Intratumoral Estrogen Disposition in Breast Cancer

Home , HSD17B10, HSD17B2

Published OnlineFirst March 9, 2010; DOI: 10.1158/1078-0432.CCR-09-2481 Published Online First on March 9, 2010 as 10.1158/1078-0432.CCR-09-2481

Clinical Human Cancer Biology Cancer Research Intratumoral Estrogen Disposition in Breast Cancer

Ben P. Haynes1, Anne Hege Straume3,4, Jürgen Geisler6, Roger A'Hern7, Hildegunn Helle5, Ian E. Smith2, Per E. Lønning3,5, and Mitch Dowsett1

Abstract

Purpose: The concentration of estradiol (E2) in breast tumors is significantly higher than that in plasma, particularly in postmenopausal women. The contribution of local E2 synthesis versus uptake of E2 from the circulation is controversial. Our aim was to identify possible determinants of intratumoral E2 levels in breast cancer patients. Experimental Design: The expression of genes involved in estrogen synthesis, metabolism, and sig- naling was measured in 34 matched samples of breast tumor and normal breast tissue, and their correlation with estrogen concentrations assessed. Results: ESR1 (9.1-fold; P < 0.001) and HSD17B7 (3.5-fold; P < 0.001) were upregulated in ER+ tumors compared with normal tissues, whereas STS (0.34-fold; P < 0.001) and HSD17B5 (0.23-fold; P < 0.001)

were downregulated. Intratumoral E2 levels showed a strong positive correlation with ESR1 expression in all patients (Spearman r = 0.55, P < 0.001) and among the subgroups of postmenopausal (r = 0.76, P < 0.001; n = 23) and postmenopausal ER+ patients (r = 0.59, P = 0.013; n = 17). HSD17B7 expression showed a significant positive correlation (r =0.59,P < 0.001) whereas HSD17B2 (r = −0.46, P = 0.0057) and

HSD17B12 (r = −0.45, P = 0.0076) showed significant negative correlations with intratumoral E2 in all patients. Intratumoral E2 revealed no correlation to CYP19, STS, and HSD17B1 expression. Multivariate models comprising ESR1 and plasma E2 predicted between 50% and 70% of intratumoral E2 variability. Conclusion: Uptake due to binding to the ER, rather than intratumoral estrogen synthesis by aroma-

tase or sulfatase, is the single most important correlate and a probable determinant of intratumoral E2.An increased expression of HSD17B7 may explain the increased ratio of E2 to estrone (E1) in breast tumors compared with normal tissue. Clin Cancer Res; 16(6); 1790–801. ©2010 AACR.

Estrogens play a pivotal role in breast cancer devel- is the most effective endocrine treatment for postmeno- opment (1), and estrogen suppression is an effective pausal women with ER+ tumors (5). therapeutic option among premenopausal as well as post- It is well established that the concentration of E2,the menopausal women harboring estrogen receptor–positive most biologically active estrogen, in breast tumors can (ER+) tumors (2, 3). In postmenopausal women, estrogens be as much as 10-fold higher than plasma levels in post- are synthesized through aromatization of circulating an- menopausal women (6). Several explanations, including drogens (mainly androstenedione and, to a minor degree, local estrogen synthesis (7–9) as well as active uptake testosterone) into estrogens (mainly E1 and, to a minor de- from the circulation (10), have been suggested. Studies at- gree, E2) in peripheral tissues (4). Inhibition of aromatase tempting to directly evaluate the contribution from local synthesis versus circulatory uptake have involved a limited number of patients and have indicated substantial interin- dividual variation (7, 11). Authors' Affiliations: 1Academic Biochemistry and 2Medicine, Royal Marsden Hospital, London, United Kingdom; 3Department of Oncology, A role for aromatase in local estrogen synthesis has Haukeland University Hospital; 4Department of Molecular Biology and been indicated by studies that reported higher levels 5Section of Oncology, Institute of Medicine, University of Bergen, of aromatase activity and mRNA in the adipose tissue Bergen, Norway; 6Institute of Medicine, University of Oslo, Faculty Division at Akershus University Hospital, Lørenskog, Norway; and directly adjacent to breast cancers compared with nor- 7Cancer Research UK Clinical Trials and Statistics Unit, Institute of mal breast tissue distant from the tumor (12, 13). Cancer Research, Sutton, Surrey, United Kingdom However, only a few weak correlations between clinical Note: Supplementary data for this article are available at Clinical Cancer outcome and biochemical measurements of aromatase Research Online (http://clincancerres.aacrjournals.org/). have been forthcoming possibly due to the low abun- Corresponding Author: Ben P. Haynes, Department of Academic Bio- dance of the enzyme (14). Immunohistochemical as- chemistry, Royal Marsden Hospital, Fulham Road, London SW3 6JJ, United Kingdom. Phone: 44-20-7808-2969; Fax: 44-2073763918; E-mail: sessment of aromatase expression has been fraught [email protected]. with difficulties. Notably, in a recent study applying a doi: 10.1158/1078-0432.CCR-09-2481 well-validated antibody, aromatase expression did not ©2010 American Association for Cancer Research. predict clinical response to letrozole in the neoadjuvant

1790 Clin Cancer Res; 16(6) March 15, 2010

Intratumoral Estrogen Disposition in Breast Cancer

showing higher E2 in tumors than in serum or nontumor Translational Relevance tissue, the concentrations reported varied widely and some studies showed no difference between premenopausal and Estrogen deprivation is a major means of endocrine postmenopausal women. This may partly be related to the – therapy of estrogen receptor positive breast cancer difficulty of accurately measuring estrogen concentrations and, through aromatase inhibition, is the most effec- in breast tissues (6, 29–32). tive treatment for postmenopausal women. Inhibition In a recent study (33), some of us confirmed elevated E2 of other enzymes involved in estrogen disposition and levels in tumor as compared with normal tissue. This relat- targeted aromatase inhibition or suppression in breast ed to ER+ tumors only. In contrast, intratumoral levels of tumors have been suggested as alternative or addition- E1 were consistently lower than normal tissue levels from al strategies. The rationale for these depends on a clear the same breast. The aim of the current study was to identify understanding of the contributors to intratumoral factors that might explain the altered estrogen disposition estradiol (E2) levels, but data on this are varied. Our between normal and malignant tissues. Thus, we correlated current study revealed significant correlations of intra- estrogen levels with mRNA levels of the estrogen receptor tumoral E2 with plasma estrogens and intratumoral (ESR1)andallthemajorestrogen-metabolizinggenes α β transcript levels of ER and some of the 17 -hydroxys- [aromatase (CYP19), STS, SULT1E1, and six isoforms of teroid dehydrogenase enzymes involved in estrogen HSD17B] by quantitative real-time PCR in these same interconversion, but not with enzymes involved in es- matched normal and malignant breast tissue samples. trogen synthesis. The data provide strongest support These data would provide correlative evidence for, or for targeting systemic, rather than intratumoral, estro- against, a role for (a) uptake due to binding to ERα,(b) gen synthesis, with HSD17B7 being a potential novel intratumoral synthesis by estrogen metabolism genes, secondary target. and/or (c) plasma estrogen concentrations as contributors to intratumoral E2 levels. In addition, the expression of two classic estrogen-dependent (reporter) genes, PGR and TFF1, was measured as markers of the intratumoral levels of E2. setting, although it was found to be a favorable prognostic biomarker (15). Materials and Methods Sulfatase (STS) catalyzes the formation of E1 from estrone sulfate (E1S), which may act as a possible reservoir Patients. Details of the patients included in this study of E1 due to its high circulating concentrations and pro- have been published elsewhere (33). In brief, premeno- longed half-life (16, 17). The much higher activity of pausal and postmenopausal breast cancer patients select- STS compared with aromatase in breast cancer has been ed for mastectomy (at the Department of Surgery, taken as suggesting that it may be more important than Haukeland University Hospital, Bergen, Norway) were aromatase for intratumoral estrogen synthesis (8). In sup- eligible but were excluded if they had taken in the pre- port of this, STS mRNA levels have been shown to be sig- vious 6 months any kind of hormone replacement ther- nificantly higher in malignant breast tissue than in normal apy or any drug known to interfere with estrogen tissue (18), and high levels of expression of STS are asso- disposition. Tissues were obtained from mastectomy ciated with poor prognosis in ER+ breast cancer (19). specimens immediately on removal and snap-frozen in E1 formed as a result of aromatase or STS activity must liquid nitrogen. About 500 mg of tumor tissue and an be reduced to E2 to achieve its full biological potency, and equivalent amount of normal tissue from each of the this interconversion is done by 17β-hydroxysteroid dehy- four breast quadrants were obtained. All tissue samples drogenase (17βHSD). Multiple isoforms of this enzyme were subsequently stored in liquid nitrogen until proces- exist, with types HSD17B1, HSD17B5, HSD17B7, and sing. Blood samples (20 mL, heparinized) for plasma HSD17B12 acting reductively to catalyze the activation hormone measurements were drawn on the morning of E1 to E2 in breast tissue (20–23). Conversely, HSD17B2, of the day of surgery following an overnight fast, and HSD17B10, and HSD17B14 are thought to be the most plasma was stored at −20°C until processing. Patients important oxidative isoforms of the enzyme in breast tis- gave their written informed consent before inclusion sue, inactivating E2 by converting it to E1 (24–26). Where- according to national regulations (33). as the overall peripheral metabolism of estrogens strongly Methods. Estrogen measurements in these matched favors E2 inactivation, the situation is opposite within tu- breast tumor, normal breast tissue, and plasma samples mor tissue where E1 is preferentially converted to E2 (27). have recently been reported by us (33). The analytic meth- Recent data from our group found HSD17B7 expression to ods, including details with respect to detection limits and be positively correlated with response to aromatase inhibi- coefficients of variations, have been published previously tion in advanced breast cancer (28). (34, 35). Although numerous studies have measured plasma es- RNA extraction and cDNA synthesis. Total RNA was ex- trogen levels in postmenopausal women, there are few tracted from breast tumor samples and a matched normal studies reporting intratumoral estrogen levels. These stud- breast tissue sample from the breast quadrant farthest ies have been mostly small, and although consistently from the tumor-containing quadrant using Trizol reagent

www.aacrjournals.org Clin Cancer Res; 16(6) March 15, 2010 1791

Haynes et al.

(Life Technologies). RNA was dissolved in diethylpyrocar- tems), and 0.5 μL of the TaqMan Gene Expression assay, or bonate-treated deionized water, then the concentration for TBP, 0.45 μmol/L of the forward primer, 0.15 μmol/L was measured using an ND1000 spectrophotometer and of the reverse primer, and 0.2 μmol/L of the probe in a adjusted to 25 ng/μL. Single-strand cDNA was synthesized total volume of 10 μL. The thermocycling conditions used from 200 ng of total RNA using SuperScript III reverse were 50°C for 2 min and 95°C for 10 min, followed by transcriptase (Invitrogen) for all genes other than CYP19 40 cycles of denaturation at 95°C for 15 s and annealing/ and ribosomal protein P2. For CYP19 and ribosomal pro- elongation at 60°C for 1 min. Genes were amplified inde- tein P2, single-strand cDNA was synthesized from 125 ng pendently in separate reaction wells as triplicates, and in of total RNA using Transcriptor reverse transcriptase all experiments, samples without template were used as (Roche) according to the manufacturer's procedure; equal negative control. The results obtained were converted into amounts of oligoT and random primers were used in the relative concentrations using an in-run standard curve and cDNA synthesis reaction mix. the observed relative concentrations were normalized to Quantitative real-time PCR. TaqMan Gene Expression the mean of the two housekeeping genes (TBP and assays (PE Applied Biosystems) were used to measure MRPL19) to allow for variation in the amount of input the RNA expression of ESR1,PGR,TFF1,HSD17B1, cDNA. HSD17B2, HSD17B5, HSD17B7, HSD17B12, HSD17B14, For CYP19, mRNA levels were quantified in triplicate re- STS (steroid sulfatase), SULT1E1 (sulfotransferase), actions along with the ribosomal protein P2 mRNA, using CYP17A1 (17-hydroxylase-lyase), and the housekeeping the Lightcycler 480 instrument (Roche). The amplification gene MRPL19. Designed primers and probe to measure primers and BlackBerry-quenched hydrolysis probes that TBP, a second housekeeping gene, were purchased from were used (TIB MOLBIOL) are given in Supplementary PE Applied Biosystems. The TaqMan Gene Expression as- Table S1. Amplification was done using the LC480 Probes say numbers and primer and probe sequences for TBP are Master (Roche) reaction mix, 0.5 μmol/L of each primer, giveninSupplementaryTableS1.PCRreactionswere 0.125 μmol/L of each hydrolysis probe, and 0.5 μLof done using an ABI Prism 7900HT Sequence Detection Sys- cDNA (3 ng). The thermocycling conditions used were tem (PE Applied Biosystems). 95°C for 5 min followed by 50 cycles of denaturation at The amplification reaction mixture consisted of 3 μL 95°C for 10 s, annealing/elongation at 55°C for 30 s, cDNA (3 ng), 5 μL TaqMan MasterMix (PE Applied Biosys- and finally 40°C for 10 s. In each run, samples without

Table 1. Clinicopathologic characteristics of patients

Parameter Premenopausal Postmenopausal

No. of patients 11 23 Age (y), mean (range) 42 (31-49) 62 (44-81) BMI, mean (range) 24 (18-43) 25 (20-38) Tumor size (cm), n (%) >0.5-1.0 — 1 (4) >1.0-2.0 1 (9) 5 (22) >2.0-5.0 9 (82) 17 (74) Unknown 1 (9) — Tumor grade, n (%) 1 1 (9) — 2 4 (36) 13 (57) 3 6 (55) 10 (43) Nodal status, n (%) Node+ 6 (55) 8 (35) Node− 5 (45) 15 (65) IHC ER, n (%) ER+ 5 (45) 17 (74) ER− 6 (55) 6 (26) IHC PR, n (%) PR+ 6 (55) 14 (61) PR− 5 (45) 9 (39) IHC HER2+, n (%) 3 (27) 1 (4)

Abbreviations: BMI, body mass index; IHC, immunohistochemistry; PR, progesterone receptor.

1792 Clin Cancer Res; 16(6) March 15, 2010 Clinical Cancer Research

Intratumoral Estrogen Disposition in Breast Cancer

Table 2. Spearman correlation of normal breast tissue estradiol levels with plasma estrogen levels and normal breast tissue gene expression

All patients Postmenopausal

Plasma E2 r (95% CI) 0.79 (0.60-0.90) 0.48 (0.05-0.76) P (n) <0.001 (29) 0.027 (21)

Plasma E1 r (95% CI) 0.82 (0.65-0.92) 0.56 (0.16-0.80) P (n) <0.001 (29) 0.008 (21)

Plasma E1S r (95% CI) 0.77 (0.56-0.89) 0.50 (0.075-0.77) P (n) <0.001 (29) 0.021 (21) ESR1 r (95% CI) −0.030 (−0.39 to 0.34) −0.25 (−0.62 to 0.20) P (n) 0.87 (31) 0.26 (22) CYP19 r (95% CI) −0.015 (−0.38 to 0.35) 0.39 (−0.05 to 0.71) P (n) 0.94 (31) 0.072 (22) HSD17B1 r (95% CI) −0.040 (−0.40 to 0.32) 0.33 (−0.12 to 0.67) P (n) 0.83 (31) 0.14 (0.22) HSD17B2 r (95% CI) 0.25 (−0.13 to 0.56) 0.083 (−0.36 to 0.50) P (n) 0.18 (31) 0.71 (22) HSD17B5 r (95% CI) −0.033 (−0.39 to 0.34) 0.19 (−0.27 to 0.58) P (n) 0.86 (31) 0.40 (22) HSD17B7 r (95% CI) −0.083 (−0.43 to 0.29) −0.13 (−0.53 to 0.33) P (n) 0.66 (31) 0.58 (22) HSD17B12 r (95% CI) −0.24 (−0.55 to 0.14) −0.056 (−0.48 to 0.39) P (n) 0.20 (31) 0.80 (22) HSD17B14 r (95% CI) −0.11 (−0.46 to 0.26) 0.13 (−0.32 to 0.53) P (n) 0.55 (31) 0.58 (22) STS r (95% CI) −0.14 (−0.48 to 0.24) 0.23 (−0.23 to 0.60) P (n) 0.45 (31) 0.31 (22) PGR r (95% CI) 0.24 (−0.13 to 0.56) −0.019 (−0.45 to 0.42) P (n) 0.19 (31) 0.93 (22) TFF1 r (95% CI) 0.48 (0.14-0.72) 0.14 (−0.32 to 0.54) P (n) 0.006 (31) 0.55 (22)

NOTE: Significant correlations (P < 0.05) are in boldface; n is the number of data points (for full details, please refer to “Patient demographics”). Abbreviation: 95% CI, 95% confidence interval. template were used as a negative control. For each analysis, Data analysis. Statistical analysis was done using SPSS. the results were converted into relative concentrations us- Correlations between individual levels of gene expression ing an in-run standard curve, and the observed relative con- and estrogen concentrations were assessed using Spear- centrations for CYP19 were normalized in accordance with man rank correlation. Differences between normal tissue the ribosomal protein P2 levels. and tumor expression were assessed using Wilcoxon's

www.aacrjournals.org Clin Cancer Res; 16(6) March 15, 2010 1793

Haynes et al.

(matched pair) sign-ranked test. Multivariate analysis was showed a significance of P < 0.10 in the univariate analysis done using linear regression with both forward and back- in the whole group of patients were considered as poten- ward selection methods; all values of biochemical para- tial predictors in multivariate analysis. To cater for the pos- meters were log transformed before analysis. Factors that sibility of unstable inclusion or exclusion of highly

Table 3. Spearman correlation of tumor estradiol levels with plasma estrogen levels and tumor gene expression

All patients Postmenopausal Postmenopausal ER+

Plasma E2 r (95% CI) 0.47 (0.11-0.72) 0.30 (−0.16 to 0.66) 0.44 (−0.082 to 0.78) P (n) 0.011 (29) 0.18 (21) 0.085 (16)

Plasma E1 r (95% CI) 0.53 (0.19-0.76) 0.51 (0.085-0.78) 0.61 (0.15-0.85) P (n) 0.0031 (29) 0.019 (21) 0.012 (16)

Plasma E1S r (95% CI) 0.28 (−0.11 to 0.59) 0.13 (−0.33 to 0.54) −0.16 (−0.62 to 0.38) P (n) 0.14 (29) 0.57 (21) 0.55 (16) ESR1 r (95% CI) 0.55 (0.24-0.75) 0.76 (0.49-0.89) 0.59 (0.13-0.84) P (n) <0.001 (34) <0.001 (23) 0.013 (17) CYP19 r (95% CI) 0.047 (−0.31 to 0.39) −0.041 (−0.46 to 0.39) −0.19 (−0.62 to 0.34) P (n) 0.79 (34) 0.85 (23) 0.47 (17) HSD17B1 r (95% CI) 0.28 (−0.074 to 0.57) 0.13 (−0.31 to 0.52) 0.071 (−0.44 to 0.54) P (n) 0.11 (34) 0.56 (23) 0.79 (17) HSD17B2 r (95% CI) −0.46 (−0.70 to −0.14) −0.65 (−0.84 to −0.31) −0.55 (−0.82 to −0.072) P (n) 0.0057 (34) <0.001 (23) 0.024 (17) HSD17B5 r (95% CI) −0.31 (−0.59 to 0.044) −0.40 (−0.71 to 0.024) −0.39 (−0.74 to 0.13) P (n) 0.076 (34) 0.057 (23) 0.13 (17) HSD17B7 r (95% CI) 0.59 (0.31-0.78) 0.55 (0.16-0.79) 0.36 (−0.17 to 0.72) P (n) <0.001 (34) 0.0071 (23) 0.16 (17) HSD17B12 r (95% CI) −0.45 (−0.69 to −0.12) −0.46 (−0.74 to −0.049) −0.38 (−0.74 to 0.14) P (n) 0.0076 (34) 0.026 (23) 0.13 (17) HSD17B14 r (95% CI) −0.12 (−0.45 to 0.24) −0.068 (−0.48 to 0.37) −0.091 (−0.56 to 0.42) P (n) 0.49 (34) 0.76 (23) 0.73 (17) STS r (95% CI) −0.19 (−0.51 to 0.17) −0.18 (−0.56 to 0.26) −0.40 (−0.75 to 0.11) P (n) 0.28 (34) 0.41 (23) 0.11 (17) SULT1E1 r (95% CI) −0.58 (−0.85 to −0.078) −0.36 (−0.81 to 0.37) — P (n) 0.024 (15) 0.31 (10) n <10 PGR r (95% CI) 0.77 (0.58-0.88) 0.80 (0.57-0.91) 0.64 (0.21-0.86) P (n) <0.001 (34) <0.001 (23) 0.006 (17) TFF1 r (95% CI) 0.46 (0.13-0.69) 0.36 (−0.077 to 0.68) −0.17 (−0.61 to 0.36) P (n) 0.0068 (34) 0.094 (23) 0.52 (17)

NOTE: Significant correlations (P < 0.05) are in boldface; n is the number of data points (for full details, please refer to “Patient demographics”).

1794 Clin Cancer Res; 16(6) March 15, 2010 Clinical Cancer Research

Intratumoral Estrogen Disposition in Breast Cancer

those in the primary analyses and therefore are not presented other than for two sets of multivariate analyses, as indicated in Results.

Results

Patient demographics Thirty-four of the 43 patients included in our previous study describing estrogen measurements (33) had suffi- cient tumor available to perform the quantitative real- time PCR assays. Of these, 23 were postmenopausal − (17 ER+ and 6 ER ) and 11 premenopausal. Further details of the patient demographics are shown in Table 1. Normal tissue RNA was not available in 3 patients (2 premenopausal and 1 postmenopausal) and plasma E2 measurements were not available in 5 patients (3 premenopausal and 2 postmenopausal). Therefore, n values varied for statistical comparisons, with all 34 samples available for the correlation of tumoral gene expression with tumor E2, 31 samples available (9 premenopausal and 22 postmenopausal) for the correlation of normal tissue gene expression with normal tissue E2,and29 samples available (8 premenopausal and 21 postmenopausal) for the correlation of plasma estrogens with normal tissue and tumor E2.

Quantitative reverse transcription-PCR All genes were detected in more than 90% of the samples by quantitative real-time PCR at reliably quantifiable levels (Ct <35) except for SULT1E1 (detected in 15 of 34 tumors; 5 of 31 normal tissues) and CYP17A1 (not detected in any of the samples). The higher sensitivity of the optimized quantitative real-time PCR assay for CYP19 allowed accurate detection at lower levels (Ct <40); using this limit, sig- nal for CYP19 RNA was observed in all samples. The mean and range of the Ct values for the genes measured are Fig. 1. Correlation of tumor gene expression and intratumoral E2 shown in Supplementary Table S1. concentration in postmenopausal patients, 17 ER+ (○) and 6 ER− (•). A, ESR1;B,HSD17B7;C,PGR. Correlation of normal breast tissue E2 with plasma estrogens and normal breast tissue gene expression Univariate analysis. Spearman correlations of normal correlated covariables, correlation matrices were devel- tissue E2 levels with normal tissue gene expression values oped. Two postmenopausal patients with ER+ tumors of ESR1, estrogen reporter genes (PGR and TFF1), and the stopped hormone replacement therapy within the 4 wk estrogen-metabolizing genes are shown for all patients and before surgery. These cases have not been excluded from for the postmenopausal subgroup separately in Table 2. the primary analyses because (a) their luteinizing hor- Premenopausal patients are not presented separately be- mone and follicle-stimulating hormone levels at surgery cause of the small number of patients (n = 9), nor are cor- showed no evidence of suppression; (b) their plasma, ma- relations with normal tissue SULT1E1 expression because it lignant, and normal tissue estrogen levels were all in the was detected in 5 of 31 cases only. middle of the range of values in the other postmenopausal Plasma E2 and E1 strongly correlated with normal tissue + cases of ER tumors; and (c) similarly, the ratios of plas- E2 in the whole group of patients (r = 0.79, P < 0.001 and ma/tumor estrogens and plasma/normal tissue estrogens, r = 0.82, P < 0.001, respectively). These correlations were which should reveal any lack of equilibration, were also in weaker but still significant in the postmenopausal sub- the middle of the range of values in other patients. This group (r = 0.48, P = 0.027 and r = 0.56, P = 0.008, respec- decision was reached without reference to the effect on tively). TFF1 was the only gene whose expression in the data analyses in relation to gene expression. Analyses normal tissue showed a significant correlation with nor- excluding these cases were not materially different from mal tissue E2 in the whole group.

www.aacrjournals.org Clin Cancer Res; 16(6) March 15, 2010 1795

Haynes et al.

Multivariate analysis. Parameters that showed a signifi- nonsignificant trend with an r value similar to that seen in cance of P < 0.10 in the univariate analysis were selected the whole group of patients (r = 0.44, P = 0.085). A stronger for the inclusion in the multivariate analysis (plasma E1, correlation was observed between plasma E1 and tumor E2 plasma E2, plasma E1S, and normal breast tissue E1) with such that significant correlations were observed in both the the exclusion of the estrogen-dependent genes (PGR and whole group of patients (r = 0.53, P = 0.0031) and in the TFF1). postmenopausal subgroups (all, r = 0.51, P = 0.019; ER+, In the whole group of patients (n = 29), forward and r = 0.61, P = 0.012). Plasma E1S did not show any signifi- backward selections of parameters gave a multivariate cant correlation with tumor E2 in any group or subgroup of model of plasma E2 as the main factor (r =0.89,P < patients. 0.001), with a small contribution from normal tissue E1 Tumor E2 levels showed a strong positive correlation (P = 0.007) as determinants of normal tissue E2 levels (cu- with ESR1 expression in all groups of patients, with the mulative r = 0.92 with both estrogens). In the postmeno- closest association observed in postmenopausal patients pausal group of patients, normal tissue E1 was the only (r =0.76,P < 0.001; Fig. 1A). This was in part due to a + − significant predictor of tissue E2 levels (r = 0.57, P = distinct difference between ER and ER patients, but a 0.007). If normal tissue E1 was omitted from this analysis, strong relationship (r = 0.59, P = 0.013) was also present + plasma E1 (r = 0.56, P = 0.009) and plasma E2 (r = 0.54, in the ER subgroup. A significant correlation was also ap- P = 0.011) were significant factors (by forward and back- parent in the small premenopausal subgroup of patients ward selections, respectively). This statistical interchange (r = 0.76, P = 0.0073; n = 11). between plasma E1 and E2 is probably caused by the very Considering the gene expression of the estrogen-associ- close correlation of these two factors (r = 0.81, P < 0.001). ated enzymes in the whole group of patients, HSD17B7 showed a significant positive correlation (Fig. 1B) and HSD17B2 HSD17B12 Correlation of breast tumor E2 with plasma estrogens and showed significant negative cor- and tumor gene expression relations with tumor E2 (Table 3). HSD17B1 expression Univariate analysis. Spearman correlations of tumor E2 did not correlate with intratumoral estrogen levels in any levels with gene expression values of ESR1, the estrogen- group of patients, and this was confirmed by repeating the metabolizing genes, and estrogen reporter genes (PGR quantitative PCR assay with a different set of primers/ and TFF1) are shown for all patients and for the postmen- probe for the HSD17B1 gene (data not shown). Gene ex- opausal and postmenopausal ER+ subgroups separately in pression levels of CYP19 and STS did not significantly cor- Table 3. Correlations with tumor expression of SULT1E1 relate with tumor E2 in any group or subgroup of patients, in the postmenopausal ER+ subgroup are not presented except in the small premenopausal subgroup where a sig- because it was detected in 6 of 17 cases only. nificant correlation was observed between tumor E2 and CYP19 r P n Plasma E2 correlated with tumor E2 in the whole group ( = 0.63, = 0.039; = 11). of patients (Spearman r = 0.47, P = 0.011). There was no Expression of the ER-dependent genes (PGR and TFF1) significant correlation in the postmenopausal group over- correlated with tumor E2 levels, with a particularly high all, but in the postmenopausal ER+ subgroup, there was a correlation with PGR expression (r = 0.80, P <0.001in

Table 4. Comparison of gene expression in matched breast tumor and normal breast tissue in postmenopausal ER+ patients (n = 17)

Gene Tumor/normal ratio, mean (range) Proportion with increased tumor expression (%)

ESR1 9.1 (1.8-31.1)* 17/17 (100) CYP19 1.4 (0.04-7.5) 10/17 (59) HSD17B1 0.58 (0.05-9.5) 7/17 (41) HSD17B2 2.9 (0.07-898) 10/17 (59) HSD17B5 0.23 (0.05-2.8)* 1/17 (6) HSD17B7 3.5 (1.5-12.9)* 17/17 (100) HSD17B12 1.0 (0.39-1.8) 11/17 (65) HSD17B14 2.1 (0.25-9.8)† 15/17 (88) STS 0.34 (0.06-1.2)* 3/17 (18) PGR 2.3 (0.14-22.7)† 14/17 (82) TFF1 156 (0.002-46375)* 16/17 (94)

NOTE: For significant genes, values are given in boldface. *P < 0.001 (Wilcoxon ranked test). †P < 0.01 (Wilcoxon ranked test).

1796 Clin Cancer Res; 16(6) March 15, 2010 Clinical Cancer Research

Intratumoral Estrogen Disposition in Breast Cancer

Fig. 2. Comparison of gene expression in breast tumors and matched normal breast tissues in postmenopausal patients, 17 ER+ (○) and 5 ER− (•). Wilcoxon ranked-test P values are quoted for the difference in gene expression between normal and tumor tissues in the ER+ subgroup. the postmenopausal subgroup; Fig. 1C). A very strong cor- HSD17B12 (P = 0.0022) replaced HSD17B2 in this mul- relation between ESR1 and PGR expression was observed tivariate model. Backward selection of the parameters in all groups of patients (r = 0.72-0.88, P < 0.001). produced an alternate multivariate model of ESR1 and A Mann-Whitney test was done to compare intratumoral plasma E2 (both P < 0.001, r = 0.80). In the postmen- estrogen levels between tumors in which SULT1E1 was ex- opausal subgroup, ESR1 was the only significant vari- pressed (15 of 34 total patients and 10 of 23 postmenopaus- able (P <0.001,r =0.85),whereasinthe al patients) and those in which it could not be detected. In postmenopausal ER+ subgroup, ESR1 was again the tumors with detectable SULT1E1, tumor E1 and E2 were sig- main determinant (P = 0.013, r =0.56)withplasma nificantly lower in the whole group of patients (2.5- and 4.5- E2 also contributing (P = 0.025, cumulative r = 0.73). fold, respectively; both P < 0.01) and in the postmenopausal If the two patients who stopped hormone replacement subgroup [2.5-fold (P < 0.01) and 3.6-fold (P < 0.05), re- therapy within the 4 weeks before surgery were excluded spectively] compared with tumors with no detectable from this analysis, HSD17B7 (P = 0.037, r = 0.56) was + SULT1E1. In contrast, tumoral E1S levels were higher in the only significant variable in the postmenopausal ER tumors expressing SULT1E1 than in those that did not, multivariate model. although this was not statistically significant [2.9-fold Correlation tables of the parameters included in multi- (P = 0.19) and 3.6-fold (P = 0.057) in the whole group of variate analysis with each other are shown in Supplementa- patients and the postmenopausal subgroup, respectively]. ry Tables S2 to S4. Of note, a strong negative correlation of Multivariate analysis. Factors that showed a significance ESR1 and HSD17B2 was observed in postmenopausal pa- of P < 0.10 in the univariate analysis (ESR1, HSD17B2, tients (r = −0.63, P = 0.0012); HSD17B2 and HSD17B12 r P HSD17B5, HSD17B7, HSD17B12, plasma E1, and plasma showed a significant positive correlation ( = 0.49, = E2) were selected for inclusion in the multivariate analysis 0.0032) in the whole group; and a positive correlation of ESR1 HSD17B7 for association with intratumoral E2 levels, with the exclu- and was seen in all groups and subgroups sion of the estrogen-dependent genes (PGR and TFF1). of patients (r = 0.54, P = 0.0011 for the whole group). In the whole group of patients, forward selection of parameters gave a multivariate model of HSD17B7 as Comparison of gene expression in breast tumor and a positive determinant (P < 0.001, r =0.61)and normal breast tissue HSD17B2 (P = 0.0043) as a negative determinant of Comparison of gene expression levels between breast tumor E2 (cumulative r =0.74).Ifthetwopatients tumors and matched normal breast tissues in postmeno- who stopped hormone replacement therapy within the pausal patients showed major differences (Table 4; 4 weeks before surgery were excluded from this analysis, Fig. 2). ESR1 expression was much higher in all 17 ER+

www.aacrjournals.org Clin Cancer Res; 16(6) March 15, 2010 1797

Haynes et al.

tumors than in matched normal tissues (mean of 9.1-fold, biological activity of the proteins involved, which may be − P < 0.001). In contrast, the 5 postmenopausal ER tumors altered by many factors including posttranslational effects all had much lower ESR1 expression than normal tissues and the availability of cofactors. Nonetheless, a number of (mean of 7.7-fold). The expression of HSD17B7 was high- strong messages emerged. er in all 17 ER+ tumors (3.5-fold; P < 0.001) and in 4 of 5 Univariate analysis revealed significant correlations in − ER tumors compared with matched normal tissues. both normal and malignant tissues between E2 concentra- HSD17B14 expression was upregulated in 15 of 17 ER+ tions and the expression of the estrogen-dependent gene − tumors (P <0.01)butnotinER tumors (upregulated TFF1, supporting the biological significance of tissue E2 in 2 of 5) compared with normal tissues. There was also levels. The only other significant correlates with normal + an indication that SULT1E1 might be upregulated in ER tissue E2 were plasma estrogens, suggesting that equilibra- tumors because it was detected in 6 of 17 tumors versus 0 tion with circulating estrogens, with little intratissular con- of 17 normal tissue samples. In contrast, the expression of version, may be the major determinant of normal tissue E2 STS and HSD17B5 was significantly lower in ER+ tumors levels. This contrasts markedly with the relationships with compared with normal tissues (2.9- and 4.3-fold, respec- E2 in malignant tissue where, in addition to plasma E2 and tively; both P < 0.001). No significant difference was ob- E1, multiple other factors including ESR1, HSD17B2, served in the expression of CYP19, HSD17B1, HSD17B2, HSD17B7,andHSD17B12 showed significant associa- or HSD17B12 between ER+ tumors and matched normal tions. In contrast, no correlation was seen between intra- tissues. Of note, the comparative expression of HSD17B2 tumoral E2 and CYP19, STS,orHSD17B1, three genes that in tumor and normal tissue was particularly variable, in some reports (7–9) have been thought to be the most with 6 tumors having >5-fold lower expression and 6 tu- closely involved with regulation of E2 levels within breast mors having >25-fold higher expression than found in tumors, providing no evidence to support a role for these matched normal tissues. The estrogen reporter genes enzymes in the establishment of intratumoral E2 levels in PGR and, to a much greater extent, TFF1 both showed these patients. A weak correlation was found between in- + − higher expression in ER tumors than in ER tumors tratumoral E2 levels and CYP19 expression in the 11 pre- and normal tissues. menopausal patients, but this probably occurred by chance because the P value (0.039) is modest, unadjusted Discussion for multiple comparisons, and based on a limited number of observations. Moreover, it does not seem to be a tena- The contribution from local estrogen synthesis versus ble finding given that the relative contribution of intratu- uptake of estrogens from the circulation to intratumoral moralsynthesiswouldbeexpectedtobehigherin estrogen levels remains controversial. Clarification of this postmenopausal women in whom no significant associa- is important in relation to the rationale for the develop- tion was found despite the higher number of samples. ment of tumor-specific aromatase inhibitors (36) and in- Many of the significant correlates of intratumoral E2 in hibitors of other enzymes involved in estrogen disposition the univariate analysis showed overlapping confidence in- (37–39). tervals such that the power of the multivariate analysis to Studying a larger sample set, from which all materials in accurately describe the independent predictors of intratu- this study were drawn, some of us have recently reported moral E2 was diminished. The strong correlation seen be- + intratumoral E2 levels to be 4.1- to 8.6-fold higher in ER tween some of the factors (e.g., ESR1 and HSD17B7, ESR1 tumors than in paired normal breast tissues, as well as sig- and HSD17B2) contributes to making the individual sig- − nificantly higher compared with ER tumors (33). In con- nificance of these factors uncertain in a multivariate mod- trast, E1 levels were reduced. In a separate study, some of el. Nevertheless, a consistent theme emerged where ESR1 us found a strong correlation between plasma E2 levels was found to be the single most important correlate of in- and the expression of four classic estrogen-dependent tratumoral E2, predicting between 50% and 70% of intra- + genes in ER tumors, and this correlation was enhanced tumoral E2 variability in all groups and subgroups of when ESR1 expression was considered as a possible co- patients. In the whole group of patients, a contribution determinant with plasma E2 (40). This suggested that to intratumoral E2 variability was also apparent from ER-dependent uptake might be more important than in- HSD17B7 and HSD17B2. tratumoral estrogen synthesis as a regulator of estrogen- These data, together with our recently published data dependent gene transcription. The current study aimed to (33, 40), are supportive of a model in which plasma E2 is examine the relationship between intratumoral E2 levels the major contributor to intratumoral E2 levels and is se- and the expression of genes for enzymes involved in estro- questered by ERα to establish the high tumor-to-plasma ra- gen synthesis and metabolism and/or ER expression to in- tio seen particularly in postmenopausal women. In tegrate these and our earlier findings into a single model addition, local interconversion of estrogens by the explaining estrogen disposition in breast tumors. A strength 17βHSD enzymes, particularly HSD17B7 and HSD17B2, of the study was the application of highly sensitive well- may play a secondary, albeit significant, role in the balance validated estrogen assays (34, 35). Our other assessments between E1 and E2 estrogens with low and high biological were based on mRNA levels, rather than protein level or activity, respectively. Although clearly supportive of this enzyme activity, and are thus an indirect indication of the model, it must be noted that our observations are based

1798 Clin Cancer Res; 16(6) March 15, 2010 Clinical Cancer Research

Intratumoral Estrogen Disposition in Breast Cancer

on correlations involving a relatively small number of pa- data indicate that the enhanced conversion of E1 to E2 may tients. Thus, other possibilities should be considered. For be mediated by the increased expression of HSD17B7. example, although we have no evidence for this, higher le- Changes in the balance of the estrogen sulfatase/sulfo- vels of androgens in tumors might also lead, by aromatiza- transferase reaction may also contribute to changes in the tion, to higher E2 levels despite there being no difference in E1/E2 balance, as we found STS expression to be ∼3-fold aromatase levels themselves. lower in breast tumors compared with normal tissues and Most studies on the 17βHSDs have primarily assessed there was an indication that sulfotransferase activity type 1 and type 2 as the mediators of intratumoral intercon- (SULT1E1 expression) might be upregulated in tumors. version of E1 and E2 and have focused on associations with We also observed that tumor E1 and E2 are significantly low- clinical outcome rather than intratumoral estrogen levels er in tumors expressing SULT1E1 than in those that do not; (9, 41). HSD17B12 has been suggested to be a more impor- one could postulate that this is due to an increased conver- tant source of intratumoral E2 than either HSD17B1 or sion of E1 to E1S in these tumors, reducing the amount of E1 HSD17B7 due to its higher expression levels in the breast available to be converted to E2. The data are in accord with (23). In contrast, in this study it seemed that HSD17B7 the observation in this same series that E1S levels were sig- (in a positive direction) and HSD17B2 expression (in a neg- nificantly higher (2-fold) in tumor than in normal tissue ative direction) were the main contributors to intratumoral andthattheratioofE1/E1Swas∼5-fold lower in tumor estrogen disposition. This is consistent with our preliminary than normal tissue (33). These findings are in contrast to report of a correlation between intratumoral HSD17B7 a previous study in which sulfatase mRNA levels were found expression and time-to-treatment failure on an aromatase to be significantly higher in malignant breast tissue than in inhibitor in 58 patients with advanced breast cancer (28). normal tissue; however, this was in a mixed patient group Few other studies have directly compared gene expres- undefined in terms of menopausal or ER status (18). A sion in matched normal and breast tumor samples, par- more recent larger study, in a primarily postmenopausal ticularly in the subset of postmenopausal ER+ patients. ER+ population, found no difference in STS expression We observed a nearly 10-fold higher expression of ESR1 between tumors and adjacent noncancerous tissues (44). in postmenopausal ER+ tumors compared with normal tis- In summary, our data favor a model in which uptake of sues. In agreement with this, an increase in ERα mRNA estrogens, particularly E2, due to binding to the ER rather levels in breast cancer versus matched nonmalignant tissue than intratumoral estrogen synthesis by aromatase or sul- in a mixed patient group has been reported previously fatase, is the single most important correlate and a proba- (42). In addition, higher immunohistochemical expression ble determinant of intratumoral E2. Thus, intratumoral E2 of ERα was observed in ER+ carcinomas than in nonmalig- seems to be derived predominantly by uptake from the nant matched tissues in two studies (43, 44). Increased ex- plasma or surrounding tissues, whereas the 17βHSD en- pression of TFF1 and PGR was seen in ER+ tumors, but not zymes (HSD17B2 and HSD17B7) play a role in the main- − in ER tumors, in comparison with paired normal tissues. tenance of the ambient intratumoral E2 levels. The data This is consistent with TFF1 and PGR being known to be suggest that the dominant target for estrogen deprivation regulated by estrogen and the presence of higher levels of should be the systemic, and not the intratumoral, produc- ESR1 in the malignant tissue than in adjacent normal tis- tion of estrogens. Inhibition of HSD17B7 might provide a sue. We did not see any significant differences in the expres- supplementary target but further data on this are needed. sion of CYP19, HSD17B1, and HSD17B2 between normal breast and malignant tissues, in agreement with the results Disclosure of Potential Conflicts of Interest of other studies of more heterogeneous patient groups (42, 44, 45). The consistent finding that CYP19 does not show No potential conflicts of interest were disclosed. increased expression in tumor compared with normal tissue may be a reflection of its very low activity within the Acknowledgments breast, which makes accurate measurement of CYP19 We thank Dr. Turid Aas at Department of Surgery, Haukeland University mRNA, protein, or enzyme activity difficult. Hospital, for recruiting the patients for this study. Of particular interest is our novel finding of 3.5-fold in- HSD17B7 creased expression of within tumors compared Grant Support with normal tissues in postmenopausal women. This may help explain the significantly greater ratio (8- to 19-fold) The Breast Cancer Research Foundation (I.E. Smith and M. Dowsett); the Norwegian Cancer Society, the “Rosa sløyfe” Breast Cancer Fund Raising, of E2/E1 in breast tumor compared with normal breast tis- HSD17B5 the Norwegian Health Region West (HelseVest), and the Innovest program sue (33). In contrast, , which can also catalyze of Excellence (all P.E. Lønning); and National Health Service funding to the the conversion of E1 to E2, was found to have lower ex- Royal Marsden National Institute for Health Research Biomedical Research pression in tumor tissue. This may reflect the greater im- Centre (M. Dowsett). The costs of publication of this article were defrayed in part by the portance of this enzyme, also known as aldo-keto payment of page charges. This article must therefore be hereby marked reductase 1C3, in androgen metabolism (21). Peripheral advertisement in accordance with 18 U.S.C. Section 1734 solely to metabolism of estrogens normally strongly favors the in- indicate this fact. activation of E2, but in women with breast cancer, E1 can Received 09/11/2009; revised 01/07/2010; accepted 01/09/2010; be preferentially converted to E2 in tumor tissue (27). Our published OnlineFirst 03/09/2010.

www.aacrjournals.org Clin Cancer Res; 16(6) March 15, 2010 1799

Haynes et al.

References 1. Key T, Appleby P, Barnes I, Reeves G, Endogenous Hormones and oid dehydrogenase isoforms (AKR1C1-1C4) of the aldo-keto reduc- Breast Cancer Collaborative Group. Endogenous sex hormones and tase superfamily: functional plasticity and tissue distribution reveals breast cancer in postmenopausal women: reanalysis of nine pro- roles in the inactivation and formation of male and female sex hor- spective studies. J Natl Cancer Inst 2002;94:606–16. mones. Biochem J 2000;351:67–77. 2. Geisler J, Lønning PE. Aromatase inhibition: translation into a suc- 22. Krazeisen A, Breitling R, Imai K, Fritz S, Möller G, Adamski J. Deter- cessful therapeutic approach. Clin Cancer Res 2005;11:2809–21. mination of cDNA, gene structure and chromosomal localization of 3. Early Breast Cancer Trialists' Collaborative Group (EBCTCG). Effects the novel human 17β-hydroxysteroid dehydrogenase type 7. FEBS of chemotherapy and hormonal therapy for early breast cancer on Lett 1999;460:373–9. recurrence and 15-year survival: an overview of the randomised 23. Luu-The V, Tremblay P, Labrie F. Characterization of type 12 17β- trials. Lancet 2005;365:1687–717. hydroxysteroid dehydrogenase, an isoform of type 3 17β-hydroxys- 4. Lønning PE, Dowsett M, Powles TJ. Postmenopausal estrogen syn- teroid dehydrogenase responsible for estradiol formation in women. thesis and metabolism: alterations caused by aromatase inhibitors Mol Endocrinol 2006;20:437–43. used for the treatment of breast cancer. J Steroid Biochem 1990; 24. Wu L, Einstein M, Geissler WM, Chan HK, Elliston KO, Andersson S. 35:355–66. Expression cloning and characterization of human 17β-hydroxyster- 5. Smith IE, Dowsett M. Aromatase inhibitors in breast cancer. N Engl J oid dehydrogenase type 2, a microsomal enzyme possessing 20 α- Med 2003;348:2431–42. hydroxysteroid dehydrogenase activity. J Biol Chem 1993;268: 6. Van Landeghem AAJ, Poortman J, Nabuurs M, Thijssen JHH. En- 12964–9. dogenous concentration and subcellular distribution of estrogens 25. He XY, Wegiel J, Yang YZ, Pullarkat R, Schulz H, Yang SY. Type 10 in normal and malignant breast tissue. Cancer Res 1985;45:2900–4. 17β-hydroxysteroid dehydrogenase catalyzing the oxidation of ste- 7. Miller WR. Importance of intratumour aromatase, and its susceptibil- roid modulators of γ-aminobutyric acid type A receptors. Mol Cell ity to inhibitors. In: Dowsett M, editor. Aromatase inhibition—then, Endocrinol 2005;229:111–7. now and tomorrow. London: Parthenon Publishing Group; 1994, p. 26. Lukacik P, Keller B, Bunkoczi G, et al. Structural and biochemical 43–53. characterization of human orphan DHRS10 reveals a novel cytosolic 8. Santner SJ, Feil PD, Santen RJ. In situ estrogen production via the enzyme with steroid dehydrogenase activity. Biochem J 2007;402: estrone sulfatase pathway in breast tumors: relative importance 419–27. versus the aromatase pathway. J Clin Endocrinol Metab 1984;59: 27. McNeill JM, Reed MJ, Beranek PA, et al. A comparison of the in vivo 29–33. uptake and metabolism of 3H-oestrone and 3H-oestradiol by normal 9. Gunnarsson C, Hellqvist E, Stål O. 17β-Hydroxysteroid dehydro- breast and breast tumour tissue in postmenopausal women. Int J genases involved in local oestrogen synthesis have prognostic sig- Cancer 1986;38:205–12. nificance in breast cancer. Br J Cancer 2005;92:547–52. 28. Haynes B, Ghazoui Z, Anderson H, et al. Identification of molecular 10. Masamura S, Santner SJ, Gimotty P, George J, Santen RJ. Mecha- predictors of response of advanced breast cancer patients to aroma- nism for maintenance of high breast tumor estradiol concentrations tase inhibition. Breast Cancer Res Treat 2008;106:S93. in the absence of ovarian function: role of very high affinity tissue 29. Blankenstein MA, Maitimu-Smeele I, Donker GH, Daroszewski J, uptake. Breast Cancer Res Treat 1997;42:215–26. Milewicz A, Thijssen JHH. On the significance of in situ production 11. James VHT, Reed MJ, Adams EF, et al. Oestrogen uptake and me- of oestrogens in human breast cancer tissue. J Steroid Biochem Mol tabolism in vivo. In: Beck JS, editor. Oestrogens and the human Biol 1992;41:891–6. breast. Proc Roy Soc Edinb 1989;95B:185–93. 30. Chetrite GS, Cortes-Prieto J, Philippe JC, Wright F, Pasqualini JR. 12. O'Neill JS, Elton RA, Miller WR. Aromatase activity in adipose tissue Comparison of estrogen concentrations, estrone sulfatase and aro- from breast quadrants: a link with tumour site. Br Med J 1988;296: matase activities in normal, and in cancerous, human breast tissues. 741–3. J Steroid Biochem Mol Biol 2000;2:23–7. 13. Bulun SE, Price TM, Aitken J, Mahendroo MS, Simpson ER. A link 31. Geisler J, Detre S, Berntsen H, et al. Influence of neoadjuvant ana- between breast cancer and local estrogen biosynthesis suggested strozole (Arimidex) on intratumoral estrogen levels and proliferation by quantification of breast adipose tissue aromatase cytochrome markers in patients with locally advanced breast cancer. Clin Cancer P450 transcripts using competitive polymerase chain reaction after Res 2001;7:1230–6. reverse transcription. J Clin Endocrinol Metab 1993;77:1622–8. 32. Geisler J, Helle H, Ekse D, et al. Letrozole is superior to anastrozole 14. Miller WR, Anderson TJ, Evans DB, Krause A, Hampton G, Dixon JM. in suppressing breast cancer tissue and plasma estrogen levels. Clin An integrated view of aromatase and its inhibition. J Steroid Biochem Cancer Res 2008;14:6330–5. Mol Biol 2003;86:413–21. 33. Lonning PE, Helle H, Duong NK, Ekse D, Aas T, Geisler J. Tissue 15. Ellis MJ, Miller WR, Tao Y, et al. Aromatase expression and out- estradiol is selectively elevated in receptor positive breast cancers comes in the P024 neoadjuvant endocrine therapy trial. Breast Can- while tumour estrone is reduced independent of receptor status. cer Res Treat 2009;116:371–8. J Steroid Biochem Mol Biol 2009;117:31–41. 16. Reed MJ, Purohit A. Sulphatase inhibitors: the rationale for the de- 34. Geisler J, Berntsen H, Lønning P. A novel HPLC-RIA method for the velopment of a new endocrine therapy. Rev Endocr Rel Cancer 1993; simultaneous detection of estrone, estradiol and estrone sulphate le- 45:51–62. vels in breast cancer tissue. J Steroid Biochem Molec Biol 2000;72: 17. Lønning PE, Johannessen DC, Thorsen T. Alterations in the produc- 259–64. tion rate and the metabolism of oestrone and oestrone sulphate in 35. Geisler J, Ekse D, Helle H, Duong NK, Lønning PE. An optimised, breast cancer patients treated with aminoglutethimide. Br J Cancer highly sensitive radioimmunoassay for the simultaneous measure- 1989;60:107–11. ment of estrone, estradiol and estrone sulfate in the ultra-low range 18. Utsumi T, Yoshimura N, Takeuchi S, Maruta M, Maeda K, Harada NJ. in human plasma samples. J Steroid Biochem Mol Biol 2008;109: Elevated steroid sulfatase expression in breast cancers. Steroid Bio- 90–5. chem Mol Biol 2000;73:141–5. 36. Simpson ER. Sources of estrogen and their importance. J Steroid 19. Miyoshi Y, Ando A, Hasegawa S, et al. High expression of steroid Biochem Mol Biol 2003;86:225–30. sulfatase mRNA predicts poor prognosis in patients with estrogen 37. Kruchten P, Werth R, Marchais-Oberwinkler S, Frotscher M, Hartmann receptor-positive breast cancer. Clin Cancer Res 2003;9:2288–93. RW. Development of a biological screening system for the evaluation 20. Poutanen M, Miettinen M, Vihko R. Differential estrogen substrate of highly active and selective 17β-HSD1-inhibitors as potential thera- specificities for transiently expressed human placental 17β-hydro- peutic agents. Mol Cell Endocrinol 2009;301:154–7. xysteroid dehydrogenase and an endogenous enzyme expressed 38. Lawrence HR, Vicker N, Allan GM, et al. Novel and potent 17β-hydro- in cultured COS-m6 cells. Endocrinology 1993;133:2639–44. xysteroid dehydrogenase type 1 inhibitors. J Med Chem 2005;48: 21. Penning TM, Burczynski ME, Jez JM, et al. Human 3α-hydroxyster- 2759–62.

1800 Clin Cancer Res; 16(6) March 15, 2010 Clinical Cancer Research

Intratumoral Estrogen Disposition in Breast Cancer

39. Day JM, Purohit A, Tutill HJ, et al. The development of steroid sulfa- meters in breast cancer patients. J Steroid Biochem Mol Biol 2009; tase inhibitors for hormone-dependent cancer therapy. Ann N Y 113:195–201. Acad Sci 2009;1155:80–7. 43. Han B, Li S, Song D, et al. Expression of 17β-hydroxysteroid dehy- 40. Dunbier AK, Anderson H, Ghazoui Z, et al. Relationship between drogenase type 2 and type 5 in breast cancer and adjacent non- plasma estradiol levels and estrogen-responsive gene expression malignant tissue: a correlation to clinicopathological parameters. in estrogen receptor-positive breast cancer in postmenopausal J Steroid Biochem Mol Biol 2008;112:194–200. women. J Clin Oncol Epub 2010 Feb 1. PubMed PMID: 20124184. 44. Honma N, Takubo K, Sawabe M, et al. Estrogen-metabolizing en- 41. Jansson AK, Gunnarsson C, Cohen M, Sivik T, Stål O. 17β-hydroxys- zymes in breast cancers from women over the age of 80 years. J Clin teroid dehydrogenase 14 affects estradiol levels in breast cancer Endocrinol Metab 2006;91:607–13. cells and is a prognostic marker in estrogen receptor-positive breast 45. Chong YM, Colston K, Jiang WG, Sharma AK, Mokbel K. The relation- cancer. Cancer Res 2006;66:11471–7. ship between the insulin-like growth factor-1 system and the oestro- 42. Suzuki M, Ishida H, Shiotsu Y, et al. Expression level of enzymes gen metabolizing enzymes in breast cancer tissue and its adjacent related to in situ estrogen synthesis and clinicopathological para- non-cancerous tissue. Breast Cancer Res Treat 2006;99:275–88.

www.aacrjournals.org Clin Cancer Res; 16(6) March 15, 2010 1801

Intratumoral Estrogen Disposition in Breast Cancer

Ben P. Haynes, Anne Hege Straume, Jürgen Geisler, et al.

Clin Cancer Res Published OnlineFirst March 9, 2010.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-09-2481

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2010/03/16/1078-0432.CCR-09-2481.DC1

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

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

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2010/03/07/1078-0432.CCR-09-2481. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.