Background
Breast cancer metastasis to distant organs such as the liver, lung, bone and brain is a leading cause of cancer-related mortality in women and is estimated to account for 626,679 deaths globally in 2018 [
1]. Patients with the human epidermal growth factor receptor-2-positive (HER2
+ve) and the triple-negative (TNBC) subtypes of breast cancer have a higher propensity to develop central nervous system (CNS) metastases [
2]. Since no clinically approved biomarkers of brain metastasis are available to prospectively identify high-risk patients, brain lesions are often diagnosed late, when symptomatic and more difficult to treat. Chemotherapy and localised therapies for brain metastases including surgery, whole-brain irradiation and stereotactic radiosurgery, while improving overall median survival, are rarely curative [
3]. Hence, more effective therapies are required for incurable brain metastases.
Trastuzumab, a humanised monoclonal antibody targeting HER2, was the first clinically approved targeted therapy for the treatment of HER2
+ve breast cancer and is now used routinely as the first-line therapy [
4]. Its introduction into the clinic has significantly improved survival outcomes for HER2
+ve breast cancer patients. However, not all patients initially respond, and for those who do respond, resistance almost inevitably develops [
5]. Moreover, trastuzumab has low permeability across the blood-brain barrier (BBB), and therefore, its benefit is attributed primarily to better control of extra-cranial disease [
6‐
9]. These observations have prompted the development of various strategies to enhance the efficacy of trastuzumab and to overcome resistance, notably with the development of potent small-molecule tyrosine kinase inhibitors (TKIs) currently in the clinic or undergoing clinical trial [
8,
10‐
13].
Neratinib is an irreversible pan-TKI that targets EGFR/HER1, HER2 and HER4. Its potent activity against HER2-overexpressing cells and tumours has been demonstrated clearly in pre-clinical studies, including against trastuzumab or lapatinib-resistant cells and tumours [
14‐
17]. Several trials aimed at identifying the best clinical setting for this inhibitor in early and metastatic breast cancer patients have been completed or are ongoing (reviewed in [
18,
19]). Its small size and inhibitory activity on ATP-binding cassette (ABC) transporters have led to the suggestion that neratinib may have a better accumulation in the brain than trastuzumab or other TKIs such as lapatinib [
20,
21]. Accordingly, its brain permeability and efficacy against brain metastases are currently under evaluation (NCT01494662). Partial results reported in HER2
+ve patients with established brain metastases who experienced progression after one or more lines of CNS local therapy showed a modest CNS objective response rate (8%) to neratinib monotherapy, but almost half of the patients treated with neratinib/capecitabine combination achieved a volumetric reduction of CNS lesions of ≥ 50% [
22,
23]. The NEfERT-T phase II trial compared the efficacy of neratinib + paclitaxel versus trastuzumab + paclitaxel in women with locally advanced or metastatic HER2
+ve breast cancer [
24]. Interestingly, while no difference was found in progression-free survival between the groups, neratinib + paclitaxel combination reduced the incidence of CNS recurrence and delayed the time to brain metastases [
24]. In the ExteNET randomised trial (NCT00878709) investigating the efficacy of neratinib after 1 year of trastuzumab adjuvant therapy, neratinib improved disease-free survival but did not reduce the cumulative incidence of CNS recurrence at the 5-year mark [
25]. Together, these studies indicate that neratinib may have a place in the clinic for the treatment of patients at risk of brain metastases, but further evaluation in pre-clinical models or in patients is required to determine the most effective treatment regimen to prevent or treat brain metastases.
A major limitation in breast cancer metastasis research is the lack of robust and clinically relevant models of brain metastasis to test novel therapies [
26‐
28]. Most models are human xenografts that lack an intact immune system and require intra-cardiac injection of a bolus of cells for efficient brain metastasis. Given the effect of TKIs on promoting trastuzumab’s antibody-dependent cell-mediated cytotoxicity [
10] and the well-documented role of inflammatory cytokines in suppressing tumour immune surveillance and drug response or promoting vessel permeability, immune-competent syngeneic models have greater clinical relevance [
29‐
31]. Syngeneic MMTV-neu and HER2
+ve Kunming mouse models circumvent the need for immune-compromised mice but are only weakly metastatic to the lung or liver and not to the brain [
32,
33]. Thus, robust syngeneic models that fully recapitulate the spontaneous spread of HER2
+ve breast cancer to the brain and other organs would significantly improve our ability to evaluate the efficacy of targeted therapies against HER2
+ve brain metastases. Here, we describe a new syngeneic mouse model of HER2
+ve breast cancer brain metastasis (TBCP-1) and its response to a panel of TKIs. We demonstrate that the high potency of neratinib is associated with its unique ability to induce cell death by ferroptosis. Furthermore, we present evidence that neoadjuvant neratinib reduces the incidence of brain metastases and provides greater survival benefit than late intervention aimed at treating advanced HER2
+ve breast cancer with established brain metastases.
Methods
Cell culture and reagents
The 67NR (provided by Dr. F. Miller, Karmanos Cancer Institute, Detroit, MI, USA), the 4T1.2 and the brain-metastatic 4T1Br4 mouse mammary carcinoma cell lines were derived from a BALB/C spontaneous mammary tumour and cultured as described previously [
34,
35]. Primary translational breast cancer program-1 (TBCP-1) cells were isolated from a spontaneously arising mammary tumour in a BALB/C mouse (SMT1 kindly donated by Dr. Judy Harmey, Royal College of Surgeons in Ireland) and adapted to culture, from nutrient-rich medium [Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% foetal bovine serum (FBS), epidermal growth factor (EGF, 10 ng/ml), insulin (5 μg/ml), glutamine (2 mM), sodium pyruvate (1 mM) and 1% penicillin/streptomycin] to low nutrient medium [DMEM, 10% FBS, sodium pyruvate (1 mM), 1% penicillin/streptomycin] over several months (Additional file
1: Figure S1A). To facilitate ex vivo imaging of metastatic lesions and quantitation of metastatic burden by genomic qPCR, TBCP-1 cells were transduced with a murine stem cell virus (MSCV)-mCherry retroviral vector as described previously [
36]. Human MCF-7, BT474 and SKBR3 lines were purchased from ATCC and cultured in DMEM supplemented with 10% foetal bovine serum (FBS), 2 mM
l-glutamine, 1 mM sodium pyruvate and 1% penicillin-streptomycin. The metastatic MDA-MB-231 HM variant was provided by Dr. Yi-Feng Hou (Fudan University, Shanghai, China) and cultured in the same medium. For routine culture, cells were passaged when sub-confluent and kept in culture for a maximum of 4 weeks.
Lapatinib ditosylate, erlotinib hydrochloride, tucatinib hydrochloride, afatinib and RAS synthetic lethal 3 (RSL3) were obtained from SelleckChem (Scoresby, VIC, Australia). Erastin and liproxstatin-1 were purchased from Sigma-Aldrich (Castle Hill, NSW, Australia). Neratinib maleate was provided by Puma Biotechnology (Los Angeles, CA, USA). All compounds were prepared at 5–10 mM stocks in DMSO and diluted to the required concentration in the appropriate assay buffer immediately before in vitro and in vivo assays.
In vitro proliferation and IC50 determination
Cell proliferation was measured using a sulforhodamine B (SRB) colorimetric assay as described previously with minor modifications [
34,
37]. Briefly, TBCP-1 (1 × 10
3) or SKBR3 (2 × 10
3) cells were seeded in triplicate wells of a 96-well plate in 100 μl of serum-containing medium and allowed to adhere at 37 °C for 6 h. Inhibitors were added in 100 μl of the same medium and proliferation measured over 5 days. DMSO (≤ 1%
v/
v) was used as vehicle control. Half maximal inhibitory concentrations (IC
50) values were determined in the same assay over 3 days with an initial cell density of 1 × 10
3 (67NR, 4T1.2), 2 × 10
3 (TBCP-1), 3 × 10
3 (SKBR3) or 5 × 10
3 (MCF-7, BT474, MDA-MB-231HM) cells/200 μl/well and IC
50 values calculated using Hill’s equation in the GraphPad Prism 6.0 software. Where indicated, photographs were taken using an Olympus CKX53 inverted microscope to show evidence of cell death.
Sub-confluent TBCP-1 cultures were incubated for 24 h in a serum-containing medium in the presence of neratinib (300 nM) or vehicle DMSO alone. RNA was extracted from replicate wells using the PureLink™ RNA isolation kit according to the manufacturer’s instructions (Thermo Fisher Scientific, Scoresby, VIC, Australia). All RNA samples were normalised to 1 μg for library preparation and then libraries normalised to 10 nM and pooled. Before sequencing, the pool was denatured with 0.2 N NaOH for 5 min at room temperature and the sample pool diluted to 1.8 pM and spiked with 2% PhiX. Sequencing was done on Illumina NextSeq 500 following the TruSeq Stranded mRNA Low Sample protocol according to the manufacturer’s instructions.
Raw reads quality control was performed by FastQC (v0.11.6). Adapter/primer sequences were clipped by trimgalore (v0.4.5). Cleaned reads were mapped to mouse genome (GRCm38.p5) by HiSat2 (v2.0.5) [
38]. Alignment quality was checked with RSeQC (v2.6.4) [
39]. The read counts for genes were computed using Subread (v1.4.6p5) [
40] and then were subjected to edgeR (v 3.18.1) [
41] for differential analysis. Genes were considered differentially expressed when the adjusted
p value of the likelihood ratio was < 0.05. Functional enrichment analysis was carried out using goana and kegga function in EdgeR with adjustment for gene length.
Immunoblotting
Expression of ERα, PR and HER2 in sub-confluent cultures of TBCP-1 cells was detected by standard immunoblotting [
37]. Primary antibodies against ERα (Santa Cruz sc-542, 1 μg/ml), PR (Santa Cruz sc-538, 1 μg/ml) or HER2 (Abcam ab2428, 1 μg/ml) and appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies were used to detect the respective proteins. An anti-GAPDH antibody (Abcam ab8245, 0.2 μg/ml) was used as a loading control.
For the expression of EGFR family of receptors and downstream signalling effectors, sub-confluent cultures were serum-starved overnight in serum-free medium supplemented with 1 mM sodium pyruvate, 2 mM glutamine and 1% penicillin/streptomycin and re-starved for 2 h in fresh serum-fee medium prior to exposure to neratinib for 1 h at 37 °C followed by the addition of EGF (100 ng/ml) (Thermo Fischer Scientific, #PHG0311) for 10 min at 37 °C. Cells were washed with ice-cold PBS and whole-cell lysates prepared in cell lysis buffer (30 mM HEPES, 5 mM EDTA, 150 mM NaCl, 1% (v/v) Triton X-100) supplemented with protease inhibitor cocktail (ROCHE, Sydney, NSW, Australia, #04693132001) and phosphatase inhibitor cocktail (Abcam, ab201112). Primary antibodies against EGFR (E235, Abcam, ab32077, 1/1000 dilution), phospho-EGFR (Y1173, Abcam ab5652, 1/1000 dilution), HER2 (ab2428, Abcam, 1/200 dilution), phospho-HER2 (Tyr877, Cell Signalling Technology, #2241, 1/1000 dilution), HER3 (ab5470, Abcam, 1/100 dilution), HER4 (E200, Abcam, ab 32375; 1/1000 dilution), MAPK (ERK1/2) (L34F12, Cell Signalling Technology, #4696, 1/1000 dilution), phospho-MAPK (Thr 202/Tyr204, Cell Signalling Technology, #9101, 1/1000 dilution), AKT (40D4, Cell Signalling Technology, #2920, 1/1000 dilution) and phospho-AKT (Ser 473, Cell Signalling Technology, #9271, 1/1000 dilution) were used to detect the respective proteins and specific binding detected using appropriate HRP-conjugated secondary antibodies and enhanced chemiluminescence (ECL) reagents (Amersham Biosciences, Castle Hill, NSW, Australia).
Ferroptosis, metabolic and apoptotic markers were analysed in whole-cell lysates from TBCP-1 sub-confluent cultures treated with DMSO (vehicle control) or neratinib (300 nM) or the BH3 mimetics ABT263 (0.5 μM) + MCL1 inhibitor S63845 (0.5 μM) for 6 h as indicated in the figure legend. Protein bands were detected with the following primary antibodies and appropriate HRP-conjugated secondary antibodies: Acyl-CoA synthetase long-chain family member 4 (ACSL4) (sc-271800, Santa Cruz Biotechnology, Santa Cruz, CA, USA, 1/1000 dilution), ferritin (ab75973, Abcam, 1/2000 dilution), transferrin receptor-1 (TFR-12-M, Alpha Diagnostics, San Antonio, TX, USA, 1/1000 dilution) and ferroportin-1 (NBP1-21502, Novus Biologicals, 1 μg/ml). Protein band intensity relative to GAPDH (Abcam ab8245, Abcam, 0.2 μg/ml) was quantitated using ImageJ software (National Institute of Health, Bethesda, MD, USA). For caspase 3 analysis (Cell Signalling Technology #9662, 1/1000 dilution), an anti-α-tubulin antibody (clone AA13, Sigma, 0.2 μg/ml) was used as a loading control.
Inductively coupled plasma mass spectrometry
Sub-confluent cultures of TBCP-1 cells (5 replicates/condition) were treated for 72 h with vehicle alone (DMSO control) or neratinib (300 nM and 500 nM) and the cells pelleted by centrifugation. Fifty microlitres of concentrated nitric acid (65% v/v, Suprapur, Merck) was added to each cell pellet overnight at room temperature. Samples were heated at 90 °C for 20 min, and final volumes made up to 500 μl 1% (v/v) nitric acid. Iron content was measured using an Agilent 7700 series ICP-MS instrument under routine multi-element operating conditions in a Helium Reaction Gas Cell. The instrument was calibrated using 0, 5, 10, 50, 100 and 500 ppb of certified multi-element ICP-MS standard calibration solutions (ICP-MS-CAL2-1, ICP-MS-CAL-3 and ICP-MS-CAL-4, Accustandard) for a range of elements. A certified standard solution containing 200 ppb of yttrium (Y89) was used as an internal control (ICP-MS-IS-MIX1-1, Accustandard). The raw ppb results were normalised to final volume and converted to ng/106 cell of metal using the formula: Final concentration (ng/106 cell) = Raw ppb value (ng/mL) × sample volume (0.5 ml)/number of cells × 106.
All procedures involving mice conformed to the National Health and Medical Research Council animal ethics guidelines and were approved by the Austin Health Animal Ethics Committee (A2016/05346) and the Peter MacCallum Animal Ethics & Experimentation Committee (E507). Female BALB/C mice (5/box) were maintained in a specific pathogen-free environment with food and water freely available and monitored daily for signs of ill health or metastatic disease.
Experimental metastasis assays were done as described previously with minor modifications [
34,
37]. For initial characterisation of metastatic spread, TBCP-1 cells (5 × 10
5) were injected into the left cardiac ventricle of 6–8-week-old female BALB/C mice. The mice were monitored daily and sacrificed after 3 weeks or earlier if signs of metastases became apparent (weight loss > 10%, ruffled fur, lethargy, rapid breathing) (Additional file
1: Figure S1B). The organs were harvested, photographed, fixed in 10% buffered formalin for 24 h and processed for histology. Tumour growth and spontaneous metastasis assays were completed as described previously [
34,
42]. For these assays, fluorescent mCherry-tagged TBCP-1 cells (1 × 10
6/20 μl PBS) were inoculated into the fourth mammary fat pad and tumour growth measured thrice weekly with electronic callipers. Tumour volume was calculated using the formula (length × width
2)/2. To document TBCP-1 distribution of spontaneous metastases, tumours were resected when they reached an average volume of ~ 400 mm
3 (~ 3 weeks), and the experiment was terminated when the mice developed signs of metastatic disease (~ 7 weeks) (Additional file
1: Figure S1C). The incidence of mice with visible lung, liver, kidney, ovary and adrenal gland metastases was measured semi-quantitatively by visual examination at necropsy. For quantitation of metastasis incidence, the brains and bones were fixed for 24 h in 4% paraformaldehide (PFA) and paraffin-embedded. The bones were decalcified in 20% ethylenediaminetetraacetic acid (EDTA) prior to embedding. Three step-sections/organs (4 μm for bone or 6 μm for brain, 100 μm apart) were stained with haematoxylin and eosin (H&E) and the presence of brain or bone lesions confirmed on an Olympus BH2 phase contrast microscope.
The efficacy of neratinib was evaluated in the neoadjuvant or metastatic setting (Additional file
3: Figure S3B-3C). For neoadjuvant therapy, mice were treated daily with control vehicle [(0.5% (
w/
v) methylcellulose, 0.4% (
v/
v) polysorbate-80] or neratinib (60 mg/kg) by oral gavage, starting when mammary tumours reached ~ 100 mm
3 (~ 1 week). Treatment was stopped and tumours resected when control tumours reached ~ 400–500 mm
3 or after a maximum of 3 weeks. In the experimental metastasis setting, treatment commenced 2 days after the intra-cardiac inoculation of the cells, and ceased after a maximum of 3 weeks, or earlier if mice showed signs of advanced metastatic disease. Mice were sacrificed individually when they developed signs of advanced metastatic disease, and survival was plotted by Kaplan-Meier analysis. The incidence of mice with brain metastases at endpoint was determined by ex vivo fluorescence imaging using a Maestro™ In-Vivo Imaging System (CRi) and, where indicated, further confirmed by histology and cytokeratin staining (see below). Relative metastatic burden in the lungs, femurs and spine of the control and treatment groups was quantitated by genomic qPCR detection of the mCherry marker gene present in tumour cells only, relative to the vimentin gene present in tumour and host cells, as described previously [
34,
37,
42].
Histology and immunohistochemical staining
Tissues were fixed in 10% buffered formalin for 24 h and processed for paraffin embedding. The sections (4 μm for the bone or 6 μm for the brain) were cut using a Leica RM 2245 Microtome and metastatic lesions in the bones or brains identified by standard haematoxylin and eosin (H&E) staining. Epithelial (cytokeratin), proliferative (Ki-67) ER, PR and HER2 status were assessed by standard IHC staining as described [
34,
43]. Primary antibody for pan-cytokeratin (Sigma C801, 2.5 μg/ml), Ki-67 (Millipore AB9260, 2 μg/ml), ERα (Santa Cruz sc-542, 1 μg/ml), PR (Santa Cruz sc-538, 1 μg/ml) and HER2 (Calbiochem 3B5, 10 μg/ml) and appropriate biotin-conjugated secondary antibodies were used to detect the respective proteins. Staining was visualised using a 3,3′-diaminobenzidine (DAB) liquid substrate system (Sigma-Aldrich). Quantitative scoring of IHC staining was done on scanned whole tumour and brain sections (
n = 3) using Aperio ImageScope software v11.1.2.760. For ERα, PR and Ki-67 scoring, the data are expressed as percentage of positive nuclei in whole tumours (excluding necrotic areas) or in delineated cytokeratin
+ve brain lesions. Pan-cytokeratin was scored as percentage of positive cells (membrane and/or cytoplasmic expression). For HER2 quantitation, the sections were scored as 3+ for strong continuous membrane expression, 2+ for moderate continuous membrane expression and 1+/0 for weak/discontinuous or no expression.
Statistical methods
Data were analysed for statistical significance using GraphPad Prism 6 software. Unless otherwise indicated, data from in vitro experiments are presented as mean ± SD and data from in vivo experiments are presented as mean ± SEM. Statistical significance between the two groups was analysed by Mann-Whitney, non-parametric test. Experiments with more than two groups were analysed with one-way analysis of variance (ANOVA) and Tukey’s multiple comparison test. Statistical analysis of tumour volume was assessed by two-way ANOVA and Bonferroni’s post-test. Survival proportions were determined using Kaplan-Meier analysis and further analysed with a Gehan-Breslow-Wilcoxon test. Incidence data were analysed by Fisher’s exact test. p < 0.05 was considered significant.
Discussion
TKIs are increasingly used in the clinic for the treatment of HER2
+ve breast cancer. Encouraging pre-clinical and clinical data indicate that TKIs, alone or in combination, may be particularly useful for the treatment of CNS metastases [
11], but further investigation is required to identify the best clinical setting for this class of inhibitors. This has been difficult, however, in the absence of robust models of HER2
+ve breast cancer that aggressively spread from the mammary gland to the brain. The TBCP-1 model characterised herein is unique in many respects and fills an important gap in pre-clinical cancer research. We show that TBCP-1 cells express high levels of HER2 but lack expression of hormone receptors. This phenotype is functionally relevant since TBCP-1 cell proliferation is inhibited by several TKIs targeting HER2 but not by anti-oestrogens (Fig.
1). Its clinical relevance is further illustrated by the fact that TBCP-1 tumours and metastases maintain expression of HER2 and response to neratinib in vivo (evidenced by inhibition of HER2 signalling, tumour growth and metastasis) (Figs.
5 and
6) and metastasise to multiple organs commonly colonised by human HER2
+ve breast cancer, including the brain (Fig.
2). To our knowledge, TBCP-1 is the only HER2
+ve breast cancer model that spreads spontaneously and avidly from the mammary fat pad to brain in immune-competent mice.
We have yet to determine whether high membrane expression of HER2 protein detected by flow cytometry, immunoblotting and immunohistochemistry occurs through gene amplification or alternative mechanisms. While there is generally good concordance between HER2 protein expression and gene amplification in breast cancer patients, high HER2 protein expression without gene amplification occurs in approximately 10% of cases [
60]. Importantly, the same study showed no difference in clinical outcome (overall survival) in patients with intermediate (2+) to high (3+) HER2 protein-expressing tumours with or without gene amplification. These observations are consistent with the aggressive nature of the TBCP-1 model. IHC analyses of HER2 status in tumours and brain metastases revealed spatially heterogeneous levels of membrane HER2 expression but overall score consistent with a HER2
+ve phenotype. Intra-tumour spatial heterogeneity has been reported also in amplified and non-amplified HER2-overexpressing human breast tumours and could impact on the efficacy of HER2-targeted therapies [
61‐
63]. Hence, it will be important in future studies to assess whether high HER2 expression and spatial intra-tumour heterogeneity may be explained by the differences in gene amplification or due to the differences in microenvironmental factors within specific regions of tumours and metastases.
Sensitivity to neratinib is closely correlated with the level of HER2 expression in human cell lines [
16]. While neratinib inhibited the proliferation of TBCP-1, BT474 and SKBR3 cells, BT474 and SKBR3 cells were more strongly inhibited than TBCP-1 cells (IC
50 ~ 1 nM, 7 nM and 117 nM, respectively). Other mouse and human TNBC or luminal cell lines tested showed varying degree of sensitivity to neratinib that largely correlated with the level of HER2 or EGFR expression, both targeted by neratinib. BT474 and SKBR3 cells express particularly high levels of HER2 due to gene amplification which might explain their greater sensitivity. Alternatively, differences in downstream signalling might contribute in part to the differential response of these cell lines to neratinib. Phosphorylation of receptor tyrosine kinases triggers the activation of multiple signalling pathways that are inhibited by TKIs [
15,
64]. In particular, Canonici et al. [
16] reported that the sensitivity of HER2-amplified cell lines to neratinib correlates with the extent of p-AKT and p-ERK inhibition. Consistent with this, neratinib almost completely blocked phosphorylation of HER2 as well as downstream activation of AKT and MAPK (ERK-1/2) in SKBR3 cells. In contrast, neratinib significantly inhibited HER2 and AKT phosphorylation but failed to inhibit ERK phosphorylation in TBCP-1 cells (Fig.
3). Interestingly, aberrant activation of downstream ERK signalling through a FOXO-dependent but Ras-independent feedback mechanism has been associated with resistance to HER2-targeting inhibitors [
65]. Further work will be required to clarify whether the same feedback loop may operate in TBCP-1 cells and contribute in part to their lower sensitivity to neratinib compared to SKBR3 cells. Our finding from the KEGG pathway analysis showing enrichment of upregulated genes associated with both MAPK and FOXO signalling pathways in neratinib-treated TBCP-1 cells (Additional file
5: Table S2) would be consistent with this possibility.
While neratinib can inhibit HER1, HER2 and HER4, TBCP-1 lack the expression of HER4, and accordingly, neratinib potently inhibited HER2 and partially reduced EGFR phosphorylation (Fig.
3). Moreover, the specific EGFR/HER1 inhibitor, erlotinib, only inhibited SKBR3 cells at high nanomolar concentrations and did not inhibit TBCP-1 proliferation even at micromolar concentrations suggesting that TBCP-1 cell proliferation is not dependent on signalling from EGFR homodimers. Similarly, afatinib, an irreversible inhibitor with ~ 25-fold higher potency against HER1 compared to HER2 [
66] only inhibited TBCP-1 at high nanomolar to low micromolar concentrations. Conversely, the selective HER2-reversible inhibitor, tucatinib (> 500-fold selectivity for HER2 compared to HER1) [
67,
68], was the second most potent inhibitor (IC
50, 191 nM) after neratinib (IC
50, 117 nM). Collectively, these observations indicate that the proliferation of TBCP-1 is critically dependent on HER2 activity and that this receptor is the primary target for neratinib in these cells.
Another key finding from our study is the potent induction of caspase-independent ferroptotic cell death induced by neratinib that could be rescued by the ferroptosis inhibitor liproxstatin-1 (Fig.
4). This response was also evidenced by an uptake of intracellular iron and altered expression of iron metabolism regulators. Amongst all TKIs tested, this property was unique to neratinib and correlated with its superior activity against mouse and human HER2
+ve cells. ACSL4, a regulator of fatty acid oxidation or lipid biosynthesis, was identified recently as a biomarker of sensitivity to ferroptosis preferentially expressed in basal-like breast cancer cell lines but also found to be elevated in HER2
+ve SKBR3 cells [
55]. These observations raise the interesting possibility that the promotion of ferroptosis could be an effective strategy to enhance the efficacy of TKIs against metastatic HER2
+ve as well as basal-like breast cancers. In addition, the work of Doll et al. [
55] and our study suggest that high levels of ACSL4 or susceptibility to ferroptosis could be predictive biomarkers of neratinib response and facilitate the patient selection. This is currently being investigated.
We were particularly interested in clarifying the properties that are required for TKI targeting EGFR family of receptors to induce ferroptosis. Prior to our study, only sorafenib, a multi-kinase inhibitor that does not target EGFR family members, had been reported to induce ferroptosis in various cancer cell lines although breast cancer lines were not investigated [
50]. Lapatinib was shown also to enhance the pro-ferroptotic activity of siramesine, a lysosome destabilising agent, but in agreement with our study, lapatinib alone was not sufficient to induce this response [
69]. Comparing the effect of neratinib on multiple mouse and human lines or to other TKIs selected for our study and results reported by others [
50,
69] allows us to draw some conclusions. Ferroptosis appears to be correlated with the level of HER2 expression and independent of ERK activation since neratinib induced this response equally well in SKBR3 and TBCP-1 cells but without significant inhibition of ERK-1/2 phosphorylation in TBCP-1 cells (Fig.
3). This conclusion was reached also for sorafenib-induced ferroptosis [
50]. None of the reversible inhibitors used herein (erlotinib, lapatinib and tucatinib) induced ferroptosis. Surprisingly, the irreversible pan-inhibitor afatinib alone was unable to illicit this response. It should be noted, however, that afatinib is significantly more potent against HER1 than HER2 [
66] and was a poor inhibitor of TBCP-1 cells (IC
50, 747 nM). On the basis of these observations, we propose that potent and sustained inhibition of HER2 is necessary to induce ferroptosis. Alternatively, targeting multiple downstream signalling pathways may partially circumvent the need for irreversible inhibition of HER2, as suggested by the induction of ferroptosis by the reversible multi-kinase inhibitor sorafenib and enhanced ferroptosis induced by the siramesine + lapatinib combination [
50,
69]. Defining the precise mechanisms by which neratinib induces ferroptosis will require further investigation.
Whether the limited efficacy of HER2 inhibitors against brain metastases is due to the poor permeability of inhibitors across the BBB or acquired resistance induced by the brain environment or both is still debated [
70]. Pre-clinical studies showing that neratinib inhibits drug efflux pumps have suggested that neratinib may have better retention in the brain [
21,
71]. However, in advanced patients with brain involvement, neratinib monotherapy showed only a modest CNS objective response rate (8%), indicating that this may not be the best clinical setting for neratinib [
23]. On the other hand, the NEfERT-T trial (NCT00915018) that compared the efficacy of neratinib + paclitaxel versus trastuzumab + paclitaxel as the first-line therapy in women with previously untreated recurrent or metastatic HER2
+ve breast cancer showed lower incidence of CNS recurrence in the neratinib-treated compared to the trastuzumab-treated group (8.3% versus 17.3%, respectively) and delayed time to CNS metastases [
24] indicating that patients at risk of brain metastasis may benefit from earlier intervention.
The spontaneous nature of the TBCP-1 model of HER2 breast cancer metastasis allowed us to test the efficacy of neratinib in a preventive neoadjuvant and late metastatic setting (Fig.
5). Results from experimental metastasis assay showed a modest improvement in survival and a trend towards the reduced incidence of brain lesions (60% versus 78% in the control group), a result roughly in line with that reported in patients [
23]. In contrast, in the neoadjuvant setting, neratinib significantly extended survival and dramatically reduced the overall metastatic burden. Remarkably, none of the mice treated with neratinib had detectable brain lesions compared to 62.5% in the control group. To our knowledge, this is the first pre-clinical demonstration that first-line neratinib neoadjuvant therapy provides significant benefit with regard to CNS recurrence. In our study, metastatic burden after neratinib treatment was analysed at the endpoint. Whether neoadjuvant neratinib reduces brain metastasis by preventing dissemination from the primary tumour, by targeting circulating tumour cells before they home to the brain or by preventing the outgrowth of perivascular micro-metastases in the brain will require more in-depth kinetic studies, including early adjuvant therapy initiated after primary tumour resection. While it is too early to know its impact on brain recurrence in HER2
+ve breast cancer patients, the I-SPY2 trial (NCT01042379) evaluating neratinib’s efficacy in the neoadjuvant setting has reported encouraging observations, with an estimated rate of pathological complete response superior to that of the trastuzumab-containing control arm (56% versus 33%, respectively) [
72]. Collectively, our results strongly argue for the use of neoadjuvant neratinib and support ongoing trials in this setting.
Conclusions
We have developed the only mouse model of spontaneous breast cancer metastasis that closely mimics the aggressive spread of HER2+ve breast cancer to the brain and other organs in immune-competent mice. The TBCP-1 model will provide a unique platform to explore new therapies against brain-metastatic HER2+ve breast cancer, including immunotherapies, an area currently understudied. The results from transcriptomic analyses and in vitro investigation identified a new mechanism of action for neratinib and demonstrated that its superior efficacy correlates with the unique ability to induce ferroptosis.
We sought to evaluate the best clinical setting for neratinib in vivo, in particular for the prevention or treatment of brain metastases. Our data show that neratinib neoadjuvant therapy effectively reduces TBCP-1 metastatic burden and extends survival. We propose that prevention of metastatic progression using a neoadjuvant treatment protocol could be more efficacious and provide a greater survival benefit to HER2+ve breast cancer patients than late intervention, particularly against the development of difficult to treat brain metastases.
Despite the potent efficacy of neoadjuvant neratinib, detection of small lung or liver nodules in some mice indicates that resistance can develop in vivo and that combination therapy may be required to completely prevent metastatic progression. Indeed, evidence that resistance can develop after prolonged exposure to neratinib is now emerging, and various mechanisms have been proposed including altered expression of EGFR family members and increased neratinib metabolism, which could reduce its bioavailability [
73,
74]. Investigating the mechanisms of resistance to neratinib-induced ferroptosis or the efficacy of combination therapies, including immunotherapy likely to enhance neratinib’s activity [
75], will require the use of an appropriate preclinical model, such as TBCP-1, in immune-competent mice.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.