..4 ,. 90.1. wfiurftth’n. .alfihfia: é dud Wémmmmnnmma J . ... 5.43:-.. ... 11.3.. . an: . .. fiffihflq A q ....u wuukfiafln khan-4...»... L ...(1 A69.» .14. .....th 4. mwglfiwfiwfigmmfiq wag» . .mvwwm 3"?" ’V -O".‘ I- " 7 Nu v. J.’.¢ . \ i l LIBRARY ‘ .14 Michigan State University This is to certify that the dissertation entitled TISSUE-SPECIFIC IN VITRO AND IN VIVO EVALUATION OF TAMOXIFEN-MEDIATED GENE EXPRESSION presented by CORA JUNG-YEE FONG has been accepted towards fulfillment of the requirements for the Ph. D degree in Biochemistry and Molecular Biology fly 4 UProt’éssor’s Signature U/az/rfiif- Date MSU is an afiinnative-action, equal-opportunity employer -.— o.-.-----.-.-‘-1-.-.-.-.-.---.—n—.-—A-c-n-n.-.-------.-.-. PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 6/07 p:lCIRC/DateDue.indd-p.1 TISSUE-SPECIFIC IN VITRO AND IN VIVO EVALUATION OF TAMOXIFEN- MEDIATED GENE EXPRESSION By Cora Jung-Yee Fong A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Biochemistry and Molecular Biology 2007 ABSTRACT TISSUE-SPECIFIC IN VITRO AND IN VIVO EVALUATION OF TAMOXIFEN- MEDIATED GENE EXPRESSION By Cora Jung-Yee Fong Estrogenic endocrine disrupting compounds (EEDCs) are an environmental and human health concern and thus have become a focus for risk assessment characterization. The United States Environmental Protection Agency (US EPA) is considering screening 87, 000 chemicals for their potential endocrine disrupting properties and is currently developing assays for this purpose. An in vitro hepatic mouse tissue culture model, Hepa-1c1c7, was thus evaluated as a system to examine estrogenic gene expression responses. Hepa-1c1c7 cells exhibit gene expression changes in response to estrogen treatment, which correlate with those of an in vivo system, such as cytoskeletal reorganization and cholesterol metabolism. However, the magnitude of the differential gene expression responses did not warrant further examination with less potent estrogenic compounds. The rodent uterotrophic assay has historically been used to evaluate estrogenic compounds and extensive literature has examined the effects of the potent estrogen mimic, 17a-ethynylestradiol (EE), on the uterus. This provided an excellent foundation for the characterization of tamoxifen (TAM)-mediated effects. The pharmaceutical tamoxifen is an estrogen receptor (ER) ligand which exhibits its anti-breast cancer effects by competing with estradiol for ER binding. In contrast, TAM elicits an estrogenic effect in endometrial tissue by promoting proliferation. Its seemingly dual nature classifies it as a selective estrogen receptor modulator (SERM). Comprehensive microarray analysis complemented with physiological and histological data illustrated that TAM elicits gene expression changes which closely resembles those of EE, although for the most part muted in magnitude. In addition, EE-specific genes were identified which were consistent with the increased EE-mediated uterotrophic response compared to that of TAM. Interestingly, historical studies have shown that mixed treatment of EE and TAM results in an inhibition of EE-mediated uterotrophy. An experimental design was developed to examine whether the mixed treatment physiology was due to global inhibition of gene expression. Surprisingly, only 10% of the genes exhibited a mixture-mediated response which differed from that of EE alone. These differential responses represented genes involved in cell growth and proliferation and were consistent with the inhibited physiology observed. These data suggest that TAM only modifies the expression of a subset of genes involved in eliciting a full uterotrophic effect under mixture conditions with EE and warrants investigation into the mechanisms of regulation involved. ACKNOWLEDGEMENTS This graduate school experience has shaped me in ways that I would never have anticipated. My anxiousness was soon overwhelmed with frustration, but I look back on this journey with much gratitude and satisfaction. I wouldn’t have changed a thing. To my advisor, Tim Zacharewski, I thank you for taking a chance on a green but eager undergraduate. He has provided me with an abundance of new learning opportunities and his academic, professional and personal guidance has been invaluable. My time at MSU has been incredibly rewarding from all the challenges which were presented to me. Thank your for your patience. My thanks also extend to the Zacharewski family; Jana, Lara and Nicholas, I’m going to miss all of you. I could not have completed my research without the help and support of the ever changing members of the Zacharewski lab. Thanks to Robert Halgren, Yue-wern Huang, Jason Matthews and Mark Fielden for providing me a good experimental foundation. Darrell Boverhof, Josh Kwekel, Ed Dere, Suntae Kim, Ania Kopec and Qi Ding, without your assistance, this book would not exist. A special thanks to Kirsten Fertuck and Lyle Burgoon for helping me see the humour in everything and maintaining my sanity, particularly at the beginning of the trip. Of course I must acknowledge the endless number of co-op and undergraduate students who have helped out with projects, sought out advice or who were just fun to hang outwith; I wish you all the best. To my committee members, Drs. David Arnosti, Pam Fraker, Patty Ganey and Bill Henry: I am grateful for all your encouragement and patience. The input you have provided over these years have helped me to develop and hone my skills as a scientist. Thanks also to the excellent departmental staff for always being helpful in keeping all sorts of paperwork straight. A great thank you goes out to my friends, old and new. Thank you, Eugenie, Nadia, Tori, Konrad & Sara, Marina, Len and Tao, for being around for the ride despite my infrequent calls and visits. Yes, I’m done now. Gabby, I am going to miss our coffee, good eating and music appreciation nights; you have been such a great support this past year. My parents have been rooting for me throughout my entire undergraduate and graduate career; thank you. Many thanks, to my sister, Grace, who has recently become a great source of entertainment. You’ve been a great help. Finally, I would like to express my gratitude to Jeremy Burt. Thank you for always being able to make me see the lighter side of things. He has been nothing but supportive, helpful and encouraging. Thanks for sticking it out for the long haul. Ihug TABLE OF CONTENTS LIST OF TABLES ..................................................................................... viii LIST OF FIGURES ..................................................................................... ix ABBREVIATIONS ..................................................................................... xii CHAPTER 1 GENE EXPRESSION RESPONSES ELICITED BY ESTROGENIC COMPOUNDS IN LIVER AND UTERUS ........................... 1 INTRODUCTION ............................................................................... 1 ESTROGENIC COMPOUNDS AND THEIR EFFECTS ..................... 3 ESTROGENS AND ESTROGEN SIGNALING .......................... 3 Nuclear Receptor Signaling .............................................. 4 DNA-Response Elements ................................................. 6 Co-regulatory Proteins ...................................................... 8 Non-classical Signaling Mechanisms ................................ 9 ESTROGENIC ENDOCRINE DISRUPTORS .......................... 10 TAMOXIFEN AS AN ENDOCRINE DISRUPTOR ................... 11 RATIONALE, HYPOTHESIS AND SPECIFIC AIMS ........................ 15 PROJECT 1: ESTROGENIC ACTIVITY ON THE LIVER ................ 15 THE HEPATIC SYSTEM ......................................................... 15 IN VITRO HEPATIC MODEL SYSTEM ................................... 16 Hypothesis ...................................................................... 17 Specific Aims .................................................................. 17 PROJECT 2: THE EFFECT OF TAMOXIFEN ON THE UTERUS... 18 Hypothesis ...................................................................... 18 Specific Aims .................................................................. 19 ESTROGEN ACTIVITY ON THE UTERUS ............................. 19 The Menstrual Cycle ....................................................... 19 The Estrous Cycle .......................................................... 20 IN VIVO MODEL AND THE UTEROTROPHIC ASSAY .......... 21 TRANSGENIC MODELS TO EXAMINE ESTROGEN RECEPTOR SIGNALING ........................................................ 21 CHEMICAL MIXTURES ................................................................... 23 MIXTURE EFFECTS ............................................................... 23 vi CHAPTER 2 EFFECTS OF CULTURE CONDITIONS ON ESTROGEN-MEDIATED HEPATIC IN VITRO GENE EXPRESSION AND CORRELATION TO IN VIVO RESPONSES ............................................................................. 27 ABSTRACT ..................................................................................... 27 INTRODUCTION ............................................................................. 29 MATERIALS AND METHODS ......................................................... 31 RESULTS ........................................................................................ 43 DISCUSSION .................................................................................. 64 CHAPTER 3 COMPARATIVE TEMPORAL AND DOSE-DEPENDENT MORPHOLOGICAL AND TRANSCRIPTIONAL UTERINE EFFECTS ELICITED BY TAMOXIFEN AND ETHYNYLESTRADIOL IN IMMATURE, OVARIECTOMIZED MICE .................................................. 72 ABSTRACT ..................................................................................... 72 INTRODUCTION ............................................................................. 74 MATERIALS AND METHODS ......................................................... 76 RESULTS ........................................................................................ 84 DISCUSSION ................................................................................ 111 CHAPTER 4 MIXTURE EFFECTS OF TAMOXIF EN AND ETHYNYLESTRADIOL ON GENE EXPRESSION IN IMMATURE, OVARIECTOMIZED MICE UTERUS .............................................................................................. 120 ABSTRACT ................................................................................... 120 INTRODUCTION ........................................................................... 122 MATERIALS AND METHODS ....................................................... 127 RESULTS ...................................................................................... 135 DISCUSSION ................................................................................ 159 CHAPTER 5 SUMMARY AND FUTURE PERSPECTIVES ........................................ 164 REFERENCES ...................................................................................... 167 vii LIST OF TABLES Chapter 2 Table 1. QRT-PCR primer list ................................................................................ 39 Table 2. Genes exhibiting high temporal gene expression and activity correlations, (p 2 0.5) ............................................................................... 60 Chapter 3 Table 1. TAM- and EE-induced uterine morphometric changes ............................. 87 Table 2. Classification of TAM and EE commonly active annotated genes ......... 102 Chapter 4 Table 1. Histological evaluations of treated uterine sections (n = 5) .................... 143 Table 2. MIX-modified, EE-induced gene list generation ..................................... 150 Table 3. MIX-modified, EE-induced gene classifications ...................................... 152 viii LIST OF FIGURES Chapter 1 Figure 1. Estrogen receptor domain structure ................................................ Figure 2. Classical mechanism of estrogen receptor signaling ...................... Figure 3. Metabolism of tamoxifen ................................................................. Chapter 2 Figure 1. Independent reference design for microarray hybridization ............ Figure 2. MTT and cell count assessment of Hepa-1c1c7 cells in serum free and stripped serum supplemented medium .................................... Figure 3. Temporal gene expression patterns: E2 in serum free media ......... Figure 4. Temporal gene expression patterns: E2 in stripped serum media .. Figure 5. Systematic comparison of in vitro and in vivo active gene lists ....... Figure 6. In vitro vs. in vivo significance P1(t) and activity index correlations Chapter 3 Figure 1. Tamoxifen-induced dose dependent and temporal changes in ........... 5 ........... 7 ......... 12 ......... 35 ......... 44 ......... 48 ......... 50 ......... 54 ......... 57 uterine weight ........................................................................................... 85 Figure 2. Uterine histology ............................................................................. ......... 89 Figure 3. Tamoxifen-induced temporal gene expression patterns ........................... 92 Figure 4. Quantitative real-time PCR verification of selected TAM-induced genes ....................................................................................................... 95 Figure 5. lmmunohistochemical detection of differential Pcna protein levels due to TAM .............................................................................................. 97 Figure 6. Temporal comparison of genes commonly activated by TAM and EE 100 Figure 7. Examples of TAM and EE differential gene expression classifications .. 104 Figure 8. Identification of unique EE and TAM differentially expressed genes ...... 106 Figure 9. Temporal expression profiles of TAM and EE-specific genes ................ 109 Chapter 4 Figure 1. Microarray hybridization design and uterotrophic assay treatment design .................................................................................................... 125 Figure 2. Dose finding: uterotrophic inhibition ....................................................... 136 Figure 3. Treatment induced uterotrophy .............................................................. 139 Figure 4. Morphometric changes elicited by EE, TAM and Mixture ....................... 140 Figure 5. Histological observations ....................................................................... 144 Figure 6. Temporal LC/MS/MS analysis of hepatic TAM and 4OH-TAM levels ..... 147 Figure 7. EE-mediated gene expression affected by TAM cotreatment ................ 151 Figure 8. Quantitative real-time PCR verification .................................................. 154 Figure 9. MIX-modified, EE-mediated gene profiles ........................................................ 158 xi 4OH-TAM AhR AF CA DBD DCC-FBS DDT DES DI DMT E2 EDC EE EEDC ELISA ER ERE ERKO FSH GnRH HAT ABBREVIATIONS 4-hydroxytamoxifen arylhydrocarbon receptor activation function concentration addition DNA binding domain dextran charcoal-coated fetal bovine serum dichloro-diphenyI-trichloroethane diethylstilbestrol displacement index N—desmethyltamoxifen 1 7beta-estrad iol endocrine disrupting compound 17alpha-ethynylestradiol estrogenic endocrine disrupting compound enzyme-linked immunosorbant assay estrogen receptor estrogen response element estrogen receptor knock out follicule stimulating hormone gonadotropin releasing hormone histone acetylase xii HMGB KO LBD LC/MS LECH LH MAPK MIX MTT NERKI NR NTC PAH PCB PND QRT-PCR SERM SRC TAM TCT US EPA UWW WT high mobility group B knock out ligand binding domain liquid chromatography/mass spectroscopy luminal epithelial cell height Iuteinizing hormone mitogen-activated protein kinase mixture treatment (EE plus TAM) 3-[4,5-dimethylthiazol-Z-yI]-2,5-diphenyl tetrazolium bromide non-classical estrogen receptor knock-in nuclear receptor no template control polyaromatic hydrocarbon polychlorinated biphenyl post-natal day quantitative real time-polymerase chain reaction selective estrogen receptor modulator steroid receptor coactivator tamoxifen Toxicogenomics Correlation Tool United States Environmental Protection Agency uterine wet weight wild type xiii CHAPTER 1 GENE EXPRESSION RESPONSES ELICITED BY ESTROGENIC COMPOUNDS IN LIVER AND UTERUS INTRODUCTION Estrogenic compounds and their impact on human health are high priorities in the research field. In 1996, enactment of the Food Quality Protection Act (FQPA) and amendments to the Safe Drinking Water Act (SDWA) required the US. Environmental Protection Agency (EPA) to develop programs for the screening of endocrine disrupting compounds (EDCs). These have been defined as synthetic or natural chemicals which have an effect on humans that is similar to those produced by naturally occurring hormones—specifically estrogen, androgen and thyroid (1). These legislative changes arose in response to observed wildlife abnormalities due to chemical exposures such as the dichloro- diphenyI—trichloroethane (DDT)-mediated feminization of male gulls (2), increases in female estrogen levels and decreases in male testosterone in alligators exposed to dicofol and DDT (3) and deformities in Xenopus embryos associated with high levels of chemical agents (4). Human exposure to EDCs is also an area of concern. For instance, world- wide polychlorinated biphenyl (PCB) contamination of fish (5-7) and led to studies assessing their toxicity and establishing safe human consumption guidelines (8). However, some compounds are examined due to their potential beneficial properties. Phytoestrogens are natural plant products demonstrated to have estrogenic properties. Genistein is an isoflavone extracted from soybean. It is structurally similar to E2 and also exerts its mild estrogenic effects through ER binding (9). In Eastern Asia, where there is a high dietary intake of soy products, the incidence of breast cancer in women is lower compared to their Western counterparts (reviewed in (10)). Phytoestrogens have been used as a natural substitute in hormone replacement therapies but have caused concern with respect to being a potential breast cancer promoter (11). In response to these environmental and health concerns, the Endocrine Disrupter Screening and Testing Advisory Committee (EDSTAC) was formed to aid in the development of an EPA program to make informed regulatory decisions on compounds. A tiered approach was adopted where by Tier 1 screening would aid in the identification of compounds which may interact with the endocrine system. Tier 2 testing would determine any adverse effects caused by the compound as well as characterize the relationship between dose and effect. At this time, compounds are being prioritized for testing and assays are being validated for the screening and testing phases. ESTROGENIC COMPOUNDS AND THEIR EFFECTS ESTROGENS AND ESTROGEN SIGNALING Endogenous estrogens are comprised of a series of steroidal compounds which are primarily associated with the regulation of female growth and development. Deficiency in humans result in ambiguous external genitalia at birth, lack of maturation of reproductive organs, polycystic ovaries (12), and delayed bone structure development with a prolonged linear bone growth (13). Estrogens also play a role in male growth and development. Estrogen deficient males show no early signs of deficiency, but are diagnosed as adults with delayed bone structure development alongside prolonged linear bone growth (reviewed in (14)). Other studies have shown estrogen to be important in the maintenance of the cardiovascular, hepatic, skeletal and renal systems as well as promoting healthy lipid profiles (reviewed in (15)). 178-Estradiol (E2) is the most abundant estrogen found in females and exerts its effects through the estrogen receptor (ER). In mammalian systems, two ER isoforms have been identified. ERa is found in uterine, ovarian, mammary, vaginal, epididymal, testicular, hepatic, adrenal and renal tissue while ERB is found in ovarian, prostate, pulmonary and cerebral tissue (16,17). To further characterize the roles of each receptor, transgenic knockout mice have been developed for both receptors—orERKO and BERKO. aERKO females exhibit immature uterine structure, enlarged, polycystic ovaries, poor mammary duct development, and smaller stature, while males experience progressive testicular tubule degradation, nonfunctional sperm, delayed cardiac depolarization, lower bone density and attenuated aggressive behavior. BERKO animals generally demonstrate normal organ structure and development, although female mice fertility is decreased (reviewed in (18)). Nuclear Receptor Signaling The ER belongs to a class of ligand activated transcription factors identified as the nuclear receptor (NR) superfamily. NR superfamily members share five distinct domain structures (19). Domain A/B contains a constitutively active activation function domain (AF 1) which interacts with other transcription cofactors. Domain C is a highly conserved DNA-binding domain (DBD) which seeks out specific response elements in DNA enhancer regions. Domain D is a hinge region separating the DBD from the ligand-binding domain (LBD), domain E, and also serves as a ligand-dependent transactivation region, AF2. ER also contains an F domain which has been implicated in ligand-dependent differential transcription activity (20). Domains A/B, D and F have the lowest similarity between ER alpha and beta isoforms, while the DBD is highly conserved with 97% similarity (21) (Figure 1). The classical mode of action of NRs involves dimerization with a partner protein followed by dimerized-complex binding to specific DNA sequences in the promoter regions of responsive genes. Co-regulating transcription factors and Figure 1 Estrogen receptor domain structure Estrogen receptor isoforms alpha and beta contain five major domains where the DNA binding domain (DBD) and ligand binding domain (LBD) share the highest amino acid sequence similarity. Activation function domains, AF1 and AF2, are located in the N8 and E domains, respectively. ERor A / B c D E F DBD LBD 97% 60% ERB A / B c o E F transcriptional machinery may then be recruited to the gene, leading to changes in basal transcriptional activation (Figure 2). Estrogens enter the cell and bind to nuclear residing ER. Ligand binding induces a conformational change causing the release of chaperone proteins such as heat shock protein 90 (Hsp90) (22). Activated ER complexes homodimerize allowing them to bind specific promoter sequences, known as estrogen response elements (EREs), of responsive genes. This DNA-bound complex may then influence changes in the transcriptional state of proximal genes through interactions with the basal transcriptional machinery and transcriptional cofactors (mechanism reviewed in (23)). DNA-Response Elements Response elements are short DNA sequences, found in the promoter and enhancer regions of primary responsive genes, which activated NR bind to with high affinity. NR specificity is determined by the nucleotide sequence. Through in vitro studies, a consensus ERE sequence (5’-GGTCAnnnTGACC-3’, where n represents any nucleotide) has been identified (24). However, examination of other estrogen responsive genes have identified nucleotide substitutions in the ERE (reviewed in (25)). Bioinformatic and high-throughput approaches have also identified putative EREs in the mouse and human genome (26,27). Interactions between EREs and ligand-bound ER may also induce receptor conformation changes. Peptidase digestion experiments have Figure 2 Classical mechanism of estrogen receptor signaling Estrogen receptor ligands diffuse into the cell and bind to the estrogen receptor located in the nucleus. Ligand binding induces a conformational change to release stabilizing chaperone proteins, such as Hsp90, and allow for dimerization - of activated receptors. ER dimers may then bind to sequence specific estrogen response elements (ERE) in the promoter region of estrogen responsive genes, recruit co—regulating proteins and transcriptional machinery to drive changes in mRNA expression. 9 ligand Hsp90 I11 .600... 0 000m .00... O... .0 cellular and physiological responses demonstrated that ligand-bound ER, incubated with different ERE sequenCes, exhibit varied electrophoretic fragment patterns (28). Similarly, phage-display ELISA assays have identified different exposed epitopes on the active ER complex when it is bound to different ERE sequences (29). Thus, ER conformation affects what subset of co-regulating proteins can be recruited and subsequently influence transcriptional changes to proximal genes. Co-regulatory Proteins Co-regulating proteins pose as a bridge between the AF2 domain of DNA- bound nuclear receptors and basal transcriptional machinery. The subset of coregulators recruited to a NR is determined in part by the specific receptor, the bound ligand, the response element sequence bound by the NR and the cell- specific expression of the coregulators (30,31). Collectively, a variety of factors play a role in nuclear receptor-mediated changes in transcription. The steroid receptor coactivator (SRC) protein family has been extensively studied. SRC proteins have a receptor-interacting domain containing two or three short, helical LXXLL motifs, where L represents leucine and X represents any amino acid (32). Some members possess histone acetylase (HAT) activity, but all recruit additional coactivators with intrinsic HAT activity, which is important in the enhancement of nuclear receptor activity through chromatin structure remodeling (reviewed in (33)). The complex of coactivator proteins facilitates the recruitment of the basal transcriptional machinery to the responsive target genes. Coregulators have also demonstrated their influence on nuclear receptor binding to specific response elements. Affinity binding assays have demonstrated that the presence of high mobility group B (HMGB) coactivators increases the affinity of estrogen-bound ER to consensus ERE sequences, and that different members of the co-regulatory family also affect the degree of affinity (34). Non-classical Signaling Mechanisms In addition to the classical NR signaling, ER can elicit activity using other pathways (23). Growth factors may initiate MAPK signaling pathways to phosphorylate specific serine residues found in the ER AF1 domain, allowing interaction with coactivators to modify gene expression (35). ER activated in this manner is capable of tethering to Fos/Jun complexes at AP-1 sites and Sp1 complexes at GC-rich regions to drive differential transcription (36,37). Recently, it has been proposed that ERs can also exist in a membrane-bound state. This form has been proposed to activate signaling cascades which are too rapid to involve genomic responses, such as the influx of extracellular calcium by mast cells (38). Estrogen signaling is a complex network and can thus be interrupted at various nodes. ESTROGENIC ENDOCRINE DISRUPTORS Xenobiotic compounds which disrupt normal estrogen signaling are known as estrogenic endocrine disruptors (EEDs). EEDs are structurally diverse and found as natural products, pharmaceuticals, industrial compounds, pesticides and other environmental contaminants. Disruption of estrogen signaling may affect enzymes involved in estrogen production or metabolism, influence ER expression levels or compete with endogenous estrogens for ER binding (reviewed in (10,39)). Although the EPA is focused on EED exposure through food and water (40), some research is focused on pharmaceuticals which are directed at disrupting endocrine systems. For example, 17a-ethnylestradiol (EE) is the main component in female contraceptives. It is structurally similar to E2 and its effects mimic that of endogenous estrogen in vivo and in vitro (41). Diethylstilbestrol (DES) is another ER-binding pharmaceutical, which was first prescribed to pregnant women in the 1940s to prevent miscarriages. In 1971, it was associated with the development of vaginal cancer in female offspring to women prescribed DES (42) and further research has demonstrated the teratogenic properties of DES. 10 TAMOXIFEN AS AN ENDOCRINE DISRUPTOR Tamoxifen (TAM) was first developed in the late 1960s and initially prescribed as a fertility drug (43). TAM was subsequently examined for potential anti-cancer activity, an application for which it proved successful (reviewed in (44)). TAM was approved in 1977 for treatment of ER-positive breast cancer. TAM exerts its effects through direct binding to the ER, thus competing with endogenous estrogens that otherwise promote proliferation and cancer progression (45). Consequently, TAM is less effective against ER-negative breast cancers. TAM is effective for suppressing cancer recurrence by 50% as well as inhibiting contralateral primary breast cancer. In addition, women identified at high risk for breast cancer have a significantly reduced risk of developing cancer with prophylactic TAM treatment (46). Three TAM metabolites also exhibit antiestrogenic activity, 4- hydroxytamoxifen (4OH-TAM), N-desmethyltamoxifen (DMT) and 4-hydroxy-N- desmethyltamoxifen (endoxifen) (Figure 3). 4OH-TAM is a potent metabolite due to its high ER binding affinity (47-50). DMT exhibits low ER binding affinity (51) but is the major human metabolite (52). Recent studies with endoxifen suggest that it my be more potent than 4OH-TAM (53,54). Moreover, human plasma concentrations indicate that endoxifen levels (12.4 ng/mL) are greater than that of 4OH-TAM (1.1 ng/mL) (55). 11 Figure3 Metabolism of tamoxifen Tamoxifen is metabolized into bioactive metabolites 4-hydroxytamoxifen, N- desmethyltamoxifen, and 4-hydroxy-N-desmethyltamoxifen. tamoxifen 4—hydroxytamoxifen 3 O CYP2D6 -—-> 0 ° ,JO \N/I/ \ l CYP3A4/5 CYP3A4/5 l I 3 O cvpzoe ’ O o f0 "‘74 I , H \T I . 4-hydroxy- N-desmethyltamoxrfen Ill-desmethyltamoxifen 12 TAM metabolism differs between species. Studies between rodents and humans have shown that TAM N-oxide, 4OH-TAM and DMT are the predominant metabolites in the mouse, while DMT is the major human metabolite in microsomal studies (56,57). In rodents, the levels and rates of TAM metabolism to 4OH-TAM and DMT were significantly different in the rat and mouse, where the rat metabolite profile more closely resembles human profiles (52). Such studies illustrate that differences metabolism between models should be considered in extrapolations for risk assessment. Despite the high therapeutic index of TAM, its adverse effects include a two-fold increase in risk to develop endometrial cancer (58). Cases of endometrial cancer have been reported as early as two years after commencement of treatment (59); however, it is unclear whether TAM is an initiator in the carcinogenesis process. Due to the seemingly opposing effects in mammary and endometrial tissues, TAM is classified as a selective estrogen receptor modulator (SERM). SERMs are pharmaceuticals with differential tissue effects and are often prescribed for specific conditions. Numerous factors influence the effects of a SERM-bound receptor such as tissue-specific ER isoform expression levels, ligand-induced ER topology, chromatin structure, and coregulator protein expression and distribution (46,60-62). A well studied factor in the SERM property of TAM is the conformation it confers on the ER. The ER-LBD is a 12- helical structure where the position of helix-12 has been identified as a key factor 13 in differentiating ligand-dependent agonistic and antagonistic effects (63). Helix- 12 acts as a lid to encase the bound ligand in the LBD. Full agonists, such as E2 and EE, induce a conformational change that closes helix-12 over the ligand binding pocket, providing an interface for coregulator protein interactions. Ligands classified as partial agonists typically have bulky side groups that protrude from the pocket displacing helix-12 from its agonist position affecting coactivator docking (64). In the case of full agonist lCl 182,780, binding causes conformational changes exposing hydrophobic surfaces that target the ER for degradation (65). Although TAM-binding causes ER-LBD to adopt a conformation with an unfavourable helix-12 position (66), which may be important in its role as an anti-estrogen in the mammary, it has been suggested that high levels of expressed steroid receptor coactivator 1 (SRC1) in uterine tissue may be a determinant in the agonistic effects of TAM in the uterus (31). However, the influence of these factors on gene expression is poorly understood and warrants further investigation. 14 RATIONALE, HYPOTHESIS AND SPECIFIC AIMS The research presented utilizes a microarray approach to comprehensively examine gene expression changes elicited by EE and TAM, alone, as well as in combination. PROJECT 1: ESTROGENIC ACTIVITY ON THE LIVER Reproductive tissues have been the focus of the majority of estrogenic studies, although many tissues not classically regarded as targets of estrogen also exhibit gene expression changes in response to estrogens. ERE-mediated transgenic mouse studies that can monitor ER-mediated gene expression have identified the liver to be one of the most estrogen-responsive tissues (67,68). Modulation of lipid transport and metabolism by estrogens in the liver has been well documented (69,70), although its mechanisms have not been fully elucidated. Xenobiotic compounds are delivered to hepatic tissue upon oral exposure; thus, it is important to examine the effects modulated by exposure EEDCs THE HEPATIC SYSTEM Although the liver is not a classical estrogen-responsive tissue, it expresses ER (16) and exhibits changes in gene expression in response to estrogens (67,68). Studies of estrogens on the liver have focused on the biliary system, where primary biliary cirrhosis (autoimmune destruction of liver bile ducts) is more prevalent in females (71), as well as on lipid profiles, where 15 hormone replacement therapy decreases cholesterol, but increases triglyceride levels (72). Microarray analysis of in vivo hepatic responses to estrogen identified changes in gene expression associated with a wide array of pathways including proliferation, cytoskeletal organization, oxidative stress and lipid metabolism (73). In response to EDSTAC’s prioritized chemical screening and testing recommendations, EPA implemented the Endocrine Disruptor Screening Program (EDSP) to develop testing assays. In addition to receptor binding studies, in vitro transcriptional assays were to be developed for compound screening. Due to its estrogen responsiveness, investigation of a comparable hepatic in vitro model was warranted. IN VITRO HEPATIC MODEL SYSTEM Cell culture models are advantageous as they reduce animal experimentation and are amenable to high-throughput testing. Homogeneous cells are expected to exhibit less variability and facilitate the investigation of cell- type specific effects which may otherwise be masked in a heterogeneous tissue. Mouse Hepa-1c1c7 hepatoma cells were selected as this line is commonly used in the field of toxicology, particularly in aryl hydrocarbon receptor (AhR) mechanism-related studies. (74,75). This model was derived from a BW 7756 hepatoma which arose in a C57L mouse and propagated in CS7L/J mice (76). Hepa-1c1c7 cells possess active ERs (77,78) and retain several liver-specific 16 functions such as synthesis and secretion of albumin (76) and transferrin (79) as well as xenobiotic detoxification activity (80). Hypothesis: Estrogenic compounds elicit species conserved time- and dose-dependent hepatic gene expression profiles between in vitro hepatoma models. Specific Aims The following specific aims were proposed to address the hypothesis: 1) Establish baseline gene expression in response to E2 in mouse, rat and human hepatoma cell lines. 2) Establish an estrogenic expression fingerprint by examining common gene expression changes elicited by structurally diverse EEDCs in a selected hepatoma model. 3) Propose an estrogen receptor-regulated biological response network. As detailed in Chapter 2, responses associated with proliferation, cytoskeletal reorganization, cholesterol transport and metabolism, fatty acid metabolism, and oxidative stress were well conserved between various models and the Hepa-1c1c7 cells. Some genes demonstrated common activation between estrogen—treated liver of CS7BU6 mice and Hepa-1c1c7 cells and exhibited temporally shifted expression patterns. Despite these similarities, the 17 magnitudes of gene expression changes elicited by the in vitro model were weak and did not warrant further development as a screening or testing system (81). Although in vivo and tissue-culture in vitro models demonstrated limited overlap in estrogenic responses, development of other in vitro systems may prove to be better in vivo predictors such as tissue slices (82). Three- dimensional architecture and signaling between different cell types may be required for accurate gene expression profile responses. PROJECT 2: THE EFFECT OF TAMOXIFEN ON THE UTERUS Due to the difficulties encountered in Project 1, a more reliable estrogen responsive model was selected—the immature, ovariectomized mouse uterus. Tamoxifen was selected as an ER ligand of interest due to its SERM properties. Tamoxifen and its role in breast cancer prevention have been well studied (reviewed in (83)); however, its increased risk in endometrial cancer remains poorly understood. In addition, previous studies have demonstrated that co- treatment of TAM and estrogen result in repression of estrogen-induced, rodent uterotrophy (84,85). However, the molecular basis of the repression has yet to be fully elucidated and a comprehensive experimental design was developed to associate gene expression to the uterotrophic response. Hypothesis Tamoxifen antagonizes EE-mediated uterotrophic responses associated with globally antagonized EE—induced gene responses. 18 Specific Aims The following specific aims were proposed to address the hypothesis: 1) Establish baseline dose response and temporal gene expression profiles following oral exposure of TAM in the 057BU6 mouse uterus. 2) Identify the optimal doses of TAM resulting in maximal inhibition of EE- induced uterotrophy. 3) Identify EE-elicited temporal gene expression affected by TAM that may contribute to the inhibition of the induced uterotrophic response. ESTROGEN ACTIVITY ON THE UTERUS Estrogen plays an integral role in the maintenance of the female reproductive cycle. The Menstrual Cycle In female primates and humans the effects of estrogen signaling on the uterus during the menstrual cycle have been extensively studied. The menstrual cycle is divided into four phases: 1) follicular phase, 2) ovulation, 3) luteal phase and 4) menstruation. During the follicular phase, gonadotropin releasing hormone (GnRH) is secreted by the hypothalamus to stimulate Iuteinizing hormone (LH) and follicle stimulating hormone (FSH) release from the anterior pituitary gland. FSH stimulates follicular maturation and positively regulates estrogen release to mediate proliferation of stromal and epithelial cells of the uterine endometrium. 19 Feedback regulation of estrogen on FSH decreases FSH secretion. This phase is complete once estrogen levels accumulate to cause an LH surge, leading to ovulation. The luteal phase is initiated by the LH surge during which the follicle develops into a corpus luteum. Throughout these phases, uterine endometrium continues to proliferate and progesterone, released by the corpus luteum, aids in its development (86). If no fertilization event occurs, the endometrium is shed during menstruation. Estrogen and progesterone levels decline, releasing inhibitory signals to diminish FSH levels and re-initializing the cycle. The Estrous Cycle Other placental mammals undergo an estrous cycle, which differs primarily from the menstrual cycle where the developed endometrium is reabsorbed rather than shed through menstruation. The estrous cycle is also separated into four phases: 1) proestrus, 2) estrus 3) metestrus and 4) diestrus. Proestrus is analogous to the follicular phase, where by signals are initiated to cause follicle maturation and endometrial proliferation. Estrogen levels peak to stimulate estrus, an LH surge and ovulation. At this stage, the uterus has reached maximal endometrial proliferation and vascularization. Estrus is the phase during which females are most sexually receptive. Decline in estrogen, FSH and LH due to no fertilization leads to metestrus where uterine epithelium begin to degenerate and a corpus luteum begins to develop. Finally, the corpus luteum matures during diestrus releasing progesterone and the uterus reverts back to an atrophic state. 20 IN VIVO MODEL AND THE UTEROTROPHIC ASSAY An immature rodent model has long been a gold standard to evaluate compound estrogenicity due to its reproducibility and reliability to identify compounds which exert its effects through an estrogenic mechanism of action (87,88). The specific model utilized in the outlined studies is an estrogen sensitive (89) immature, ovariectomized, C57BU6 female mouse. An immature mouse provides a low background system in which estrogen treatment can exhibit maximal physiological effects. Ovariectomizing allows continued development of organs for analysis without the confounding effects of circulating estrogens. Moreover, the mouse genome annotation is extensive, comprising approximately 28 000 unique transcripts (90), which aids the construction of estrogen-modulated pathways in gene expression analysis experiments. The uterotrophic assay consists of three daily doses of compound through subcutaneous injection or oral gavage. Compounds are classified as estrogenic if increases in uterine wet weight (UWW), due to a combination of increased cellular hypertrophy, hyperplasia and water imbibition, are observed. Histological hallmarks of uterotrophy include increased luminal epithelial cell height (LECH), increased luminal circumference, luminal epithelial invagination, stromal edema, and increased glandular epithelium (91). TRANSGENIC MODELS TO EXAMINE ESTROGEN RECEPTOR SIGNALING To further elucidate the roles of estrogen signaling in the uterus, several transgenic models have been developed. aERKO mouse uteri maintain all 21 uterine cell types, however, tissue strata are smaller when compared to wild type (92). These mice are infertile and do not undergo uterotrophy upon estrogen treatment (92). ERB is not as prominently expressed in the uterus as its alpha counterpart (93,94) and BERKO mice are subfertile, due to diminished ova maturation and subsequent release from the ovaries (95). These models illustrate the importance of ERa in uterine development and function. More recently non-classical ERor knock-in (NERKI) mice have been developed. A single ER allele mutant, that does not bind DNA, was developed (96) and introduced into embryonic stem cells to create NERKI mice (97). NERKI mice have a double alanine mutation (AA) in the zinc finger region of the ER DBD and were used to characterize non-classical ER signaling where the receptor is required to tether to other DNA bound proteins to influence transcription of estrogen-responsive, non-ERE containing genes (96). NERKI (AA/+) mice exhibit uterotrophy upon E2 and TAM treatment, but a smaller increase in UWW compared to treated WT mice, and NERKI females are infertile (97). True non-classical signaling cannot be examined in the NERKI mice as a wild type ERor allele is still present. Thus, NERKI (AA/+) males were crossed with ERa +/- females to generate mice with no classical ER signaling capabilities (AA/-) and compared with orERKO (-/-) and wild type (+/+) mice (98). Uteri of AAI- mice demonstrated a physiology intermediate to those of K0 and WT mice where AAI- uterine wet weight, radius, inner circular muscle and luminal epithelial height were significantly greater than those of KO mice, but significantly less than 22 those of WT (98). Responses to estrogen treatment further illustrated the roles of classical and non-classical ER signaling in mouse uterus. Increases in luminal epithelial cell height occurs in AA/- mice suggesting that non-classical signaling is adequate to stimulate this response; however, stromal proliferation was only stimulated in WT mice, indicating its dependency on the classical ER mechanism (98). CHEMICAL MIXTURES The field of risk assessment examines the effects of compounds on human health and environmental organisms. Data from these studies provide information to agencies which prioritize these potential hazards and determine methods to regulate high risk factors. Efforts to elucidate the mechanisms of action of individual compounds allows for the association of adverse affects by specific chemicals and classes of chemicals. However, wildlife and human exposure to compounds primarily occur as complex mixtures; thus, efforts to examine mixture effects for risk assessment purposes are warranted. An approach has been developed by the EPA Superfund Program Office which involves the identification and characterization of individual chemicals before examining mixtures (99). MIXTURE EFFECTS Effects by mixtures can generally be classified as additive, antagonistic or synergistic. Additive effects are those where the combined treatment results in 23 an effect which is comparable to the sum of the responses elicited by each individual treatment; for this reason, they are also known as concentration- addition (CA) effects. Additive effects are also considered to be non-interactive as each compound induces an expected degree of response despite the presence of another compound (100). Studies of various species have demonstrated that mixtures of compounds, particularly pesticides, acting through similar modes of action result in additive responses for growth and lethal endpoints (101,102). These studies also demonstrated that the majority of mixtures comprised of compounds with differing modes of action elicit concentration-addition; however, some demonstrated Iess-than-additive effects while others resulted in greater-than-additive effects (101,102). Although CA predictions are likely adequate for the purposes of risk assessment, identifying and characterizing the combinations of compounds which elicit different-than- additive effects is important. Synergistic effects are those which exhibit responses greater than the sum of the individual responses. These responses are of interest as the use of the CA theory to risk assessment underestimates the potential adverse affects demonstrated by compound mixtures eliciting synergy. Studies of pesticide mixtures in the environment suggest that atrazine herbicides in combination with organophosphate insecticides resulted in greater-than-additive effects on the locomotive ability of certain invertebrate species (103). Moreover, this mixture represents compounds which exhibit potentiation, in which one compound that 24 does not exhibit high toxicity, atrazine, increases the expected toxic effect of the second compound, organophosphate. Other studies have identified compounds that demonstrate synergism through the mechanistic examination of specific signaling pathways. It has been shown that the aryl hydrocarbon receptor (AhR) signaling pathway and its influence on cytochrome P450, family 1 (CYP1A) induction are important mediators in xenobiotic toxicity in mammals (reviewed in (104)). Studies indicate that CYP1A knockdown in killifish and zebrafish embryos result in greater-than- additive effects in the presence of polyaromatic hydrocarbons (PAHs), but not dioxin-like compounds (105). Thus the presence of PAHs with potential CYP1A inhibiting compounds in the environment appears to be a greater hazard for some aquatic species and should be re-evaluated where current risk assessments suggests application of an additive model (106). Less-than-additive effects are also known as negative interactions or antagonistic effects. Identification of these effects allow for prioritization of compounds with respect to risk assessment. For example, Aroclors are commercial mixtures of PCBs that are immunosuppressive. However, some combinations containing greater concentrations of coplanar PCBs, such as 3,3’,4,4’-tetrachlorobipheyl, 2,3,3’,4,4’-pentachlorobiphenyl and 2,3’,4,4’,5’- pentachlorobiphenyl, result in a Iess-than-additive effect (107). Thus, sites containing high levels of these particular congeners may need to be assessed differently from other PCB contaminated areas with congeners exhibiting additive properties. 25 Currently, the additive model is still the assumed model for untested chemicals, particularly at concentrations which demonstrate no observable adverse effect (108). It has been recommended that tests should be carried out rather than blindly accepting the assumption (109). Due to the complexity of chemical mixtures, models to predict the effects of compounds are continually being developed and refined (110-113). These models need to take into consideration the mode of action of the compounds, the close and temporal range of responses exhibited by each compound (111,114), the phamtacodynamics of the compounds on various tissues (115), and the toxicodynamic and toxicokinetic between the compounds of interest (116). It is clear that the data collected through mixture studies will be invaluable to the field of risk assessment; however, additional research is necessary in developing study designs to examine effects and generating statistical models for predictive toxicology. Moreover, few studies examine temporal effects of mixture treatments; thus, development of appropriate temporal study designs and establishment of accurate temporal models are warranted. The approach utilized in Chapter 4 offers an experimental design which can be critically evaluated for future studies of mixed-compound affects on gene regulation. 26 CHAPTER 2 EFFECTS OF CULTURE CONDITIONS ON ESTROGEN-MEDIATED HEPATIC IN VITRO GENE EXPRESSION AND CORRELATION TO IN VIVO RESPONSESL ABSTRACT Refinement of in vitro systems for predictive toxicology is important in order to develop high-throughput early toxicity screening assays and to minimize animal testing studies. This study assesses the ability of mouse Hepa-1C1c7 hepatoma cell model under differing culture conditions to predict in vivo estrogen- induced hepatic gene expression changes. Custom mouse cDNA microarrays were used to compare Hepa-1CIC7 temporal gene expression profiles treated with 10 nM 178-estradiol (E2) in serum free and charcoal-stripped serum supplemented medium at 1, 2, 4, 8, 12 and 24 hrs. Stripped serum- supplemented rnedium increased the number gene expression changes and overall responsiveness likely due to the presence of serum factors supporting proliferation and mitochondrial activity. Data from both experiments were compared to a gene expression time course study examining the hepatic effects of 100 jig/kg 17a-ethynylestradiol (EE) in C57BU6 mice at 2, 4, 8, 12, 18 and 24 1Data contained in this Chapter have been published. Fong CJ, Burgoon LD, Zacharewski TR. 2005. Comparative microarray analysis of basal gene expression in mouse Hepa-1c1c7 wild-type and mutant cell lines. Toxicol Sci. 86(2):342-53. 27 hrs. Only 18 genes overlapped between the serum free and in vivo studies, whereas 238 genes were in common between Hepa-1c1c7 cells in stripped serum data and C57BU6 liver samples. Stripped serum cultured cells exhibited E2-elicited gene expression changes associated with proliferation, cytoskeletal re-organization, cholesterol uptake and synthesis, increased fatty acid [3- oxidation and oxidative stress, which correlated with in vivo hepatic responses. These results demonstrate that E2 treatment of Hepa-1c1c7 cells in serum supplemented medium modulate responses in selected pathways which appropriately model estrogen-elicited in vivo hepatic responses. 28 INTRODUCTION Predicting human toxicity typically involves the extrapolation of in vitro and non-human model data (117,118). Ideally surrogate models will reflect in vivo human responses by replicating appropriate pharmacodynamic and pharrnacokinetic interactions. Conventional wisdom suggests that predictive accuracy improves by minimizing the extrapolation to humans, and therefore models that most closely resemble human responses are preferred. Increasing pressure to develop early high-throughput toxicity screening assays and to reduce animal testing has renewed efforts to assess the limitations of existing systems for predicting human toxicity to more accurately define the role of in vitro data in decision-making. Cells in culture have many advantages as well as some significant limitations for toxicity screening early in development. In general, in vitro models are amenable to high-throughput screening which can be used to prioritize commerce chemicals and drug candidates requiring further toxicity testing or warranting further development. Cells in culture also provide a homogeneous population that facilitates studies examining the effects of different conditions (i.e., serum free, hypoxia, co-cultures) which are not experimentally feasible using in vivo models. In addition, in vitro models are expected to be less variable and allow cell-specific effects to be examined that may otherwise be masked in a multicellular target organ. However, they are also sensitive to the culturing environment, which could influence their response and potentially compromise their ability to predict in vivo effects. In this study the effects of serum free and 29 dextran charcoal-coated (DCC) stripped serum supplemented medium on 17D- estradiol (E2)-elicited mouse hepatic Hepa-1c1c7 gene expression were compared to in vivo hepatic responses. Although not considered a classical target organ, the liver is an estrogen- responsive tissue (67,73,119—121). Most hepatic responses are mediated through estrogen receptor (ER) alpha(16), although alternative mechanisms of estrogen activity have been reported (122-127). Hepa-1C1C7 cells possess active ERS (77,78) and retain several liver-specific functions (e.g., synthesis and secretion of albumin (76) and transferrin (79) in addition to xenobiotic detoxification as evidenced by its high aryl hydrocarbon hydroxylase activity (80))- This study involved time course experiments to identify the effects of culture condition on in vitro gene expression following treatment with E2 using cDNA microarrays. Comparisons of in vitro data to hepatic gene expression studies of estrogen treated C57BU6 mice were then conducted to assess whether Hepa-1c1c7 responses are able to model in vivo hepatic responses to estrogen. 30 MATERIALS AND METHODS Cell viability and growth rate Hepa-1c1c7 cells (gift from O. Hankinson, UCLA, Los Angeles, CA) were maintained at 375°C and 5% CO; in phenol red free DMEM/F12 medium (lnvitrogen, Carlsbad, CA) supplemented with 5% fetal bovine serum (FBS) (Serologicals Corporation, Norcross, GA), 50 pg/mL gentamycin, 2.5 pg/mL amphotericin B, 100 U/mL penicillin and 100 pg/mL streptomycin (lnvitrogen). For the M‘l'l' (3-[4,5-dimethylthiazoI-Z-yH-Z,5-diphenyl tetrazolium bromide) colorimetric assay of cell viability, 10 000 cells were seeded in a 96-well tissue culture plate (Corning, Acton, MA) in 100 pl 5% FBS medium and grown for 48 hrs. Wells containing medium only were used as a blank control. Medium was aspirated and replaced with 5% FBS medium or serum free medium 24 hrs before treatment with 10 nM 17B-estradiol (1,3,5[10]-estratriene-3-17(3-diol) (E2) (Sigma, St. Louis, MO) or 0.1% DMSO (JT Baker, Phillipsburg, NJ) treatment. MTT solution (Sigma) was added to cells and absorbencies read 3, 6, 12, 24, 36 and 48 hrs after E2 treatment at 595 nm on an Emax 96-well microplate reader (Molecular Devices, Sunnyvale, CA) and captured using Softmax software (Molecular Devices). Colorimetric readings (n = 4) were normalized to the blank wells. Two-way ANOVA analysis followed by Tukey’s HSD post hoc test were performed to detect time-matched differences between culture conditions (a = 0.05). For direct cell counts, 3 x 105 cells were seeded in T25 culture flasks (Corning). Medium was changed to 5% FBS or serum free 24 hrs prior to E2 or 31 DMSO treatment. Cells were then trypsinized and counted, in duplicate, using a hemocytometer (Hausser Scientific Co., Horsham, PA) 6, 12, 24, and 48 hrs after treatment. Experiments were completed in quadruplicate. Statistics were calculated using SAS v.9.1 (Cary, NC). Repeated measures ANOVA was performed followed by a Tukey’s HSD post hoc test to detect time-matched differences between culture conditions (or = 0.05). Hepa-1c1 c7 time course treatment and RNA isolation regimens Hepa-1 C1 c7 cells were seeded (1.5 x 106 cells) and grown for 48 hrs in 150 mm culture plates (Coming) in 5% FBS medium. Serum free medium (serum free experiments) or 5% DCC-FBS medium (stripped serum experiments) was then replaced 24 hrs prior to 10 nM E2 or 0.1% DMSO treatment. Cells were harvested by scraping in the presence of Trizol (lnvitrogen). RNA was isolated as per manufacturer’s instructions at 1, 2, 4, 8, 12, and 24 hrs after treatment and resuspended in RNA Storage Solution (Ambion Inc., Austin, TX). RNA quality and purity was examined by running 2 pg of total RNA on a denaturing 1% agarose gel and by examining an Ame/230 ratio. Samples were stored at -80°C until further use. Experiments were completed in triplicate using cells between passage 8 and 12. Animal handling, husbandry and treatment Animals were treated as previously described (73). Briefly, female C57BU6 mice, ovariectomized by the vendor on postnatal day (PND) 20, were 32 obtained from Charles River Laboratories (Raleigh, NC) on PND 26. Groups of five mice were house in polycarbonate cages with cellulose fiber Chip bedding (Aspen Chip Laboratory Bedding, Northeastern Products, Warrensberg, NY) at 23°C and 30—40% humidity and a 12 hr light/dark cycle (0700 — 1900 hr). Animals had access to deionized water and Harlan Teklad 22/5 Rodent Diet 8640 (Madison, WI) ad libitum and were acclimatized for 4 days prior to treatment. Animals were weighed and orally gavaged with 100 )4ng 17a- ethynylestradiol (17oI-Ethynyl-1,3,5(10)-estratriene-3,17B-diol; EE) (Sigma) dissolved in 0.1 mL sesame oil or vehicle alone. Doses were prepared based on average animal weight. Animals were sacrificed 2, 4, 8, 12, 18 and 24 hrs treatment at which time necropsies were performed to remove hepatic tissues. Tissues were snap frozen in liquid nitrogen and stored at -80°C until further processing. All procedures were performed with the approval of the Michigan State University All-University Committee on Animal Use and Care. Frozen tissues were homogenized in the presence of Trizol reagent, RNA was isolated as per manufacturer’s instructions and resuspended in RNA Storage Solution. RNA quality and purity was examined by running 2 mg of total RNA on a denaturing 1% agarose gel and by examining an Ame/280 ratio. 33 Microarray processing Custom in-house cDNA arrays comprising of 6376 features, representing 4858 unique genes (print version Mm. 6), or 13361 features (print version Mm. 7), representing 7952 unique genes (Unigene Build 144), were Spotted on epoxy coated glass Slides (SCHOTT Nexterion, Germany) using an Omnigrid arrayer (GeneMachines, San Carlos, CA) and 16 (4 x 4) Chipmaker 2 pins (Mm. 6) or 48 (4 x 12) Telechem Chipmaker 3 pins (Mm. 7) in a TeleChem CHP3 printhead head (Telechem International Inc., Sunnyvale, CA) by the Research Technology Support Facility at Michigan State University (128). Serum free studies were conducted using Mm. 6 arrays, while stripped serum supplemented studies were completed on Mm. 7 arrays, a more comprehensive version of Mm. 6 arrays. Selected clones were obtained from EPAMAC (129), Research Genetics, the National Institute of Aging and Lion Biosciences. Detailed protocols for processing of microarrays are available at the deach Home Page (130). Independent reference study designs were used (Figure 1) to assess three biological replicates of treatment effects. All microarray studies incorporated 6 time points and utilized 12 arrays, including dye swaps, for each biological replicate for a total of 36 microarrays each experiment. Briefly, 20 pg of RNA was reverse transcribed to incorporate Cy3- or Cy5-conjugated dUTP. Cy3 and Cy5 labelled samples were mixed, purified and resuspended in 48 pl of hybridization buffer (56% formamide, 32% 20X SSPE, 8% 50X Denhardt’s Solution, 4% 20% SDS, 20 pg poly(A), 20 pg mouse COT-1 DNA, 10 pg yeast 34 Figure 1 Independent reference design for microarray hybridization. Arrows represent a Single microarray in which two labelled samples, Cy3 (tail) and Cy5 (arrow head), are hybridized. Directionally opposing arrow pairs represent a dye swap where estrogen-treated (T) and vehicle treated (V) samples are reciprocally labelled and hybridized on two individual arrays. Each replicate temporal study (e.g., samples collected at 1, 2, 4, 8, 12, and 24 hrs) involved 12 hybridizations (2 per time point) for a total of 36 arrays per time course study (n = 3 independent animals). An identical design was used to assess in vivo gene expression where samples collected at 2, 4, 8, 12, 18 and 24 hrs were hybridized. T1 T2 T4 T8 T12 T24 II II II II II II V1 V2 V4 V8 V12 V24 35 tRNA carrier) for overnight 42°C hybridization on printed arrays. Slides were washed in SSC solutions contain decreasing concentrations of SDS, dried andscanned using a 428 Affymetrix Scanner (Santa Clara, CA). Images were examined, features identified and intensity values determined using GenePix v.5.1 (Molecular Devices). All data was stored in deach (130), a Minimum lnforrnation About Microarray Experiments (MlAME)-compliant relational database (131) running under Windows 2003/Oracle 109 that currently supports microarray data storage, retrieval, and querying as well as facilitates data analysis, Sharing and reporting (132,133) All arrays within this study were compared to a historical data set of established high quality arrays. Parameters that were assessed included background signal intensity, feature signal intensity, feature vs. background Signal intensity ratios, the number of features with background intensities greater than the feature intensity for each array, and relationships between feature and background signal intensities. All arrays met the standards of the quality control parameters (1 34). Statistical, filter and cluster analysis of microarray data Microarray data were analyzed using a semi-parametric approach (135). Model-based t-values were calculated from normalized data, comparing treated and vehicle responses at each time-point. Empirical Bayes analysis was used to calculate posterior probabilities (P1(t)-value) of activity on a per gene and time point basis using the model-based t-value (135). A P1(t) score cut-off was 36 initially used to identify differentially expressed transcripts between treatment groups. Feature subsets were associated with functional annotation using Entrez Gene (136) and Gene Ontology (137). General temporal patterns were identified using k-means clustering (GeneSpring v7, Silicon Genetics, Redwood City, CA). Temporal gene expression correlations (activity index) and temporal P1(t) activity correlations (significance index) between in vitro and in vivo studies were calculated using Pearson’s correlation at overlapping time points (La, 2, 4, 8, 12 and 24 hrs). Correlation indices were plotted on a Cartesian plane, to visualize the relationship of the same gene in the two model systems, through an in-house developed Toxicogenomics Correlation Tool (TCT). A third dimension of information is provided through the Displacement Heat Map function of TCT where time displacement for a gene between the in vitro and in vivo models is visualized through the color intensity of the point. A Displacement Index (DI) is derived by: i) identifying the number of time points that exhibit opposite activities in between models (e.g. in vitro model meets P1(t) > 0.9 cut-off whereas in vivo does not, or vice versa), ii) identifying the number of time points which are being compared, iii) calculating the quotient by dividing the value of step i by the value of step ii. A range of values representing non- displaced (DI = 0) to highly displaced (DI = 1) results in a gradient from light to dark color intensity, respectively (133). TCT can be licensed through arrangement with the Office of Intellectual Property at Michigan State University. 37 Quantitative RT-PCR RNA aliquots from each replicate were set aside for microarray verification by SYBRTM Green quantitative real-time PCR (QRT-PCR). Briefly, 2 pg of RNA were primed by an anchored oligo—dT and reverse transcribed using Superscript II (lnvtrogen, Carlsbad, CA) in a 40 pl reaction as described by the manufacturer. The cDNA solution was diluted 4-fold and 3 pl was used in a 30 pl PCR reaction containing 1X SYBR Green PCR buffer, 3 mM MgCl2, 0.33 mM dNTPs, 0.5 IU AmpliTaq Gold (Applied Biosystems, Foster City, CA) and 0.15 mM forward and reverse primer. All primers were designed by submitting clone sequences into Primer3 (138) to obtain an amplicon of approximately 125bp (Table 1). All primer and QRT-PCR reaction conditions were submitted and stored within the Real- Time PCR Subsystem of deach (133). PCR amplification was conducted in 96-well MicroAmp Optical plates (Applied Biosystems) on an Applied Biosystems PRISM 7000 Sequence Detection System under the following conditions: 10 min denaturation and enzyme activation at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. A 30 min dissociation protocol, after amplification, was conducted to assess primer specificity and product uniformity. Each plate contained duplicate standards of purified PCR product of known template concentration over the range of eight orders of magnitude to generate a log template concentration 38 NzE ooF 8m8om980mm893m 88899908888 mm Ecol—22 mow: NEqu 68.03828 020.5336 8 F F 882828880899 090808888832 5 Em PIS—z 0 5mm 9E0 520:. m:_.om.mE_-Emm 6858628 0.8805 mm? 88898888.... .8m888888 488322 usage .8ng -m 858.988 42 888888099 8.888.888 583422 82: Er. 88: 852-..... 88. .m 8? 508808888. 888988988. 83 3422 v85 83m“. 590:. 8.2.5 new Em. P 8.3828 .m 2.2.958 .F 03 «8888888988 62068988388 «mammal—22 who? rm ...:0 2E9. 693. mEoEooSo F mmcom 352:8 2: 90228808882 8988898888 89 SIEZ £89 390 -Sm. .o .8859 5.2.8 P N: «8882888898 88286898888 NvRooIEZ ~83 3:00 9.9m ._ 83 683.002.. < mo.oEo: 8? 898888882 288288.888 808322 «$3 880 N 0.26 5.22.3 =8 5 v 86989888889 88888888008. mmmmoolzz v EN F mmuo amazcm @000 E 888898868 8888688899 88322 88? 380 5 5.96 or; 8888889880 88888888 5.2.89.2 88? 2:80 5 5.96 82 .F 520:. 889958962 8? 8888888.... «858888588 858422 ER? 3%. m E 8.308858 8? 8.5888888. 2862888808 833422 3me 2.2 F 5.5.2:. 889.8 mm? 6888888808. 88080828838 mmmnoolfiz 53 w £o< 2.583030 .33 .czom 38 SEE OBOE”. LOEEn. Emzco”. 8.88.4 o. 698% 952 950 mum 896”. meow meow Nozcm .m: LOEEQ moan—MEG ... 23¢... 39 no? .00.. 03 mm. m: m: 00.. ..oF m: vor 03 m: 0.8000.08.008008 002800008088.8 008808820220 08090890000880 80000880909800 09089800008800 820.900.09.808 00008808090880 0900088888080 8800900880800 08088000009000. 00809080000888 00008880008008 990.900.880.90 80008830089008 0005080052000 08229088000000 008.808.000.880 00098880880000 90.009090090090 90000800890 8. 008008908080000 .800800080900000 88.008.00.088. 00... Buzz 00008 .22 .00. 20:22 503422 0:08:22 00.003422 0.002022 5.08 .22 0.00.0122 0000.01.22 0.003422 000031.22 3N FN Nmoow owmom n 500 203. omNNN ooNNN 05m: 0009 800 F m ..nwom V00... _. wtow F.0ow 0000 ESE 80.2 05.2 05.2 ~20. :0. ...... .0090: 0.0.00 -0.0000. .0. 000.0 ”Zn... 00.0000. 00.0.0.-0.._..00 F 000.0.000.._>00-O .0.0.0 .02. .0 0.0.0.0 .0...00 .0.0.0 F 0000.08.02.90... 0.02:...0 P 500.00. .0 000.0 .0 2.0.0500 .0.0000. .00.000 0 .0900... .1 000.0 .F 2.0.0.000 .0.0000. .0080: N .0900... .... 000.0 4 2.000.000 .0.0000. .00.000 0 0.008... P .0.00. 0.26.0 00...-0..000. 000.0800: .200000000000 00.0000 002.000.0020 F 0000.50 .4. 0E>~00oo-.>.0.0.0.>0.00. 0-8.22-0 m 00. .000.0.000..-w 00250.20 00:00:00 .. 0.0a» 8.0.0.000 40 .6389 £99509. 2: 8%538882? oamsngoame 8531.22 93% .65 23ch 26.39, (m 539.855 .6 Bumzuom m9. owommoommmmmmmummmmm ommfimmmommfimmmoom wwELoIEZ ommom .3.me Ucm .mozuwcmbficgm 25350 .P 2%... 41 standard curve. No template controls (NTC) were included on each plate where unknown samples with a Ct value within 2 SD of the mean Ct values of the NTCs were considered below the limits of detection. Plots were visualized and thresholds determined using ABI Prism 7000 808 Software (Applied Biosystems). Results were normalized to a geometric mean of B-actin (Actb), glyceraldehyde—3-phosphate dehydrogenase (Gapd) and hypoxanthine guanine phosphoribosyl transferase (Hprt) mRNA levels to control for differences in RNA loading, quality and cDNA synthesis. Statistical significance of expression differences were assessed using a factorial ANOVA followed by Tukey’s HSD post hoc analysis to examine treatment and treatment over time effects using SAS version 9.1. R, version 1.9.1, was used to compute the Pearson’s correlation coefficient between DNA microarray data and QRT-PCR results. 42 RESULTS Hepa-1c1c7 cell viability and proliferation in serum free medium E2 effects on Hepa-1c1c7 cells were examined in serum free medium to minimize exposure to serum borne estrogens. Serum starvation synchronizes cells at G1 and was used in this study to optimize the detection of potential proliferative responses which may otherwise be masked in an asynchronous population. Published reports have demonstrated that estrogen induces synchronized uterine proliferation in the immature, ovariectomized rodent model. However, limiting factors provided by serum may compromise cellular viability and responsiveness. Consequently, medium supplemented with dextran-coated charcoal (DCC) stripped serum was also examined. DCC serum stripping removes steroids and other small molecules that can pass through the dextran coating and bind to the activated charcoal, which is then discarded (139). Cellular distress may be detected through morphological changes exhibited by the cells. However, no significant morphological differences were observed after four days in either 5% FBS supplemented or serum free medium (data not shown). In addition cell viability and proliferation in serum free medium was assessed, by monitoring mitochondrial activity using the MTT assay, and by direct cell counting. MTT time course assays indicate that Hepa-1c1c7 cells exhibited a 3-fold increase in mitochondrial activity (p < 0.05) in 5% FBS supplemented medium over a period of 48 hrs indicative of cellular proliferation (Figure 2A). In serum free conditions, MTT activity was significantly reduced (p < 43 Figure 2 MTT and cell count assessment of Hepa-1c1c7 cells in serum free and stripped serum supplemented medium. MTT and cell count assays were used to assess the viability of Hepa-1c1c7 cells incubated in a serum free environment. A) Parallel MTT time course assays were conducted in 5% FBS supplemented medium (black bars), serum free medium treated with DMSO (open bars), and serum free medium treated with 10 nM E2 (grey bars). Wells were seeded with 10 000 cells, medium changed after 24 hrs, treated and assayed at 3, 6, 12, 24, 36 and 48 hrs after treatment. Only cells in 5% FBS exhibited a significant increase in mitochondrial activity a(p < 0.05) at all time points beyond 3 hrs. Viability of serum starved cells were significantly different b(p < 0.05) from time-matched 5% FBS cultured cells. B) Parallel direct cell count time course assays were conducted in 5% FBS supplemented medium (squares), serum free medium (triangles), serum free medium treated with DMSO (diamonds) and serum free medium treated with E2 (circles). T25 flasks were seeded with 300 000 cells and directly counted at 6, 12, 24 and 48 hrs. Cells cultured in 5% FBS exhibited a significant increase in number from at 48 hrs relative to 3 hrs “(p < 0.05). Cell numbers of serum starved cells are significantly different b(p < 0.05) from time-matched 5% FBS cultured cells. 44 Figure 2 A MTTAssay a -5%FBS 35' a a IZISF-DIVSO 30' a a -SF-EZ 325 C «862.0- b 1.5- b “O - b bb bb b 81.0- b b “5' I | | | | I W 3 6 12 24 36 43 Time(hours) DirectCeIICout -E-5%FBs Ascooooo- fi-SF-mtwed E -0-SF-DNBO 2 mm -O—SF-E2 3 a venomo- C .2 Emmo- ‘5 b 8 1m '-- _?_ 356E533 B c .. O o c 1 I I ' I ' 0 10 Z) I!) 40 m 60 Tina (hows) 45 0.05) compared to serum supplemented conditions but no net loss of cells was detected over 48 hrs. E2 did not induce a significant change in MTT activity in a serum free environment when compared to time-matched vehicle treated cells under comparable conditions, although the trend suggested E2 may induce activity beyond 48 hrs. Direct counts indicate that Hepa-1c1c7 cells cultured in serum supplemented medium are actively proliferating (Figure 28). This increase in cell number may account for the increased mitochondrial activity observed in the MTT assay by cells in the serum supplemented condition. No increase in cell number was observed using serum free conditions and neither E2 nor DMSO enhanced cell number over time as suggested by the MTT assay. These results indicate that Hepa-1c1c7 cells maintained in serum free conditions are viable and do not appear to undergo proliferation. Furthermore, viability and proliferation are not affected by E2 treatment, thus concerns regarding E2 induction of proliferation and confluency will not confound gene expression analysis. Temporal E2-mediated changes of gene expression in a serum free environment cDNA microarrays were used to investigate E2-elicited Hepa-1c1c7 cell gene expression changes in serum-free and DCC-stripped serum supplemented medium. Empirical Bayes analysis identified 245 active features (P1(t) > 0.999, Mm. 6), representing 167 unique Entrez Gene annotated genes following E2 46 treatment compared to time-matched DMSO treated controls (Supplemental Table S1 (130)). P1(t) values were used to rank and prioritize features for further investigation. Gene expression changes ranged from 2.1-fold induction (e.g., decorin - Dcn) to 2.17-fold repression (e.g., chemokine (C-C motif) receptor 5 - Ccr5). Five K-means clusters best represented the temporal profiles of these active genes as A) up-regulated beyond 8 hrs, B) down-regulated at 4 hrs, C) down-regulated at 1 and 8 hrs, D) down-regulated at 8 hrs and up-regulated at 12 hrs and E) up-regulated at 4 hrs (Figure 3). Functional annotation for the 167 active genes was identified through Gene Ontology and complemented with reports in the published literature. Genes with roles in transcriptional regulation were most frequently represented in addition to those involved in cell proliferation and differentiation, cytoskeletal organization, and transport and metabolism of lipids and carbohydrates. The effects of DMSO alone were also examined under serum free conditions where comparisons were made with untreated samples. Interestingly, DMSO elicited transcriptional changes primarily associated with proliferative arrest and increased solute regulation, which were not observed with E2 treatment (Supplemental Table 82) when compared to untreated cells. These results indicate that gene expression differences due to E2 treatment can not be attributed to DMSO. 47 Figure 3 Temporal gene expression patterns: E2 in serum free medium Five k-means clusters were identified to concisely represent the general temporal patterns exhibited by 246 active features treated with 10 nM E2 in serum free medium. Each line represents a single feature with its fold-change (x = induction; / = repression) determined by comparison to the time-matched vehicle control. Black pseudolines indicate the general profile represented in each cluster. Fold Change 1'24 6 1'2 211 iéh is 1'2 24 1'21: 6 1'2 2'4 Time (hours) 48 Temporal E2-mediated gene expression changes in stripped serum supplemented medium Although serum free medium provides a nearly depleted steroid environment and synchronizes the cells at G1, these conditions may compromise responsiveness due to the lack of serum factors that facilitate gene expression (23,140). When E2 elicited gene expression effects were examined in medium supplemented with DCC-stripped serum, 1882 unique features (P1(t) > 0.999; Mm. 7) representing 1134 unique annotated genes were identified as differentially expressed (Supplemental Table S3). The magnitude of transcriptional changes ranged from 2.06-fold induction (e.g., cytochrome P450, family 1, subfamily a, polypeptide 1; Cyp1a1) to 2.08-fold repression (Accession lD: CR517543). Five K-means clusters best described the temporal gene expression elicited by E2 as A) induced at 2 hrs, B) induced at 8 hrs and repressed at 24 hrs, C) repressed at 24 hrs, D) sustained induction between 2 - 8 hrs and E) repressed at 8 hrs (Figure 4). These clusters exhibited different profiles when compared to the five clusters identified for E2 treated Hepa-1c1c7 cells in serum free conditions in terms of which genes were responsive and their temporal pattern of gene expression. Most genes were induced at one or more time points while only 60 features were repressed. As observed in serum free conditions, gene expression changes in stripped serum supplemented conditions included functional annotation associated with transcriptional regulation, cell proliferation, cytoskeletal organization, and lipid transport and metabolism. 49 Figure 4 Temporal gene expression patterns: E2 in stripped serum medium Five k-means clusters best represent the temporal expression patterns exhibited by 1882 active features elicited following treatment with 10 nM E2 in stripped serum medium. Each line represents a single feature with its fold-change (x = induction; / = repression) determined by comparison to the time-matched vehicle control. Black pseudolines indicate the general expression profile represented in each cluster. The patterns represented in this study differ from the general patterns exhibited by l-lepa-1c1c7 cells treated in serum free medium in genes represented within the cluster and the shape of the temporal expression profile. 50 Figure 4 x2 Fold Change x2 Time (hours) 51 Cells cultured in stripped serum medium exhibited E2-induced changes in a greater number of genes compared to cells cultured in serum free medium. However, this can be attributed to the more comprehensive Mm.7 version of the array. Only 30 unique annotated genes were found to be active between serum free and stripped serum supplemented medium studies suggesting that serum borne factors influence E2-mediated transcription (141,142). Factors influencing gene expression may include the lower mitochondrial activity of Hepa-1c1c7 cells in serum free conditions, and possible non-additive interactions between E2 and serum components such as growth factors, which could activate other signaling pathways. Quantitative real-time PCR verification QRT—PCR was used to verify microarray data of selected genes representing different cluster profiles and functional pathways. Pearson’s correlations were used to quantitatively assess the level of agreement between microarray and QRT—PCR gene expression profiles. Correlations were classified as either good (p 2 0.5), moderate (0.5 > p > 0.1) or poor (p S 0.1). Of the 32 genes examined (Table 1), 19 were classified as good, 7 as moderate and 6 as poor. In some cases, poor correlations were the result of changes in gene name annotation. Primers were originally designed for a specific gene represented by the clone printed on the array, and not the sequence of the clone. For example, accession numbers for clones may be reassigned upon Unigene database rebuilds, and therefore the initial primer set may no longer amplify the gene of 52 interest. For example, deach Clone ID 87 which was originally identified as representing Lamb3 (laminin B3), but is not currently associated with an official gene. The correlation for this primer set was classified as poor. Overall, QRT- PCR verified the gene expression changes detected in the microarray assay. In vivo vs. in vitro gene expression comparison ln-life study of temporal hepatic responses to estrogen was conducted in C57BU6 mice orally gavaged with 100 )1ng 17a-ethynyl estradiol (EE) for 2, 4, 8, 12, 18 and 24 hrs. EE is an orally active estrogen used in contraceptives that elicits responses comparable to endogenous E2 (143). Empirical Bayes analysis identified 1582 features, representing 1007 active unique annotated genes (P1(t) > 0.9999) (Supplemental Table 4). The results were comparable to a previous intralaboratory study using a model-based t-test for microarray analysis (73). Both serum free and stripped serum supplemented conditions were assessed to evaluate the in vivo predictive value of Hepa-1c1c7 cells. 6376 features were in common between the serum free (Mm. 6) and in vivo (Mm. 7) studies due to different array platform versions (Figure 5). Between the 254 and 1582 active features of the respective studies, only 23, representing 18 annotated genes, were in common between the serum free (Mm. 6) and in vivo 53 Figure 5 Systematic comparison of in vitro and in vivo active gene lists Data sets for E2 treated Hepa-1c1c7 in serum free (245 features) and stripped serum (1882 features) (P1(t) 2 0.999) were compared to the EE treated C57BU6 hepatic tissue data set (1582 features; P1(t) 2 0.9999). Only 15 annotated genes were active in both serum free in vitro and in vivo studies; while 238 genes were active in both DCC-treated stripped serum in vitro and in vivo studies. 54 me m P $50 “.9385. 8:00 .6990: Rm mm .8233“. 56:60 3553 coEEoO rams News ammo A SE 888 A 85 a c< U a 5m 2 won 9. 03263. 8563 o_nm__m>< $.28“. 629m 835m 669.88: .2: 2 8878.8... 9%. s 669.8 3938-8: 9%: 8F 82. 20:3 9.5an 9.3 E U mew ammo A SE a onmo m.nm__m>< woeammn. mm: 523 668.88: .2: 2 no we 7.82.. 0b.: S m 959". 55 (Mm. 7) studies indicating that E2-treated Hepa-1c1c7 cells in a serum free environment poorly model EE-elicited hepatic gene expression in C57BU6 mice. In contrast, 13,361 features (Mm. 7) were available for comparison between the stripped serum and in vivo studies. Microarray studies identified 1881 and 1532 features, in the stripped serum and in vivo studies, respectively. Comparing these active gene lists identified 337 active features, representing 238 genes, that were in common between Hepa1c1c7 cells in serum stripped medium and CS7BU6 hepatic tissue. Specific biological pathways were not over-represented by these genes, but associated functions included cellular proliferation, cell signaling, cytoskeletal organization, lipid metabolism, and intracellular communication. The in-house developed Toxicogenomics Correlation Tool (TCT) was used to identify genes exhibiting similar and different temporal gene expression and P1(t) patterns between Hepa-1c1c7 cells in stripped serum medium and CS7BL/6 liver. Each data point represents a single gene. lts position on the Cartesian plane represents how the similarity of the temporal response of that gene in the two models as reflected in the gene expression (activity index) and P1(t) values (significance index). In general, Pearson’s correlations for in vitro and in vivo gene expression data exhibited a positive relationship (i.e., data point distribution along the positive x-axis, Figure 6A) indicating that these genes respond in similar directions and magnitude over time in estrogen-treated Hepa- 1c1c7 cells and C57BU6 liver. However, P1(t) correlations (significance index; y-axis) span both the positive and negative axes indicating variability across time 56 Figure 6 In vitro vs. in vivo significance P1(t) and activity index correlations The TCT plot is a visualization tool which allows groups of genes with similar temporal activity and/or significance between in vitro and in vivo models to be quickly identified. (A) Plot of significance index (i.e., P1(t)) coefficients vs. activity index (i.e., gene expression) coefficients for 337 active clones (238 genes) in E2- treated Hepa-1c1c7 cells maintained in stripped serum (in vitro) and EE-treated C56BU6 mouse hepatic tissue (in vivo). Each data point represents a single feature where the in vitro and in vivo P1(t) values and gene expression patterns have been compared through Pearson’s correlation analyses. The inset box, upper right, encloses 27 features (17 genes) with the highest P1(t) and gene expression correlation coefficients (p 2 0.5) (Table 2) identifying in vitro and in vivo responses with highly similar response profiles and P1(t) values across 2, 4, 8, 12 and 24 hrs. Shading intensity of the data point indicates the degree of time displacement of P1(t)-values for a single gene when comparing between models. Darker points identify genes with a greater number of time points exhibiting P1(t)- value discrepancies between models while lighter points identify genes with fewer time related discrepancies as calculated through a Displacement Index (Dl) value. Data points labelled 1 through 4 are graphically described in (B) to further illustrate differences in temporal in vitro and in vivo activity and significance profiles. 57 Figure 6A Activity Correlation Index e1 ' " e1 0. O Expression Correlation Index Sq 58 Figure GB 1 SW 2 ”"53 6- +!!! vitro 1'3, «rein VIVO * * 5' 1.21 * c .A.‘ c 1.0- a ' s‘ * (I 3 3i #0 x‘ 3 as- * “ too-......... "5 2- ‘ I: ‘:.,,. * 2 0'8. ‘ " ‘ ‘5 .*_..——-|\* * ""0".“ u’ 0.7'l “0* I" 1‘ as “a: b I I I 0.5 I I 0 10 20 so 0 10 Z) Tm(hous) “790013) 13. 1.5)- 1.1- % 3125- 5 1.01 5 g S o O 'u 0.9- 1’ '5 '6 1.00- “- II. 0.8- 0.7 r r q 0.75 0 1O 20 3 0 Trre(hou‘s) 59 Table 2. Genes exhibiting high temporal gene expression and activity correlations, (p 2 0.5) Gene Gene Entrez Expression Activity Gene Name Symbol Gene ID Correlation Correlation signal transducer and Stat5afil 20850 0.81 0.99 activator of transcription 5A histocompatibility 2, H2-Bf 14962 0.81 0.57 complement component factor B protein C Proc 19123 0.76 0.88 uridine monophosphate Umpk 80914 0.74 0.84 kinase LIM domain only 6 Lmo6 54630 0.73 0.84 syncollin Sycn 68416 0.70 0.67 PHD finger protein 5A th5a2 68479 0.70 0.56 Bcl2-like 10 Bcl2l10 12049 0.66 0.57 enoyl Coenzyme A hydratase Echch 52430 0.60 0.82 domain containing 2 FK506 binding protein-like Fkbpl 56299 0.59 0.71 B-cell leukemia/lymphoma 2 Bcl21 12043 0.57 0.62 fyn-related kinase Frk 14302 0.54 0.62 Fraser syndrome 1 homolog Fras1 231470 0.53 0.77 (human) pleckstrin homology, Sec7 Pscd3 19159 0.53 0.88 and coiled-coil domains 3 degenerative spermatocyte Degs1 13244 0.52 0.61 homolog 1 (Drosophila) voltage-dependent anion Vdac12 22333 0.52 0.57 channel1 tumor necrosis factor Tnfrsf11b 18383 0.51 0.69 receptor superfamily, member 11b (osteoprotggerin) 1Genes containing bone fide, functional ERE sequences. 2Genes containing putative ERE sequences as defined by Bourdeau et al. 2004. 60 between model systems. These results suggest some degree of conservation of estrogen-induced signaling pathways where transcript levels exhibit conserved regulation between the stripped serum in vitro and in vivo systems. Using a correlation threshold of p 2 0.5 for both parameters 27 features, representing 17 annotated genes, were identified with comparable significance and temporal gene expression patterns (Table 2). Generally, these genes were induced between 2-12 hrs and down-regulated by 24 hrs. The functional pathways associated with these genes varied but supported several responses known to be modulated by estrogens (73,144,145). Because these genes exhibited high correlations between models, they were suspected to be primary targets of ER-mediated responses and a search for estrogen response elements (EREs) in their regulatory regions was conducted. Stat5a (signal transducer and activator of transcription 5A) and 8ch (B-cell leukemia/lymphoma 2) contain functional EREs, and computational searches identified putative EREs in the regulatory regions of th5a (PHD finger protein 5A) and Vdac1 (voltage- dependent anion channel 1) (26). The remaining genes may also be candidates for primary ER-mediated modulation. Activity indices were spread across both positive and negative axes and appeared to be more heavily distributed into the negative, thus these latter genes were examined to investigate causes contributing to poor temporal P1(t) correlation. Displacement analysis was conducted to identify the degree of non- overlapping significance at similar times between models on a per gene basis. The greater number of time points where a gene does not meet the P1(t) > 0.9 61 cut-off in both models, a greater displacement index is exhibited by that gene. Displacement indices are displayed as a third dimension (i.e. color) of the TCT plot where features exhibiting greater temporal displacement (P1(t) > 0.9) between models are represented by points with greater color intensities, while features with little or no temporal P1(t) displacement are of lighter intensity (Figure 6A). The P1(t) cut-off was lowered to include data approaching significance to achieve a more comprehensive impression of which data were truly incidental, despite the fact that each gene met the initial parameters in both model systems. From the 337 active features in common between models, displacement analysis identified 327 features exhibiting temporally displaced P1(t) values. The trend of color intensity is distributed from light to dark from +1.0 to -1.0 (top to bottom) on the significance index axis. This result is not unexpected since negative significance indices indicate poor correlation of P1(t)- values over time. Selected data points, labeled 1 through 4 (Figure 6A), were examined to further illustrate data represented on the TCT plot. Data point 1 represents Stat5a, which demonstrated indices of activity, significance and displacement of 0.807, 0.998 and 0.2, respectively. Graphically, in vitro and in vivo expression profiles follow similar patterns and attains a P1(t) > 0.9 in both models at all but one time point (Figure 6B1). Similarly for data point 2, Dhrs3 transcripts had similar expression profiles over time but in opposite directions translating to activity, significance and displacement indices of -0.810, 0.885 and 0, respectively (Figure 6B2). However, most genes did not exhibit such stark 62 similar or opposing responses between models. Data points 3, Synj2 (Dl = 0.8), and 4, Pla1a (Dl = 1.0), have activity and significance profiles that are temporally displaced (Figure 6B3, 6B4). Interestingly, the expression profiles suggest that an in vitro response temporally precedes a similar response in vivo. Of the data points demonstrating negative significance indices, 111, representing 80 genes, were first responsive in vitro by at least one time point prior to a similar in vivo response (Supplementary Table S5). This temporal shift in response may partially be attributed to differences in absorption, distribution and metabolism between in vitro and in vivo models. However, eight features exhibit in vivo gene expression prior to evidence of significant Hepa—1c1c7 expression, while the remaining 87 features had divergent expression patterns between models or do not exhibit temporal displacement. These results indicate that factors beyond differences in pharrnacokinetics can also affect in vitro and in vivo gene expression profiles. 63 DISCUSSION In this study the utility of Hepa-1c1c7 cells as a potential in vitro model to predict in vivo hepatic estrogen responses was examined under serum free and stripped serum conditions. Serum free medium minimizes the effects of serum- borne steroids, synchronizes cells in G1 arrest and more closely mimics the hormonal milieu of an ovariectomized mouse model. Theoretically, cell synchronization under conditions devoid of steroids should enhance the detection of expression responses for those genes involved in cell cycle and proliferation. However, serum free gene expression was significantly compromised as only 18 genes demonstrated overlap with Hepa-1c1c7 cells cultured in a stripped serum environment. The lack of common responses may be attributed to the lack of serum factors. Epidermal growth factor (EGF) potentiates estrogenic responses in mouse uterus and its signaling pathway has been coupled to several ER-dependent effects (145,146). Furthermore, the promoter regions of classical E2-responsive genes, such as p82 and lactoferrin, contain active response elements which requires activation by growth factors and other signaling molecules in order to co-operatively elicit a robust response (147,148). In vivo transgenic reporter models and microarray studies clearly demonstrate that the liver is estrogen responsive, although it does not exhibit the same gross physiology changes as the rodent uterus (67,73). Under stripped serum conditions, Hepa-1c1c7 cells exhibited a more robust gene expression response to E2, when compared to cells maintained in serum free medium, but 64 did not reflect the diverse response observed in vivo. However, specific stripped serum in vitro responses associated with proliferation, cytoskeletal reorganization, cholesterol transport and metabolism, fatty acid metabolism, oxidative stress and carbohydrate synthesis, were consistent with reported in vivo effects on gene expression. Cellular proliferation Several E2-elicited gene expression changes are indicative of proliferation. At 2 hrs Kit and F03 oncogene transcripts, early proliferation indicators, were up-regulated followed by the induction of genes involved in G1 9 S transition such as Calmodulin 3 (Calm3), G1 9 S phase transition 1 (Gspt1), polo-like kinase 2 (PIk2), protein phosphatase 3, catalytic subunit, alpha isoform (Ppp3ca), and protein kinase, cAMP dependent regulatory, type 1, alpha (Prkar1a)). All except Prkar1a, which is up-regulated in proliferative cancer lines (149), have been associated with estrogen-mediated action (150-152), and possess putative EREs (26). Moreover, E2 induction of epidermal growth factor receptor (ngr) (153), c-fos induced growth factor (Figf) (154), platelet derived growth factor, alpha (Pdgfa) (155) and placental growth factor (P91) (156) are consistent with cell proliferation. Despite these events no significant E2-induced proliferation was observed, consistent with the lack of proliferation of human HepG2 cells (157) and mouse liver (73), suggesting that these genes are not significant for proliferation or that other signaling responses are involved that negate E2-elicited proliferation signals in hepatic tissue. For example, G1 -) S 65 transition gene integrin beta 1 (ltgb1) was down-regulated, contrary to E2 induction seen in MCF-7 cells (158), along with the down regulation of G2 9 M transition genes, protein phosphatase 1D magnesium-dependent, delta isoform (Ppm1d) and protein kinase, cAMP dependent regulatory, type II beta (Prkar2b), which have not been previously reported to be E2 responsive. Interestingly, these responses were also repressed in mouse uterine tissue following the uterotrophic response (159). Cytoskeletal organization Proliferation also involves the rearrangement of actin filaments and microtubules for cellular reformation through polymerizing and depolymerizing reactions. Upon estrogen treatment, actin monomer genes, actin, alpha2, smooth muscle, aorta (Acta2) (160), and actin polymerizing genes, actin related protein 2/3 complex, subunit 5 (Al‘pC5) were induced along with FYVE, RhoGEF and PH domain containing 1 (ng1) whose protein product interaction with Cdc42 GTPase may activate actin filament restructuring (161). Some Arpc homologues, which define actin filament polarity, have been shown to be estrogen responsive (162) and may suggest a role for induced actin related protein 2/3 complex subunit transcripts (Arch, Arpc4 and Arp05). Estrogen has been shown to modulate tubulin polymerization at the protein level (163). This study identifies E2-induction of tubulin monomer transcripts, tubulin, beta 4 (Tubb4) and tubulin, gamma 2 (Tung). These subunits may be remodeled by depolymerizing gene, kinesin family member 2A 66 (Kif2a) and through microtubule interacting organization genes microtubule-actin crosslinking factor 1 (Macf1), microtubule-associated protein 1 light chain 3 beta (Map1lc3b) and microtubule-associated proteins (Mtap2 and Mtap4) which are induced following estrogen exposure. Estrogen-induction of keratin (i.e., Krt1-17, Krt1-19, Krt2-7 and KrtZ-8) may also be preparatory for morphological changes (164,165). However, all were repressed by 24 hrs, possibly in response to the lack of proliferation. Myosin genes Myh6, Myl4, Myl7 and Myo1b are cytoskeletal components involved in cellular motility and their induction may be a possible morphological determination factor following estrogen exposure. Thus far, estrogen has only been reported to modulate myosin heavy chain expression (166). Despite numerous cytoskeletal reorganization gene expression events, E2 did not induced dramatic changes in cellular morphology (data not shown). It has been suggested that these changes may be in anticipation of pending physiological alterations that require additional signaling (73). In contrast, similar gene expression changes in the uterus yield a dramatic physiological response (144,159). Cholesterol transport and metabolism Estrogen lowers serum cholesterol by decreasing LDLzHDL (low density lipoprotein : high density lipoprotein) ratios through increased cellular cholesterol uptake, thus retarding atherosclerotic progression (70). Early induction of LDL receptor transcripts such as Lrp1, Lrp10 and estrogen responsive Vldlr (167) 67 suggests enhanced cholesterol uptake by Hepa-1c1c7 cells. However, secreted components of the VLDL cholesterol carrier, apolipoproteins ApocZ and Apoe, are down-regulated, an effect that is contrary to estrogen’s anti-atherosclerotic activity, in mouse liver (168), primates and human HepG2 cells (169). Lipoprotein lipase (Lpl) cleavage of VLDL is ApocZ-dependent; however, its repressed transcript levels also suggest reduced VLDL synthesis. These inconsistencies between pro- and anti-atherosclerotic signals may be due to the lack of circulating cholesterol in vitro, and therefore its carrier protein expression becomes unnecessary. Estrogen treatment also affects the cholesterol synthesis pathway. ER- mediated induction of ngcr, the rate-limiting enzyme involved in cholesterol synthesis, increases buffering by building resistance to dietary cholesterol (170). Although ngcr was not detected to be significant in this data set, cholesterol synthesis genes 3-hydroxy-3-methylglutaryl-Coenzyme A synthases (ngcs1 and ngcs2) and phosphomevalonate kinase (vak) were up-regulated. However, squalene epoxidase (Sq/e), a down-stream synthesis gene, was repressed possibly providing feedback to inhibit cholesterol synthesis. Only vak and Sqle have been reported to be estrogen responsive (171), but all contain a response element in their promoter for sterol regulatory element binding factor 1 (Srebf1), which is ER regulated (172). Srebf1 is an endoplasmic reticulum protein that is regulated through Scap cleavage. Scap activity is determined by its release from insulin induced gene 1 (Insig1), which was induced at 2 hr by E2, also reported in the rat uterus (173), and repressed at 24 68 hr, in the presence of low cholesterol. This potential regulatory network is consistent with the early up-regulation of cholesterol synthesis genes to support cell membrane synthesis for E2 induced growth and proliferation (174). Fatty acid oxidation Estrogen is important for lipid homeostasis and regulates a number of [3- and w-oxidation genes as demonstrated by hepatic lipid accumulation in aromatase null mice (175,176). At 8 hrs, several mitochondrial fatty acid oxidation genes were induced. This includes acyl-Coenzyme A dehydrogenase, short chain (Acads), which initiates oxidation, acetyl-Coenzyme A dehydrogenase, long-chain (Acadl), which breaks down branched and saturated fatty acids, and hydroxyacyl-Coenzyme A dehydrogenase/3-ketoacyI-Coenzyme A thiolase/enoyl-Coenzyme A hydratase (trifunctional protein), beta subunit (Hadhb) which hydrates trans double bonds. In addition, peroxisome straight- chain oxidation transcripts acyl-Coenzyme A oxidase 1, palmitoyl (Acox1), butyryl Co-enzyme A synthetase 1 (Bucs1) and enoyl coenzyme A hydratase 1, peroxisomal (Ech1) transcript levels were induced. Although these genes have not been shown to be estrogen responsive, it is consistent with other studies showing that estrogen treatment increases fatty acid oxidation in mice and rats (177,178). 69 Oxidative stress E2 hydroxylation and subsequent oxidation to quinines along with the induction of peroxisomal B-oxidation of fatty acids may generate reactive oxygen species (ROS) (179,180). Early E2-induced defensive responses include transaldolase 1 (Taldo1), which protects against ROS intermediates, and catalase (Cat) which neutralizes peroxide synthesis (175). Furthermore, the elimination of reactive metabolites is facilitated by induced glutathione-S- transferases (Gsta2, Gsta3 and Gstm3) with glutathione replenishment supported by glutathione reductase 1 (637) induction, although in vivo it was repressed by EE (73). These responses to oxidative stress are accompanied by increases in cytochrome P450 enzymes Cyp2$1 and Cyp4b1 for further metabolism and elimination, consistent with reports of other E2-induced isoforms (73,181,182). Although Hepa-1c1c7 cells maintained in stripped serum conditions are responsive to E2, the transcriptional changes did not reflect the diversity of estrogen-induced responses reported in the mouse liver. Their limited capability to model in vivo estrogen elicited responses is not surprising and can be attributed to many factors (e.g., hepatoma vs. normal tissue, 2-D vs. 3-D environment, homogeneous cell population vs. multicellular tissue, lack of systemic immunological effects, and differing pharmacodynamic and pharrnacokinetic capacity). Gene-specific discrepancies between models may be a consequence of switch-like responses where treatment triggers a cell from a 70 gene expression “off” state to an “on” status (183). Consequently, genes may not be identified as active if an insufficient number of cells are triggered. Thus, models may exhibit divergent responses if gene-specific thresholds differ and are not met in both systems. Nevertheless, several responses associated with proliferation, cytoskeletal reorganization, cholesterol transport and metabolism, fatty acid metabolism, and oxidative stress were well conserved. More importantly, most genes commonly expressed between the stripped serum in vitro model and C57BU6 liver samples were temporally co-expressed or exhibited a temporal shift in which in vitro responses preceded an equivalent in vivo response. Therefore, Hepa-1c1c7 cells in stripped serum conditions can serve as an appropriate model to further investigate selected in vivo estrogen- mediated hepatic mechanisms. 71 CHAPTER 3 COMPARATIVE TEMPORAL AND DOSE-DEPENDENT MORPHOLOGICAL AND TRANSCRIPTIONAL UTERINE EFFECTS ELICITED BY TAMOXIFEN AND ETHYNYLESTRADIOL IN IMMATURE, OVARIECTOMIZED MICEZ. ABSTRACT Uterine temporal and dose-dependent histopathologic, morphometric and gene expression responses to the selective estrogen receptor modulator tamoxifen (TAM) were comprehensively examined to further elucidate its estrogen receptor-mediated effects. These results were systematically compared to the effects elicited by the potent estrogen receptor ligand 17a- ethynylestradiol (EE) to identify pathways similarly and uniquely modified by each compound. Three daily doses of 100 pg/kg TAM elicited a dose—dependent increase in uterine wet weight (UWW) in immature, ovariectomized CS7BU6 mice at 72 hrs with concurrent increases in luminal epithelial cell height (LECH), luminal circumference and glandular epithelial tubule number. Significant UVWV and LECH increases were detected at 24 hrs after a single dose of 100 pg/kg TAM. cDNA microarray analysis identified 2235 differentially expressed genes following a single dose of 100 pg/kg TAM at 2, 4, 8, 12, 18 and 24 hrs, and at 72 2 Data contained in this chapter have been published. Fong CJ, Burgoon LD, Williams KJ, Forgacs AL, Zacharewski TR. 2007. Comparative temporal and dose-dependent morphological and transcriptional uterine effects elicited by tamoxifen and ethynylestradiol in immature, ovariectomized mice. BMC Genomics 8:151. 72 hrs after three daily doses (3x24 hrs). Functional annotation of differentially expressed genes was associated with cell growth and proliferation, cytoskeletal organization, extracellular matrix modification, nucleotide synthesis, DNA replication, protein synthesis and turnover, lipid metabolism, glycolysis and immunological responses as is expected from the uterotrophic response. Comparative analysis of TAM and EE treatments identified 1209 common, differentially expressed genes, the majority of which exhibited similar profiles despite a temporal delay in TAM elicited responses. However, several conserved and treatment specific responses were identified that are consistent with proliferation (Fos, Cdkn1a, Anapc1), and water imbibition (Slc3033, Slc30a5) responses elicited by EE. Overall, TAM and EE share similar gene expression profiles. However, TAM responses exhibit lower efficacy, where responses unique to EE are consistent with greater proliferation potential and water imbibition. 73 INTRODUCTION Tamoxifen (TAM) treatment is an adjuvant therapy prescribed for estrogen receptor positive breast cancers. TAM and its metabolites, 4-hydroxytamoxifen (4OH-TAM), N-desmethyltamoxifen (DMT) and 4-OH-N-desmethyltamoxifen (endoxifen), exhibit antiestrogenic activities by competitively inhibiting the binding of potent agonists to the estrogen receptor (ER) thus antagonizing their proliferative effects (53,184-186). Despite the high therapeutic index of TAM for breast cancer, there are concerns regarding the increased occurrence of uterine cancer as early as 2 years after initiating treatment (187). Although there is no direct evidence that it initiates or promotes uterine cancer, TAM exhibits partial ER-agonist activity by inducing uterotrophy in immature and ovariectomized rodents (188,189). Consequently, a more comprehensive comparison to full agonists is warranted to further elucidate the uterine gene expression effects responsible for its partial agonist activity. TAM is classified as a selective estrogen receptor modulator (SERM) as a result of its differential effects in breast and uterine tissues (190). A number of factors influence the specificity and efficacy of SERM-bound, ER-mediated gene expression, and the subsequent physiological effects. This includes differences in tissue-specific ER isoform expression levels, ligand-induced ER topology, chromatin structure, and coactivator expression and distribution (46,60), thus making the ER an ideal target for drug discovery and development. For example, raloxifene, a second-generation SERM, has been approved for osteoporosis and studies also support its use for breast cancer (191). 74 The uterotrophic assay is a well established method to evaluate the estrogenicity of a compound as measured by ER-mediated increases in uterine wet weight, making it an ideal model for comparing Won-ethynylestradiol (EE) and TAM elicited effects (87). The uterotrophic response also provides well characterized phenotypic hallmarks that facilitate the interpretation of gene expression changes and their function. Early studies have shown that TAM elicits a weaker uterotrophic response than 17B-estradiol (E2) in an immature rodent model (47), however, the mechanisms for its partial agonist activity are not well understood. Genome-wide expression analysis, phenotypically anchored to tissue level effects, provides a comprehensive strategy to identify differential gene expression important in the ER-induction of uterine wet weight. In this report, we extend previous studies examining ER-mediated induction of uterine wet weight (73,144,159) by identifying conserved and divergent uterine tissue and gene expression responses elicited by TAM when compared to EE, an orally active full agonist that mimics the effects of E2 (41). Comparative analysis found conserved gene expression responses that exhibited lower efficacy, consistent with the weak agonist activity of TAM, as well as divergent responses unique to EE that partially explain the lack of TAM-induced water imbibition. 75 Methods Animal husbandry and treatment Female C57BU6 mice, ovariectomized by the vendor on postnatal day (PND) 20, were obtained from Charles River Laboratories (Raleigh, NC) on PND 25. Groups of five mice were housed in polycarbonate cages bedded with cellulose fiber chips (Aspen Chip Laboratory Bedding, Northeastern Products, Warrensberg, NY) in a 23°C environment with 30-40% humidity and a 12 h light/dark cycle (0700 — 1900 h). Animals had access to deionized water and Harlan Teklad 22/5 Rodent Diet 8640 (Madison, WI) ad libitum and acclimatized for 4 days prior to treatment. For the dose response study, animals (n = 5) were orally gavaged with 0.1 mL of 1, 3, 10, 30, 100, 300 or 1000 ug/kg b.w. tamoxifen (2 99% pure, trans-2-[4-(1,2-Diphenyl-1-butenyl)phenoxy]-N,N- dimethylethylamine) (Sigma Chemicals, St. Louis, MO), 100 ug/kg b.w. 17a- ethynylestradiol (EE; 17a-Ethynyl-1,3,5(10)-estratriene-3,178-diol) (Sigma) or sesame oil vehicle (Sigma) alone. Standard uterotrophic regimen was followed (87), consisting of three daily doses followed by sacrifice 24 hrs after the final treatment, (3 x 24 hrs). Doses were prepared based on average animal weight. For the time course study, animals (n = 5) were orally gavaged once or three times daily (3x24) with 100 pg/kg b.w. TAM or vehicle alone and sacrificed at 2, 4, 8, 12, 18 and 24 hrs after treatment in addition to 3x24 hrs treatment group. Animals were sacrificed by cervical dislocation and animal body weights were recorded. The uterus was transected at the border of the cervix, and stripped of extraneous connective tissue and fat. Whole uterine weights were recorded 76 before (wet weight) and after blotting (blotted weight) under pressure with absorbent tissue. A 6-8 mm section of uterine horn was not blotted and placed in 10% neutral buffered formalin (NBF) for histological preparation while the remainder was snap frozen in liquid nitrogen and stored at -80°C for RNA extraction. All procedures were performed with the approval of the Michigan State University All-University Committee on Animal Use and Care. Histological processing, morphometric and pathological analysis Samples stored in 10% NBF were allowed to fix for at least 24 hrs at room temperature then placed into tissue cassettes and stored in 30% ethanol holding solution at 4°C. Paraffin embedding, 5 pm sectioning, mounting and hematoxylin and eosin staining were completed by the Michigan State University Laboratory for Anatomical Histology and Molecular Sciences according to standard techniques (192). Pathological assessments were evaluated according to standardized National Toxicology Program (NTP) pathology codes. Morphometric analysis was performed on midhorn uterine cross sections for all animals (n = 5 per treatment group) using Scion Image analysis software (Scioncorp, Frederick, MD). Histological markers of uterotrophy, including luminal epithelial cell height (LECH), luminal circumference and number of endometrial glands were quantified for each slide. Statistical analysis of morphometric data was assessed by Dunnett’s or two-way ANOVA followed with Tukey’s HSD post hoc analysis to examine dose dependent and temporal effects, respectively (SAS version 9.1). 77 RNA isolation Briefly, 1.0 mL of Trizol (lnvitrogen, Carlsbad, CA) was added to the frozen uterine tissue in a 2.0 mL microfuge tube and homogenized in the presence of steel beads by a Mixer Mill 300 homogenizer (Retsch, Germany). Total RNA was isolated and extracted according to the manufacturer’s protocol and resuspended in The RNA Storage Solution (Ambion, Austin, TX). RNA samples were quantified spectrophotometrically (A250) and assessed for quality by Azeo/Azao ratio as well as inspected using denaturing agarose gel electrophoresis. Microarray hybridization and analysis Custom in-house cDNA arrays consisting of 13,361 features, representing 7,952 unique genes (Unigene Build 144), were spotted on epoxy coated glass slides (SCHOTT Nexterion, Germany) using an Omnigrid arrayer (GeneMachines, San Carlos, CA) and Telechem Chipmaker 3 pins in a TeleChem CHP3 printhead head (Telechem International Inc., Sunnyvale, CA) by the Research Technology Support Facility at Michigan State University (128). Selected clones were obtained from EPAMAC (129), Research Genetics, the National Institute of Aging and Lion Biosciences. Detailed protocols for processing of microarrays are available at the deach Home Page (130). An independent reference study design was used to assess treatment effects (73). For the dose response study, each treatment group was hybridized to a single vehicle pool utilizing 14 arrays, including dye swaps, and 3 biological 78 replicates for a total of 42 arrays. For the time course study, each time-matched treated and vehicle sample was competitively hybridized utilizing 14 arrays, including dye swaps with 3 biological replicates for a total of 42 arrays. The Genisphere 900 3DNA Array Detection (Genisphere lnc., Hatfield, PA) indirect incorporation kit was used to generate cDNA samples for hybridization. Briefly, 1 ug of RNA was reverse transcribed in the presence of an oligo-tagged primer specifically targeted for Cy3- or Cy5- conjugated dendrimers. The cDNA was resuspended in 58 pL of 2X Formamide-Based Hybridization Buffer and hybridized overnight on arrays sealed in a light-shielded, humid chamber submerged in a 42°C water bath incubation. Slides were then washed in SSC solutions containing decreasing concentrations of SDS, spin-dried and re- hybridized with a Cy3sz5 (1:1) dendrimer mixture in formamide based buffer to indirectly incorporate dyes at the Cy3— and Cy5-dendrimer—tagged cDNA hybridized on the first day. Slides were washed and dried as previously described, and scanned at 635 nm (Cy3) and 532 nm (Cy5) using a 428 Affymetrix Scanner (Santa Clara, CA). Images were examined, features identified and intensity values recorded using GenePix v.5.1 (Molecular Devices). Microarray quality control, statistical analysis and gene list filtering All arrays in this study were compared to a historical data set of high quality arrays. Parameters assessed included background signal intensity, feature signal intensity, feature vs. background signal intensity ratios, the number of features with background intensities greater than the feature intensity for each array, and relationships between feature and background signal intensities. All 79 arrays surpassed the quality control parameters established in this laboratory (193). Data were normalized using a semi-parametric approach (194) and model-based t-values were calculated comparing time-matched treated and vehicle samples. Posterior probabilities of activity [P1(0-value] were then calculated on a per-gene and per-time point basis using an Empirical Bayes analysis (135). Gene lists were initially filtered based on posterior probability (P1(t) > 0.999) and fold—change cut-off (|fold changel > :1: 1.5) resulting in an active gene list on which further functional analysis was conducted. All raw and analyzed data were stored in deach (130), a Minimum Information About Microarray Experiments (MIAME)-supportive relational database (131) running under Linux/Oracle 109. deach currently supports microarray data storage, retrieval, and querying as well as facilitates data analysis, sharing and reporting (133). Active gene lists exclusive to TAM and EE were also generated. Data for the EE time course has previous been published (159). The TAM unique gene list was generated based on relaxed criteria (P1(t) > 0.9 and |fold changel > :I: 1.4 cut-off) to obtain a liberal EE-mediated gene list which was then excluded from the original TAM unique gene list using P1(t) > 0.999 and |fold changel > :t 1.5 criteria. The EE unique gene list was generated using a reciprocal approach (i.e., relaxed criteria (P1(t) > 0.9 and |fold changel > :1: 1.4 cut-off) to obtain a liberal TAM-mediated gene list which was then excluded from the original EE unique gene list using P1(t) > 0.999, and |fold changel > d: 1.5 criteria). This 80 approach ensured that genes marginally missing the cutoffs were not included in the compound-unique list. Estrogen response element searches were completed by comparing Gene Symbols to the computationally identified list compiled by Bourdeau et al. (26). Quantitative RT-PCR Aliquots of RNA isolated from each of the five replicates were set aside for SYBRT"l Green quantitative real-time PCR (QRT-PCR) verification. EE-treated, temporal mouse uteri RNA were previously isolated (159). An oligo—dT anchored Superscript II (lnvitrogen) reverse transcriptase reaction was carried out on 1 pg of RNA, in a 20 pL reaction, from each biological sample as per manufacturer’s instructions. Samples were diluted four-fold and 3 pL used in a 30 uL real-time reaction mix containing 1X SYBR Green PCR buffer, 3 mM MgCl2, 0.33 mM dNTPs, 0.5 IU AmpliTaq Gold (Applied Biosystems, Foster City, CA) and 0.15 mM forward and reverse primer. All primers were designed by submitting cDNA microarray clone sequences into Primer3 (138) to obtain an amplicon of approximately 125bp (Supplemental Table 6). PCR amplification was conducted in 96-well MicroAmp Optical plates (Applied Biosystems) on an Applied Biosystems PRISM 7000 Sequence Detection System under the following conditions: 10 min denaturation and enzyme activation at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. After amplification, a 30 min dissociation protocol was conducted to assess primer specificity and product uniformity. Each plate contained duplicate standards of purified PCR product of 81 known template concentration over eight orders of magnitude to generate a log template concentration standard curve. No template controls (NTC) samples were included on each plate such that experimental samples within 2 standard deviations of the NTCs are considered below the limits of detection. Plots were visualized and thresholds determined using ABI Prism 7000 SDS Software (Applied Biosystems). Results were normalized to a geometric mean of beta- actin (Actb), glyceraldehyde-3-phosphate dehydrogenase (Gapd) and hypoxanthine guanine phosphoribosyl transferase (Hprt) mRNA levels to control for differences in RNA loading, quality and cDNA synthesis. Statistical significance of expression differences between vehicle and TAM treated samples were assessed by two-way ANOVA followed by Tukey’s HSD post hoc analysis to examine treatment and treatment over time effects (SAS version 9.1). Correlation analyses of QRT-PCR and microarray data generated using the correlation function of R v2.1.0. lmmunohistochemistry Rabbit polyclonal antibodies specific for PCNA were purchased from Abcam, Inc. (Cambridge, MA) and staining localized using manufacturer’s instructions for the Vectastain Elite ABC Kit (Vector Laboratories, Burlingame, CA). Briefly, paraffin-embedded uterine sections were placed on glass slides, deparaffinized in xylene and re-hydrated through a series of decreasing ethanol concentration washes ending in ddH20. Endogenous peroxidases were quenched in 0.3% H202 in methanol solution (30 min) followed by boiling (15 82 min) in a 10 nM sodium citrate solution (pH 6.0) for antigen retrieval. To minimize nonspecific background staining, sections were blocked with normal goat serum (Vector Laboratories) for 20 min. The slides were incubated for 1 hr with the primary rabbit anti-PCNA polyclonal antibody (1:500 dilution in PBS), followed by 30 min each with biotinylated goat anti-rabbit antibody (Vector Laboratories) (1:400) and ABC reagent (Vector Laboratories). A single PBS rinse was performed between incubations with each antibody. Localization of antigen was obtained using Vector® NovaRED (Vector Laboratories). The sections were counterstained with hematoxylin. 83 RESULTS Uterine weight Increases in uterine wet weight (UWW) in rodents after three daily subcutaneous doses of TAM is well documented (84,195). Dose-dependent increases in uterine weight (ECso = 33.7 pg/kg) were observed following three consecutive daily oral treatments of TAM (Figure 1A), however induction plateaued at 5-fold, compared to 11-fold with an equivalent dose of 100 pg/kg Wei-ethynylestradiol (EE) (159). Comparison of wet and blotted uterine weights indicated no significant water imbibition in TAM-treated uteri. However, blotted EE-treated uteri were larger, consistent with past reports that TAM induces a less efficacious uterotrophic effect (196). In order to establish a temporal profile, the uterotrophic effects of 100 pg/kg TAM were also investigated at 2, 4, 8, 12, 18, 24 and 3x24 hrs. A significant 2.5-fold increase was observed at 24 hrs after a single 100 pg/kg TAM dose (Figure 1B) which was delayed compared to the significant increase seen with 100 (1ng EE at 18 hrs (159). Morphometric analysis and histopathology Luminal epithelial cell height (LECH), luminal circumference and number of endometrial glands are hallmarks of estrogen action in the rodent which correlate with UWW induction (197). Significant dose-dependent increases in LECH and luminal circumference were initially detected at 30 1.1ng TAM (Table 1A). Interestingly, LECH was not significantly different between 100 84 Figure 1 Tamoxifen-induced dose dependent and temporal changes in uterine weight Graphs illustrate fold—change increases in uterine wet (open) and blotted (solid) weight. A) Tamoxifen elicits a dose dependent uterotrophic response (E050 = 33.7 pg/kg) and achieves maximal induction of approximately 5-fold following three daily doses (3 x 24 hrs) of 100 pig/kg TAM. Significant increases (p < 0.05, n = 5) are denoted by an asterisk (*). In contrast, 100 ug/kg EE (positive control) maximally induced uterine wet weight 11-fold (*, p < 0.05, n = 5) with significant water imbibition (if; p < 0.05, n = 3), while TAM only achieved 50% uterotrophic efficacy and no water imbibition. B) A single dose of 100 ug/kg TAM significantly increased uterine wet weight as early as 24 hrs after administration. No significant water imbibition was observed at any time point. 85 Figure 1 A LlerusVlbight-Dosefbsponse Fold Change B Fold Change 12.5- 10.04 51 ‘l‘ p: 9 N ‘l' p 1’ “99°‘E‘P‘PT‘QP‘P A I ? V *# 1 3 10 so 100 30010005: TAMdosefilglkg) Llerus Weight - Time Course * 2 * Ill 4 8 12 18 243x24 Trrre(Irs) 86 Table 1. TAM- and EE-induced uterine morphometric changes A) Dose Response (3 x 24 hr) Luminal Avg. Number of TAM Dose Epithelial Cell Luminal Glandular (Hg/kg) Height (pm) Circumference (mm) Tubules 0 8.75 :t 0.86 0.77 1: 0.14 1 1 8.99 :I: 1.00 0.72 :l: 0.12 0 3 10.91 :I: 2.97 1.17 t 0.41 1 10 107311.15 1.17:0.29 3 30 15.12 :I: 1.55* 1.87 :I: 0.26* 5* 100 24.58 :I: 2.79* 3.60 :I: 0.27* 10* 300 27.08 :I: 3.79* 2.68 1: 1.19* 5* 1000 31.30 :I: 2.25* 3.05 :1: 0.73* 5* 100 EE 28.94 :I: 3.35* +++“I 4 3) Time Course (100 W) Luminal Epithelial Cell Luminal Time (hrs) Height (pm) Circumference (mm) 2 9.98 :t 1.68 0.79 1: 0.19 4 8.61 :l: 1.58 0.80 :1: 0.06 8 10.06 :1: 2.50 0.96 i 0.29 12 9.46 1: 1.28 0.99 :l: 0.21 18 9.18:1.03 1.291042 24 11.08 :I: 1.94* 1.22 :l: 0.42 3x24 28.61 i 7.50* 2.85 :l: 1.83* * Statistically different from time matched vehicle (p < 0.05) a Lumen larger than 100x field of view, accurate measurements could not be made at 40x magnification Time course vehicle samples are not significantly different from each other. 87 pg/kg EE and TAM, although the luminal circumference of EE uteri was greater with more pronounced invagination of the luminal glandular epithelium (Figure 2). There was also mild to moderate hypertrophy in the stromal nuclei at 10 uglkg TAM with moderate epithelial hypertrophy and hyperplasia at 30 pg/kg TAM, which was marked at higher doses. Mild edema was noted for all samples beginning at 100 pg/kg TAM. Marked to severe stromal nuclei hypertrophy and epithelial hypertrophy and hyperplasia, all with mild edema, was observed at 100 pg/kg EE. Mild to moderate stromal edema was observed as early as 12 hrs following after a single 100 pg/kg TAM dose, while increased UWW and LECH were not significant until 24 hrs (Table 1B). No significant increase in luminal circumference was observed in the first 24 hrs after treatment. Uterine endometrial glands synthesize and secrete fluids in preparation for conceptus, implantation and growth. Significant increases in the number of glands was observed at 30 rig/kg TAM (Table 1A) in the absence of a dose responsive increase, which may be an artifact of histological sampling of the uterine horn. Similarly, EE-treated uteri exhibited an increased number of endometrial glands that was not statistically significant. Uterine gene expression changes elicited by tamoxifen Differentially expressed genes in the dose and time dependent studies were identified based on their empirical Bayes posterior probability of activity [P1(t)-value] on a per-gene, per-time point basis. P1(t)-values approaching 1.0 indicate a greater likelihood of treatment-related differential gene expression. 88 Figure 2 Uterine histology Hematoxylin and eosin stained sections of uterine tissue at 100x magnification after three daily doses of A) sesame oil, B) 1 mg/kg TAM and C) 100 ug/kg EE. TAM and EE treatment induced increases in luminal epithelial cell height. Luminal circumference is increased to a greater degree by EE than TAM. Bars represent 20 pm. 89 Figure 2 90 Using P1(t) > 0.999 and |fold changel 2 1.5 as selection criteria, a prioritized list of 2941 features, representing 2235 unique Entrez Gene annotated genes, were identified in the temporal study with 55% of the genes exhibiting induction and 45% repression (Supplemental Table 1). Differential expression levels ranged from 14.3-fold repression (tight junction protein 4, ij4) to 28.1-fold induction (arginase 1, Arg1), further demonstrating the responsiveness of the uterus to tamoxifen. Using the same selection criteria (P1(t) > 0.999 and |fold changel of 2 1.5) at a minimum of three doses, to ensure dose responsiveness, 1630 features, representing 1036 unique Entrez Gene-annotated genes, exhibited dose dependent expression (Supplemental Table 2). Of the 1036 genes exhibiting a dose-dependent response at 3x24 hrs and of the 738 differentially expressed genes at 3x24 hrs in the time course study, 691 genes (94%) were in common, demonstrating good reproducibility between experiments. Differentially expressed genes were associated with cell growth and proliferation, cytoskeletal organization, extracellular matrix modification, nucleotide synthesis, DNA replication, protein synthesis and turnover, lipid metabolism, glycolysis and immunological responses. The temporal changes in gene expression were best represented using five k-means clusters: A) induced at 12 and 24 hrs, B) induced and sustained from 24 — 72 hrs, C) induced late at 72 hrs, D) repressed between 8 - 24 hrs and E) repressed and sustained from 24 - 72 hrs (Figure 3). The majority of TAM-elicited differential expression occurred after 12 hrs with only 42 features (26 genes) exhibiting differential gene expression between 2 and 8 hrs, in marked contrast to EE studies where 91 Figure 3 Tamoxifen-induced temporal gene expression patterns Five k-means clusters best represent the general temporal patterns for the 2941 features differentially expressed following TAM treatment. Note the 8 hr delay in gene expression response especially in comparison to EE elicited gene expression is speculated to be due to the delayed absorption of TAM. Inset numbers indicate the number of features represented by each cluster. Black pseudolines indicate the general profile represented within each cluster. 92 an vu 2.5 o N vu 0;? o .1. ,TL eBuaqo [)le n 0.52... 93 significant gene expression changes occurred prior to 8 hrs (144,159,198). The temporal pattern of differential gene expression correlates with the histology results which indicate a delayed response in comparison to EE. Eleven genes, representative of affected pathways and exhibiting different temporal gene expression patterns (i.e. cytoskeletal organization (Krt2-4), signal transduction (Igf1), immunological responses (I17), acid-base homeostasis (Car3) and lipid transport (Fabp5, Vldlr)), were verified by QRT-PCR and exhibited good agreement with microarray results. Correlation coefficients ranged from 0.46 to 0.97 (mean = 0.80) (Figure 4). lmmunohistochemistry (IHC) was also used to assess and localize PCNA protein expression following TAM treatment (Figure 5). Microarray results indicate a 2.5-fold increase in Pcna transcript levels between 12 - 18 hrs after treatment with IHC, confirming elevated protein expression in epithelial and stromal cells in 12 hr TAM treated samples when compared to time matched controls. Comparison of common temporal TAM and EE gene expression data Temporal TAM data were compared to an analogous EE study using the same immature, ovariectomized C57BU6 mouse model (159). Employing the P1(t) > 0.999 and |fold change] 2 1.5 criteria, 2657 unique annotated genes were differentially expressed following treatment with 100 pg/kg EE, of which 1209 were also activated by TAM (Supplemental Table 3). Agglomerative hierarchical clustering of common genes by treatment and time indicates that the 12 hr TAM 94 Figure 4 Quantitative real-time PCR verification of selected TAM-induced genes Overall, the microarray results for 14 TAM- and EE-induced genes were verified using QRT-PCR. The verified genes represent various affected pathways and different temporal patterns of expression. Overall, there was good correlation (average p = 0.8) between microarray (lines) and QRT-PCR (bars) data. Examples for six of the genes are illustrated. Statistically significant QRT-PCR differences (p < 0.05, n = 4) due to treatment are denoted by an asterisk (*). 95 eggs EVE: 8.882924" 8.882984“ a . a m J .... ...l- .1 L My... W 8 m. d m . 3 o o .w w 2m u 99W .—.5 w .. .. r f mama F mass 2 ...—65 EVE; W58: ... 56832.2..." _C .. A A w. u. m m N P P 3 w m m D. m m w 3 a 998.6 0.8 v 0.59". 96 Figure 5 Immunohistochemical detection of differential Pcna protein levels due to TAM Twelve-hour vehicle (A) and TAM (B) treated uteri sections were immunohistochemically stained (NovaRED®) with Pcna specific antibodies. Treated samples have darker nuclear staining, indicating greater levels of Pcna protein expression, in agreement with the histological assessment and changes in gene expression associated with cell proliferation. Increased Pcna expression is more pronounced in the luminal and glandular epithelium, and stroma (arrows). Tissues were counter-stained with hematoxylin. Images are representative of four biological replicates. Bars represent 20 pm. Color representation of this figure may be found in: Fong CJ, Burgoon LD, Williams KJ, Forgacs AL, Zacharewski TR. 2007. Comparative temporal and dose-dependent morphological and transcriptional uterine effects elicited by tamoxifen and ethynylestradiol in immature, ovariectomized mice. BMC Genomics 8:151. 97 Figure 5 98 response is most similar to the 4 hr EE response, followed closely by 8 hr TAM (Figure 6). Interestingly, TAM and EE exhibit similar gene expression profiles at 24 and 72 hrs, suggesting that the delay in some TAM-elicited responses is not maintained at later time points. Expression profiles were compared for the 1209 differentially expressed genes that were regulated by TAM and EE. These genes were categorized as Similar, more Efficacious by EE or TAM, or Ambiguous (Table 2). A total of 793 genes (66%) exhibited expression profiles that were similar in pattern and efficacy when a temporal shift, due to delayed TAM response, was considered. Interestingly, 28 genes that were differentially expressed at least 2-fold more by EE when compared to TAM (i.e., EE Efficacious genes) were associated with cell growth, regulation of transcription and protein metabolism and transport including Fos (6.4-fold by EE; 4.1-fold by TAM) and Inhbb (7.6-fold by EE; 3.2-fold by TAM). These genes are involved in cell cycle regulation and cellular growth, respectively, and possibly support the greater physiological effect exhibited by EE. In contrast, 19 genes were modulated 2-fold or greater by TAM, including an (3.6-fold by EE; 5.5-fold by TAM), which is associated with proliferation inhibition. In general, efficacious TAM elicited responses were associated with receptor-mediated signal transduction, ion transport and protein metabolism. Gene expression comparisons between the two studies were also verified by QRT-PCR. As previously reported, gene expression data is subject to compression (199), and therefore the sensitivity of QRT-PCR data is often greater when compared to microarray data Thus, some genes classified. 99 Figure 6 Temporal comparison of genes commonly activated by TAM and EE Hierarchical clustering of 1209 TAM- and EE-regulated genes (y-axis) identifies subsets of similar profiles according to time and treatment (x-axis). The dendrogram indicates that early responses (4 hrs) to ethynylestradiol (E) are most similar to 8 and 12 hrs tamoxifen (T) responses demonstrating temporally displaced TAM activation consistent with the delayed absorption of TAM. However, temporal displacement of TAM elicited responses is not maintained as EE and TAM responses cluster together at 24 and 72 hrs. Color representation of this figure may be found in: Fong CJ, Burgoon LD, Williams KJ, Forgacs AL, Zacharewski TR. 2007. Comparative temporal and dose-dependent morphological and transcriptional uterine effects elicited by tamoxifen and ethynylestradiol in immature, ovariectomized mice. BMC Genomics 8:151. 100 Figure 6 34.1 no.9 03966 062.820: m8_.o 22222 .2222 2.2222222. 222222 ..2 22.222 :2 22: F2. . ,222222222222221 22222222222222.222.22222522.22 _. 2222222222l2a2222222. 22 222222 22222222222222. 2222222... 222222222 2850 um ecu. s__<._. o2>=o< 2:.oEEom 4E 12T 8T 8E 12E 18E 24E 24T 18T 2T 72E 72T 4T 2E Treatment Group 101 Table 2. Classification of TAM and EE commonly active annotated genes Classification Category Total Genes Similar (8) EE Efficacious (EED TAM Efficacious (TEf) Ambiguous (A) Definition Similar profiles exhibit patterns which are comparable in direction and magnitude across time; this also takes into account temporally shifted responses. Efficacious responses demonstrate similar directional responses, but one compound elicits a greater induction or repression, by at least 2- fold, than the other; this category also includes temporally shifted responses. Gene pairs which did not fall into the previous three categories were labeled as Ant—biguous 102 Number of Annotated Genes 1209 793 28 19 369 as Similar may also be classified as EE— or TAM-Efficacious. For example, microarray data suggested that Cdkn1a response to TAM and EE were comparable, but through QRT-PCR EE induced an 8-fold response compared to a 3.5-fold induction by TAM (Figure 7). TAM and EE responsive genes were also examined for estrogen response elements (EREs) in their promoter regions by comparison to a list of computationally identified sequences (26). EREs were found in 176 TAM-active genes and 218 EE-active genes, with 133 regulated by both compounds. Only 10% of TAM or EE differentially expressed genes possessed an ERE suggesting that other trans-acting factors may also be involved or that EREs were outside of the search regions. Annotation information in public repositories is constantly evolving, thus gene names may have changed or new genes may have been added since the publication. As a result, some genes may be misclassified regarding their ERE status. TAM- and EE-specific gene expression data Gene expression changes unique to either TAM or EE may be another factor contributing to their different uterotrophic responses. An additional filtering method was used to identify genes more likely to be unique to EE treatment which involved excluding an extended list of TAM-regulated genes obtained by relaxing the TAM criteria to P1(t) > 0.9 and |fold change| 2 1.4 from the standard criteria (P1(t) > 0.999; |fold change| 2 1.5) of EE (Figure 8A). The same approach was also used to obtain a list of genes unique to TAM (Figure 88). 103 Figure 7 Examples of TAM and EE differential gene expression classifications Examples of representative genes classified as Similar or Efficacious based on microarray data only. QRT-PCR analysis confirmed the classifications of these genes. In some cases (e.g., Cdkn1a) a gene classified as Similar may also be classified as EE-Efficacious based on QRT-PCR results due to data compression inherent in microarray data. Statistically significant differences (p < 0.05, n = 4) due to treatment are denoted by an asterisk (*). 104 Microarray Fold Change a O ‘— moious Tfl‘l - TAM Tme (l‘I's) TAM Effi afiueqo p|o;| Had-1.80 Microarray Fold Chang e (D In 1' ('0 N 1- <3 l- l! +< R'- H:l ‘2 “ I fi 5’ f: 2% V N D W afiueuo mos Had-180 -k F—flifldfi " Hulk“ carious 12 18 2A TmthS) Fos - EE Fos-TAM EEEffi afiueuo P|0:l Trre (its) “KI—'5 *l—r22:m-§xII-N s a 2 <= afiueuopioa 3x24 Sinilar Cdkn1a - TAM Tme ("5) E l- aBueuo mos Figure 7 Microarray Fold Change momomomo ddddr‘r‘dci Tim ("3) Cdm1a-EE it I—IJ... «a. 3an 33353352: drs'uioici F‘- eBueug plOJ Had-180 Figure 8 Identification of unique EE and TAM differentially expressed genes Treatment specific differentially expressed genes were identified by excluding a list obtained using a more relaxed criteria (P1(t) > 0.9; |fold change| 2 1.4) for one treatment from the differentially expressed genes identified using the standard criteria (P1(t) > 0.999; |fold change| 2 1.5) of the second treatment to identify gene expression changes that were more likely to be unique to one treatment. (A) A liberal list of TAM-induced genes identified, using a relaxed criteria of P1(t) 2 0.9 and lfold change| 2 :l: 1.4, was excluded from the EE differentially expressed gene list using the standard selection criteria of P1(t) 2 0.999 and |fold change| 2 i 1.5 to identify 240 genes more likely to be differentially expressed by EE alone. (B) Using a similar approach, a list of 60 genes more likely to be differentially expressed by TAM alone was generated. Lists of EE and TAM specific genes are provided in Supplemental Tables 4 and 5. 106 Figure 8 A EE-modulated genes TAM-modulated genes P1(t) > 0.999 P1 (t) > 0.9 |fold change| > 1.5 |fold change| > 1.4 unique genes TAM-modulated genes EE-modulated genes P1(t) > 0.999 P1(t) > 0.9 |fold change| > 1.5 |fold change| > 1.4 60 TAM- unique genes 107 This ensures that those genes significant in both treatments and approaching significance in the other treatment are not considered as unique, thus increasing the likelihood of identifying treatment-specific differential gene expression responses. For example, to identify unique EE responses, the 2417 differentially expressed TAM genes that satisfy the P1(t) > 0.9 and |fold change| > 1.4 were excluded from the 2657 differentially expressed EE genes (P1(t) > 0.999; |fold change| 2 1.5) to identify 240 genes unique to EE treatment (Fig. 83; Supplemental Table 4). Similarly, genes more likely unique to TAM were identified by excluding the 2175 differentially expressed EE genes with a P1(t) > 0.9 and |fold change| > 1.4 that were in common with the 2235 differentially expressed TAM genes (P1(t) > 0.999; |fold change| 2 1.5) to identify 60 genes more likely unique to TAM (Supplemental Table 5). Treatment-specific responses exhibited profiles distinctly different in pattern and magnitude from their counterpart (Figure 9) even when taking delays, due to TAM, into consideration. The pathways represented within unique EE-responsive genes include apoptosis regulators (Bok and Pdcd6) and water imbibition (qu8 and SI022a7), consistent with the physiological effects observed. Fewer unique TAM- responsive genes were identified. There was no overrepresentation of any functional pathway consistent with its weaker uterotrophic response. These data suggest that differentially regulated subsets of genes exist that contribute to the distinctive uterotrophic response elicited by each treatment. 108 Figure 9 Temporal expression profiles of TAM and EE-specific genes Graphical representation of genes exhibiting compound-specific responses demonstrated profiles which were distinctly different in pattern and magnitude compared to its non-responsive counterpart. These examples further illustrate that the filtering conditions used were adequate to identify differential responses by TAM and EE. 109 225 82F 2322.: 2.1.5 8.: fiavwmwuwmwmfid évwwwuwmwmwd fimqwmwmwmwmd . .06 n h .F .3“. .54. in 3m .30 .23 flu in n. .m._.B in 8.2.... are see .3 3 .5 3 H88“. H.522 8:8 ouoam-m_ $58: 2228.: 2822952 «wravmwwuwmmmd «gammuwmwmg vutxmvwemwmwm a... .mdd / h ....a .... w 25w m. , . D. .N.—.D. D. 30 3% m .w “Wm M a a 6N8 K... mu 3 2225.26". 2333 5.5.382 $28 050% ... S—(P . m 239E 110 DISCUSSION A comparative approach was used that integrates the gross organ, histopathological, and morphometric uterine effects of EE and TAM with their dose response and temporal gene expression profiles to further elucidate the molecular basis of the partial agonist activity of TAM. TAM treatment induces a 5-fold increase in gross uterine weight following three daily doses compared to an 11-fold increase with EE. In addition, no significant water imbibition was induced by TAM. These effects are well documented and are the basis for the classification of TAM as a partial agonist (84,195,196,200). Moreover, TAM induces a delayed increase in uterine weight when compared to EE which may be partially attributed to its weaker agonist activity but is more likely a reflection of slower absorption (52,201,202). In contrast, peak serum levels of EE are detected within two hours of treatment (203). At equi-efficacious doses of TAM and EE (i.e. 100 vs. 20 pg/kg, respectively), comparable effects on UWW, luminal circumference and glandular epithelial were observed (data not shown), suggesting both treatments proceed through similar changes to achieve uterotrophy. However, at higher doses, TAM does not elicit a comparable gamut of responses as seen with higher doses of EE. Surprisingly, TAM increased luminal epithelial thickness (188), due to cellular hypertrophy and hyperplasia, that was not significantly different from EE, but mediated a smaller increase in luminal circumference with more endometrial glands compared to EE. Although these results appear contradictory, glandular epithelium may arise from the luminal epithelium and appear as highly 111 invaginated regions of the lumen that generate a large secretory surface area (204). Thus, despite fewer endometrial glands in EE samples, its glandular area is greater due to the increased luminal glandular surface area which was not observed in the TAM treated samples. Temporal tamoxifen-elicited gene expression profiles were examined following a single dose as well as after three daily doses of 100 pg/kg TAM. Only 9 features, representing 6 annotated genes, exhibited differential expression at 2 and 4 hrs after TAM treatment compared to 1234 EE genes at the same time points (159), consistent with the delayed histological effects. Of these early TAM responses, only Esr1 and Car3 have been reported to be induced by estrogen (159,205). At 12 hrs, 683 genes were differentially expressed in response to TAM, of which 541 genes were also affected by EE between 2 and 8 hrs (159). Agglomerative hierarchical clustering suggests that genes affected by TAM and EE exhibited comparable gene expression changes despite the delay in TAM responses. Genes regulated by TAM and EE represent a variety of pathways including cell cycle regulation, cytoskeletal re-organization, nucleotide metabolism, immune and complement activation and lipid transport and metabolism, and have previously been associated with eliciting the uterotrophic response (144,159,198,206-208). Similarities in their gene expression profiles suggest that the uterotrophic response involves a defined subset of genes mediated by the ER. Furthermore, greater than 75% of TAM-activated genes with putative EREs (26), were also activated by EE. However, differences in 112 efficacy and responsive genes may partially explain uterotrophic response differences. Despite temporal delays, many genes were regulated by both EE and TAM. Most of these commonly active genes exhibited comparable fold changes suggesting that they do not significantly influence the magnitude of the uterotrophic response. For instance, both treatments equally repressed uterotrophic supportive pro-apoptotic caspases (Casp2 and Casp6) (reviewed in (209)). Although these genes were responsive to EE and TAM, others demonstrated quantitative differences in their expression behavior. Twenty-eight genes, including the proliferation supportive genes Cdkn1a, Fos and Inhbb, exhibited greater EE efficacy consistent with their previously reported estrogen- induced expression (210-212) resulting in a full uterotrophic agonist response. In contrast, 22 genes more highly induced by TAM included GZ/M inhibitor (Sin/14- 3-30), which has been associated with human endometrial carcinomas (213) to reduce proliferation. Many of these quantitative differences in gene expression efficacy are consistent with the potent agonist activity of EE and the weak agonist activity of TAM. There were also treatment-specific gene expression effects. Tentatively, 240 and 60 modulated genes were identified as unique to EE or TAM, respectively. In general, these responses were consistent with uterotrophic activity elicited by EE and TAM. For example, QRT-PCR verified the early induction of mitotic gene, Anapc1 by EE (data not shown). Also, the treatment specific repression of pro-apoptotic Bel-2 member, Bok, and the induction of 113 Pdcd6, an apoptosis regulator, associated with proliferating tissues (214) are consistent with the greater efficacy of EE. Bok has previously been shown to be EE responsive in uteri, whereas Pdcd6 approached the statistical cut-off in a previous study (144). For TAM, QRT-PCR confirmed decreased expression of Sipa1 (data not shown), a repressed response at 24 hrs associated with decreased proliferation (215) that may reduce hyperplasia. DNA synthesis and replication pathways were also differentially regulated. Sustained up-regulation of dNDP phosphorylating genes, Nme1 and Nme6 (216), suggest salvage pathways are emphasized for nucleotide synthesis rather than de novo processes. Consistent with this view Prps1, the first step in purine biosynthesis, is repressed during the same period. These genes are similarly modulated by TAM and EE, suggesting that proliferation may deplete resources for de novo synthesis. Only Nme1 has been previously shown to be EE responsive in rodent uteri (144,159). However, EE uniquely inhibited the de novo pyrimidine synthesis gene, Dhodh [18 - 72 hrs], and induced the nucleotide recycling gene, Nt5m [18 and 72 hrs] (217) suggesting an involvement of salvage pathways to support EE-induced proliferation which have not previously been reported to be estrogen responsive. Water imbibition is a characteristic uterine response to estrogens, involving the increased flow of water to the lumen mediated by aquaporins and ion transporters (218). It does not appear to be a factor in TAM-induced uterine weight increases, as blotted weights were not significantly different from wet weights. qu1 and qu5 are comparably regulated by TAM and EE, while qu8 114 induction was specific to EE (QRT-PCR verified, data not shown). qu8 is a known contributor to water imbibition (219) and its EE-specific response suggests it may play a larger role in the process of a full uterotrophic response. The lack of ion transporter regulation may also be a contributing factor in the absence of TAM-induced water imbibition. The EE induction of zinc transporter, Slc30a3 [12 hrs], which causes ion uptake into various vesicle compartments (220,221) may facilitate stromal edema and has been shown to be responsive to estrogen where it is down-regulated in brain tissue (222). Organic anion transporter, SI022a7, was repressed by EE from 18 - 72 hrs in the uteri suggesting anion retention in the stroma that may also be important for edema. SI022a7 is an importer in the basolateral membrane of kidney tubule epithelia (reviewed in (223)), and is estrogen responsive in the kidney (224). Differential regulation of ATP production genes is also consistent with the greater uterotrophic efficacy of EE. Transcripts associated with oxidative phosphorylation (OXPHOS) complex I, Ndufb8 [8 — 24 hrs], and complex Ill, Uqcr [8 - 18 hrs] and Uqcrh [4 - 18, 72 hrs], were all up-regulated. Although not previously been reported as responsive, collectively, the EE modulation of OXPHOS components is consistent with greater energy demands required to support increasing hypertrophic and hyperplastic activity induced by EE compared to TAM. Other TAM gene expression studies have been conducted using in vitro breast cancer models, primarily MCF-7 cells. Comparisons of differentially expressed gene lists identified minimal to no overlap of TAM responses between 115 in vitro human breast tissue and in vivo mouse uterus (225,226). Only the induction of Uqcrb (227), Nqo1 (228), Tff1, Mapt (229), Pctk3, Wnt4 (230), Myb, Cdc6, CchO, Mcm2, Fos and Mybl2 (231) and repression of ch1, Tgfa (228), Rap1ga1, Blnk, Tm4sf1, Matn2, Ifi30, Tgfb3 and Smpd1 (229) correlated with the changes observed in the current study. Moreover, there are examples of divergent gene expression changes such as inverse responses for an2 (228), Ctsh, Selenbp1, Nfrkb, Cyp1a1 (229), Prps1 and Tmsb4x (230). The long term uterine effects of TAM have also been examined in mice following neonatal exposure. Mice were treated for four consecutive days after treatment and uteri samples examined at various months after dosing (232). 001131 exhibited persistent up-regulation months after treatment and was also induced in our short term study. Several factors, such as model and tissue differences, likely contribute to the minimal overlap including differences in array platforms and genome coverage, study design, and data analysis. For example, E2 and 4OH- TAM were utilized in the in vitro studies while EE and TAM were administered to the mice. Despite the minimal overlap between the models, the activities of TAM, when compared to E2 were comparable. In vitro and in vivo, the gene expression changes elicited by 4OH-TAM were similar to those mediated by E2 in MCF-7 cells. Furthermore, the magnitude of gene expression changes due to 4OH-TAM was attenuated compared to E2 (229,231). Although 4OH-TAM and EE induced similar cell cycle genes, down-stream mechanisms were also regulated to prevent 4OH-TAM mediated cell cycle progression (231). Some of 116 these pathways may play a role in the partial uterotrophic response elicited by TAM in treated mice. Differences in chemical structure may contribute to ligand specific responses. TAM belongs to the stilbene/triphenylethylene family while EE is steroidal. Each has unique binding modes resulting in different ER conformations (186), binding affinities (233,234), ligand-induced binding domain topographies (235), coactivator recruitment capabilities (236,237), gene-specific thresholds of activation, and efficacies (238). Specifically, 4OH-TAM induces a different conformational change in the ER compared to E2, influencing interactions with different coactivators. Electrophoretic mobility shift assay and crystallographic examination (66) have shown that 4OH-TAM-bound ER could not bind a GRIP1 coactivator LXXLL peptide due to helix-12 interference at the binding cleft, which was recruited by E2. Consequently coactivator recruitment may influence receptor complex interactions with response element variants (34) which has been shown with other structurally diverse ligands and nuclear receptors (239,240). In addition, differences in absorption, distribution, metabolism and excretion (ADME) between ligands and species, likely contribute to divergent physiological and gene expression characteristics. It is well documented that TAM metabolism differs significantly between humans and rodents, for example, TAM N-oxide, 4OH-TAM and DMT are the predominant metabolites in the mouse, while DMT is the major human metabolite in microsomal studies (52,56,57). In rodents, the levels and rates of TAM metabolism to 4OH-TAM and 117 DMT were significantly different in the rat and mouse, where the rat metabolite profile more closely resembles human profiles (52). A cytochrome P450 2D6 polymorphism in humans further illustrates the potential effects of differences in metabolism on TAM activity. 4-OH-N— desmethyltamoxifen (endoxifen) is a recently identified TAM metabolite, found at higher levels than 4OH-TAM in patient serum, generated by CYP2D6 activity. It exhibits similar ER binding affinity, and comparable breast cancer cell proliferation and estrogen-induced p82 mRNA expression inhibition activities compared to 4OH-TAM (53). However, patients expressing specific CYPZD6 polymorphisms (i.e., CYP206*3, *4, *5 and *10) that impaired or abolished CYPZD6 metabolism have a nearly 2-fold higher risk of breast cancer recurrence (241). Collectively, these studies illustrate the significant differences in TAM metabolism between models that compromise the extrapolation of rodent data for use in human risk assessment. Conclusions Despite the comprehensive time course and dose response studies, an assessment of the gene expression effects and their roles in uterine responses could not be achieved due to limited genome coverage on our custom cDNA arrays and incomplete functional annotation for the represented genes. However, comparative TAM and EE studies using comparable designs and models identified conserved functionally annotated gene expression changes that are consistent with the measured uterotrophic response. Qualitatively, TAM 118 and EE gene expression profiles are similar; however, there are quantitative differences in efficacy, consistent with the partial agonist activity of TAM. Despite the evidence for these qualitative and quantitative differences in gene expression, demonstration that these changes have causal roles in the partial uterotrophic response elicited by TAM is required. The relevance of the differences between estrogen and TAM and the association with endometrial cancer (46,242,243) also needs further investigation. 119 CHAPTER 4 MIXTURE EFFECTS OF TAMOXIFEN AND ETHYNYLESTRADIOL ON GENE EXPRESSION IN IMMATURE, OVARIECTOMIZED MICE UTERUS. ABSTRACT Tamoxifen (TAM), the primary treatment for estrogen receptor (ER) positive breast cancer, has been associated with an increased incidence of endometrial cancer in post-, but not pre-menopausal women. TAM elicits a partial ER- mediated uterotrophic response in immature rodents when compared to ethynylestradiol (EE), a potent ER agonist. However, cotreatment with 1000 pg/kg TAM antagonizes the uterotrophic effect induced by 30 pg/kg EE. To further investigate the antiestrogenicity of TAM in the uterus, immature, ovariectomized C57BU6 mice were treated with a single oral dose of EE, TAM, EE+TAM or vehicle. Uteri were subsequently examined at 2, 4, 8, 12, 18, 24 hrs or after three daily treatments (3x24 hrs). Significant increases in uterine wet weight (UWW) were observed at 18 hrs for EE, TAM, and EE+TAM. However, EE+TAM induction of UW was significantly lower when compared to EE- induced uterotrophy at 3x24 hrs. This inhibitory effect is also reflected in decreases in luminal circumference, yet EE-induced luminal epithelial cell height was unaffected by cotreatment with TAM. Analysis using a 2x2 factorial cDNA microarray study design identified 290 genes differentially expressed following EE treatment. However, only a subset of EE-elicited changes in gene expression was affected by TAM cotreatment, consistent with the antiestrogenic response. 120 These data suggest that the mechanism of TAM antagonism of EE-induced UWW involves the selective inhibition EE-induced genes. 121 INTRODUCTION Tamoxifen (TAM) is an adjuvant and prophylactic therapy prescribed for estrogen receptor alpha (ERa) positive breast cancers. Due to the opposing effects of TAM in different tissues, it has been classified as a selective estrogen receptor modulator (SERM). In adjuvent therapy, it suppresses breast cancer recurrance by 50% and reduces the occurrence of contralateral primary breast cancer by 50%; when used as a prophylactic, TAM also reduces cancer occurrence in high risk populations (46). Despite its high therapeutic index, TAM also elicits undesirable effects in postmenopausal women including a two-fold risk increase in endometrial cancer (58). TAM and its active metabolites, 4- hydroxytamoxifen (4OH-TAM), N-desmethyltamoxifen (DMT), and 4-hydroxy-N- desmethyltamoxifen (endoxifene), elicit these effects by directly binding to the ligand binding domain (LBD) of ERa. Interestingly, TAM elicits a gene expression profile similar to estrogen in both MCF-7 cells and uterine tissue, albiet with lower efficacy (229,231,244). ER conformational changes in response to ligand binding affect its subsequent activities. Structrural resolution of the ligand binding domain occupied with 178-estradiol (E2) have elucidated a ligand-trapping conformation involving helix-12; whereas selective antagonists, such as raloxifene, position helix-12 in an orientation where the C-terminal domain of the ER interferes with ER transactivation (64,245). Protease sensitivity assays have also demonstrated that different ER surfaces are accessible to degradation depending on the bound ligand (246), while phage display assays that probe for different epitopes indicate 122 that ligands induce different ER topologies (235). In addition, FRET analysis demonstrated that different short-peptide fragments prefered to bind to different ligand-bound ER complexes (247). Ligand induced conformations have been implicated in the spectrum of coactivator proteins which may interact with the active receptor complex. Electrophoretic mobility shift assays have demonstrated the recruitment of GRIP1 co-activator to 178-estradiol (E2)-bound ER, but not by 4OH-TAM-bound receptor (66). Colorimetric phage ELISA assays have demonstrated that ER conformation may be influenced by both the ligand and the sequence of the gene-specific estrogen response element (ERE) (29). Moreover, coactivator recruitment may influence which activated receptors bind to specific promoter sequences. For example, DNA footprinting has shown that high mobility group B (HMGB) coactivator proteins enhance ER binding to EREs (34). It is evident that ligand-induced topology influences the gene-specific transcriptional activation of a number of steroid hormone receptors including the ER (reviewed in (33)). However, elucidating the influence of ligand structure on receptor conformation and transcriptional activitiy warrants further investigation. Although TAM and estrogen individually induce agonistic effects on the uterus in the immature, ovariectomized rodent, cotreatment at appropraite ratios elicit an antagonistic effect. For example, TAM significantly repressed uterine weight after 28 days in intact adult mice, but did not elicit reductions after daily treatments six months (248). TAM also antagonizes the E2 induction of uterotrophy (51), as well as other endpoints such as progesterone receptor (249) 123 and Fos expression (211), reporter gene assays (250) and peroxidase enzyme activity (51 ). Collectively, these studies demonstrate that TAM elicits a unique ER complex conformation affecting its tissue-specific agonist and antagonist activities. This report extends our previous studies examining the ER-mediated changes in gene expression elicited by EE and TAM alone that are associated with the induction of uterine wet weight (144,159,244), by examining their effects following cotreatment. A temporal two-by-two factorial microarray hybridization design, with complementary histopathology, was used to comprehensively examine differential gene expression associated with the antagonism of EE-induced uterine wet weight by TAM cotreatment (Figure 1A) (251). Interestingly, only a select subset of EE-induced genes was affected by TAM cotreatment. Antagonized responses were associated with specific genes within cell growth and proliferation pathways that could be correlated with the anti-uterotrophic effect. Results from this study further elucidate the mechanisms involved in the antagonist activities of TAM. 124 Figure 1 Microarray hybridization design and uterotrophic assay treatment design A) Differential gene expression between all treatment combinations was examined using the two by two factorial hybridization design (251) to minimize the number of arrays required per biological replicate. Each arrow represents an array and the Cy3 (head) and Cy5 (tail) dyes incorporated. B) Preliminary dose finding experiments examined uterine wet weight (UWW) 24 hrs after three orally administered daily doses. Mice (n = 5) were treated with vehicle, 30 pg/kg EE, 1000 pg/kg TAM or 30 pg/kg plus 1000 pg/kg TAM. Animals were sacrificed at 2, 4, 8, 12, 18 and 24 hrs after a single oral gavage or at 72 hrs after three daily doses. TAM was initially closed 8 hrs before EE to compensate for the delayed TAM-elicited responses associated with metabolism and distribution (244) to facilitate equal competition for ER availability. 125 Figure 1 A ll Microarray analysis was "as completed in triplicate for 2, 4, m 12, 24 and 72 hr time points for a total of 90 arrays. 53' [a B TREATMENT REGIMEN .r‘t‘ Vehicle Vehicle Vehicle EE kl": TAM TAM TAM EE+TAM "I‘ITT—l—l 1 is, -8 0 L2 4 8 12 18 24 48 72I SACRIFICE TIME (hrs) 126 MATERIALS AND METHODS Animal husbandry and treatment Female C57BU6 mice, ovariectomized by the vendor on postnatal day (PND) 20, were obtained from Charles River Laboratories (Raleigh, NC) on PND 25. Animals (n = 5) were housed in polycarbonate cages bedded with cellulose fiber chips (Aspen Chip Laboratory Bedding, Northeastern Products, Warrensberg, NY) in a 23°C environment with 30-40% humidity and a 12 h light/dark cycle (0700 — 1900 h). Animals had access to deionized water and Harlan Teklad 22/5 Rodent Diet 8640 (Madison, WI) ad libitum and acclimatized for 4 days prior to treatment. To account for the delayed gene expression responses (244), animals (n = 5 per group) were primed at -8 hrs with 1000 pg/kg TAM (TAM and mixture (MIX) groups) or sesame oil (vehicle and EE groups) (Figure 18). At 0 hrs, animals were dosed with 30 pg/kg EE (EE and MIX groups) or sesame oil (TAM and vehicle groups). Four groups (n = 5) of mice were also treated with sesame oil, 30 pg/kg EE (Sigma), 1000 pg/kg TAM (Sigma) or 30 pg/kg EE and 1000 pg/kg TAM at 24 and 48 hrs to represent the 3 x 24 hr treatment group. Doses were prepared based on average animal weight. Animals were sacrificed by cervical dislocation and body weights were recorded. The uterus was transected at the border of the cervix, and stripped of extraneous connective tissue and fat. Whole uterine weights were recorded before (wet weight) and after blotting (blotted weight) with absorbent tissue. A 6-8 mm section of unblotted uterine horn was placed in 10% neutral buffered formalin (NBF) for histology. The remainder was snap frozen in liquid nitrogen and stored 127 at -80°C for RNA extraction. Liver sections from the left lobe were snap frozen for LC/MS/MS analysis. All procedures were performed with the approval of the Michigan State University All-University Committee on Animal Use and Care. Histological processing, morphometric and pathological analysis Samples stored in 10% NBF were allowed to fix for at least 24 hrs at room temperature then placed into tissue cassettes and stored in 30% ethanol holding solution at 4°C. Paraffin embedding, sectioning (5 pm), mounting and hematoxylin and eosin staining were completed by the Michigan State University Laboratory for Anatomical Histology and Molecular Sciences (192) using standard techniques. Pathological assessments were evaluated according to standardized National Toxicology Program (NTP) pathology codes. Morphometric analysis was performed on midhorn uterine cross sections for all animals (n = 5 per treatment group) using Scion Image analysis software (Scioncorp, Frederick, MD). Histological markers of uterotrophy, including luminal epithelial cell height (LECH), luminal circumference and number of endometrial glands were quantified for each slide. Statistical analysis of morphometry data was assessed by Dunnett’s or two-way ANOVA followed with Tukey’s HSD post hoc analysis to examine dose dependent and temporal effects, respectively (SAS version 9.1). 128 RNA isolation Briefly, 1.0 mL of Trizol (lnvitrogen, Carlsbad, CA) was added to the frozen uterine tissue in a 2.0 mL microfuge tube and homogenized in the presence of steel beads by a Mixer Mill 300 homogenizer (Retsch, Germany). Total RNA was isolated and extracted according to the manufacturer’s protocol and resuspended in The RNA Storage Solution (Ambion, Austin, TX). RNA samples were quantified spectrophotometrically (A260) and assessed for quality by Azao/Azao ratio as well as inspected using denaturing agarose gel electrophoresis. Microarray hybridization and analysis Custom in-house cDNA arrays consisting of 13,361 features, representing 7,952 unique genes (Unigene Build 152), were spotted on epoxy coated glass slides (SCHOTT Nexterion, Germany) using an Omnigrid arrayer (GeneMachines, San Carlos, CA) and Telechem Chipmaker 3 pins in a TeleChem CHP3 printhead head (T elechem International Inc., Sunnyvale, CA) at the DNA Sequencing and Gene Expression Analysis facility at Michigan State University (128)). Selected clones were obtained from EPAMAC (129), Research Genetics, the National Institute of Aging and Lion Biosciences. Detailed protocols for processing of microarrays are available at the deach Home Page (130). A two by two factorial hybridization design was used to assess treatment effects (251) such that all treatment groups could be compared to each other 129 (Figure 1A). Four time-matched samples, of each treatment group, were hybridized to six slides to generate a single replicate of data. Three biological replicates were completed for 2, 4, 12, 24 and 3 x 24 hr time points for a total of 90 arrays. The Genisphere 900 3DNA Array Detection (Genisphere lnc., Hatfield, PA) indirect incorporation kit was used to generate cDNA samples for hybridization. Briefly, 1 pg of RNA was reverse transcribed in the presence of an oligo-tagged primer specifically targeted for Cy3- or Cy5- conjugated dendrimers. The cDNA was resuspended in 58 pL of 2X Formamide-Based Hybridization Buffer and hybridized overnight on arrays sealed in a light-shielded, humid chamber submerged in a 42°C water bath. Slides were then washed in SSC containing decreasing concentrations of SDS, spin-dried and re-hybridized with a Cy3:Cy5 (1:1) dendrimer mixture in formamide based buffer to indirectly incorporate dyes at the Cy3- and Cy5-dendrimer tagged cDNA hybridized on the first day. Slides were washed and dried as previously described, and scanned at 635 nm (Cy3) and 532 nm (Cy5) using a Molecular Devices Genepix 4100A scanner (Sunnyvale, CA). Images were examined, features identified and intensity values recorded using GenePix v.5.1 (Molecular Devices). Microarray quality control, statistical analysis and gene list filtering All arrays were compared to a historical data set of high quality arrays. Parameters assessed included background signal intensity, feature signal intensity, feature vs. background signal intensity ratios, the number of features 130 with background intensities greater than the feature intensity for each array, and relationships between feature and background signal intensities (134). Data were normalized using a semi-parametric approach (194). Model- based t—values were calculated comparing all time-matched treated and vehicle samples. Posterior probabilities of activity [P1(t)-value] were then calculated on a per-gene and per-time point basis using an Empirical Bayes analysis (135). Gene lists were filtered to identify genes which demonstrate differential expression between EE and mixture treatment. At each time point, both EE vs. V (EV) and MIX vs. V (MV) lists were identified based on posterior probability (P1(t) > 0.9999) and fold-change cut-off (|fold change| > 1.5) and then compared to identify differential expression between EV and EE vs MIX (EM) where P1(t) > 0.9999). All raw and analyzed data were stored in deach (httpzlldbzach.fst.msu.edu), a Minimum Information About Microarray Experiments (MIAME)-supportive relational database (133). QRT-PCR Aliquots of RNA isolated from each of the five biological replicates were set aside for SYBRTM Green quantitative real-time PCR (QRT-PCR) verification. EE-treated, temporal mouse uteri RNA were previously isolated (159). An oligo- dT anchored Superscript II (lnvitrogen) reverse transcriptase reaction was carried out on 1 pg of RNA, in a 20 pL reaction, from each biological sample as per manufacturer’s instructions. Samples were diluted four-fold and 3 pL used in a 30 pL real-time reaction mix containing 1X SYBR Green PCR buffer, 3 mM 131 MgClz, 0.33 mM dNTPs, 0.5 lU AmpliTaq Gold (Applied Biosystems, Foster City, CA) and 0.15 mM forward and reverse primer. All primers were designed by submitting cDNA microarray clone sequences into Primer3 (138) to obtain an amplicon of approximately 125bp (Table 1). PCR amplification was conducted in 96-well MicroAmp Optical plates (Applied Biosystems) on an Applied Biosystems PRISM 7000 Sequence Detection System under the following conditions: 10 min denaturation and enzyme activation at 95°C, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. After amplification, a 30 min dissociation protocol was conducted to assess primer specificity and product uniformity. Each plate contained duplicate standards of purified PCR product of known template concentration over eight orders of magnitude to generate a log template concentration standard curve. No template controls (NTC) samples were included on each plate such that experimental samples within 2 standard deviations of the NTCs are considered below the limits of detection. Plots were visualized and thresholds determined using ABI Prism 7000 SDS Software (Applied Biosystems). Results were normalized to Rpl7 mRNA levels to control for differences in RNA loading, quality and cDNA synthesis. Expression differences were assessed using a two-way ANOVA followed by Tukey's HSD post hoc analysis to examine treatment and treatment over time effects (SAS version 9.1). Correlation analyses of QRT-PCR and microarray data were generated using the correlation function of R v2.1.0. 132 LCIMSIMS Liver tissue was homogenized with ddHZO in a 1:20 dilution using a handheld Polytron homogenizer (Kinematica, Switzerland). One mL ddHZO, 200 pL 1N NaOH and 1 ng [‘5N, 1302] tamoxifen (Sigma), as an internal standard, was added to 1 mL of homogenate. The mixture was extracted in an ether:methanol (95:5 vlv) solution and evaporated at 55°C under a stream of N2. Residue was resuspended in 200 pL acetonitrilezammonium acetate (65:35 v/v) and stored at -20°C in amber sample vials until use. Appropriate standards were also prepared for quantitative interpolation of TAM and 4OH-TAM concentrations. Extracted samples were analyzed at the MSU Mass Spectrometry Facility (128). Samples were injected into the LC-20AD (Shimadzu; Columbia, MD) HPLC system with the SIL—5000 Injector (Shimadzu) and separated on an Atlantis dC18 3mm column (Waters Corporation; Milford, MA) using a 60:40 (vlv) methanolz100mm ammonium acetate (pH = 3) solution. Electrospray ionization mass spectrometry was carried out on a Quattro micro API instrument (Waters Corporation) and data analyzed using Mass Lynx v4.0 software (Waters Corporation). Bioinformatic Promoter Word Search Regulatory sequences of genes were obtained from the UCSC Genome Browser for mature RefSeq mRNA accessions and stored in the deach database (130). The sequences obtained extended from -5000 kb, upstream from the transcriptional start site, through the 5’ untranslated region. A sliding window method (252) was implemented to create a library of 5 to 10 nucleotide 133 words from these sequences. To identify over-represented 5 to 10 nucleotide motifs from the active gene lists determined through microarray analysis, an empirical Bayes implementation of the Wilcoxon’s Rank Sum Test was executed to calculate posterior probabilities (135,253). Queries of the Transfac database (254) were conducted to identify potential binding proteins associated with some over-represented short sequences. 134 RESULTS Dose finding studies Preliminary studies were conducted to identify optimal EE and TAM doses to investigate possible additive, synergistic and antagonistic uterotrophic tissue and gene expression responses. We have established that the oral £050 for the uterotrophic response elicited by EE and TAM are 22.1 and 33.7 pg/kg, respectively, and that 100 pg/kg EE induced a maximal uterine wet weight (UWW) response (~10-fold) in the immature, ovariectomized CS7BL/6 (159,244). TAM also exhibited a pronounced temporal delay in gene expression when compared to EE (244). In order to accommodate this delay, and to ensure equal competitive binding by TAM and EE for the ER, a modified treatment regimen was used that dosed the animals with TAM 8 hrs prior to EE (Figure 1B). Preliminary dose range finding studies were conducted at 72 hrs to identify the optimal EE:TAM ratio that would maximize the antagonism of EE- induced uterotrophy by TAM. Cotreatment with 0.1 and 1.0 mg/kg TAM significantly repressed UWW induction by treatments of 0.03 and 0.06 mg/kg EE (Figure 2). Consequently, 0.03 mg/kg EE and 1.0 mg/kg TAM (1:33 ratio) were selected to further investigate the additive, synergistic and antagonistic uterine responses following cotreatment (MIX). 135 Figure 2 Dose finding: uterotrophic inhibition EE (0, 10, 30 and 60 pg/kg) was co-treated with TAM (0.1, 1, 100 mglkg) to determine the optimal doses resulting in inhibition of EE-induced uterotrophy. 30 pg/kg EE and 1 mglkg TAM (1:33 ratio) were selected for further examination. The asterisk (*) indicates significant (p < 0.05) inhibition relative to EE alone. 136 2955 88 ES mm. 93: 8.0 mm 9.? 8.0 mm. 9.? 5o 6 959 kph; > For ...Fhrd > ...3 ._._.._.F.o > For he .Cuo > 29235ch N 0.52”. 137 Treatment Effects on Uterine Weight Significant (p < 0.05) increases in UVWV were observed at 18, 24 and 72 hrs after treatment with 0.03 mglkg EE, and at 8, 18, 24 and 72 hrs with 1.0 mglkg TAM (Figure 3). However TAM only elicited a 4.0-fold increase compared to the 8.1- fold increase induced by EE at 72 hrs. Although, cotreatment of 0.03 mglkg EE with 1.0 mglkg TAM still increased UWW from 12 - 72 hrs compared to vehicle, cotreatment-induced UWW was inhibited approximately 50% at 72 hrs compared to EE treatment alone. Morphometric analysis and histopathology Increases in luminal epithelial cell height (LECH) and luminal circumference are hallmarks of estrogenicity in the uterus (88). LECH was significantly induced 3.7-, 3.5- and 3.3-fold by EE, TAM and MIX treatment, respectively, compared to time-matched vehicle controls at 72 hrs (Figure 4A). There was no significant difference in LECH between EE and TAM at 72 hrs, and TAM cotreatment did not antagonize EE induced LECH. Luminal circumference was induced 3.1- and 2.9-fold at 24 hrs, and 8.0- and 4.9-fold at 72 hrs for EE and TAM, respectively (Figure 4B). Mixture treatment repressed luminal circumference by 54% compared to EE alone at 72 hrs, but was not significantly different from TAM alone. 138 Figure 3 Treatment induced uterotrophy Uterine wet weight (UWW) was measured at 2, 4, 8, 12, 18, 24 and 72 hrs after treatment (n = 5). “a” indicates a significant increase in UWW compared to the time-matched vehicle control. “b” indicates a significant difference in UWW compared to the time-matched EE-treated sample. TAM inhibited EE-induced UWW only after three daily treatments. WetVlbight 45- a 40. = Vehicle 35. - Estrogen ‘5 30. E: Tarmxifen Ev 25. - Mxtue d0 5 2°- , d) a 15' I 10- § . 5- 3 ° 0. k 139 Figure 4 Morphometric changes elicited by EE, TAM and Mixture Morphometric measurements of luminal epithelial cell height (LECH) and luminal circumference were made on all uteri sections. “a indicates a significant increase in LECH or luminal circumference compared to the time-matched vehicle control. “b” indicates a significant difference in luminal circumference compared to the time-matched EE-treated sample. TAM inhibited EE-induced luminal circumference only after three daily treatments. There were no significant differences in LECH between treatments at any time point examined. 140 Figure 4 aa ‘.‘-‘~“--~k n “~“‘k m fi“~“‘k 18 I I I 12 8 Trre (fls) Lurinal Circurrferenoe B .m \NNL‘NNNNN. n 33 V‘N“: m ‘§5 18 ‘g 12 Tim ("5) ‘;“L 8 s 5 7 I m a w. :55 3:335:85 141 Temporal Histological Changes Temporal- and dose-dependent histological changes in the uterus induced by EE and TAM in the immature, ovariectomized CS7BU6 mouse has been previously reported (159,244) The same assessment was used to characterize the histological changes elicited by vehicle, EE, TAM and EE+TAM treatment using the modified treatment regimen. (Table 1). Mild to moderate stromal edema was observed at 2 hrs in the TAM and MIX groups, likely due to early priming. All treatment groups exhibited mild to moderate hypertrophy in stromal nuclei by 4 hrs, with mild to moderate epithelial hyperplasia in the MIX treatment at 8 hrs. At 12 hrs, EE induced mild to moderate uterine stromal edema, mild stromal cell hypertrophy, and moderate endometrial hyperplasia, while TAM elicited qualitatively similar changes to the uterine architecture. MIX treatment induced comparable uterine morphology relative to EE and TAM treatments alone. After 24 hrs, EE and TAM alone elicited marked increases in uterine edema, stromal cell hypertrophy, and endometrial hyperplasia, and were not histologically distinguishable. Comparable changes were also present 24 hrs after MIX treatment. The severity of the uterotrophic response continued to 72 hrs after EE and TAM treatments alone. In contrast, MIX elicited changes in the uterus were attenuated compared to EE and TAM treatments alone, as evident in the areas of stromal hypertrophy and endometrial hyperplasia (Figure 5). Overall, EE, TAM and MIX treated uteri exhibit similar histological changes. Only at 72 hrs is there evidence of a diminished response elicited by MIX when compared to EE and TAM treatments alone. 142 Table 1. Histological evaluations of treated uterine sections (n = 5) Time Treatment Stromal Stromal Epithelial Myometrial (hrs) group edema nuclei hyperplasia hypertrophy hypertrophy V - - - - 2 E - - - - T mild - - - M mild - none - mild - - moderate V - - - - 4 E mild none - mild - - T moderate mild - - M mild - moderate - - moderate V _ - - - 8 E mild - mild - — moderate T marked mild mild - - moderate M marked — mild mild - - severe moderate V - - - - 12 E mild - moderate mild - - moderate moderate T moderate mild moderate - M moderate moderate moderate - V - - - - 18 E mild — mild - mild — - moderate moderate moderate T moderate - mild moderate - marked M moderate mild moderate - V _ - - - 24 E moderate moderate marked mild T moderate moderate marked mild M mild — mild — moderate — mild moderate moderate marked V - - - - 72 E moderate — marked severe mild marked T moderate marked severe mild M moderate moderate marked — mild severe 143 Figure 5 Histological observations Hematoxylin and eosin staining of uteri treated three times daily with (A) vehicle, (B) 30 jig/kg ethynylestradiol, (C) 1000 [19le tamoxifen and (D) 30 jig/kg ethynylestradiol plus 1000 ug/kg tamoxifen (MIX). All treatments elicited a uterotrophic effect, however MIX attenuated proliferative effect compared to EE- and TAM-indufid responses. Images are representative of five biological replicates; bar represents 30 pm. 144 v I. u. “halted ... m 959“. 145 LCIMSIMS analysis of liver TAM and 4OH-TAM levels TAM and 4OH-TAM levels were determined using LC/MS/MS in liver samples from the same animals due to the limited amount of uterine tissue available. Extracts from a previous study (244) demonstrated that TAM can be detected in hepatic tissues 2 hrs after treatment, with 4OH-TAM reaching a plateau by 4 hrs and decreasing after 12 hrs (Figure 6A). In the current cotreatment study with TAM-priming, comparable levels of TAM and 4OH-TAM were detected in hepatic liver extracts (Figure 68). Approximately 70 ng/mL TAM were detected at 2 hrs in TAM and MIX treated liver extracts. Peak levels of130 ng/mL were detected at 8 hrs that decreased to 50 ng/mL by 24 hrs. TAM levels were not significantly different between TAM and MIX hepatic extracts at any time point. However, 4OH-TAM levels were significantly higher in MIX (208 ng/mL) compared to TAM (92 ng/mL) at 2 hrs which converged to 100 ng/mL at 4 hrs. 4OH-TAM levels were not significantly different between TAM and MIX groups at any other time point. It was not possible to determine EE levels due to the low doses administered and the inefficiency of EE ionization and detection using LCIMSIMS. Uterine gene expression changes demonstrating mixture effects Differentially expressed genes were identified based on their empirical Bayes posterior probability of activity [P1(t)-value] on a per-gene, per-time point basis (Supplemental Table 1). P1(t)-values approaching 1.0 indicate a greater likelihood of treatment-related differential gene expression. EE-induced gene 146 Figure 6 Temporal LCIMSIMS analysis of hepatic TAM and 4OH-TAM levels Hepatic TAM and 4OH-TAM were extracted from a previous study (244) to determine tissue levels using LC/MS/MS. A) TAM was detected (*p < 0.05 compared to time—matched vehicle) at 2 hrs after treatment. B) 4OH-TAM levels peaked at 4 hrs (*p < 0.05 compared to time-matched vehicle) and plateaued at 12 hrs before steadily decreasing over time. TAM and 4OH-TAM were also extracted from liver samples from the current study to determine hepatic tissue levels using LCIMSIMS. TAM and 4OH-TAM were not significantly different between vehicle and EE treatments. C) TAM levels in TAM and MIX treated samples are significantly different from time-matched vehicle and EE controls (*p < 0.05), but not significantly different between TAM and MIX treatments at any time point. D) 4OH-TAM levels are significantly different from time-matched vehicle controls and EE treated animals (ap < 0.05). At 2 hrs, TAM and MIX demonstrate significantly different 4OH-TAM levels (”p < 0.05) but not beyond 4 hrs after treatment. 147 I! (a E 0 m Tm (Imus) 148 TPMTreatment Analysis MJdure Treatmentflnalysis A TAM C TAM €70“ * gm 8 -B-Vd'l'cb 360- * -I-V 3, 550. +TAM 5 1: * = 3 40' :8 100 a 30' «I s a 5 20' g g 10- § ol 0 °m248121824720 24812182472 Tireflws) MM) B 401+TAM D 40mm 3 3 g m +v g 300 g * * +TAM g g * g 2‘”- ‘c‘ g 100 3 E C O O o 24812182472” 24812182472 Tirea'lou's) expression affected by TAM co-treatment was identified using a two-step process. All genes were first filtered to identify 2518 EE-elicited gene expression changes (EE vs. V: P1(t) 2 0.9999; fold change 2 1.5) across all time points. These 2518 genes where then screened for modulation by TAM cotreatment (EE vs. MIX: P1(t) 2 0.9999) to identify only 290 unique, annotated genes exhibiting a MIX-treatment effect, representing potential non-additive interactions (Table 2). Gene expression changes were further examined by comparing EE vs. V and MIX vs. V to classify potential non-additive interactions as: A) EE-induced expression repressed by MIX, 8) EE-induced expression augmented by MIX, C) EE-repressed expression diminished by MIX , and D) EE-repressed expression further repressed by MIX (Figure 7, Table 3). The distribution of genes across time appears to shift from categories A, B and C (2 - 12 hrs) to primarily categories B and C (24 and 72 hrs). Note that a potential non-additive interaction may occur at several time points. For example, fos-like antigen 2 (Fos/2) is a category A gene at 2 and 4 hrs. A gene may also exhibit different non-additive patterns at different times, such as inhibin beta-B (Inhbb) which is a category A gene at 2 hrs but is classified as a category B at 24 and 72 hrs. Functional categorization of microarray data The majority of EE-elicited differentially expressed genes affected by TAM cotreatment identified at 2 and 4 hrs are associated with cell growth and proliferation including oncogenes such as myelocytomatosis oncogene (Myc), Jun oncogene (Jun) and FBJ osteosarcoma oncogene (Fos). Genes involved in 149 Table 2. MIX-modified, EE-induced gene list generation EE-induced genes * MIX-modified, * Time P1(t) 2 0.9999 EE-induced genes (hours) Fold-Change z 1.5 P1(t) 2 0.9999 2 49 25 4 336 87 12 1946 128 24 1534 79 72 591 48 Total Unique Genes ** 2518 290 * Number of unique, Entrez Gene—annotated genes at indicated time point ** Number of unique Entrez Gene-annotated genes across all time points NB: Genes may be active across multiple time points, thus the sum of each column is greater than total unique genes in each category. 150 Figure 7 EE-mediated gene expression affected by TAM cotreatment Only 209 out of 2518 EE-elicited gene expression changes (P1(t) 2 0.9999; fold change 2 1.5) were affected by TAM cotreatment. These genes were classified as: A) EE-induced expression repressed by MIX, B) EE-induced expression augmented by MIX, C) EE-repressed expression diminished by MIX, and D) EE- repressed expression that Is further repressed by MIX. The numbers within each panel indicates the number of genes exhibiting the pattern. Note that some genes exhibited different TAM cotreatment expression patterns at different time points. A 109 B 87 151 Table 3. MIX-modified, EE-induced gene classifications A B C D Time MlX- MIX- MIX— MIX- (hours) repression augmentation diminution augmentation of EE- of EE- of EE- of EE- induction induction repression repression 2 9 15 1 0 4 58 1 28 0 12 48 29 50 1 24 0 45 34 0 72 5 20 23 0 Total Unique 109 87 106 1 Genes* * A total of 290 MIX-modified, EE-mediated genes were identified. Some genes demonstrated different expression patterns at different time points, thus the sum of Total Unique Genes across all four categories is greater than 290. 152 the cell cycle, cyclin-dependent kinase inhibitor 1A (Cdkn1a) and branched chain aminotransferase 1, cytosolic (Bcat1), as well as guanine nucleotide binding protein-like 3 (Gn/3) and activating transcription factor 4 (NM) that are associated with proliferation, were also affected by TAM cotreatment. Other affected functional categories included lipid metabolism [peroxisomal trans-2- enoyl-CoA reductase (Pear) and carnitine palmitoyltransferase 2 (Cpt2)], immune response [interferon gamma inducible protein 30 ([1730) and chemokine (C-X-C motif) ligand 12 (Cxcl12)], and ion binding and transport [selenoprotein K (Selk) and solute carrier family 23 (nucleobase transporters), member 2 (Sl023a2)]. Eleven of these genes, representing different categories of MIX-mediated changes from Table 3, exhibited good correlations between microarray and QRT- PCR data (Figure 8). Bioinformatic promoter word search analysis At a single time point, categories A through D adequately describes the relationship between EE and MIX treatment; however, temporal patterns elicited by MIX treatment may also provide insight to its regulation. MIX-modified, EE- mediated genes were categorized according to their MIX vs. V fold-change temporal patterns (Figure 9) and bioinforrnatic promoter word searches were conducted on each temporal pattern group to identify putative sequence elements over-represented in MIX-mediated, EE-induced genes. Despite seven distinct temporal patterns, only three returned positive TRANSFAC® hits 153 Figure 8 Quantitative real-time PCR verification Microarray results for 11 genes were verified using QRT—PCR. These genes represent various affected pathways and different temporal patterns of expression. Overall, there was good agreement between microarray (left) and QRT-PCR (right) data. Examples for four of the genes, A) Fos and Inhbb, B) Ccl21b and Ndufb9, are illustrated demonstrating different patterns of MIX- modified, EE-mediated changes. Statistically significant QRT-PCR differences (p < 0.05, n = 4) due to treatment are denoted by an asterisk (*). 154 Figure 8A W—PCRData Tln‘e (fl's) l:lV Mcroarray Data The (“5) T””IIIIA e a: at. ‘11 efiuqu mod .0 Fos Inhb 155 Figure 88 QRT-PCRDaIa McroanayData afiueuo pics .0 F S o O 156 eobcué—E WI]IIIIIIIA ,, . E v ”III/”nilllllz i Time (113) 'II’””””III’A W””””’III4 '”I”’l’l”l" Ll efiueuo p105 Ndufb9 N h 3’. V N (Supplemental Table 2). These positive hits were associated with known binding factors (posterior probability 2 0.95) as reported by experiments reported in the TRANSFAC® database. The most common binding factors associated with the over-represented sequences include C/EBP, Sp1 and hepatocyte nuclear factors (HNFs). Further studies are required to elucidate the role of these factors in MIX- modified, EE-mediated gene transcription. 157 Figure 9 MIX-modified, EE-mediated gene profiles Bioinformatic word searches were completed on seven categories of MIX vs. V fold-change, temporal profiles demonstrated by 209 MIX-modified, EE-mediated genes. Three categories returned positive TRANSFAC hits, associated with a binding transcription factor. 90 genes 19 genes 168 hits 31 genes 1 hit 23 genes 18 genes [E] 14 genes Fold Change Time 158 DISCUSSION The rodent uterotrophic assay is a well established model to study the physiological and morphological effects elicited by estrogenic compounds (91). It has been used to examine the differential uterine gene expression elicited by EE (144,159,198,255) and more recently, the partial agonist effects of TAM (244). Previous studies have demonstrated that TAM inhibits estrogen-induced increases in UWW (84,85), however the effects of cotreatment on gene expression have not been comprehensively examined. In this study, we have used our previously reported EE and TAM differential gene expression data (144,159,244). to further investigate the inhibition of EE—induced uterotrophy by TAM co-treatment using the same model, study design and analysis methods. Moreover, we are also able to re-examine many widely held hypotheses regarding the mechanisms involved in the anti-estrogenicity of TAM. This study demonstrates that TAM inhibited EE induction of UW by approximately 50% in immature, ovariectomized C57BL6 mice, comparable to the levels of suppression previously reported (84,200,256). Histologically, TAM co-treatment inhibited EE-induction of luminal circumference but did not antagonize EE-induced LECH, suggesting that the antagonism of proliferation is cell type-specific. Differential gene expression data also indicates that the antagonism is not global since the majority of EE elicited responses were not affected by TAM cotreatment. Of the 2518 EE-elicited differential gene responses, only 290 were affected by TAM cotreatment, with 214 exhibiting 159 repression and 76 exhibiting enhanced responses, relative to EE treatment alone. Consequently, only a small subset of EE-elicited differential gene expression is affected by TAM, thus indicating that competition for ER binding TAM (51), and down-regulation of ER gene expression (257), does not sufficiently explain the more complex interactions resulting in the inhibition of EE- induced UWW increases (258). Examination of the functions of EE-elicited. gene expression affected by TAM is consistent with the inhibition of EE-induced UWW. For example, several genes associated with growth and proliferation were repressed by TAM at early time points (Figure 9a), including Myc, Jun, and Fos. The proliferation- regulating, uterine-expressed transcription factors, Fos/2 (259), Ets1 and Ets2 (260), as well as estrogen-responsive proliferation-associated thbp4 (261,262), uterotrophy-associated Gnl3 (263) and stromal cell differentiation regulator 80083 (80033; 4hrs) (264) were also repressed. Group A proliferation related genes including StxZ (265), estrogen responsive CIu, mouse uterus-expressed Popch (266) and Gja1 found in human myometrium (267) were also all found to be repressed at later time points. The EE elicited repression of some genes was also minimized (Fig. 9c). Growth arrest specific 1 (Gas1) expression, which is repressed by Myc (268), is consistent with the inhibition of EE-induced Myc, thus consistent with the repression of uterotrophy by Gas1. In addition, the endometrial expression of Cirbp, which exhibits an inverse relationship with proliferation (269), was de-repressed by TAM, also consistent with the antagonism of UW increases by EE. 160 Furthermore, TAM enhanced the induction and repression of some EE- elicited gene expression changes (Groups B and D). Although these responses appear counter intuitive, several of these changes are consistent with the repression of the EE-induced uterotrophic effect. For example, over-expression of Atf4 impairs mammary proliferation and development (270) and Cdkn1a is known to promote growth arrest and apoptotic pathways (271). These responses provide further support for a transcriptional role in MIX-repression of EE-induced uterine weight. However, there were also late differential gene expression responses by EE that were enhanced by TAM cotreatment that are consistent with proliferation (Group B). Crip1, which is up-regulated in proliferating mammary luminal epithelial cells (272), Cdc2l1 (273) and endometrium expressed Tgfa (274), all exhibited enhanced differential expression at later time points. This enhanced expression may be an attempt to over compensate for the limited induction of UW in response to the majority of gene expression changes that were otherwise unaffected under TAM co-treatment. Cytoskeletal reorganization is integral to estrogen-mediated restructuring of proliferating tissue (198). Several genes associated with the cytoskeleton including Bicd2 (275), 00th (276) and Mfap5 (277) were induced by EE and repressed following TAM cotreatment (Group A), consistent with the inhibition of uterotrophy. Although, these genes have not been identified to be ER-regulated, their differential expression serves to prepare the tissue for proliferation. Binding studies indicate that 33:1 TAM to EE ratio is insufficient to displace greater than 50% of estrogen bound to ER (278). Furthermore, 161 estrogen is ~200,000-fold more potent than TAM in eliciting a DNA synthesis response in mouse uterus (279). In addition, some genes such as Fos and Ndufb9 (Figure 10), exhibited intermediate behavior where cotreatment induced a response that was greater than TAM alone but less than EE alone. Collectively, these results suggest that the inhibitory effects of TAM are not simply a result of TAM saturation of the ER. SERM activity is based on the ability to differentially affect various tissue types (31,280). This study is the first to demonstrate that TAM also elicits selective in vivo gene expression responses within the uterus. Estrogen and 4OH-TAM cotreatment studies in MCF-7 cells have identified genes that exhibit comparable patterns of antagonism. For example, Group A genes Fos/2, Asns (225) and Fos (229), Group C genes Il1r1, Tm4sf1, and M30 (229) exhibited similar gene expression behavior in MCF-7 cells and C57BU6 uterine tissue. Differences in study design, microarray platforms, gene representation on the arrays, and data analysis are significant factors that limit the number of genes affected by TAM co-treatment in both models. For example, there are significant differences in ER protein levels (31), tissue specific co-regulating factor availability (237,246) as well as gene-specific thresholds of activation (238) that likely confound comparisons between human breast cancer MCF-7 cells and mouse uterine gene expression profiles. 162 Conclusions: This study represents the first comprehensive in vivo investigation of the anti—estrogenic effects of TAM on uterine gene expression. Repression of EE- induced uterotrophy, by TAM co-treatment, did not globally repress all EE- mediated gene expression. In contrast, only a select subset was affected which include genes associated with cellular growth and proliferation, consistent with an anti-uterotrophic effect. However, comparative studies in the rat or more sophisticated transgenic approaches are required to conclusively demonstrate the importance of these potential targets in uterine proliferation and growth and as critical TAM targets for the inhibition of EE-induced increases in UVWV. 163 CHAPTER 5 SUMMARY AND FUTURE PERSPECTIVES The preceding studies have examined mouse in vitro liver and in vivo uterus as models to examine gene expression changes elicited by estrogenic compounds. Although serum deprived Hepa-1c1c7 cells only demonstrated changes of small magnitude to a potent estrogen, the responses under serum- supplemented conditions did correlate with the diverse responses found in vivo. Identification of a more appropriate in vitro model would facilitate high-throughput screening of less potent estrogenic compounds with respect to risk assessment and pharmaceutical analysis. Furthermore, characterization of suitable rodent and human in vitro models would allow comparative analyses not feasible in vivo. The uterus of the immature, ovariectomized mouse is an excellent model to characterize changes elicited by TAM and examine mixture effects between TAM and EE, due to its well characterized physiological, histological and gene expression responses to estrogens. TAM is a known partial agonist, thus it was not surprising to identify numerous differentially expressed genes associated with cell growth and proliferation. These studies also identified genes which may contribute to the limited uterotrophic effect compared to EE, and through genes uniquely activated by TAM. Foundational microarray studies of EE, as a positive control, on rodent uterine and hepatic systems have established a baseline, which have been extended to included responses elicited by TAM. These results further support the identification and development of estrogen receptor—specific biomarkers 164 suitable for high-throughput screening. Collecting differential gene expression data elicited by structurally diverse ER ligands may also be used to investigate structure and function relationships important in target-specific pharmaceuticals. The benefits of intra-lab microarray studies are also demonstrated through these studies. Early establishment of a comprehensive study design minimizes the need to repeat foundational studies, and facilitate comparisons to other compounds of interest. Also, utilizing the same model, experimental procedures, microarray platform, and data analysis methods reduces the variables that may confound comparisons and data interpretation in future studies examining other ligands, tissues or model systems. The EE and TAM mixture study demonstrated that only a subset of genes exhibit differential expression when compared to independent treatment. Moreover, the transcriptional changes, elicited by the EE and TAM mixture, correlate with the observed physiological changes. These results present new questions regarding the regulation of responsive genes that exhibit differential regulation following treatment of EE alone, TAM alone and EE and TAM cotreatment. These mixture studies also further elucidated the characteristics of SERMs. SERMs were initially defined as compounds eliciting differing responses between organs. For example, TAM exhibits anti-estrogenic activities in mammary tissue and partial agonist activities in the uterus. These results indicate that TAM elicits differential gene expression regulation within the uterus, many of which are similar to those elicited by full agonist, EE. Thus, the activities 165 SERMs, and other compounds, can be more accurately classified using microarray technology. These types of data may be beneficial to the development and characterization of new, target-specific drugs. Executing a mixture study revealed novel experimental considerations. This includes the use of additional concentrations of the compounds as well as alternate ratio combinations in order to more comprehensively assess the effects of a mixture. In addition, there may be different responses if other endpoints, such as rate of DNA replication, were selected, which could be important for risk assessment and drug development. Such considerations are applicable to all mixture studies as standard experimental designs have yet to be defined. Consequently, this thesis demonstrates the importance of a solid foundation of experimental procedures, based on a comprehensive design, to optimize future comparative efforts. The analysis of TAM data is enriched due to the availability of well-established EE data as a baseline for comparison. Furthermore, the modifications to the experimental design, to facilitate the comparison between two compounds, have resulted in sound, reproducible data. These results demonstrate the utility of the approach that may be used as a basis for future two-compound mixture studies. 166 10. 11. 12. 13. 14. 15. REFERENCES US. EPA. (1998) Federal Register 63, 71542-71568 Fry, D. M., and Toone, C. K. (1981) Science 213, 922-924 Guillette, L. J., Jr., Gross, T. S., Masson, G. R., Matter, J. M., Percival, H. F., and Woodward, A. R. (1994) Environ Health Perspect102, 680-688 Burkhart, J. G., Helgen, J. C., Fort, D. J., Gallagher, K., Bowers, D., Propst, T. L., Gernes, M., Magner, J., Shelby, M. D., and Lucier, G. (1998) Environ Health Perspect106, 841-848 Hagel, P., and Tuinstra, L. G. (1978) Bull Environ Contam Toxicol19, 671- 676 Watanabe, l., Yakushiji, T., Kuwabara, K., Yoshida, S., Maeda, K., Kashimoto, T., Koyama, K., and Kunita, N. (1979) Arch Environ Contam Toxicol 8, 67-75 Schwartz, P. M., Jacobson, S. W., Fein, G., Jacobson, J. L., and Price, H. A. (1983) Am J Public Health 73, 293-296 Cordle, F., Locke, R., and Springer, J. (1982) Environ Health Perspect 45, 171-182 Cooke, P. S., Selvaraj, V., and Yellayi, S. (2006) J Nutr136, 704-708 Cheek, A. O., Vonier, P. M., Oberdorster, E., Burow, B. C., and McLachlan, J. A. (1998) Environ Health Perspect106 Suppl 1, 5-10 Rice, 8., and Whitehead, S. A. (2006) Endocr Relat Cancer 13, 995-1015 Ito, Y., Fisher, C. R., Conte, F. A., Grumbach, M. M., and Simpson, E. R. (1993) Proc Natl Acad Sci U S A 90, 11673-11677 Conte, F. A., Grumbach, M. M., Ito, Y., Fisher, C. R., and Simpson, E. R. (1994) J Clin Endocrinol Metab 78, 1287-1292 Jones, M. E., Boon, W. C., Mclnnes, K., Maffei, L., Carani, C., and Simpson, E. R. (2007) Nat Clin PractEndocn’nol Metab 3, 414-421 Punyadeera, C., Verbost, P., and Groothuis, P. (2003) J Steroid Biochem Mol Biol 84, 393-410 167 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. Kuiper, G. G., Carlsson, B., Grandien, K., Enmark, E., Haggblad, J., Nilsson, S., and Gustafsson, J. A. (1997) Endocrinology 138, 863-870 Harris, H. A. (2007) Mol Endocrinol 21, 1-13 Curtis Hewitt, 8., Couse, J. F., and Korach, K. S. (2000) Breast Cancer Res 2, 345-352 Moras, D., and Gronemeyer, H. (1998) Curr Opin Cell Biol 10, 384-391 Nichols, M., Rientjes, J. M., and Stewart, A. F. (1998) Embo J 17, 765-773 Saunders, P. T. (1998) Rev Reprod 3, 164-171 Sabbah, M., Radanyi, C., Redeuilh, G., and Baulieu, E. E. (1996) Biochem J 314 ( Pt 1), 205-213 Hall, J. M., Couse, J. F., and Korach, K. S. (2001) J Biol Chem 276, 36869-36872 Klein-Hitpass, L., Schorpp. M., Wagner, U., and Ryffel, G. U. (1986) Cell 46, 1053-1061 Klinge, C. M. (2001) Nucleic Acids Res 29, 2905-2919 Bourdeau, V., Deschenes, J., Metivier, R., Nagai, Y., Nguyen, D., Bretschneider, N., Gannon, F., White, J. H., and Mader, S. (2004) Mol Endocrinol18, 1411-1427 Carroll, J. S., Meyer, C. A., Song, J., Li, W., Geistlinger, T. R., Eeckhoute, J., Brodsky, A. S., Keeton, E. K., Fertuck, K. C., Hall, G. P, Wang, Q., Bekiranov, S., Sementchenko, V., Fox, E. A., Silver, P. A., Gingeras, T. R., Liu, X. S., and Brown, M. (2006) Nat Genet 38, 1289-1297 Klinge, C. M., Jernigan, S. C., Smith, S. L., Tyulmenkov, V. V., and Kulakosky, P. C. (2001) Mol Cell Endocrinol 174, 151-166 Hall, J. M., McDonnell, D. P., and Korach, K. S. (2002) MolEndocrinol16, 469-486 Zhou, G., Cummings, R., Li, Y., Mitra, S., Wilkinson, H. A., Elbrecht, A., Hermes, J. D., Schaeffer, J. M., Smith, R. G., and Moller, D. E. (1998) Mol Endocn'nol12, 1594-1604 Shang, Y., and Brown, M. (2002) Science 295, 2465-2468 Heery, D. M., Kalkhoven, E., Hoare, S., and Parker, M. G. (1997) Nature 387, 733-736 168 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. Edwards, D. P. (2000) J Mammary Gland Biol Neoplasia 5, 307-324 Das, D., Peterson, R. C., and Scovell, W. M. (2004) MolEndocn'noI18, 2616-2632 Kato, S. (2001) Breast Cancer 8, 3—9 Kushner, P. J., Agard, D. A., Greene, G. L., Scanlan, T. S., Shiau, A. K., Uht, R. M., and Webb, P. (2000) J Steroid Biochem Mol Biol 74, 311-317 Dong, J., Tsai-Morris, C. H., and Dufau, M. L. (2006) J Biol Chem 281, 18825-18836 Zaitsu, M., Narita, S., Lambert, K. C., Grady, J. J., Estes, D. M., Curran, E. M., Brooks, E. G., Watson, C. S., Goldblum, R. M., and Midoro-Horiuti, T. (2007) Mol lmmunol 44, 1977-1985 Witorsch, R. J. (2002) Regul Toxicol Phannacol 36, 118-130 U.S.EPA. (2005) Federal Register 70, 56449-56465 Van den Belt, K., Berckmans, P., Vangenechten, C., Verheyen, R., and Witters, H. (2004) Aquat Toxicol 66, 183-195 Herbst, A. L., Ulfelder, H., and Poskanzer, D. C. (1971) N Engl J Med 284, 878—881 Klopper, A. (1976) Br Med J 2, 414-416 Jordan, V. C., and Brodie, A. M. (2007) Steroids 72, 7-25 Omoto, Y., and Hayashi, S. (2002) Breast Cancer 9, 308-311 Lewis, J. S., and Jordan, V. C. (2005) Mutat Res 591, 247-263 Jordan, V. 0., Collins, M. M., Rowsby, L., and Prestwich, G. (1977) J Endocrinol 75, 305-316 Brodie, A., Sabnis, G., and Macedo, L. (2007) J Steroid Biochem Mol Biol Zheng, A., Kallio, A., and Harkonen, P. (2007) Endocrinology 148, 2764- 2777 Fritah, A., Redeuilh, G., and Sabbah, M. (2006) J Endocrinol 191, 613-624 Robertson, D. W., Katzenellenbogen, J. A., Long, D. J., Rorke, E. A., and Katzenellenbogen, B. S. (1982) J Steroid Biochem 16, 1-13 169 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. Robinson, S. P., Langan-Fahey, S. M., Johnson, D. A., and Jordan, V. C. (1991) Drug Metab Dispos 19, 36-43 Johnson, M. D., Zuo, H., Lee, K. H., Trebley, J. P., Rae, J. M., Weatherman, R. V., Desta, Z., Flockhart, D. A., and Skaar, T. C. (2004) Breast Cancer Res Treat 85, 151-159 Garber, K. (2005) J Natl Cancer Inst 97, 412-413 Stearns, V., Johnson, M. D., Rae, J. M., Morocho, A., Novielli, A., Bhargava, P., Hayes, D. F., Desta, Z., and Flockhart, D. A. (2003) J Natl Cancer Inst 95, 1758-1764 Lim, C. K., Yuan, Z. X., Lamb, J. H., White, I. N., De Matteis, F., and Smith, L. L. (1994) Carcinogenesis 15, 589-593 White, I. N. (2003) Curr Drug Metab 4, 223-239 Houssami, N., Cuzick, J., and Dixon, J. M. (2006) Med J Aust 184, 230- 234 Bergman, L., Beelen, M. L., Gallee, M. P., Hollema, H., Benraadt, J., and van Leeuwen, F. E. (2000) Lancet 356, 881-887 Smith, C. L., Nawaz, Z., and O'Malley, B. W. (1997) Mo/Endocn'nol11, 657-666 Badia, E., Duchesne, M. J., Semlali, A., Fuentes, M., Giamarchi, C., Richard-Foy, H., Nicolas, J. C., and Pons, M. (2000) Cancer Res 60, 4130-4138 Dowers, T. S., Qin, 2. H., Thatcher, G. R., and Bolton, J. L. (2006) Chem Res Toxicol 19, 1125-1 137 Celik, L., Lund, J. D., and Schiott, B. (2007) Biochemistry 46, 1743-1758 Pike, A. C., Brzozowski, A. M., Hubbard, R. E., Bonn, T., Thorsell, A. G., Engstrom, 0., Ljunggren, J., Gustafsson, J. A., and Carlquist, M. (1999) Embo J 18, 4608-4618 Wu, Y. L., Yang, X., Ren, Z., McDonnell, D. P., Norris, J. D., Willson, T. M., and Greene, G. L. (2005) Mol Cell 18, 413-424 Shiau, A. K., Barstad, D., Loria, P. M., Cheng, L., Kushner, P. J., Agard, D. A., and Greene, G. L. (1998) Cell 95, 927-937 Ciana, P., Di Luccio, G., Belcredito, S., Pollio, G., Vegeto, E., Tatangelo, L., Tiveron, C., and Maggi, A. (2001) MolEndocrino/15, 1104-1113 170 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. Ciana, P., Raviscioni, M., Mussi, P., Vegeto, E., Que, l., Parker, M. G., Lowik, C., and Maggi, A. (2003) Nat Med 9, 82-86 Williams, C. M. (2004) Proc Nutr Soc 63, 153-160 Blum, A., and Cannon, R. 0., 3rd. (1998) Curr Opin Lipidol 9, 575-586 Alvaro, D., Mancino, M. G., Onori, P., Franchitto, A., Alpini, 6., Francis, H., Glaser, S., and Gaudio, E. (2006) World J Gastroenterol12, 3537- 3545 Lewis, S. (2007) Am Heart J 153, 182-188 Boverhof, D. R., Fertuck, K. C., Burgoon, L. D., Eckel, J. E., Gennings, C., and Zacharewski, T. R. (2004) Carcinogenesis 25, 1277-1291 Hoffman, E. C., Reyes, H., Chu, F. F., Sander, F., Conley, L. H., Brooks, B. A., and Hankinson, O. (1991) Science 252, 954-958 Reyes, H., Reisz-Porszasz, S., and Hankinson, O. (1992) Science 256, 1 193-1 195 Bernhard, H. P., Darlington, G. J., and Ruddle, F. H. (1973) Dev Biol 35, 83-96 Jeong, H. G., and Lee, S. S. (1998) Cancer Lett 133, 177-184 Zacharewski, T., Harris, M., and Safe, S. (1991) Biochem Phannacol 41, 1931-1939 Darlington, G. J., Bernhard, H. P., Miller, R. A., and Ruddle, F. H. (1980) J Natl Cancer Inst 64, 809-819 Hankinson, O. (1979) Proc Natl Acad Sci U S A 76, 373-376 Fong, C. J., Burgoon, L. D., and Zacharewski, T. R. (2005) Toxicol Sci 86, 342-353 Werle-Schneider, G., Wolfelschneider, A., von Brevern, M. C., Scheel, J., Storck, T., Muller, D., Glockner, R., Bartsch, H., and Bartelmann, M. (2006) Int J Toxicol 25, 379-395 Jordan, V. C. (2006) Br J Pharmacol 147 Suppl 1, 8269-276 Jordan, V. C., Dix, C. J., Naylor, K. E., Prestwich, G., and Rowsby, L. (1978) J Toxicol Environ Health 4, 363-390 Jordan, V. C., and Gosden, B. (1983) Endocrinology 113, 463-468 171 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. Dassen, H., Punyadeera, C., Kamps, R., Klomp, J., Dunselman, G., Dijcks, F ., de Goeij, A., Ederveen, A., and Groothuis, P. (2007) Cell Mol Life Sci 64, 1009-1032 Reel, J. R., Lamb, l. J., and Neal, B. H. (1996) Fundam Appl Toxicol 34, 288-305 Owens, J. W., and Ashby, J. (2002) Crit Rev Toxicol 32, 445-520 Greenman, D. L., Dooley, K., and Breeden, C. R. (1977) J Toxicol Environ Health 3, 589-597 (2007) Diet, P., Schmidt, 8., and Vollmer, G. (2002) J ChromatogrB Analyt Techno! Biomed Life Sci 777, 191-202 Lubahn, D. B., Moyer, J. S., Golding, T. S., Couse, J. F., Korach, K. S., and Smithies, O. (1993) Proc Natl Acad Sci U S A 90, 11162-11166 Kuiper, G. G., and Gustafsson, J. A. (1997) FEBS Lett 410, 87-90 Couse, J. F ., Curtis Hewitt, 8., and Korach, K. S. (2000) J Steroid Biochem Mol Biol 74, 287-296 Krege, J. H., Hodgin, J. B., Couse, J. F., Enmark, E., Warner, M., Mahler, J. F., Sar, M., Korach, K. S., Gustafsson, J. A., and Smithies, O. (1998) Proc Natl Acad Sci U S A 95, 15677-15682 Jakacka, M., Ito, M., Weiss, J., Chien, P. Y., Gehm, B. D., and Jameson, J. L. (2001) J Biol Chem 276, 13615-13621 Jakacka, M., Ito, M., Martinson, F., lshikawa, T., Lee, E. J., and Jameson, J. L. (2002) MolEndocn'nol16, 2188-2201 O'Brien, J. E, Peterson, T. J., Tong, M. H., Lee, E. J., Pfaff, L. E., Hewitt, S. C., Korach, K. S., Weiss, J., and Jameson, J. L. (2006) J Biol Chem 281, 26683-26692 U.S.EPA. (1989) Vol. 1, United States Environmental Protection Agency, Washington, DC Berenbaum, M. C. (1981) Adv Cancer Res 35, 269-335 Deneer, J. W. (2000) Pest Management Science 56, 516-520 Broderius, S. J., Kahl, M. D., Elonen, G. E., Hammen'neister, D. E., and Hoglund, M. D. (2005) Environ Toxicol Chem 24, 3117-3127 172 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. Schuler, L. J., Trimble, A. J., Belden, J. B., and Lydy, M. J. (2005) Arch Environ Contam Toxicol 49, 173-177 Nebert, D. W., Dalton, T. P., Okey, A. B., and Gonzalez, F. J. (2004) J Biol Chem 279, 23847-23850 Billiard, S. M., Timme-Laragy, A. R., Wassenberg, D. M., Cockman, C., and Di Giulio, R. T. (2006) Toxicol Sci 92, 526-536 Barron, M. G., Carts, M. G., Heintz, R., and Rice, S. D. (2004) Toxicol Sci 78, 60-67 Harper, N., Connor, K., Steinberg, M., and Safe, S. (1995) Fundam Appl Toxicol 27, 131-139 U.S.EPA. (2000), United States Environmental Protection Agency, Washington, DC Stork, L. G., Gennings, C., Carchman, R. A., Carter, W. H., Jr., Pounds, J., and Mumtaz, M. (2006) Risk Anal 26, 1601-1612 Mowat, F. S., and Bundy, K. J. (2002) Chemosphere 49, 499-513 Gennings, 0., Carter, W. H., Jr., Carney, E. W., Charles, G. D., Gollapudi, B. B., and Carchman, R. A. (2004) Toxicol Sci 80, 134-150 Moser, V. 0., Simmons, J. E., and Gennings, C. (2006) Toxicol Sci 92, 235-245 Charles, G. D., Gennings, C., Tornesi, B., Kan, H. L., Zacharewski, T. R., Bhaskar Gollapudi, B., and Carney, E. W. (2007) ToxicolAppI Phannacol 218, 280-288 Jonker, M. J., Svendsen, C., Bedaux, J. J., Bongers, M., and Kammenga, J. E. (2005) Environ Toxicol Chem 24, 2701-2713 Andersen, M. E., and Dennison, J. E. (2004) Environ Toxicol Pharm 16, 1- 11 Rider, C. V., and LeBlanc, G. A. (2005) Toxicol Sci 87, 520-528 Horrocks, C., Halse, R., Suzuki, R., and Shepherd, P. R. (2003) Curr Opin Drug Discov Devel 6, 570-575 Baker, V. A. (2001) Toxicol In Vitro15, 413-419 Diel, P., Walter, A., Fritzemeier, K. H., HegeIe-Hartung, C., and Knauthe, R. (1995) J Steroid Biochem Mol Biol 55, 363-373 173 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. Lemmen, J. G., Arends, R. J., van Boxtel, A. L., van der Saag, P. T., and van der Burg, B. (2004) J Mol Endocrinol 32, 689-701 Zajchowski, D. A., Kauser, K., Zhu, Q, Webster, L., Aberle, S., White, F. A., 3rd, Liu, H. L., Humm, R., MacRobbie, J., Ponte, P., HegeIe-Hartung, C., Knauthe, R., Fritzemeier, K. H., Vergona, R., and Rubanyi, G. M. (2000) J Biol Chem 275, 15885-15894 Cato, A. C., Nestl, A., and Mink, S. (2002) Sci STKE 2002, RE9 Hammond, G. L. (2002) Obstet Gynecol Clin North Am 29, 411-423 Paech, K., Webb, P., Kuiper, G. G., Nilsson, S., Gustafsson, J., Kushner, P. J., and Scanlan, T. S. (1997) Science 277, 1508-1510 Revelli, A., Massobrio, M., and Tesarik, J. (1998) Endocr Rev 19, 3-17 Watson, C. S., Pappas, T. C., and Gametchu, B. (1995) Environ Health Perspect103 Suppl 7, 41-50 Wong, C. W., McNally, C., Nickbarg, E., Komm, B. S., and Cheskis, B. J. (2002) Proc Natl Acad Sci U S A 99, 14783-14788 Research Technology Support Facility [http:l/www.genomics.msu.edu] Rockett, J. C., and Dix, D. J. (1999) Environ Health Perspect107, 681-685 deACH Home Page: Zacharewski Laboratory [http:l/dbzach.fst.msu.edu] Brazma, A., Hingamp, P., Quackenbush, J., Sherlock, G., Spellman, P., Stoeckert, C., Aach, J., Ansorge, W., Ball, C. A., Causton, H. C., Gaasterland, T., Glenisson, P., Holstege, F. C., Kim, I. F., Markowitz, V., Matese, J. C., Parkinson, H., Robinson, A., Sarkans, U., Schulze-Kremer, S., Stewart, J., Taylor, R., Vilo, J., and Vingron, M. (2001) Nat Genet 29, 365-371 Mattes, W. B., Pettit, S. D., Sansone, S. A., Bushel, P. R., and Waters, M. D. (2004) Environ Health Perspect 112, 495-505 Burgoon, L. D., Boutros, P. C., Dere, E., and Zacharewski, T. R. (2005) Toxicol Sci Burgoon, L. D., Eckel-Passow, J. E., Gennings, C., Boverhof, D. R., Burt, J. W., Fong, C. J., and Zacharewski, T. (2005) Nucleic Acids Research 33, e172 Eckel, J. E., Gennings, C., Chinchilli, V. M., Burgoon, L. D., and Zacharewski, T. R. (2004) J Biopharrn Stat 14, 647-670 174 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. 151. 152. Gene Home [http:l/www.ncbi.nlm.nih.gov/sites/entrez?db=gene] the Gene Ontology [http:lewgeneontology.org] Primer3 Input (primer3/input.htm version 0.3.0 modified for Wl) [http:l/frodowi.mit.edu/cgi-bin/primer3/primer3_www.cgi] Wilkinson, R. F. (1993) in Art To Science: In Tissue Culture (HyClone Laboratories, l., ed) Vol. 12, pp. 1 - 10, HyClone Laboratories, Inc., Logan lgnar-Trowbridge, D. M., Pimentel, M., Parker, M. G., McLachlan, J. A., and Korach, K. S. (1996) Endocrinology 137, 1735-1744 Katzenellenbogen, B. S., and Norman, M. J. (1990) Endocrinology 126, 891-898 Cappelletti, V., Fioravanti, L., Miodini, P., and Di Fronzo, G. (1996) Tumour Biol 17, 290-298 Hyder, S. M., Chiappetta, C., and Stancel, G. M. (1999) J Phannacol Exp Ther 290, 740-747 Fertuck, K. C., Eckel, J. E., Gennings, C., and Zacharewski, T. R. (2003) Physiol Genomics 15, 127-141 Curtis, S. W., Washburn, T., Sewall, C., DiAugustine, R., Lindzey, J., Couse, J. F., and Korach, K. S. (1996) Proc Natl Acad Sci U S A 93, 12626-12630 Nelson, K. G., Takahashi, T., Bossert, N. L., Walmer, D. K., and McLachlan, J. A. (1991) Proc Natl Acad Sci U S A 88, 21-25 Nunez, A. M., Berry, M., Imler, J. L., and Chambon, P. (1989) Embo J 8, 823-829 Shi, H., and Teng, C. (1996) MoIEndocrinol10, 732-741 Tortora, G., Budillon, A., Yokozaki, H., Clair, T., Pepe, S., Merlo, G., Rohlff, C., and Cho-Chung, Y. S. (1994) Cell Growth Differ 5, 753-759 Woodman, O. L., Missen, M. A., and Boujaoude, M. (2004) J Cardiovasc Phamiacol 44, 155—163 Wolf, G., Hildenbrand, R., Schwar, C., Grobholz, R., Kaufmann, M., Stutte, H. J., Strebhardt, K., and Bleyl, U. (2000) Pathol Res Pract 196, 753-759 Rider, V., Foster, R. T., Evans, M., Suenaga, R., and Abdou, N. l. (1998) Clin lmmunol Immunopathol 89, 171-1 80 175 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. Salvatori, L., Ravenna, L., Felli, M. P., Cardillo, M. R., Russo, M. A., Frati, L., Gulino, A., and Petrangeli, E. (2000) Endocrinology 141, 2266-2274 Currie, M. J., Hanrahan, V., Gunningham, S. P., Morrin, H. R., Frampton, 0., Han, C., Robinson, B. A., and Fox, S. B. (2004) J Clin Pathol 57, 829- 834 Gray, K., Eitzman, B., Raszmann, K., Steed, T., Geboff, A., McLachlan, J., and Bidwell, M. (1995) Endocrinology 136, 2325-2340 Albrecht, E. D., Robb, V. A., and Pepe, G. J. (2004) J Clin Endocrinol Metab 89, 5803-5809 Tarn, S. P., Archer, T. K., and Deeley, R. G. (1985) J Biol Chem 260, 1670-1675 Dabrowska, M., Palka, J., Wolczynski, S., Pietruczuk, M., and Osada, J. (2001) Gynecol Obstet Invest 51, 248-253 Kwekel, J. C., Burgoon, L. D., Burt, J. W., Harkema, J. R., and Zacharewski, T. R. (2005) Physiol Genomics 23, 327-342 Hsu, C. Y., and Frankel, F. R. (1987) J Biol Chem 262, 9594-9600 Whitehead, |. P., Abe, K., Gorski, J. L., and Der, C. J. (1998) Mol Cell Biol 18, 4689-4697 Szelei, J., Soto, A. M., Geck, P., Desronvil, M., Prechtl, N. V., Weill, B. C., and Sonnenschein, C. (2000) J Steroid Biochem Mol Biol 72, 89-102 Kipp, J. L., and Ramirez, V. D. (2003) Neuroendocrinology 77, 258-272 Choi, l., Gudas, L. J., and Katzenellenbogen, B. S. (2000) Mol Cell Endocrinol164, 225-237 Vijayalakshmi, V., and Gupta, P. D. (1994) Exp Cell Res 214, 358-366 Piccone, C. M., Brazeau, G. A., and McCormick, K. M. (2005) Exp Physiol 90, 87-93 Naciff, J. M., Jump, M. L., Torontali, S. M., Carr, G. J., Tiesman, J. P., Overrnann, G. J., and Daston, G. P. (2002) Toxicol Sci 68, 184-199 Srivastava, R. A., Srivastava, N., Averna, M., Lin, R. C., Korach, K. 8., Lubahn, D. B., and Schonfeld, G. (1997) J Biol Chem 272, 33360-33366 Archer, T. K., Tam, S. P., Deugau, K. V., and Deeley, R. G. (1985) J Biol Chem 260, 1676-1681 176 170. 171. 172. 173. 174. 175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. Ness, G. C., and Chambers, C. M. (2000) Proc Soc Exp Biol Med 224, 8- 19 Watanabe, H., Suzuki, A., Mizutani, T., Khono, S., Lubahn, D. B., Handa, H., and lguchi, T. (2002) Genes Cells 7, 497-507 Lopez, 0., Sanchez, M. D., Shea-Eaton, W., and McLean, M. P. (2002) Endocrinology 1 43, 21 55-2168 Wu, X., Pang, S. T., Sahlin, L., Blanck, A., Norstedt, G., and Flores- Morales, A. (2003) Biol Reprod 69, 1308-1317 Singh, R. P., Kumar, R., and Kapur, N. (2003) J Environ Pathol Toxicol Oncol 22, 75-92 Toda, K., Takeda, K., Akira, S., Saibara, T., Okada, T., Onishi, S., and Shizuta, Y. (2001) J Steroid Biochem Mol Biol 79, 11-17 Nemoto, Y., Toda, K., Ono, M., Fujikawa-Adachi, K., Saibara, T., Onishi, S., Enzan, H., Okada, T., and Shizuta, Y. (2000) J Clin Invest 105, 1819- 1825 Hatta, H., Atomi, Y., Shinohara, S., Yamamoto, Y., and Yamada, S. (1988) Horm Metab Res 20, 609-611 Grist, M., Wambolt, R. B., Bondy, G. P., English, D. R., and Allard, M. F. (2002) Can J Physiol Phannacol 80, 1001-1007 Dannan, G. A., Porubek, D. J., Nelson, S. D., Waxman, D. J., and Guengerich, F. P. (1986) Endocrinology 118, 1952-1960 Liehr, J. G., and Roy, D. (1990) Free Radic Biol Med 8, 415-423 Lee, S. S., Jeong, H. G., and Yang, K. H. (1998) Biochem Mol Biol Int 45, 775-781 Sakuma, T., Takai, M., Endo, Y., Kuroiwa, M., Ohara, A., Jarukamjorn, K., Honma, R., and Nemoto, N. (2000) Arch Biochem Biophys 377, 153-162 Broccardo, C. J., Billings, R. E., Chubb, L. 8., Andersen, M. E., and Hanneman, W. H. (2004) Toxicol Sci 78, 287-294 Mandlekar, S., Hebbar, V., Christov, K., and Kong, A. N. (2000) Cancer Res 60, 6601-6606 Jordan, V. C. (1993) Br J Pharmacol 110, 507-517 177 186. 187. 188. 189. 190. 191. 192. 193. 194. 195. 196. 197. 198. 199. 200. Wijayaratne, A. L., Nagel, S. C., Paige, L. A., Christensen, D. J., Norris, J. D., Fowlkes, D. M., and McDonnell, D. P. (1999) Endocrinology 140, 5828-5840 Fisher, D. E. (1994) Cell 78, 539-542 Carthew, R, Edwards, R. E., Nolan, B. M., Tucker, M. J., and Smith, L. L. (1999) Toxicol Sci 48, 197-205 Kohlerova, E., and Skarda, J. (2004) J Vet Med A Physio! Pathol Clin Med 51. 209-217 Osborne, C. K., Zhao, H., and Fuqua, S. A. (2000) J Clin Oncol18, 3172- 3186 Vogel, V. G., Costantino, J. P., Wickerham, D. L., Cronin, W. M., Cecchini, R. S., Atkins, J. N., Bevers, T. B., Fehrenbacher, L., Pajon, E. R., Jr., Wade, J. L., 3rd, Robidoux, A., Margolese, R. G., James, J., Lippman, S. M., Runowicz, C. D., Ganz, P. A., Reis, S. E., McCaskill-Stevens, W., Ford, L. G., Jordan, V. C., and Wolmark, N. (2006) Jama 295, 2727-2741 Human Pathology @ Michigan State University [http:l/www.humanpathologymsu.edu/histology] Burgoon, L. D., Eckel-Passow, J. E., Gennings, C., Boverhof, D. R., Burt, J. W., Fong, C. J., and Zacharewski, T. R. (2005) Nucleic Acids Res 33, e172 Eckel, J. E., Gennings, C., Therneau, T. M., Burgoon, L. D., Boverhof, D. R., and Zacharewski, T. R. (2005) Bioinfonnatics 21, 1078-1083 Carthew, R, Edwards, R. E., and Nolan, B. M. (1999) ToxicoIAppI Phannacol 158, 24-32 Jordan, V. C., and Dix, C. J. (1979) J Steroid Biochem 11, 285-291 Loeb, L., Suntzeff, V., and Burns, E. L. (1938) Am J Cancer 34, 413-427 Moggs, J. G., Tinwell, H., Spurway, T., Chang, H. S., Pate, l., Lim, F. L., Moore, D. J., Soames, A., Stuckey, R., Currie, R., Zhu, T., Kimber, |., Ashby, J., and Orphanides, G. (2004) Environ Health Perspect112, 1589- 1606 Yuen, T., Wurmbach, E., Pfeffer, R. L., Ebersole, B. J., and Sealfon, S. C. (2002) Nucleic Acids Res 30, e48 Black, L. J., and Goode, R. L. (1980) Life Sci 26, 1453-1458 178 201. 202. 203. 204. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. Fromson, J. M., Pearson, S., and Bramah, S. (1973) Xenobiotica 3, 693- 709 Fromson, J. M., Pearson, S., and Bramah, S. (1973) Xenobiotica 3, 711- 714 Kiriwat, O., and Fotherby, K. (1983) Contraception 27, 153-160 Gray, C. A., Bartol, F. F., Tarleton, B. J., Wiley, A. A., Johnson, G. A., Bazer, F. W., and Spencer, T. E. (2001) Biol Reprod 65, 1311-1323 Sanyal, A., Riggs, B. L., Spelsberg, T. C., and Khosla, S. (2005) J Cell Biochem 94, 88-97 Boverhof, D. R., Burgoon, L. D., Tashiro, C., Chittim, B., Harkema, J. R., Jump, D. B., and Zacharewski, T. R. (2005) Toxicol Sci 85, 1048-1063 Naciff, J. M., Overrnann, G. J., Torontali, S. M., Carr, G. J., Tiesman, J. P., Richardson, B. D., and Daston, G. P. (2003) Toxicol Sci 72, 314-330 Yamashita, S., Takayanagi, A., and Shimizu, N. (1996) Endocrinology 137, 5468-5475 Fan, M., Yan, P. S., Hartman-Frey, C., Chen, L., Paik, H., Oyer, S. L., Salisbury, J. 0., Cheng, A. S., Li, L., Abbosh, P. H., Huang, T. H., and Nephew, K. P. (2006) Cancer Res 66, 11954-11966 Lai, M. D., Jiang, M. J., and Wing, L. Y. (2002) Endocrine 17, 233-239 Kirkland, J. L., Murthy, L., and Stancel, G. M. (1993) Mol Phannacol43, 709-714 Mylonas, l., Jeschke, U., Wiest, l., Hoeing, A., Vogl, J., Shabani, N., Kuhn, C., Schulze, S., Kupka, M. S., and Friese, K. (2004) Histochem Cell Biol 122, 461-471 Nakayama, H., Sano, T., Motegi, A., Oyama, T., and Nakajima, T. (2005) Pathol Int 55, 707-715 Krebs, J., Saremaslani, P., and Caduff, R. (2002) Biochim Biophys Acta 1600, 68-73 Fernandez-Zapico, M. E., Gonzalez-Paz, N. C., Weiss, E., Savoy, D. N., Molina, J. R., Fonseca, R., Smyrk, T. C., Chari, S. T., Urrutia, R., and Billadeau, D. D. (2005) Cancer Cell 7, 39-49 Mehus, J. G., Deloukas, P., and Lambeth, D. O. (1999) Hum Genet 104, 454-459 179 217. 218. 219. 220. 221. 222. 223. 224. 225. 226. 227. 228. 229. 230. Rampazzo, C., Kost-Alimova, M., Ruzzenente, B., Dumanski, J. P., and Bianchi, V. (2002) Gene 294, 109-117 Jablonski, E. M., McConnell, N. A., Hughes, F. M., Jr., and Huet-Hudson, Y. M. (2003) Biol Reprod 69, 1481-1487 Koyama, N., lshibashi, K., Kuwahara, M., lnase, N., lchioka, M., Sasaki, S., and Marumo, F. (1998) Genomics 54, 169-172 Suzuki, T., Ishihara, K., Migaki, H., Matsuura, W., Kohda, A., Okumura, K., Nagao, M., Yamaguchi-lwai, Y., and Kambe, T. (2005) J Biol Chem 280, 637-643 McMahon, R. J., and Cousins, R. J. (1998) J Nutr128, 667-670 Lee, J. Y., Kim, J. H., Hong, S. H., Cherny, R. A., Bush, A. l., Palmiter, R. D., and Koh, J. Y. (2004) J Biol Chem 279, 8602-8607 Burckhardt, B. C., and Burckhardt, G. (2003) Rev Physio! Biochem Phannaco! 146, 95-158 Ljubojevic, M., Balen, D., Breljak, D., Kusan, M., Anzai, N., Bahn, A., Burckhardt, G., and Sabolic, l. (2006) Am J Physiol Renal Physiol lnoue, A., Yoshida, N., Omoto, Y., Oguchi, S., Yamori, T., Kiyama, R., and Hayashi, S. (2002) J Mo! Endocrinol 29, 175-192 del Carmen Garcia Molina Wolgien, M., da Silva, l. D., Villanova, F. E., Yumi Otsuka, A., Borra, R. C., Lima Reis, L. F., Carvalho, A. F., Baracat, E. C., and Gebrim, L. H. (2005) EurJ Gynaeco! Oncol 26, 501-504 Scafoglio, C., Ambrosino, C., Cicatiello, L., Altucci, L., Ardovino, M., Bontempo, P., Medici, N., Molinari, A. M., Nebbioso, A., Facchiano, A., Calogero, R. A., Elkon, R., Menini, N., Ponzone, R., Biglia, N., Sismondi, P., Bortoli, M. D., and Weisz, A. (2006) J Cell Biochem Gadal, F., Starzec, A., Bozic, C., Pillot-Brochet, C., Malinge, S., Ozanne, V., Vicenzi, J., Buffat, L., Perret, G., Iris, F., and Crepin, M. (2005) J Mol Endocrinol 34, 61 -75 Frasor, J., Stossi, F., Danes, J. M., Komm, B., Lyttle, C. R., and Katzenellenbogen, B. S. (2004) Cancer Res 64, 1522-1533 Frasor, J., Chang, E. C., Komm, B., Lin, C. Y., Vega, V. 3., Liu, E. T., Miller, L. D., Smeds, J., Bergh, J., and Katzenellenbogen, B. S. (2006) Cancer Res 66, 7334-7340 180 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243. 244. Hodges, L. C., Cook, J. D., Lobenhofer, E. K., Li, L., Bennett, L., Bushel, P. R., Aldaz, C. M., Afshari, C. A., and Walker, C. L. (2003) Mol Cancer Res 1, 300-311 Green, A. R., Styles, J. A., Parrott, E. L., Gray, 0., Edwards, R. E., Smith, A. G., Gant, T. W., Greaves, P., Al-Azzawi, F., and White, I. N. (2005) Exp Toxicol Pathol 56, 255-263 Robertson, D. W., Katzenellenbogen, J. A., Hayes, J. R., and Katzenellenbogen, B. S. (1982) J Med Chem 25, 167-171 Dowsett, M., and Haynes, B. P. (2003) J Steroid Biochem Mol Biol 86, 255-263 Giambiagi, N., and Pasqualini, J. R. (1988) J Steroid Biochem 30, 213-217 McDonnell, D. P., Chang, C. Y., and Norris, J. D. (2000) J Steroid Biochem Mol Biol 74, 327-335 Webb, P., Nguyen, P., and Kushner, P. J. (2003) J Biol Chem 278, 6912- 6920 Geum, 0., Sun, W., Paik, S. K., Lee, C. C., and Kim, K. (1997) Mo! Reprod Dev 46, 450-458 Kazmin, D., Prytkova, T., Cook, C. E., Wolfinger, R., Chu, T. M., Beratan, D., Norris, J. 0., Chang, C. Y., and McDonnell, D. P. (2006) Mol Endocrinol 20, 1201-1217 Meijer, O. C., Kalkhoven, E., van der Laan, S., Steenbergen, P. J., Houtman, S. H., Dijkmans, T. F., Pearce, D., and de Kloet, E. R. (2005) Endocrinology 146, 1438-1448 Goetz, M. P., Knox, S. K., Suman, V. J., Rae, J. M., Safgren, S. L., Ames, M. M., Visscher, D. W., Reynolds, 0., Couch, F. J., Lingle, W. L., Weinshilboum, R. M., Fritcher, E. G., Nibbe, A. M., Desta, Z., Nguyen, A., Flockhart, D. A., Perez, E. A., and lngle, J. N. (2007) Breast Cancer Res Treat101, 113-121 Deligdisch, L., Kalir, T., Cohen, C. J., de Latour, M., Le Bouedec, G., and Penault-Llorca, F. (2000) Gynecol Oncol 78, 181-186 Mourits, M. J., De Vries, E. G., Willemse, P. H., Ten Hoor, K. A., Hollema, H., and Van der Zee, A. G. (2001) Obstet Gynecol 97, 855-866 Fong, C. J., Burgoon, L. D., Williams, K. J., Forgacs, A. L., and Zacharewski, T. R. (2007) BMC Genomics 8, 151 181 245. 246. 247. 248. 249. 250. 251. 252. 253. 254. 255. 256. 257. 258. 259. Brzozowski, A. M., Pike, A. C., Dauter, 2., Hubbard, R. E., Bonn, T., Engstrom, 0., Ohman, L., Greene, G. L., Gustafsson, J. A., and Carlquist, M. (1997) Nature 389, 753-758 McDonnell, D. P., Clemm, D. L., Hermann, T., Goldman, M. E., and Pike, J. W. (1995) Mol Endocrinol 9, 659-669 Weatherman, R. V., Chang, C. Y., Clegg, N. J., Carroll, D. 0., Day, R. N., Baxter, J. D., McDonnell, D. P., Scanlan, T. S., and Schaufele, F. (2002) Mo! Endocrinol 16, 487-496 Sourla, A., Luo, S., Labrie, C., Belanger, A., and Labrie, F. (1997) Endocrinology 138, 5605-5617 Castellano-Diaz, E., Gonzalez-Quijano, M. l., Liminana, J. M., and Diaz- Chico, B. N. (1989) J Steroid Biochem 33, 133-139 Mclnerney, E. M., and Katzenellenbogen, B. S. (1996) J Biol Chem 271, 24172-24178 Yang, Y. H., and Speed, T. (2002) Nat Rev Genet 3, 579-588 Tajima, F. (1991) J Mol Evol 33, 470-473 Efron, B., and Tibshirani, R. (2002) Genet Epidemiol 23, 70-86 Gene Regulation [http:l/www.gene- regulation.com/pub/databases.html#transfac] Naciff, J. M., Overrnann, G. J., Torontali, S. M., Carr, G. J., Khambatta, Z. S., Tiesman, J. P., Richardson, B. D., and Daston, G. P. (2007) Toxicol Sci 97, 467-490 Jordan, V. C., Rowsby, L., Dix, C. J., and Prestwich, G. (1978) J Endocrinol 78, 71-81 Koibuchi, Y., lino, Y., Uchida, T., Andoh, T., Horii, Y., Nagasawa, M., Horiguchi, J., Maemura, M., Takei, H., Yokoe, T., and Morishita, Y. (2000) Oncol Rep 7, 135-140 Mnif, W., Pascussi, J. M., Pillon, A., Escande, A., Bartegi, A., Nicolas, J. C., Cavailles, V., Duchesne, M. J., and Balaguer, P. (2007) Toxicol Lett 170, 19-29 Bamberger, A. M., Milde-Langosch, K., Rossing, E., Goemann, C., and Loning, T. (2001) J Cancer Res Clin Oncol127, 545-550 182 260. 261. 262. 263. 264. 265. 266. 267. 268. 269. 270. 271. 272. 273. 274. Kilpatrick, L. M., Kola, l., and Salamonsen, L. A. (1999) Biol Reprod 61, 120-126 Laping, N. J., Olson, B. A., and Zhu, Y. (2001) J Am Soc Nephrol12, 883- 890 Wang, D. Y., McKague, B., Liss, S. N., and Edwards, E. A. (2004) Environ Sci Technol 38, 6396-6406 Beekman, C., Nichane, M., De Clercq, S., Maetens, M., Floss, T., Wurst, W., Bellefroid, E., and Marine, J. C. (2006) Mo! Cell Biol 26, 9291-9301 Dimitriadis, E., Stoikos, 0., Tan, Y. L., and Salamonsen, L. A. (2006) Endocrinology 147, 3809-3817 Wang, Y., Wang, L., lordanov, H., Swietlicki, E. A., Zheng, Q., Jiang, S., Tang, Y., Levin, M. S., and Rubin, D. C. (2006) J Clin Invest 116, 1535- 1546 Andree, B., Hillemann, T., Kessler-lcekson, G., Schmitt-John, T., Jockusch, H., Arnold, H. H., and Brand, T. (2000) Dev Biol 223, 371-382 Cluff, A. H., Bystrom, B., Klimaviciute, A., Dahlqvist, C., Cebers, G., Malmstrom, A., and Ekman-Ordeberg, G. (2006) Reprod Biol Endocrinol 4, 24 Lee, T. C., Li, L., Philipson, L., and Ziff, E. B. (1997) Proc Nat/Acad Sci U S A 94, 12886-12891 Hamid, A. A., Mandai, M., Fujita, J., Nanbu, K., Kariya, M., Kusakari, T., Fukuhara, K., and Fujii, S. (2003) IntJ Gynecol Pathol 22, 240-247 Bagheri-Yarrnand, R., Vadlamudi, R. K., and Kumar, R. (2003) J Biol Chem 278, 17421-17429 Qiao, L., lVchinstry, R., Gupta, S., Gilfor, D.,’Windle, J. J., Hylemon, P. 8., Grant, 8., Fisher, P. B., and Dent, P. (2002) Hepatology 36, 39-48 Mackay, A., Jones, 0., Dexter, T., Silva, R. L., Bulmer, K., Jones, A., Simpson, P., Harris, R. A., Jat, P. S., Neville, A. M., Reis, L. F., Lakhani, S. R., and O'Hare, M. J. (2003) Oncogene 22, 2680-2688 Wilker, E. W., van Vugt, M. A., Artim, S. A., Huang, P. H., Petersen, C. P., Reinhardt, H. C., Feng, Y., Sharp, P. A., Sonenberg, N., White, F. M., and Yaffe, M. B. (2007) Nature 446, 329-332 Ejskjaer, K., Sorensen, B. S., Poulsen, S. S., Mogensen, O., Forman, A., and Nexo, E. (2005) Mol Hum Reprod11, 543-551 183 275. Hoogenraad, C. C., Wulf, P., Schiefermeier, N., Stepanova, T., Galjart, N., Small, J. V., Grosveld, F ., de Zeeuw, C. l., and Akhmanova, A. (2003) Embo J 22, 6004-6015 276. Uetake, Y., Terada, Y., Matuliene, J., and Kuriyama, R. (2004) Cell Motil Cytoskeleton 58, 53-66 277. Lemaire, R., Bayle, J., Mecham, R. P., and Lafyatis, R. (2007) J Biol Chem 282, 800-808 278. Levin, E., Tomchinsky, S., and Lopez, S. (1990) J Steroid Biochem Mol Bio! 37, 681 -686 279. Zhang, H., McElrath, T., Tong, W., and Pollard, J. W. (2005) J Endocrinol 184, 129-140 280. Jordan, V. C. (2007) Nat Rev Cancer 7, 46-53 184 I ll " ll I! ll l l l LIBRAR l ll l [IT/l" l l 129flwfl:9 iii lllll‘ll” ll 4329 56 3