Published OnlineFirst November 14, 2014; DOI: 10.1158/0008-5472.CAN-13-3455 Cancer Molecular and Cellular Pathobiology Research
Tropomodulin 1 Expression Driven by NF-kB Enhances Breast Cancer Growth Taku Ito-Kureha1,2, Naohiko Koshikawa3, Mizuki Yamamoto4, Kentaro Semba4, Noritaka Yamaguchi5, Tadashi Yamamoto2, Motoharu Seiki6, and Jun-ichiro Inoue1
Abstract
Triple-negative breast cancers (TNBC), which include the basal- regulated directly by NF-kB and was significantly higher in TNBC like and claudin-low disease subtypes, are aggressive malignan- than other breast cancer subtypes. TMOD1 elevation is associated cies for which effective therapeutic targets are lacking. NF-kB with enhanced tumor growth in a mouse tumor xenograft model activation has an established role in breast malignancy, and it is and in a 3D type I collagen culture. TMOD1-dependent tumor higher in TNBC than other breast cancer subtypes. On this basis, growth was correlated with MMP13 induction, which was medi- we hypothesized that proteins derived from NF-kB target genes ated by TMOD1-dependent accumulation of b-catenin. Overall, might be molecular targets for TNBC therapy. In this study, we our study highlighted a novel TMOD1-mediated link between conducted a microarray-based screen for novel NF-kB–inducible NF-kB activation and MMP13 induction, which accounts in part proteins as candidate therapeutic targets, identifying tropomo- for the NF-kB–dependent malignant phenotype of TNBC. Cancer dulin 1 (TMOD1) as a lead candidate. TMOD1 expression was Res; 75(1); 1–11. 2014 AACR.
Introduction higher proliferative activity as reflected in high Ki-67 expression when compared with luminal-like subtypes (5). Therefore, it is Gene-expression analyses have defined five breast cancer necessary to find molecular signatures and signaling pathways subtypes (luminal-like, ERBB2-enriched, basal-like, claudin- that contribute to the malignancy of TNBCs. low, and normal breast-like), each of which is thought to be The NF-kB family of transcriptional factors plays a critical derived from a distinct differentiation stage of mammary epi- role in inflammation, immunoregulation, and cell differentiation thelial cells, thereby displaying unique prognostic features (6, 7). This family consists of five members, including p50, p52, (1–3). The luminal-like subtype is characterized as either an þ RELA (p65), RELB, and c-REL, which form homomeric or hetero- estrogen receptor–positive (ER ) or a progesterone receptor– þ meric dimers to activate transcription of the target genes. NF-kBis positive (PR ) phenotype, whereas the basal-like and claudin- made transcriptionally inactive by being sequestered in the cyto- low subtypes constitute the majority of triple-negative cancers plasm when it forms complexes with the IkB family, including (TNBCs; ER ,PR , and ERBB2 ), show a higher malignancy IkBa,IkBb,IkBe, and the p105 and p100 precursors of p50 and than other subtypes, and exhibit a poor prognosis against p52, respectively. Nuclear translocation of NF-kB can be driven by various methods of therapy. Triple-negative breast cancers two distinct signaling pathways. In the canonical pathway, a large (TNBC) were identified on the basis of low/absent expression number of stimuli, including various cytokines and bacterial and of luminal differentiation markers, and enrichment of epithe- viral products, induce IkB kinase (IKK) b-catalyzed phosphory- lial-to-mesenchymal transition (EMT) and stem cell markers. lation and proteasomal degradation of IkBa, followed by nuclear Therefore, although luminal-like cells appear more differenti- translocation of mainly p50-RELA heterodimers (8). The nonca- ated and form tight cell–cell junctions, TNBCs appear less nonical pathway is activated by receptors that are crucial in the differentiated and have a more mesenchymal-like appearance. formation of lymphoid organs and lymphocyte development, TNBCs are much more frequently highly invasive (4) and show such as the lymphotoxin b receptor, the receptor activator of NF-kB (RANK), and CD40. This pathway induces the IKKa- catalyzed phosphorylation of the C-terminal half of p100 1 Division of Cellular and Molecular Biology, Institute of Medical Sci- that sequesters RelB in the cytoplasm, which leads to polyubi- ence, University of Tokyo, Tokyo, Japan. 2Cell Signal Unit, Okinawa Institute of Science and Technology, Okinawa, Japan. 3Division of quitination-dependent processing of p100 to p52 and the nuclear Cancer Cell Research, Institute of Medical Science, University of Tokyo, translocation of the p52-RELB heterodimers (9). Accumulating 4 Tokyo, Japan. Department of Life Science and Medical Bio-science, evidence indicates that aberrant NF-kB activation leads to tumor- Waseda University,Tokyo, Japan. 5Department of Molecular Cell Biol- ogy, Graduate School of Pharmaceutical Sciences, Chiba University, igenesis and cancer malignancy through the expression of genes Chiba, Japan. 6Graduate School of Medicine, Kochi University, Kochi, involved in survival, metastasis, and angiogenesis (6, 10). We Japan. have previously reported that TNBCs undergo constitutive and Corresponding Author: Jun-ichiro Inoue, Division of Cellular and Molecular strong activation of NF-kB, whose activation is transient in Biology, Institute of Medical Science, University of Tokyo, 4-6-1 Shirokanedai, normal cells upon various physiologic stimuli (11, 12). Further- Minato-ku, Tokyo 108-8639, Japan. Phone: 813-5449-5275; Fax: 813-5449-5421; more, the adenovirus-mediated expression of a nondegradable E-mail: [email protected] IkBa super-repressor (IkBaSR), in which Ser-32 and Ser-36 (the doi: 10.1158/0008-5472.CAN-13-3455 residues phosphorylated by IKKb) were substituted with alanine, 2014 American Association for Cancer Research. blocked NF-kB activation, and thereby inhibited the growth of
www.aacrjournals.org OF1
Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2014 American Association for Cancer Research. Published OnlineFirst November 14, 2014; DOI: 10.1158/0008-5472.CAN-13-3455
Ito-Kureha et al.
several TNBC cell lines. This result strongly suggests that consti- 3 mg/mL) and seeded into 24-well plates (500 mL each, 5 103 tutive NF-kB activation plays a critical role in TNBC malignancy. cells/mL). After an 8-day culture, the cells were extracted from the In this report, we demonstrate that tropomodulin 1 (TMOD1) collagen gels, and the numbers of extracted cells were measured is a novel NF-kB target gene and that TMOD1 protein, which with a hemocytometer. In selected experiments, cells were inhibits the elongation and depolymerization of actin filaments incubated with 5 mmol/L MMI270 (kindly provided by Dr. M. by binding to the pointed end of the actin filament (13–15), could Nakajima, Novartis Pharma) or the recombinant matrix metallo- be involved in the enhancement of the in vivo growth of TNBC. proteinase (MMP) inhibitors, TIMP-1 (10 mg/mL) or TIMP-2 (5 mg/mL; Daiichi Fine Chemical Co., Ltd.). Materials and Methods Invasion assay Cell culture Transwell inserts (upper chamber) with 8-mm pore size were MDA-MB-436, MDA-MB-231, and BT-549 were obtained from coated with collagen gel solution and air-dried. Cells were added the ATCC, and resuscitated from early passage liquid nitrogen to each Transwell. After 20 hours, cells that had migrated through stocks as needed. Cells were cultured for less than 3 months before the collagen gel and adhered to the other side of the insert were reinitiating cultures and were routinely inspected microscopically fixed and stained with 0.5% crystal violet. for stable phenotype. Microarray analysis for identification of NF-kB target genes Subcellular fractionation The preparation of recombinant adenoviruses was performed Cells are suspended in buffer A (10 mmol/L HEPES-KOH as described previously (11). RNA samples were prepared from pH 7.9 at 4 C, 1.5 mmol/L MgCl2, 10 mmol/L KC1, 0.5 MDA-MB-436 cells 24 hours after infections. Total RNA labeled mmol/L dithiothreitol, 0.2 mmol/L PMSF) and incubated for with Cy3 or Cy5 was hybridized to 3D-Gene Human Oligo chip 15 minutes on ice followed by centrifugation. The supernatant 25 k (25,370 distinct genes, Toray Industries Inc.). Genes with was stored as a cytoplasmic extract. The pellet was resuspended in Cy3/Cy5 normalized ratios greater than 2.0 were defined. Micro- buffer C (20 mmol/L HEPES-KOH pH 7.9, 25% glycerol, 420 array data have been deposited in the Gene Expression Omnibus mmol/L NaCl, 1.5 mmol/L MgCl2, 0.2 mmol/L EDTA, 0.5 mmol/ (GEO) under the accession number GSE56812. L dithiothreitol, and 0.2 mmol/L PMSF) and incubated on ice for 20 minutes. Cellular debris was removed by centrifugation, and Analysis of TMOD1 expression in breast cancers the supernatant was used as a nuclear extract. DNA microarray analysis of 35 human breast cancer cell lines (CIBEX database accession no. CBX20; ref. 16) was used to Tumor formation assay analyze TMOD1 expression in human breast cancer cell lines. 6 m MDA-MB-231 cells (10 ) in 100 L of RPMI1640 (Wako) A hierarchical clustering analysis of publicly available cDNA containing 50% Matrigel (BD Biosciences) were s.c. injected into microarray data derived from 251 primary mammary tumor BALB/c Slc-nu/nu female mice (5-week-old, Japan SLC, Inc.). samples (GEO Series accession no. GSE3494; ref. 17) was per- Tumor growth was monitored twice a week. Animal research formed using the GeneSpring software. complied with protocols approved by the Committee for Animal Experimentation of Waseda University. Statistical analysis Statistically significant differences between the mean values Chromatin immunoprecipitation assay were determined using the Student t test ( , P < 0.001; , P < A chromatin immunoprecipitation (ChIP) assay was per- 0.01; and , P < 0.05). The values represent the means of triplicate formed as described previously (12). Anti-RELA antibody (Santa samples SD. Cruz Biotechnology) was used for immunoprecipitation. The immunoprecipitation efficiency of the TMOD1 promoter region ( 713 to 486) was analyzed by real-time PCR using FastStart Results 0 Universal SYBR Green Master (Roche). The primer set used was 5 - Screening of novel NF-kB target genes in TNBCs 0 0 ATGTGGATCTGCTGCTTCCT-3 and 5 -TTAAGCACCTACTGCA- To search for novel NF-kB target genes involved in the malig- 0 TACA-3 . nancy of TNBC, the claudin-low subtype cell line MDA-MB-436 was used because its NF-kB activation level is significantly higher Luciferase assay than those in other claudin-low cell lines (11). The cells were Cells were transfected with 1 mg of pTOPflash-Luc or infected with adenovirus-expressing IkBaSR to block constitutive pFOPflash-Luc and 0.001 mg of pRL-Renilla Luciferase Reporter activation of NF-kB (11, 18). To avoid the selection of genes plasmid and harvested 48 hours after transfection. Reporter whose expressions are nonspecifically blocked by adenovirus activities were measured with a luminometer (LB9507; Berthold) infection, GFP-expressing adenovirus was used (11). A subse- using a dual luciferase reporter assay system (Promega). T-cell quent gene-expression analysis using a DNA microarray led to the factor/lymphoid enhancer factor (TCF/LEF) activity was defined identification of 48 genes whose expressions were reduced more as the ratio of TOPflash:FOPflash reporter activities. than 2-fold in IkBaSR-expressing cells when compared with both mock- and GFP-infected cells (Fig. 1A and Supplementary Proliferation assay Table S2). Twenty-two known NF-kB target genes, including For two-dimensional (2D) culture, cells (104/well) were cul- RELB, TNFAIP3 (also known as A20), and BIRC3 (also known tured for 4 days, the viable cell number was then determined by as cIAP2), were included in the 48 genes, indicating the validity trypan blue exclusion. For three-dimensional (3D) culture, cells of our procedure. Among the remaining 26 genes, four genes, were suspended with chilled collagen gel solution (Nitta Gelatin; including TMOD1, chromosome 15 open reading frame 48
OF2 Cancer Res; 75(1) January 1, 2015 Cancer Research
Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2014 American Association for Cancer Research. Published OnlineFirst November 14, 2014; DOI: 10.1158/0008-5472.CAN-13-3455
NF-kB–Induced Tropomodulin Enhances Breast Tumor Growth
ABinvestigated whether the TMOD1 expression was upregulated a k κ α κ α by TNF stimulation, which induces NF- B activation. Real-time I B -SR vs. Mock I B -SR vs. GFP MDA-MB-436 TMOD1 Virus: Mock GFP IκBα-SR RT-PCR analysis revealed that the expression of mRNA was increased within the first 4 hours after TNFa stimulation in TMOD1 MDA-MB-436 and MDA-MB-231 cells (Fig. 3A). Because the 385 48 432 C15orf48 TMOD1 promoter has five putative NF-kB–binding sites, we next CH25H analyzed the binding of RELA (also known as p65) to the TMOD1 VNN2 promoter region using a ChIP assay. RELA was recruited to the TMOD1 promoter in both cell lines (Fig. 3B). Moreover, TMOD1 RelB expression was also reduced by RELA downregulation or treat- 18S RNA 22 genes : Known NF-κB target genes ment with Bay11-7082, an inhibitor of IkBa phosphorylation, in 26 genes : Semiquantitative RT-PCR to check various claudin-low cell lines (Fig. 3C and Supplementary Fig. their NF-κB–dependent expression S1). These results indicated that the transcription of TMOD1 was k Figure 1. directly regulated by NF- B. k TMOD1 Screening for a novel NF-kB target gene. A, the scheme for screening novel Because NF- B is highly activated in TNBCs and is an NF-kB target genes. B, semiquantitative RT-PCR to analyze the effects of NF-kB target gene, we next checked whether TMOD1 is preferen- IkBaSR expression on the expression of the candidate genes. tially expressed in TNBCs. On the basis of our DNA microarray analysis of 35 human breast cancer cell lines (Supplementary (C15orf48), cholesterol 25-hydroxylase (CH25H), and vanin Materials and Methods; ref. 16), TMOD1 expression is significantly 2(VNN2), were selected as candidates for novel NF-kB target higher in TNBCs (Fig. 3D). Nevertheless, TMOD1 expression is genes based on the semiquantitative RT-PCR analysis (Fig. 1B). varied among claudin-low cell lines used in this study (Supple- mentary Fig. S1), which may account for their characteristic behav- TMOD1 overexpression enhanced in vivo growth of MDA-MB- ior. Furthermore, we used a published dataset of human breast 231 cells in the nude mouse xenograft model cancer specimens to investigate TMOD1 expression in primary To address the involvement of the four candidate genes in breast tumors. We first categorized 251 breast tumors (17) into tumor progression in vivo, we used the claudin-low subtype cell basal-like, claudin-low, ERBB2-enriched, and luminal-like sub- line MDA-MB-231 because this cell line can form tumors in the types based on a hierarchical clustering analysis of their gene- xenograft model more efficiently than MDA-MB-436 and BT-549. expression profiles (Fig. 3E, left). TMOD1 expression is significant- Each gene was introduced by gene transfer with the retrovirus ly higher in TNBCs than in ERBB2-enriched and luminal-like breast vector, and then the population of cells that stably expressed each tumors (Fig. 3E, right). In normal mouse mammary epithelial cells, gene was selected by puromycin treatment to generate 231- Tmod1 expression is significantly higher in basal and mature TMOD1, 231-C15orf48, 231-CH25H, and 231-VNN2. The con- luminal cells than luminal progenitor cells (Fig. 3F). Because it trol vector was used to generate 231-control (Fig. 2A–D). Because has been reported that luminal-like subtype tumors originate from infection efficiency was about 20% and no cell cloning, was mature luminal cells and basal-like subtype tumors from luminal performed, the effects observed here are not limited to one specific progenitor cells, high TMOD1 expression in TNBCs does not reflect cell clone, and it is also unlikely that the selected population was gene-expression characteristics of normal epithelial cells. Together, biased compared with the parental cell population. These cells these results point to TMOD1 as a promising specific therapeutic were then injected bilaterally into the fat pads of BALB/c nu/nu target of TNBCs. TMOD1 was originally found as an actin-capping mice, and tumor growth was subsequently measured using cali- protein that binds to the N-terminus of tropomyosin (19). There- pers. Through this in vivo tumor growth screening, we found that fore, pointed-end capping by TMOD1 helps to maintain the the overexpression of TMOD1 resulted in enhanced tumor growth constant lengths of the actin filaments in skeletal muscle and in in nude mice, whereas the overexpression of the other three genes the red cell membrane skeleton. However, the role of TMOD1 in had no significant effect (Fig. 2A–D). Tumors formed by injecting breast cancer development remains to be elucidated. 231-TMOD1 appeared within a month, and the TMOD1-over- expressing tumors were about 2-fold larger compared with those The TMOD1-induced enhancement of proliferation of MDA- of control tumors 40 days after engraftment (Fig. 2A). In addition, MB-231 in the 3D culture system is mediated by the MMP family tumor weights were also significantly increased by TMOD1 over- To elucidate the molecular mechanism of the TMOD1-depen- expression when compared with control tumors (Fig. 2E). His- dent enhancement of in vivo tumor growth (Fig. 2A), we first tologic analysis revealed that 231-TMOD1 tumors often had a analyzed the effects of TMOD1 overexpression on cell growth in a necrotic area in their central portion, and a higher number of 2D normal culture system. In 2D culture, 231-control and 231- mitotic cells than 231-control tumors, but otherwise showed no TMOD1 showed similar rates of cell proliferation (Fig. 4A). In differences in morphology and angiogenesis (Fig. 2F). These data contrast, TMOD1 overexpression resulted in enhanced cell pro- indicate that TMOD1 is involved in promoting the in vivo growth liferation in the type I collagen 3D culture system (Fig. 4B), a of this claudin-low breast cancer cell line. model culture system for in vivo tumor cell growth (20, 21). Because the proportion of the sub-G1 population was not reduced TMOD1 gene expression was directly regulated by NF-kB, and by TMOD1 overexpression (Fig. 4C), the enhancement of prolif- TNBCs showed higher TMOD1 expression than other breast eration was due to enhanced cell division rather than reduced cancers apoptosis. Although less prominent than in the case of MDA-MB- Whereas we found that TMOD1 expression was reduced by 231 cells, similar 3D culture-specific and TMOD1 overexpres- blocking NF-kB activation (Fig. 1B), it was unclear whether the sion–dependent growth advantages were observed in two other transcription of TMOD1 was directly regulated by NF-kB. We first distinct claudin-low cell lines, MDA-MB-436 and BT-549
www.aacrjournals.org Cancer Res; 75(1) January 1, 2015 OF3
Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2014 American Association for Cancer Research. Published OnlineFirst November 14, 2014; DOI: 10.1158/0008-5472.CAN-13-3455
Ito-Kureha et al.
AB 231-Control (n = 6) 600 231-TMOD1 (n = 8) 600 231-Control (n = 4) ) )
** 3 3 ** 231-C15orf48 (n = 4)
400 231-Control 231-TMOD1 400 TMOD1 231-Control 231-C15orf48 * C15orf48 Tubulin 200 ** 200 Tubulin * * * * Tumor volume (mm volume Tumor Tumor volume (mm volume Tumor 0 0 Figure 2. 10 20 30 40 day 10 20 30 40 day TMOD1 enhances tumor growth in a mouse xenograft model. A to D, control MDA-MB-231 cells (231- CE control) and the cells expressing one of the four candidate genes 0.6 (231-TMOD1, -C15orf48, -CH25H, or
231-Control 231-TMOD1 ** 600 -VNN2) were subcutaneously ) 231-Control (n = 4) 0.5 3 231-CH25H (n = 4) 0.4 implanted into the fat pads of nude n 0.3 mice; , number of injections. 400 Expression of each candidate protein 231-Control 231-CH25H 0.2 CH25H was determined by Western blotting 0.1
Tumor weight weight (g) Tumor using the anti-FLAG antibody. E, Tubulin 200 0 tumors formed in mice (left); scale bars, 10 mm. Weights of tumors 40 Tumor volume (mm volume Tumor days after injection (right). F, sections 0 of tumors from mice injected with 231- 10 20 30 40 day control or 231-TMOD1 were stained 231-Control 231-TMOD1 with hematoxylin and eosin. N, a necrotic area; arrowheads, mitotic cells; scale bars, 100 mm (top) and 50 mm (bottom). , P < 0.01; D 231-Control (n = 4) F 231-Control 231-TMOD1 , P < 0.05. 600 231-VNN2 (n = 4) ) 3 NN 400 231-Control 231-VNN2 VNN2 Tubulin 200 Tumor volume (mm volume Tumor 0 10 20 30 40 day
(Supplementary Fig. S2A–S2C). These results suggest that nant MMP inhibitors, TIMP1 and TIMP2: TIMP-1 is a potent TMOD1 overexpression could induce some protease activities inhibitor of secreted MMPs but inefficiently inhibits membrane- that degrade type I collagen. Consistent with this idea, 231- anchored MMPs, whereas TIMP2 is an efficient inhibitor of both TMOD1 showed higher invasive activity than 231-control membrane-anchored and secreted MMPs (25, 26). Both TIMP1 (Fig. 4D). The invasive growth capacity of malignant cancer cells and TIMP2 significantly suppressed the growth rates of 231- is linked to the abilities of cells to degrade extracellular matrix TMOD1, bring it down to those of the 231-control in the 3D (ECM), and members of the MMP family display specific proteo- culture system (Fig. 4F), whereas both inhibitors had negligible lytic activities against components of ECM (22, 23), which led us effect on the growth rates of MDA-MB-231 in the 2D culture to hypothesize that some members of the MMP family might be system, irrespective of TMOD1 expression (Fig. 4E). These results induced by the TMOD1 overexpression. To examine involvement indicate that both secreted and membrane-anchored types of of the MMP family, we tested the effects of MMI270, an inhibitor MMP are involved. Given that the ECM in the 3D culture is for both secreted and membrane-anchored MMPs (24), on the composed of type I collagen, TMOD1 could induce expression growth of 231-TMOD1 and 231-control in the 2D and 3D culture of MMP members that can degrade collagen or gelatin, a partial systems. In 3D culture, MMI270 treatment resulted in a reduction hydrolytic product of collagen. in growth of both 231-control and 231-TMOD1 to similar levels (Fig. 4B), whereas MMI270 had little effect on the growth of either MMP13 induction is likely involved in the TMOD1-mediated in 2D culture (Fig. 4A). These results strongly suggest that the enhancement of 3D growth TMOD1-induced enhancement of tumor cell proliferation in the To evaluate the function of endogenous TMOD1 in TNBCs, the 3D culture system is mediated by MMP-dependent processes. To effect of TMOD1 downregulation on the 3D growth was tested. identify which members of the MMP family were involved in the The effective knockdown of TMOD1 in MDA-MB-231 cells was TMOD1-induced enhancement of 3D growth, we used recombi- achieved using two distinct TMOD1 shRNAs (Fig. 5A). Although
OF4 Cancer Res; 75(1) January 1, 2015 Cancer Research
Downloaded from cancerres.aacrjournals.org on October 1, 2021. © 2014 American Association for Cancer Research. NF-kB–Induced Tropomodulin Enhances Breast Tumor Growth
ABC MDA-MB-231 MDA-MB-231 MDA-MB-436 –1,000 –717 –607 –589 –394 –38 +1 0.4 Figure 3. *** TNFα: 0122444 01224h 0.15 0.3 TMOD1 expression is directly ** *** regulated by NF-kB. A, the amounts TMOD1 ** 0.2 of TMOD1 mRNA were measured by 0.1 18S rRNA 0.1 RT-PCR (top). The bands were 6 fi / 18S rRNA TMOD1 0 quanti ed by densitometry (bottom). 5 0.05
Expression levels of 18S rRNA were 4 (%) ChIP/input siRNA: used for normalization. B, a schematic 3
0 Control of the TMOD1 promoter highlighting 2 RELA-1 RELA-2 1 Antibody : RELA Fold change the region used in the ChIP assay 0 RELA RELA (top). The primers used (arrows) and Control Control Tubulin putative NF-kB–binding sites (closed rectangles) are indicated. The amounts of precipitated DNA relative MDA-MB-436MDA-MB-231 to the input DNA were determined D E by real-time RT-PCR (bottom). C, after RELA siRNA transfection, TMOD1 and RELA expression was measured by real-time RT-PCR and ** immunoblotting. D, the TMOD1 mRNA 8 * ** expression in each subtype is shown * 4 n.s * using a box and whisker plot, which is 6
expression ESR1 3 depicted with boxes showing the PGR 4 K8 expression median and the 25th and 75th K18 GATA3 2 percentiles, with whiskers showing ERBB2 2 FOXC1 TMOD1
K5 TMOD1 the minimal and maximal values. E, K14 1 EGFR hierarchical cluster analysis of the 0 CDH1 Basal Luminal CLDN3 published gene-expression profiles of CLDN4 0 CLDN7
Relative -like -like 251 human primary breast tumors n = 14 n = 21 OCLN Relative
Basal performed with GeneSpring software -like Ratio (log2 scale) Claudin Luminal-like ERBB2-enriched Luminal -low (left). Each cluster was classified on ERBB2 -like Basal-like Claudin-low -enriched the basis of the expression of –4.40 4.4 subtype-specific genes. The bottom colored bar, the tumor subtypes. The F 67.7% Krt14 Krt18 Esr1 TMOD1 mRNA levels among primary Luminal progenitor 1.2 breast cancers are shown (right) as in (LP) cells – + + 0.8 D. F, expression levels of K14 K18 ER Keratin 14 (Krt14), Keratin 18 (Krt18), CD61 0.4 estrogen receptor 1 (Esr1)andTMOD1 CD61 Mature luminal (ML) cells 0 mRNA were measured by real-time CD49f 66.7% LP ML B LP ML B LP ML B
– + + expression Relative RT-PCR. Those of b-actin mRNA were K14 K18 ER used for normalization. The sorting CD49f 2 procedure for various mouse mammary epithelial cell populations 1.5 (left) and confirmation of each CD24
92.4% expression 1 population (top right) were shown. Basal (B) cells
CD49f Relative