ARTICLES SATB1 reprogrammes gene expression to promote breast tumour growth and metastasis Hye-Jung Han1, Jose Russo2, Yoshinori Kohwi1* & Terumi Kohwi-Shigematsu1* Mechanisms underlying global changes in gene expression during tumour progression are poorly understood. SATB1 is a genome organizer that tethers multiple genomic loci and recruits chromatin-remodelling enzymes to regulate chromatin structure and gene expression. Here we show that SATB1 is expressed by aggressive breast cancer cells and its expression level has high prognostic significance (P , 0.0001), independent of lymph-node status. RNA-interference-mediated knockdown of SATB1 in highly aggressive (MDA-MB-231) cancer cells altered the expression of .1,000 genes, reversing tumorigenesis by restoring breast-like acinar polarity and inhibiting tumour growth and metastasis in vivo. Conversely, ectopic SATB1 expression in non-aggressive (SKBR3) cells led to gene expression patterns consistent with aggressive-tumour phenotypes, acquiring metastatic activity in vivo. SATB1 delineates specific epigenetic modifications at target gene loci, directly upregulating metastasis-associated genes while downregulating tumour-suppressor genes. SATB1 reprogrammes chromatin organization and the transcription profiles of breast tumours to promote growth and metastasis this is a new mechanism of tumour progression. Metastasis is the final step in solid tumour progression and is the most common cause of death in cancer patients1. Metastasis is a multi-step process: invasion of tumour cells into the adjacent tissues, entry of tumour cells in the systemic circulation (intravasation), survival in circulation, extravasation to distant organs, and finally growth of cancer cells to produce secondary tumours2,3. How tumour cells become metastatic is largely unknown. It was widely believed that metastatic cells are rare and evolve during late stages of tumour progression from a series of genetic changes that enable the cells to progress through the sequential steps that finally result in growth in distant organ microenvironments. Recently, however, gene expres- sion analysis of human breast carcinomas with known clinical outcomes has revealed profiles that are associated with disease pro- gression and has identified groups of genes whose characteristic expression pattern can predict the risk of metastatic recurrence4���9. The detection in some primary tumours of such poor-prognosis gene ���signatures��� indicates that a large number of cells in the primary tumours already have such a gene expression pattern. Therefore, in addition to the traditional view of metastasis as an evolving process of rare variant clones, the poor-prognosis signatures suggest that cells in some primary tumours are predisposed to metastasis10. In fact, there might be an active molecular mechanism underlying such events. How gene expression profiles are established in these tumour cells such that they acquire metastatic properties is unknown. Here we show that the protein SATB1 is necessary for breast cancer cells to become metastatic, and when ectopically expressed in non- metastatic cells, can induce invasive activity in vivo. We also show that SATB1 expression in breast cancer cells establishes gene expres- sion profiles consistent with invasive tumours. SATB1 is a nuclear protein that functions as a ���genome organizer��� essential for proper T-cell development11. SATB1 constitutes a functional nuclear architecture that has a ���cage-like��� protein distribution surrounding heterochromatin. This architecture is referred to as ���the SATB1 regu- latory network���, as SATB1 regulates gene expression11���14 by recruiting chromatin remodelling/modifying enzymes and transcription fac- tors13,14 to genomic DNA, which it tethers via specialized DNA sequences highly potentiated for unpairing (base unpairing regions, or BURs)15���17. On T-helper 2 cell activation, SATB1 becomes expressed and folds the cytokine-gene locus into dense loops for rapid induction of multiple cytokine genes18. In breast cancer cells, we find that once SATB1 is expressed, it coordinates expression of a large number of genes to induce metastasis. Removal of SATB1 from aggressive breast cancer cells not only reverses metastatic phenotypes but also inhibits tumour growth, indicating its key role in breast cancer progression. SATB1 expression correlates with poor prognosis We examined SATB1 expression in 24 breast epithelial cell lines, including normal human mammary epithelial cells (HMECs), 5 immortalized derivatives, 13 non-metastatic cancer cell lines and 5 metastatic cancer cell lines. Both SATB1 messenger RNA and protein were detected only in metastatic cancer cell lines, correlating SATB1 expression with aggressive tumour phenotypes (results from repre- sentative cell lines shown in Fig. 1a). SATB2, a close homologue of SATB1, was expressed in both malignant and non-malignant cell lines (Supplementary Fig. 1a). Among 28 human primary breast tumours, SATB1 protein was detected in all 16 poorly differentiated infiltrating ductal carcinomas (P , 0.0001). Low-level SATB1 expression was found in some mode- rately differentiated tumour samples (7 out of 12), and no SATB1 was detected in control samples taken from adjacent non-malignant tis- sues (representative data shown in Fig. 1b and Supplementary Table *These authors contributed equally to this work. 1Life Sciences Division, Lawrence Berkeley National Laboratory, University of California, Berkeley, California 94720, USA. 2Breast Cancer Research Laboratory, Fox Chase Cancer Center, Philadelphia, Pennsylvania 19111, USA. Vol 452|13 March 2008|doi:10.1038/nature06781 187 NaturePublishing Group ��2008

highly metastatic MDA-MB-231 cell line. Expression of SATB1 pro- tein became hardly detectable by immunoblotting and its mRNA levels were substantially reduced by 70% and 90% in both SATB1- shRNA1 and SATB1-shRNA2 expressing cells, respectively (Fig. 2a). SATB1 expression remained unaltered in MDA-MB-231 cells expres- sing a control shRNA whose sequence did not match any known human gene. SATB1 knockdown decreased proliferation of SATB1 shRNA1 and SATB1 shRNA2 cells on Matrigel compared with the parental cell line and control shRNA cells (Fig. 2b). Furthermore, the invasive capacity in vitro of SATB1 shRNA cells was reduced by 80���85% (Fig. 2c). Consistently, depletion of SATB1 prevented colony formation of these cell lines in soft agar, indicating that anchorage-dependent growth was restored (Fig. 2d). Tissue organization and polarity are typically disrupted in mam- mary epithelial tumours in vivo. Therefore, we examined cell mor- phology of SATB1-depleted MDA-MB-231 cells. We observed major differences in cell morphology between control shRNA and SATB1 shRNA1 cells grown on Matrigel. Where control cells exhibited a spindle-like fibroblastic morphology, SATB1 shRNA cells had a cobble-stone-like morphology (Fig. 2e, left panel). We also observed similar cell morphology and reduction in invasive activity when SATB1 was depleted from the highly invasive BT549 breast tumour cell line (Supplementary Fig. 2a���d). When non-transformed mammary epithelial cells (MCF-10A cells) were cultured on Matrigel and analysed using markers of acinar formation���such as filamentous actin (F-actin), b-catenin, the basal extracellular matrix (ECM) receptor, and integrin a6���glandular-like structures (acini) formed, with a hollow lumen surrounded by polarized epithelial cells19���21 (Fig. 2e, right panel). Control shRNA cells, however, formed disorganized structures lacking basal polarity (Fig. 2e, right panel). In contrast, SATB1 shRNA1 cells cultured on Matrigel displayed normal acinar structures, showing uniform and polarized nuclei, cortically organized F-actin, basally distributed integrin a6, and b-catenin that localized to the lateral cell���cell junctions, as exhibited by MCF-10A cells. Similar results were confirmed with SATB1 shRNA2 cells. Therefore, SATB1 knockdown in MDA-MB-231 breast cancer cells restores polarized cellular structures found in normal mammary epithelial cells. SATB1 depletion reverses cancer metastasis We next evaluated the in vivo effects of SATB1 depletion on meta- stasis. We used an assay called experimental metastasis, in which we injected SATB1 shRNA1, SATB1 shRNA2, or control shRNA cells (1 3 106 cells) into the lateral tail vein of 6-week-old athymic mice 0 2 4 6 8 10 12 MDA-MB-231 Control shRNA SATB1 shRNA1 SATB1 shRNA2 0 3 7 1 5 9 Number of viable cells ( �� 10 5 ) Days b Relative invasiveness (%) 120 100 80 60 40 20 0 MDA-ControlSATB MB-231 shRNA 1 shRN A1 SATB1 shRNA2 c d On Matrigel 5 days 10 days e Relative level (%) 100 �� 2.5 100 30.4 �� 3.7 12.4 �� 5.2 MDA-MB-2 31 Control shRNA SATB1 shRNA1 SATB1 shRNA2 F-actin ��-Catenin Integrin ��6 Control acini MCF-10A Control shRNA SATB1 shRNAs a SATB1 shRNAs Number of colonies 474 �� 52 59 �� 13 Control shRNAs Control shRNA SATB1 ��-Tubulin Protein mRNA SATB1 kDa 100 75 50 On Matrigel Figure 2 | SATB1 depletion restores cell polarity and reduces aggressive phenotypes of MDA-MB-231 cells in vitro. a, Reduced SATB1 expression, determined by immunoblot and quantitative RT���PCR analyses, in SATB1 shRNA (1 and 2) cells, compared with controls (parental cell line MDA-MB- 231 and control shRNA cells a-tubulin levels as loading control). b, Reduced proliferation of SATB1 shRNA (1 and 2) cells grown on Matrigel compared with controls. c, Chemoinvasion assay of SATB1 shRNA (1 and 2) cells, compared with controls (error bars indicate s.e.m., n 5 3 experiments). d, Representative photographs of soft agar colony formation 20 days after culture of control shRNA and SATB1 shRNA (1 and 2) cells, with mean colony counts from three dishes. e, Left panel: Phase contrast micrographs of control shRNA (top row) or SATB1 shRNA1 cells (bottom row) cultured on Matrigel. Scale bar, 40 mm. Right panel: morphologies of SATB1 shRNA1 cells (acinar structure) and controls grown on Matrigel and stained for F-action, b-catenin or integrin a6 (green) and DAPI (blue). MCF-10A cells show the typical acinar structure. Scale bars, 15 mm. SATB1 nuclear distribution in MDA-MB-231 cells is shown in Supplementary Fig. 2e. a b Number of metastatic nodules per lung 180 160 140 120 100 80 60 40 20 0 Mouse ID 1 2 3 4 1 2 3 1 2 6 Thy 4 5 6 1 2 3 4 6 5 6 1 2 3 4 5 6 1 2 3 4 5 6 Control shRNA SATB1 shRNA1 SATB1 shRNA2 Average number of metastatic nodules 145 �� 5.2 13.33 �� 8.0 0.83 �� 0.8 Control shRNA SATB1 shRNA1 SATB1 shRNA2 H&E SATB1 GAPDH Tumour volume (mm 3 ) 1,000 800 600 400 200 0 MDA-MB-231 Control shRNA SATB1 shRNA1 pool SATB1 shRNA1 7 13 19 28 36 39 Days single clone Control shRNA SATB1 shRNA Figure 3 | SATB1 is necessary for lung colonization and tumour growth. a, Number of metastases in lungs of mice (n 5 6 per group) 9 weeks after tail- vein injection of control shRNA, SATB1 shRNA1 or SATB1 shRNA2 cells (left), with mean nodule per lung values shown below. Expression levels of human SATB1 in lungs was analysed by RT���PCR, with human GAPDH as a loading control and mouse thymocytes (Thy) as a negative control. Representative lungs and haematoxylin and eosin (H&E) staining of metastatic tumour (M) and normal (N) lung tissues are shown (right). b, Reduced tumour volumes in fat pads of nude mice injected with SATB1 shRNA1 cells (single clones or pools), compared to controls (parental cell line MDA-MB-231 or control shRNA cells). Each data point is the mean value (6s.e.m.) of five���six primary tumours. Photographs of representative mice and tumours are shown, along with haematoxylin and eosin staining of tumour (T) and normal (N) breast tissue. Scale bar, 50 mm. NATURE|Vol 452|13 March 2008 ARTICLES 189 NaturePublishing Group ��2008