BEX2 Is Overexpressed in a Subset of Primary Breast Cancers
and Mediates Nerve Growth Factor/Nuclear Factor-KB
Inhibition of Apoptosis in Breast Cancer Cell Lines
Ali Naderi,1Andrew E. Teschendorff,
1Juergen Beigel,
1Massimiliano Cariati,
1
Ian O. Ellis,2James D. Brenton,
1and Carlos Caldas
1
1Cancer Genomics Program, Department of Oncology, University of Cambridge, Hutchison/Medical ResearchCouncil Research Center, Cambridge, United Kingdom and 2Department of Histopathology, NottinghamCity Hospital NHS Trust and University of Nottingham, Nottingham, United Kingdom
Abstract
We have identified a novel subtype of estrogen receptor (ER)-positive breast cancers with improved outcome after tamox-ifen treatment and characterized by overexpression of thegene BEX2. BEX2 and its homologue BEX1 have highlycorrelated expression and are part of a cluster enriched forER response and apoptosis genes. BEX2 expression is inducedafter estradiol (E2) treatment with a peak at 3 h, suggestingBEX2 is an estrogen-regulated gene. BEX2 belongs to a familyof genes, including BEX1, NGFRAP1 (alias BEX3), BEXL1 (aliasBEX4), and NGFRAP1L1 (alias BEX5). Both BEX1 andNGFRAP1 interact with p75NTR and modulate nerve growthfactor (NGF) signaling through nuclear factor-KB (NF-KB) toregulate cell cycle, apoptosis, and differentiation in neuraltissues. In breast cancer cells, NGF inhibits C2-inducedapoptosis through binding of p75NTR and NF-KB activation.Here, we show that BEX2 expression is necessary andsufficient for the NGF-mediated inhibition (through NF-KBactivation) of C2-induced apoptosis. We also show that BEX2modulates apoptosis of breast cancer cells in response to E2(50 nmol/L) and tamoxifen (5 and 10 Mmol/L). Furthermore,BEX2 overexpression enhances the antiproliferative effect oftamoxifen at pharmacologic dose (1 Mmol/L). These datasuggest that a NGF/BEX2/NF-KB pathway is involved inregulating apoptosis in breast cancer cells and in modulatingresponse to tamoxifen in primary tumors. [Cancer Res2007;67(14):6725–36]
Introduction
The heterogeneity of breast cancer poses a significant challengein the diagnosis, prognostication, and treatment of the disease.Indeed, using expression analysis, breast cancers can be sub-classified into at least six subtypes: luminal-like subgroups A, B,and C; basal-like; normal breast-like; and HER-2–like (1–3).Different subclasses of estrogen receptor–positive (ER+) tumors
might reflect different cells of origin (4), and importantly, iden-tifying different subgroups of ER+ tumors may have clinicalimplications: within tamoxifen-treated tumors, those that areER+/progesterone receptor–negative (PR�) and express HER-1 andHER-2 have worse clinical outcomes than ER+/PR+ tumors (5).The heterogeneity of ER+ breast cancer is the result of complex
interactions between estrogen response and growth factor receptorsignaling pathways, including the insulin-like growth factor andepidermal growth factor families (6). The nerve growth factor(NGF) pathway has also been implicated in the survival andproliferation of breast cancer cells (7). NGF treatment of breastcancer cells activates nuclear factor-nB (NF-nB), resulting inapoptosis inhibition after treatment with a ceramide homologue,C2 (8, 9). In breast cancer cells, the effects of NGF on apoptosis andproliferation are mediated through different receptors: p75NTRand p140TrkA, respectively (8, 9).In this study, using expression microarray analysis of 135
primary breast tumors, we identified a subset of ER+ breast cancerwith overexpression of BEX2 and BEX1. BEX1 was first identified inblastocytes using differential display analysis, and subsequentdatabase homology searches identified other family members[BEX2, NGFRAP1 (BEX3), BEXL1 , and NGFRAP1L1], all mapping toXq22.1-23 (10, 11). NGFRAP1 (BEX3) encodes NADE, whichinteracts with the death domain of p75NTR and mediatesapoptosis in neural cells in response to NGF (12). BEX1 alsoencodes a small adaptor-like protein that interacts with p75NTRand inhibits NF-nB activation in PC12 cells to regulate cell cyclearrest (13). BEX1 and BEX2 have been reported to be silenced bypromoter methylation in malignant gliomas (14). We thereforeinvestigated whether BEX2 and BEX1 could also be involved inNGF signaling in breast cancer cells.
Materials and Methods
Expression Microarray ExperimentsThe samples used were from a cohort of 135 frozen breast tumors
collected at Nottingham City Hospital NHS Trust between 1986 and 1992.
The institution research ethics committee approved this study. Total RNAextraction, RNA amplification, and indirect labeling were carried out as
described before (15–17). Oligonucleotide microarrays containing 22,575
features were used (Agilent Human 1A 60-mer Oligo Microarray, Agilent
Technologies). Hybridization, scanning, feature extraction, and datanormalization were done as previously published (15).
Microarray Data AnalysisA total of 307 slides was compared for correlations between dye reversal
pairs using Spotfire DecisionSite 8.0, and for each biological specimen, only
those genes with a positive correlation were selected for further analysis. We
further removed those genes with >20% missing data across all samples and
Note: Supplementary data for this article are available at Cancer Research Online(http://cancerres.aacrjournals.org/).
Current address for A.E. Teschendorff, J. Beigel, J.D. Brenton, and C. Caldas: CancerResearch UK Cambridge Research Institute, Li Ka Shing Centre, Robinson Way,Cambridge CB2 0RE, United Kingdom.
Requests for reprints: Carlos Caldas or Ali Naderi, Cancer Genomics Program,Department of Oncology, University of Cambridge, Hutchison/MRC Research Center,Hills Road, Cambridge CB2 2XZ, United Kingdom. Phone: 44-1223-404420; Fax: 44-1223-331753; E-mail: [email protected] or [email protected].
I2007 American Association for Cancer Research.doi:10.1158/0008-5472.CAN-06-4394
www.aacrjournals.org 6725 Cancer Res 2007; 67: (14). July 15, 2007
this resulted in a matrix of 1,203 genes. Missing data points were imputedby the k-means nearest neighbor method as described before (18).
Microarray data are deposited in ArrayExpress database (accession number
E-UCON-1).
Independent component analysis (ICA; ref. 19) was implemented usingthe R package mlica.3 Supervised analysis by t test/ANOVA and similarity
ranking were done by Spotfire DecisionSite 8.0. Significance analysis of
microarrays (SAM)4 and prediction analysis of microarrays (PAM)5 were
done following the instructions of the software. The significance analysis ofPearson’s correlations using Monte-Carlo simulations was done applying
the R statistical language.6 Analysis of the BEX expression signature
function was done using Ingenuity Pathways Analysis software (Ingenuity
Systems). Biostatistical analysis was done with Statistical Package for theSocial Sciences version 12.0.1 (SPSS). Single sample prediction (SSP)
classifier was derived as described before (20).
Real-Time PCR Analysis on TumorsReverse transcription-PCR (RT-PCR) to assess the expression levels of
BEX1 and BEX2 on tumor samples was done using gene-specific Taqman
assays (Applied Biosystems). Housekeeping genes HPRT1 and RPLP0 were
used as controls. Experimental procedures were done following themanufacturer’s instruction (Applied Biosystems). Relative gene expressions
were calculated as described before (21).
Analysis of Apoptosis with C2 and NGF TreatmentsCell culture and apoptotic assays with MCF-7 and MDA-MB-231 cells
were done as described in refs. 8, 9. Briefly, cells were grown and treated in
the following groups (in triplicate): (a) control group (C2�/NGF�), notreatment; (b) C2�/NGF+, treatment with 200 ng/mL h-NGF (R&D Systems);(c) C2+/NGF�, treatment with ceramide analogue C2 (Sigma) at 20 Amol/L;and (d) C2+/NGF+, treatments with both h-NGF and C2. Apoptosis was
scored after staining with Hoechst 33258 (Sigma) as described in refs. 8, 9.
RT-PCR Analysis of BEX Expression in Cell Lines withDifferent TreatmentsExpression levels for BEX1 and BEX2 genes were measured using gene-
specific Taqman assays. BEX relative expression is equal to BEX expression
in the treated group/average BEX expression in the control group.
NGF/C2 treatments. The MCF-7 cell line was treated with C2 and NGFin four groups as described above. After 24 h of incubation, cells wereharvested. Experiments were carried out in four replicates.
Estradiol treatment. MCF-7 cells grown in medium containing phenol
red–free DMEM (Invitrogen) and 10% charcoal/dextran-treated serum
(HyClone) were incubated with 17-h-estradiol (Sigma) at 1 nmol/Lconcentration. Cells were harvested at 0, 1, 3, 6, and 12 h for RT-PCR
measurement of BEX1 and BEX2 expression. Four replicate experiments
were carried out for each treatment condition.NF-KB inhibition. MCF-7 cells were serum deprived for 24 h and then
treated separately with 18 Amol/L SN50 NF-nB inhibitor peptide (Tebu-Bio)in the presence or absence of C2/NGF. Untreated MCF-7 cells and
treatment with C2/NGF were used as controls. After another 6, 12, and18 h of incubation, cells were harvested.
Experiments with BEX2-Transfected CellsBEX2 transfection. The BEX2 construct in pDream2.1/LIC expression
vector (cytomegalovirus promoter and a Flag tag) was obtained fromGenScript Corp. Transfection of MCF-7 and MDA-MB-231 cells was carried
out in four replicates using ExGen 500 reagent (Fermentas Life Sciences) as
instructed by the manufacturer. Cells were cotransfected with 1.2 Ag of
green fluorescent protein (GFP)-carrying vector with either 1.6 Ag of theempty or BEX2-carrying pDream2.1/LIC vector (BEX2+). RT-PCR and
Western blotting with anti-FLAG antibody (Sigma) at 1:500 dilution ofprimary antibody were used to confirm overexpression.
C2/NGF experiments. Cells were grown for 24 h and then incubated
with C2 and NGF for the apoptotic assays described above.
Tamoxifen treatment.MCF-7 cells were grown on coverslips. Tamoxifen(Sigma) treatment was carried out at 1, 5, and 10 Amol/L concentrations inserum-free medium for 24 h and untreated cells were used as controls. To
assess the effect of BEX2 overexpression, either BEX2+ or empty vector was
transfected in four replicates followed by tamoxifen treatment at 5 and10 Amol/L. Apoptosis was measured using Hoechst and Annexin V-FITC
staining. Annexin V-FITC assay was done using Annexin V-FITC
fluorescence microscopy kit (BD Biosciences) following the manufacturer’s
instructions.
Experiments with BEX2 Gene Knockdown CellsGeneration of BEX2 knockdown cells. BEX2 knockdown (BEX2-KD) in
MCF-7 and MDA-MB-231 cell lines was carried using SMARTpool smallinterfering RNA (siRNA; four oligonucleotides) reagents following the
manufacturer’s instructions (Dharmacon, Inc.). Cells transfected with
siCONTROL Non-Targeting siRNA (Dharmacon) and grown under the same
conditions were used as controls. All siRNA silencing experiments weredone in four replicates.
C2/NGF experiments. After transfection, MCF-7 and MDA-MB-231 cells(knockdown and controls) were incubated at 37jC for 24 h followed by
treatments with C2 and NGF as above.Estradiol treatment. MCF-7 cells were treated in the following four
groups for 48 h: (a) serum starvation, (b) serum starvation and 17-h-estradiol at 50 nmol/L, (c) serum starvation in BEX2-KD cells, and (d)serum starvation in BEX2-silenced cells and 17-h-estradiol at 50 nmol/L.After 48 h of incubation, cells were harvested, fixed in 4% formaldehyde, and
then stained with Hoechst to score apoptosis (percentage).
Tamoxifen treatment. MCF-7 cells (controls and knockdown) weretreated with NGF and tamoxifen (5 and 10 Amol/L). Apoptosis was assessedafter 24 h using Annexin V-FITC assay as described above.
Assessment of Proliferation in Tamoxifen-Treated CellsMCF-7 and MDA-MB-231 cells were grown in 96-well plates to 50%
confluence. Both cell lines were then transfected and treated in the
following groups: (a) siCONTROL transfection followed by NGF and (b)
BEX2-KD followed by NGF. MCF-7 cells were further studied in the
following groups: (a) empty vector transfection followed by tamoxifen at1 Amol/L and (b) BEX2 transfection followed by tamoxifen at 1 Amol/L.siCONTROL or empty vector–transfected cells without any treatment were
used as controls. Proliferation was measured 24 and 72 h after thetreatments using Vybrant 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazo-
lium bromide (MTT) proliferation assay kit (Invitrogen) following the
manufacturer’s instructions. All experiments were done in eight biological
replicates.
NF-KB Activity Assaysp50 NF-KB. MCF-7 cells were cultured in the following five groups: (a)
in eight biological replicates following the manufacturer’s instructions. Foreach experimental group, the ratio of phosphorylated p65/total p65 was
obtained and relative ratios were calculated as treatment group/control
group.
Results
BEX1 and BEX2 are classifiers of ER+ breast cancer.Expression microarray analysis revealed that BEX2 was the genewith the highest frequency of significant log2 ratios (P < 0.05)across the cohort: >98% of the samples showed differentialexpression relative to the common reference pool (SupplementaryFig. S1). This indicated that the expression of BEX2 variedsignificantly across most samples. Cancer outlier profile analysis(COPA), a methodology recently proposed to identify biologicallyrelevant genes (22), showed that BEX2 and its close homologue,BEX1 , ranked among the top genes (Fig. 1A). The mutual presenceof these two genes in COPA analysis was striking and led us tofurther investigate their expression profiles.BEX2 expression relative to the common reference pool
separated the breast tumors in two significantly different (P < 1� 10�5; Fig. 1B) groups with overexpression (n = 20) and under-expression (n = 115) of the gene. BEX1 expression also variedsignificantly between the two groups (P < 1 � 10�5), reflectingthe strong correlation (Pearson’s correlation coefficient = 0.94)in expression patterns of both genes (Fig. 1C). Most interestingly,all samples with BEX overexpression were ER+ (P < 0.001;Fig. 1C).To validate the expression microarray results, RT-PCRwas used to
confirm the expression levels of BEX1 and BEX2 in a subset ofsamples from the original cohort. The correlation coefficientsbetween RT-PCR results and expression microarray ratios were 0.87(BEX2 ; n = 40) and 0.6 (BEX1 ; n = 40; P < 0.001; SupplementaryFig. S2). BEX1 expression was very low in BEX� samples, explainingthe wide difference in log ratios between BEX+ and BEX� groups inthe microarray data.BEX1 and BEX2 genes are associated with an estrogen
response expression mode. To gain insight into the possiblebiological role and function of the BEX� genes, we applied ICA.ICA decomposes the expression matrix into a set of expressionmodes, which have been shown to provide better representationsof underlying biological processes or pathways than otherdimensional reduction methods, such as principal componentanalysis (PCA; refs. 19, 23). The explanation for this is that ICAinfers expression modes that are as mutually statisticallyindependent as possible. As such modes provide closer approx-imations to the underlying biologically relevant processes thatgave rise to the data in the first place. In other words, ICAapproximates the expression matrix as a linear superposition ofindependent biological processes that are inferred using thecriterion of statistical independence through a nonlinear decorre-lation step. In contrast, PCA only does a linear decorrelation andso does not decompose the expression matrix into theindependent biological processes that generated the data. OurICA analysis identified an expression mode with strong represen-tation of BEX1 and BEX2 (Fig. 1D ; Supplementary Table S1).BEX1 showed the highest weight in this mode, and BEX2 wasrepresented as one of the top 20 genes. Known estrogen responsegenes, such as GATA3, TFF1, LIV-1 , and AREG (24, 25), as well asgenes previously shown to be expressed in breast cancers, such asLACRT (26), DCD (27), and BMP7 (28), were also highly activated.Not surprisingly, using the SSP classifier (20), 60% (n = 12) of BEX
overexpressers were luminal subtype A, 30% (n = 6) were luminalsubtype B, and 10% (n = 2) were normal subtype.Projection of the sample expression values along the BEX mode
showed significant differential activation between ER+ and ER�
samples (P = 6.4 � 10�9; Supplementary Fig. S3). Thus, weconcluded that this ICA mode and the genes highly activated inthis mode, including the BEX genes, are associated with ER status.We verified that the BEX ICA mode and its association withestrogen response were robust under repeated runs of thealgorithm as well as under variations in the number of inferredmodes (data not shown). Analysis of publicly available data setsalso showed a majority of the genes in this mode to be differentiallyexpressed in ER+ versus ER� breast cancers (data not shown).7
A distinct BEX+ cluster of genes in breast cancer. Havingestablished the significance of the expression of BEX genes inbreast cancer using both COPA and ICA, we asked whether therewere genes significantly correlating with BEX1 and BEX2 (hereafterdesignated ‘‘BEX+ cluster’’). To investigate this, we used Pearson’scorrelation coefficients and tested significance of the coefficientsby comparison with null distributions obtained using Monte-Carlosimulation. The gene correlations were computed to the averageexpression profile of BEX1 and BEX2 because both genes werehighly correlated (>0.9). The correlation analysis and Monte-Carlosimulation identified 35 genes (Fig. 2A) with significant correla-tions to the expression of BEX1 and BEX2 (BEX+ cluster), withabsolute correlation values ranging from approximately 0.25 to 0.4(P < 0.0025, q < 0.06). There were nine of these genes also present inthe BEX ICA mode. To validate the BEX+ cluster, a similar analysiswas done in an independent data set (29). BEX1 , together with 23other genes from the cluster, was present in the external data set.Sixteen of the overlapping genes (f70%) had significant correla-tions with BEX1 (P < 0.04; Fig. 2B) and showed the same directionof expression changes. Considering the differences in methodologyand variation in the platform used, this overlap supports theexistence of a distinct BEX+ cluster.Supervised analysis defines a developmental/apoptosis/cell
proliferation BEX expression signature in breast cancer. Toidentify genes that could differentiate between tumors expressinghigh levels of BEX genes (BEX+ ; n = 20) from the remainder oftumors (BEX� ; n = 115), the samples were divided in two groupsbased on the expression ratios of BEX2 (BEX+ log2 > 0; BEX�
log2 < 0). This subdivision was validated using an unsupervisedk-means algorithm (k = 2; data not shown).The first supervised analysis used a class prediction algorithm
based on the nearest shrunken centroid method5 (30). Thisalgorithm generated an optimal classifier consisting of 37 geneswith a correct classification rate of 85% (Fig. 2C ; SupplementaryFig. S4). Notably, 60% of the genes from the BEX+ cluster were alsomembers of the optimal classifier.To validate these results, further supervised analysis using
t test/ANOVA and SAM were conducted (Supplementary ExcelFiles S1 and S2). The results showed that approximately 2/3 of thegenes overlap among all supervised methods and with the BEX+
cluster.The expression patterns in comparison with BEX2 of the 68
genes from the ANOVA-derived signature were analyzed usingsimilarity ranking of correlation coefficients (Supplementary Fig. S5;
7 http://www.oncomine.org/main/index.jsp
BEX2 Is Overexpressed in a Subset of Breast Cancers
www.aacrjournals.org 6727 Cancer Res 2007; 67: (14). July 15, 2007
Supplementary Excel File S3) within the expression matrix of 1,203genes filtered for analysis (see Materials and Methods). BEX1ranked closest to BEX2 with 0.9 similarity and the other 31 withpositive correlation ranked ordered 3 to 33 of 1,203. Likewise, the36 genes with negative correlation with BEX genes ranked the mostdistant (e.g., C10orf81 ranked 1,203 and TNFSF7 ranked 1,201 with�0.28 and �0.27 similarities, respectively).These findings show that there is a distinct expression signature
differentiating BEX+ and BEX� samples, which is reproducibleusing different supervised methods.To gain further insights into the potential functional significance
of the BEX expression signature, the Ingenuity Pathways Analysissoftware was used. The stability of results was tested by analyzingsignatures obtained using SAM, ANOVA, and PAM. Significantassociations (P < 0.001) were detected between the BEX signatureand development, apoptosis, and cell proliferation functions(Supplementary Table S2).BEX2 expression in MCF-7 breast cancer cells increases with
estrogen treatment. Given that BEX1 and BEX2 expression aresignificantly higher in a subset of ER+ breast cancers, weinvestigated the effect of estrogen on the expression of both genesin the ER+ cell line MCF-7. MCF-7 cells were treated with estradiol(E2) at 1 nmol/L over 12 h. Cells were harvested at 1, 3, 6, and12 h time points, and BEX1 and BEX2 expression ratios relative tobaseline untreated cells were measured using RT-PCR (Fig. 2D).BEX2 expression was induced at 60 min, peaked at 3 h (4.1 F 0.2),and returned to baseline by 12 h. In contrast, E2 did not have anyeffect on the expression of BEX1 , which was not expressed atbaseline in MCF-7 cells (data not shown). These findings indicatethat BEX2 expression increases in response to E2 treatment in thebreast cancer cell line MCF-7.BEX2 expression in MCF-7 breast cancer cells increases with
C2-induced apoptosis and with NGF treatment. The expressiondata suggested a potential link between BEX2 and BEX1 with ERresponse and apoptosis. We therefore asked whether the expressionof BEX genes in the MCF-7 ER+ breast cancer cell line is modulatedin response to C2 and NGF using RT-PCR (Fig. 3A). MCF-7 cellswere previously shown to express p75NTR (9).Apoptosis in MCF-7 cells were analyzed in four groups as
described in Materials and Methods. We confirmed the previouslypublished results: no significant change compared with controlsin the C2�/NGF+ group, gross apoptosis in the C2+/NGF� group,and rescue from apoptosis in the C2+/NGF+ group (Supplemen-tary Fig. S6).The BEX2 expression ratios relative to control cells were as
follows: (a) 2.62F 0.44 in response to NGF, (b) 4.4F 0.80 in responseto C2, and (c) 8.9 F 1.61 in response to C2 + NGF treatments(significantly higher than other groups; P < 0.001; Fig. 3A). BEX1expression was not detected in MCF-7 either at baseline or inresponse to the treatments (data not shown).These data show that, in MCF-7 cells, BEX2 expression increases
slightly with NGF stimulation, increases significantly with C2treatment (which induces apoptosis), and increases further withrescue from C2-induced apoptosis by NGF treatment. In contrast,BEX1 was not expressed either at baseline or with any of thetreatments used. We therefore manipulated BEX2 expression levelsin breast cancer cells and analyzed its effects on apoptosis.BEX2 overexpression rescues breast cancer cells from C2-
induced apoptosis mimicking NGF treatment. We askedwhether BEX2 overexpression could rescue breast cancer cells(MCF-7 and MDA-MB-231) from C2-induced apoptosis in the
absence of NGF. Despite being ER�, MDA-MB-231 is also rescuedfrom C2-induced apoptosis by NGF (9).Both cell lines were cotransfected with a BEX2-FLAG expression
vector and a GFP expression vector to assess transfectionefficiency. As a negative control, similar experiments usedcotransfection with an empty vector and GFP. Overexpression ofBEX2 in transfected cells was confirmed using RT-PCR andWestern blot with anti-FLAG antibody because none of thecommercially available BEX2 antibodies worked in our hands(Fig. 3B). BEX2 overexpression significantly reduced the C2-induced apoptosis in both cell lines (P = 0.0001; Fig. 3C ;Supplementary Fig. S7). These data show that BEX2 overexpressionrescues cells from C2-induced apoptosis in two different breastcancer cell lines and is sufficient to produce an antiapoptotic effectsimilar to that observed with NGF treatment.Knockdown of BEX2 inhibits the NGF antiapoptotic re-
sponse in breast cancer cells. We next asked whether theexpression of BEX2 is necessary for the NGF antiapoptotic effect.siRNA was used to knock down BEX2 in MCF-7 and MDA-MB-231cell lines, and in parallel experiments, nontargeted siRNA (controls)was used as a control. RT-PCR confirmed that BEX2 transcriptlevels were suppressed by f90% in BEX2 siRNA-treated cellscompared with controls (Fig. 3B). BEX2-KD significantly inhibitedthe antiapoptotic effect of NGF after C2 treatment (P = 0.001) inboth cell lines (Fig. 3D) compared with controls. These resultsshow that BEX2 expression is required for NGF-mediatedantiapoptotic response in two different breast cancer cell lines.BEX2 modulates the apoptotic responses of MCF-7 cells to
estrogen and tamoxifen. Ceramide is a physiologic mediator ofprogrammed cell death. Although C2 ‘‘mimics’’ the proapoptoticactivity of cytokines known to interact with the ER response, thedirect relevance of the ‘‘NGF/BEX2 pathway’’ to apoptosis inresponse to estrogen modulation was not addressed in theexperiments described above. Indeed, tamoxifen treatment inaddition to its antiproliferative effects induces apoptosis in breastcancer cells (31), and E2 has an antiapoptotic activity in breastcancer cell lines in response to various stressors (32, 33). Wetherefore treated MCF-7 cells with tamoxifen or E2, using dosespreviously shown to modulate apoptosis, and compared responseswith and without manipulation of BEX2 expression.Treatment of MCF-7 cells with tamoxifen at 5 and 10 Amol/L
resulted in significant apoptosis (P < 0.01), which was not seen at1 Amol/L (Fig. 4A and B ; Supplementary Fig. S8). Apoptosis wassignificantly reduced by both NGF treatment and BEX2 over-expression (P < 0.03; Fig. 4A and B). Treatment with the NF-nBinhibitor SN50 and BEX2-KD removed the protective effects ofBEX2 overexpression and NGF, respectively (P < 0.01; Fig. 4A andB). These data indicate that activation of the NGF/BEX2 pathwayinhibits tamoxifen-induced apoptosis and this effect of BEX2 ismediated through NF-nB.MCF-7 cells subjected to serum starvation (48 h) and treated
with E2 (50 nmol/L) showed a significant reduction of apoptosiscompared with untreated cells (P = 0.01; Fig. 4C). In contrast, inBEX2-KD MCF-7 cells, E2 was not as effective as an antiapoptoticagent (Fig. 4C). These results show that BEX2 is necessary for theE2 antiapoptotic activity in serum-starved MCF-7 cells.BEX2 modulation of cell proliferation. Having shown that
BEX2 plays a role in modulating apoptosis in breast cancer cells inresponse to NGF and tamoxifen, we next analyzed whether it couldalso play a role in modulating cell proliferation using the MTTassay.
Cancer Research
Cancer Res 2007; 67: (14). July 15, 2007 6728 www.aacrjournals.org
Figure 1. Expression pattern of BEX genes in primary breast cancers and ICA of breast cancer expression data reveals a BEX mode. A, results of COPA showing top10 ranked genes. Range of COPA scores for the top 10 genes in each centile and scores of BEX genes. B, correlation of BEX2 log ratios between dye reversalexperiments [sample/reference (LogS/R ) and reference/sample (LogR/S ) are log10 based]. P value is calculated using t test/ANOVA. C, heat map showing theexpression of BEX1 and BEX2 across the 135 samples (log10 ratio scales: �1.5 to 1.5). Columns, ER status of the samples. D, graphic showing relative weights ofgenes in BEX ICA mode. Bold italic, BEX1, BEX2 , and known estrogen response genes.
BEX2 Is Overexpressed in a Subset of Breast Cancers
www.aacrjournals.org 6729 Cancer Res 2007; 67: (14). July 15, 2007
Figure 2. BEX cluster and expression signature in breast cancer. A, bar plot of BEX cluster 35 genes identified by correlation analysis. Each column identifies a geneand the direction of the column and its height reflect the nature of the correlation (P and q values of Pearson’s correlation and Monte-Carlo simulation). B, barplot of overlapping BEX cluster genes with significant correlations to BEX1 (P < 0.04) in van ’t Veer et al.’s study. C, centroid of BEX expression signature by PAM.Weights for the up-regulated and down-regulated genes. Red, genes that overlap with ICA BEX mode. D, graphic of BEX2 relative expression ratios in responseto E2 (at 1 nmol/L) in MCF-7 cell line. *, P < 0.01, compared with untreated cells using Mann-Whitney U test.
Cancer Research
Cancer Res 2007; 67: (14). July 15, 2007 6730 www.aacrjournals.org
As previously described (34), NGF significantly (P < 0.01)increased cell proliferation after 1 and 3 days of treatment. Thiswas not significantly altered by BEX2-KD (P > 0.3; Fig. 4D). Inaddition, BEX2-KD and BEX2 overexpression had no measura-ble effect in the proliferation of these two breast cancer cell lines(P > 0.3; data not shown). These data indicate that BEX2 is notinvolved in NGF-mediated proliferation.MCF-7 cells treated with tamoxifen at 1 Amol/L (pharmacologic
dose) for 1 or 3 days showed significantly impaired proliferation(Fig. 4D). BEX2 overexpression further reduced proliferation rate incells tamoxifen treated for 24 h (P = 0.01; Fig. 4D) but not in cellstamoxifen treated for 72 h (P > 0.3). These data indicate that BEX2
modulates the antiproliferative activity of tamoxifen in the MCF-7breast cancer cell line but this effect seems to depend on theduration of tamoxifen exposure.BEX2 mediates NF-nB activation in the NGF antiapoptotic
pathway. The antiapoptotic effect of NGF in breast cancer celllines treated with C2 is mediated through the activation of NF-nB(9). To determine whether BEX2 is directly involved in this pathwayin MCF-7 cells, we measured the binding of p50 NF-nB to DNA andthe phosphorylation of p65 NF-nB using ELISA assays. NGFtreatment had no effect on p50 NF-nB binding (Fig. 5A) andincreased ratio of phosphorylated p65 NF-nB/total p65 NF-nB(P < 0.01; Fig. 5B). BEX2 overexpression significantly increased both
Figure 3. BEX2 and regulation of apoptosis. A, BEX2 relative expression ratios by RT-PCR in MCF-7 cell line. Columns, BEX2 expression ratios. C2�/NGF+,NGF treatment alone; C2+/NGF�, C2 treatment alone; C2+/NGF+, C2 and NGF treatments. *, P < 0.001, compared with untreated cells using Mann-Whitney U test.B, RT-PCR and Western blot to confirm BEX2 expression in MCF-7 and MDA-MB-231 cells. For RT-PCR experiments, transfection with BEX2 vector (left ) andknockdown using siRNA (right ) are shown. Four replicate experiments were done for each condition and �DDCT measurements are shown for each set of experiments.Western blot was done using BEX2 -transfected MCF-7 (MCF-BEX2 ) and MDA-MB-231 (MDA-BEX2 ) cells. Empty vector control cells (MDA-EV and MCF-EV ).Anti-FLAG antibody (Sigma) at 1:500 dilution was used (BEX2 construct carries a FLAG tag at the 5¶-end). C, apoptosis in MCF-7 and MDA-MB-231 cellsoverexpressing BEX2. Columns, percentage of apoptosis. EV, cells transfected with empty vector; BEX2+, cells transfected with BEX2 vector; MDA, MDA-MB-231.D, apoptosis in MCF-7 and MDA-MB-231 cells with BEX2 -KD. CT, control siRNA.
BEX2 Is Overexpressed in a Subset of Breast Cancers
www.aacrjournals.org 6731 Cancer Res 2007; 67: (14). July 15, 2007
p50 NF-nB DNA binding and ratio of phosphorylated p65 NF-nB/total p65 NF-nB (P = 0.01; Fig. 5A and B). Treatment with C2/NGFalso significantly increased p50 NF-nB DNA binding (P < 0.01;Fig. 5A), but BEX2-KD, which reverts apoptosis rescue (see Fig. 3D),
dramatically reduced p50 NF-nB binding (1 F 0.18; P = 0.01;Fig. 5A) after C2/NGF treatment. The rise in the ratio ofphosphorylated p65 NF-nB/total p65 NF-nB after NGF treatmentis also significantly reduced in BEX2-KD cells (P = 0.03; Fig. 5B).
Figure 4. BEX2 modulation of tamoxifen/E2 induced apoptosis and tamoxifen antiproliferative effect. A, apoptosis of MCF-7 in response to tamoxifen using Hoechststaining. TAM, tamoxifen at 10 Amol/L for 24 h; NGF, at 200 ng/mL for 24 h; BEX2 (+), BEX2 transfection. *, P < 0.01 for NGF and TAM versus TAM, BEX2transfection and TAM versus empty vector and TAM, and BEX2 transfection and TAM + SN50 versus empty vector and TAM. Columns, mean; bars, SE. B, apoptosisof MCF-7 in response to tamoxifen using Annexin V-FITC assay. TAM, tamoxifen at 0, 1, 5, and 10 Amol/L concentrations for 24 h; NGF, at 200 ng/mL. *, P < 0.03 forTAM versus TAM-5 and BEX2 transfection; **, P < 0.01 for all other experimental conditions. All P values are calculated using Mann-Whitney U test. C, apoptosisof MCF-7 in response to serum starvation and E2. Percentage of apoptotic MCF-7 cells is shown in each treatment group. SS, serum starvation for 48 h; E2, at50 nmol/L; *, P = 0.01 for E2 treatment versus no treatment. D, effect of BEX2 on proliferation changes mediated by NGF and tamoxifen using MTT assay.BEX2 transfection, BEX2 overexpression; TAM, tamoxifen at 1 Amol/L; MCF, MCF-7. Incubation was done for 1 and 3 d. DOD measured as absorbancedifference at 520 nm between the treatment groups and untreated controls. *, P value is for TAM versus TAM/BEX2 overexpression after 1-d incubation usingMann-Whitney U test.
Cancer Research
Cancer Res 2007; 67: (14). July 15, 2007 6732 www.aacrjournals.org
These data show that BEX2 expression is both necessary andsufficient for NGF-mediated NF-nB activation.Notably, total p65 protein level by ELISA (and REL-A/p65 gene
by RT-PCR) did not significantly change after either NGF treat-ment or BEX2 overexpression (data not shown). This indicatesthat the effect of NGF/BEX2 on p65 NF-nB is mediated throughan increase in phosphorylation, not a change of protein/geneexpression.BEX2 expression is increased by NF-KB inactivation. The
ELISA experiments showed that NF-nB activation in C2/NGFtreated cells is extremely sensitive to and dependent on BEX2expression. This suggested that modulation of BEX2 expression isa regulator of the pathway. We therefore analyzed whether BEX2expression levels change when NF-nB is prevented from relocatingto the nucleus despite stimulation of the pathway. MCF-7 cellswere treated with SN50, a cell-permeable polypeptide that inhibitstranslocation of the NF-nB active complex into the nucleus, andthis inhibited the antiapoptotic effect of NGF in C2-exposed cells(data not shown). SN50 treatment alone did not change BEX2expression (Fig. 5C). Overnight treatment with C2/NGF/SN50 for18 h led to a 21-fold F 3.3 increase (P < 0.001; Fig. 5C) in BEX2expression. Similar results were obtained after 6 and 12 h of SN50treatment: relative expressions of 18 F 2.5–fold and 20 F 3–fold,respectively (data not shown). This shows that, upon simultaneousactivation of the pathway with NGF/C2 treatment and inhibitionof NF-nB DNA binding using SN50, there is a significant furtherincrease of BEX2 expression compared with activation of thepathway alone.Increased BEX2 expression was not observed with other
proapoptotic stimuli, such as tamoxifen treatment (10 Amol/L) orserum starvation (data not shown), indicating that the effect ofNGF/NF-nB modulation (e.g., with C2 and SN50) on BEX2expression is not the result of a nonspecific apoptotic response.BEX2+/ER+ breast cancers have a better response to
tamoxifen therapy. Having established that in our cohort allBEX+ cases were ER+ breast cancers, we asked whether there wasany influence of BEX2 status on outcome among patients treatedwith tamoxifen therapy. We first divided the cases (n = 34) intoBEX2 overexpression or BEX2 underexpression using one of twocriteria: (a) log2 ratios normalized to the median value on themicroarrays, showing at least 2-fold expression difference (n = 24),and (b) RT-PCR of samples with <2-fold expression difference bymicroarray analysis (n = 10), showing DCT < �1 SD or DCT > +1 SDof the mean (n = 6). Using these criteria, a total of 30 cases wasavailable for survival analysis and this showed a significantly better(P = 0.03; Fig. 6A ; Supplementary Table S3) disease-free interval in
Figure 5. BEX2 and NF-nB activation. A, measurement of p50 NF-nBDNA binding in MCF-7 cells using ELISA. Fold p50 NF-nB DNA binding iscompared with untreated cells (C2/NGF and silencing experiments) or emptyvector–transfected cells (transfection experiment). C2, 10 Amol/L; NGF,200 ng/mL. *, P = 0.01 for C2/NGF versus BEX2 -KD and C2/NGF and forBEX2 overexpression versus empty vector transfection. B, measurement ofphosphorylated p65 NF-nB in MCF-7 cells using ELISA. Relative ratio forphosphorylated p65 NF-nB/total p65 NF-nB is compared with untreated cells(NGF and silencing experiments) or empty vector–transfected cells (transfectionexperiment). *, P = 0.03 for NGF versus BEX2 -KD and NGF; **, P = 0.01for BEX2 overexpression versus empty vector transfection. C, BEX2 foldexpression by RT-PCR in MCF-7 cells treated with C2+/NGF+ and inhibitionof NF-nB with SN50 treatment. *, P < 0.001 for C2+/NGF+/SN50+ versusC2+/NGF+. Expression ratios are relative to the average BEX2 expression inuntreated MCF-7 cells. All P values are calculated using Mann-Whitney U test.Columns, mean; bars, SE.
BEX2 Is Overexpressed in a Subset of Breast Cancers
www.aacrjournals.org 6733 Cancer Res 2007; 67: (14). July 15, 2007
patients with BEX2 overexpression [n = 16; 95% confidence interval(95% CI), 121–161 months] versus those with BEX2 under-expression (n = 14; 95% CI, 63–123 months).
Discussion
The evidence for the involvement of BEX1 and BEX2 genes inhuman cancer has emerged with the report that both genes aresilenced by promoter methylation in human gliomas (14). BEXL1 ,another BEX family member, is epigenetically silenced in ovariancancer (35). The data presented here reveal overexpression of BEX1and BEX2 in a subset of ER+ breast cancers. We could not detectany correlation in our breast cancers between BEX2 methylationand expression (data not shown). We also reviewed arraycomparative genomic hybridization data available for the tumorswith BEX overexpression and found no alterations at the genomicDNA level (data not shown). BEX2 and BEX1 were also highlightedby outlier analysis (COPA), which recently was used to identifysignificant cancer genes (22). We noted that BEX1 was alsosignificantly overexpressed in ER+ versus ER� tumors in twopublicly available data sets, and interestingly, in both data setsBEX1 ranks among the top genes in COPA analysis7 (36, 37). ICA, acompletely unsupervised algorithm, identified a mode thatincluded BEX1 and BEX2 . This mode correlated strongly withER+ status (P = 6.4 � 10�9) and included several well-knownestrogen response genes. The correlation of BEX1 and BEX2 withestrogen response genes was further supported using supervisedanalysis. One of these genes, GATA3 , is a transcription factor withknown correlation to ER expression (24). Other estrogen responsegenes included LIV-1, TFF1 , and AREG (24, 38, 39). Interestingly,several BEX-associated genes (GATA3, LIV-1, TFF1, SCUBE2 , andKIAA0882) are present in the luminal subtype A cluster, which isassociated with estrogen response (2), and indeed, 60% of BEXoverexpressers were luminal A. The pattern of BEX2 increasedexpression in response to E2 treatment (Fig. 2D) is compatible withthat of other estrogen early up-regulated genes (40).BEX2 interacts with LMO2 and this interaction may regulate the
transcriptional activity of LMO2 through its binding to NSCL2 (41).
NGFRAP1 encodes NADE, which interacts with p75NTR (12, 42–44).More recently, BEX1 was also shown to interact with p75NTR andthe data from this group suggest that NGF uses BEX1 to regulatecell cycle arrest (13). The results from these published experimentsshowing that BEX1 and NGFRAP1 encode small adaptor proteinsthat interact with p75NTR constitute a link between these genes,NGF signaling, and regulation of apoptosis and differentiation inneural cells. This provided us with a clue about a possiblemechanism for involvement of the BEX gene family in breastcancer in light of the published data showing that NGF inhibits theapoptotic response to C2 in cell lines through p75NTR and NF-nB.C2 is a ceramide analogue that induces massive apoptosis of breastcancer cells (45, 46). For these studies, we analyzed BEX2 becauseBEX1 is not expressed in the breast cancer cell lines used. Ourexperiments show that BEX2 is necessary for the antiapoptoticfunction of NGF in C2-treated breast cancer cells (Fig. 3D) andBEX2 overexpression produces an antiapoptotic effect similar tothat observed with NGF treatment (Fig. 3C). Interestingly, inneuronal cells, ceramide is generated as a second messenger ofsphingomyelin cycle activation by NGF, downstream of p75NTR(47). This probably explains the moderate rise in BEX2 expressionwe observed in breast cancer cells treated with C2 alone (Fig. 3A).BEX2-KD inhibited and BEX2 overexpression enhanced NF-nBactivation (both p50 and p65 components; Fig. 5A and B),indicating that BEX2 is in the NGF/NF-nB pathway upstream ofNF-nB in the modulation of apoptosis. In contrast, BEX2 seems notto be involved in NGF-mediated proliferation (Fig. 4D), which isnot surprising because this effect is mediated through a separatereceptor (p140TrkA) and downstream pathway (mitogen-activatedprotein kinase) in breast cancer cells (8, 9).Our data indicate that BEX2 is also important for apoptosis in
response to ER modulation by E2 and tamoxifen by modulatingNF-nB. NF-nB activity has been shown to play a role in hormoneindependence and resistance to tamoxifen and raloxifene in breastcancer (48–50). NF-nB is also activated in anti–estrogen (Fulves-trant)-resistant breast cancer cells, and this resistance can bereversed using NF-nB inhibitors (51). Our experiments with MCF-7
Figure 6. Outcome with tamoxifen therapyand proposed NGF/BEX2 /NF-nB apoptoticpathway. A, Kaplan-Meier disease-freesurvival curve in tamoxifen-treated caseswith high versus low BEX2 expression.Cum Survival, cumulative survival; DFI,disease-free interval in months; N, numberof events/number of cases in each group.B, schematic presentation of NGF/BEX2 /NF-nB pathway. +, stimulatory effect; �,inhibitory effect; dashed line, hypotheticalfeedback loop.
Cancer Research
Cancer Res 2007; 67: (14). July 15, 2007 6734 www.aacrjournals.org
cells provide further evidence for this role and suggest direct BEX2involvement: NF-nB activation by NGF treatment or BEX2 over-expression resulted in reduced tamoxifen-induced apoptosis andthe protective effect of BEX2 overexpression was removed withNF-nB inhibition (Fig. 4A and B).We propose a possible explanation to reconcile the cell line
results with the observed better prognosis of tamoxifen-treatedpatients who were also BEX2 overexpressers. We consistentlyobserved higher BEX2 expression on simultaneous activation of thepathway (with NGF/C2) and inhibition of NF-nB DNA binding(with SN50) compared with pathway activation alone (Fig. 5C). Thisshows that NF-nB DNA binding is upstream of BEX2 expression,possibly creating a feedback loop. Others have shown that nuclearextracts from ER+ tumors with better response to tamoxifen havemuch lower NF-nB DNA-binding activity and ER stimulationresults in direct inhibition of NF-nB DNA binding (52–54). We alsoobserved that BEX2 overexpression enhanced the antiproliferativeeffect of tamoxifen after 24 h of treatment (Fig. 4D). It is thereforeconceivable that the better prognosis in tamoxifen-treated BEX2+
patients is the combined result of relative NF-nB inactivation
leading to BEX2 overexpression, which in turn enhances thetamoxifen antiproliferative effect.In summary, overexpression of BEX genes is a novel classifier of
ER+ breast tumors, and potentially, BEX2 overexpression couldserve as a marker to identify tumors with better response totamoxifen therapy. We propose a model where BEX2 is part of boththe NGF/NF-nB and estrogen response pathways, which regulateapoptosis in breast cancer cells (Fig. 6B). In addition, BEX2cooperates in inhibiting proliferation of breast cancer cells treatedwith tamoxifen at pharmacologic dose. These findings suggest thetantalizing possibility of the NGF/BEX2/NF-nB pathway being atherapeutic target in breast cancer.
Acknowledgments
Received 11/29/2006; revised 4/30/2007; accepted 5/11/2007.Grant support: Cancer Research UK.The costs of publication of this article were defrayed in part by the payment of page
charges. This article must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.
References1. Sorlie T, Perou CM, Tibshirani R, et al. Geneexpression patterns of breast carcinomas distinguishtumor subclasses with clinical implications. Proc NatlAcad Sci U S A 2001;98:10869–74.2. Sorlie T, Tibshirani R, Parker J, et al. Repeatedobservation of breast tumor subtypes in independentgene expression data sets. Proc Natl Acad Sci U S A2003;100:8418–23.3. Sotiriou C, Neo SY, McShane LM, et al. Breast cancerclassification and prognosis based on gene expressionprofiles from a population-based study. Proc Natl AcadSci U S A 2003;100:10393–8.4. Dontu G, El-Ashry D, Wicha MS. Breast cancer, stem/progenitor cells and the estrogen receptor. TrendsEndocrinol Metab 2004;15:193–7.5. Arpino G, Weiss H, Lee AV, et al. Estrogen receptor-positive, progesterone receptor-negative breast cancer:association with growth factor receptor expression andtamoxifen resistance. J Natl Cancer Inst 2005;97:1254–61.6. Nicholson RI, Gee JM. Oestrogen and growth factorcross-talk and endocrine insensitivity and acquiredresistance in breast cancer. Br J Cancer 2000;82:501–13.7. Dolle L, Adriaenssens E, El Yazidi-Belkoura I, LeBourhis X, Nurcombe V, Hondermarck H. Nerve growthfactor receptors and signaling in breast cancer. CurrCancer Drug Targets 2004;4:463–70.8. El Yazidi-Belkoura I, Adriaenssens E, Dolle L, DescampsS, Hondermarck H. Tumor necrosis factor receptor-associated death domain protein is involved in theneurotrophin receptor-mediated antiapoptotic activityof nerve growth factor in breast cancer cells. J Biol Chem2003;278:16952–6.9. Descamps S, Toillon RA, Adriaenssens E, et al. Nervegrowth factor stimulates proliferation and survival ofhuman breast cancer cells through two distinctsignaling pathways. J Biol Chem 2001;276:17864–70.10. Brown AL, Kay GF. Bex1, a gene with increasedexpression in parthenogenetic embryos, is a member ofa novel gene family on the mouse X chromosome. HumMol Genet 1999;8:611–9.11. Alvarez E, Zhou W, Witta SE, Freed CR. Character-ization of the Bex gene family in humans, mice, and rats.Gene 2005;357:18–28.12. Mukai J, Hachiya T, Shoji-Hoshino S, et al. NADE,a p75NTR-associated cell death executor, is involvedin signal transduction mediated by the commonneurotrophin receptor p75NTR. J Biol Chem 2000;275:17566–70.13. Vilar M, Murillo-Carretero M, Mira H, Magnusson K,Besset V, Ibanez CF. Bex1, a novel interactor of the p75
neurotrophin receptor, links neurotrophin signaling tothe cell cycle. EMBO J 2006;25:1219–30.14. Foltz G, Ryu GY, Yoon JG, et al. Genome-wide analysisof epigenetic silencing identifies BEX1 and BEX2 ascandidate tumor suppressor genes in malignant glioma.Cancer Res 2006;66:6665–74.15. Naderi A, Teschendorff AE, Barbosa-Morais NL, et al.A gene-expression signature to predict survival in breastcancer across independent data sets. Oncogene 2007;26:1507–16.16. Naderi A, Ahmed AA, Barbosa-Morais NL, Aparicio S,Brenton JD, Caldas C. Expression microarray reproduc-ibility is improved by optimising purification steps inRNA amplification and labelling. BMC Genomics 2004;5:9.17. Naderi A, Ahmed AA, Wang Y, Brenton JD, Caldas C.Optimal amounts of fluorescent dye improve expressionresults in tumor specimens. Mol Biotechnol 2005;30:151–4.18. Troyanskaya O, Cantor M, Sherlock G, et al. Missingvalue estimation methods for DNA microarrays. Bio-informatics 2001;17:520–5.19. Liebermeister W. Linear modes of gene expressiondetermined by independent component analysis. Bio-informatics 2002;18:51–60.20. Hu Z, Fan C, Oh DS, et al. The molecular portraits ofbreast tumors are conserved across microarray plat-forms. BMC Genomics 2006;7:96.21. Ma XJ, Wang Z, Ryan PD, et al. A two-gene expressionratio predicts clinical outcome in breast cancer patientstreated with tamoxifen. Cancer Cell 2004;5:607–16.22. Tomlins SA, Rhodes DR, Perner S, et al. Recurrentfusion of TMPRSS2 and ETS transcription factor genesin prostate cancer. Science 2005;310:644–8.23. Lee SI, Batzoglou S. Application of independentcomponent analysis to microarrays. Genome Biol 2003;4:R76.24. Lacroix M, Leclercq G. About GATA3, HNF3A, andXBP1, three genes co-expressed with the oestrogenreceptor-a gene (ESR1) in breast cancer. Mol CellEndocrinol 2004;219:1–7.25. Vendrell JA, Magnino F, Danis E, et al. Estrogenregulation in human breast cancer cells of newdownstream gene targets involved in estrogen metab-olism, cell proliferation and cell transformation. J MolEndocrinol 2004;32:397–414.26. Weigelt B, Bosma AJ, van ’t Veer LJ. Expression of anovel lacrimal gland gene lacritin in human breasttissues. J Cancer Res Clin Oncol 2003;129:735–6.27. Porter D, Weremowicz S, Chin K, et al. A neuralsurvival factor is a candidate oncogene in breast cancer.Proc Natl Acad Sci U S A 2003;100:10931–6.28. Schwalbe M, Sanger J, Eggers R, et al. Differential
expression and regulation of bone morphogeneticprotein 7 in breast cancer. Int J Oncol 2003;23:89–95.29. van ’t Veer LJ, Dai H, van de Vijver MJ, et al. Geneexpression profiling predicts clinical outcome of breastcancer. Nature 2002;415:530–6.30. Tibshirani R, Hastie T, Narasimhan B, Chu G.Diagnosis of multiple cancer types by shrunkencentroids of gene expression. Proc Natl Acad Sci U S A2002;99:6567–72.31. Zhang GJ, Kimijima I, Onda M, et al. Tamoxifen-induced apoptosis in breast cancer cells relates todown-regulation of bcl-2, but not bax and bcl-X(L),without alteration of p53 protein levels. Clin Cancer Res1999;5:2971–7.32. Razandi M, Pedram A, Levin ER. Plasma membraneestrogen receptors signal to antiapoptosis in breastcancer. Mol Endocrinol 2000;14:1434–1447.33. Fernando RI, Wimalasena J. Estradiol abrogatesapoptosis in breast cancer cells through inactivationof BAD: Ras-dependent nongenomic pathways requiringsignaling through ERK and Akt. Mol Biol Cell 2004;15:3266–84.34. Chiarenza A, Lazarovici P, Lempereur L, Cantarella G,Bianchi A, Bernardini R. Tamoxifen inhibits nervegrowth factor-induced proliferation of the human breastcancerous cell line MCF-7. Cancer Res 2001;61:3002–8.35. Chien J, Staub J, Avula R, et al. Epigenetic silencing ofTCEAL7 (Bex4) in ovarian cancer. Oncogene 2005;24:5089–100.36. van de Vijver MJ, He YD, van ’t Veer L, et al. A gene-expression signature as a predictor of survival in breastcancer. N Engl J Med 2002;347:1999–2009.37. Wang Y, Klijin JG, Zhang Y, et al. Gene-expressionprofiles to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005;365:671–9.38. Taylor KM, Morgan HE, Johnson A, Hadley LJ,Nicholson RI. Structure-function analysis of LIV-1, thebreast cancer-associated protein that belongs to a newsubfamily of zinc transporters. Biochem J 2003;375:51–9.39. Martinez-Lacaci I, Saceda M, Plowman GD, et al.Estrogen and phorbol esters regulate amphiregulinexpression by two separate mechanisms in humanbreast cancer cell lines. Endocrinology 1995;136:3983–92.40. Carroll JS, Meyer CA, Song J, et al. Genome-wideanalysis of estrogen receptor binding sites. Nat Genet2006;38:1289–97.41. Han C, Liu H, Liu J, et al. Human Bex2 interacts withLMO2 and regulates the transcriptional activity of anovel DNA-binding complex. Nucleic Acids Res 2005;33:6555–65.42. Mukai J, Shoji S, Kimura MT, et al. Structure-functionanalysis of NADE: identification of regions that mediate
BEX2 Is Overexpressed in a Subset of Breast Cancers
www.aacrjournals.org 6735 Cancer Res 2007; 67: (14). July 15, 2007
nerve growth factor-induced apoptosis. J Biol Chem2002;277:13973–82.43. Mukai J, Suvant P, Sato TA. Nerve growth factor-dependent regulation of NADE-induced apoptosis.Vitam Horm 2003;66:385–402.44. Park JA, Lee JY, Sato TA, Koh JY. Co-induction ofp75NTR and p75NTR-associated death executor inneurons after zinc exposure in cortical culture ortransient ischemia in the rat. J Neurosci 2000;20:9096–103.45. Gill ZP, Perks CM, Newcomb PV, Holly JM. Insulin-like growth factor-binding protein (IGFBP-3) predis-poses breast cancer cells to programmed cell death in anon-IGF-dependent manner. J Biol Chem 1997;272:25602–7.46. Pirianov G, Danielsson C, Carlberg C, James SY,Colston KW. Potentiation by vitamin D analogs of TNFaand ceramide-induced apoptosis in MCF-7 cells is
associated with activation of cytosolic phospholipaseA2. Cell Death Differ 1999;6:890–901.47. Dobrowsky RT, Werner MH, Castellino AM, Chao MV,Hannun YA. Activation of the sphingomyelin cyclethrough the low-affinity neurotrophin receptor. Science1994;265:1596–9.48. deGraffenried LA, Chandrasekar B, Friedrichs WE,et al. NF-nB inhibition markedly enhances sensitivity ofresistant breast cancer tumor cells to tamoxifen. AnnOncol 2004;15:885–90.49. Pratt MA, Bishop TE, White D, et al. Estrogenwithdrawal-induced NF-nB activity and bcl-3 expressionin breast cancer cells: roles in growth and hormoneindependence. Mol Cell Biol 2003;23:6887–900.50. Liu H, Lee ES, Gajdos C, et al. Apoptotic action of17h-estradiol in raloxifene-resistant MCF-7 cells in vitroand in vivo . J Natl Cancer Inst 2003;95:1586–97.51. Riggins RB, Zwart A, Nehra R, Clarke R. The nuclear
factor nB inhibitor parthenolide restores ICI 182,780(Faslodex; fulvestrant)-induced apoptosis in antiestro-gen-resistant breast cancer cells. Mol Cancer Ther 2005;4:33–41.52. Kalaitzidis D, Ok J, Sulak L, Starczynowski DT,Gilmore TD. Characterization of a human REL-estrogenreceptor fusion protein with a reverse conditionaltransforming activity in chicken spleen cells. Oncogene2004;23:7580–7.53. Hsu SM, Chen YC, Jiang MC. 17h-estradiol inhibitstumor necrosis factor-a-induced nuclear factor-nBactivation by increasing nuclear factor-nB p105 levelin MCF-7 breast cancer cells. Biochem Biophys ResCommun 2000;279:47–52.54. Zhou Y, Eppenberger-Castori S, Marx C, et al.Activation of nuclear factor-nB (NFnB) identifies ahigh-risk subset of hormone-dependent breast cancers.Int J Biochem Cell Biol 2005;37:1130–44.
Cancer Research
Cancer Res 2007; 67: (14). July 15, 2007 6736 www.aacrjournals.org
