Sumoylation Pathway Is Required to Maintain the Basal Breast Cancer Subtype

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Cancer Cell Article

Maria V. Bogachek,1 Yizhen Chen,1 Mikhail V. Kulak,1 George W. Woodﬁeld,1 Anthony R. Cyr,1 Jung M. Park,2 Philip M. Spanheimer,1 Yingyue Li,1 Tiandao Li,1,3 and Ronald J. Weigel1,2,4,* 1Department of Surgery, University of Iowa, Iowa City, IA 52242, USA 2Department of Anatomy and Cell Biology, University of Iowa, Iowa City, IA 52242, USA 3The Genome Institute, Washington University in St. Louis, St. Louis, MO 63108, USA 4Department of Biochemistry, University of Iowa, Iowa City, IA 52242, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.ccr.2014.04.008

SUMMARY

The TFAP2C/AP-2g transcription factor regulates luminal breast cancer genes, and loss of TFAP2C induces epithelial-mesenchymal transition. By contrast, the highly homologous family member, TFAP2A, lacks transcriptional activity at luminal gene promoters. A detailed structure-function analysis identiﬁed that sumoylation of TFAP2A blocks its ability to induce the expression of luminal genes. Disruption of the sumoylation pathway by knockdown of sumoylation enzymes, mutation of the SUMO-target lysine of TFAP2A, or treatment with sumoylation inhibitors induced a basal-to-luminal transition, which was dependent on TFAP2A. Sumoylation inhibitors cleared the CD44+/hi/CD24À/low cell population characterizing basal cancers and inhibited tumor outgrowth of basal cancer xenografts. These ﬁndings establish a critical role for sumoylation in regulating the transcriptional mechanisms that maintain the basal cancer phenotype.

INTRODUCTION patterns of gene expression in breast cancer are predictive of clinical phenotype, little is known about the transcriptional Breast cancer has an incidence of 226,000 and accounts for mechanisms responsible for establishing the characteristic approximately 40,000 deaths annually in the United States (Sie- expression profile. Since many of the ERa-associated genes gel et al., 2012). There has been an improvement in survival for are not part of the ERa pathway, the coexpression of these women with breast cancer, although patients with locally genes suggests the existence of transcriptional mechanisms advanced or metastatic disease continue to have a poor prog- common to luminal genes. nosis. The clinical subtypes of breast cancer are defined by The triple-negative breast cancer subtype is a heterogeneous the expression of estrogen receptor-alpha (ERa) and progester- group that represents 10%–20% of breast cancers (Bertucci one receptor (PgR) and the amplification and overexpression of et al., 2012; Lehmann et al., 2011). The triple-negative subtypes c-ErbB2/HER2. The four common molecular subtypes of breast have an aggressive clinical course and do not respond to therapy cancers include the Luminal A (ERa/PgR+, HER2À), Luminal B effective for cancers that express ERa or HER2. Hence, there (ERa/PgR+, HER2+), HER2 (ERa/PgRÀ, Her2+), and triple- has been intense research focus on understanding the molecular negative (ERa/PgRÀ, HER2À)(Carey et al., 2006; Sørlie et al., characterization of this group with the goal of defining novel 2001). The luminal breast cancer subtypes (comprising approx- molecular targets (Bertucci et al., 2012). Detailed molecular imately 75% of breast cancer in postmenopausal women) are profiling has allowed further subclassification of the triple-nega- characterized by the expression of a set of ERa-associated tive breast cancer phenotypes into at least six distinct sub- genes (Sørlie et al., 2001). Although it is well established that types, including basal-like 1, basal-like 2, immunomodulatory,

Signiﬁcance

Clinical breast cancer subtypes are characterized by patterns of gene expression that predict outcome and response to therapy. Luminal breast cancers express steroid hormone receptors and tend to be hormone responsive. By comparison, basal breast cancers are not hormone responsive, display an expansion of a CD44+/hi/CD24À/low cell population, and have a worse prognosis. Herein, we show that sumoylation of the TFAP2A transcription factor is required to maintain the basal breast cancer phenotype. Disruption of SUMO conjugation of TFAP2A was associated with a loss of the CD44+/hi/CD24À/low cell population and an associated inability for basal cancer lines to form tumor xenografts. Inhibiting the sumoylation pathway may be an effective treatment strategy for basal breast cancer.

748 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

mesenchymal-like, mesenchymal stem-like, and luminal crest development, TFAP2A and TFAP2C appear to have com- androgen receptor subtypes (Lehmann et al., 2011). Other pro- plementary and overlapping roles (Hoffman et al., 2007). How- posed subclassifications of the triple-negative breast cancer ever, in breast cancer models, TFAP2C was found to have a phenotype have identified a claudin-low subgroup characterized unique role in regulation of ESR1/ERa gene expression, which by the relatively reduced expression of genes involved in cell was functionally distinct from the effects of TFAP2A (Woodfield adhesion and formation of tight junctions (Herschkowitz et al., et al., 2007). Furthermore, recent findings have highlighted a crit- 2007; Valentin et al., 2012). Basal-like breast cancers are further ical role for TFAP2C in maintaining the luminal phenotype distinguished from luminal cancers by frequent mutations of through the induction of luminal-associated genes and repres- TP53, gene expression patterns characteristic of epithelial-to- sion of basal-associated genes (Cyr et al., 2014). It remains to mesenchymal transition (EMT) and an increase in the percentage be seen if TFAP2A has a similar effect on the expression of of cancer stem cells (CSCs) (Bertucci et al., 2012; Valentin et al., luminal genes. Furthermore, if the role of TFAP2C were function- 2012). ally distinct, it would be of critical importance to understand the TFAP2C (AP-2g) is a member of the developmentally regu- molecular basis for transcriptional specificity of luminal gene lated family of AP-2 factors that include five members—TFAP2A regulation. We expect that mechanisms regulating patterns of (AP-2a), TFAP2B (AP-2b), TFAP2C (AP-2g), TFAP2D (AP-2d), gene expression in breast cancer would provide important and TFAP2E (AP-2ε)(Bosher et al., 1996; Feng and Williams, insight into strategies for drug development. With these consid- 2003; Moser et al., 1995; Williams et al., 1988; Zhao et al., erations in mind, we sought to confirm the functional differences 2001). TFAP2C binds to a GC-rich consensus sequence in the between TFAP2A and TFAP2C in regulation of luminal gene promoters of target genes through a helix-loop-helix motif in expression and to determine the molecular basis for functional the DNA binding domain (Eckert et al., 2005). Analysis of a chro- specificity of TFAP2C-mediated gene regulation. matin immunoprecipitation with direct sequencing (ChIP-seq) data set for TFAP2C defined the consensus site as the nine RESULTS base sequence SCCTSRGGS (S = G/C, r = A/G) (Woodfield et al., 2010), which closely matches the previously defined Functional Specificity of TFAP2C for the Luminal Gene optimal in vitro binding site (McPherson and Weigel, 1999). Expression Cluster AP-2 factors are expressed early in differentiation of the ecto- Previous studies in luminal breast cancer cell lines demonstrated derm and specify cell fates within the epidermis and neural crest that knockdown of TFAP2C downregulated ERa, whereas (Hoffman et al., 2007; Li and Cornell, 2007). Within the adult knockdown of TFAP2A failed to have a similar effect on ERa mammary gland, TFAP2C is expressed in the luminal and myoe- expression (Cyr et al., 2014; Woodfield et al., 2007). In order to pithelial cells (Cyr et al., 2014; Friedrichs et al., 2005, 2007). validate the unique functional role of TFAP2C in primary human Overexpression of TFAP2A or TFAP2C in mouse mammary cancer, we obtained fresh tumor tissue from patients with epithelial cells (MMECs) results in lactation failure with hypopla- ERa-positive breast cancer. Tumor-derived breast cancer cells sia of the alveolar mammary epithelium during pregnancy (Ja¨ ger were transduced with lentiviruses encoding TFAP2A, TFAP2C, et al., 2003; Zhang et al., 2003). Conditional knockout of the or nontargeting small hairpin RNA (shRNA). As seen in Figures mouse homolog of TFAP2C, Tcfap2c, in MMECs promoted 1A–1C and Figure S1A (available online), knockdown of TFAP2C, aberrant growth of the mammary tree, leading to a reduction in but not TFAP2A, repressed expression of ERa, confirming that the luminal cell population and concomitant gain of the basal TFAP2C has unique functional effects with regard to ESR1/ cell population at maturity (Cyr et al., 2014). In tumor models, ERa gene regulation and that the cell line models are reflective both TFAP2A and TFAP2C are important to cell proliferation, of gene regulation in primary human breast cancer. Using establishment of colonies in soft agar, cell migration, and xeno- MCF-7 cells, a more expansive examination of luminal gene graft outgrowth (Orso et al., 2008). targets was performed. The luminal breast cancer subtype ex- In breast cancer, AP-2 factors regulate expression of both ERa presses a set of luminal-associated genes including ESR1/ and Her2. TFAP2C regulates expression of ERa as well as other ERa, MUC1, FGFR4, KRT8, RET, MYB, FOXA1, and GATA-3 ERa-associated genes characteristic of luminal breast cancer (Kao et al., 2009). Knockdown of TFAP2C repressed expression (Cyr et al., 2014; deConinck et al., 1995; McPherson et al., of luminal genes, as noted by analysis of RNA (Figures 1D and 1997; Woodfield et al., 2007). TFAP2A and TFAP2C induce S1B) and protein (Figure 1E), whereas knockdown of TFAP2A expression of the cloned HER2/ErbB2 promoter (Begon et al., had minimal or no effect. The basal target genes, MMP14, 2005; Bosher et al., 1996; Delacroix et al., 2005; Yang et al., CALD1, and CD44, which are overexpressed in basal cancers, 2006). TFAP2C bound to the HER2 promoter, and knockdown were repressed by TFAP2C but not TFAP2A. By contrast, a of TFAP2C reduced HER2 expression (Ailan et al., 2009). In known TFAP2A target gene, CDKN1A/p21-CIP (Scibetta et al., BT474 breast carcinoma cells, TFAP2A and TFAP2C coordi- 2010; Woodfield et al., 2007), was responsive to TFAP2A only nately regulate HER2 expression (Allouche et al., 2008), and a (Figures 1D and 1E). Hence, although TFAP2A has the ability to correlation has been established between AP-2 expression induce certain genes, it lacks functional activity with regard to and the expression of HER2 in primary breast cancers (Allouche the luminal-associated gene expression cluster. et al., 2008; Pellikainen et al., 2004; Turner et al., 1998). One possibility for the functional differences of TFAP2A and Several critical questions remain to be addressed. There is TFAP2C might be due to differences in the ability for the factors 83% similarity between TFAP2A and TFAP2C, with 76% identity to bind to the regulatory regions of the luminal cluster genes. In in the carboxyl half of the proteins containing the DNA binding addition to the genes examined earlier, the FREM2 gene was and dimerization domains (McPherson et al., 1997). In neural identified as a specific TFAP2C target gene, which was highly

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 749 Cancer Cell Sumoylation Required for Basal Breast Cancer

A D TFAP2A TFAP2C E Primary Tumors 2.5 1.2 1.0 p < 0.05 2.0 p < 0.05 0.8 1.5 siRNA: NT A C 0.6 1.0 2.0 0.4 * 0.5 0.2 TFAP2A Relative Expression TFAP2A Relative Expression * * TFAP2C 0.0 0.0 1.5 siRNA: NT A C siRNA: NT A C ESR1 / ERα MUC1 1.5 1.5 TFAP2C p < 0.05 1.0 1.2 1.2 p < 0.05 0.9 0.9 ESR1 / 0.5 0.6 0.6 ERα RNA Expression RNA 0.3 * 0.3 * Relative Expression * Relative Expression * 0.0 0.0 0.0 siRNA: NT A C siRNA: NT A C MUC1 shRNA: NT A C KRT8 1.2 FGFR4 1.2 p < 0.05 1.0 1.0 p < 0.05 0.8 0.8 FGFR4 0.6 * 0.6 0.4 0.4 Luminal 0.2 0.2 Relative Expression Relative Expression KRT8 B 0.0 0.0 * 2.0 Tumor 1 siRNA: NT A C siRNA: NT A C Tumor 2 RET MYB Tumor 3 1.2 1.5 1.0 p < 0.05 RET 1.6 1.2 0.8 p < 0.05 0.9 0.6 * 0.6 1.2 0.4 * MYB 0.2 0.3 Relative Expression Relative Expression * 0.0 0.0 0.8 siRNA: NT A C siRNA: NT A C FOXA1 * 1.2 FOXA1 1.2 GATA3 p < 0.05 1.0 1.0 0.4 p < 0.05 0.8 0.8

ESR1/ERα RNA Expression ESR1/ERα RNA 0.6 0.6 GATA3 0.0 0.4 0.4 * 0.2 * 0.2 ShRNA: NT A C Relative Expression Relative Expression 0.0 0.0 siRNA: NT A C siRNA: NT A C MMP14

MMP14 CALD1 Basal 3.0 75 * * 2.5 60 CALD1 2.0 p < 0.05 45 1.5 p < 0.05 30 C 1.0 0.5 15 CD44 Relative Expression Relative Expression ESR1 / ERα 0.0 0 siRNA: NT A C siRNA: NT A C

GAPDH CDKN1A 4.5 CD44 1.5 CDKN1A * p < 0.05 ShRNA: NT A C 3.75 1.2 3.0 p < 0.05 0.9 2.25 0.6 GAPDH 1.5 0.3

0.75 Relative Expression Relative Expression * 0.0 0.0 siRNA: NT A C siRNA: NT A C

Figure 1. Functional Specificity of TFAP2C for the Luminal Cluster Genes (A) Primary ERa-positive breast cancer cells derived from patient samples were transduced with lentiviral vectors encoding shRNA specific for nontargeting (NT) TFAP2A (indicated by A) or TFAP2C (indicated by C). Knockdown of TFAP2A and TFAP2C was confirmed compared to NT; *p < 0.05, compared to NT. (B) ERa RNA was assessed by RT-PCR; data for all three tumor isolates demonstrate that knockdown of TFAP2C specifically repressed ERa expression; *p < 0.05, compared to NT. (C) Western blot for ERa protein confirmed that ERa protein expression was repressed by knockdown of TFAP2C only. (D) Functional effects on RNA expression of luminal genes, the basal genes, MMP14, CALD1, and CD44, and the TFAP2A-specific target gene CDKN1A/p21-CIP in MCF-7 cells after knockdown of either TFAP2A or TFAP2C; data demonstrate functional specificity of TFAP2C, with statistical differences shown comparing knockdown of TFAP2A versus TFAP2C for all genes (p < 0.05); *p < 0.05, compared to NT control. (E) Western blot confirmed functional specificity for TFAP2C in regulation of luminal and basal genes. Error bars indicate SEM. See also Figure S1. responsive to changes in TFAP2C expression but was unrespon- An analysis of the genomic binding of TFAP2A and TFAP2C to sive to TFAP2A (Figure S1). Using ChIP-seq the binding of the regulatory regions of the luminal-associated genes, ESR1/ TFAP2A and TFAP2C was compared in MCF-7 cells (Figure 2A). ERa, FOXA1, and FREM2, demonstrated colocalization of the

750 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

A ESR1 / ERα FOXA1 FREM2 94 48 59 TFAP2A

220 185 140 TFAP2C

ESR1 FOXA1 FREM2 100 kb 5 kb 5 kb

BCESR1 / ERα FOXA1 FREM2 30 30 # 15 * # # * # * * Vector: EVHA-C HA-A 25 25 12

TFAP2A 20 # 20 * 9 # * TFAP2C 15 15 6 HA 10 10 5 5 3 #

# Enrichment Relative ChIP Relative ChIP Enrichment Relative ChIP ACTIN Enrichment Relative ChIP 0 0 0 Off Target On Target Off Target On Target Off Target On Target

EV TFAP2A-HA TFAP2C-HA # p < 0.05 compared to EV * p < 0.05 compared to Off Target

2.5 DEF1.4 TFAP2A K10 2.0 1.2 100 200 300 400 AA 1.0 TFAP2A AD Di / DBD 1.5 0.8 TFAP2C AD Di / DBD 1.0 0.6 * * 0.4 AAD-CDBD AD Di / DBD * * * 0.5 0.2 * * * * CAD-ADBD AD Di / DBD FREM2 Expression ESR1 / ERα Expression 0.0 0.0 aad-cdbd AD Di / DBD siRNA: --- TFAP2C siRNA: - NT - TFAP2C - cad-adbd AD Di / DBD Vector: Vector: -- - pcDNA pcDNA pcDNA pcDNA TFAP2C TFAP2C TFAP2C aad-cdbd cad-adbd AD-CDBD AD-CDBD CAD-ADBD CAD-ADBD TFAP2C-mut TFAP2C-mut TFAP2C-mut A A

Figure 2. ChIP of TFAP2A and TFAP2C with Functional Specificity of TFAP2C Mapped to Amino Terminus (A) ChIP-seq demonstrates identical binding pattern comparing TFAP2A and TFAP2C to luminal target genes ESR1/ERa, FOXA1, and FREM2; red dot indicates peak analyzed in detail in (C). (B) Western blot of MCF-7 cells transfected with empty vector (EV) or HA epitope-tagged AP-2 constructs, TFAP2C (HA-C) or TFAP2A (HA-A), and probed with antibody shown. (C) Real-time ChIP was performed with anti-HA antibody, and precipitated chromatin was amplified at off-target and on-target locations for ESR1/ERa, FOXA1, and FREM2 (Woodfield et al., 2010). Data confirm specific binding of TFAP2A and TFAP2C to peaks identified by ChIP-seq with minimal binding to off-peak sites. (D) Schematic of TFAP2A (blue) and TFAP2C (yellow) showing homologous regions and chimeric AP-2 proteins generated (all chimeras generated using TFAP2C- mut, which is construct insensitive to the siRNA). AD, activation domain; Di, dimerization domain; DBD, DNA binding domain. Assignment of functional domains was as described elsewhere (Williams and Tjian, 1991a, 1991b). (E) Using endogenous ERa RNA expression as functional assay, MCF-7 cells were transfected with siRNA and expression vector as diagrammed as in (D). The data show that rescue of ERa transcriptional activity maps to the amino half of the TFAP2C protein. (F) This experiment is identical to the one described in (E), except that this one uses expression of endogenous FREM2 and maps functional effect to the first 128 amino acids of TFAP2C. *p < 0.05, compared to normalized expression in untransfected control. Error bars indicate SEM. See also Figure S2. two factors. ChIP-seq data for the other luminal target genes field et al., 2010). The data demonstrate that TFAP2A and examined demonstrated identical binding patterns for TFAP2A TFAP2C bind to the AP-2 sites in the luminal target genes with and TFAP2C in the promoter regulatory regions, and these approximately equal binding affinity. Hence, the functional differ- data agreed with other published ChIP-seq data for MCF-7 cells ences between TFAP2A and TFAP2C cannot be attributed to dif- (Figure S2). Epitope-tagged AP-2 constructs were used to ferences in genomic binding. confirm the identical chromatin binding patterns (Figures 2B and 2C). Hemagglutinin (HA)-tagged constructs for TFAP2A Functional Specificity of TFAP2C Localized to Amino and TFAP2C were transfected into MCF-7 cells, and ChIP was Terminus performed with anti-HA antibody with amplification at on-target We sought to identify the domain of TFAP2C responsible for and off-target sites for ESR1/ERa, FOXA1, and FREM2 (Wood- regulation of the luminal cluster genes. A knockdown/knockin

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 751 Cancer Cell Sumoylation Required for Basal Breast Cancer

A B TFAP2A TFAP2C Both Interactive Interactive n=21 n=67 n=61 TFAP2C-Interactive 3.0 * TFAP2A-Interactive 22.0 2.5 * Named Genes (53) Named Genes (53) 21.0 2.0 ACTA2 PIAS1 ACLY HPS1 20.0 1.5 ACY1 PLA2G4B AKT2 IFI27 19.0 1.0 ALDH16A POTEF * APOL4 ITGB4 * ALDH16A1 PPME1 AQP1 KHSRP 18.0 FREM2 Expression 0.5 RANBP9 ALDOA ATXN2L LOX 0.0 ARSA PRNPIP C18orf1 MAN2C1 T6 siRNA: ZYX ATF7IP RERE CCT7 MORG1 1.6 TAD3 STA TMF1 TRIP6 * KHSRPSAFB2 SF3B1 ZFP36L1 ATP5A1 RHOA CD248 MRPL38 1.4 SER C1QC RPL15 Named Genes (19) CD97 PAPLN 1.2 CDK5 RPSA ACTN4 FCGBP COL3A1 PCBD1

CPE SERHL CCDC80 FHL2 COL6A2 PLSCR3 FREM2 Expression 1.0 Both-Interactive CTBP1 SNTA1 CFH IFI27 CTHRC1 RMND5B 0.8 DBF4B THY COL15A1 LAMA5 EEF1A1 RPL38 * 2.5 * DMRT3 TIMM50 COL1A1 LOXL1 EEF1B2 SAFB2 0.6 * COL1A2 MT2A 2.0 DNAJC19 TMPRSS5 EEF1C SERTAD3 0.4 GLUL TOX3 DPP7 TNXA EFEMP SF3B1 1.5 GNB2 TPT1 EYA2 TNXB EFEMP1 SHARPIN 0.2 HCCA2 TUBB2A FAM14A WISP2 EFEMP2 SHISA5 0.0 1.0 HYI Ubc9/UBE2I FBLN1 FABP4 STAT6 siRNA: THY PIAS1 RERE FREM2 Expression 0.5 IFI35 UNKL FALN2 TMF1 ATF7IP CTBP1 ZMYM2 ZNF326 KLF10 WBSCR16 * FAM14A TRIP6 Ubc9/UBE2I 0.0 LDHA WDR68 FBLN2 UXT siRNA: A2 EY FHL2 LGALS1 ZMYM2 FBLN5 XRCC6 LOXL1 WISP2 MOBKL2C ZNF326 FIP1L1 ZFP36L1 MORN1 ZNF641 FLVCR1 ZYX MPST FN1 NOMO3 HLA-C PELI1 HN1L

Figure 3. Yeast Two-Hybrid Identifies Sumoylation Pathway Regulating Activity of TFAP2A on FREM2 (A) Yeast two-hybrid screen using TFAP2A or TFAP2C as bait identified potential AP-2 interacting factors. Proteins from named genes are shown that were uniquely pulled out using either TFAP2A (blue) or TFAP2C (yellow) or were pulled out with both factors (green) as bait. (B) A set of 21 factors was chosen for screening by knockdown with specific siRNAs in MCF-7 cells and assaying for effects on expression of endogenous FREM2 compared to nontargeting (NT) siRNA (normalized to 1.0); *p < 0.05, compared to NT. Two proteins identified as TFAP2A-interacting factors in yeast two-hybrid screen (PIAS1 and Ubc9/UBE2I) significantly induced FREM2 expression with knockdown. Error bars indicate SEM. system was developed in which the expression of endogenous small ubiquitin-like modifier (SUMO)-conjugating enzyme (Ihara TFAP2C was knocked down by small interfereing RNA (siRNA) et al., 2008; Johnson and Blobel, 1997), and PIAS1 is a SUMO and rescued by cotransfection with expression vectors for either E3 ligase (Leitao et al., 2011). Ubc9 was previously shown to TFAP2A or TFAP2C, engineered to be resistant to the siRNA. bind to TFAP2A and TFAP2C (Eloranta and Hurst, 2002). As Chimeric AP-2 proteins were created where regions of TFAP2A both proteins are part of the sumoylation pathway, the findings were substituted with the homologous region of TFAP2C (Fig- implicated sumoylation as a potential mechanism accounting ure 2D). With endogenous ERa expression used as a marker for a functional block of TFAP2A in regulation of the luminal for activation, rescue of ERa expression was localized to the gene cluster. Glutathione S-transferase (GST) pull-down and amino half of TFAP2C (Figure 2E). FREM2 was more robust as coimmunoprecipitation confirmed that Ubc9 bound to both a marker for TFAP2C-specific activation and allowed localization TFAP2A and TFAP2C (Figures 4A–4C). Knockdown of Ubc9 of luminal-specific activation to the first 128 amino acids of the increased endogenous FREM2 expression (Figure 4D). Whereas activation domain (Figure 2F). knockdown of TFAP2A alone had no effect, knockdown of TFAP2A abrogated the effect of Ubc9 knockdown, indicating Sumoylation Pathway Inhibits Transcriptional Activity that, in the absence of the sumoylation pathway, FREM2 expres- of TFAP2A sion became responsive to TFAP2A. It is interesting that western The luminal cluster gene promoters appear to share a common blot analysis of TFAP2A often demonstrates a doublet (Figures transcriptional mechanism, and we sought to understand the 4D and S1G; Woodfield et al., 2010). Knockdown of Ubc9 molecular basis for the functional block of TFAP2A. The block reduced the relative amount of the upper band (Figure 4D), sug- of TFAP2A activity at luminal promoters might involve either a co- gesting the potential for sumoylation of TFAP2A. One major site activator (or coactivators) specific for TFAP2C or promoter-spe- for sumoylation is lysine 10, and the TFAP2A isoform 1A, which cific corepressors of TFAP2A. A set of potential AP-2 cofactors was used to generate the chimeric AP-2 proteins, contains the were identified using yeast two-hybrid in which either TFAP2C K10 SUMO site (Berlato et al., 2011). To formally demonstrate or TFAP2A was used as bait (Figure 3A). A set of factors was cho- sumoylation of TFAP2A, SUMO-1, -2, and -3 were expressed sen based on previous findings suggesting a potential role in in vitro with TFAP2A. Wild-type TFAP2A was sumoylated by all gene regulation. The functional effect of each cofactor was as- three SUMO proteins, but the TFAP2A mutant K10R had signif- sessed by serial knockdown using specific siRNAs and assaying icantly reduced sumoylation (Figure 4E). In MCF-7 cells, wild- for expression of endogenous FREM2 (Figure 3B). Of 21 AP-2 type TFAP2A, but not K10R mutant, was sumoylated in vivo binding partners tested, knockdown of two TFAP2A-interactive with all three SUMO proteins (Figure 4F). Immunoprecipitated factors, PIAS1 and Ubc9/UBE2I, significantly induced endoge- TFAP2A was evaluated by western blot with anti-SUMO2/3 anti- nous FREM2 expression in MCF-7 cells. Ubc9 is a unique E2 body and demonstrated the endogenous sumoylated form of

752 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

A B C GST GFP DAPI Overlay Pulldown IP: IgG IP: GFP

Ubc9 +++-+ GFP-TFAP2A ++- - Empty ---- + GFP-TFAP2C - ++-

GST ---- + TFAP2A-GFP WB: Ubc9

TnT 100 μm 100 μm 100 μm GST-TFAP2A - --+ - Reactions GST-TFAP2C ---- + WB: TFAP2A

WB: Ubc9 WB: TFAP2C

TFAP2C-GFP 100 μm 100 μm 100 μm

DE4.0 F 3.5 * 3.0

2.5 Reactions SUMO Isoform 1 2 331 2 2.0 Empty + - - + - - + - - TnT TFAP2A wt +++ --- 1.5 TFAP2A wt - + - - + - - + - 1.0 TFAP2A K10R ---+++ 0.5 TFAP2A K10R - - + - - + - - +

FREM2 Expression 0.0 TFAP2A + * siRNA: NT Ubc9 TFAP2A WB: TFAP2A Ubc9 WB: TFAP2A* Ubc9

Co-Expressed:SUMO1 SUMO2 SUMO3 WB: GAPDH TFAP2A

GAPDH

GHReactions ISUMO Isoform - 1 2 3 SUMO Isoform TnT IgG α-AP2A Load - 1 2 3 TFAP2C + + + + IP: TFAP2C + + + +

EP1 P2 E * * WB: TFAP2C WB: WB: TFAP2C SUMO 2/3 WB: GAPDH

Figure 4. Sumoylation Functionally Linked to AP-2 Activity (A) Ubc9 binds to TFAP2A and TFAP2C in GST pull-down. TnT, in vitro transcription-translation. (B) MCF-7 cells were transfected with expression vectors for green fluorescent protein (GFP)-AP-2 fusion proteins demonstrating nuclear expression of GFP fusion proteins with colocalization with DAPI nuclear stain. (C) Coimmunoprecipitation of GFP-TFAP2A and GFP-TFAP2C confirms protein-protein interaction between Ubc9 and both AP-2 proteins. IP, immunoprecipitation; IgG, immunoglobulin G. WB, western blot was probed with antibody to the protein shown. (D) Expression of endogenous FREM2 RNA in MCF-7 cells transfected with siRNA (normalized to nontargeting [NT]), indicated with western blots below. Data show that FREM2 expression is not responsive to TFAP2A; however, knockdown of Ubc9 induced FREM2 expression, and induction is blocked with knockdown of TFAP2A, showing that FREM2 expression responds to TFAP2A in the absence of sumoylation pathway; *p < 0.05, compared to NT. Error bars indicate SEM. (E) Sumoylation using in vitro assay demonstrates wild-type TFAP2A is sumoylated by SUMO-1, -2, or -3, whereas TFAP2A K10R mutant has significantly reduced sumoylation. The asterisk with arrow indicates the SUMO-conjugated form of the protein. (F) MCF-7 cells cotransfected with expression vector for TFAP2A or K10R mutant construct for SUMO-1, -2, or -3. Data show that wild-type TFAP2A is sumoylated in vivo, whereas the K10R mutant is not. The asterisk with arrow indicates the SUMO-conjugated form of the protein. (G) Protein from MCF-7 cells was immunoprecipitated (IP) using IgG or anti-TFAP2A antibody; protein was eluted from beads (E), and prewashes (P1 and P2) were assayed; load is unprecipitated extract; western blot was probed with anti-SUMO2/3 antibody. (H) TFAP2C was sumoylated in vitro with SUMO-1, -2, or -3; an asterisk indicates sumoylated form of TFAP2C. (I) MCF-7 cells transfected with expression vectors for SUMO-1, -2, or -3, and western blot probed with TFAP2C shows evidence for sumoylation with SUMO-1; an asterisk indicates location of sumoylated TFAP2C. See also Figure S3.

TFAP2A in MCF-7 cells (Figure 4G). Although sumoylated forms cells. Transfection of TFAP2A-K10R induced expression of of TFAP2C were identiﬁed in vitro with SUMO-1, -2, or -3 (Fig- luminal-associated genes (RET, MUC1, and FREM2), whereas ure 4H), expression in MCF-7 cells in vivo was only able to transfection of wild-type TFAP2A had no effect (Figure 5A). By identify sumoylation of TFAP2C with coexpression of SUMO-1 contrast, transfection of wild-type TFAP2A or K10R mutant (Figure 4I). The sumoylated forms of AP-2 were estimated to induced expression of CDKN1A/p21-CIP. The induction of be 70 ± 5 kDa (Figures S3A–S3D). Sumoylation of TFAP2A was FREM2 protein with expression of TFAP2A-K10R was conﬁrmed similarly demonstrated in the basal line BT549, which was by western blot, without changes in expression of TFAP2C (Fig- increased with peroxide (Figures S3E and S3F). ure 5B). Although the effects on luminal gene expression were reproducible, the overall effects were modest, since parental Inhibiting SUMO Conjugation of TFAP2A Induced a Basal MCF-7 cells express TFAP2C as well as the luminal gene targets. to Luminal Transition We have shown that stable knockdown of TFAP2C converts To demonstrate the functional effects of sumoylation, wild-type MCF-7 cells from a luminal to basal-like phenotype (Cyr et al., and the K10R mutant of TFAP2A were transfected into MCF-7 2014). Hence, a MCF-7 cell clone with stable knockdown of

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 753 Cancer Cell Sumoylation Required for Basal Breast Cancer

ABC 12 2.0 *p < 0.005, compared to Empty p=0.005 10 4.0 3.5 1.5 * * 8 * 3.5 3.0 3.0 2.5 1.0 6 2.5 2.0 2.0 4 1.5 1.5 0.5

1.0 ESR1/ERα 1.0 2

RET Expression 0.5 FREM2 Expression 0.5 FREM2 Relative Intensity 0.0 0.0 0.0 0 Vector: Vector: TFAP2A 300 Empty Empty K10R K10R TFAP2A TFAP2A TFAP2A TFAP2A 250 3.5 * 2.5 TFAP2C 3.0 * 2.0 200 * 2.5 * FREM2 2.0 1.5 150

1.5 KRT8 1.0 GAPDH 100 1.0 0.5 MUC1 Expression 0.5 Vector: Empty TFAP2A TFAP2A 50 CDKN1A Expression CDKN1A K10R 0.0 0.0 0 Vector: Vector: Empty Empty K10R K10R 5 TFAP2A TFAP2A TFAP2A TFAP2A * FREM2 10 * ESR1 / ERα D 1.0 E CD44 4 8 * 0.8 -5 3 p < 10 6 * * * * 0.6 p = 0.001 4 2 FREM2

0.4 2 RNA Expression RNA 1 Expression 0.2

ESR1 / ERα RNA ESR1 / ERα RNA 0 0 0.0 FREM2 TFAP2A ESR1 / ERα ESR1 / ERα CD44 CD44 GAPDH

PIAS1 TFAP2C Ubc9

GAPDH K10R Empty GAPDH GA: TFAP2A siRNA: NTUbc9 PIAS1 Cells: sKD-NT sKD-C sKD-C Cell Line: sKD-NT sKD-C

Figure 5. Functional Effects of TFAP2A-K10R Mutant and Sumoylation Inhibition (A) Expression of luminal genes in MCF-7 cells transfected with wild-type TFAP2A or K10R mutant. K10R mutant induced luminal genes but wild-type TFAP2A did not. Both wild-type and K10R mutant TFAP2A induced CDKN1A/p21. (B) Protein expression from western blots performed in triplicate, with an example of western blot below showing FREM2 protein expression induced by K10R but not wild-type TFAP2A. (C) Expression of luminal cluster genes ESR1/ERa, KRT8, and FREM2 in sKD-C cells transfected with empty vector (EV) or expression vector for TFAP2A or K10R- TFAP2A mutant, with expression normalized to EV. K10R induced expression of luminal target genes, whereas wild-type TFAP2A did not. (D) Knockdown of Ubc9 or PIAS1 reactivated ERa and repressed CD44 expression in sKD-C cells. (E) Treatment of sKD-C cells with GA reactivated FREM2 and ERa and repressed CD44 mRNA normalized to lowest value (top) and protein (bottom). For all panels, an asterisk indicates p < 0.05 compared to normalized signal of 1.0. For CD44 expression in (E), GA treated and untreated were also signiﬁcantly different from each other, p < 0.05. Error bars indicate SEM.

TFAP2C (sKD-C) was utilized compared to a cell clone with a activation of ERa mRNA and protein expression and repressed nontargeting shRNA (sKD-NT). Overexpression of TFAP2A- expression of the basal-associated gene CD44 (Figure 5D). To K10R in sKD-C basal-like cells signiﬁcantly induced the luminal conﬁrm the effect of the sumoylation pathway, the small mole- genes ESR1/ERa, KRT8, and FREM2, whereas wild-type cule inhibitor of sumoylation, ginkgolic acid (GA) (Fukuda et al., TFAP2A had no effect (Figure 5C). Inhibiting sumoylation by 2009), was examined for its effect on ERa, FREM2, and CD44 knockdown of either PIAS1 or Ubc9 in sKD-C cells resulted in re- expression. Treatment of sKD-C cells with GA induced ERa

754 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

and FREM2 expression and repressed CD44 expression com- formed xenografts with a median time of 8 weeks. To prove that pared to vehicle treatment without altering TFAP2C expression the drugs were not cytotoxic, BT20 xenografts were inoculated (Figure 5E). in nude mice, and the mice were gavaged with AA or vehicle. As noted in Figure 7E, mice gavaged with AA failed to form Sumoylation Inhibitors Clear the CD44+/hi/CD24–/low Cell tumors, whereas vehicle-gavaged mice formed xenografts as Population expected. We created a breast carcinoma cell line (IOWA-1T) The findings suggest that the sumoylation pathway plays a crit- from the basal tumor-derived cells used in Figure 6A. The ical role in maintaining the basal breast cancer phenotype. One IOWA-1T cell line rapidly forms locally advanced tumors in of the characteristics of basal breast cancers is the relatively nude mice (R.J.W. and M.V.B., unpublished data). Identical ex- high percentage of the CD44+/hi/CD24À/low cell population. To periments using the IOWA-1T basal cell line confirmed that either confirm the generality of the findings, basal breast cancer cell pretreating the cancer cells or gavaging mice with AA repressed lines were treated with GA or another inhibitor of sumoylation, tumor initiation of xenografts (Figure 7F). anacardic acid (AA) (Fukuda et al., 2009). As previously reported, we confirmed that AA inhibited SUMO conjugation of proteins DISCUSSION globally and the SUMO-conjugated form of TFAP2A specifically (Figure S4). AA was also noted to decrease slightly the overall Transcriptional Regulation of Luminal Gene Expression expression of TFAP2A. As seen in Figure 1, knockdown of in Breast Cancer TFAP2C induced a 1.5-fold increase in TFAP2A, indicating that In the presence of an intact sumoylation pathway, TFAP2C tran- TFAP2C moderately represses TFAP2A. The finding that AA scriptionally regulates the expression of luminal genes such as induced a slight decrease in TFAP2A is consistent with the ERa/ESR1, whereas TFAP2A is functionally inactive at luminal SUMO-unconjugated form of TFAP2A acquiring TFAP2C-like gene promoters. A model depicting the role of AP-2 factors in repression activity. Treatment of the basal breast cancer cell establishing gene expression patterning in breast cancer is pre- lines BT-20 and BT-549, as well as sKD-C cells with sumoylation sented in Figure 8. Furthermore, the data herein support the inhibitors, abrogated expression of CD44 and significantly conclusion that this model is reflective of gene regulation in pri- reduced the CD44+/hi/CD24À/low population (Figure 6A). In addi- mary ERa-positive cancer. Recent evidence indicates that tion, cells from a primary basal breast cancer were treated TFAP2C likely functions in concert with other transcription fac- in vitro in parallel. The cancer was from a patient with locally tors, including ERa and FOXA1, to regulate many of the genes advanced breast cancer that was refractory to conventional in the luminal expression cluster (Cyr et al., 2014; Jozwik and chemotherapy. Remarkably, treatment with GA or AA cleared Carroll, 2012; Tan et al., 2011). Other studies have attempted the CD44+/hi/CD24À/low population from the cells harvested to define mechanisms common to the regulation of luminal from the primary tumor (Figure 6A, last panel). By contrast, GA/ breast cancer genes (Joshi et al., 2012). GATA-3 regulates the AA treatment of MCF-10A cells, a normal breast cell line model, differentiated luminal breast cancer phenotype (Fang et al., had no effect on CD44 expression. 2009), and recent findings indicate an important role of To further demonstrate the role of AP-2 in repression of CD44, FOXM1 in mediating mammary luminal cell differentiation the effects of GA and AA were examined with knockdown of through GATA-3 (Carr et al., 2012). Previous studies have also TFAP2A. As seen in Figures 6B and 6C, knockdown of TFAP2A implicated FOXA1 (Bernardo et al., 2013), Elf5 (Chakrabarti alone had no effect on CD44 expression in sKD-C, BT549, and et al., 2012), BRCA1 (Bai et al., 2013), and ErbB3 (Balko BT20 cells. However, the ability for GA and AA to repress et al., 2012) in maintaining the luminal mammary phenotype. It CD44 expression in the basal breast cancer cell lines was is interesting that prostate-derived ETS factor (PDEF) mediates completely eliminated by knockdown of TFAP2A. Since drug ef- luminal differentiation and also correlates with expression in fects might not be specific for a single pathway, GA and AA might luminal breast cancer, suggesting a strong link between devel- induce changes in CD44 expression through several mecha- opment of luminal mammary cells and oncogenesis of luminal nisms. To prove that the effects on CD44 expression were medi- breast cancer (Buchwalter et al., 2013). Similarly, TFAP2C par- ated through the sumoylation pathway, Ubc9 and PIAS1 were ticipates in luminal mammary development and luminal gene knocked down by siRNA in the basal cell lines BT549 and expression in breast cancer (Cyr et al., 2014), further strength- BT20, and the effects on CD44 expression were assessed. As ening the link between the processes of luminal differentiation seen in Figures 7A and 7B, knockdown of either Ubc9 or and oncogenesis. PIAS1 similarly repressed CD44 expression. Furthermore, knockdown of Ubc9 or PIAS1 eliminated the SUMO-conjugated Sumoylation Blocked the Activity of TFAP2A at Luminal form of TFAP2A (and slightly reduced the overall level of TFAP2A) Gene Promoters in both basal (BT549) and luminal (MCF-7) cells (Figures 7C Several lines of evidence indicate that sumoylation plays a key and 7D). role in establishing the functional differences between TFAP2C The CD44+/hi/CD24À/low cell population is associated with a and TFAP2A and accounts for the functional block of TFAP2A subset of breast carcinoma cells that are critical for the initiation at luminal gene promoters (see Figure 8). First, we have demon- of tumor xenografts. To address the effects of GA and AA on strated sumoylation of TFAP2A at lysine 10 in vitro and in vivo. tumor initiation, BT20 cells were pretreated with GA or AA Second, blocking the sumoylation pathway, either by knock- compared to vehicle prior to inoculation of xenografts in nude down of critical enzymes in the sumoylation pathway or with mice. As noted in Figure 7E, pretreatment of the cells repressed the use of small molecule inhibitors of sumoylation, allowed the formation of tumor xenografts, whereas vehicle-treated cells TFAP2A to induce expression of luminal genes such as

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 755 Cancer Cell Sumoylation Required for Basal Breast Cancer

A MCF10A sKD-C BT549 BT20 Basal Cancer

VC GA AA VC GA AA VC GA AA VC GA AA VC GA AA CD44 GAPDH

5 5 5 5 5 10 10 10 10 10

4 104 104 104 10 104 Vehicle 3 103 103 103 10 103 Control

2 102 102 102 10 102 10 10 10 10 7.2% 99.9% 30.2% 10 10.5% 96.6% 2345 2345 2345 2345 2345 10 1010 10 10 10 1010 10 10 10 1010 10 10 10 1010 10 10 10 1010 10 10

5 5 5 5 5 10 10 10 10 10

4 104 104 104 10 104 Ginkgolic 3 3 3 103 3 10 10 10 10 Acid CD24 2 102 102 102 10 102 10 10 10 10 7.4% 1.9% 0.1% 10 0.2% 0.1% 2345 2345 2345 2345 2345 10 1010 10 10 10 1010 10 10 10 1010 10 10 10 1010 10 10 10 1010 10 10

5 5 5 5 5 10 10 10 10 10

104 104 104 104 104 Anacardic 3 3 3 103 3 10 10 10 10 Acid 2 102 102 102 10 102 10 10 10 10 11.4% 1.6% 5.6% 10 0.2% 0.2% 2345 2345 2345 2345 2345 10 1010 10 10 10 1010 10 10 10 1010 10 10 10 1010 10 10 10 1010 10 10 CD44

B sKD-C C BT549 BT20

CD44 CD44 CD44 CD44 TFAP2A TFAP2A TFAP2A TFAP2A

GAPDH GAPDH GAPDH GAPDH siNT ++siNT ++ siNT +++ siNT +++ siTFAP2A ++ siTFAP2A ++ siTFAP2A +++ siTFAP2A +++ GA ++ AA ++ GA ++ GA ++ AA ++ AA ++

Figure 6. Sumoylation Inhibitors Cleared CD44+/hi/CD24–/low Cell Population (A) Treatment of s-KD-C, BT-549, BT-20, or cells derived from a primary basal cancer (Basal Cancer) with GA or AA inhibited CD44 expression by western blot (top row) and signiﬁcantly reduced the CD44+/hi/CD24-/low population by FACS analysis (lower panels) but had no effect on the normal breast cell line MCF-10A. VC, vehicle control. (B and C) Western blots showing that CD44 repression by GA and AA treatment was dependent on expression of TFAP2A since knockdown of TFAP2A with siRNA abrogated effect of sumoylation inhibitors in sKD-C cells (B) and basal cell lines BT549 and BT20 (C). See also Figure S4.

ESR1/ERa and FREM2 and repress expression of the basal gene of CDKN1A/p21-CIP. Although there is evidence that TFAP2C CD44. Finally, mutation of the SUMO target lysine of TFAP2A can be sumoylated, we were not able to demonstrate that su- conferred the ability to induce expression of luminal cluster moylation has functional effects on the transcriptional activity genes. Taken together, the data indicate that sumoylation of of TFAP2C. Under conditions where sumoylation of TFAP2A endogenous TFAP2A blocks this factor from regulating luminal blocked its activity at luminal gene promoters, TFAP2C remained gene expression. Furthermore, it is clear that the functional effect active despite evidence for similar levels of sumoylation. Further of sumoylation is a luminal gene-speciﬁc effect since wild-type studies concerning details of sumoylation of TFAP2C may or K10R mutant TFAP2A was active in transcriptional activation uncover subtle effects at speciﬁc promoters.

756 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

ABBT549 BT20 The Role of Sumoylation in Cancer-Related Gene Regulation 1.2 1.0 Sumoylation involves the posttranslational modification of pro- 1.0 0.8 teins through the covalent attachment of SUMO proteins to 0.8 0.6 lysine residues in target proteins (Bettermann et al., 2012; Cube- 0.6 CD44 CD44 n˜ as-Potts and Matunis, 2013). At least four SUMO proteins have 0.4 0.4 * * * Ubc9 * Ubc9 0.2 * been described, SUMO-1 through SUMO-4. The enzymatic RNA Expression RNA 0.2 RNA Expression RNA PIAS1 PIAS1 * pathway involves several steps beginning with ATP-dependent 0.0 * 0.0 * CD44 CD44 activation of the SUMO protein by the E1 heterodimer ASO1- * p<0.05 * p<0.05 UBA2, continuing with the transfer of SUMO to the cysteine res- Ubc9 vs NT Ubc9 vs NT idue of the E2 enzyme Ubc9, and finally the enzymatic transfer of PIAS1 PIAS1 the SUMO tag to the target protein by the E3 ligase, e.g., PIAS1. GAPDH GAPDH Sumoylation of key regulatory proteins influences several siRNA: NT Ubc9 PIAS1 siRNA: NT Ubc9 PIAS1 aspects of oncogenesis and cancer progression (Bettermann et al., 2012). There is a growing body of literature reporting the C D BT549 MCF7 sumoylation of transcription factors and the effects on transcrip-

60 * 60 tional regulation (Gill, 2003), including sumoylation of androgen 50 * TFAP2A 50 TFAP2A receptor, glucocorticoid receptor, C/EBP, Smad4, Myb, Ets-1, 40 40 Pdx1, Sp3, p300, CREB, and p53 (Bettermann et al., 2012). In 20 Ubc9 20 Ubc9 most cases, sumoylation represses transcriptional activity. 80 PIAS1 80 PIAS1 Mechanisms resulting in transcriptional repression by sumoyla- 40 GAPDH 40 GAPDH tion may include effects of protein stability, altered cellular local-

NT ization or DNA binding, modulation of corepressor binding, and MW NT Ubc9 PIAS1 MW Ubc9 PIAS1 altered association with chromatin-modifying enzymes such as siRNA siRNA histone deacetylases (Gill, 2003; Girdwood et al., 2003). Holm- strom et al. (Holmstrom et al., 2008) showed that transcriptional E BT20 inhibition by sumoylation occurred at compound, but not single, Pre-Treat Gavage sites and was related to the ability for sumoylation to destabilize 1.0 1.0 AA/GA AA the transcription factor-chromatin interaction. A mechanism Pretreat Gavage 0.8 0.8 whereby sumoylation destabilizes TFAP2A binding to certain regulatory regions may provide a mechanism for our finding of 0.6 0.6 promoter-specific repression. Since the function of TFAP2A at 0.4 0.4 promoters for genes such as CDKN1A/p21-CIP is SUMO insen- Vehicle Gavage VehicleTx 0.2 0.2 sitive, it is possible that promoter regulatory structure common Tumor-Free Survival Tumor-Free Tumor-Free Survival Tumor-Free to luminal genes may account for SUMO-specific effects. p=0.0128 p=0.00253 0.0 0.0 0 5 10 15 051015 Many luminal genes contain closely linked promoter elements Weeks post-inoculation Weeks post-inoculation for AP-2, ERa, and FOXA1 (Tan et al., 2011), and the interaction of these factors may be sensitive to sumoylation (Figure 8). F IOWA-1T Consistent with previous studies, sumoylation of TFAP2A was Pre-Treat Gavage increased by peroxide and confirms that oxidative stress can 1.0 1.0 AA AA Gavage increase the sumoylation of factors (Bossis and Melchior, Pretreat 0.8 0.8 2006; Ryu et al., 2010). Recent studies have suggested that the sentrin-specific protease 1 (SENP1) may be involved in 0.6 0.6 SUMO deconjugation of factors in breast cancer (Abdel-Hafiz

0.4 0.4 and Horwitz, 2012; Chen et al., 2013). It is intriguing to consider

Overall Survival Vehicle Tx 0.2 0.2 Vehicle Gavage Tumor-Free Survival Tumor-Free p=0.002 p=0.025 0.0 0.0 05101520 25 30 024681012 (E) At left, tumor-free survival of nude mice (n = 5 per group) inoculated with Days post-inoculation Days post-inoculation BT20 cells pretreated for 48 hr with either GA or AA compared to no pretreatment; pretreated cells failed to form tumors. At right, tumor-free survival of Figure 7. Knockdown of Sumoylation Enzymes Repressed CD44 nude mice (n = 5 per group) inoculated with BT20 cells, with mice gavaged with and Blocked SUMO Conjugation of TFAP2A AA versus vehicle. (A and B) Knockdown of Ubc9 and PIAS1 by siRNA repressed expression of (F) At left, IOWA-1T cells were pretreated with AA or vehicle, and mice were CD44 in BT549 (A) and BT20 (B) cells showing same effect as GA and AA. Error followed until requiring euthanasia due to tumor size (Overall Survival). Pre- bars indicate SEM. treatment with AA inhibited tumor formation; n = 5 mice per group, p = 0.002. (C and D) Endogenous TFAP2A was examined by western blot in BT549 (basal) At right, nude mice were inoculated with 1 3 106,53 105, or 2.5 3 105 IOWA- (C) and MCF-7 (luminal) (D) cells. The SUMO-conjugated form of TFAP2A is 1T cells, and mice were gavaged with either vehicle or AA; vehicle-gavaged seen in both cell types (denoted by an asterisk), and knockdown of either Ubc9 mice developed tumors in 5, 10, and 12 days, respectively. Mice gavaged with or PIAS1 signiﬁcantly reduced SUMO-conjugated TFAP2A. MW, molecular AA failed to form tumors over the course of the experiment; n = 3 mice per weight markers. group, p = 0.025.

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 757 Cancer Cell Sumoylation Required for Basal Breast Cancer

Figure 8. Schematic Model of Gene Regu- Scenario 1: SUMO intact, TFAP2C Active lation in Breast Cancer Subtypes In scenario 1 with an intact SUMO pathway and active TFAP2C, breast cancer cells are main- TFAP2C TFAP2A SUMO tained in the luminal phenotype. Transcriptional regulation of many luminal gene promoters are ERα FOXA1 coordinately regulated by TFAP2C, ERa, and FOXA1, as well as potentially other cofactors. As MET diagrammed in scenario 2, loss of TFAP2C in the presence of the SUMO pathway induces EMT, characterized by repression of luminal genes and Scenario 2: SUMO intact, TFAP2C Inactive induction of basal-associated genes. As shown in scenario 3, inhibition of the SUMO pathway in the absence of TFAP2C activity, SUMO-unconju- TFAP2C TFAP2A SUMO gated TFAP2A induces mesenchymal-to-epithelial transition (MET), characterized by repression ERα FOXA1 of basal-associated gene expression and induction of luminal-associated genes. EMT

Scenario 3: SUMO Blocked, TFAP2C Inactive the CD44+/hi/CD24À/low tumor-initiating cell population suggests that this class Luminal Basal TFAP2C TFAP2A of agents may have an effect on basal Cancer Cells Cancer Cells breast cancers either alone or in combi- ERα FOXA1 nation with conventional chemotherapy. Since the CD44+/hi/CD24À/low population MET deﬁnes the tumor-initiating cells in many types of carcinomas, it is possible that SUMO inhibitors may have clinical effects in a wide range of carcinomas. It is inter- that SENP1 may regulate transcriptional activity of AP-2 factors esting that sumoylation inhibitors did not affect MCF10A cells, in breast cancer by inducing SUMO deconjugation. Further work which are commonly used as a model for normal breast cells. is needed to elucidate details of the transcriptional mechanisms Hence, the sumoylation pathway appears to be critical for main- whereby sumoylation speciﬁcally inhibits the functional activity taining the basal breast cancer subtype and is not a general of certain factors such as TFAP2A. mechanism regulating CD44 expression in normal breast cells.

+/hi –/low Sumoylation Inhibitors Clear the CD44 /CD24 Cell EXPERIMENTAL PROCEDURES Compartment Basal breast cancers are characterized by a relatively high Cell Lines percentage of cells expressing the characteristic markers The human breast cancer cell lines were derived and maintained as described elsewhere (Cyr et al., 2014). Under a protocol approved by the University of CD44+/hi/CD24À/low, which include tumor-initiating cells associ- Iowa Institutional Review Board and with informed consent, primary cancer ated with the outgrowth of tumor xenografts (Al-Hajj et al., cells were obtained from surgical resection specimens, and cell suspensions +/hi À/low 2003; Iqbal et al., 2013). The CD44 /CD24 population is were prepared with gentle collagenase/hyaluronidase (Stemcell Technologies) relatively chemoresistant and becomes enriched after chemo- (Ponti et al., 2005). The IOWA-1T cell line was established from a primary basal therapy (Lee et al., 2011). Stable knockdown of TFAP2C in tumor (R.J.W. and M.V.B., unpublished data). luminal cancer cells induced EMT, characterized by the repression of luminal gene expression, activation of basal-associated ChIP-Seq genes, and an increased population of cells expressing the ChIP-Seq was performed as described elsewhere (Woodﬁeld et al., 2010). CD44+/hi/CD24À/low markers (Figure 8)(Cyr et al., 2014). In the Sumoylation Assays present study, SUMO inhibition allowed TFAP2A to acquire Sumoylation in vitro was conducted using the SUMOlink SUMO-1 Kit TFAP2C-like repression activity, inhibiting CD44 expression, and SUMOlink SUMO-2/3 Kit (Active Motif). SUMO plasmids and +/hi À/low clearing cells expressing CD44 /CD24 markers, and pcDNA3.1-TFAP2A or K10R mutant cells were used for in vitro protein blocking the outgrowth of cancer xenografts. Of particular clin- production. MCF-7 cells were transfected for 48 hr with 2 mg SUMO- ical relevance, sumoylation inhibitors were able to efﬁciently expressing plasmids (Feng et al., 2013), which were kindly provided by clear the CD44+/hi/CD24À/low population in a primary basal breast Dr. Xiaolu Yang (University of Pennsylvania). As indicated in some experiments, the proteasome inhibitor MG132 was added as described elsewhere cancer obtained from a patient with a locally advanced breast (Chu and Yang, 2011). cancer that was refractory to conventional chemotherapy. The +/hi À/low high percentage of cells expressing the CD44 /CD24 Western Blots markers was likely due to selection from the treatment with Western blots were performed as described elsewhere (Cyr et al., 2014; Kulak chemotherapy. The remarkable effect of SUMO inhibitors to clear et al., 2013).

758 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

AP-2 Constructs Allouche, A., Nolens, G., Tancredi, A., Delacroix, L., Mardaga, J., Fridman, V., AP-2 constructs were amplified using previously cloned cDNAs for template Winkler, R., Boniver, J., Delvenne, P., and Begon, D.Y. (2008). The combined (McPherson and Weigel, 1999). Gateway TFAP2A and TFAP2C clones were immunodetection of AP-2alpha and YY1 transcription factors is associated inserted in-frame into pG-LAP1 (Torres et al., 2009) via LR clonase reaction with ERBB2 gene overexpression in primary breast tumors. Breast Cancer using Gateway LR Clonase II Enzyme Mix (Invitrogen). Res. 10, R9. Bai, F., Smith, M.D., Chan, H.L., and Pei, X.H. (2013). Germline mutation of GA and AA Treatment Brca1 alters the fate of mammary luminal cells and causes luminal-to-basal 5 2 Cells were plated (2.5 3 10 /10 cm ) and treated with 10 mM GA or AA (Sigma) mammary tumor transformation. Oncogene 32, 2715–2725. for 2–4 days and collected for quantitative PCR, western blot and fluores- Balko, J.M., Miller, T.W., Morrison, M.M., Hutchinson, K., Young, C., Rinehart, cence-activated cell sorting (FACS) analysis. C., Sa´ nchez, V., Jee, D., Polyak, K., Prat, A., et al. (2012). The receptor tyrosine kinase ErbB3 maintains the balance between luminal and basal breast epithe- Flow Cytometry lium. Proc. Natl. Acad. Sci. USA 109, 221–226. FACS analysis was performed as described elsewhere (Cyr et al., 2014; Begon, D.Y., Delacroix, L., Vernimmen, D., Jackers, P., and Winkler, R. (2005). Roederer and Hardy, 2001)(http://www.flowjo.com/v8/html/distancing.html). Yin Yang 1 cooperates with activator protein 2 to stimulate ERBB2 gene expression in mammary cancer cells. J. Biol. Chem. 280, 24428–24434. Yeast Two-Hybrid Assay Yeast two-hybrid for AP-2 factors was performed as described elsewhere Berlato, C., Chan, K.V., Price, A.M., Canosa, M., Scibetta, A.G., and Hurst, (McPherson et al., 2002). H.C. (2011). Alternative TFAP2A isoforms have distinct activities in breast cancer. Breast Cancer Res. 13, R23. Tumor Xenografts Bernardo, G.M., Bebek, G., Ginther, C.L., Sizemore, S.T., Lozada, K.L., Following a vertebrate animal protocol approved by the University of Iowa Miedler, J.D., Anderson, L.A., Godwin, A.K., Abdul-Karim, F.W., Slamon, Institutional Animal Care and Use Committee, xenografts were generated by D.J., and Keri, R.A. (2013). FOXA1 represses the molecular phenotype of basal inoculating 5 3 106 BT20 cells or indicated number of IOWA-1T cells into breast cancer cells. Oncogene 32, 554–563. nude mice as described elsewhere (Woodfield et al., 2007). All experiments Bertucci, F., Finetti, P., and Birnbaum, D. (2012). Basal breast cancer: a com- conformed to the regulatory standards reviewed in the animal protocol. plex and deadly molecular subtype. Curr. Mol. Med. 12, 96–110. Bettermann, K., Benesch, M., Weis, S., and Haybaeck, J. (2012). SUMOylation Statistical Analysis in carcinogenesis. Cancer Lett. 316, 113–125. Statistical analysis was performed using the two-sided Student’s t test for Bosher, J.M., Totty, N.F., Hsuan, J.J., Williams, T., and Hurst, H.C. (1996). A continuous variables. Comparisons for xenografts were performed using the family of AP-2 proteins regulates c-erbB-2 expression in mammary carcinoma. log rank test. Oncogene 13, 1701–1707. Bossis, G., and Melchior, F. (2006). Regulation of SUMOylation by reversible ACCESSION NUMBERS oxidation of SUMO conjugating enzymes. Mol. Cell 21, 349–357.

The ChIP-seq data are accessible in the Gene Expression Omnibus database Buchwalter, G., Hickey, M.M., Cromer, A., Selfors, L.M., Gunawardane, R.N., under accession number GSE44257. Frishman, J., Jeselsohn, R., Lim, E., Chi, D., Fu, X., et al. (2013). PDEF pro- motes luminal differentiation and acts as a survival factor for ER-positive breast cancer cells. Cancer Cell 23, 753–767. SUPPLEMENTAL INFORMATION Carey, L.A., Perou, C.M., Livasy, C.A., Dressler, L.G., Cowan, D., Conway, K., Supplemental Information includes Supplemental Experimental Procedures Karaca, G., Troester, M.A., Tse, C.K., Edmiston, S., et al. (2006). Race, breast 295 and four figures and can be found with this article online at http://dx.doi.org/ cancer subtypes, and survival in the Carolina Breast Cancer Study. JAMA , 10.1016/j.ccr.2014.04.008. 2492–2502. Carr, J.R., Kiefer, M.M., Park, H.J., Li, J., Wang, Z., Fontanarosa, J., DeWaal, ACKNOWLEDGMENTS D., Kopanja, D., Benevolenskaya, E.V., Guzman, G., and Raychaudhuri, P. (2012). FoxM1 regulates mammary luminal cell fate. Cell Rep. 1, 715–729. This work was supported by NIH grants R01CA109294 (principal investigator Chakrabarti, R., Wei, Y., Romano, R.A., DeCoste, C., Kang, Y., and Sinha, S. [PI]: R.J.W.) and T32CA148062 (PI: R.J.W.) and a generous gift from the Kristen (2012). Elf5 regulates mammary gland stem/progenitor cell fate by influencing Olewine Milke Breast Cancer Research Fund. P.M.S. was supported by NIH notch signaling. Stem Cells 30, 1496–1508. grant T32CA148062. This research was supported in part through computa- Chen, C.H., Chang, C.C., Lee, T.H., Luo, M., Huang, P., Liao, P.H., Wei, S., Li, tional resources provided by the University of Iowa. F.A., Chen, R.H., Zhou, X.Z., et al. (2013). SENP1 deSUMOylates and regulates Pin1 protein activity and cellular function. Cancer Res. 73, 3951–3962. Received: July 22, 2013 Chu, Y., and Yang, X. (2011). SUMO E3 ligase activity of TRIM proteins. Revised: November 12, 2013 Oncogene 30, 1108–1116. Accepted: April 11, 2014 Published: May 15, 2014 Cuben˜ as-Potts, C., and Matunis, M.J. (2013). SUMO: a multifaceted modifier of chromatin structure and function. Dev. Cell 24, 1–12. REFERENCES Cyr, A.R., Kulak, M.V., Park, J.M., Bogachek, M.V., Spanheimer, P.M., Woodfield, G.W., White-Baer, L.S., O’Malley, Y.Q., Sugg, S.L., Olivier, A.K., Abdel-Hafiz, H.A., and Horwitz, K.B. (2012). Control of progesterone receptor et al. (2014). TFAP2C governs the luminal epithelial phenotype in mammary transcriptional synergy by SUMOylation and deSUMOylation. BMC Mol. Biol. development and carcinogenesis. Oncogene. Published online January 27, 13, 10. 2014. http://dx.doi.org/10.1038/onc.2013.569. Ailan, H., Xiangwen, X., Daolong, R., Lu, G., Xiaofeng, D., Xi, Q., Xingwang, H., deConinck, E.C., McPherson, L.A., and Weigel, R.J. (1995). Transcriptional Rushi, L., Jian, Z., and Shuanglin, X. (2009). Identification of target genes of regulation of estrogen receptor in breast carcinomas. Mol. Cell. Biol. 15, transcription factor activator protein 2 gamma in breast cancer cells. BMC 2191–2196. 9 Cancer , 279. Delacroix, L., Begon, D., Chatel, G., Jackers, P., and Winkler, R. (2005). Distal Al-Hajj, M., Wicha, M.S., Benito-Hernandez, A., Morrison, S.J., and Clarke, ERBB2 promoter fragment displays specific transcriptional and nuclear bind- M.F. (2003). Prospective identification of tumorigenic breast cancer cells. ing activities in ERBB2 overexpressing breast cancer cells. DNA Cell Biol. 24, Proc. Natl. Acad. Sci. USA 100, 3983–3988. 582–594.

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 759 Cancer Cell Sumoylation Required for Basal Breast Cancer

Eckert, D., Buhl, S., Weber, S., Ja¨ ger, R., and Schorle, H. (2005). The AP-2 Lee, H.E., Kim, J.H., Kim, Y.J., Choi, S.Y., Kim, S.W., Kang, E., Chung, I.Y., family of transcription factors. Genome Biol. 6, 246. Kim, I.A., Kim, E.J., Choi, Y., et al. (2011). An increase in cancer stem cell Eloranta, J.J., and Hurst, H.C. (2002). Transcription factor AP-2 interacts with population after primary systemic therapy is a poor prognostic factor in breast 104 the SUMO-conjugating enzyme UBC9 and is sumolated in vivo. J. Biol. Chem. cancer. Br. J. Cancer , 1730–1738. 277, 30798–30804. Lehmann, B.D., Bauer, J.A., Chen, X., Sanders, M.E., Chakravarthy, A.B., Fang, S.H., Chen, Y., and Weigel, R.J. (2009). GATA-3 as a marker of hormone Shyr, Y., and Pietenpol, J.A. (2011). Identification of human triple-negative response in breast cancer. J. Surg. Res. 157, 290–295. breast cancer subtypes and preclinical models for selection of targeted ther- apies. J. Clin. Invest. 121, 2750–2767. Feng, W., and Williams, T. (2003). Cloning and characterization of the mouse AP-2 epsilon gene: a novel family member expressed in the developing olfac- Leitao, B.B., Jones, M.C., and Brosens, J.J. (2011). The SUMO E3-ligase tory bulb. Mol. Cell. Neurosci. 24, 460–475. PIAS1 couples reactive oxygen species-dependent JNK activation to oxidative cell death. FASEB J. 25, 3416–3425. Feng, Z.J., Gurung, B., Jin, G.H., Yang, X.L., and Hua, X.X. (2013). SUMO modification of menin. Am. J. Cancer Res. 3, 96–106. Li, W., and Cornell, R.A. (2007). Redundant activities of Tfap2a and Tfap2c are Friedrichs, N., Ja¨ ger, R., Paggen, E., Rudlowski, C., Merkelbach-Bruse, S., required for neural crest induction and development of other non-neural ecto- 304 Schorle, H., and Buettner, R. (2005). Distinct spatial expression patterns of derm derivatives in zebrafish embryos. Dev. Biol. , 338–354. AP-2alpha and AP-2gamma in non-neoplastic human breast and breast can- McPherson, L.A., and Weigel, R.J. (1999). AP2alpha and AP2gamma: a com- cer. Mod. Pathol. 18, 431–438. parison of binding site specificity and trans-activation of the estrogen receptor 27 Friedrichs, N., Steiner, S., Buettner, R., and Knoepfle, G. (2007). promoter and single site promoter constructs. Nucleic Acids Res. , 4040– Immunohistochemical expression patterns of AP2alpha and AP2gamma in 4049. the developing fetal human breast. Histopathology 51, 814–823. McPherson, L.A., Baichwal, V.R., and Weigel, R.J. (1997). Identification of Fukuda, I., Ito, A., Hirai, G., Nishimura, S., Kawasaki, H., Saitoh, H., Kimura, K., ERF-1 as a member of the AP2 transcription factor family. Proc. Natl. Acad. 94 Sodeoka, M., and Yoshida, M. (2009). Ginkgolic acid inhibits protein Sci. USA , 4342–4347. SUMOylation by blocking formation of the E1-SUMO intermediate. Chem. McPherson, L.A., Loktev, A.V., and Weigel, R.J. (2002). Tumor suppressor Biol. 16, 133–140. activity of AP2alpha mediated through a direct interaction with p53. J. Biol. 277 Gill, G. (2003). Post-translational modification by the small ubiquitin-related Chem. , 45028–45033. modifier SUMO has big effects on transcription factor activity. Curr. Opin. Moser, M., Imhof, A., Pscherer, A., Bauer, R., Amselgruber, W., Sinowatz, F., Genet. Dev. 13, 108–113. Hofsta¨ dter, F., Schu¨ le, R., and Buettner, R. (1995). Cloning and characteriza- 121 Girdwood, D., Bumpass, D., Vaughan, O.A., Thain, A., Anderson, L.A., tion of a second AP-2 transcription factor: AP-2 beta. Development , 2779– Snowden, A.W., Garcia-Wilson, E., Perkins, N.D., and Hay, R.T. (2003). P300 2788. transcriptional repression is mediated by SUMO modification. Mol. Cell 11, Orso, F., Penna, E., Cimino, D., Astanina, E., Maione, F., Valdembri, D., 1043–1054. Giraudo, E., Serini, G., Sismondi, P., De Bortoli, M., and Taverna, D. (2008). Herschkowitz, J.I., Simin, K., Weigman, V.J., Mikaelian, I., Usary, J., Hu, Z., AP-2alpha and AP-2gamma regulate tumor progression via specific genetic 22 Rasmussen, K.E., Jones, L.P., Assefnia, S., Chandrasekharan, S., et al. programs. FASEB J. , 2702–2714. (2007). Identification of conserved gene expression features between murine Pellikainen, J., Naukkarinen, A., Ropponen, K., Rummukainen, J., Kataja, V., mammary carcinoma models and human breast tumors. Genome Biol. 8, R76. Kellokoski, J., Eskelinen, M., and Kosma, V.M. (2004). Expression of HER2 40 Hoffman, T.L., Javier, A.L., Campeau, S.A., Knight, R.D., and Schilling, T.F. and its association with AP-2 in breast cancer. Eur. J. Cancer , 1485–1495. (2007). Tfap2 transcription factors in zebrafish neural crest development and Ponti, D., Costa, A., Zaffaroni, N., Pratesi, G., Petrangolini, G., Coradini, D., ectodermal evolution. J. Exp. Zoolog. B Mol. Dev. Evol. 308, 679–691. Pilotti, S., Pierotti, M.A., and Daidone, M.G. (2005). Isolation and in vitro prop- Holmstrom, S.R., Chupreta, S., So, A.Y., and In˜ iguez-Lluhı´, J.A. (2008). SUMO- agation of tumorigenic breast cancer cells with stem/progenitor cell proper- mediated inhibition of glucocorticoid receptor synergistic activity depends on ties. Cancer Res. 65, 5506–5511. stable assembly at the promoter but not on DAXX. Mol. Endocrinol. 22, 2061– Roederer, M., and Hardy, R.R. (2001). Frequency difference gating: a multivar- 2075. iate method for identifying subsets that differ between samples. Cytometry 45, Ihara, M., Stein, P., and Schultz, R.M. (2008). UBE2I (UBC9), a SUMO-conju- 56–64. gating enzyme, localizes to nuclear speckles and stimulates transcription in Ryu, J., Cho, S., Park, B.C., and Lee, H. (2010). Oxidative stress-enhanced mouse oocytes. Biol. Reprod. 79, 906–913. SUMOylation and aggregation of ataxin-1: implication of JNK pathway. Iqbal, J., Chong, P.Y., and Tan, P.H. (2013). Breast cancer stem cells: an Biochem. Biophys. Res. Commun. 393, 280–285. update. J. Clin. Pathol. 66, 485–490. Scibetta, A.G., Wong, P.P., Chan, K.V., Canosa, M., and Hurst, H.C. (2010). Ja¨ ger, R., Werling, U., Rimpf, S., Jacob, A., and Schorle, H. (2003). Dual association by TFAP2A during activation of the p21cip/CDKN1A pro- Transcription factor AP-2gamma stimulates proliferation and apoptosis and moter. Cell Cycle 9, 4525–4532. impairs differentiation in a transgenic model. Mol. Cancer Res. 1, 921–929. Siegel, R., Naishadham, D., and Jemal, A. (2012). Cancer statistics, 2012. CA Johnson, E.S., and Blobel, G. (1997). Ubc9p is the conjugating enzyme for the Cancer J. Clin. 62, 10–29. ubiquitin-like protein Smt3p. J. Biol. Chem. 272, 26799–26802. Sørlie, T., Perou, C.M., Tibshirani, R., Aas, T., Geisler, S., Johnsen, H., Hastie, Joshi, H., Nord, S.H., Frigessi, A., Børresen-Dale, A.L., and Kristensen, V.N. T., Eisen, M.B., van de Rijn, M., Jeffrey, S.S., et al. (2001). Gene expression (2012). Overrepresentation of transcription factor families in the genesets patterns of breast carcinomas distinguish tumor subclasses with clinical impli- underlying breast cancer subtypes. BMC Genomics 13, 199. cations. Proc. Natl. Acad. Sci. USA 98, 10869–10874. Jozwik, K.M., and Carroll, J.S. (2012). Pioneer factors in hormone-dependent Tan, S.K., Lin, Z.H., Chang, C.W., Varang, V., Chng, K.R., Pan, Y.F., Yong, E.L., cancers. Nat. Rev. Cancer 12, 381–385. Sung, W.K., and Cheung, E. (2011). AP-2g regulates oestrogen receptor-medi- 30 Kao, J., Salari, K., Bocanegra, M., Choi, Y.L., Girard, L., Gandhi, J., Kwei, K.A., ated long-range chromatin interaction and gene transcription. EMBO J. , Hernandez-Boussard, T., Wang, P., Gazdar, A.F., et al. (2009). Molecular 2569–2581. profiling of breast cancer cell lines defines relevant tumor models and provides Torres, J.Z., Miller, J.J., and Jackson, P.K. (2009). High-throughput a resource for cancer gene discovery. PLoS ONE 4, e6146. generation of tagged stable cell lines for proteomic analysis. Proteomics 9, Kulak, M.V., Cyr, A.R., Woodfield, G.W., Bogachek, M., Spanheimer, P.M., Li, 2888–2891. T., Price, D.H., Domann, F.E., and Weigel, R.J. (2013). Transcriptional regula- Turner, B.C., Zhang, J., Gumbs, A.A., Maher, M.G., Kaplan, L., Carter, D., tion of the GPX1 gene by TFAP2C and aberrant CpG methylation in human Glazer, P.M., Hurst, H.C., Haffty, B.G., and Williams, T. (1998). Expression of breast cancer. Oncogene 32, 4043–4051. AP-2 transcription factors in human breast cancer correlates with the

760 Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. Cancer Cell Sumoylation Required for Basal Breast Cancer

regulation of multiple growth factor signalling pathways. Cancer Res. 58, Woodfield, G.W., Horan, A.D., Chen, Y., and Weigel, R.J. (2007). TFAP2C con- 5466–5472. trols hormone response in breast cancer cells through multiple pathways of estrogen signaling. Cancer Res. 67, 8439–8443. Valentin, M.D., da Silva, S.D., Privat, M., Alaoui-Jamali, M., and Bignon, Y.J. (2012). Molecular insights on basal-like breast cancer. Breast Cancer Res. Woodfield, G.W., Chen, Y., Bair, T.B., Domann, F.E., and Weigel, R.J. (2010). Treat. 134, 21–30. Identification of primary gene targets of TFAP2C in hormone responsive breast carcinoma cells. Genes Chromosomes Cancer 49, 948–962. Williams, T., and Tjian, R. (1991a). Analysis of the DNA-binding and Yang, J.W., Lee, E.Y., and Kang, K.W. (2006). ErbB2 overexpression in p53-in- activation properties of the human transcription factor AP-2. Genes Dev. 5, activated mammary epithelial cells. FEBS Lett. 580, 6501–6508. 670–682. Zhang, J., Brewer, S., Huang, J., and Williams, T. (2003). Overexpression of Williams, T., and Tjian, R. (1991b). Characterization of a dimerization motif in transcription factor AP-2alpha suppresses mammary gland growth and 251 AP-2 and its function in heterologous DNA-binding proteins. Science , morphogenesis. Dev. Biol. 256, 127–145. 1067–1071. Zhao, F., Satoda, M., Licht, J.D., Hayashizaki, Y., and Gelb, B.D. (2001). Williams, T., Admon, A., Lu¨ scher, B., and Tjian, R. (1988). Cloning and expres- Cloning and characterization of a novel mouse AP-2 transcription factor, AP- sion of AP-2, a cell-type-specific transcription factor that activates inducible 2delta, with unique DNA binding and transactivation properties. J. Biol. enhancer elements. Genes Dev. 2 (12A), 1557–1569. Chem. 276, 40755–40760.

Cancer Cell 25, 748–761, June 16, 2014 ª2014 Elsevier Inc. 761