Dual endothelin receptor inhibition enhances T-DM1 efficacy in brain metastases from HER2-positive breast cancer

The effective treatment of cerebral metastases from HER2-positive breast cancer remains an unmet need. Recent studies indicate that activated astrocytes and brain endothelial cells exert chemoprotective effects on cancer cells through direct physical interaction. Here we report that the endothelin axis mediates protection of HER2-amplified brain metastatic breast cancers to the anti-HER2 antibody–drug conjugate ado-trastuzumab emtansine (T-DM1). Macitentan, a dual inhibitor of endothelin receptors A and B, improves the efficacy of T-DM1 against breast cancers grown in the brain. We show that direct contact of brain stroma with cancer cells is required for protection to T-DM1. Our data suggest that targeting the endothelin axis may be beneficial when anti-signaling agent and cytotoxic agent are combined. These findings may contribute to the development of therapeutic approaches with enhanced efficacy in the brain microenvironment.


INTRODUCTION
The brain microenvironment has a key role in metastatic tumor progression. Microglia, activated astrocytes, and brain endothelial cells may support survival and growth of brain metastases (BM). 1 Studies indicate that astrocytes and brain endothelial cells have chemoprotective effects on cancer cells through an endothelin pathway-mediated mechanism. 2 Specifically, physical interaction between metastatic cancer cells and brain stromal cells increased expression of endothelins (ET) in astrocytes, activating the endothelin pathway in cancer cells and promoting their survival. 2 Treatment with macitentan, a dual inhibitor of ET A and ET B receptors abolished the astrocyte-and endothelial cell-mediated chemoprotection, resulting in apoptosis increase after paclitaxel chemotherapy in brain metastasis models from breast and lung cancer. 3 We recently demonstrated that the antibody-drug conjugate ado-trastuzumab emtansine (T-DM1) has preclinical activity against BM from HER2-positive breast cancer. 4 T-DM1 increased mitotic catastrophe through the DM1 component, facilitated by selective trastuzumab-mediated targeting. Whether endothelin pathway activation plays any role in this setting is unknown. We hypothesized that dual ET A and ET B inhibition can enhance the activity of T-DM1 against BM from HER2-positive breast cancer.

RESULTS
We treated mice bearing human HER2-amplified BT474 brain tumors, engineered to express secretable Gaussia luciferase (BT474-Gluc), 4 with T-DM1 alone or in combination with macitentan. Adding macitentan to T-DM1 delayed brain tumor growth. Median mouse survival improved from 42 days in the T-DM1 group to 53 days for T-DM1 plus macitentan (HR = 0.79, p = 0.042) (Fig. 1a). 5 Macitentan monotherapy did not affect tumor growth or survival. We next investigated the mechanism of synergy between macitentan and T-DM1. Western blot analysis of protein extracted from BT474-Gluc BM three days after treatment initiation revealed decreased AKT phosphorylation, as well as decreased expression of the anti-apoptotic protein Bcl2 (Fig. 1b). 5 Significant apoptosis increase was found in the T-DM1 plus macitentan group (one-way ANOVA, p < 0.001) (Fig. 1c, d). 5 Previous reports suggested that physical interaction between cancer cells and brain-stromal cells is required for endothelinmediated chemoprotection. 2 We used organotypic brain-slice cultures to evaluate the role of cell-cell contact (Supplemental Fig.  1A). 5 BT474-Gluc cells were either injected directly onto organotypic brain slices or grown in brain-slice conditioned media. Adding macitentan to T-DM1 delayed BT474-Gluc growth significantly, when growing in direct contact with the brain slice ( Fig. 1e). 5 However, when cancer cells were exposed to brain-slice conditioned media without physical contact, macitentan did not enhance T-DM1 activity (Supplemental Fig. 1B-C). 5 These data confirm that direct contact with cancer cells is required for stromal cell-mediated protection.
To determine molecular changes in vivo, we investigated ET A and ET B expression in BT474-Gluc tumors implanted in the brain or mammary fat pad (MFP), as well as in breast cancer MDA-MB-361-Gluc brain tumors. As expected, HER3 was overexpressed in brain samples (Fig. 1f). 5,6 No clear difference in endothelin receptor expression between the isogenic BT474-Gluc tumors in brain and MFP was seen (Fig. 1f). 5 However, we observed differential expression of ET A and ET B between BT474-Gluc and MDA-MB-361-Gluc brain tumors. Moreover, we investigated HER2downstream signaling three days after treatment initiation, when cells were sensitive to T-DM1 plus macitentan, and at the endpoint of the study, after cells became treatment-resistant. Western blot revealed a decrease in expression and phosphorylation of HER2 and HER3, compared to control on day 3 (Fig. 1g). 5 Of note, a striking decrease in HER2 and HER3 expression and activation was noticed at the study endpoint (Fig. 1g). 5 Short-term treatment with T-DM1 alone did not clearly affect HER2 phosphorylation, compared to control. However, HER2 and HER3 were also strongly decreased at the study endpoint (Supplemental Fig. 1D-E). 5 Despite a decrease of total AKT for the combination of T-DM1 with macitentan at the endpoint, AKT phosphorylation was still active, whereas ERK signaling remained activated (Fig. 1g). 5 Moreover, Bcl2 expression was restored in tumors resistant to T-DM1 plus macitentan (Fig. 1h). 5 These effects were associated with a strong PTEN reduction (Fig. 1h). 5 PTEN regulates pro-survival signaling, 7 and mediates Bcl2 expression at the translational level. 8 The brain microenvironment can promote PTEN loss in cancer cells. 9 Thus, PTEN loss may promote resistance of BT474-Gluc brain tumors to treatment with T-DM1 plus macitentan.
We next investigated the effects of dual endothelin-receptor inhibition on the efficacy of HER2-signaling inhibition. We combined macitentan with the HER2/EGFR small-molecule inhibitor neratinib. The rationale for choosing neratinib was two-fold: i) it inhibits irreversibly erbB1, HER2 and erbB4, and ii) data suggest drug penetration across the BBB. 10 Our studies showed minimal tumor growth delay for neratinib alone (Supplemental Fig. 1F), 5 which is not-surprising, since HER2-targeted therapies have shown limited efficacy against BM. 6,11 Adding macitentan to neratinib did not improve survival, compared to control-treated mice (Supplemental Fig. 1G). 5 These data support the hypothesis that the role of the endothelin axis may be more specific to chemoprotection in BM.
T-DM1 was effective against brain metastases from HER2amplified breast cancer independent of the PIK3CA mutation status. 4 We investigated the effects of adding macitentan to T-DM1 on its efficacy against MDA-MB-361-Gluc BM, which are HER2-amplified but also harbor an activating PIK3CA mutation . f Western blots were performed in BT474-Gluc tumors growing in the brain or the mammary fat pad (MFP), and in MDA-MD-361-Gluc brain tumors. HER3 was used as control, since it has been previously established that the brain microenvironment increases HER3 expression 6 g, h. Western blots were performed in BT474-Gluc brain tumors, collected 48 h after treatment with T-DM1 + macitentan (T-DM1 + mac. short) and at the endpoint of the study (T-DM1 + mac. long). Tumors from mice treated with unspecific IgG were collected and used as control. i Schematic of brain-stroma mediated protection to the HER2-targeted ADC T-DM1. Direct contact of reactive astrocytes and brain endothelial cells with HER2-amplified breast cancer cells can reduce the activity of ado-trastuzumab emtansine (T-DM1) through an endothelin-pathway-mediated mechanism. Dual endothelin receptor inhibition with macitentan increases T-DM1 induced apoptosis and enhances the activity of the ADC in the brain microenvironment V. Askoxylakis et al.
(E545K). Our studies confirmed the efficacy of T-DM1, however, adding macitentan did not improve its activity (Supplemental Fig.  1H). 5 Similar to the BT474 model, comparison of HER2 signaling revealed a decrease in HER2 and HER3 expression and phosphorylation after the MDA-MD-361-Gluc tumors became resistant to treatment. Of note, AKT and ERK phosphorylation were enhanced in MDA-MB-361 tumors (Supplemental Fig. 1I), 5 an effect that was stronger when macitentan was added to T-DM1 (Supplemental Fig. 1I). 5 The underlying mechanisms of the increased signaling in response to dual endothelin-receptor inhibition in MDA-MB-361 brain tumors remain unclear. Endothelin receptor expression analysis of MDA-MD-361-Gluc brain tumors showed decreased expression of ET B , compared to BT474-Gluc brain tumors. This could provide an additional explanation for the lack of activity of macitentan, allowing the hypothesis that endothelin-mediated resistance in the brain microenvironment may not only be mediated by homo-and hetero-receptor expression and compensatory oncogenic downstream signaling, but also by receptor activation through increased ligand expression by brain stromal cells. Indeed previous studies show that peritumoral astrocytes overexpress endothelins in more than 80% of human brain metastases. 12 Collectively, these data indicate that the effects of dual endothelin-receptor inhibition may vary between tumors with different characteristics, thus underscoring the need for personalized treatment.

DISCUSSION
In conclusion, our studies show a potential role of the endothelin axis in protection of HER2-amplified brain metastatic cells to an anti-HER2 antibody-drug-conjugate (Fig. 1i). We show in clinically relevant brain-metastasis models that dual endothelin-receptor inhibition improved T-DM1 activity, and that this effect required direct contact of brain-stroma with cancer cells. Dual endothelinreceptor blockade did not overcome resistance to a HER2signaling inhibitor in the brain microenvironment in the absence of cytotoxic treatments. Conversely, it did not improve sensitivity to cytotoxic therapy in HER2-amplified brain metastases with activating PIK3CA mutation, characterized by absence of HER2downstream signaling inhibition. These findings suggest that the beneficial effects of endothelin inhibitors on T-DM1 involve both the inhibition of intrinsic signaling pathways and the cytotoxic component. Considering that effective treatment of brain metastases from HER2-positive breast cancer remains an unmet need, our findings may contribute to the development of therapeutic approaches with enhanced efficacy in the brain microenvironment. Further evaluation of endothelin signaling and the impact of endothelin-receptor inhibition in brain metastases from HER2-positive breast cancer is warranted.

Statistical analysis
Statistical analysis was performed using Prism 6 (GraphPad Software Inc., La Jolla, CA). All statistical tests were two-sided. A difference was considered statistically significant when P < 0.05. Comparisons of continuous outcomes were performed using one-way ANOVA test with Bonferroni-Holms correction for categorical and t test for dichotomous variables. Kaplan-Meier analysis and log-rank test were used for survival analysis. Measurements were taken from distinct samples, except from in vivo tumor growth measurements, where the same sample was measured repeatedly.

DATA AVAILABILITY
The raw image and graph data generated and analysed during this study are described in the following data record: https://doi.org/10.6084/m9.figshare.7462034. All the data files are available on request from the R.K.J. at the following email: PAtoDrJain@steele.mgh.harvard.edu.