Estrogen Activation by Steroid Sulfatase Increases Colorectal Cancer Proliferation via GPER

Context: Estrogens affect the incidence and progression of colorectal cancer (CRC), although the precise molecular mechanisms remain ill-defined. Objective: The present study investigated prereceptor estrogen metabolism through steroid sulphatase (STS) and 17β-hydroxysteroid dehydrogenase activity and subsequent nongenomic estrogen signaling in human CRC tissue, in The Cancer Genome Atlas colon adenocarcinoma data set, and in in vitro and in vivo CRC models. We aimed to define and therapeutically target pathways through which estrogens alter CRC proliferation and progression. Design, Setting, Patients, and Interventions: Human CRC samples with normal tissue-matched controls were collected from postmenopausal female and age-matched male patients. Estrogen metabolism enzymes and nongenomic downstream signaling pathways were determined. CRC cell lines were transfected with STS and cultured for in vitro and in vivo analysis. Estrogen metabolism was determined using an ultra-performance liquid chromatography–tandem mass spectrometry method. Primary Outcome Measure: The proliferative effects of estrogen metabolism were evaluated using 5-bromo-2′-deoxyuridine assays and CRC mouse xenograft studies. Results: Human CRC exhibits dysregulated estrogen metabolism, favoring estradiol synthesis. The activity of STS, the fundamental enzyme that activates conjugated estrogens, is significantly (P < 0.001) elevated in human CRC compared with matched controls. STS overexpression accelerates CRC proliferation in in vitro and in vivo models, with STS inhibition an effective treatment. We defined a G-protein–coupled estrogen receptor (GPER) proproliferative pathway potentially through increased expression of connective tissue growth factor in CRC. Conclusion: Human CRC favors estradiol synthesis to augment proliferation via GPER stimulation. Further research is required regarding whether estrogen replacement therapy should be used with caution in patients at high risk of developing CRC.


Introduction
Controversy surrounds the role estrogens play in CRC (1). Observational studies from the Women's Health Initiative (WHI) suggest pre-menopausal women have a 20% reduction in CRC compared to age-matched men (2). These gender differences plateau as women become post-menopausal. However, women on exogenous hormone-replacement therapy (HRT = conjugated estrogen (estrone sulfate (E 1 S)) plus medroxyprogesterone), maintain protection against CRC (3), and elevated endogenous plasma estrogen concentrations also protect against CRC incidence (4). Conversely, other studies suggest greater endogenous plasma estrone (E 1 ) concentrations in post-menopausal women increase CRC risk (5); similarly, women with estrogen-dependent breast cancer have a higher risk of developing CRC (6).
Importantly, women taking HRT at the time of CRC diagnoses are more likely to present with advanced-stage disease (7), suggesting either the symptoms associated with HRT use leads to delayed clinical diagnosis, or that HRT increases CRC development and proliferative rates.
As HRT, and thus estrogens, may influence CRC proliferation, the local colonic tissue activation of estrogens via steroid sulfatase (STS) and 17-hydroxysteroid dehydrogenases (HSD17B) must be important (8). The expression of STS, the fundamental enzyme desulfating circulating estrogens to their active forms ( Figure 1A), is prognostic for CRC survival (9), and mRNA expression of HSD172, which catalyzes estradiol (E 2 ) to E 1 , is down-regulated in human CRC tissue (10) suggesting estrogen metabolism as important in CRC progression. However, little is known about HSD171, HSD177, and HSD1712 expression, all of which activate E 1 to E 2 (11,12).
Questions also remain regarding how estrogens act in CRC. ER has either low (13) or no (14) expression in both normal colon and CRC, although splice variants do exist (15).
Furthermore, loss of the pro-apoptotic ER, which implies subsequent dominance of other ERs, defines CRC progression (16). Indeed, a recent meta-analysis has confirmed the loss of ER expression as CRC develops (17). However, no human CRC studies have examined the G-protein coupled estrogen receptor (GPER), an endoplasmic reticulum membrane-bound receptor with high E 2 -binding affinity (18) with known pro-proliferative actions in breast (19) and endometrial cancer (20).
Here we aimed to determine how estrogen metabolism and action impacts CRC. By examining key estrogen metabolizing enzymes in matched normal and cancerous human colorectal tissue, and then translating findings to in vitro and in vivo systems, we demonstrate for the first time that CRC exhibits dysregulated estrogen metabolism with STS activity and estrogen reductase pathways elevated in CRC. We go on to show greater STS activity increases estrogen-stimulated CRC proliferation in vitro and in vivo through GPER-activation via increased expression of connective tissue growth factor (CTGF), a known modulator of GPER action (21). Finally, we demonstrate that GPER expression is elevated in human CRC tissue, with this significantly correlating to increased CTGF expression. Thus, both STS and GPER inhibition may represent novel therapeutic targets for patients with CRC.

Estrogenic enzymes favor E 2 metabolism in human CRC.
Immunohistochemical studies show STS expression is increased in human CRC (9).
However, as STS expression does not correlate with enzyme activity, and no data exists on STS activity in human colon, we determined STS activity in human CRC and histopathologically unchanged colonic mucosa located at least 10-20 cm away from cancerous lesions (see patient characteristics for 64 participants: Supplementary  Figure 1B). Although not formally tested, plotting the data suggests a more pronounced effect in females ( Figure 1B). Increased STS activity does not correlate with increasing STS mRNA expression (STS dCT) in either normal or cancerous tissues ( Figure 1C) (calculated correlation coefficient (p-value) g 0.27 (0.07) and 0.04 (0.77) for normal and cancerous samples, respectively). Indeed, RNASeq data (RNASeq V2) analyzed from The Cancer Genome Atlas (TCGA) colon adenocarcinoma (COAD) dataset showed no significant change in STS expression from normal to cancerous tissue (Supplementary Figure   1). As STS activity is altered by various post-translational modifications (22) this suggests determining only STS expression does not represent in situ colon activity. Furthermore, STS activity does not correlate to Duke's or T-staging (Table 1), indicating increased STS activity is most likely an early event in tumor formation.
As STS desulfates circulating and peripheral E 1 S to E 1 , we next determined the expression of enzymes that oxidize E 2 to E 1 (HSD17B2) and reduce E 1 to E 2 (HSD171, HSD177, and HSD1712) in the same human CRC samples. HSD17B2 mRNA significantly (p<0.01) decreased in CRC tissue compared to matched controls (Figure 1D, Supplementary Table 2,   Supplementary Table 3 for raw dCT values). HSD171 mRNA was not detectable (data not shown). HSD177 and HSD1712 mRNA were significantly increased in female and male CRC tissue compared to matched controls (Supplementary Table 2, Supplementary Table 3 for raw dCT values, and Figure 1E, 1F). This data was supported by further analysis of the TCGA COAD data, which demonstrated a significant decrease in mRNA expression of HSD17B2, and increased expression of HSD17B7 and HSD17B12, in colon cancer (Supplementary Figure 1). Immunoblotting ( Figure 1G) and subsequent densitometry analysis ( Figure 1H) of normal and cancerous tissue confirmed lack of HSD171 and increased HSD17B12 expression. HSD177 protein expression showed a trend towards increased expression in CRC. In contrast to mRNA data ( Figure 1F), HSD172 protein was not decreased in CRC compared to controls ( Figure 1H). HSD174, which oxidizes E 1 to E 2 , expression was not determined as previous studies have shown this as significantly downregulated in human CRC (23). Taken together our data suggests CRC up-regulates pathways favoring E 1 S hydrolysis and subsequent E 2 synthesis.
Estrogen metabolizing enzyme expression defines CRC estrogenic proliferative response.
As human CRC exhibited dysregulated estrogen metabolism we hypothesized CRC cell lines expressing E 2 synthesis pathways may be more responsive to estrogen signaling. Thus, we determined expression patterns of key estrogen metabolizing enzymes in selected CRC cells.
Compared to human CRC tissue, HCT116 and HT-29 cells exhibited similar HSD17 mRNA (data not shown) and protein ( Figure 2A) expression (i.e. lack of HSD171, presence of HSD177 and HSD1712, limited HSD172 expression). In contrast, Caco2 cells have low HSD177 and HSD1712, and higher HSD172 expression. Colo205 cells had low or no HSD17Bs mRNA (not shown) and protein expression and thus these cells were not used in further testing. When incubated for 72h with E 1 ( Figure 2B, raw absorbance data Supplementary Figure 2A) or E 2 ( Figure 2C, raw absorbance data Supplementary Figure 2B) in charcoal-stripped FBS (sFBS) media, HCT116 and HT-29 cells had increasing dosedependent proliferative rates compared to sFBS media controls. Caco2 cells failed to respond to E 1 or E 2 stimulation.
Using LC-MS/MS we next examined how CRC cells metabolized estrogens over 24h.
HCT116 cells did not significantly metabolize E 2 to other oestrogen metabolites, HT-29 cells metabolized E 2 to unknown metabolites, and Caco2 cells rapidly oxidized E 2 to E 1 ( Figure   2D and 2E) indicative of its high HSD17B2 reductase expression. This suggests oxidation of E 2 via HSD172, expressed in Caco2 but not HCT116 and HT-29 cells, impacts local E 2 availability and consequently the ability of Caco2 cells proliferation to E 2 . This further implies that peripheral estrogen metabolism in CRC may define the tumors responsiveness to estrogen action.

STS over-expression augments E 1 S-and E 2 S-stimulated proliferation in CRC.
As STS activity was significantly increased in human CRC samples (Fig. 1B) Figure   3B).
In full media, HCT116[sts] proliferation significantly increased compared to HCT116 [vo] cells ( Figure 3A), with this augmented growth blocked by the non-cytotoxic, specific STS inhibitor STX64. Incubation of these same cells in sFBS media supplemented with E 1 , E 2 , or E 1 S (at 100nM) for 72h, a significant (p<0.001) growth difference was observed between Figure 3C). This demonstrates greater STS desulfation of E 1 S, leading to increased E 1 liberation driving proliferation. When these cells were grown for 72h in sFBS supplemented with E 2 S (100nM), all proliferated in response to E 2 S, with the greatest increase seen in HCT116[sts] cells compared to sFBS controls ( Figure 3B). STX64 blocked this increased growth suggesting estrogen desulfation as an important regulator in CRC proliferation.

STS over-expression increases CRC xenograft growth.
As  Figure 3C), leading to a greater tumor burden by day 21 post implantation ( Figure 3D and Figure 3E). Dosing of STX64 (20mg/kg, p.o., thrice weekly) initially completely stagnated (days 3-18) HCT116[sts] growth ( Figure   3C), although tumors were proliferating by day 24. Despite tumor STS activity being almost completely ablated by STX64 treatment ( Figure 3F), HCT116[vo] xenograft growth was not affected by STS inhibition. This suggests once STS is over-expressed, CRC may rely more heavily on estrogen desulfation as pro-proliferative.

Estrogens increase proliferation through GPER signaling in CRC.
As controversy surrounds how estrogens elicit their effects in CRC (24), we investigated whether GPER was expressed in human CRC and if GPER-stimulation augmented CRC proliferation. In contrast to HCT116 and HT-29 cells, Caco2 and Colo205 cells express ERNone of the CRC cell lines tested expressed ER, but all expressed GPER ( Figure 4A).
Others have shown (21) in breast cancer that GPER stimulation with E 2 can increase proliferation and increase the expression of various downstream regulators of survival and migration ( Figure 4B). Thus, we next examined whether the specific GPER agonist G1 (72h treatment) stimulated HCT116, HT-29 and Caco2 cell proliferation, as measured by BRdU incorporation, compared to sFBS controls ( Figure 4C). G1 induced a dose-dependent stimulation in proliferation, with this effect more pronounced in HCT116 and HT-29. These results mimicked proliferative effects by E 1 and E 2 (see Figure 2B and Figure 2C).
Intriguingly, Caco2 cells modestly responded to G1 treatment in contrast to their lack of increased proliferation in response to E 2 ( Figure 2B) supporting the notion that rapid E 2 oxidization in Caco2 limits estrogenic effects. However, when GPER is stimulated by G1, Caco2 cells can increase proliferation through this pathway. In HCT116 and HT-29 cells, the GPER antagonist G15 (1M) blocked both E 2 -and G1-stimulated proliferation over 72h compared to controls (Supplementary Figure 4A).
To further delineate GPER action in CRC, we also determined how E 2 and G1 affected downstream molecular regulators of GPER action (21). Figure 4B illustrates the key genes that we examined in CRC cells, namely FOS, EGR1, ATF3, CTGF, DUSP1, and TNF. All these genes are upregulated in response to GPER stimulation in breast cancer cell lines (21). In HT-  Figure 4F), were significantly elevated in response to E 2 (100nM for 24h) and G1 (100nM for 24h) compared to sFBS treated cells. In HCT116 cells, EGR1, ATF3 and CTGF were significantly elevated in response to treatment. As CTGF gave the largest response to E 2 and G1 stimulation, we further examined its protein expression in response to treatment. In HCT116 and HT-29 cells, E 2 (100nM) and G1 (100nM) increased CTGF protein expression after 24h as measured by immunoblotting ( Figure 4D).
To confirm the importance of GPER or CTGF in mediating the pro-proliferative effects of E 2 , we next did transient knockdown of these two proteins using siRNA and determined their response to E 2 treatment. In HCT-116 cells, siRNA of GPER and CTGF provided protein knockdown for 96h, the length of time required for subsequent proliferation studies ( Figure   4E). Knockdown of GPER and CTGF significantly (p<0.001 and p<0.05, respectively) inhibited the proliferation driven by E 2 and G1 in HCT-116 ( Figure 4F). Intriguingly, when we again moved into an in vivo model of CRC, the use of the GPER antagonist G15 (at 50g/kg i.p. thrice weekly) also significantly (p<0.01) inhibited HCT116[sts] xenograft growth implanted into female nude mice ( Figure 4G). Indeed, interrogation of the TCGA COAD data-set indicated that although all estrogen receptors (ER, ER, and GPER) were significantly (p<0.0001) down-regulated in CRC compared to normal controls (Supplementary Figure 6A), GPER still has the highest expression in CRC. Further analysis of the TCGA dataset demonstrated that patients with CRC tumors expressing high GPER had significantly (p=0.0431) poorer outcomes compared to low to mid expression levels ( Figure   4H). Indeed, CRC with high ER also had significantly (p=0.0265) worse outcomes (Supplementary Figure 6C), suggesting the importance of these pro-proliferative pathways in CRC. High ER expression did not affect CRC patient outcomes (Supplementary Figure 6D).  Figure 6B) and protein was significantly (p<0.001) increased in CRC as determined by relative densitometry ( Figure 5B). Relative intensity of immunoblots for GPER and CTGF highlighted a significant (p=0.0042) positive correlation between GPER and CTGF expression in cancerous tissue but not matched normal controls ( Figure 5C). As GPER stimulation increases CTGF expression our results indicate greater estrogen availability through STS activity in these tumors may lead to increased GPER stimulation and CTGF expression ( Figure 5D).

Discussion
Here we demonstrate a critical role for pre-receptor local estrogen metabolism and action in the proliferation of CRC. For the first time we show estrogen synthesis pathways, via STS, HSD17B7 and HSD17B12, are elevated in CRC and estrogens stimulate CRC growth through a GPER-mediated mechanism. Of particular importance is STS, a key regulator in estrogen activation. When over-expressed in HCT116 cells STS drives greater tumor proliferation in in vitro and in vivo models. Finally, we demonstrate E 2 acts through GPER signaling, most likely via CTGF, in CRC, and that both GPER and CTGF are increased in human CRC. Our results suggest that inhibiting GPER or estrogen metabolism may be a novel therapeutic option for this malignancy.
Controversy exists on estrogens role in CRC development and progression. The Women's Health Initiative (26) has highlighted various questions on how estrogens and progestins impact cancer. Epidemiological studies indicate estrogens as protective against CRC development. However, how estrogens impact CRC once it has developed is poorly defined.
It has been suggested that whilst initially protective, estrogens may be mitogenic in CRC (24) through changes in local estrogen metabolism and receptor availability. Indeed, estrogens promote tumorigenesis in colitis-associated CRC (27), and E 2 increases LoVo cell line proliferation via up-regulation of fatty acid synthesis (28). However, few studies have investigated enzymes involved in estrogen metabolism in CRC, and the ones that have overlooked key 17HSDs and STS activity. Furthermore, although evidence strongly suggests that ER down regulation, and thus the loss of this pro-apoptotic pathway, is an important turning-point in CRC development (16), whether GPER expression or stimulation impacts CRC has not previously been determined.
We show that STS activity is significantly elevated in human CRC and that STS overexpression stimulates CRC cell proliferation. Previous findings had indicated increased STS expression as prognostic for CRC survival (9), however this study did not measure STS activity. This is an important distinction. STS is subject to post-translational modifications affecting activity and we show here colon STS activity and expression do not correlate.
Furthermore, analysis of the TCGA COAD database demonstrated no significant changes in STS expression in colon cancer compared to normal controls. Although eventual patient outcomes have not been determined, we show STS activity does not correlate to Duke's staging or tumor T-stage implying increased STS activity is most likely an early event in CRC development, and thus its prognostic significance is questionable (9). Along with increased aromatase expression, elevated STS activity is a hallmark of estrogen-dependent cancers (29). STS inhibition is currently in Phase II clinical trials in patients with hormonedependent breast cancer, after it had shown promise in pre-clinical studies against E 2 Sstimulated breast cancer in vivo as well as in Phase I trials (30). As aromatase expression is not detectable in the human colon (9), local desulfation of circulating E 1 S may act as the primary route for estrogen availability in CRC.
Once desulfated, HSD171, HSD177, and HSD1712 reduce E 1 to E 2 , with HSD172 catalyzing reverse oxidation. Supporting our findings, TCGA COAD analysis and others (23) have shown that HSD17B2 expression is down-regulated in CRC, however our data here indicates no change in HSD172 protein expression suggesting this pathway may remain active. Although HSD171 is the prime reducer of E 1 (31) we demonstrate this enzyme is absent in CRC. Interestingly, HSD177 and HSD1712 expression are significantly up-regulated in CRC compared to matched normal controls, with this effect mimicked at the protein level, and our findings are supported by TCGA COAD data analysis. Thus, CRC may favor E 2 synthesis. Recently, pre-clinical studies show inhibition of HSD17B7 in hormonedependent breast cancer blocks E 1 to E 2 synthesis and thus has therapeutic potential (32). As intratumoral E 1 and E 2 concentrations in CRC tissue pertains to a poor prognosis (9), inhibiting these enzymes in CRC may be therapeutically beneficial.
As ER and ER are not present in the CRC cell lines tested, a question arose, how do estrogens act in CRC? Limited data on colonic GPER expression exists: GPER stimulation may affect colonic motility in mice (33) and its expression may influence abdominal pain severity in IBD (34). We demonstrate GPER protein, but only limited mRNA, is expressed in human CRC tissue and cell lines. GPER protein expression is elevated in human CRC tissue compared to matched normal controls, in contrast to mRNA which is decreased. This may imply that GPER protein degradation pathways may be altered in CRC, effectively allowing for GPER protein retention. Stimulation of GPER by E 2 or the specific agonist G1 increased CRC proliferation in vitro, with this effect blocked by GPER inhibition in in vitro and in vivo CRC models. In contrast to our findings, recent research has shown GPER stimulation by G1 decreases proliferation of various CRC cell lines, including HCT116 (35). However, these studies used higher doses of G1 (up to 10 M) compared to our 100 nM dose, and, unlike the work presented here, the studies are not performed in stripped-media (i.e. no/low estrogen) conditions. Thus, this suggests there may be a biphasic response to G1 and estrogens with regards GPER stimulation in CRC, with low doses increasing proliferation and high doses inducing apoptosis. This biphasic response is also evident with ER stimulation in breast cancer (36). GPER deficiency results in multiple physiological alterations including obesity, cardiovascular dysfunction, insulin resistance and glucose intolerance (37), and there is much interest in its pro-proliferative effects in breast cancer. In breast cancer patients, GPER expression is associated with increased primary tumor size and the prevalence of distant metastasis (38). Indeed, GPER-stimulation by tamoxifen is a potential pathway of tamoxifenresistant hormone-dependent breast cancer (39) and intriguingly, breast cancer patients treated with tamoxifen are more likely to develop CRC (40). Our results strongly implicate E 2 -GPER-mediated action, through CTGF, in CRC proliferation. As the loss of ER defines CRC development (16), it will be of interest to further examine GPER action in the context of ER expression to determine whether an estrogen receptor "switch" occurs during CRC progression.
Furthermore, in CRC cell lines, the expression of CTGF, a known downstream regulator of GPER action (21), was elevated by E 2 and G1 treatment. A correlation is evident between GPER and CTGF expression in human CRC tissue. CTGF is up-regulated in some CRC patients (41), although its expression reduces in latter-stage disease (42). Analysis of the TCGA COAD dataset also suggests that high CTGF is related to poor patient survival, although others have shown high CTGF expression correlates to improved CRC survival rates (41). This implies a complicated relationship between E 2 -stimulation of GPER, increased proliferation, CTGF-mediated effects, and patient outcomes. However, in general, dysregulation of CTGF expression is linked to poor outcomes in many human cancers (43).
In conclusion, we have identified a new estrogen-driven proliferative pathway in CRC.
Increased STS activity leads to greater estrogen desulfation, thereby increasing HSD17B substrate availability for subsequent E 2 synthesis, followed by GPER activation and CTGF up-regulation. These findings identify STS, 17BHSD7, 17BSHD12, and GPER as potential new therapeutic targets for CRC.

Experimental Procedures
Compounds STX64 (Irosustat) was from Sigma-Aldrich Ltd. (Dorset, UK). G1 and G15 were from Torcis Bioscience (Abingdon, UK). E 1 S, E 1 , E 2 S, and E 2 were from Sigma-Aldrich. [4-14 C] E 1 (1 x 10 4 dpm, Perkin-Elmer) was included to monitor procedural losses. Samples were incubated at 37 o C after which the product, E 1 , was separated from E 1 S by partition with toluene. 3 H and 14 C radioactivity was measured by liquid scintillation spectrometry. Mass of E 1 S hydrolysed was calculated from 3 H counts detected corrected for procedural losses.

Human tissue and cell culture
Results were determined as pmol product formed/h/mg protein.

qRT-PCR analysis
From human samples 30mg tissue was homogenised in RLT buffer containing mercaptoethanol. cDNA was manufactured with SENSIFast kit (Bioline) using 1g mRNA as per the manufacturers' instructions. From cell lines mRNA was purified using RNeasy kits (QIAGEN) as per the manufacturers' instructions. mRNA samples were reverse transcribed to form cDNA using Tetro cDNA Synthesis Kit (Bioline Reagents Ltd.).
Expression of specific mRNAs was determined on a 7500 Real-time PCR system (Applied Biosystems) using the QuantiTect Probe RT-PCR kit (Qiagen). Relative expression was determined using the 2 -∆∆Ct method. Taqman assays are described (Supplementary Table 4

LC-MS/MS
Estrogens were measured by uHPLC-MS/MS. After addition of internal standard steroids (E 1 S-d4, E 2 S-d4 (Cambridge Isotopes) and 13C E 2 (Sigma)) samples were extracted using solid phase extraction (C18 Isolute SPE columns 500mg, Biotage). Estrogens were quantified relative to a calibration series (0.5-500ng/L) via tandem mass spectrometry. A Waters Xevo mass spectrometer with an electrospray ionization source was used with an attached Acquity liquid chromatography system. Estrogens were eluted from a HSS C18 SB 1.8um, 2.1 x 30mm column using a methanol/water gradient system with 0.3mM ammonium fluoride added to the aqueous phase. The coefficient of variation for all assays was less than 20%.

siRNA Design and Transfection
The siRNA oligonucleotides and transfection reagents were purchased from Dharmacon, Inc.

In vivo xenograft studies
Six week old athymic, female CD-1 nude mice (nu−/nu−) were purchased from Charles River. All experiments were carried out under conditions that complied with institutional guidelines. Five million HCT116 cells were injected s.c. into the right flanks of the animal.
Tumor volumes were calculated using the formula (length × width 2 /2). At the conclusion of dosing, animals were terminated and their tumors removed, weighed, and stored at -80 o C.

Proliferation assays
Cell proliferation was measured by the CyQuant cell proliferation (Thermo Fisher Scientific) and BrdU incorporation assays (Roche Applied Science) as per the manufacturer instructions.
Prior to experiments, cells were placed into stripped-FBS phenol-red free medium (Thermo Scientific) with 5mM L-glutamine for 72h to clear any remaining estrogens in the media.
Cells were cultured in flat-bottom 96-well plates in either complete or stripped FBS phenolfree growth media containing estrogens and subsequent assays performed.

Statistics
For human data the population analyzed was described using summary statistics and relationships between STS activity and STS expression investigated by plotting the data and   whereas HSD17B7 is expressed, and HSD17B12 expression is increased (n = 16), with little change in HSD17B2 expression (n = 16). For relative intensity data, a 2-tailed Student's t test was used. All data represents mean ± s.d. Caco-2 cells rapidly metabolized E 2 to E 1 . (n = 3 independent experiments). All data represents mean ± s.d.  (Kaplan-Meier survival analysis (Log-rank method). All data represents mean ± s.d.