Effects of Estrogen on Global Gene Expression: Identification of Novel Targets of Estrogen Action1

[CANCER RESEARCH 60, 5977–5983, November 1, 2000] Advances in Brief

April H. Charpentier, Andrzej K. Bednarek, Rachael L. Daniel, Kathleen A. Hawkins, Kendra J. Laflin, Sara Gaddis, Michael C. MacLeod, and C. Marcelo Aldaz2 Department of Carcinogenesis, University of Texas, M. D. Anderson Cancer Center, Smithville, Texas 78957

Abstract It is known that the effect of E2 is mediated through its ability to bind the ERs, ␣ and ␤, which are basically ligand-activated transcrip- The important role played by the sex hormone estrogen in disease and tion factors. Recently, numerous ER-associated proteins, coactivators, physiological processes has been well documented. However, the mecha- and repressors have been identified that are of importance in regulat- nisms by which this hormone elicits many of its normal as well as pathological effects are unclear. To identify both known and unknown ing the ER interaction with the basal transcription machinery (2). The genes that are regulated by or associated with estrogen action, we per- mitogenic effects of E2 have been in large part attributed to its ability formed serial analysis of gene expression on estrogen-responsive breast to increase the expression of key cell cycle regulatory genes (3). cancer cells after exposure to this hormone. We examined approximately However, regulation of cell proliferation is just one aspect of interest

190,000 mRNA transcripts and monitored the expression behavior of in E2 studies. Of much importance is the identification of “novel” 12,550 genes. Expression levels for the vast majority of those transcripts downstream E2 effectors, regardless of their association with prolif- were observed to remain constant upon 17␤ estradiol (E ) treatment. Only 2 eration. Furthermore, the potential exists for such newly identified E2 approximately 0.4% of the genes showed an increase in expression of targets to become biomarkers of relevance in the monitoring of >3-fold by 3 h post-E treatment. We cloned five novel genes (E2IG1-5), 2 estrogen-related disease conditions such as breast cancer and osteo- which were observed up-regulated by the hormonal treatment. Of these porosis. the most highly induced transcript, E2IG1, appears to be a novel member of the family of small heat shock proteins. The E2IG4 gene is a new In this report we describe a comprehensive gene expression anal- member of the large family of leucine-rich repeat-containing proteins. On ysis of the effects of E2 using a classical E2-responsive human model. the basis of architectural and domain homology, this gene appears to be a To perform the comparative gene expression profiling we used SAGE, good candidate for secretion in the extracellular environment and, there- a powerful global gene expression technique that allows for the fore, may play a role in breast tissue remodeling and/or epithelium-stroma quantitative evaluation of all cellular mRNA populations (i.e., the interactions. Several interesting genes with a potential role in the regula- transcriptome; Ref. 4). As a result of this study, various novel E2 tion of cell cycle progression were also identified to increase in expression, targets, both known and previously unknown, were identified, which including Pescadillo and chaperonin CCT2. Two putative paracrine/auto- will stimulate additional studies into the mechanisms of E2 action. crine factors of potential importance in the regulation of the growth of breast cancer cells were identified to be highly up-regulated by E2: Materials and Methods stanniocalcin 2, a calcium/phosphate homeostatic hormone; and inhibin-␤ B, a TGF-␤-like factor. Interestingly, we also determined that E2IG1 and Human Breast Cancer Cell Lines. The MCF-7 cell line batch used in these stanniocalcin 2 were exclusively overexpressed in estrogen-receptor-posi- studies is derived directly from the original MCF-7 cell line. The ability of E2 tive breast cancer lines, and thus they have the potential to serve as breast to induce cell proliferation was verified by fluorescence-activated cell sorting cancer biomarkers. This data provides a comprehensive view of the analysis and cell counts. MCF-7 cells were maintained in Iscove’s MEM changes induced by E2 on the transcriptional program of human E2- (Biofluids) without phenol red and supplemented with glutamine, 50 ␮g/ml responsive cells, and it also identifies novel and previously unsuspected gentamicin, and 10% FBS (BioWhittaker). Other breast cancer cell lines used gene targets whose expression is affected by this hormone. included ZR-75-1, SKBR3, MDA-MB-157, MDA-MB-435, MDA-MB-453, and T47D grown in10% FBS/DMEM, and UACC-812 grown in 10%FBS/L- Introduction 15. We also used normal human mammary epithelial cells (HME-87), normal The sex steroid hormone estrogen plays an essential role in the mammary organoids (B43), and normal bulk breast tissues (HMG), as de- development of various tissues and in the maintenance of numerous scribed previously (5). SAGE. To obtain mRNA samples for SAGE, MCF-7 cells were seeded physiological processes. However, it has also been well documented into 150-mm plates (1.5 ϫ 106 cells/plate) and allowed to reach a logarithmic that estrogen plays a critical role in the etiology and progression of growth phase in culture media supplemented with FBS. At 40% confluency, human breast and gynecological cancers (1). cells were incubated for 48 h in culture media supplemented with 10% Ϫ charcoal stripped serum (Hyclone). Cells were then treated with 10 8 M Received 6/16/00; accepted 9/15/00. 17␤-estradiol (Sigma) or vehicle control (equivalent amounts of ethanol). The costs of publication of this article were defrayed in part by the payment of page Samples were collected at time 0 (i.e., untreated), 3, and 10 h after starting charges. This article must therefore be hereby marked advertisement in accordance with treatment. Total RNA was isolated using the Qiagen total RNA Maxi kit 18 U.S.C. Section 1734 solely to indicate this fact. 1 Supported in part by the Susan G. Komen Breast Cancer Foundation, postdoctoral (Qiagen) and mRNA was purified using the Oligotex mRNA mini kit (Qiagen) award to A. H. C., and National Institute of Environmental Health Sciences Center Grant following the manufacturer’s protocol. SAGE library generation and sequenc- ES07784. ing was performed as described previously by Velculescu et al. (4). Statistical 2 To whom requests for reprints should be addressed, at University of Texas, M. D. analysis of and comparison between different time points and controls were Anderson Cancer Center, Science Park Research Division, P. O. Box 389, Smithville, TX 78957. Phone: (512) 237-9530; Fax: (512) 237-2475; E-mail: [email protected]. performed as described by Zhang et al. (6) and by using statistical functions tmc.edu. available in the SAGE 3.0 software (kindly provided by Dr. K. Kinzler, The 3 ␤ The abbreviations used are: E2, 17 estradiol; FBS, fetal bovine serum; SAGE, serial John Hopkins University) for P calculations and Monte Carlo simulations. analysis of gene expression; RT-PCR, reverse transcription-PCR; HSP, heat shock pro- Northern Blot Analysis. Samples for Northern blot analysis were tein; CCT2, chaperonin-containing t-complex; STC, stanniocalcin; INHBB, inhibin-␤ B; ER, estrogen receptors; GB, GenBank; EST, expressed sequence tag; CAV1, caveolin 1; obtained from MCF-7 cells handled as described for SAGE and treated with Ϫ8 NDKA, nucleoside diphosphate kinase A, or Nm23H1; PES1, Pescadillo. 10 M E2 for 0, 3, 6, 10, 15, and 24 h and with vehicle control for 24 h before 5977

Downloaded from cancerres.aacrjournals.org on September 26, 2021. © 2000 American Association for Cancer Research. EFFECTS OF ESTROGEN ON GENE EXPRESSION ANALYZED BY SAGE harvesting. RNA isolation and Northern blotting was performed following standard procedures. Probes for hybridization were obtained by RT-PCR from breast cDNA libraries and their sequence was confirmed before use. Expres- sion of ER-␣ mRNA was confirmed using RT-PCR. Cloning and Computer Analysis of Novel Transcripts. Novel transcripts were cloned from a human placenta cDNA library (Rapid Screen; Origene Technologies). To screen this library we used the same primers as for generating Northern probes. Each cDNA clone was sequenced in its entirety, and the SAGE tag was confirmed. The predicted amino acid sequence for each transcript was analyzed using BLASTP and PSIBLAST algorithms, and the identification of protein family domains was determined using the Pfam domain models.4 For protein cellular localization analysis we used the PSORT algorithm (7).

Results Summary of Global Gene Expression Findings The SAGE method generates short sequences (i.e., transcript tags) specific to each expressed gene. The proportion of each tag in the overall tag population is representative of the proportion of each mRNA in the original mRNA population. Expression patterns are then deduced from the abundance of individual tags within each sample set. In the following study we generated a SAGE database of E2- Ϫ8 responsive MCF-7 breast cancer cells treated with 10 M E2. The MCF-7 E2-dependent system was chosen because it is one of the best and most widely used models for the study of the effects of E2 on human cells (8).

MCF-7 cells were cultured in conditions devoid of E2 for 48 h before the start of the experiments. mRNA was then collected for

SAGE at time points of 3 (E2 3 h) and 10 (E2 10 h) h after E2 treatment. The sampled time points allowed us to analyze changes occurring in the transcriptome of the MCF-7 cells prior to entry into S phase, as determined by fluorescence-activated cell sorting (data not shown).

The SAGE profiles for E2-treated cells (E2 3 h and E2 10 h), Fig. 1. A, scattergrams comparing the levels of expression of the 12,550 transcripts Ϫ8 were compared with SAGE profiles from untreated MCF-7 cells detected in MCF-7 at time points 0, 3, and 10 h after treatment with 10 M E2. The (E 0 h) and a 3-h, vehicle-treated control. A total of 188,367 Y and X axes indicate the total number of tags detected for each of the indicated 2 conditions. Each white dot represents a transcript, and its location in the plot will transcript tags (approximately 61,000 tags for each E2 0h,E2 3h, depend on the relative expression in the conditions compared. As can be observed by and E2 10 h) plus additional tags from vehicle control were the excellent r squares obtained, the three samples are remarkably similar, indicating sequenced and analyzed. These tags identified a total of 12,550 that relatively few genes are found to be differentially expressed in a significant way. B, summary of the ratios of abundance of SAGE transcript tags expressed in MCF-7 different transcripts. Of these, the vast majority (83%) showed cells in the presence (10-h time point) versus the absence of E2. To avoid division by matches in GB databases to either known genes or anonymous zero, we used a tag value of 1 for any tag that was not detectable in one of the samples. ESTs. The remaining 17% of the tags (i.e., 2,100 transcripts) Ratios were rounded to the nearest integer. Negative integers represent a decrease in tag numbers upon E2 treatment, and positive integers represent an increase in tag showed no database matches. numbers upon treatment. The number of tags displaying each ratio is plotted on the Y Comparison of the SAGE tag libraries from MCF-7 treated and axis. untreated cells demonstrated a remarkable similarity between expression profiles. Fig. 1A illustrates scatter plots representing the relative expression of all transcripts analyzed, derived from pairwise compar- more than one sequence, and 3 transcripts showed no reliable matches isons of the three SAGE libraries. The excellent correlation coeffi- in the GB databases (Table 1). One transcript (the arginino succinate cients illustrates both the reproducibility of SAGE using three differ- synthase gene) was also observed to increase in the vehicle control- ent samples and RNA isolates as well as the close similarity of the treated cells. samples under analysis in this particular study. The vast majority of The SAGE database containing the complete list of the 12,550 Ϸ transcripts did not change in expression; 81.4% of them ( 10,214 transcripts identified in the MCF-7 cells, as well as their relative levels transcripts) exhibited a Ͻ2-fold difference in expression upon E 2 of expression under the various conditions of E treatment, can be treatment (Fig. 1B). 2 viewed at our website.5 This MCF-7 SAGE database displays, among Comparative statistical analyses of the tag libraries was performed other features, the Ps for each individual comparison in relative to estimate the relative likelihood that a detected difference in expres- transcript expression and active “tag links” connected to the recently sion would be seen by chance for each individual tag, given the size of the SAGE libraries under study (6). Only 50 transcripts demon- described SAGEmap National Center for Biotechnology Information strated an increase in expression at approximately Ն3-fold at the P databases in which expression levels of each tag in various tissues and Ͻ0.001 level of significance. Of these transcripts, 37 tags identified cell lines can be evaluated (9). known genes, 8 corresponded to anonymous EST clusters, 2 matched

4 Internet address: http://pfam.wustl.edu. 5 Internet address: http://sciencepark.mdanderson.org/ggeg. 5978

Table 1 Transcripts induced by E2 treatment in MCF-7 cells Tag E2 0h E2 3h E2 10h Accession No. Locus (a) Description Novel genes CCTGGCCTAA 0 18 4 AF191017 E2IG1 Estrogen-induced gene 1 GTCAGATGTC 1 14 6 NM_016565 E2IG2 Estrogen-induced gene 2 AGCTGTGTAA 0 12 2 NM_014366 E2IG3 Estrogen-induced gene 3 GGCATCAGGG 0 12 13 AF191019 E2IG4 Estrogen-induced gene 4 GTTATTGAGG 0 5 11 NM_014367 E2IG5 Estrogen-induced gene 5 Chaperones GGCTCCCACT 0 21 22 NM_007355 HSPCB Heat shock 90kD protein 1, ␤ TACTAGTCCT 0 5 12 AF028832 HSPCA Heat shock 90kD protein 1, ␣ GAAGCTTTGC 4 25 20 AF028832 HSPCA Heat shock 90kD protein 1, ␣ TACCAGTGTA 1 9 17 NM_002156 HSPD1 Heat shock 60kD protein 1 (chaperonin) CCAGGAGGAA 30 84 162 NM_006597 HSPA10 Heat shock 70kD protein 10 (HSC71) CTTGAGCAAT 6 18 21 NM_002014 FKBP4 FK506-binding protein 4 Cell cycle progression role CTGTGACACA 0 6 10 NM_006431 CCT2 Chaperonin-containing TCP1, subunit 2 (␤) AAAGTCTAGA 32 113 71 NM_001758 CCND1 Cyclin D1 TCCTCAAGAT 7 13 26 NM_004450 ERH Enhancer of rudimentary (Drosophila) homolog AGGTGCAGAG 0 14 9 U78310 PES1 Pescadillo (zebrafish) homolog 1 CGGATAACCA 7 17 29 NM_006191 PA2G4 Proliferation-associated 2G4, 38kD TGCTTCATCT 4 31 38 NM_006325 RAN RAN, member RAS oncogene family TTGTTGTTGA 12 32 40 NM_001743 CALM2 Calmodulin 2 (phosphorylase kinase-␦) Paracrine/autocrine factors CCCGAGGCAG 0 10 2 NM_003714 STC2 Stanniocalcin 2 ATTAGTCAGA 1 17 8 M31682 INHBB Inhibin-␤ B (activin AB betapolypeptide) Tumor-associated proteins CAGATTGTGA 0 9 17 NM_006510 RFP RET finger protein ACTGTCTCCA 7 27 19 NM_003287 TPD52L1 Tumor protein D52-like 1 (D53) TCCTGTAAAG 6 11 25 NM_001753 CAV1 Caveolin-1, caveolae protein, 22kD CTGGCCCTCG 13 94 186 NM_003225 TFF1 Trefoil factor 1(PS2 gene) GGCAGAGGAC 27 41 67 NM_000269 NME1 NDKA (Nm23H1) Protein degradation AAGGAATCGG 0 11 4 NM_002796 PSMB4 Proteasome subunit, ␤ type, 4 GCCAAGTTTG 1 11 3 NM_002807 PSMD1 Proteasome 26S subunit, non-ATPase, 1 Protein synthesis role/ribosome-related CTGGCAGATT 1 3 14 NM_006802 SF3A60 Splicing factor 3a, subunit 3, 60kD GGGGCAGGGC 10 20 31 NM_001970 EIF5A Eukaryotic translation initiation factor 5A CCTGCTCCCT 0 5 12 NM_003486 SLC7A5 Solute carrier family 7 member 5 TGTGTTGAGA 70 224 250 NM_001402 EEF1A1 Eukaryotic translation elongation factor 1 ␣ 1 GTGTTAACCA 22 85 70 NM_002948 RPL15 Ribosomal protein L15 TACCAGCACA 4 23 21 AF123534 NOP58 Nucleolar protein NOP5/NOP58 Chromatin-related GACGTGTGGG 1 2 16 NM_002106 H2AFZ H2A histone family, member Z GGGTCAAAAG 8 27 30 NM_002107 H3F3A H3 histone, family 3A Miscellaneous CGTGTTAATG 4 21 33 NM_003418 ZNF9 Zinc finger protein 9 (CNBP, sterol reg.binding prot.) CAGCCTTGGA 5 18 26 NM_005105 RBM8 RNA-binding motif protein 8 (ribonucleoprotein) CTGAGACAAA 3 23 17 NM_001207 BTF3 Basic transcription factor 3 TCTTGTGCAT 4 23 20 NM_005566 LDHA Lactate dehydrogenase A AACGTGCAGG 1 19 11 NM_000050 ASS Argininosuccinate synthase (see text) GAATCAGAAG 4 20 17 NM_001879 MASP1 Mannan-binding lectin serine protease 1 CACAGGCAAA 9 28 16 D13630 KIAA0005 KIAA0005 gene GCGGACGAGG 0810 PDCD5 or TAP Two matches TGTAAGAAAA 6826 MCM6 or SUOX Two matches ACCCCCAAGG 0 3 10 ESTs, Weakly similar to Strabismus [D. melanogaster] GCAGCTAATT 1 2 13 ESTs ACGTGGTGAT 4 12 20 ESTs similar to Leydig cell tumor prot. (Rattus N.) GAACCCAAAG 0 10 3 No reliable match AACCAATCTG 0 16 2 No reliable match CGCCGCCGGT 2 7 27 No reliable match a Ն Bold, genes identified by tags increasing at 10-fold at 3h post-E2.

Cloning and Characterization of Novel E2-induced Genes number from 0 to 18 tags, was named E2IG1. The corresponding cDNA is 2007 bp long containing a predicted open reading frame of As indicated, several of the transcript tags shown in Table 1 588 bp. The central portion of this 196AA protein is homologous to identified anonymous EST clusters or had no matches in GB data- a highly conserved HSP-␣ crystalline domain common to all HSP20 bases. On our follow-up studies to the SAGE analysis, we first Ͼ family members (10). Furthermore, E2IG1 shows a 54% homology to focused on those transcript tags that increased 10-fold upon E2 treatment. As a result, we isolated a series of five novel cDNAs from HSP27, suggesting that E2IG1 is a novel member of the small HSP a human placenta cDNA library using transcript-specific PCR prim- family. On the basis of matching sequence with sts TIGR–A002J47, we can predict that this gene maps to chromosome 12 (D12S366– ers. We named these novel genes “E2IGs” 1–5, for “E2-induced genes.” Full-length cDNA sequences for each of the novel transcripts D12S340 interval). have been reported to the GB databases of the National Center for E2IG4 corresponds to the SAGE tag GGCATCAGGG, which Biotechnology Information under accession nos. AF191017, increased from 0 tags to 12 tags by 3 h post-E2 exposure in MCF7 AF242180, AF191018, AF191019, and AF191020 for E2IG1, E2IG2, cells. This cDNA is 2542 bp long, showing a 1059-bp open reading E2IG3, E2IG4, and E2IG5, respectively. frame encoding for a 353AA protein. This protein is characterized

The gene cloned from tag CCTGGCCTAA, which increased in by the presence of a leucine-rich repeat in the NH2-terminal 5979

Among these transcripts we found abundant representation of protein members of a large class of molecular chaperones known as HSPs. In our studies, the most significant increases in gene expression were observed for HSP90 (both ␣ and ␤ chains), HSP60 (chaperonin), CCT2, HSC71 (HSP70 family), FKBP4 (immunophilin/p59/HSP56),

and the E2-induced gene described in this report, E2IG1 (HSP20 family; Table 1). Cell Cycle Progression-related Genes. SAGE tags identifying

CCT2 were found to increase from 0 to 10 tags by 10 h post-E2 treatment. This chaperone complex plays a fundamental role in cell cycle progression because it is in charge of the proper folding and maturation of cyclin E molecules (12).

Among well-known critical cell cycle targets of E2, the tags identifying cyclin D1 were observed to increase from 32 to 113 SAGE tags (i.e., 3.5-fold) by3hofhormonal treatment. The tag identifying the human homologue to the zebrafish developmental gene PES1 also demonstrated a marked increase in expres-

sion after E2 treatment (from 0 to 14 tags; Table 1). PES1 is predicted to encode a protein of 582AA that is highly conserved from yeast to humans (13). On the basis of the presence of a BRCT (BRCA1 COOH terminus) domain in PES1, we can speculate that PES1 may be of relevance in cell cycle regulation (14). Ran/TC4, another transcript up-regulated by the treatment, is a small GTP-binding protein member of the Ras superfamily, which is essential for the translocation of RNA and proteins through the nuclear pore complex. This GTPase may play a role in releasing steroid receptors from the nuclear pore complex and in connecting DNA synthesis with the onset of mitosis (15, 16). Tags identifying other relevant cell cycle progression modulators include those identifying calmodulin subunits, CALM1 and CALM2. CALM1 expression increased from 30 to 50 tags (data not shown), and Ͻ CALM2 increased from 12 to 40 tags (P 0.001) upon E2 treatment (Table 1). It is known that this important calcium-binding protein Fig. 2. The E2IG4 transcript. The predicted amino acid sequence of the E2IG4 plays an essential role for quiescent cells to enter the cell cycle (17). transcript is depicted, including leucine repeats (dark box), leucine-rich repeat in the Calmodulin also regulates phosphorylation and induces conforma- NH2-terminal (light box), signal peptide (open box), and conserved amino acid sequences (bold). The last CATG and tag sequence are also depicted in bold. tional changes of the ER (18, 19). We also observed an increase in tag numbers identifying subunits of the proteasome complex, a multicatalytic intracellular protease domain and eight additional leucine-rich repeats (Fig. 2). Interest- system that targets key proteins for degradation (20). The ubiquitin- ingly, E2IG4 demonstrates domain and structural homology to proteasome pathway also appears to be a major mechanism implicated extracellular matrix leucine-rich proteoglycans such as decorin in the turnover of the ER in an E2-dependent manner, which would (PGS2) and biglycan (PGS1). The E2IG4 protein is likely to reside agree with our findings (21). in the cell membrane or extracellular compartments (57% proba- Paracrine-Autocrine Factors. Among interesting gene targets not bility; Ref. 7). E2IG4 also contains a typical cleavable signal associated previously with E2 effects, we observed a 10-fold increase peptide at AA 1–16, which makes it a candidate protein for in expression for STC2 (Table 1). STC2 (also known as stanniocalcin- intracellular transport and extracellular secretion. In addition, related peptide STCrP) has amino acid sequence homology to the E2IG4 contains various potential phosphorylation sites throughout previously identified STC1 (22, 23). However, the SAGE tag match- its amino acid sequence. On the basis of matching DNA sequence ing STC1 did not increase in numbers, nor did we observe up- from sts AA009735 to E2IG4, it is possible to predict the mapping regulation by Northern analysis (data not shown). of this gene to human chromosome 11, (D11S911–D11S4172 An additional autocrine/paracrine factor, INHBB (activin-␤-b sub- Ͼ interval). unit) was identified to increase 15-fold upon E2 treatment (Table 1). The E2IG2 cDNA, encodes for a 97AA protein with some homol- Originally inhibin was identified as a gonadal hormone that inhibits ogy to two yeast proteins. The E2IG3 gene appears to be a nuclear the secretion of follicle-stimulating hormone (whereas activin stimu- 560AA GTP-binding protein and the E2IG5 cDNA encodes a 148AA lates the secretion of follicle-stimulating hormone) by the pituitary protein with significant homology to a transformation-dependent rat gland (24). Activins/inhibins belong to the TGF-␤ superfamily (25). protein (pIL2; Ref. 11). More specific information on these novel The only inhibin-related transcript tag found up-regulated by E2 transcripts is available on our Internet site.5 treatment was that of the INHBB subunit. Tumor-associated Genes. SAGE identified an increase in expres- E -Induced Expression of Previously Known Genes sion for various tumor associated proteins, RFP (ret finger protein), 2 D52L1 (D53 tumor protein), Trefoil factor (TFF1 or PS2), CAV1, and Chaperones. In addition to the novel genes described above, we NDKA, among others (Table 1). show in Table 1 all genes that were found to increase in expression at RFP was first identified for its elevated expression in a variety of Ն Ͻ 3-fold with E2 treatment at the P 0.001-level of significance. rodent tumor cell lines (26). This protein is located in the nucleus and 5980

Downloaded from cancerres.aacrjournals.org on September 26, 2021. © 2000 American Association for Cancer Research. EFFECTS OF ESTROGEN ON GENE EXPRESSION ANALYZED BY SAGE because of its zinc finger domain is believed to bind DNA. However the function of this protein is, at present, unknown. Tumor protein D52L1, also known as D53, increased from 7 tags in untreated cells to 27 tags in the 3-h E2-treatment profile. D53 was originally cloned from a breast carcinoma cDNA library and has a 52% homology to D52 (27). Interestingly, and in agreement with our SAGE findings, previous studies have shown that both D52 and D53 transcription in breast carcinoma cells are dependent upon E2 presence in MCF-7 cells (27).

The well-known E2 regulated gene PS2 (Trefoil factor; Ref. 28) was also detected to increase 14-fold by SAGE from 13 tags at0hto

186 tags by 10-h-post-E2 treatment. Interestingly, we observed a significant increase in tags identifying the putative metastasis suppressor gene NDKA/Nm23H1 (2.5-fold increase; P ϭ 0.000003). It is likely that such an increase in expression is related to its fundamental housekeeping role in maintaining the nucleoside diphosphate/nucleotide triphosphate cellular balance (29). The tags identifying CAV1 were also shown to increase ϳ4-fold (Table 1). CAV1 is an integral membrane protein and the main component of caveolae membranes (a plasma membrane specializa- tion). It was recently shown that CAV1 functions as a membrane adaptor to link integrin subunits to the tyrosine kinase FYN (30).

Validation of SAGE Findings. The effects of E2 treatment on the expression of specific genes were confirmed by means of Northern analysis. To this end, MCF-7 cells were grown in conditions devoid of

E2 for 48 h and then exposed to E2 treatment for various lengths of time as shown in Fig. 3. These representative transcripts demonstrated different patterns of expression. Some transcripts increased and re- mained high for the entire course of treatment, whereas others increased rapidly in early time points and then tapered off in later time points (Fig. 3). As can be observed, E2IG1 and E2IG4 already increased considerably by3hofE2 treatment. E2IG5 showed only a modest increase in expression levels as detected by total RNA North- ern blot analysis (data not shown). The reason for the apparent discrepancy between the fold values of increased expression detected by SAGE versus those detected by Northern analysis in this last case is unclear. Fig. 3. Validation of E2-induced expression of specific transcripts in MCF-7 cells. Representative Northern blots are shown for several transcripts identified by SAGE to be It is worth mentioning that Northern blot analysis also confirmed expressed at higher levels in E2-treated samples as compared with untreated (0 h) samples. SAGE observations in transcripts that demonstrated a Ͻ3-fold differ- MCF-7 cells were grown in conditions devoid of E2 for 48 h, and at time 0 cells were Ϫ8 treated with 10 M E2 over a 24-h time course. Controls included MCF-7 cells main- ence; e.g., for the NDKA transcript, SAGE detected 27 tags in control tained in the presence of FBS (first lane) and MCF-7 cells treated with equivalent amounts compared with 67 tags in the E2-treated (Table 1), and Northern blot of vehicle (ethanol) for 24 h (24 control). All Northern analyses were performed on total analysis confirmed this fold difference with remarkable accuracy to a RNA with the exception of INHBB, in which mRNA was used (bottom two rows). A probe for glyceraldehyde-3-phosphate dehydrogenase was used as a control for RNA loading. 2.5-fold change. In addition, SAGE also identified a 3.5-fold increase in cyclin D upon E2 treatment in close agreement with Northern and Western data reported previously (3). Our studies led us to identify and clone five novel genes shown to

increase upon E2 treatment. We also identified a series of previously Transcript Expression in ER؉ and ER؊ Breast Cancer unsuspected targets of E2 effects. Cell Lines Our cloning effort was focused on those SAGE tags identifying anonymous ESTs which increased above 10-fold upon E treatment. Northern blot analysis was also used to assess the level of expres- 2 Among these, E2IG1 was identified as a putative small HSP and sion for E2IG1, E2IG2, E2IG3, E2IG4, and STC2 in a small panel of which bears 54% homology to HSP27. Interestingly, the HSP27 ERϩ and ER- cell lines. The expression of ER-␣ for each of the cell expression has been associated with the presence of ER in breast and lines was confirmed by RT-PCR (Fig. 4). Both E2IG1 and STC2 endometrial carcinomas (31). It was remarkable in our study that there demonstrated selective expression in ERϩ cell lines (Fig. 4), whereas was selective overexpression of E2IG1 in ERϩ breast cancer cell E2IG2, E2IG3, and E2IG4 demonstrated varied expression in the lines (Fig. 4). These findings suggests that E2IG1 expression is different cell lines (data not shown). The selective expression in ERϩ dependent on the presence of a functional ER. On the basis of these cell lines for STC2 and E2IG1 suggests their expression is dependent observations and on the rapid and dramatic up-regulation of E2IG1 on the presence of a functional ER. after E2 treatment, we can speculate that this gene is a prime candidate Discussion to be a direct effector of E2. Because E2IG1 is a putative new member of the small HSP family, it may play a specific chaperonic role related

In this study we have used a powerful global gene expression to either the ER itself or to some point downstream of the E2-induced methodology to identify novel direct or associated targets of E2 action. signaling cascade. In addition, the described selective overexpression 5981

of cyclin E as a key player promoting the G1-to-S phase transition in breast cancer cells has been shown (39). The demonstrated ability of CCT2 to control in turn the maturation of cyclin E, and hence its critical function in the cell cycle, appears to provide an additional

layer of influence in cell cycle progression by E2. PES1 was another gene observed to be up-regulated upon E2 treatment that also may play a role in cell cycle regulation. The highly conserved PES1 displays a region of approximately 100 amino acids (322–415) with homology to the BRCT domain superfamily. BRCT domains are usually found in proteins that play critical roles in cell cycle control and DNA repair (14). The possibility that PES1 is playing a role in either of these processes is of interest and therefore deserves additional investigation. Fig. 4. Northern blot analysis of various breast cancer cell lines. ERϩ cell lines including MCF-7, T47D, HME-87, UACC812, and ZR751 and estrogen ER- cell lines Among the identification of novel targets of E2 action, it is impor- including MDA MB157, MDA MB435, MDA MB453, and SKBR3 were used for tant to stress the dramatic increase in expression observed for STC2. Northern blot analysis on the transcript E2IG1 and STC2. RT-PCR was used to confirm Although STC was initially thought to be unique to fish, the two the expression of ER␣, bottom panel. mentioned homologues of STC have now been identified in mamma- lians (i.e., STC1, 61% homology, and STC2, 30–38% homology). In of E2IG1 in ERϩ breast tumor cells make it a good potential marker humans, the function of these hormones is unknown. However, in fish for E2-dependent breast carcinomas. it was observed that STC functions as a potent regulator of calcium A second very intriguing novel target of E2 treatment is the E2IG4 and phosphate homeostasis preventing hypercalcemia in a similar protein. This putative protein is a new member of the large family of fashion to calcitonin in mammals (40). It has been proposed that leucine-rich repeat-containing proteins. On the basis of sequence mammalian STCs act as regulators of calcium/phosphate homeostasis homology and motif distribution analysis, it is possible to speculate in a paracrine and/or autocrine, rather than endocrine, fashion, as that E2IG4 may constitute a novel extracellular matrix component. suggested by their widespread expression in tissues (41–43). Some Analysis of the E2IG4 amino acid sequence showed some moderate evidence suggests that STC2 acts in a manner opposite to STC1 on homology to GAC1, a leucine-rich protein amplified in gliomas (32). regulating calcium/phosphate concentrations (23). The fact that STC2 E2IG4 also showed domain and leucine-repeat distribution homology, appears to be a secreted protein and is selectively overexpressed in as well as general features of similarity, to several cleaved extracel- ERϩ cell lines suggests that it may have potential use as a serum- lular proteins such as: platelet glycoprotein V precursor; PGS2 (bone detectable prognostic marker for breast cancer. In addition, because of proteoglycan II precusor), also known as decorin; and PGS1 (bigly- its potential role as a calcium/phosphate regulator and its wide ex- can). Both decorin and biglycan belong to the small interstitial pro- pression, including in bone, STC2 becomes a prime target for addi- teoglycans family and are found in the extracellular matrix. Interest- tional study because it may provide a novel link between the effects ingly in rat, decorin has been isolated from cervix uteri, and it appears of E2 and bone remodeling. to be hormonally regulated (33). SAGE also identified another interesting E2-regulated, autocrine/ The presence of a typical signal peptide cleavage site in the AA paracrine factor, INHBB. This protein subunit functions either as a sequence of E2IG4 also makes this protein a good candidate for homodimer (activin B) or a heterodimer with ␤-A or -␣ subunits transport to the cell membrane and secretion to the extracellular (forming activin AB or inhibin B, respectively). The elevation of ␣ ␤ environment. The combination of the described E2IG4 structural inhibin subunits and -B, after E2 treatment has been observed in features and homologies plus the observed up-regulation induced by granulosa cells (44). It has also been shown that ERϩ breast cancer

E2 suggests that this protein could play a role in hormonally regulated, lines are growth-inhibited by activin B (45). Furthermore, the expres- extracellular matrix remodeling and/or epithelium-stromal interac- sion of inhibin/activin subunits and corresponding receptors has been tions in breast tissue. detected in MCF-7 cells (46). Theoretically, the increased expression

Among the SAGE tags increasing in abundance upon E2 treatment, of INHBB would tilt the homo/heterodimer balance toward generation several identified various members of large families of proteins with of activin B, inasmuch as INHBB was the only inhibin subunit found chaperonic function. There is evidence to suggest that the expression to increase upon E2 treatment. This appears to point to a paradoxical ␤ of HSPs may play a role in breast cancer and in E2 regulation, effect of E2 because activins, along with the rest of the TGF- family, however the induced expression of HSPs is not unique to E2 (34–36). are thought to contribute to maintaining a negative-growth regulatory HSP90 also functions as a stabilizer of critical signal transducers, function in ERϩ breast cancer cells, which would be in opposition to including cell cycle and developmental regulators. These include the ability of E2 to stimulate cell growth (45). Additional study is Src-family-kinases, raf serine/threonine kinases, calmodulin, dioxin- therefore required to better understand the role of modifications in the receptor, cyclin-dependent kinases, and steroid-hormone receptors activin/inhibin homeostasis in E2-responsive breast cancer cells and (37). HSP90 together with HSP70 play important roles in the matu- its putative effect in regulating cell growth. ration of the steroid receptor in achieving a hormone-binding compe- In this report we have summarized some of the most prominent tent state as well as in regulating the receptor cytoplasmic-nuclear effects induced by E2 affecting the transcriptional program of E2- trafficking (15). The HSP90-binding protein, FKBP4, was also ob- responsive cells at a global level. As mentioned, several of the served increasing in expression upon E2 treatment. Both HSP90 and observed changes in gene expression will be common to those in- FKBP4 are known to bind unliganded, steroid receptor complexes duced by other mitogens. On the other hand, numerous changes (38). observed are likely to be directly induced by this important female Interestingly SAGE also identified up-regulation of CCT2, a chap- steroid hormone. Follow-up studies on the mechanisms of expression erone protein which has recently been shown to bind newly synthe- regulation for the novel targets described, as well as for previously sized cyclin E, mediating its maturation (folding) into a form that can unsuspected effectors of E2, are now in order. associate with cyclin-dependent kinase 2 (12). The important function It is also important to stress that several of these genes have a very 5982

Downloaded from cancerres.aacrjournals.org on September 26, 2021. © 2000 American Association for Cancer Research. EFFECTS OF ESTROGEN ON GENE EXPRESSION ANALYZED BY SAGE good potential to serve as diagnostic/prognostic tools for the moni- 24. McLachlan, R. I., Robertson, D. M., de Kretser, D., and Burger, H. G. Inhibin: a toring of estrogen-related disease conditions such as breast cancer. non-steroidal regulator of pituitary follicle stimulating hormone. Baillieres Clin. Endocrinol. Metab., 1: 89–112, 1987. 25. Massague, J. The transforming growth factor-␤ family. Annu. Rev. Cell Biol., 6: References 597–641, 1990. 26. Isomura, T., Tamiya-Koizumi, K., Suzuki, M., Yoshida, S., Taniguchi, M., 1. Pike, M., Spicer, D., Dahmoush, L., and Press, M. Estrogens, progesterones, normal Matsuyama, M., Ishigaki, T., Sakuma, S., and Takahashi, M. RFP is a DNA binding breast cell proleferation, and breast cancer risk. Epidemiol. Rev., 15: 17–35, 1993. protein associated with the nuclear matrix. Nucleic Acids Res., 20: 5305–5310, 1992. 2. McKenna, N. J., Xu, J., Nawaz, Z., Tsai, S. Y., Tsai, M. J., and O’Malley, B. W. 27. Byrne, J. A., Mattei, M. G., and Basset, P. Definition of the tumor protein D52 Nuclear receptor coactivators: multiple enzymes, multiple complexes, multiple func- (TPD52) gene family through cloning of D52 homologues in human (hD53) and tions. J. Steroid. Biochem. Mol. Biol., 69: 3–12, 1999. mouse (mD52). Genomics, 35: 523–532, 1996. 3. Prall, O., Sarcevic, B., Musgrove, E., Watts, C., and Sutherland, R. Estrogen-induced 28. Jakowlew, S. B., Breathnach, R., Jeltsch, J. M., Masiakowski, P., and Chambon, P. activation of Cdk4 and Cdk2 during G -S phase progression is accompanied by 1 Sequence of the pS2 mRNA induced by estrogen in the human breast cancer cell line increased cyclin D1 expression and decreased cyclin-dependent kinase inhibitor MCF-7. Nucleic Acids Res., 12: 2861–2878, 1984. association with cyclin E-Cdk2. J. Biol. Chem., 272: 10882–10894, 1997. 29. Schaertl, S., Konrad, M., and Geeves, M. A. Substrate specificity of human nucleo- 4. Velculescu, E., Zhang, L., Volgelstein, B., and Kinzler, W. Serial analysis of gene side-diphosphate kinase revealed by transient kinetic analysis. J. Biol. Chem., 273: expression. Science (Washington DC), 270: 484–487, 1995. 5662–5669, 1998. 5. Brenner, A. J., Stampfer, M. R., and Aldaz, C. M. Increased p16 expression with first 30. Wary, K. K., Mariotti, A., Zurzolo, C., and Giancotti, F. G. A requirement for senescence arrest in human mammary epithelial cells and extended growth capacity caveolin-1 and associated kinase Fyn in integrin signaling and anchorage-dependent with p16 inactivation. Oncogene, 17: 199–205, 1998. 6. Zhang, L., Zhou, W., Velculescu, V., Kern, S., Hruban, R., Hamilton, S., Vogelstein, cell growth. Cell, 94: 625–634, 1998. B., and Kinzler, K. Gene expression profiles in normal and cancer cells. Science 31. Ciocca, D. R., Oesterreich, S., Chamness, G. C., McGuire, W. L., and Fuqua, S. A. (Washington DC), 276: 1268–1272, 1997. Biological and clinical implications of heat shock protein 27,000 (Hsp27): a review. 7. Nakai, K., and Kanehisa, M. A knowledge base for predicting protein localization J. Natl. Cancer Inst., 85: 1558–1570, 1993. sites in eukaryotic cells. Genomics, 14: 897–911, 1992. 32. Almeida, A., Zhu, X. X., Vogt, N., Tyagi, R., Muleris, M., Dutrillaux, A. M., 8. Levenson, A. S., and Jordan, V. C. MCF-7: the first hormone-responsive breast Dutrillaux, B., Ross, D., Malfoy, B., and Hanash, S. GAC1, a new member of the cancer cell line. Cancer Res., 57: 3071–3078, 1997. leucine-rich repeat superfamily on chromosome band 1q32.1, is amplified and over- 9. Lal, A., Lash, A. E., Altschul, S. F., Velculescu, V., Zhang, L., McLendon, R. E., expressed in malignant gliomas. Oncogene, 16: 2997–3002, 1998. Marra, M. A., Prange, C., Morin, P. J., Polyak, K., Papadopoulos, N., Vogelstein, B., 33. Kokenyesi, R., and Woessner, J. F., Jr. Purification and characterization of a small Kinzler, K. W., Strausberg, R. L., and Riggins, G. J. A public database for gene dermatan sulphate proteoglycan implicated in the dilatation of the rat uterine cervix. expression in human cancers. Cancer Res., 59: 5403–5407, 1999. Biochem. J., 260: 413–419, 1989. 10. Kim, K. K., Kim, R., and Kim, S. H. Crystal structure of a small heat-shock protein. 34. Shyamala, G., Gauthier, Y., Moore, S. K., Catelli, M. G., and Ullrich, S. J. Estrogenic Nature (Lond.)., 394: 595–599, 1998. regulation of murine uterine 90-kilodalton heat shock protein gene expression. Mol. 11. Glaichenhaus, N., and Cuzin, F. A role for ID repetitive sequences in growth- and Cell. Biol., 9: 3567–3570, 1989. transformation-dependent regulation of gene expression in rat fibroblasts. Cell, 50: 35. Conroy, S. E., and Latchman, D. S. Do heat shock proteins have a role in breast 1081–1089, 1987. cancer? Br. J. Cancer, 74: 717–721, 1996. 12. Won, K. A., Schumacher, R. J., Farr, G. W., Horwich, A. L., and Reed, S. I. 36. Shyamala, G., Schweitzer, M., and Ullrich, S. J. Relationship between 90-kilodalton Maturation of human cyclin E requires the function of eukaryotic chaperonin CCT. heat shock protein, estrogen receptor, and progesterone receptor in human mammary Mol. Cell. Biol., 18: 7584–7589, 1998. tumors. Breast Cancer Res. Treat., 26: 95–100, 1993. 13. Allende, M. L., Amsterdam, A., Becker, T., Kawakami, K., Gaiano, N., and Hopkins, 37. Rutherford, S. L., and Lindquist, S. Hsp90 as a capacitor for morphological evolution. N. Insertional mutagenesis in zebrafish identifies two novel genes, Pescadillo and Nature (Lond.), 396: 336–342, 1998. dead eye, essential for embryonic development. Genes Dev., 10: 3141–3155, 1996. 38. Peattie, D. A., Harding, M. W., Fleming, M. A., DeCenzo, M. T., Lippke, J. A., 14. Bork, P., Hofmann, K., Bucher, P., Neuwald, A. F., Altschul, S. F., and Koonin, E. V. Livingston, D. J., and Benasutti, M. Expression and characterization of human A superfamily of conserved domains in DNA damage-responsive cell cycle check- FKBP52, an immunophilin that associates with the 90-kDa heat shock protein and is point proteins. FASEB J., 11: 68–76, 1997. a component of steroid receptor complexes. Proc. Natl. Acad. Sci. USA, 89: 10974– 15. DeFranco, D. B., Ramakrishnan, C., and Tang, Y. Molecular chaperones and sub- 10978, 1992. cellular trafficking of steroid receptors. J. Steroid. Biochem. Mol. Biol., 65: 51–58, 39. Keyomarsi, K., and Pardee, A. B. Redundant cyclin overexpression and gene ampli- 1998. fication in breast cancer cells. Proc. Natl. Acad. Sci. USA, 90: 1112–1116, 1993. 16. Ren, M., Drivas, G., D’Eustachio, P., and Rush, M. G. Ran/TC4: a small nuclear 40. Wagner, G. F., Guiraudon, C. C., Milliken, C., and Copp, D. H. Immunological and GTP-binding protein that regulates DNA synthesis. J. Cell Biol., 120: 313–323, 1993. biological evidence for a stanniocalcin-like hormone in human kidney. Proc. Natl. 17. Rasmussen, C. D., Means, R. L., Lu, K. P., May, G. S., and Means, A. R. Charac- Acad. Sci. USA, 92: 1871–1875, 1995. terization and expression of the unique calmodulin gene of Aspergillus nidulans. 41. Varghese, R., Wong, C. K., Deol, H., Wagner, G. F., and DiMattia, G. E. Compar- J. Biol. Chem., 265: 13767–13775, 1990. ative analysis of mammalian stanniocalcin genes. Endocrinology, 139: 4714–4725, 18. Migliaccio, A., Rotondi, A., and Auricchio, F. Calmodulin-stimulated phosphoryla- 1998. tion of 17 ␤-estradiol receptor on tyrosine. Proc. Natl. Acad. Sci. USA, 81: 5921– 42. Yoshiko, Y., and Maeda, N. In situ hybridization analysis of stanniocalcin mRNA 5925, 1984. expressing cells in the mouse kidney. Mol. Cell. Endocrinol., 141: 37–40, 1998. 19. Biswas, D. K., Reddy, P. V., Pickard, M., Makkad, B., Pettit, N., and Pardee, A. B. 43. Yoshiko, Y., Son, A., Maeda, S., Igarashi, A., Takano, S., Hu, J., and Maeda, N. Calmodulin is essential for estrogen receptor interaction with its motif and activation Evidence for stanniocalcin gene expression in mammalian bone. Endocrinology, 140: of responsive promoter. J. Biol. Chem., 273: 33817–33824, 1998. 1869–1874, 1999. 20. Spataro, V., Norbury, C., and Harris, A. L. The ubiquitin-proteasome pathway in 44. Turner, I. M., Saunders, P. T., Shimasaki, S., and Hillier, S. G. Regulation of inhibin cancer. Br. J. Cancer, 77: 448–455, 1998. subunit gene expression by FSH and estradiol in cultured rat granulosa cells. Endo- 21. Nawaz, Z., Lonard, D. M., Dennis, A. P., Smith, C. L., and O’Malley, B. W. crinology, 125: 2790–2792, 1989. Proteasome-dependent degradation of the human estrogen receptor. Proc. Natl. Acad. 45. Kalkhoven, E., Roelen, B. A., de Winter, J. P., Mummery, C. L., van den Eijnden-van Sci. USA, 96: 1858–1862, 1999. Raaij, A. J., van der Saag, P. T., and van der Burg, B. Resistance to transforming 22. Chang, A. C., and Reddel, R. R. Identification of a second stanniocalcin cDNA in growth factor ␤ and activin due to reduced receptor expression in human breast tumor mouse and human: stanniocalcin 2. Mol. Cell. Endocrinol., 141: 95–99, 1998. cell lines. Cell Growth Differ., 6: 1151–1161, 1995. 23. Ishibashi, K., Miyamoto, K., Taketani, Y., Morita, K., Takeda, E., Sasaki, S., and 46. Ying, S. Y., and Zhang, Z. Expression and localization of inhibin/activin subunits and Imai, M. Molecular cloning of a second human stanniocalcin homologue (STC2). activin receptors in MCF-7 cells, a human breast cancer cell line. Breast Cancer Res. Biochem. Biophys. Res. Commun., 250: 252–258, 1998. Treat., 37: 151–160, 1996.

5983

Downloaded from cancerres.aacrjournals.org on September 26, 2021. © 2000 American Association for Cancer Research. Effects of Estrogen on Global Gene Expression: Identification of Novel Targets of Estrogen Action

April H. Charpentier, Andrzej K. Bednarek, Rachael L. Daniel, et al.

Cancer Res 2000;60:5977-5983.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/60/21/5977

Cited articles This article cites 46 articles, 19 of which you can access for free at: http://cancerres.aacrjournals.org/content/60/21/5977.full#ref-list-1

Citing articles This article has been cited by 53 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/60/21/5977.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/60/21/5977. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.