Subnuclear Distribution of Progesterone Receptors A and B in Normal and Malignant Endometrium

Transcription

1 X/04/$15.00/0 The Journal of Clinical Endocrinology & Metabolism 89(3): Printed in U.S.A. Copyright 2004 by The Endocrine Society doi: /jc Subnuclear Distribution of Progesterone Receptors A and B in Normal and Malignant Endometrium REBECCA L. ARNETT-MANSFIELD, ANNA DEFAZIO, PATRICIA A. MOTE, AND CHRISTINE L. CLARKE Westmead Institute for Cancer Research (R.L.A.-M., A.deF., P.A.M., C.L.C.), University of Sydney at the Westmead Millennium Institute; and Department of Gynaecological Oncology (A.deF.), Westmead Hospital, Westmead, New South Wales 2145, Australia The nuclear progesterone receptor (PR), which is expressed as two proteins with different functions (PRA and PRB), is expressed in the normal endometrium and endometrial cancer. Our previous work has shown that there is a disruption to the relative PR isoform expression in progression to malignancy in endometrial cancer in which expression of a single isoform is common. A consistent feature in these studies was discrete punctate distribution of PRA and PRB in the nuclei of endometrial cancer cells. In this study PRA and PRB distribution within the nucleus was examined in vivo in the normal endometrium during the menstrual cycle, and in endometrial cancer, by dual immunofluorescence and confocal microscopy using cohorts in which PRA and PRB expression levels have previously been characterized. In the normal endometrium, PR was distributed evenly within the nucleus and also localized in discrete subnuclear foci. In the proliferative phase, even PR distribution was predominant and both PR isoforms were colocated and distributed evenly. In the secretory phase, there was a marked increase in the proportion of nuclei containing PR distributed into discrete foci, and PRB was the predominant isoform in nuclear foci. There was an inverse relationship between even and focal PR distribution in the menstrual cycle, suggesting that hormonal fluctuations THE OVARIAN HORMONES, estrogen and progesterone, play crucial roles in controlling the normal function of the endometrium. During the normal menstrual cycle, there is proliferation, differentiation and degeneration of the endometrium in response to the fluctuations of steroid hormone concentrations. In the proliferative phase, estrogens stimulate proliferation of the epithelial and stromal components of the endometrium, whereas in the secretory phase, progesterone is involved in glandular differentiation and glycogenesis as well as inhibition of estrogen-mediated proliferation (1). Endometrial carcinomas are the most common malignancy of the female genital tract and the third most common cancer in Western women. They make up 97% of all uterine Abbreviations: AR, Androgen receptor; CI, confidence interval; ER, estrogen receptor; FIGO, Fédération Internationale de Gynécologie et d Obstétrique; FITC, fluorescein isothiocyanate; GFP, green fluorescent protein; GR, glucocorticoid receptor; HRT, hormone replacement therapy; MR, mineralocorticoid receptor; OR, odds ratio; PR, progesterone receptor; RNA pol II, RNA polymerase II; SRC, steroid receptor coactivator; TXR, Texas red. JCEM is published monthly by The Endocrine Society ( endo-society.org), the foremost professional society serving the endocrine community. were involved in movement of PR into focal nuclear locations. In endometrial cancers colocalization of PRA and PRB was infrequent, and there was no relationship between even and focal PR isoform distribution, unlike the normal endometrium. PRA was predominantly evenly distributed in endometrial cancers, whereas PRB was focal. Even PRB distribution in endometrial cancer was not often noted. Multivariate analysis showed that PRA expression was highly predictive of even nuclear distribution in endometrial cancers and PRB expression of distribution into foci, and these associations were independent of total PRA and PRB levels. Nuclear distribution of PR isoforms was associated with clinical grade, where tumors of high grade had significantly fewer nuclei containing even PRA distribution and focal PRB distribution, compared with tumors of low grade. In the normal endometrium, localization of PR into nuclear foci coincides with high progesterone levels, suggesting that altered intranuclear PR distribution is hormonally regulated. Nuclear distribution may be an important component of gene regulation in target tissues, and disruptions in PR distribution in endometrial cancer could affect the function of PR and contribute to aberrant hormonal responses. (J Clin Endocrinol Metab 89: , 2004) cancers and arise from the glands within the endometrium (2). As well as having roles in the normal endometrium, estrogen and progesterone are also implicated in endometrial cancer. Estrogen exposure, when in excess and in the absence of progesterone influence, causes continued stimulation of the endometrium and is strongly associated with increased endometrial carcinoma risk (2 6). Progesterone exposure attenuates endometrial cancer risk, and this is attributed to interruption of continued estrogenic stimulation of the endometrium. Endometrial hyperplasia, caused by unopposed estrogenic stimulation, can be reversed by treatment with progestins (7, 8). Progestin treatment also provides protection against the stimulatory effects of estrogenic treatments: hormone replacement therapy (HRT) using combinations of estrogens and progestins yields a lower risk of endometrial carcinoma than that associated with estrogen alone (9). Progestational agents are the most common type of hormone treatment for endometrial hyperplasia and endometrial cancer (2, 8, 10). Progesterone action is mediated in target tissues by a specific nuclear receptor, the progesterone receptor (PR). The expression of PR in endometrial glands is under the control of estrogen and progesterone, where estrogen induces PR 1429

2 1430 J Clin Endocrinol Metab, March 2004, 89(3): Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms synthesis and progesterone down-regulates the expression of its own receptor (1). The PR is derived from a single gene encoding two proteins (11) termed PRB and PRA, which are structurally different in that PRA is a truncated form of PRB, lacking 164 amino acids at the N terminus. In vitro evidence to date suggests that the PR isoforms have different abilities to activate progestin-regulated promoters and that their activities may change depending on the target cell and gene promoter (12). Our previous work has shown that PRA and PRB are coexpressed in the nucleus of target cells of the human endometrium (13). PRA and PRB are also coexpressed in cell nuclei of endometrial carcinomas; however, expression of only one isoform, either PRA or PRB, is common (14). Because PRA and PRB are proposed to have different functions, and because it has been established that PRA and PRB exist at different relative levels in various tissue types (13, 15 18), the ratio of PRA/PRB is likely to affect tissue responsiveness to progesterone. There is emerging evidence of discrete subnuclear locations of nuclear receptors, such as estrogen receptor (ER) (19, 20), androgen receptor (AR) (21), glucocorticoid receptor (GR) (22, 23), and mineralocorticoid receptor (MR) (24). The significance of discrete nuclear localization of these receptors is unknown, but it is now evident that the nucleus is a highly organized structure containing numerous specialized subnuclear structures and many nuclear components localized in discrete domains. Ribosomal RNA genes and other components involved in ribosomal biogenesis are localized in the nucleolus (25, 26); transcription is localized at the surface of euchromatin territories (27, 28); and there is concentration of small nuclear ribonucleoproteins (snrnps) and other factors involved in splicing in discrete nuclear domains (interchromatin granules or speckles) (28, 29). Distinct clusters of replication sites can be seen during S-phase (30 33). Moreover, an increasing number of nuclear bodies with unknown function, such as coiled bodies, gems and promyelocytic leukemia (PML) bodies, also show discrete localization in the nucleus (27, 29, 34, 35). These nuclear bodies are thought to either increase efficiency of transcription and RNA processing by bringing together molecules of related function, or, alternatively, they are proposed to be storage regions for inactive factors that function at sites of lower concentration (35). Various proteins have been localized in clusters, discrete domains or foci, some forming components with known nuclear bodies, such as p53 (36), and the steroid receptor coactivator (SRC), and GR interacting protein-1 (37), which localize in the PML nuclear body. In previous work examining PR isoform expression in endometrial carcinoma (14), a consistent feature we observed was a discrete punctate distribution of PRA and PRB in the nuclei of endometrioid endometrial cancer cells. Specific subnuclear localization of PRA and PRB may contribute to the different transcriptional activities of the two PR isoforms. The aim of this study was to explore the subnuclear distribution of PRA and PRB in human endometrial tissue by determining PRA and PRB distribution in the normal endometrium during the menstrual cycle, and in endometrial cancer, using cohorts in which PRA and PRB expression levels have previously been characterized (13, 14). Human tissues Materials and Methods All archival, formalin-fixed, and paraffin-embedded tissues were acquired from the Department of Tissue Pathology at Westmead Hospital. Patients with normal endometrial tissue (n 23) had undergone hysterectomy or endometrial biopsy at Westmead Hospital with all endometrial samples being reported as being morphologically normal. The characteristics of the normal endometrial cohort have been described previously (13). The endometrial cancer cohort (n 46) were patients diagnosed with endometrioid endometrial adenocarcinoma between 1994 and 1999 at Westmead Hospital. The endometrial cancer cohort including tumor grading and identification of coexisting areas of normal glands and hyperplasia has been described previously (14). Both the normal endometrium and endometrial carcinoma cohorts provided a valuable tool for studying PR isoform subnuclear distribution because the total PR concentration and individual PRA and PRB expression levels within the same tissue sections of each case have been previously characterized in detail (13, 14). Control tissues consisted of normal colon and normal myometrium. The work in this manuscript was conducted in accordance with the National Health and Medical Research Council of Australia s National Statement on Ethical Conduct in Research Involving Humans (1999) and approved by the Western Sydney Area Health Service Human Research Ethics Committee. Tissue sectioning and antigen retrieval Formalin-fixed, paraffin-embedded archival specimens were cut at 6 m and mounted onto SuperFrost Plus slides (Menzel-Glaser, Germany; supplied by Lomb Scientific, Taren Point, Australia) coated with Mayer egg albumin adhesive (38). Sections were dried at 37 C for 72 h followed by storage at 4 C until use. Antigen retrieval was by autoclaving as described previously (38). Briefly, slides were placed in 0.01 m sodium citrate buffer (ph 6.0) and autoclaved using a Tuttnauer 2540 EKA autoclave (Tuttnauer Co. Ltd., Jerusalem, Israel) at 121 C, 15 psi for 30 min and then allowed to cool in the sodium citrate buffer for 30 min. Dual immunofluorescent staining Sections were stained for PRB and then PRA as described previously (13). Briefly, to detect PRB, sections were incubated with a mouse antihuman PR monoclonal antibody that detects PRB only (hpra6) (39) and with a biotinylated goat antimouse antibody (Dako, Sydney, NSW, Australia) and Texas red (TXR)-avidin (Vector Laboratories, Burlingame, CA). Sections were incubated with a mouse monoclonal antibody (hpra7), which detects only PRA in formalin-fixed tissues (13, 40), and with a biotinylated goat antimouse antibody (Dako) and fluorescein isothiocyanate (FITC)-avidin (Calbiochem, Sydney, NSW, Australia). Sections were mounted with Vectashield mountant for fluorescence (Vector Laboratories) and stored in the dark at 4 C. It is important to note that the distributions of PR described in this paper were not seen using immunoperoxidase staining due to the physical masking of the subnuclear distribution by nature of the complex used to reveal PR. The dual immunofluorescent technique used in this study is selective for PRA and PRB (40) and reflects relative levels of the two isoforms (13). Under dual fluorescent excitation, PRB proteins that were labeled with TXR appeared orange; PRA proteins, labeled with FITC, appeared green, and nuclei expressing equivalent levels of PRA and PRB were yellow. Control sections were treated and stained in the same way as the test sections. Controls included adjacent sections to each tumor sample stained using antibody diluent (PBS/0.5% Triton X-100) in place of both primary antibodies to control for nonspecific staining, and to replace the second sequence primary antibody to ensure no cross-reactivity between the two staining sequences. Human colon and myometrial tissues were used for a negative and positive control respectively. Analysis of PR expression by fluorescence microscopy PR staining was examined using an Olympus BX 40 microscope fitted with filters to detect both TXR (BP ) and FITC (BP ) fluorescence simultaneously and each of the two fluorochromes separately (Olympus Optical Co. Ltd., Tokyo, Japan). The whole section was examined in detail under both individual fluorochrome excitations and

3 Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms J Clin Endocrinol Metab, March 2004, 89(3): also using the dual filter to identify the distribution of PR. Two distributions for PR within the nucleus were noted: 1) even (diffuse or fine granular staining throughout the nucleus) and 2) focal (discrete foci within the nucleus). Even and focal staining were scored separately and the proportion of nuclei with even or focal staining determined. Cases with more than 70% of nuclei showing even or focal staining were designated as high, and cases with less than 30% of nuclei as low. Cases that were not scored as either high or low were designated as moderate. The number of PR foci per nucleus was determined at 600 magnification by counting foci in more than 200 nuclei per case and expressed as less than 4, 4 6 and more than six foci per nucleus. Total PR distribution was determined by examining the localization pattern of TXR and FITC fluorescence under the dual filter, and individual PR isoform distribution was determined by examining the localization pattern of fluorescence under the individual fluorochrome excitations. Analysis of subnuclear PR distribution by confocal microscopy Distribution of PR within nuclei was studied using confocal microscopy [OptiScan F900e personal confocal system fitted with a two-line Kr-Ar laser (Optiscan Pty. Ltd., Notting Hill, VIC, Australia), which provides excitation at 488 nm and/or 568 nm, attached to an Olympus BX 40 fluorescent microscope]. Four areas within each section identified as being representative of PR distribution within the entire section were analyzed further by three-dimensional sequence acquisition with a step size of 0.2 m. The Optiscan software (F900e version 1.6.0) imageprofiling tool was used to obtain X-Y (within the image plane) and Z profiles (vertically through three-dimensional sequence of images) of staining patterns, and the size of PR foci was measured. Statistical analysis The SPSS statistical program (SPSS Inc., Chicago, IL) was used to perform all statistical tests. Spearman s rank correlation was used to test for an association between the proportion of nuclei with even PR distribution and the different phases of the menstrual cycle. Fisher s exact test was used to determine an association between the proportion of nuclei with PR distributed in discrete foci and the number of PR foci per nucleus with phases of the menstrual cycle. Spearman s rank correlation was also used to test for a correlation between PR distributed evenly in the nucleus and PR distributed into nuclear foci and further to test for a correlation between the proportion of nuclei with PR distributed into foci and the number of PR foci per nucleus. The Wilcoxon signed ranks test (matched pairs) was used to determine whether the distribution of PRA was statistically different from PRB case by case. The Kruskal- Wallis one-way ANOVA was used to test whether the proportion of nuclei with even PR distribution, the proportion of nuclei with PR distributed in discrete foci, and the number of PR foci per nucleus were similar between tumor groups expressing different PR isoform ratios in endometrial cancer. The Wilcoxon signed ranks test (matched pairs) was used to determine whether the proportion of nuclei with even PR distribution, proportion of nuclei with PR distributed in discrete foci, and the number of PR foci per nucleus changed between normal endometrial glands or hyperplasia and the matched adjacent tumor. Fisher s exact test was used to test whether the proportion of nuclei with even PR distribution, proportion of nuclei with PR distributed in discrete foci, and the number of PR foci per nucleus were similar between hyperplasia and normal endometrial glands in the endometrial cancer cohort. Spearman s rank correlation was used to test for correlations between the proportion of nuclei with even PR distribution, proportion of nuclei with PR distributed in discrete foci, and the number of PR foci per nucleus with clinical grade and HRT. Multiple logistic regression analysis was used to test whether PR concentration, PR isoform expression, and clinical grade were independent predictors of PR distribution. A bestfitting multiple logistic regression model was obtained using backward stepwise variable selection with entry criteria P 0.05 and removal P 0.1. Results Two cohorts were used in this study. The first cohort (n 23) comprised normal endometrial tissue during the menstrual cycle as previously described (13) (Table 1). In this study, the menstrual cycle phases examined were: proliferative (n 5), early secretory (n 3), midsecretory (n 6), and late secretory (n 9). The second cohort (n 46) comprised tissue from endometrial cancer patients, previously described (14) (Table 1). This endometrial cancer cohort has been described in terms of PR content and relative PR isoform expression (14). Briefly, most endometrial tumors (68%) had low PR levels (14) and expression of only one PR isoform was common [expression of PRA only or PRB only 27 of 46 (58%); Table 1]. The proportion of tumors with expression of either PRA or PRB only was the same [14 of 46 (30%) and 13 of 46 (28%), respectively; Table 1]. In the normal glands of the endometrium, PRA and PRB are expressed at similar levels (13, 14), whereas in endometrial cancer, few tumors expressed comparable levels of PRA and PRB [10 of 46 (22%); Table 1]. We have shown previously that there were inverse associations between the clinical grade of the tumor and relative PR isoform expression in which loss of one isoform was associated with higher tumor grades (14). The FIGO (Fédération Internationale de Gynécologie et d Obstétrique) grade distribution of this study (Table 1 and Ref. 14) is similar to the distribution of FIGO grade in previous endometrial cancer cohorts (2). In 12 endometrial cancer cases, the sections contained adjacent histologically normal glands with no evidence of hyperplasia or in situ malignancy. Ten cases TABLE 1. Endometrial cohorts No. of cases % Endometrium during the menstrual cycle cohort (n 23) a Early proliferative (d 5 7) 1 Mid proliferative (d 8 10) Late proliferative (d 11 14) 2 } Early secretory (d 16 19) 3 13 Mid secretory (d 20 24) 6 26 Late secretory (d 25 28) 9 39 Endometrial cancer cohort (n 46) b Age (yr) Average, 66 (range, 36 93) Median, 67 Premenopausal, Postmenopausal, HRT b Never Ever 8 17 Unknown Cancer cases containing adjacent normal glands b Cancer cases containing adjacent hyperplasic glands b PR isoform expression b PRA only PRA PRB 9 20 PRA PRB PRB only FIGO grade b Nuclear grade b a Previously published (13). Previously published (14).

4 1432 J Clin Endocrinol Metab, March 2004, 89(3): Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms contained adjacent complex atypical hyperplasia within the same section. Most patients (40 of 46, 87%) were postmenopausal ( 50 yr of age), and of the patients for which data on HRT use was available (35 of 46, 76%), most patients (27 of 35, 77%) had never used HRT (Table 1). Distribution of PR in endometrial cells Dual immunofluorescence revealed two different distributions of PR in the nuclei of endometrial glands, an even distribution, in which PR was distributed throughout the whole nucleus in a diffuse fine granular pattern (Fig. 1, A and G), and a focal distribution, in which PR was located in discrete nuclear foci (Fig. 1, D and I). Even and focal PR distribution were not mutually exclusive. Both focal and even PR distribution could be seen together in the same nucleus or in separate nuclei in the same case. Confocal analysis showed that in cases with even distribution, PR was seen throughout the nucleus (Fig. 1G), and a Z-profile, taken where indicated by the white line in Fig. 1G, confirmed this observation (Fig. 1H). The dark area in the center of the nuclei is the nucleolus, which is negative for PR (41), and this was also noted in the Z-plane image. When PR was distributed FIG. 1. PRA and PRB distribution in the normal endometrium during the menstrual cycle and endometrial cancer. PR isoform distribution was determined by dual immunofluorescent histochemistry and confocal microscopy in the glands of the normal endometrium. A C, PR distribution in the proliferative phase of the menstrual cycle, endometrial gland with nuclei containing predominately even PRA and PRB distribution; dual excitation (A), FITC (PRA) (B), and TXR (PRB) (C). D F, PR distribution in the secretory phase of the menstrual cycle, endometrial gland with nuclei predominantly containing PR localized into discrete foci; dual excitation (D), FITC (PRA) (E), and TXR (PRB) (F). G, Detail of nucleus from A (indicated by arrow), dual detection by confocal microscopy. H, Z-section of nucleus through the line shown in G, dual detection by confocal microscopy. I, Detail of nucleus from D (indicated by arrow), dual detection by confocal microscopy. J, Z-section of nucleus through the line shown in I, dual detection by confocal microscopy. K M, Endometrial cancer expressing both PRA and PRB; dual excitation (K), FITC (PRA) (L), and TXR (PRB) (M).

5 Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms J Clin Endocrinol Metab, March 2004, 89(3): into foci, these PR foci were randomly distributed throughout the nucleus (Fig. 1I). This was confirmed by the Z-plane image (Fig. 1J) taken where indicated by the white line in Figure 1I. The Z-plane image allowed the PR foci above and below the plane, shown in Figure 1I (along the line), to be visualized. By combining X-Y plane and X-Z plane profiles, length measurements were made of more than 200 foci from multiple nuclei within each section. The median length of foci was 0.75 m with interquartile range of m. PR isoform distribution in the normal endometrium during the menstrual cycle The nuclear distribution of PR isoforms was determined in the endometrium at different stages of the menstrual cycle. Normal endometrial specimens were examined for proportion of nuclei with even PR distribution (e.g. Fig. 1A), proportion of nuclei with PR distributed into discrete foci (e.g. Fig. 1D), and the number of foci per nucleus (e.g. Fig. 1I). These end points were compared with PRA and PRB concentration as previously determined in these cases using dual immunofluorescence (13). During the menstrual cycle, the relative expression of PRA to PRB alters from being similarly expressed in the proliferative phase to a predominance of PRA in the early secretory phase to PRB predominant expression in the mid secretory phase (Fig. 2A and Ref. 13). Even PR distribution The proportion of nuclei with PR evenly distributed decreased significantly throughout the menstrual cycle (P 0.01, Spearman s rank correlation). In the proliferative and early secretory phase, PR was predominantly distributed evenly throughout the nucleus and a majority of cases contained a high proportion of nuclei with this PR distribution (Fig. 2A). In the proliferative phase, both PR isoforms were highly expressed at similar levels in the glands (Fig. 1, B and C). During this phase both PRA and PRB were predominantly evenly distributed and there was no statistical difference between proportion of nuclei with even PRA or PRB distribution (P 1.00, Wilcoxon signed ranks test; Fig. 2A). The coexpression of PRA and PRB evenly throughout the nucleus is further illustrated in Fig. 1A, which shows even yellow fluorescence in most nuclei. In the early secretory phase, PR levels decreased, and FIG. 2. PR isoform distribution in the normal endometrium during the menstrual cycle. PRA and PRB distribution was determined by dual immunofluorescence histochemistry and confocal microscopy. Relative PR isoform expression levels for each menstrual cycle phase were measured previously on these cases (13), and the mean value is shown in arbitrary units. Each column represents 100% of cases within that menstrual cycle phase. Proportion of nuclei with even PRA or PRB distribution and proportion of nuclei with PRA or PRB distributed into foci was scored as low ( 30%, open bar), moderate (diagonal lines) or high ( 70%, solid bar). The number of foci per nucleus was scored as less than 4 (open bar), 4 6 (diagonal lines), or more than 6 (solid bar). A, Comparison of the proportion of nuclei with even PRA and even PRB distribution throughout the menstrual cycle. B, Comparison of the proportion of nuclei with PRA localized in discrete foci and PRB localized in discrete foci throughout the menstrual cycle. C, Comparison of the number of PRA foci per nucleus and PRB foci per nucleus throughout the menstrual cycle.

6 1434 J Clin Endocrinol Metab, March 2004, 89(3): Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms although there was more PRA than PRB expressed, both isoforms were predominantly evenly distributed. The proportion of nuclei with even PRB and even PRA distribution was not statistically different (P 1.00, Wilcoxon signed ranks test; Fig. 2A). There was a predominance of PRB expression in the midsecretory phase and a decrease in even PR distribution, with fewer cases showing a predominance of even distribution [PRA: 1 of 6 (17%); PRB: 2 of 6 (33%)]. In the late secretory phase, PR expression was low (Fig. 2 and Ref. 13), and all cases [9 of 9 (100%)] showed a low proportion of nuclei with even PRA or even PRB distribution. Focal PR distribution The proportion of nuclei containing PR distributed into discrete foci increased significantly from the proliferative phase to the secretory phase (P 0.04, Fisher s exact test). Both PRA and PRB were localized in foci, but at all stages of the menstrual cycle, PRB was always distributed into foci in a significantly higher proportion of nuclei than PRA (P 0.01, Wilcoxon signed ranks test; Fig. 2B). In the proliferative phase, cases showed a low proportion of nuclei containing PR distributed into discrete foci, and there were no cases with a high proportion of nuclei with PRA or PRB distributed into foci (Fig. 2B). In the early secretory phase, there was an increase in the proportion of nuclei with PR distributed into foci. This increase was seen for both PRA and PRB. By the midsecretory phase PRB distribution in nuclear foci was predominant, and most cases had a high proportion of nuclei with focal PRB [4 of 6 (67%); Fig. 1D, 2B]. Figure 1, D F, illustrates that foci contained more PRB than PRA because the foci were red under dual excitation (Fig. 1D). PRA levels were low in these foci as shown by detection of low fluorescence under FITC excitation (Fig. 1E). The alteration of PR distribution during the menstrual cycle from predominantly even in the proliferative phase to focal in the secretory suggests an association between the different distributions. There was a significant inverse relationship between the proportion of nuclei with PR distributed evenly and the proportion of nuclei with PR distributed into nuclear foci (P 0.04, Spearman s rank correlation, not shown). This suggests that even distribution of PR in the proliferative phase was replaced by focal distribution in the secretory phase. Number of PR foci per nucleus There was a strong positive association between the number of PR foci per nucleus and the proportion of nuclei with PR foci (P 0.01, Spearman s rank correlation). This relationship was seen for both PRA foci (P 0.01, Spearman s rank correlation) and PRB foci (P 0.01, Spearman s rank correlation). The number of foci per nucleus (Fig. 2C), at different stages of the menstrual cycle, closely resembles the pattern described for the proportion of nuclei with PR foci (Fig. 2B). PRB was always distributed in more discrete foci per nucleus than PRA (P 0.01, Wilcoxon signed ranks test). In the early and midsecretory phase, in which foci were most prevalent, 3 of 9 (33%) cases contained more than six PRB foci per nucleus, whereas 0 of 9 (0%) cases has this abundance of PRA foci (Fig. 2C). PR isoform distribution in endometrial cancer Even PR distribution. In tumors containing both PRA and PRB, even PR distribution was predominant [13 of 19 (68% of cases) had a high proportion of nuclei with even distribution; Fig. 3A]. This was also the case in tumors expressing only PRA, and there was no statistical difference in proportion of nuclei with evenly distributed PR between PRA-onlyexpressing tumors and tumors containing both PRA and PRB (P 0.17, Kruskal-Wallis test; Fig. 3A). However, tumors expressing PRB only seldom showed even PR distribution, and this was statistically different [10 of 13 (77% of cases had a low proportion of nuclei with even distribution); P 0.01, Kruskal-Wallis test; Fig. 3A] from tumors expressing either both PR isoforms or PRA only. This suggests that PRA was the major contributor to even PR distribution because this was the predominant distribution when PRA was present, whether at similar levels with PRB, greater levels than PRB, or when PRA was expressed alone, in the absence of PRB. Focal PR distribution The distribution of PR into discrete nuclear foci, in endometrial cancer, was also dependent on which isoform was expressed in the tumor. In tumors expressing both PRA and PRB, focal distribution was common [10 of 19 (53% of cases had a high proportion of nuclei with focal distribution)]. There were few PRA-only-expressing tumors that contained a high proportion of nuclei with PR distributed into nuclear foci [2 of 14 (14%)]. PR distribution into foci was more frequent when PRB was expressed in the tumor [15 of 32 (47% of cases, containing PRB with or without PRA, had a high proportion of nuclei with focal distribution)], compared with tumors that expressed PRA alone. This was statistically significant (P 0.01, Kruskal-Wallis; Fig. 3B). There was no statistical difference in the proportion of nuclei with foci between tumors expressing both PR isoforms or PRB only (P 0.74, Kruskal-Wallis; Fig. 3B). This suggests PRB to be the main component in PR distribution into nuclear foci because this was the predominant distribution when PRB was present with PRA or in absence of PRA. There was a strong correlation between the number of PR foci per nucleus (not shown) and the proportion of nuclei with PR foci, in which the number of foci per nucleus increased with increasing proportions of nuclei with PR foci (P 0.01, Spearman s rank correlation). This relationship was seen for both PRA foci (P 0.01, Spearman s rank correlation) and PRB foci (P 0.01, Spearman s rank correlation). In the normal endometrium during the menstrual cycle, there was a significant inverse relationship between the proportion of nuclei with PR distributed evenly and the proportion of nuclei with PR distributed into foci as shown. Interestingly, in the malignant endometrium cohort, this association was not noted (P 0.67, Spearman s rank correlation). This suggests that in the malignant cohort, even and focal distribution of PR were independent of each other. In malignant tissues, PR-positive cells that did not show PR distributed either evenly or in foci had a patchy distribution of PR throughout the nucleus. This patchy distribution was not seen in the normal endometrium.

7 Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms J Clin Endocrinol Metab, March 2004, 89(3): PR isoform levels and PR distribution The distribution of PR was not associated with PR concentration. PRA-only and PRB-only tumors have lower PR levels than tumors expressing both PR isoforms (Fig. 3 and Ref. 14); however, even PR distribution was common both in PRA-only-expressing tumors and tumors expressing both PRA and PRB. Also focal distribution was more frequent in tumors containing PRB, either alone or with PRA. This shows that distribution of PR was not associated with PR concentration, which was supported by multivariate analysis (Table 2), and this was tested further in tumors with similar levels of PRA and PRB. FIG. 3. PR distribution in endometrial cancer. PRA and PRB distribution was determined by dual immunofluorescence histochemistry (14). The proportion of cases expressing PRA only, or expressing PRA and PRB either in similar levels or with PRA predominance, or expressing only PRB is shown in Table 1. Total PR levels were measured previously (14) and the mean value is shown in arbitrary units. Each column represents 100% of cases within that PR isoform expression group. Proportion of nuclei with even PR distribution and proportion of nuclei with PR distributed into foci was scored as low ( 30% of nuclei), moderate, or high ( 70% of nuclei). A, Proportion of nuclei with even PR distribution in tumors expressing either PRA only, PRA and PRB, or PRB only. Open bar, low; diagonal lines, moderate; solid bar, high. B, Proportion of nuclei with PR localized in discrete foci in tumors expressing either PRA only, PRA and PRB, or PRB only. Open bar, low; diagonal lines, moderate; solid bar, high. C, Predominant distribution of individual PR isoforms in tumors with similar levels of PRA and PRB (Table 1, n 10). Solid bars, PRA distribution; open bars, PRB distribution. PR distribution in tumors expressing similar levels of both PR isoforms Ten tumors expressed similar levels of PRA and PRB (Table 1). The predominant distribution of the two PR isoforms was determined (Figs. 1, K M, and 3C) and demonstrated that PRA and PRB seldom colocated within the nucleus in endometrial cancers. PRA was predominantly evenly distributed (Fig. 1L) because most [8 of 10 (80%)] tumors expressing similar levels of PRA and PRB showed even PRA distribution in a high proportion of nuclei (Fig. 3C). PRB, however, was seldom distributed evenly (Fig. 1M) because only 1 of 10 (10%) tumors contained a high proportion of nuclei with PRB distributed evenly (Fig. 3C). This difference was statistically significant (P 0.01, Wilcoxon signed ranks test). PRB was predominantly focal in tumors expressing similar levels of PRA and PRB because most [8 of 10 (80%)] of these tumors contained a high proportion of nuclei with PRB in discrete nuclear foci, whereas only 1 of 10 (10%) of tumors contained a high proportion of nuclei with PRA in foci (Fig. 3C). The difference in focal distribution between PRA and PRB was statistically significant (P 0.01, Wilcoxon signed ranks test). Similar findings were seen for the number of PR foci per nucleus, with PRB foci per nucleus being more numerous than PRA foci (P 0.01, Wilcoxon signed ranks test; not shown). Importantly, these observations in malignant endometrial cases in which both PR isoforms were expressed at similar levels and the overall concentration of PR was moderate to high (14) demonstrate that the preferential localization of PRB in foci and PRA evenly throughout the nucleus was not associated with PR concentration. This further reflects the lack of association found by multivariate analysis between PR concentration and even or focal PR distribution (Table 2). PR isoform distribution in nonmalignant areas of endometrial cancer cases The demonstration in endometrial cancers that PRA and PRB were distributed differently within the nucleus, PRA being even and PRB focal irrespective of the relative concentration of each isoform, was in contrast to what was found in the normal endometrium. The main difference between the distributions noted in the normal and cancer tissues was the lack of PRB evenly distributed in cancer. To determine whether this difference was related to the difference in the physiological environment between the normal and cancer

8 1436 J Clin Endocrinol Metab, March 2004, 89(3): Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms TABLE 2. Multiple logistic regression analysis of PR distribution in endometrial cancer 95% CI Coefficient SE OR P value b Lower Upper Even a Independent predictors Nuclear grade PRA expression (alone or with PRB) Variables not required PR level 0.34 FIGO grade 0.64 Foci c Independent predictors FIGO grade PRB expression (alone or with PRA) Variables not required PR level 0.48 Nuclear grade 0.82 PR distribution was determined by dual immunofluorescence. PR level and PR isoform expression were previously described (14). Multiple logistic regression analysis was used to test whether PR concentration, FIGO grade, nuclear grade, and PR isoform expression were independent predictors of PR distribution. Values refer to the independent predictors in the best fitting model for each PR distribution. Statistical significance level of P a Where more than 70% of nuclei contain PR distributed evenly. b P value adjusted for independent predictors. c Where more than 70% of nuclei contain PR distributed into nuclear foci. cohort, the normal cohort being premenopausal, whereas the cancer cohort was primarily postmenopausal, PRA and PRB distributions were determined in a subgroup of cases that contained adjacent glands, within the same section as the tumor, identified as being histologically normal (n 12) or complex atypical hyperplasia (n 10). These areas were compared with the matched malignant area within the same section (Fig. 4, A and B). Even PR distribution In the normal glands, even PR distribution was predominant, and PRA and PRB were similarly evenly distributed [9 of 12 (75%) and 8 of 12 (67%) of normal areas, respectively, had a high proportion of nuclei with even distribution (P 1.00, Wilcoxon signed ranks test; Fig. 4A)], and this was similar to the observations made in the normal cohort (Fig. 2). In the adjacent malignant areas of these cases, PRA distribution remained predominantly even [7 of 12 (58%) malignant areas had a high proportion of nuclei with even PRA distribution; Fig. 4A; P 0.63, Wilcoxon signed ranks test]. In contrast, there was a reduction in the number of tumors with PRB distributed evenly, compared with the matched normal glands [3 of 12 (25%) malignant areas had a high proportion of nuclei with even PRB distribution; Fig. 4A]. However, this did not reach significance (P 0.060, Wilcoxon signed ranks test). In hyperplasia there was a significant difference in distribution between PRA and PRB (P 0.01, Wilcoxon signed ranks test; Fig. 4B). PRA was predominantly evenly distributed [8 of 10 (80% of hyperplastic areas had a high proportion of nuclei with even PRA distribution); Fig. 4B]. However, even distribution of PRB was less common in hyperplasias, with most cases having a low proportion of nuclei with PRB distributed evenly [1 of 10 (10%); Fig. 4B]. In the malignant areas adjacent to the hyperplasia, there was a nonsignificant reduction in the number of cases with a high proportion of nuclei with PRA distributed evenly [3 of 10 (30% of cases); P 0.13, Wilcoxon signed ranks test; Fig. 4B]. Even PRB distribution was less common in the malignant areas matched to adjacent hyperplasia, and this was not statistically different [2 of 10 (20% of malignant areas had low proportion of even PRB distribution; P 1.00, Wilcoxon signed ranks test; Fig. 4B]. There was no statistical difference between normal and hyperplastic areas in the proportion of nuclei with PRA distributed evenly (P 1.00, Fisher s exact test; Fig. 4, A and B). However, the proportion of nuclei with even PRB distribution was statistically different between the normal and hyperplastic areas (P 0.01, Fisher s exact test; Fig. 4A and B). The comparison of normal and hyperplastic distributions were not matched because there were insufficient cases containing coexisting normal, hyperplastic, and malignant areas. Focal PR distribution Across all histologies (normal, hyperplastic, and malignant), there were always more cases with a high proportion of nuclei containing PRB distributed into discrete nuclear foci than PRA (not shown). PRA was not highly focal and distributed into discrete nuclear foci predominantly in only 2 of 12 (17%) normal areas and not in any hyperplasias or malignant areas (not shown). No association was observed between even and focal PR in the adjacent normal and hyperplastic areas within the endometrial cancer cohort (P 0.18 and P 0.56, respectively, Spearman s rank correlation, not shown). The observation that even PRB distribution was less common in hyperplastic and invasive tumor areas did not coincide with a significant increase in focal PRB. This suggests that there may be a disruption to the even distribution of PRB, and furthermore, because this occurs in premalignant lesions, the disruption may be early in the progression to malignancy. In summary, comparison of PRA and PRB distribution in coexisting normal and invasive and hyperplastic and invasive lesions showed that PRA and PRB were evenly distributed in normal glands adjacent to endometrial cancer, con-

9 Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms J Clin Endocrinol Metab, March 2004, 89(3): compared with higher grades (P 0.02, Spearman s rank correlation; Fig. 5A). However, there was no association between even PRB distribution and FIGO grade (P 0.38, Spearman s rank correlation; Fig. 5A). The proportion of nuclei with PR distributed into nuclear foci also correlated with FIGO grade where FIGO grade 1 tumors were associated with higher proportions of nuclei with focal PRA (P 0.03, Spearman s rank correlation; Fig. 5B), and this was also the case with focal PRB. There were significantly more FIGO grade 1 cases with a high proportion of nuclei containing focal PRB than higher grades (P 0.01, Spearman s rank correlation; Fig. 5B). The associations seen above were also true for nuclear grade. Nuclear grade was associated with FIG. 4. Even PR isoform distribution in the normal, premalignant, and malignant endometrium. PRA and PRB distribution was determined by dual immunofluorescence histochemistry and confocal microscopy in endometrial cancer cases with adjacent normal glands (n 12) or adjacent complex atypical hyperplasia (n 10) (14). The distribution of PR isoforms in nonmalignant regions was compared with the matched malignant area. Each line represents a case and indicates the change in distribution between the nonmalignant and matched malignant area. Cases in which PR was predominantly evenly distributed ( 70% of nuclei) were designated as predominant. Cases that did not have predominantly even PR distribution were designated as nonpredominant. A, Even PR distribution in normal glands, compared with distribution in matched malignant area. B, Even PR distribution in hyperplasia, compared with distribution in matched malignant area. sistent with their colocalization in even distribution in the normal cohort. However, even distribution of PRB was less common in hyperplasias and malignant areas, compared with normal areas, confirming the observation that lack of even PRB distribution was a feature of endometrial cancers rather than an effect related to systemic or hormonal factors. Association of PR isoform distribution in the malignant endometrium with clinical features The proportion of nuclei with even PR distribution was associated with FIGO grade in endometrial cancer, in which FIGO grade 1 tumors contained significantly more cases with higher proportions of nuclei with even PRA distribution, FIG. 5. PR isoform distribution and clinical grade. PRA and PRB distribution was determined by dual immunofluorescence histochemistry and confocal microscopy. Proportion of nuclei with even PR distribution and proportion of nuclei with PR distributed into foci was scored as low (30% of nuclei, open bar), moderate (diagonal lines), or high ( 70% of nuclei, solid bar). A, Comparison of even PRA or PRB distribution and FIGO grade. B, Comparison of PRA or PRB localized in discrete foci and FIGO grade. C, Comparison of even PRA or PRB distribution and nuclear grade.

10 1438 J Clin Endocrinol Metab, March 2004, 89(3): Arnett-Mansfield et al. Subnuclear Distribution of PR Isoforms proportion of nuclei with even PRA distribution where nuclear grade 1 tumors contained significantly more cases with a high proportion of nuclei with even PRA distribution (P 0.03, Spearman s rank correlation; Fig. 5C). Only nuclear grade 1 cases had a high proportion of nuclei with even PRB distribution and even PRB was less common in higher nuclear grades, but this was not significant (P 0.08, Spearman s rank correlation; Fig. 5C). In summary, even PRA distribution was associated with low nuclear and FIGO grade and distribution of PRB into foci was associated with low FIGO grade. PR distribution was compared with menopausal status and HRT usage. No association was noted (not shown). Previously we have shown that PR concentration is lower in endometrial tumors with higher clinical grade and that expression of a single PR isoform (PRA or PRB) is also associated with higher clinical grade (14). Furthermore, PR concentration is lower in endometrial tumors expressing a single PR isoform. Therefore, it was important to establish whether these variables confound the associations with PR distribution reported in this study. Multivariate analysis was used to test whether PR concentration, PR isoform expression, and clinical grade were independent predictors of PR distribution. Even PR distribution The adjusted odds ratio (OR) for the model showed there was a negative association between nuclear grade and even PR distribution where tumors with a low nuclear grade predicted a predominantly even distribution. Multivariate analysis revealed low nuclear grade to be an independent predictor of even PR distribution [P 0.01; OR: 0.13; 95% confidence interval (CI): ; Table 2]. Coexpression of PRA and PRB did not independently predict an even distribution of PR (P 0.81, adjusted for nuclear grade and expression of PRA, not shown), but an association was seen with PRA expression and even PR distribution in which expression of PRA alone or with PRB predicted a predominantly even PR distribution (P 0.01; OR: 24.92; 95% CI: 2.24 to 100; Table 2). PR level and FIGO grade were not independent predictors of even PR distribution (P 0.34 and P 0.64, respectively, adjusted for nuclear grade and expression of PRA). Focal PR distribution The distribution of PR into nuclear foci was independently predicted by FIGO grade (P 0.01; OR: 0.22; 95% CI: ; Table 2) where there was a negative association, demonstrating that lower FIGO grade was associated with a predominantly focal PR distribution. Coexpression of PRA and PRB did not independently predict focal PR distribution (P 0.95, adjusted for FIGO grade and expression of PRB, not shown), but an association was seen with PRB expression and PR distribution into foci where expression of PRB, alone or with PRA, predicted that most nuclei would contain PR distributed into foci (P 0.01; OR: 11.60; 95% CI: ; Table 2). PR level and nuclear grade were not independent predictors of focal PR distribution (P 0.48 and P 0.82, respectively, adjusted for FIGO grade and expression of PRB). As stated earlier, nuclear grade and PR isoform expression are correlated with PR levels (14). After adjusting for nuclear grade and PRA expression, there was no statistical association between PR levels and even PR distribution, and after adjusting for FIGO grade and PRB expression, there was no association between PR levels and focal PR distribution. This demonstrated that nuclear distribution of PRA and PRB was independent of PR concentration in endometrial cancer. In summary, multiple logistic regression analysis showed that the nuclear grade of the tumor and PRA expression were highly predictive of even nuclear distribution, whereas FIGO grade and PRB expression were highly predictive of distribution into foci. These associations were independent of total PRA and PRB levels. Discussion We have revealed in this study that PR isoforms were distributed differently within the nuclei of endometrial cells. Two distributions for PR were noted in the endometrium: an even distribution, in which PR was distributed throughout the whole nucleus in a diffuse fine granular pattern, and a focal distribution, in which PR was located in discrete nuclear foci. This phenomenon was seen for both PRA and PRB and was present in normal endometrial glands throughout the menstrual cycle as well as in premalignant and malignant endometrial cells. Moreover, the two distributions could coexist within the same nucleus. Hormone regulation of PR distribution in the normal endometrium The human uterus during the menstrual cycle provides the opportunity to examine the expression and localization of hormone receptors under well-described hormone conditions as they occur naturally in vivo. In this study we found striking alterations in the distribution of PRA and PRB, in the nuclei of endometrial cells, during the menstrual cycle. Throughout the proliferative phase and early secretory phase, PRA and PRB were distributed evenly throughout the nucleus. In the early secretory phase, PRB distribution into nuclear foci began to be evident. During the midsecretory phase, both PRA and PRB were less evenly distributed, compared with the early secretory phase, and PRB, distributed into nuclear foci, became predominant in this phase. Estrogen is the dominant hormone in the proliferative phase of the human menstrual cycle, whereas progesterone levels are very low (42), and this hormone environment was associated with an even distribution of PRA and PRB throughout the nucleus. In the secretory phase, progesterone levels are high, PR distributed into nuclear foci were predominant, and there was an inverse relationship between even and focal distribution throughout the cycle. Taken together, this suggests that PR distribution altered from even to focal under the influence of secretory phase hormones and therefore that the nuclear distribution of PR isoforms in the endometrium is hormonally regulated. Evidence of hormone-regulated distribution of other nuclear receptors has been documented in vitro. The nuclear distribution of ER, fused to green fluorescent protein (GFP), changed under the influence of hormones in breast cancer cells, from an even