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 [6, 19, 20, 25]. 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 [26–28]. In contrast, peak serum levels of EE are detected within two hours of treatment .
At equi-efficacious doses of TAM and EE (i.e. 100 vs. 20 μg/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 , 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 invaginated regions of the lumen that generate a large secretory surface area . 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 μg/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 , consistent with the delayed histological effects. Of these early TAM responses, only Esr1 and Car3 have been reported to be induced by estrogen [16, 31]. 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 . 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 [15, 16, 22, 32–34]. 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 that possessed an ERE, were also activated by EE. However, differences in 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 ). 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 [36–38] resulting in a full uterotrophic agonist response. In contrast, 22 genes more highly induced by TAM included G2/M inhibitor (Sfn/14-3-3σ), which has been associated with human endometrial carcinomas  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 Bcl-2 member, Bok, and the induction of Pdcd6, an apoptosis regulator, associated with proliferating tissues  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 . For TAM, QRT-PCR confirmed decreased expression of Sipa1 (data not shown), a repressed response at 24 hrs associated with decreased proliferation  that may reduce hyperplasia.
DNA synthesis and replication pathways were also differentially regulated. Sustained up-regulation of dNDP phosphorylating genes, Nme1 and Nme6 , suggest salvage pathways are emphasized for nucleotide synthesis rather than de novo processes where 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 [15, 16]. 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]  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 . It does not appear to be a factor in TAM-induced uterine weight increases as blotted weights were not significantly different from wet weights. Aqp1 and Aqp5 are comparably regulated by TAM and EE, while Aqp8 induction was specific to EE (QRT-PCR verified, data not shown). Aqp8 is a known contributor to water imbibition  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 [46, 47] may facilitate stromal edema and has been shown to be responsive to estrogen where it is down-regulated in brain tissue . Organic anion transporter, Slc22a7, was repressed by EE from 18 - 72 hrs in the uteri suggesting anion retention in the stroma that may also be important for edema. Slc22a7 is an importer in the basolateral membrane of kidney tubule epithelia (reviewed in ), and is estrogen responsive in the kidney .
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 III, 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 in vitro human breast tissue and in vivo mouse uterus [51, 52]. Only the induction of Uqcrb , Nqo1 , Tff1, Mapt , Pctk3, Wnt4 , Myb, Cdc6, Cdc20, Mcm2, Fos and Mybl2  and repression of Xrcc1, Tgfa , Rap1ga1, Blnk, Tm4sf1, Matn2, Ifi30, Tgfb3 and Smpd1  correlated with the changes observed in the current study. Moreover, there are examples of divergent gene expression changes such as inverse responses for Pfn2 , Ctsh, Selenbp1, Nfrkb, Cyp1a1 , Prps1 and Tmsb4x . 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 . Col1a1 exhibited persistent up-regulation months after treatment and was also induced in our short term study. Several factors, in addition to model 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 [55, 57]. Although 4OH-TAM and EE induced similar cell cycle genes, other down-stream mechanisms were also regulated to prevent 4OH-TAM mediated cell cycle progression . Some of these mechanisms may play a roll in the partial uterotrophic response elicited by TAM in treated mice.
Differences in chemical structure may also 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 , binding affinities [59, 60], ligand-induced binding domain topographies , coactivator recruitment capabilities [62, 63], gene-specific thresholds of activation, and efficacies . 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  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  which has been shown with other structurally diverse ligands and nuclear receptors [67, 68].
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 [28, 69, 70]. 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 .
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 pS2 mRNA expression inhibition activities compared to 4OH-TAM . However, patients expressing specific CYP2D6 polymorphisms (i.e., CYP2D6*3, *4, *5 and *10) that impaired or abolished CYP2D6 metabolism have a nearly 2-fold higher risk of breast cancer recurrence . 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.