XBP 1 Promotes Triple Negative Breast Cancer By Controlling the HIF 1 α Pathway

Cancer cells induce a set of adaptive response pathways to survive in the face of stressors due to inadequate vascularization1. One such adaptive pathway is the unfolded protein (UPR) or endoplasmic reticulum (ER) stress response mediated in part by the ER-localized transmembrane sensor IRE12 and its substrate XBP13. Previous studies report UPR activation in various human tumors4-6, but XBP1's role in cancer progression in mammary epithelial cells is largely unknown. Triple negative breast cancer (TNBC), a form of breast cancer in which tumor cells do not express the genes for estrogen receptor, progesterone receptor, and Her2/neu, is a highly aggressive malignancy with limited treatment options7, 8. Here, we report that XBP1 is activated in TNBC and plays a pivotal role in the tumorigenicity and progression of this human breast cancer subtype. In breast cancer cell line models, depletion of XBP1 inhibited tumor growth and tumor relapse and reduced the CD44high/CD24low population. Hypoxia-inducing factor (HIF)1α is known to be hyperactivated in TNBCs 9, 10. Genome-wide mapping of the XBP1 transcriptional regulatory network revealed that XBP1 drives TNBC tumorigenicity by assembling a transcriptional complex with HIF1α that regulates the expression of HIF1α targets via the recruitment of RNA polymerase II. Analysis of independent cohorts of patients with TNBC revealed a specific XBP1 gene expression signature that was highly correlated with HIF1α and hypoxia-driven signatures and that strongly associated with poor prognosis. Our findings reveal a key function for the XBP1 branch of the UPR in TNBC and imply that targeting this pathway may offer alternative treatment strategies for this aggressive subtype of breast cancer. We determined UPR activation status in several breast cancer cell lines (BCCL). XBP1 expression was readily detected in both luminal and basal-like BCCL, but was higher in the latter which consist primarily of TNBC cells and also in primary TNBC patient samples (Fig. 1a, b). PERK but not ATF6 was also activated (Extended Data 1a) and transmission electron microscopy revealed more abundant and dilated ER in multiple TNBC cell lines (Extended Data 1b). These data reveal a state of basal ER stress in TNBC cells. XBP1 silencing impaired soft agar colony forming ability and invasiveness (Extended Data 1c) of multiple TNBC cell lines, indicating that XBP1 regulates TNBC anchorageindependent growth and invasiveness. We next used an orthotopic xenograft mouse model with inducible expression of two XBP1 shRNAs in MDA-MB-231 cells. Tumor growth and metastasis to lung were significantly inhibited by XBP1 shRNAs (Fig. 1c-e, Extended Data 1d-g). This was not due to altered apoptosis (Caspase 3), cell proliferation (Ki67) or hyperactivation of IRE1 and other UPR branches (Fig. 1e, Extended Data 1h, i). Instead, XBP1 depletion impaired angiogenesis as evidenced by the presence of fewer intratumoral blood vessels (CD31 staining) (Fig. 1e). Subcutaneous xenograft experiments using two other TNBC cell lines confirmed our findings (Extended Data 1j, k). Importantly, XBP1 Chen et al. Page 2 Nature. Author manuscript; available in PMC 2014 October 03. N IH -P A A uhor M anscript N IH -P A A uhor M anscript N IH -P A A uhor M anscript silencing in a patient-derived TNBC xenograft model (BCM-2147) significantly decreased tumor incidence (Fig. 1f, Extended Data 1l, m). TNBC patients have the highest rate of relapse within 1-3 years despite adjuvant chemotherapy7, 8. To examine XBP1's effect on tumor relapse following chemotherapeutic treatment, we treated MDA-MB-231 xenograft bearing mice with doxorubicin and XBP1 shRNA. Strikingly, combination treatment not only blocked tumor growth but also inhibited or delayed tumor relapse (Fig. 2a). Tumor cells expressing CD44high/CD24low have been shown to mediate tumor relapse in some instances11-13. To test whether XBP1 targeted the CD44high/CD24low population, we examined the mammosphere-forming ability of cells derived from treated tumors (day 20). Mammosphere formation was increased in doxorubicin treated tumor cells, while tumors treated with doxorubicin plus XBP1 shRNA displayed substantially reduced mammosphere formation (Fig. 2b), a finding confirmed using another chemotherapeutic agent, paclitaxel (Extended Data 2a, b). Hypoxia activates the UPR, and XBP1 knockdown also dramatically reduced mammosphere formation in hypoxic conditions (Extended Data. 2b). Furthermore, CD44 expression was reduced in XBP1-depleted tumors (Extended Data 2c). To further interrogate XBP1's effect on CD44high/CD24low cell function, we used mammary epithelial cells (MCF10A) carrying an inducible Src oncogene (ER-Src), where v-Src is fused with the estrogen receptor ligand binding domain14. Tamoxifen (TAM) treatment results in neoplastic transformation and gain of a CD44high/CD24low population that has been previously associated with tumor-initiating properties15. In transformed MCF10A-ERSrc cells, XBP1 splicing was increased in CD44high/CD24low population (Fig. 2c), while XBP1 silencing reduced the CD44high/CD24low fraction (Extended Data 2d, e) and markedly suppressed mammosphere formation (Extended Data 2f), phenotypes not attributable to a direct effect of XBP1 on cell viability (Extended Data 2g, h). Furthermore, limiting dilution experiments demonstrated loss of tumor-seeding ability in XBP1-depleted cells (Fig. 2d). CD44high/CD24low cells sorted from TNBC patient samples confirmed increased XBP1 splicing and other UPR markers, and XBP1 silencing impaired mammosphere-forming ability (Fig. 2e, f, Extended Data 3a). Conversely, XBP1s overexpression in CD44low/ CD24high cells resulted in gain of mammosphere-forming ability and increased resistance to doxorubicin treatment (Extended Data 3b, c). Strikingly, patient derived CD44low/CD24high cells overexpressing XBP1s, but not control parental cells, initiated tumor formation in immunodeficient mice (Extended Data 3d, e). These data establish a critical role of XBP1 in CD44high/CD24low cells within TNBC. ChIP-seq and motif analysis of XBP1 in MDA-MB-231 cells revealed statistically significant enrichment of both the HIF1α and XBP1 motifs (Fig. 3a, Extended Data 4a), suggesting frequent colocalization of HIF1α and XBP1 to the same regulatory elements. HIF1α is hyperactivated in TNBCs, required for the maintenance of CD44high/CD24low cells9, 10, 16, 17 and regulated in response to microenvironmental oxygen levels. XBP1 ChIPseq was therefore also performed in MDA-MB-231 and Hs578T cells cultured under hypoxia and glucose deprivation conditions for 24h. Exposure to these stressors increased XBP1 splicing, resulting in a corresponding increase in signal intensity (Extended Data 4b-f) Chen et al. Page 3 Nature. Author manuscript; available in PMC 2014 October 03. N IH -P A A uhor M anscript N IH -P A A uhor M anscript N IH -P A A uhor M anscript and further enrichment of HIF1α motifs in TNBC (Fig. 3a), but interestingly not in luminal breast cancer cells (Extended Data 4g). HIF1α motif enrichment in the XBP1 ChIP-seq dataset suggested that XBP1 and HIF1α might interact within the same transcriptional complex. Co-IP experiments revealed a physical interaction of HIF1α, but not HIF2α, with XBP1 in 293T cells co-expressing HIF1 and XBP1s cultured under hypoxic conditions, also observed with endogenous proteins in two TNBC cell lines: MDA-MB-231 and Hs578T (Fig. 3b, Extended Data 4h-j). Subcellular fractionation revealed that this interaction occurs in the nucleus, and that unspliced XBP1u protein was not detectable (Extended Data 4k, l). GST pull-down experiments showed that HIF1α interacts with the XBP1s N-terminus b-zip domain (Extended Data 4m, n). We next established that XBP1 and HIF1α co-occupied several well-known HIF1α targets using ChIP-qPCR (Extended Data 5a-c). ChIP-re-ChIP assays using anti-XBP1s followed by anti-HIF1α antibodies confirmed that XBP1s and HIF1α simultaneously co-occupy these common targets (Extended Data. 5d). DNA-pull down assays with an HIF1α18 specific probe precipitated XBP1s in MDA-MB-231 nuclear extracts under hypoxia, indicating their presence in the same complex (Extended Data 5e, f). XBP1 depletion by two independent shRNA constructs dramatically reduced hypoxia response element (HRE) luciferase activity under hypoxia (Fig. 3c). Conversely, XBP1s expression dose-dependently transactivated the HRE reporter (Extended Data 5g, h), confirming that XBP1 augments HIF1α activity. When we profiled the differential transcriptome induced by XBP1 silencing in MDAMB-231 cells, gene set enrichment analysis identified significant enrichment of HIF1α mediated hypoxia response pathway genes (Fig. 3d, Extended Data 6a). XBP1 depletion downregulated HIF1α targets VEGFA, PDK1, GLUT1, and DDIT4 expression in both normoxic and hypoxic conditions (Extended Data 6b), and these results were validated in breast cancer xenografts (Fig. 3e) and Hs578T cells (Extended Data 6c). However, XBP1 depletion in luminal tumors did not affect these targets (Extended Data 6d). To further explore the consequences of this cooperation, we examined how XBP1 or HIF1α loss affected the transcription of common target genes. We found that high occupancy by XBP1 was associated with increased occurrence of the HIF1α motif across the genome in TNBC (Fig. 3f). While XBP1 depletion had no immediate effect on HIF1α expression, it substantially attenuated concurrent HIF1α and RNA polymerase II occupancy (Extended Data 6e-g, Fig. 3g). Similarly, XBP1 and RNA Polymerase II occupancy at co-bound sites was likewise reduced in the absence of HIF1α under hypoxic conditions (Fig. 3h, Extended Data 6h-l). These results indicate that the assembly of the XBP1-HIF1α complex on target promoters is crucial for their transcription, via the recruitment of RNA polymerase II. To establish whether HIF1α contributes to XBP1's function in TNBC, we performed rescue experiments using a HA