Bcl3 Selectively Promotes Metastasis of ERBB2-Driven Mammary Tumors

HER2-overexpressing tumors are a clinically aggressive subtype of breast cancer resulting in poor prognosis and increased incidence of metastases. Current therapy for HER2-positive breast cancer involves the use of receptor targeted drugs such as the humanized monoclonal antibody, trastuzumab (Herceptin), which, in combination with chemotherapy, can delay disease progression in patients with metastatic disease (1, 2). However, because of prevalent de novo and acquired resistance mechanisms, relapse is common. Therefore, there is a need to identify new or synergistic therapeutic molecular targets for HER2-positive breast cancer and to identify mechanisms to suppress disease progression in this poor prognosis patient group.

A recent study of primary human breast tumor cells identified an interleukin (IL)-6/Stat3 autocrine loop in HER2-overexpressing breast cancers that enhanced the aggressive HER2 tumor phenotype in vivo (3). Furthermore, Stat3 has been shown to promote malignancy through the transcriptional regulation of inflammatory mediators in a mouse model of HER2-positive disease (mmtv-Neu; ref. 4). Previously, we had identified a panel of genes from murine mammary epithelial cells that were responsive to Stat3 (5). One of these Stat3-responsive genes, Bcl3, is a proto-oncogene originally described in B-cell lymphomas that has previously been shown to be constitutively expressed in a small study of breast cancer tissues (6, 7). The precise role for Bcl3 in tumor pathology is currently unknown, but studies in cell lines suggest that it upregulates cell proliferation and survival (8–12). Similarly, there is no known role for Bcl3 in either the normal or neoplastic mammary gland, yet is a cofactor of the inflammatory mediator NF-κB (6), which is associated with mammary epithelial cell survival (13) and malignant progression in HER2- and EGFR1-positive mammary tumors (14–16). We have previously proposed that Bcl3 may act as a link between Stat3 and NF-κB signaling in vivo (5, 17).

Here, we have used mouse models of HER2-positive breast cancer to determine for the first time the direct oncogenic effect of Bcl3 in vivo. Stat3-deficient ErbB2 tumors exhibited a reduction in Bcl3 expression, whereas Bcl3 deficiency alone, both in transgenic mice in vivo and in xenografts, significantly reduced metastatic tumor burden yet did not affect primary tumor growth. We show that these effects are mediated by Bcl3-dependent regulation of motility/invasion-related genes in a tumor cell–autonomous manner. These results highlight a novel disease-modifying role for Bcl3 in vivo, which was not predicted from earlier in vitro studies. In the light of these results and the current limitations in targeting Stat3 and NF-κB pathways directly, we discuss the potential for Bcl3 as a new alternative therapeutic target for HER2-positive breast cancer.

All animal experiments were carried out in accordance with UK Government Home Office guidelines. The generation of mice with a mammary-specific deletion of Stat3 in normal (Blg-Cre/Stat3^fl/−) and neoplastic mammary tissues (Stat3/N₂: Stat3^fl/fl/(NIC)mmtv-N2/mmtv-Cre) have been described previously (4, 18). Bcl3/N₂ mice were generated from mmtv-N₂ (19) and Bcl3^−/− (20) lines. Animals were inspected for mammary tumors by palpation, and tumors were measured twice weekly with calipers. Animals were euthanized when tumors reached 2 cm and tissues either snap frozen in liquid nitrogen for protein or RNA analysis or fixed in formalin for histology. For pregnancy cycle and involution studies, mammary tissue from fourth abdominal glands was collected at the indicated time points and formalin fixed for histology. For whole mounts, fourth abdominal glands were incubated in Carnoy solution for 2 to 4 hours and then stained in Carmine solution overnight. For secondary tumor studies, 1:30 serial sections were assessed for the entire depth of the tissue and stained for histology.

The MG1361 murine mammary tumor cell line (21) and the EPH4 normal murine mammary epithelial cell line were obtained from Prof. C. Watson (Department of Pathology, Cambridge University, Cambridge, UK). The N202A and N202E murine mammary tumor cell lines were gifts from Prof. P.-L. Lollini (Sezione di Cancerologia) and the 4T1 murine mammary tumor cell line was a gift from Dr. R. Anderson (Department of Pathology, University of Melbourne, Melbourne, VIC, Australia). The human breast cancer cell lines, MDA-MB-231 and SKBR3, were a gift from Dr. Julia Gee (Department of Pharmacy, Cardiff University, Cardiff, UK). The HCC1954 human breast cancer cell line was a gift from Dr. Mohamed Bentires-Alj (Friedrich Miescher Institute for Biomedical Research, Basle, Switzerland) and the ZR-75-1 human breast cancer cell line was a gift from Prof. Wen Jiang (Wales College of Medicine, Cardiff University). Eph4 and human lines cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% v/v FBS (Sigma), penicillin (50 units/mL; Invitrogen), streptomycin (50 units/mL; Invitrogen), and l-glutamine (2 mmol/L; Invitrogen). The N202A and N202E cell lines were maintained in the same media as above but supplemented with 15% v/v FBS. The MG1361 cell line was maintained in Williams Media E (Invitrogen) supplemented with l-glutamine (2 mmol/L), penicillin (50 units/mL), streptomycin (50 μg/mL), 15% v/v FBS, and 5 mL nonessential amino acids (Sigma). The 4T1 cell line was maintained in RPMI (Invitrogen) supplemented with 10% v/v FBS, penicillin (50 units/mL; Invitrogen), streptomycin (50 units/mL; Invitrogen), and l-glutamine (2 mmol/L; Invitrogen). All cell lines were incubated at 37°C and 5% CO₂. Cells were reverse-transfected with ON-TARGET plus SMARTpool siRNA (Dharmacon) targeting the genes indicated and an irrelevant control RNA, using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's protocols.

RNA was isolated from frozen tissues and cells by RNeasy (Qiagen) and cDNA obtained using Revertaid (Fermentas) reverse transcriptase and random primers (Promega) according to manufacturer's instructions. Quantitative PCR was carried out using SyBr green as previously described (22). Primers, available on request, were generated against the following mouse sequences: Nme1 (NM_008704.2); Nme2 (NM_008705.4); Nme3 (NM_019730.1); Timp1 (NM_011593.2); Timp2 (NM_011594.3); Arhgdib (NM_007486.1); Bcl3 (NM_033601.3); Cyclin D1 (NM_007631.2); and the human sequence Bcl3 (NM_005178.4).

Frozen mammary tumor tissues and cell pellets were lysed using radioimmunoprecipitation assay (RIPA; ref. 22) buffer and run on SDS-PAGE gels. Proteins were detected using antibodies directed against Neu (1:200, Abcam, Cat# Ab2428), EGFR (1:1000, Cell Signalling Technologies, Cat# 2232), Vimentin (1:200, Santa Cruz, Cat# SC-7557), E-cadherin (1:500, Upstate, Cat# 07-697), Snail (1:500, Santa Cruz. Cat# Sc-10433), and Tubulin (1:10000, Abcam, Cat# Ab6160).

RNA was extracted and treated with Deoxyribonuclease I Amplification Grade (Invitrogen) to remove any contaminating gDNA. Microarray analysis was conducted using an Illumina MouseRef-8 version 2 BeadChip. Comparisons between data sets were conducted by the using the R-package Limma (23). Samples were ranked according to adjusted P values obtained from t tests corrected for multiple testing.

Mammary glands or tumors were fixed in 4% paraformaldehyde before paraffin embedding and sectioning. Ten-micrometer sections were incubated in 1:20 anti-Ki67 (Vector Labs, Mouse monoclonal, VP-K452) in 20% normal rabbit serum in PBST, 1:200 anti-cleaved caspase-3 (Asp 175, Cell Signalling, Rabbit polyclonal, 9661) in PBST, or 1:100 anti-Bcl3 (H-146, Santa Cruz, sc-13038) in 20% normal goat serum in PBS, with prior avidin/biotin block (Vector Laboratories) at 4°C overnight. Primary antibody was incubated with biotinylated secondary antibodies [rabbit anti-mouse (Dako) 1:200 in 20% normal rabbit or goat serum in PBS(T) for Ki67 and Bcl3 respectively or goat anti-rabbit (Dako) 1:200 in 5% bovine serum albumin (Sigma) in PBST for cleaved caspase-3] and detected using ABC complex (Vector Laboratories) and DAB chromogen (Vector Laboratories). Nuclei were counterstained with hematoxylin before mounting.

Migration assays were conducted using modified Boyden chambers with filter inserts for 24-well dishes containing 8-μm pores (Becton Dickinson). Forty-eight hours after transfection, 70,000 cells were plated into 400 μL of media supplemented with 0.2% serum in the upper chamber, and the lower chamber was filled with 750 μL of media supplemented with 10% serum. After 24 hours in culture, cells were fixed in 70% ice-cold ethanol overnight before staining with Harris hematoxylin and eosin (H&E). Filters were mounted on glass slides and counted by microscopic inspection. The same procedure was used for invasion assays apart from BD BioCoat Growth Factor Reduced Matrigel Invasion chambers were used and cells were incubated for 48 hours. For each experiment, cells were counted in 2 fields of view from 2 membranes for each condition.

Transfected cells were washed in PBS/2% FBS, resuspended in 100 μL of fluorescein isothiocyanate (FITC)-conjugated anti-CD24 (BD Pharmingen) or phycoerythrin (PE)-conjugated anti-Sca1 (BD Pharmingen) and left on ice for 30 minutes in the dark. Cells were washed twice in PBS/2% FBS before fluorescence-activated cell sorting (FACS) using a FACSAria Flow Cytometer (BD Biosciences) and analysis conducted with FlowJo software. Gates were set to exclude more than 99% of cells labeled with isoform-matched control antibodies conjugated with the corresponding fluorochromes.

Aldefluor reagents (Stemcell Technologies) were prepared according to manufacturer's instructions. Transfected cells were suspended at 1 × 10⁶ cells/mL with 5 μL of activated Aldefluor substrate per mL. All samples were incubated for 45 minutes (±DEAB) at 37°C. FACS was conducted as above with gates set to exclude more than 99% of control cells incubated with DEAB.

Cells were resuspended in mammosphere culture medium [serum-free epithelial growth medium (MEBM, Lonza), supplemented with B27 (Invitrogen), 20 ng/mL EGF (Sigma), 5 mg/mL insulin (Sigma), 0.0008% v/v β-mercaptoethanol (Sigma), and 1 mg/mL hydrocortisone (Sigma)], counted and seeded into ultra-low attachment plates (Corning) at a density of 4,000 cells/mL and incubated for 7 days at 37°C and 5% CO₂, after which mammospheres were counted. To determine the self-renewal capacity of cells, mammospheres were collected by gentle centrifugation at 1,100 rpm and dissociated into single cells before being reseeded into ultra-low attachment plates at a density of 4,000 cells/mL. Cells were then incubated for 7 days at 37°C and 5% CO₂ before counting.

MG1361 cells were transfected with Bcl3 or control siRNA for 48 hours and harvested using Tryspin/EDTA, washed, and disaggregated into single-cell suspensions before being resuspended at 4 × 10⁶ cells/mL in serum free L-15 media (Invitrogen), 200 μL of cell suspension was injected into the lateral tail vein of 6- to 8-week-old female virgin BALBc/SCID mice. Animals were left for 21 days before being culled by cervical dislocation. Lungs were dissected photographed and fixed in 4% formalin before being processed for H&E staining.

The Mann–Whitney U test was used to determine statistical differences between nonparametric data sets. The Student t test was used to determine statistical differences between normally distributed data sets and between data sets with sample sizes of n = 3. The χ² test was used to determine statistical differences between observed frequencies and expected frequencies of data.

We have previously reported that expression of the NF-κB cofactor Bcl3 is dependent on Stat3 in mammary epithelial cells (5). This was confirmed for both Bcl3 mRNA and protein levels in normal mammary tissues (Fig. 1A) and in ErbB2-positive mammary tumors (Fig. 1B). Bcl3 protein was predominantly expressed in the majority of nuclei of tumor cells, whereas in normal mammary epithelium, Bcl3 was preferentially localized within the cytoplasm of luminal epithelial cells and a minority of epithelial nuclei. Bcl3 expression was also observed to be constitutively elevated in ErbB2-positive mouse mammary (Fig. 1C) and human breast tumor cell lines (Supplementary Fig. S1). We wished to determine, therefore, whether Bcl3 contributed directly to the progression of ErbB2 mammary tumors in vivo and furthermore whether this correlated with the tumorigenic effects of Stat3 previously observed in this tumor subtype.

Bcl3-deficient mice (20) were crossed with the mmtv-ErbB2/neu (N₂) transgenic mouse line, a model of metastatic breast cancer (19). Bcl3^−/−N₂ mice developed focal primary mammary adenocarcinomas that were comparable with those formed in wild-type N₂ mice both in terms of their growth rate (Fig. 2A) and in histologic appearance (Supplementary Fig. S2). This was confirmed at the cellular level where molecular markers of apoptosis, mitosis, and ErbB/ErbB2 expression were unchanged in Bcl3^−/−N₂ tumors compared with wild-type N₂ controls (Fig. 2B and C; Supplementary Fig. S2). Furthermore, the cell-cycle regulator cyclin D1, a known NF-κB/Bcl3–responsive gene (11), was not significantly altered by the absence of Bcl3 expression (Fig. 2D).

Similar results were also observed during early tumorigenesis whereby the incidence, histology, and cell turnover kinetics of preneoplastic mammary epithelial hyperplasias were unaffected by Bcl3 loss (Supplementary Fig. S3).

Thus, unlike previous reports on the effects of Stat3 (4) and components of the IKK/NF-κB pathway (24, 25) in ErbB2 tumors, loss of Bcl3 did not affect mammary tumor growth in vivo.

In addition to the potentiation of tumor growth, Stat3 was previously shown to promote malignancy in ErbB2-positive tumors (4). The role of NF-κB in the metastatic progression of ErbB2 tumors in vivo has not been previously reported, yet there is a body of evidence, in breast cancer cell lines in vitro, to suggest that NF-κB signaling might contribute directly to malignancy (14–16, 24, 26). N₂ mice develop lung metastases with high frequency (19). To determine the influence of Bcl3 in ErbB2 mammary tumor progression, the lungs of Bcl3^+/+N₂ and Bcl3^−/−N₂ mice harboring late-stage primary mammary tumors of the same size (4,000 ± 30 mm³) were investigated for the presence of secondary tumors by gross examination and histology. As previously described (19), lung metastases with identical cellular morphology to the primary mammary tumor arose with high incidence in N₂ mice with large primary lesions (Fig. 3A). Bcl3 deficiency resulted in a 40% reduction in the number of animals with metastatic lesions within the lung (Fig. 3B), and in those remaining animals with metastases, a 75% reduction in tumor burden (Fig. 3C). In contrast to the primary mammary lesions, immunohistochemistry of these secondaries revealed a significant reduction in mitotic index (Ki-67) and a modest increase in apoptosis (cleaved caspase-3; cc3), resulting in a 3.6-fold decrease in cell turnover index (CTI; ref.27) in Bcl3^−/−N₂ metastases (38% Ki-67 vs. 0.93% cc3) compared with Bcl3^+/+N₂ controls (52% Ki-67 vs. 0.36% cc3; Fig. 3D–F).

Thus, deletion of Bcl3 exerted dual effects on ErbB2 tumor cells resulting in a dramatic reduction in metastatic tumor burden: a decrease in secondary tumor colonization (unrelated to primary tumor size or growth) and a reduction in metastatic tumor growth—a property not seen at the primary site.

NF-κB and Stat3 are differentially regulated during the course of adult mammary gland development and play key roles in epithelial homeostasis within the mammary gland (18, 28–30). Targeted inhibition of components of these signaling pathways is therefore detrimental to normal mammary (and other tissue) function and as such diminishes their use for therapeutic intervention in breast cancer, despite their well-characterized role in tumorigenesis.

Bcl3 exhibits a distinct expression pattern in the mouse mammary gland during the adult pregnancy cycle (Fig. 4A; refs. 5, 13) compared with its cognate NF-κB–binding partners p50 (Nfkb1) and p52 (Nfkb2). To determine whether Bcl3 affected the mammary epithelial compartment in vivo, we compared mammary ductal morphology and epithelial architecture between Bcl3^−/− mice and their littermate controls during pubertal development and in the adult pregnancy cycle (Fig. 4B; Supplementary Fig. S4). Bcl3^−/− mice exhibited no differences in epithelial architecture or tissue homeostasis in the adult mammary gland. Luminal epithelial cells exhibited the characteristic morphologic appearance of secretory cells during lactation and litters were reared successfully with no signs of reduced vitality, indicating that epithelial function was also unaffected. Epithelial regression and remodeling during postlactational involution was also undisturbed, despite the transient increase in Bcl3 at the onset of involution in wild-type Bcl3 animals (Fig. 4C and D).

Global suppression of Bcl3 therefore had no effect on normal mammary epithelial homeostasis, highlighting a tumor-specific role for Bcl3 in ErbB2-overexpressing cancers.

The substantial decrease in metastasis incidence in Bcl3^−/−N2 animals suggested that Bcl3 promoted the dissemination and/or colonization of tumor cells to distal sites, a complex process that involves interaction between tumor cells and the neighboring stromal environment. This raised the question of whether Bcl3 mediated its effects directly within the tumor cells or whether the absence of Bcl3 in the surrounding tissues of the Bcl3-null mice indirectly mediated the reduction in tumor metastasis.

To address this, Bcl3 was inhibited in an ErbB2-positive mammary tumor cell line derived from an activated Neu/ErbB2 mammary tumor (MG1361; ref. 21) and transplanted into recipient animals with normal levels of Bcl3. Introduction of Bcl3-deficient cells into the bloodstream of recipient mice, resulted in an 80% reduction in metastases to the lung compared with control xenografts, consistent with the phenotype observed in transgenic animals (Fig. 5A and B). Thus, Bcl3 mediated its prometastatic effect intrinsically within the tumor cells. The transient nature of Bcl3 inhibition in this model however meant that it was not possible to assess the effect of Bcl3 on cell growth in vivo. Consequently, we assessed the role of Bcl3 on cell turnover in vitro (Fig. 5C). No differences either in cell number or viability were observed following suppression of Bcl3 in adherent monolayer culture, suggesting that the effect on metastatic tumor growth and CTI observed in vivo may be mediated indirectly by the tumor microenvironment within the lung.

Tumor incidence and metastatic properties of HER2/ErbB2-positive tumors have previously been ascribed to the expansion of a tumor-initiating (stem/progenitor) cell subpopulation, mediated by HER2 signaling. Furthermore canonical NF-κB signaling has been shown to influence the size of the stem/progenitor population in a 7,12-dimethylbenz(a)anthracene (DMBA) mammary tumor model in mice (31). The stem/progenitor markers Sca1 and CD24 have independently been shown to enrich for tumor-initiating cells in mmtv-Neu and mmtv-NeuT tumor cell populations respectively (32, 33). We used a combination of functional stem/progenitor assays and surrogate markers in flow cytometry to assess whether the effect of Bcl3 inhibition on metastasis of MG1361 cells correlated with changes in stem/progenitor subpopulations. We found no changes in stem/progenitor numbers to explain the reduction in secondary tumor seeding observed in Bcl3-deficient tumor cells (Supplementary Fig. S5).

To identify the underlying mechanism responsible for the cell-autonomous effect of Bcl3 on secondary tumor burden, several key cellular properties associated with metastasis were assessed in Bcl3-deficient tumor cells.

Prometastatic tumor cells disseminated in the blood stream must overcome proapoptotic stresses imparted by cell detachment to seed distal sites. We therefore determined the sensitivity of MG1361 cells to cell detachment in vitro. No differences in cell viability were observed between Bcl3-deficient and wild-type tumor cells (Fig. 6A), suggesting that Bcl3 did not affect the ability of detached cells to undergo anoikis.

Metastasis is also associated with epithelial-to-mesenchymal transition (EMT) of tumor cells, a genetically programmed event that encompasses changes in cell shape, motility, and adhesion. Boyden chamber assays determined that Bcl3 promoted cell motility of ErbB2-positive cells. Bcl3 suppression resulted in a 61% reduction in the migratory capacity of MG1361 cells (Fig. 6B), with those cells that had migrated expressing normal levels of Bcl3 mRNA (Supplementary Fig. S6). Furthermore, global gene expression analysis revealed that a number of motility-associated genes were differentially expressed following siRNA-mediated suppression of Bcl3 in MG1361 cells (Table 1), including 4 members of the Nme (NM23) family of nucleoside diphosphate kinase/transcription co-activator genes, of which Nme1, in particular has been shown to be involved in the inhibition of cell motility and metastases in various model systems including breast cancer (34–36). Quantitative PCR was used to confirm the differential expression of these genes (Fig. 6C). Similarly, the Rho GDP dissociation inhibitor, Arhgdib, was confirmed to be upregulated following Bcl3 suppression (Fig. 6C). This inhibitor of GDP–GTP exchange has previously been implicated in the dynamic regulation of skeletal filaments within tumor cells that controls cell motility (37, 38) and is associated with malignant progression in breast cancer (39). To confirm the relevance of these transcriptional changes, we investigated the motility of ErbB2-positive mammary tumor cells following suppression of Nme 1, 2, and 3 and Arhgdib and also observed their effect on Bcl3-dependent motility (Fig. 6D). Thus, RNA interference (RNAi)-mediated suppression of Arhgdib, Nme1, and Nme2 increased motility of MG1361 cells as expected. No significant effect was observed with Nme3. Furthermore, knockdown of Arhgdib, Nme1, and Nme2 gene expression each partially restored the loss of motility observed with Bcl3 RNAi, suggesting that the negative regulation of both Nme and Arhgdib contributed to the observed mobilization of tumor cells by Bcl3 (Fig. 6D).

Analysis of the EMT markers E-cadherin, N-cadherin, and their upstream transcriptional regulator Snail failed to identify significant changes in gene expression following Bcl3 RNAi, suggesting that Bcl3 specifically influenced cell motility but did not regulate cell adhesion associated with EMT (Supplementary Fig. S7).

An additional property of metastatic cells is the ability to invade tissues through the local production of proteases either directly from the tumor itself or in a juxtacrine manner through cross-talk with the adjacent microenvironment. We assessed the invasive property of Bcl3-depleted MG1361 cells in Matrigel-coated Transwells and observed a significant decrease in the ability of Bcl3-deficient cells to migrate through the matrix (Fig. 6E). However, as the relative decrease in invasive cells was equivalent to the reduction in cell motility, it is likely that the motility phenotype was primarily responsible for the observed effect in the invasion assay. Despite this, microarray data verified by quantitative PCR identified a Bcl3-dependent decrease in Tissue Inhibitor of Metalloprotease 1 and 2 (Timp1, Timp2), key inhibitors of tumor-associated MMPs (Fig. 6C) (40). Taken together these data indicate that Bcl3 plays a tumor-specific role in promoting cell migration/invasion of ErbB2-positive mammary tumors.

As its original identification at the locus of a chromosomal breakpoint in chronic lymphocytic leukemia, Bcl3 has been shown to be aberrantly expressed in several solid tumor types, including breast, endometrial, colorectal, and nasopharangeal carcinomas (7, 41–43). Despite this, the role of Bcl3 in endogenous tumor pathology is poorly defined. Previous studies have focused on the effects of Bcl3 in tumor cell lines and have identified roles in proliferation (8–11) and cell survival (12) in vitro, whereas one study showed that ectopic expression of Bcl3 in an estrogen receptor (ER)-positive breast cancer cell line (MCF-7) promoted hormone-independent growth in xenografts (44). Here, we addressed the role of Bcl3 in endogenous tumor development in vivo, using a transgenic mouse model of spontaneous breast cancer progression. We describe a distinct disease-modifying role for Bcl3 in ErbB2-positive mammary tumors in which Bcl3 has no significant role in growth of the primary tumor but profoundly affects the intrinsic ability of mammary tumor cells to colonize at distal sites. This correlated with an increase in cell migration, mediated by Bcl3-dependent regulation of several motility-associated genes, each of which contributed to the motility phenotype.

Bcl3, which mediates gene transcription through interaction with the NF-κB subunits p50 and p52, appears to alter a distinct subset of malignancy-associated genes from its NF-κB dimeric partners (14, 16). Thus, global suppression of NF-κB signaling in mammary epithelial cells had previously been found to impact on the metastasis-associated genes CXCR4, VEGF1, and MMP9, distinct from the motility inhibitors Nme1, Nme2, and Arhgdib identified here to be responsive to Bcl3. There is much interest in the use of NF-κB inhibitors to suppress tumor progression in cancer, but global inhibition of NF-κB signaling has a profound effect on homeostasis of normal tissues, particularly on cells of the immune system (45). In the mammary gland, tissue-specific inhibition of NF-κB activity has deleterious effects on lobuloalveolar remodeling during pregnancy (30) and post-lactational involution (28). Therapeutic targeting of NF-κB in disease is therefore likely to have severe consequences on normal tissues. An alternative strategy of NF-κB targeted therapy has therefore been proposed, which involves a more selective inhibition of NF-κB signaling through the putative targeting of cofactors or upstream regulators of NF-κB (45–47). Thus, the selective regulation of malignancy-associated gene targets by Bcl3 and the fact that mice with constitutive Bcl3 deficiency are viable with only minor immunologic defects (20, 48), and no detrimental effects on normal mammary gland morphology or function (Fig. 4) make it a candidate for therapeutic intervention of disease progression in patients with breast cancer.

We found that the majority of Bcl3-responsive genes identified were upregulated following Bcl3 suppression, consistent with a recent study in HeLa cells (49). This supports previous observations that Bcl3 can suppress as well as activate gene transcription (50). A recent study showing enhanced stability of the transcriptional inhibitor CtBP through a protein–protein interaction with Bcl3 (51) provides a possible explanation for the repressive function of Bcl3. The fact that Bcl3 does not affect promotility genes suggests that Bcl3 de-represses cell motility in tumor cells with prometastatic oncogenic lesions. This is supported by recent unpublished data from our laboratory showing that the migratory capacity of nontumorigenic cells is not affected by Bcl3 suppression (Soukupova, unpublished). A recent study by Merkhofer and colleagues showed that downregulation of canonical NF-κB activity in an ERBB2-positive human breast cancer cell line resulted in a decrease in invasive properties in vitro but had no effect on cell proliferation (26). ERBB2-driven cell proliferation in these cells was attributed to phosphoinositide 3-kinase (PI3K) signaling independently of NF-κB, suggesting that these 2two signaling pathways have different roles downstream of ERBB2. It is possible therefore that in the current study of ERBB2-positive tumors, Bcl3 is able to modify canonical NF-κB signaling to specifically exert a prometastatic rather than proproliferative effect on this tumor subtype.

Loss of Stat3 has also previously been shown to delay the metastatic progression of ERBB2-positive tumors both via direct cell-autonomous mechanisms and through inflammatory mediators (4). Stat3, like NF-κB, is upregulated in many malignancies including breast cancer (52) but is paradoxically proapoptotic during mammary involution (18). As Bcl3 is a downstream transcriptional target of Stat3, we suggest that Bcl3 acts as a molecular link between Stat3 and NF-κB and consequently may contribute to the metastatic phenotypes mediated by these transcription pathways. Furthermore, like NF-κB, as constitutive inhibition of Stat3 is detrimental to normal tissue homeostasis and cellular immunity in the adult (53, 54), it will be interesting in the future to determine the effect of conditionally suppressing Bcl3 in breast cancer models where Stat3 and/or NF-κB activities are aberrantly activated to determine its therapeutic efficacy.

In addition to the profound reduction in metastatic tumor burden, we also observed a significant decrease in the ratio of proliferating versus apoptotic cells in the secondary lesions, an effect not seen at the primary site. We postulate that the microenvironment of the respective tumors may play a role in this, possibly through the activation of a distinct NF-κB/Bcl3 signaling pathway in the lesions within the lung, that are more focused on cell turnover. While we have yet to determine the identity of this pathway, it is clear that if mirrored in humans, this potential reduction in tumor growth at the secondary site could contribute significantly to the overall reduction in tumor burden in patients with HER2-positive breast cancer.

Current work is therefore focusing on the mechanism of secondary tumor growth in this model and whether the beneficial effects of Bcl3 suppression are recapitulated in human breast cancers.

No potential conflicts of interest were disclosed.

Conception and design: R.W.E. Clarkson

Development of methodology: A. Wakefield, J. Soukupova, R.W.E. Clarkson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Wakefield, A. Montagne, R. French, W. Muller, R.W.E. Clarkson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Wakefield, J. Soukupova, A. Montagne. J.J. Ranger, R.W.E. Clarkson

Writing, review, and/or revision of the manuscript: A. Wakefield, R.W.E. Clarkson

Study supervision: R.W.E. Clarkson

J. Soukupova is currently a Cardiff University President's PhD Scholar. The authors thank Emma Blain, Dawn Croston, Natalie Card, and Sabine Schmidt for technical assistance and Prof. Alan Clarke for critical assessment of the manuscript.

This study was funded by the Breast Cancer Research Trust UK, Breast Cancer Campaign and seedcorn funding from Cardiff University School of Biosciences. R.W.E. Clarkson was funded by a Research Council-UK Research Fellowship. A. Wakefield was supported by an MRC PhD studentship.

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.