The CARMA3–Bcl10–MALT1 Signalosome Drives NFκB Activation and Promotes Aggressiveness in Angiotensin II Receptor–Positive Breast Cancer Free

While significant strides have been made in the treatment of breast cancer, resulting in a decline in mortality over the past two decades (1), progress has been incomplete, and the medical armamentarium for treating this disease still relies heavily on nonspecific, cytotoxic chemotherapeutics. Changing this scenario necessitates the development of more innovative and targeted therapies based on an increasingly nuanced understanding of breast cancer subtypes and their molecular drivers. One of the most significant advances in this regard has been in the development of therapeutic antibodies against ERBB2 (HER2), which drives 20%–25% of breast cancers (2). However, most breast cancers (∼75%) progress independently of HER2, so there remains a pressing need to discover the molecular determinants for these cancers, thereby uncovering new therapeutic targets for the next era of personalized medicine. This is particularly applicable to triple-negative breast cancers, for which there is no targeted therapy, and for luminal B cancers that may behave aggressively and prove resistant to hormonal therapy, despite their expression of estrogen receptor (ER; refs. 3, 4).

Several years ago, we collaborated in a multidisciplinary effort to identify new receptors that might drive oncogenic progression in unique subsets of breast cancer (5). Using a bioinformatics approach termed Meta-Cancer Outlier Profile Analysis (MetaCOPA), we sought to nominate candidate oncogenes from multiple independent breast cancer profiling datasets, based on aberrantly high expression. We analyzed 31 published datasets comprising nearly 3,200 independent microarray experiments and found that the second most consistently high-scoring gene (after HER2) is AGTR1, which encodes the angiotensin II type I receptor. AGTR1 is a member of the G protein–coupled receptor (GPCR) superfamily and is best known for its role in vascular biology (6, 7). On the basis of MetaCOPA analyses, we estimated that AGTR1 overexpression occurs in 15%–20% of breast cancers, and found that AGTR1 and HER2 overexpression define mutually exclusive subsets of disease.

Despite the identification of AGTR1 as a potential pathogenic driver in a subset of breast cancers, little is known regarding the mechanisms by which AGTR1 might affect signaling in breast cancer cells, or the phenotypic consequences of its overexpression. AGTR1 is capable of engaging a number of signaling pathways in vascular endothelial and smooth muscle cells, some of the cell types where it has been best studied (6, 8). As a result, it seems likely that AGTR1 overexpression coopts several pathways for aberrant activation in breast cancer. Indeed, Oh and colleagues have clearly shown both aberrant ERK and SMAD3/4 activity in MCF7 breast cancer cells engineered to overexpress AGTR1 (9).

With the current work, we utilized newer profiling databases, including The Cancer Genome Atlas (TCGA) and METABRIC, to further interrogate AGTR1 as an oncogene in breast cancer. In addition, we sought to explore the hypothesis that AGTR1 might mimic the actions of HER2 with regard to activation of NFκB, one of the major downstream mediators driving pathogenesis of HER2⁺ breast cancer. This hypothesis was particularly compelling since AGTR1 and HER2 overexpression are mutually exclusive in breast cancer (5), suggesting that the two receptors direct redundant pathways as a means of promoting tumor progression. In support of this hypothesis, we find that AGTR1 harnesses a unique signaling pathway for activation of NFκB, which involves assembly of the CARMA–Bcl10–MALT1 signalosome, best known as a critical regulator of immune responses in lymphocytes (10, 11). In breast cancer cells, AGTR1-dependent activation of this NFκB pathway initiates a distinct set of responses, causing cells to adopt a proliferative, migratory, invasive, and proangiogenic phenotype.

AGTR1 has long been successfully targeted in the practice of cardiology by therapeutics that include both receptor antagonists [angiotensin receptor blockers (ARB) such as losartan] and inhibitors of ligand production [angiotensin converting enzyme inhibitors (ACE inhibitors) such as captopril; ref. 12]. In addition, novel inhibitors of MALT1 are now being described, including some that have a history of use in psychiatric disorders (e.g., the phenothiazines, mepazine, and thioridazine; refs. 13, 14). As a result, there exists an opportunity to explore repurposing of these legacy drugs in the novel arena of breast cancer therapy, provided we appropriately identify and select breast cancer patients with AGTR1 overexpression who might benefit from this combination therapy. The work described here provides preclinical validation for this concept and motivation to pursue this goal.

A detailed description of reagents and their sources can be found in the Supplementary Methods.

BT549, HCC1500, ZR75-1, Hs578T, Hs606T, CRL-7548, and MDA-MB231 cells were obtained directly from ATCC, with cell line identities confirmed by short tandem repeat (STR) profiling by the source. Frozen aliquots of cells were prepared upon receipt and all cell lines were passaged for less than 6 months. SKBR3 cells were kindly provided by Dr. Ira Bergman (Department of Pediatrics, University of Pittsburgh, Pittsburgh, PA) and the identity of this line was authenticated by STR profiling at the University of Arizona Genetics Core (UAGC, Tucson, AZ). Primary human umbilical vein endothelial cells (HUVEC) were obtained from Lonza and were maintained in culture for no more than 7 passages. BT549, HCC1500, ZR75-1, and SKBR3 cells were grown in phenol red–free RPMI1640 media (catalog no: 11835030, Gibco), whereas MDA-MB-231 were grown in DMEM-Glutamax media, both supplemented with 10% FBS, 1% penicillin/streptomycin (Gibco), and MycoZap Prophylactic (catalog no: VZA-2032, Lonza). HUVECs were grown in VascuLife EnGS Endothelial Complete Medium (catalog no: LL-0002, Lifeline Technology). Lenti-Pac 293Ta cells (catalog no: CLv-PK-01) were purchased from Genecopoeia for lentiviral packaging. These cells were grown in DMEM-Glutamax media. All cells were grown at 37°C in a 5% CO₂ incubator. Cell lines were regularly monitored for mycoplasma contamination using the mycoplasma MycoAlert Detection Kit (catalog no: LT07-318, Lonza). All cell lines were periodically reauthenticated by STR profiling using one of two services (ATCC or UAGC). Unless otherwise indicated, cells were serum starved for 6–12 hours prior to treatments with either Ang II (1 μmol/L), TNFα (10 ng/mL) or PBS vehicle, ± inhibitors as appropriate (IKK-VI, 1–5 μmol/L; losartan, 5 μmol/L).

ZR75-1 cells were transfected with either pReceiver-AGTR1-FLAG (ZR751-AGTR1) or pReceiver-FLAG (ZR751-Neo) using Lipofectamine 2000 (Thermo Fisher Scientific). After 48–72 hours, 0.4 mg/mL geneticin (G418; catalog no: 10131027, Thermo Fisher Scientific) was added to the media and cells were cultured for two weeks. Resulting G418-resistant clones were pooled and expanded further in the presence of 0.4 mg/mL G418.

Lentiviral plasmids harboring either control or Bcl10 shRNAs were transfected into 293Ta packaging cells using Lipofectamine 2000. Lentiviral particles were harvested, concentrated, and used to transduce ZR75-AGTR1 cells for 24 hours. Selection was accomplished by culturing in the presence of puromycin and G418. Immunoblot analyses were used to verify either the maintenance or loss of Bcl10 in the resulting pools of stably transduced cells.

ON-TARGET plus SMARTpool siRNAs targeting CARMA3 (catalog no: L-004395-00-0020), Bcl10 (catalog no: L-004381-00-0020), and MALT1 (cat no: L-005936-00-0020) were obtained from GE Dharmacon. Nontargeting siRNA pools (catalog no: D-001810-10-50) were used as controls. SMARTpool siRNAs (20 nmol/L) were reverse transfected into BT549 or ZR75-AGTR1 cells using Lipofectamine RNAiMAX (Thermo Fisher Scientific). Knockdown efficiencies were assessed using immunoblots and real-time quantitative RT-PCR assays (TaqMan) for the intended targets after 48–72 hours.

Cell lysates were prepared with RIPA buffer (catalog no: 89901, Thermo Fisher Scientific) containing HALT Protease and phosphatase inhibitor cocktail (catalog no: 78440, Thermo Fisher Scientific), loaded onto Bio-Rad 4%–15% gradient SDS-PAGE TGX gels, and transferred to 0.2-μm nitrocellulose membranes (Bio-Rad). Blots were then probed with the indicated primary antibodies (listed in Supplementary Methods) and developed using Pierce ECL Plus Western Blotting Substrate (catalog no: 32134). Total RNA was isolated from cell cultures and evaluated by RT-PCR using TaqMan gene expression assays (Thermo Fisher Scientific) as described in Supplementary Methods. NFκB reporter assays were performed as described previously (15) and as detailed in the Supplementary Methods.

BT549 and ZR75 cells were plated on glass-bottom 35-mm dishes (D35-20-0-N, Cellvis), 1 × 10⁵ cells/dish. After treatment ± Ang II for 1 hour, cells were fixed with 2% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS. Cells were then blocked for 60 minutes and incubated overnight with rabbit anti-RelA antibody (1:400), followed by goat anti-rabbit (Alexa Fluor 488) secondary antibody (1:400) for 1 hour. Confocal microscopy was performed using a Zeiss LSM 710 with a 63× oil objective. Images were collected and processed using Zen software (Carl Zeiss, Inc.).

Two-dimensional cell proliferation was measured using the IncuCyte Live Cell Imaging System, according to the manufacturer's instructions (Essen BioScience). Spheroid growth was also measured in the IncuCyte and spheroid volume was quantified using SpheroidSizer, a MATLAB-based open-source application. Migration assays were performed following the IncuCyte ZOOM 96-well Scratch Wound Cell Migration Assay protocol. Finally, invasion assays were performed using a modified Boyden chamber assay protocol. All methods are described in detail in Supplementary Methods.

BT549 and ZR75-AGTR1 cells were transiently transfected with control or Bcl10 siRNA in 60-mm dishes (4 × 10⁵ cells/dish). After 48 hours, cells were serum-starved overnight and treated ± Ang II for another 24 hours before collecting the resulting conditioned medium from the dishes. The collected medium was centrifuged at 1,000 rpm for 10 minutes and filtered through a 0.2-μm membrane to remove intact cells or cellular debris. The conditioned medium was further concentrated using a Millipore Amicon ULTRA (3 kDa) filter (UFC800324, EMD Millipore), normalized, and immediately used in the chemotaxis assay.

Neutralizing antibodies were added to some samples of conditioned medium as a cocktail of α-IL6 (250 ng/mL), α-IL8 (600 ng/mL), α-IL1β (5 ng/mL), α-VEGF-A (10 ng/mL), α-INHBA (100 ng/mL), and α-SerpinE1 (400 ng/mL), and allowed to incubate for 30 minutes before proceeding to the chemotaxis assay. In parallel, isotype-matched IgG1 and IgG2b was added to control conditioned media at equivalent concentration.

Chemotaxis assays were performed using IncuCyte ClearView 96-well chemotaxis plates (catalog no: 4582, Essen BioScience), which contain transwell membranes punctuated with 96 laser-etched pores of 8 μmol/L diameter. Briefly, both sides of the ClearView membranes were coated with Matrigel (50 μg/mL for 30 minutes at 37°C, followed by 30 minutes at room temperature). Next, 2 × 10³ HUVECs, resuspended in 60 μL of 0.5% FBS VascuLife EnGS (LifeLine Cell Technology) endothelial cell medium (complete medium diluted 1:4 in EnGS basal medium), were added to the top chamber of ClearView 96-well inserts. Cells were allowed to adhere to the top surface of the ClearView membranes for 30 minutes at room temp. Concentrated conditioned medium (100 μL) from either BT549 or ZR75-AGTR1 cells was diluted 1:1 with 1% FBS-containing endothelial cell media (complete medium diluted 1:2 in EnGS basal medium) and the resulting 200 μL of diluted conditioned medium was added to each bottom chamber. Images were obtained of both the top and bottom surfaces of each ClearView membrane every 2 hours for 4–6 days. The IncuCyte Chemotaxis Analysis software module was used to quantify migration of HUVEC cells toward the conditioned medium.

All animal procedures were performed in accordance with the NIH and institutional guidelines, and were approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh. The in vivo angiogenesis plug and traditional xenograft assays, as well as the CAM assay were performed as described in Supplementary Methods.

RNA was extracted from BT549 cells 72 hours after siRNA transfection using the RNeasy Plus Kit (Qiagen) and quantified using NanoDrop, while quality was assessed using RNA Nano chips (catalog no: 5067-1511) on an Agilent 2100 Bioanalyzer. RNA Samples were then evaluated using the nCounter PanCancer Progression Panel (NanoString) according to the manufacturer's directions. Alternatively, samples were evaluated using a custom-designed probe panel consisting of 72 test genes and 6 housekeeping genes (see Supplementary Table S1). Briefly, 100 ng of total RNA was hybridized overnight at 65°C, then run on a NanoString Prep Station at maximum sensitivity. Cartridges were scanned on a NanoString Digital Analyzer at 555 fields of view. Raw count data was normalized using the nSolver analysis software version 3.0, which normalizes samples according to positive and negative control probes and the geometric mean of six housekeeping probes. Genes with normalized counts less than 20 were considered as background and were not included in the analysis. Ingenuity Pathway Analysis (IPA) analyses of NanoString data were performed using the Core Analysis function of the IPA software, v.31823283 (Qiagen Bioinformatics).

Publicly available gene expression data were obtained from cited studies via CBioPortal (www.cbioportal.org), the UCSC Xena Browser (http://xena.ucsc.edu/), and the NCBI Gene Expression Omnibus (www.ncbi.nlm.nih.gov/geo). Kernel density plots were generated using the CGDS-R and gtools packages in R (v3.0.2). Heatmaps were generated using the Xena Browser and Morpheus (https://software.broadinstitute.org/morpheus). GSEA analyses were performed using the GSEA software package (GSEA v2.2.3) and molecular signatures available from the Broad Institute. Clinical outcome analyses, including Kaplan–Meier analyses, were performed using GraphPad Prism (v7.01) and Minitab (v17.1.0) software packages.

Statistical analyses were performed with GraphPad Prism software. P values were calculated using the Student t test (two-sided) or by analysis of one-way ANOVA, followed by Bonferroni post test as appropriate. Significance was determined at P < 0.05.

We previously demonstrated that AGTR1 overexpression characterizes a subset of HER2-negative breast cancers, based on MetaCOPA analysis of 31 independent microarray datasets published between 1999 and 2006 (5). Since that time, additional large datasets have become available, most notably the TCGA dataset, which includes a complete analysis of >1,000 breast cancers (16). To determine whether the phenomenon of AGTR1 overexpression is recapitulated in the newer TCGA dataset, we evaluated RNA expression data using kernel density plot analysis. Results showed a striking bimodal distribution pattern for AGTR1 expression, with approximately 19.5% of cases falling within a group characterized by distinctly high levels of AGTR1 mRNA (Fig. 1A). Density plot analysis performed on the nearly 2,000 invasive breast cancer cases profiled by METABRIC revealed a similar bimodal pattern, with 17.8% of cases showing aberrantly high AGTR1 (Supplementary Fig. S1A). Importantly, we found that AGTR1 and HER2 overexpression are mutually exclusive, with essentially no cases in the TCGA collection showing significantly high coexpression of the two receptors (Fig. 1B). When sorted by PAM50 subtype, AGTR1⁺ cases clustered within the ER⁺ (luminal A and luminal B) subgroups (Fig. 1C). Finally, we analyzed selected TCGA cases (originating from University of Pittsburgh) using RNA in situ hybridization (RNAscope) to confirm that high levels of AGTR1 mRNA revealed by the TCGA microarray analysis were in fact due to aberrant expression of the mRNA within malignant epithelial cells, as opposed to inflammatory cells or other stromal cells of the tumor microenvironment (Supplementary Fig. S1B).

If AGTR1 overexpression functionally drives protumorigenic behavior, much like HER2, we might expect to find evidence that AGTR1⁺ breast cancers are among the more aggressive luminal cancers. We therefore queried the TCGA collection of invasive ductal carcinomas to determine if AGTR1 levels correlate with TNM stage at diagnosis, and found a significant association between high AGTR1 expression and axillary node metastases (Fig. 1D). In addition, we applied gene set enrichment analysis (GSEA) to TCGA cases stratified by AGTR1 expression, using the van 't Veer 56 geneset (17), one of the first and most powerful multigene panels for prognostication. Strikingly, AGTR1⁺ tumors showed dramatic enrichment for genes linked to poor prognosis and development of distant metastasis within 5 years (Fig. 1E). We also queried the entirely separate Pusztai dataset, which includes detailed clinical outcome data (18), and found that AGTR1 expression inversely correlates with pathologic complete response (pCR) to neoadjuvant combination chemotherapy (Fig. 1F). Indeed, retrospective analysis of all 133 patients within the Pusztai dataset demonstrates that measurement of AGTR1 alone would have served as a statistically meaningful predictor of treatment resistance. Particularly notable is the fact that no patient with an AGTR1 level in the upper quartile experienced pCR (Fig. 1G). Finally, Kaplan–Meier curves generated from both the TCGA dataset and from the Gyorffy meta-analysis (19) demonstrate that AGTR1 overexpression correlates strongly with reduced breast cancer survival (Fig. 1H and I).

Because HER2 and AGTR1 overexpression are mutually exclusive in breast cancer, we hypothesized that the most critical pathogenic signaling pathways for AGTR1 would likely be the pathways that are redundantly activated by HER2. While each receptor activates a range of intracellular signaling pathways relevant to cancer progression, both are capable of triggering the canonical NFκB pathway and this represents an important node of potential signal convergence for these two receptors. NFκB is a critically important transcription factor controlling expression of genes that drive breast cancer proliferation, survival, migration, invasion, and peritumoral angiogenesis (20, 21). Thus, we looked for evidence that AGTR1 overexpression drives an NFκB gene expression signature in breast cancer by specifically querying NFκB gene targets that are known to be expressed in epithelial cells and that have been highlighted in the breast cancer literature. Strikingly, many of the genes that meet these criteria are substantially upregulated in AGTR1⁺ invasive breast cancers from the TCGA collection (Fig. 2A). These include genes that affect pathogenic processes such as survival and proliferation (e.g., BCL2, CCND1, DUSP6, TNFRSF10C), migration, and invasion (e.g., IL1β, PCSK6, CSF1, JAG1, ZEB1, SAA1), stemness (e.g., CD44, BMPR1B, DCLK1, JAG1), and angiogenesis (e.g., IL1β, NPY1R, IL6ST/GP130) among others.

Given the association between AGTR1 overexpression and upregulation of key pathogenic NFκB target genes in clinical samples, we sought to directly test whether AGTR1 drives NFκB activity in established cell model systems. Although we had previously identified several breast cancer cell lines with high endogenous AGTR1 expression (5), we revisited the analysis using the newer expression profiling data available through the Cancer Cell Line Encyclopedia (22). Results showed that the BT549 cell line ranks near the top of all breast cancer lines when stratified for AGTR1 expression; in contrast, ZR75-1 cells have essentially undetectable levels of AGTR1 mRNA (Supplementary Fig. S1C). Established cell lines with significant AGTR1 overexpression are uniformly HER2-negative, but interestingly many also lack ER, unlike what we find in clinical samples of primary breast tumors, where AGTR1 tends to be coexpressed with ER. The reason for this difference is unclear but may relate to alterations in ER expression that can occur with prolonged culture in vitro. In contrast, the ZR75-1 cell line belongs to the ER⁺/HER2⁻ luminal subcategory (23), and therefore the same subcategory of breast cancer in which AGTR1 overexpression can be seen clinically. To simulate acquired AGTR1 overexpression in this subcategory, we created derivative ZR75-1 lines by stably transfecting the parental line with an AGTR1 expression vector (ZR75-AGTR1), or an empty vector as a negative control (ZR75-neo) (Supplementary Fig. S1D). Finally, we experimentally confirmed AGTR1, ER, PR, and HER2 status in all lines by quantitative RT-PCR (Supplementary Fig. S1E and S1F). Together, the BT549 and ZR75-AGTR1 lines provide complementary models of high-level endogenous and exogenous AGTR1 expression, respectively. For secondary analyses and to confirm observations made with these principal models, we selected the Hs578T and Hs606T cell lines, which, like BT549 cells, are among the breast cancer lines with the highest endogenous AGTR1 expression (Supplementary Fig. S1C).

In ZR75-1 cells, we found that overexpression of AGTR1 alone confers significant NFκB activation, as evidenced by moderate, constitutive RelA (p65) nuclear translocation in ZR75-AGTR1 cells relative to control ZR75-neo cells (Fig. 2B and C). Compared with ZR75-neo cells, BT549 cells, which express high levels of endogenous AGTR1, also show enhanced nuclear RelA at baseline (Supplementary Fig. S2A). In ZR75-AGTR1 cells, this increase in basal nuclear RelA is reflected by a >20-fold increase in NFκB-dependent transcriptional activity at baseline, as measured by expression of a transfected NFκB-luciferase reporter (Fig. 2D). Despite the elevated basal NFκB activity, treatment of ZR75-AGTR1 cells with Angiotensin II ligand (Ang II) results in even greater NFκB activation (∼4-fold more than ZR75-AGTR1 cells under basal conditions and >100-fold more than the level observed in ZR75-neo cells; Fig. 2B–D). BT549 cells similarly respond to Ang II with an increase in NFκB nuclear translocation and NFκB-luciferase reporter activity (Supplementary Fig. S2A and S2B). For both ZR75-AGTR1 and BT549 cells, we confirmed that this ligand-dependent NFκB activation proceeds through AGTR1 and the canonical NFκB pathway by demonstrating that it is completely blocked via pretreatment with either losartan (an AGTR1-specific antagonist), or IKK-VI (an IKKβ-specific inhibitor; Supplementary Fig. S2B).

The acute, ligand-induced burst of NFκB signaling is best demonstrated by Western blot analysis of p-IκB. Accordingly, we found that while ZR75-neo cells show no detectable increase in p-IκB following Ang II stimulation, both ZR75-AGTR1 and BT549 cells respond to Ang II with an acute increase in p-IκB that is similar in magnitude to that achieved by stimulating cells with TNFα (Fig. 2E). Hs578T and Hs606T cells, which like BT549 cells also express high endogenous AGTR1, respond to Ang II with similar time-dependent generation of p-IκB (Supplementary Fig. S3A). Importantly, the magnitude of Ang II-dependent p-IκB induction in BT549 cells is comparable with that seen constitutively in SKBR3 cells, which overexpress HER2 (Fig. 2F).

To rule out the possibility that the basal NFκB activity in cells expressing high AGTR1 levels might be due to an autocrine loop, with cells producing and secreting Ang II ligand, we performed ELISAs on media collected from both ZR75-AGTR1 and BT549 cells. Ang II was undetectable in media taken from these cells following 1–3 days in culture, while Ang II was clearly detected in media taken from the culture of CRL-7548 placental cells, which are known to produce and secrete low levels of Ang II (Fig. 2G; ref. 24). These results are in keeping with previous work indicating that many GPCRs, including AGTR1, exhibit measurable basal signaling activity that is enhanced further by ligand stimulation (25–27). Thus, overexpression of AGTR1 confers both significant, ligand-independent (basal) NFκB activity while sensitizing cells to further bursts of NFκB activity initiated by exposure to the Ang II ligand. These bursts of NFκB activity are capable of reaching levels seen in HER2⁺ breast cancer cells.

Our laboratory previously described a signaling pathway in vascular endothelial cells that mediates NFκB activation downstream of AGTR1 (28, 29). This signaling pathway is analogous to the well-known pathway that is operative in immune cells downstream of antigen receptors, and relies on PKC-dependent assembly of a signaling module comprised principally of the triad of CARMA3 (a scaffolding protein), Bcl10 (a small linker protein), and MALT1 (an effector protein), also known as the CBM signalosome (10, 11, 30, 31). In endothelial cells, AGTR1 triggers the CBM signalosome to direct NFκB-dependent proinflammatory responses that contribute to endothelial dysfunction and vascular disease (15, 29, 32, 33).

We therefore asked whether the same CBM signalosome drives NFκB activation in breast cancers harboring AGTR1 overexpression. To this end, we used siRNA to individually target each component of the CBM signalosome in both BT549 and ZR75-AGTR1 cells, and tested the impact on Ang II-dependent p-IκB induction. Results demonstrated that in both cell models, knockdown of either CARMA3, Bcl10, or MALT1 completely abrogates the p-IκB response (Fig. 3A). Hs578T and Hs606T cells show a similar complete dependence on the CBM signalosome for Ang II-dependent p-IκB induction (Supplementary Fig. S3B). Notably, the inhibitory effect of knocking down the CBM components is comparable with the effect of treating cells with either losartan or the IKKβ inhibitor, IKK-VI (Fig. 3B). These results suggest that the AGTR1–CBM–NF-κB signaling axis has impact that extends beyond its role in vascular inflammation, and into the arena of solid tumor pathogenesis.

We next tested the impact of disabling the CBM signalosome on NFκB-dependent gene expression reprograming by targeting Bcl10, the central adaptor protein within the complex. Using siRNA, we knocked down Bcl10 in BT549 cells and assessed alterations in gene expression using a custom NanoString codeset designed to simultaneously measure 72 NFκB regulated mRNAs, the products of which all have known roles in cancer progression (Supplementary Table S1). We designed the custom codeset to measure selected genes that we had previously found to be induced in AGTR1⁺ human breast cancer specimens (see Fig. 2A), as well as additional potential NFκB targets. Of the 72 genes, we found that 41 can be reliably quantified in BT549 cells by the NanoString assay and cluster within several discrete groups. First, a substantial number are strongly induced by Ang II in BT549 cells, particularly after 24 hours of stimulation, but are not induced in cells with siRNA-mediated Bcl10 knockdown (Group A; Fig. 3C). A second group is composed of genes that required Bcl10 for both basal and Ang II-induced expression (Group B). These results are consistent with our observation that AGTR1 overexpression mediates both basal and Ang II-responsive NFκB activity in breast cancer cells. Thus, genes in group B appear to be particularly sensitive targets, and are induced by basal AGTR1/CBM/NFκB signaling activity, while those in Group A require ligand-activated AGTR1/CBM/NFκB signaling to be significantly induced. Finally, smaller subsets of genes are induced by Bcl10 knockdown (Group C) or are not substantially impacted (Group D) (Fig. 3C).

We then tested the impact of Bcl10 knockdown on the broader gene expression landscape, by using the NanoString PanCancer Progression codeset, which quantifies expression of 770 genes of significance to solid tumor pathogenesis, most of which are not necessarily NF-κB target genes. Knockdown of Bcl10 in BT549 cells caused a dramatic shift in the gene expression program, with many genes being either up- or downregulated (Fig. 3D). This result suggests that disruption of the Bcl10/NFκB signaling pathway has far-reaching effects that ultimately impact multiple gene networks through complex and potentially indirect mechanisms. To better understand the overall pathophysiologic impact of Bcl10 knockdown, we performed IPA with the NanoString data and found evidence for disruption of at least three cancer hallmark processes: (i) proliferation and survival, (ii) invasion and metastasis, and (iii) tumor angiogenesis (Fig. 3E). Specifically, Bcl10 knockdown leads to upregulation of tumor suppressor pathways (PTEN and p53), as well as enhanced apoptosis signaling (Fig. 3E, left). Conversely, Bcl10 knockdown leads to decreased activity in growth factor and motility signaling networks (Fig. 3E, middle). Finally, Bcl10 knockdown suppresses key angiogenesis pathways, including those for IL6, IL8, and VEGF (Fig. 3E, right).

Given the broad impact of Bcl10 knockdown on the protumorigenic gene expression signature in AGTR1⁺ breast cancer cells, we sought to experimentally determine which protumorigenic phenotypes, highlighted by the pathway analysis, are most clearly dependent on the CBM/NFκB pathway. In addition, because we had observed that Bcl10 knockdown impacts both basal and Ang II–dependent gene expression, we wanted to determine which CBM/NFκB-driven phenotypes are evident in the basal state, and which require Ang II stimulation.

We first tested the impact of pharmacologic NFκB inhibition on the cell-intrinsic properties of proliferation, migration, and invasion in the basal state (without Ang II stimulation). In this setting, proliferation was the phenotype most significantly impacted by NFκB inhibition; results showed that treatment of either ZR75-AGTR1 or BT549 cells with IKK-VI, an IKKβ inhibitor, dramatically impairs cell proliferation as monitored in 2D cultures (Fig. 4A and B). Next, we tested the effect of CBM disruption and found that proliferation is likewise impaired upon siRNA-mediated CARMA3 or MALT1 knockdown (Fig. 4C and D). We explored this phenomenon further using 3D spheroid assays, and found that IKK-VI also exerts a dramatic inhibitory effect on spheroid growth (Fig. 4E and F). Disruption of the CBM complex, via knockdown of any of the three components, in either ZR75-AGTR1 or BT549 cells similarly inhibits spheroid growth (Fig. 4G and H; Supplementary Fig. S4A and S4B). Finally, we evaluated the additional AGTR1⁺ cell lines, Hs578T and Hs606T, and found that basal proliferation in these cells is again abrogated by knockdown of the CBM components (Supplementary Fig. S4C and S4D). As these 2D and 3D proliferation assays were performed in the absence of exogenous Ang II, the results suggest that signaling through the CBM/NFκB pathway occurs under basal conditions in AGTR1⁺ breast cancer cells, to impart a positive proliferative advantage.

Because stimulating AGTR1⁺ breast cancer cells with exogenous Ang II induces NFκB levels beyond those seen in the basal state (see Fig. 2), we asked whether additional protumorigenic phenotypes are revealed under this ligand-stimulated condition, to complement the increase in proliferation that occurs in the absence of added ligand. We found that Ang II stimulation significantly increases migration of both ZR75-AGTR1 and BT549 cells in a kinetic wound-healing assay and that this is completely abrogated by Bcl10 knockdown (Fig. 5A–D). Furthermore, Ang II stimulation promotes invasion of both cell lines through Matrigel, as assessed in a transwell invasion assay, and this is similarly blocked by Bcl10 knockdown (Fig. 5E–H). These findings are recapitulated in Hs578T and Hs606T cells; both of these AGTR1⁺ lines respond to Ang II with an increase in migration and invasion that is abrogated upon Bcl10 knockdown (Supplementary Fig. S5A–S5D). Taken together, these results indicate that AGTR1 overexpression confers both a constitutive proliferative advantage to breast cancer cells as well as a ligand-responsive migration and invasion advantage, all of which critically depend on signaling through the CBM/NFκB axis.

Because our gene expression data strongly implicated the CBM pathway in regulating an angiogenic gene signature, we asked whether AGTR1⁺ cells exert a proangiogenic effect through paracrine stimulation of neighboring endothelial cells. To this end, we established a real-time, kinetic assay for monitoring endothelial chemotaxis. In this assay, we utilized specialized IncuCyte transwells that consist of top and bottom chambers separated by an impermeable surface punctuated by regularly spaced, 8-μm channels that are large enough to allow for endothelial cell transmigration (Fig. 6A). We plated primary HUVECs in the top chamber, along with regular media, and used the real-time imaging capabilities of the IncuCyte system to measure movement of HUVECs through the channels to the bottom chamber when filled with conditioned media obtained from AGTR1⁺ breast cancer cells. Because the system can simultaneously image cells in the top and bottom chambers, applying a pseudocolor (blue and yellow, respectively) to distinguish the two populations, it is possible to quantify endothelial cell migration over time.

In control experiments, we found that media that has never been exposed to breast cancer cells has minimal impact on the migration of endothelial cells to the bottom chamber, regardless of whether or not this control media contains exogenously added Ang II (Fig. 6B and D, top row of images; quantified in Fig. 6C and E). In contrast, conditioned media obtained from the culture of either ZR75-AGTR1 or BT549 cells, transfected with nontargeting siRNA, strongly enhances endothelial chemotaxis, but only when the breast cancer cells have been stimulated by Ang II (Fig. 6B–E). Conditioned media from breast cancer cells with siRNA-mediated knockdown of Bcl10, however, is unable to induce endothelial chemotaxis above baseline (Fig. 6B–E). These data suggest that Ang II stimulation of AGTR1⁺ breast cancer cells results in secretion of one or more substances that support endothelial chemotaxis, and that this is a Bcl10-dependent process.

On the basis of our gene expression profiling results employing Bcl10 knockdown (Fig. 3C and D), there are numerous candidate molecules that could fulfill the role of an endothelial chemotactic factor (Supplementary Table S2). To begin deconvoluting this list with the goal of identifying the most important factor(s), we prepared a mix of neutralizing antibodies against six particularly promising candidates, IL1β, IL6, IL8, SERPINE1 (PAI-1), VEGFA, and INHBA. We then added the neutralizing antibody cocktail, or isotype-matched control IgG cocktail, to conditioned media from Ang II–treated BT549 cells, and then utilized the media in the chemotaxis assay. Remarkably, we found that the neutralizing antibody cocktail specifically and completely inhibits the induced endothelial chemotaxis (Fig. 6F). Further work will be required to determine which of these six factors might be most important for the Bcl10-dependent endothelial chemotaxis phenomenon, or if a specific combination of factors is critical.

Finally, we tested if the Ang II/Bcl10-dependent endothelial chemotaxis observed in vitro correlates with cancer cell–induced angiogenesis in an in vivo model. To this end, we mixed Ang II–treated breast cancers cells with concentrated, growth-factor depleted Cultrex to create a semisolid suspension that can be implanted subcutaneously in athymic nude mice. These implanted “plugs,” containing breast cancer cells, then serve as a matrix for the potential ingrowth of host vasculature over the next 12–14 days, dependent on angiogenic signals provided by the embedded cancer cells. Compared with plugs of control ZR75-neo cells, we found that ZR75-AGTR1 plugs show grossly obvious, enhanced vascularization (Fig. 7A). We confirmed this observation by quantifying plug hemoglobin content, a measurement of functional vascular content (Fig. 7B). Plugs formed with ZR75-AGTR1 cells engineered to stably express a nontargeting shRNA are similar to those composed of parental ZR75-AGTR1 cells. In contrast, plugs formed from ZR75-AGTR1 cells expressing Bcl10 shRNA are visibly pale and grossly lacking in vascularization, more similar to ZR75-neo plugs, a finding confirmed by measurement of hemoglobin content (Fig. 7C and D). Of importance, all plugs were prepared using cells treated with Ang II at the time of subcutaneous implantation, as we had found that AGTR1⁺ cancer cells must be stimulated with Ang II to drive endothelial chemotaxis in vitro.

Taken together, the in vitro and in vivo results suggest that AGTR1⁺ breast cancers are capable of supporting tumor angiogenesis through the CBM/NFκB signaling axis. This cell-extrinsic effect likely synergizes with the cell-intrinsic effects of the CBM pathway in supporting tumor cell proliferation, migration, and invasion.

Results presented above indicate that the CBM/NFκB signaling axis controls a range of specific phenotypes in the context of AGTR1⁺ breast cancer. These include cancer cell proliferation, migration, and invasion, as well as induced angiogenesis. To determine whether the CBM complex controls overall tumorigenesis as an integrated outcome of these specific phenotypes, we performed traditional xenograft growth experiments. We found that xenografts consisting of ZR75 cells with enforced expression of AGTR1 (ZR75-AGTR1) grow significantly faster than those consisting of control ZR75 cells, ultimately producing larger, heavier tumors (Fig. 7E–G). Knockdown of Bcl10 via stable shRNA expression dramatically reverses tumor growth, while control shRNA expression has no effect (Fig. 7E–G). In keeping with results from the angiogenesis plug assay, ZR75-AGTR1 xenografts show gross evidence of enhanced vascularization, as compared with ZR75-neo xenografts (Fig. 7E). Moreover, knockdown of Bcl10 appears to prevent efficient vascularization of mature xenografts, in addition to abrogating tumor growth.

To evaluate tissue invasion, we turned to the chick chorioallantoic membrane (CAM) model. Here, we layered Ang II–treated ZR75-AGTR1 cells, with or without Bcl10 knockdown, on the surface of the living chick membranes and observed the resulting interaction between cancer cells and the host tissue after a short, four-day period. Results showed that ZR75-AGTR1 cells expressing control shRNA induce a remarkably robust host tissue response characterized by acute neutrophilic inflammation, edema, fibroblast proliferation, reactive epithelial metaplasia of the surface epithelium, and overall superficial connective tissue disruption (Fig. 7H). These features result in an irregular tumor cell/host tissue interface that resembles the invasive, leading front of an aggressive infiltrating carcinoma. In contrast, ZR75-AGTR1 cells expressing Bcl10 shRNA remain completely quiescent on the surface of the CAM and show almost no discernable effect on the underlying host membrane (Fig. 7H). Taken together, these two in vivo models suggest that the CBM signalosome plays a central role in the overall growth and vascularization of AGTR1⁺ breast cancers and in instigating local, tissue-destructive, inflammatory responses in the surrounding host parenchyma that may facilitate invasion.

NFκB activation is well-described in aggressive breast cancer, particularly for the HER2⁺ intrinsic subtype (20, 21, 34). Yet the regulation of NFκB and its potential pathogenic role in luminal breast cancer is not well characterized. Our group previously identified a subset of luminal breast cancers that harbor dramatic overexpression of AGTR1, a GPCR best known for its role in the vasculature (5). With the current work, we now provide a mechanistic understanding of how AGTR1 overexpression recapitulates aggressive features of HER2⁺ cancer by harnessing a signaling complex composed of the proteins CARMA3, Bcl10, and MALT1 (CBM signalosome) to affect NFκB activation.

Importantly, we demonstrate that the CBM pathway exerts not only cell-intrinsic effects on tumor cell proliferation, migration, and invasion, but also cell-extrinsic effects by impacting nonneoplastic cells of the tumor microenvironment. Specifically, we show that conditioned media from Ang II–stimulated, AGTR1⁺ breast cancer cells induces endothelial cell chemotaxis in vitro and promotes tumor angiogenesis in vivo. Numerous CBM/NFκB responsive target genes are likely to mediate the cell-intrinsic and cell-extrinsic effects, with different subsets of genes controlling each (see schematic, Fig. 7I). On the basis of our in silico and experimental analyses, there are a large number of secreted substances that could be released from AGTR1⁺ tumor cells in a CBM/NFκB–dependent manner and act in paracrine fashion on neighboring endothelial cells (Fig. 7I). These include IL1β, IL8, IL6, INHBA, SERPINE1, and VEGF, all of which have prominent roles in regulating tumor angiogenesis, along with many other candidates (35, 36). We are now working to systematically deconvolute the secretome of AGTR1⁺ breast cancer cells, which should provide further insights into this paracrine effect. In addition, it will be important to determine if any of these secreted substances influence other cells of the tumor microenvironment, such as adipocytes, fibroblasts, or cells of the immune system including neutrophils, T cells, macrophages, and myeloid-derived suppressor cells, particularly in light of the fact that our CAM assays demonstrate that ZR75-AGTR1 cells induce a markedly robust host inflammatory response. IL1β represents an intriguing candidate effector in this regard, as it has well-recognized roles in modulating the immune contexture of breast cancer to promote immunologic tolerance and support both local and metastatic growth (37–39). IL1β may also affect resident stromal cells of the breast; a recent report demonstrated that IL1β production in breast cancer stimulates adipocyte-derived VEGF that in turn drives angiogenesis (40). Thus, CBM/NFκB–dependent IL1β production could initiate a cascade of events affecting multiple, interacting cells of the tumor microenvironment.

While our study demonstrates a pathogenic role for CBM-dependent NFκB activation downstream of AGTR1 in a specific subset of breast cancers, we speculate that the CBM signalosome may act as a central signaling node for other subsets of breast cancer characterized by overexpression or hyperactivity of related GPCRs. Several GPCRs implicated as oncogenic drivers in breast cancer, including protease-activated receptor 1 (PAR1), C-X-C chemokine receptor 4 (CXCR4), and the lysophosphatidic acid receptors (LPAR1 and LPAR3) each require the CBM signalosome to elicit NFκB activation in a variety of contexts (11, 31). The common mechanism appears to be that each receptor activates specific PKC isoforms that phosphorylate CARMA3, allowing for CBM complex assembly. Thus, it will be important to address whether ligand-dependent activation of these related GPCRs can drive the aggressive phenotypes of proliferation, invasion, and migration in a manner that similarly requires CBM signaling. Finally, Lin and colleagues recently reported the surprising finding that specific tyrosine kinase receptors can also harness the CBM signalosome (41, 42). Using both lung and breast models, these authors demonstrated that both heregulin and EGF induce NFκB via a PKC–CBM pathway. In both cases, CBM signaling was required for in vitro cell proliferation, migration, and invasion. Furthermore, the authors found that genetic MALT1 deficiency blocked in vivo tumorigenesis in two elegant models of HER2- and EGFR-driven breast and lung cancer. Taken together, our studies and those of the Lin group suggest that the CBM complex represents a central signaling node upon which multiple receptor-mediated signals may converge to drive pathogenic NFκB activation in solid tumors.

We speculate that therapeutically targeting MALT1 may be a rational and effective approach for treating a range of breast cancers, given the prominent role of the CBM signaling pathway in AGTR1⁺ breast cancer and possibly in those driven by related GPCRs, or even HER family members. Recently, several potent small-molecule inhibitors of MALT1 have been characterized that block the protease function of MALT1, necessary for its full induction of NFκB target genes (43, 44). These inhibitors include members of the phenothiazine class (mepazine, thioridazine, promazine) as well as a novel compound, MI-2. Importantly, these MALT1 inhibitors are effective at blocking growth of diffuse large B-cell lymphomas that are driven by activating mutations in CARMA1 (the lymphocyte homolog of CARMA3), thereby providing proof of concept for their effectiveness in malignancies characterized by unmitigated CBM/NFκB activity (43–45). By analogy, we suspect that these MALT1 inhibitors may also show efficacy against solid tumors such as the AGTR1⁺ subset of breast cancer. Interestingly, multiple groups have shown that mice harboring homozygous knock-in of a protease dead MALT1 allele, thereby mimicking complete and sustained pharmaceutical MALT1 inhibition, eventually develop a striking autoimmune phenotype (13, 46). This phenotype is due to the development of an imbalance in T cells, wherein loss of MALT1 protease activity preferentially reduces differentiation of Treg cells while leaving Th1 and Th2 effector cells relatively intact and functional (13, 46). This result suggests that use of a MALT1 inhibitor in the setting of a solid tumor, such as AGTR1⁺ breast cancer, might have dual efficacy in that it could (i) abrogate CBM-dependent, protumorigenic activity within tumor cells and (ii) simultaneously alter the immune contexture to favor antitumor immunity. In this context, MALT1 inhibition would theoretically leverage the dual role of MALT1 in both cancer cells and lymphocytes to synergistic benefit.

In summary, the work described herein characterizes a novel subtype of breast cancer, defined by AGTR1 overexpression in the absence of HER2 expression. These tumors, while largely confined to the luminal A and B categories, show surprisingly aggressive behavior and poor outcome. Our mechanistic studies demonstrate that the CBM signalosome drives a range of pathogenic phenotypes in the AGTR1⁺ subtype and nominate MALT1 as a potential therapeutic target. Given that the phenothiazine class of MALT1 inhibitors has been used clinically for other indications, the opportunity exists for their relatively rapid redeployment in the setting of AGTR1⁺ breast cancer. Finally, as the CBM signalosome may also function downstream of related GPCRs, or even HER2, in other subtypes of breast cancer, MALT1 inhibitors could have even broader therapeutic potential.

P.C. Lucas is a director of pathology at the NSABP Foundation. S.A. Tomlins is a consultant for and receives honoraria from Roche/Ventana Medical Systems, Almac Diagnostics, Jansenn, AbbVie, Astellas/Medivation, and Sanofi. S.A. Tomlins has sponsored research agreements with Compendia Biosciences/Life Technologies/ThermoFisher Scientific, Astellas, and GenomeDX. D.R. Rhodes and S.A. Tomlins are co-founders and equity holders in Strata Oncology. A.M. Chinnaiyan and D.R. Rhodes are co-founders of Compendia Biosciences. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Ekambaram, N.E. Hubel, P.C. Delekta, D.R. Rhodes, L.M. McAllister-Lucas, P.C. Lucas

Development of methodology: P. Ekambaram, J.-Y. (Lloyd) Lee, N.E. Hubel, S. Yerneni, S.A. Tomlins, A.V. Lee, L.M. McAllister-Lucas, P.C. Lucas

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Ekambaram, J.-Y. (Lloyd) Lee, D. Hu, S. Yerneni, P.G. Campbell, L.R. Klei, V.J. Concel, P.C. Delekta, A.V. Lee, L.M. McAllister-Lucas, P.C. Lucas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Ekambaram, J.-Y. (Lloyd) Lee, N.E. Hubel, D. Hu, S. Yerneni, N. Pollock, S.A. Tomlins, N. Priedigkeit, A.V. Lee, S. Oesterreich, L.M. McAllister-Lucas, P.C. Lucas

Writing, review, and/or revision of the manuscript: P. Ekambaram, J.-Y. (Lloyd) Lee, L.R. Klei, A.M. Chinnaiyan, S.A. Tomlins, D.R. Rhodes, N. Priedigkeit, A.V. Lee, S. Oesterreich, L.M. McAllister-Lucas, P.C. Lucas

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.-Y. (Lloyd) Lee, N.E. Hubel, P.G. Campbell, N. Pollock, L.R. Klei, V.J. Concel, A.M. Chinnaiyan, N. Priedigkeit, A.V. Lee, L.M. McAllister-Lucas, P.C. Lucas

Study supervision: L.M. McAllister-Lucas, P.C. Lucas

The authors thank Kimberly Fuhrer for her work in optimizing the RNA ISH analysis of AGTR1 expression and Anthony Green for expert assistance with histopathology. This work was supported in part by funding from the NIH (R01 HL082914, to P.C. Lucas and L.M. McAllister-Lucas) and by a grant from the Cancer Research Fund of the University of Michigan Comprehensive Cancer Center (to P.C. Lucas). J.-Y. Lee was supported by a RAC Fellowship Award from the Children's Hospital of Pittsburgh of UPMC. Dong Hu was supported by a gift from the NSABP Foundation. This project also used the University of Pittsburgh Cancer Institute (UPCI) Tissue and Research Pathology Services (TARPS) core facilities that are supported in part by award P30 CA047904.

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.