
Patient samples
RANK and RANKL expression was assessed in tumor samples from three different cohorts of patients with HER2-positive breast cancer.
Treatment-naive cohort
Patients with primary and operable HER2-positive breast cancer (
n = 197) diagnosed from 2003 to 2010 at the Nottingham City Hospital, Nottingham, UK. Tumor samples were collected at surgery prior to any neoadjuvant treatment. Histological grade was assessed by the Nottingham Grading System [
19] and other clinicopathological factors such as tumor size, lymph node (LN) status, estrogen receptor (ER), progesterone receptor (PR), and HER2 expression, as well as patient age and disease progression, were analyzed before including the samples into the TMAs, prepared as previously described [
20].
Anti-HER2-resistant cohort
Patients treated with trastuzumab-based primary chemotherapy and residual disease at surgery (
n = 43) diagnosed at the Catalan Institute of Oncology (ICO), Bellvitge University Hospital in l’Hospitalet de Llobregat, and Dr. Josep Trueta University Hospital in Girona (Spain) between 2005 and 2014 and described in [
21]. The selection criterion included patients with early or locally advanced HER2-positive breast cancer (including inflammatory breast cancer) who had received neoadjuvant treatment with trastuzumab-based chemotherapy and had residual invasive disease following surgery (i.e., who had not achieved a pathological complete response at surgery). Neoadjuvant chemotherapy was based on anthracyclines and taxanes given concurrently with weekly trastuzumab for 24 weeks followed by surgery. For all patients, hematoxylin and eosin (H&E)-stained slides from formalin-fixed paraffin-embedded (FFPE) tumor blocks were examined to determine representative areas of the invasive tumor. ER, PR, and HER2 positivity were assessed in the initial tumor core biopsies as well as in the residual disease. For each patient, different clinical and histopathological features such as age, and histological grade (Nottingham Grading System) were obtained.
SOLTI-1114 PAMELA cohort
Patients treated with neoadjuvant dual-blockade trastuzumab and lapatinib (
n = 151) and in which biopsy paired samples were prospectively obtained. The main results of the PAMELA neoadjuvant phase II study have been previously reported [
22] and the completed study is registered in ClinicalTrials.gov (number NCT01973660). In this trial, patients with early HER2-positive breast cancer were treated with neoadjuvant lapatinib (1000 mg daily) and trastuzumab (8 mg/kg i.v. loading dose followed by 6 mg/kg) for 18 weeks. Patients with hormonal receptor-positive breast cancer received letrozole or tamoxifen according to menopausal status. FFPE tumor samples at baseline, at day 14 of treatment, and at surgery were collected according to standard protocols.
Gene expression analyses
RNA samples of the PAMELA trial from tumors at baseline (
n = 151) were previously analyzed [
22]. Here, the nCounter platform (NanoString Technologies, Seattle, WA, USA) analyzed RNA of 101 residual tumors from surgical samples of the PAMELA trial. A minimum of 100 ng of total RNA was used to measure the expression of 550 genes, including
RANK and
RANKL, and 5 housekeeping genes (
ACTB,
MRPL19,
PSMC4,
RPLP0, and
SF3A1). Expression counts were then normalized using the the nSolver 4.0 software and custom scripts in R 3.4.3. For each sample, we calculated the PAM50 signature scores (basal-like, HER2-E, luminal A and B, normal-like) and the risk of recurrence score [
23]. Intrinsic molecular subtypes were identified using the research-based PAM50 predictor as previously described [
22,
24].
Immunohistochemistry and tissue microarray scoring
Immunohistochemistry (IHC) in TMAs was performed using anti-human mouse monoclonal RANK (N-1H8 Amgen) and human RANKL (M366 Amgen) as described in [
9]. RANK or RANKL staining was scored on a scale of 0 to 3 for intensity (0 = no staining, 1 = weak, 2 = moderate, 3 = intense) and for the percentage of positively stained tumor cells (0–100) as previously reported [
25]. The result of multiplying staining intensity by positive cell percentage is the
H-score value, ranging from 0 to 300. TMA cores were scored for RANK and RANKL with the assistance of the breast cancer pathologists from the Bellvitge Hospital, if tumor cells represented > 15% of the total TMA core area. Patients were stratified according to RANK or RANKL
H-scores as being protein-positive (
H-score ≥ 1) or protein-negative (
H-score = 0). Breast tumors from patient-derived xenografts were used as positive and negative controls. Experimental data from our laboratory in breast cancer cells and patients’ samples [
26] confirmed that cells in which RANK protein expression is not detected by IHC/western blot may still respond to RANKL stimulation or denosumab inhibition [
11,
26,
27]. This is probably due to the “fragility” of the RANK epitope and the limited sensitivity of the current tools to detect RANK protein expression. Thus, even with an
H-score ≥ 1, we are likely underestimating samples with a functional RANK signaling pathway.
Statistical analyses were performed with the support of IDIBELL and Nottingham University Statistical Assessment Services. The ER/PR/HER2 status, grade, and tumor stage were known for each case included in the TMAs. Associations between IHC scores and clinicopathological parameters were evaluated using Pearson’s chi-squared test.
Cell lines and cell culture
The cell lines BT474 parental (BT474) and BT474 with lapatinib resistance (BTLR) were described in [
28]. SKBR3 parental (SKBR3) and SKBR3 lapatinib resistant (SKLR) lines were described in [
29]. The cell line HCC1954 was obtained from ATCC (CRL-2338). BT474 cells were grown in DMEM + GlutaMAX (Gibco) supplemented with 2 mM
l-glutamine (HyClone), 1× penicillin-streptomycin solution (P/S, Gibco), and 7.5% fetal bovine serum (FBS, Gibco). SKBR3 cells were grown in McCoy’s 5A + GlutaMAX supplemented with 2 mM
l-glutamine, 1 mM sodium pyruvate (HyClone), 1× P/S, and 5% FBS. HCC1954 cells were grown in RPMI medium 1640 + GlutaMAX supplemented with 2 mM
l-glutamine, 1× P/S, and 5% FBS. The cells were grown at 37 °C in 5% CO
2 humidified incubators. For RANKL treatments, cells were incubated in the presence of 100–300 ng/ml of RANKL. Cell lines were routinely tested for mycoplasma contamination.
Viral transduction
To ectopically express RANK, the RANK gene (TNFRSF11A) was cloned into the lentiviral vector pSD-69 (kindly provided by S. Duss and M. Bentires-Alj) under the control of hPGK promoter. As a control (ctrl), we used an empty pSD-69 plasmid generated by removing the BamHI-SalI fragment containing CcdB and CmR genes. Knockdown of RANK endogenous expression was achieved by shRNA lentiviral delivery using pGIPZ vectors containing shRNAs against human RANK (RHS4531, Dharmacon), and shRNAs sequences #3 (TATCTTCTTCATTCCAGCT) and #4 (ATTCTTCCTTGAACTTCCC) were selected based on their ability to silence RANK expression. As a control, we used pGIPZ expressing a verified non-targeting sequence (RHS4346 Dharmacon). Lentiviruses were prepared in HEK293T cells transfected with psPAX2 (Addgene #12260) and pMD2.G (Addgene #12259) by the calcium phosphate method. Virus-containing supernatants were centrifuged at 1500 rpm for 5 min and filtered with 0.45-μm filters (Millipore). The medium from 1-cm2 production cells was used to infect 2-cm2 recipient cells at roughly 33% confluence, adding fresh medium (1:1) and 8 μg/ml polybrene (Millipore). Approximately 90% infection efficiency was verified 3 days after transduction by detection of GFP expressed from pGIPZ plasmids. Transduced cells were selected with 1.5 μg/ml puromycin (Sigma), starting 3 days after infection, and subsequently maintained with 1 μg/ml puromycin in the growth media.
Cell proliferation
To determine cell proliferation, 1000–4000 cells per well in 100 μl were seeded in 96-well plates. After 24 h, 100 μl of medium with or without the indicated concentrations of lapatinib (0–16 μM) was added, and cells were incubated for 4 days. The relative number of viable cells each day was determined by adding 50 μl of diluted CCK-8 reagent according to the manufacturer’s protocol (Sigma).
Western blot
Cells were seeded at approximately 33% confluence in 6-well plates. The following day, they were washed and incubated in a medium without FBS. The next day, the medium was changed to 1.8 ml medium with or without 1 μM lapatinib followed by a 2-h incubation. Subsequently, 0.2 ml of medium with or without 300 ng/ml of RANKL (RANKL-LZ Amgen) was added to the wells. Ten minutes later, the extracts for immunoblots were prepared with modified RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1% NP-40, 0.25% sodium deoxycholate) containing 1× PhosSTOP and complete protease inhibitor cocktail (Roche), and protein concentrations determined with DC protein assay reagents (BIO-RAD). Fifteen micrograms of protein were resolved by SDS-PAGE and blotted into Immobilon-P 0.45 μm membranes (Millipore). Antibodies against the following proteins were used for probing: RANK (R&D Systems AF683), p-HER2 (#2249), HER2 (#2165), p-EGFR (#3777), EGFR (#4267), p-ERK1/2 (#9101), ERK1/2 (#9102), p-AKT (#4051), AKT (#9272), p-p65 (#3033), p65 (#8242), p-IκB (#9246), IκB (#9242) (from Cell Signaling), β-actin (sc-47778), and tubulin (Abcam ab21058).
Immunoprecipitation
Upon transiently transfecting HEK293 cells with affinity-tagged versions of full-length RANK (RANK-V5 in pLenti6/V5-DEST, Invitrogen), full-length HER2 (FLAG-HER2 [
30]), an amino (742-NTF) [
30], or carboxy-terminal fragment of HER2 (611-CTF) [
31], cells were washed twice with ice-cold PBS and proteins were extracted with 20 mM Tris-HCl pH 7.4, 137 mM NaCl, 2 mM EDTA, 10% glycerol, 1% NP-40 supplemented with 50 μg/ml leupeptin, 50 μg/ml aprotinin, 0.5 mM sodium orthovanadate, and 1 mM phenylmethylsulfonyl fluoride. Equal amounts of extracts were incubated for 3 h with immunoglobulin G (Abcam ab171870), FLAG (Sigma F3165), HA (Abcam ab9110), V5 (Thermo Scientific #R961-25), HER2 (32H2 in house antibody described in [
32]), or trastuzumab (Hoffmann-La Roche) antibodies. Then, protein A agarose beads (Calbiochem IP02) were added for 2 h. Immunoprecipitates were washed thoroughly with lysis buffer and boiled in reducing SDS loading buffer to be analyzed by Western blot.
RNA isolation and RT-qPCR
Cells were seeded at approximately 33% confluence in 6-well plates. The next day, the medium was changed to medium with or without 100 ng/ml RANKL followed by an additional 24 h incubation period. To analyze mRNA expression levels, total RNA was purified with Maxwell RSC simplyRNA Tissue kit (AS1340 Promega). For each sample, cDNA was retrotranscribed from 1 μg of RNA using 200 U SuperScript II plus random hexamer oligos following the manufacturer’s protocol (Invitrogen); cDNA from 20 ng RNA for each sample was analyzed by SYBR green real-time PCR (Applied Biosystems) with 10 μM primers using a LightCycler® 480 thermocycler (Roche). Analyses were performed in triplicates using the LightCycler® 480 software (Roche). Peptidylprolyl isomerase A, PP1A, was used as the reference gene. The primer sequences used in the analyses are as follows: PP1A (fw ATGGTCAACCCCACCGTT, rev TCTGCTGTCTTTGGGACCTTG), RANK (fw GCAGGTGGCTTTGCAGAT, rev 5’GCATTTAGAAACATGTACTTTCCTG), BIRC3 (fw GGTAACAGTGATGATGTCAAATG, rev TAACTGGCTTGAACTTGACG), ICAM1 (fw AACTGACACCTTTGTTAGCCACCTC, rev CCCAGTGAAATGCAAACAGGAC), TNFα (fw AAGCTGTAGCCCATGTTGT, rev TGAGGTAACAGGCCCTCTGAT), and IL8 (fw CTGCGCCAACACAGAAATTA, rev CATCTGGCAACCCTACAACA).
Discussion
A crosstalk between RANK and EGFR signaling has been described in the context of osteoclast differentiation [
44], as well as in breast cancer for a particular RANK truncated isoform [
45]. In the mammary gland, we found that pharmacological inhibition of RANKL decreases tumorigenesis and lung metastases in the MMTV-ErbB (Neu) transgenic mouse model [
9]. In the same line, MMTV-ErbB mice with a heterozygous RANK deletion showed decreased pulmonary metastasis than RANK WT MMTV-ErbB controls [
15]. In addition, RANKL treatment increased lung metastases in both FVB/N and MMTV-ErbB animals [
15]. More recently, a review [
46] followed by an article with experimental data [
47], suggested the combination of RANK and HER2 signaling inhibition as a new strategy for the treatment of HER2-positive breast carcinomas.
In this study, we have shown that
RANK gene expression increased after dual treatment with lapatinib and trastuzumab in HER2-positive tumor samples from the PAMELA clinical trial [
22] and in HER2-positive breast cancer cell lines. These observations would point to increased RANK signaling in patients treated with anti-HER2 drugs. We also observed that the percentage of patients with RANK tumor expression doubled in the context of HER2 resistance when compared to treatment-naive HER2-positive breast tumors. Furthermore, both SKBR3- and BT474 HER2-positive cell lines with acquired lapatinib resistance displayed increased RANK expression and pathway activation compared to their respective lapatinib-sensitive controls. Thus, our combined analyses of HER2-positive breast cancer samples and cell lines demonstrate that RANK expression is higher in HER2-resistant breast cancer. RANK loss moderately sensitized lapatinib-resistant cells to the drug, and overactivation of RANK signaling increased lapatinib resistance in HER2-positive cell lines (SKBR3, BT474, and HCC1954). Based on these results, one could speculate that activation of RANK signaling may allow breast cancer cells to survive anti-HER2 therapies and the benefit of combining denosumab with HER2 inhibitors as postulated by [
47].
NF-κB signaling has been shown to enhance ErbB2-induced tumor growth both in vitro and in immune-competent mice [
48,
49]. Increased NF-κB activation downstream of RANK [
50] may also contribute to lapatinib resistance. Hyperactive NF-κB signaling has been proposed as a possible resistance mechanism after lapatinib treatment in HER2-positive [
51] and triple-negative breast cancer [
52,
53]. In HER2-positive breast cancer, lapatinib-resistant cells show increased NF-κB levels and do not respond to single HER2 or NF-κB inhibitors, but to a combination of both [
51]. The NF-κB expression is normally linked to invasive high-grade tumors, and several NF-κB inhibitors are currently being investigated [
54,
55]. Chen and colleagues showed that lapatinib treatment induced a constitutive activation of NF-κB through Src-dependent p65 and IκBα phosphorylation, sensitizing the cells to proteasome inhibitors [
52]; our data suggest that increased RANK being a well-known regulator of NF-κB may also play a role, although we cannot discard the contribution of other RANK-driven downstream pathways. The phosphorylation of IκBα, leading to its degradation and resulting in p50/p65 heterodimer nuclear translocation, is mediated by the IKK complex (comprising IKKα, IKKβ, and IKKγ/NEMO) [
56,
57]. HER2 itself was shown to activate NF-κB via the canonical pathway involving IKKα in HER2-positive and ER-negative breast cancer cells [
58]. IKKα also mediates NF-κB activation in mammary cells during pregnancy and after RANKL stimulation [
59]. In our study, we did not observe clear changes in p65 phosphorylation after stimulation with ErbB ligands and the treatment with lapatinib could not counteract p65 phosphorylation driven by RANKL treatment in RANK-overexpressing HER2-positive cell lines, providing an alternative survival path for these cells.
Importantly, we have shown RANK binding to HER2 by co-immunoprecipitation experiments. Accordingly, Zoi et al. recently showed the interaction of RANK with ErbB family members by proximity ligation assays [
47]. In this publication, the authors claim that the number of RANK/HER2 dimers in cells correlates with HER2 expression levels. Also, denosumab, trastuzumab, and/or pertuzumab treatment reduces the number of RANK/HER2 dimers whereas RANKL stimulation leads to an increased number of RANK/HER2 dimers [
47]. Finally, their data show that RANKL addition decreases the efficacy of HER2 inhibitors [
47]. In our hands, a direct interaction between RANK and HER2, independent of EGF, was observed. RANKL stimulation of HER2-positive breast cancer cells overexpressing RANK decreases HER2 phosphorylation, indicating that RANKL influences ErbB2 signaling.
RANKL was shown to promote migration in breast cancer cells after activation of the ERK and AKT pathways [
60]. We have also found increased phosphorylation of ERK1/2 and AKT after RANKL treatment in SKBR3 and BT474 cell lines, with either physiological or increased RANK levels by receptor overexpression. Interestingly, we observed that RANKL-mediated induction of ERK1/2 and AKT phosphorylation was completely abrogated after lapatinib treatment in SKBR3 and BT474 cells, again independently of RANK receptor expression levels. These observations and the fact that RANK and HER2 interact suggest that lapatinib inhibits not only EGFR/HER2 tyrosine phosphorylation but also RANK signaling driven by RANKL (e.g., ERK1/2 and AKT). Importantly, in addition to the direct interaction between RANK and HER2, we observed that RANK signaling is functionally linked to the ErbB2 pathway. Additional research is required to address whether the direct RANK/HER2 interaction contributes to the enhanced resistance to lapatinib observed after activation of RANK signaling.
Taken together, we showed that anti-HER2 treatment and resistance acquisition both raised RANK expression levels in HER2-positive clinical breast tumors and cell lines. Also, enhanced RANK signaling increased lapatinib resistance in HER2 breast cancer cells. We found that RANK and HER2 physically and functionally interact. Altogether, these results hint to a dual RANK and HER2 inhibition therapy for RANK-expressing HER2-positive breast cancer patients, whose benefit remains to be tested.
Conclusions
In summary, we showed that RANK is expressed in HER2-positive breast cancer samples, particularly in patients resistant to anti-HER2 blocking therapy. The RANK expression is often associated with younger age, hormone receptor-negative status, and higher histological grade and proliferation index. Moreover, in HER2-positive breast cancer samples from the PAMELA trial, RANK expression increased upon treatment with lapatinib and trastuzumab. This was confirmed in vitro in several HER2-positive human breast cancer cell lines suggesting that RANK signaling may contribute to the development of lapatinib resistance. Indeed, RANK-overexpressing HER2-positive cell lines showed increased resistance to lapatinib and higher NF-κB pathway activation. Finally, we demonstrated that RANK physically and functionally interacted with HER2 suggesting a RANK/HER2 crosstalk. Together, these results suggest that inhibition of RANK signaling may improve the response to anti-HER2 therapies in RANK-positive, HER2-positive breast cancer patients.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.