| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Lilly Research Labs, Indianapolis, Indiana 46285
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
estradiol receptor, female reproductive tract, mechanisms of hormone action, uterus, vagina
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The differential effects of E compared to those of various SERMs in the uterus have been well characterized at the tissue level. The interaction of SERMs with uterine E-receptor (ER) leads to a range of responses in the ovariectomized (Ovx) rat model, from no effect or small increases in uterine wet weight or epithelial cell height to robust, proliferative increases depending on the individual SERM tested. Also, SERMs exhibit a spectrum of E-antagonist activity in the uterus, ranging from partial antagonism of E's proliferative effects to a complete antagonism for some SERMs. Estrogen, on the other hand, has a profound proliferative effect on the uterus at much lower doses [1–4].
In humans, the benzothiophene SERM, raloxifene (Ral), produced a neutral or antiestrogenic effect on the endometrium and uterus [5, 6]. In post-hoc analyses of data from an osteoporosis treatment trial, women assigned to Ral had a significantly lower incidence of pelvic organ prolapse (POP) compared with those assigned to placebo [7], and no evidence was found for an increased risk of urinary incontinence (UI) in the Ral group compared with the placebo group [8, 9]. In contrast, hormone replacement therapy, once thought to be primarily beneficial to the postmenopausal woman's pelvic tissues [10, 11], was shown to exacerbate POP recurrence [12, 13] and UI in postmenopausal women [14, 15]. Additionally, the clinical development of levormeloxifene, a SERM from a different structural class than Ral, was halted because of an increased risk of POP [16, 17].
At the gene expression level, differences in coactivator recruitment associated with the ER may be responsible for the dramatic differences in the cellular response to E versus that to a given SERM [18]. However, much remains to be learned about the molecular events that occur as a consequence of activation of the ER with different ligands. Differential examination of uterine RNA changes in response to E compared to a SERM may help to elucidate differences in signaling downstream of the ER. Examination of RNA changes that persist over time also may help to understand what molecular differences result from long-term treatment with E versus that of a SERM and suggest cellular mechanisms for their different effects on clinical outcomes related to the uterus. Toward this end, we used microarrays to profile changes in response to E and Ral in the highly E-responsive uterus of Ovx rats following 1 day or 5 wk to reflect acute or chronic ER modulation with fully efficacious doses of compounds. To our knowledge, the present study provides the first detailed look at the molecular changes that occur on loss of ovarian function in the rat uterus and what genes are restored to pre-OVX levels following treatment with E versus treatment with Ral.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Sprague Dawley rats (age, 6 mo; Harlan, Indianapolis, IN) were used in both microarray studies. Rats were group housed and maintained on a 12L:12D photoperiod at 22°C with ad libitum access to food and water. Rats were randomly selected into four groups (n = 5 rats/group): a sham-operated group treated with vehicle, 20% hydroxy propyl-ß-cyclodextrin (Sigma); an Ovx group treated with vehicle; an Ovx group administered E at 0.1 mg kg–1 day–1 (17
-ethinyl estradiol; Sigma); and Ovx animals administered 1.0 mg kg–1 day–1 of Ral (Eli Lilly & Co). The first study was initiated 12 days post-surgery with compounds administered subcutaneously and tissues were collected 24 h after dosing (Table 1). The second study was initiated at 5 days post-Ovx, and compounds were administered orally for 35 days (40 days postsurgery), after which tissues were collected 24 h postdosing. The doses chosen were shown to be fully efficacious in this model [19]. At the time of killing, anesthetized rats were subjected to cardiac puncture and asphyxiated by CO2 inhalation. The uterus was excised just above the vagina, weighed, and frozen in liquid nitrogen or prepared for explant culture as described below. Subsequent studies used Sprague Dawley rats of various age (Table 1) that were housed and fed as described above. Levormeloxifene and lasofoxifene were synthesized as described previously [20, 21] and were dosed orally in study 6 in addition to Ral at 3, 0.3, and 0.03 mg kg–1 day–1. Animals were fasted the night before termination of each study, and all animal procedures were reviewed before implementation by an internal animal welfare committee to ensure compliance with National Institutes of Health guidelines.
|
RNA Isolation and Northern Blot Analysis
Tissues remained in liquid nitrogen until they were mechanically homogenized in Ultraspec RNA Isolation reagent (Biotecx, Houston, TX) according to the manufacturer's instructions. For Northern blot analysis, 25 µg of total RNA from each animal were loaded onto a 1% formaldehyde/agarose gel and transferred to a Nytran (Schleicher and Schuell, Keene, NH) filter following electrophoretic separation. The RNA was then cross-linked to the membrane in a UV Stratalinker 1800 (Stratagene, La Jolla, CA) and hybridized as described previously [22]. Probes used in RNA analysis were generated by random priming of approximately 200-to 500-base pair cDNA fragments with a Boehringer Mannheim High-Prime DNA Labeling Kit (Roche) and [33-P]dCTP (3000 Ci/mmol; Amersham, Piscataway, NJ). Washed membranes were exposed to phosphorimaging plates, and relative mRNA levels were quantified on a Molecular Dynamics Storm (Sunnyvale, CA). Gene expression was normalized to either 18S ribosomal or cyclophilin expression, and the group means and SEMs were calculated.
Microarray Analysis
Affymetrix rat genome U34A microarrays (Santa Clara, CA) were used to determine transcript abundance from total RNA samples of individual uterine samples. Total RNA was labeled according to manufacturer's instructions (Affymetrix GeneChip Expression Technical Manual). Briefly, all samples were cleaned using RNeasy spin columns (Qiagen, Valencia, CA). Double-stranded cDNA was synthesized from 10 µg of total RNA using Superscript II cDNA synthesis kit (Invitrogen, Carlsbad, CA), and the T7-(dT)24 primer containing a T7 promoter (5' GGCCAGTGAATTGTAA-TACGACTCACTATAGGGAGGCGG(dT)24 3' Genset Corp., Evry, France). Phase Lock Gel (1.5-ml) tubes (Eppendorf, Westbury, NY) were used to clean cDNA following phenol/chloroform/isoamyl alcohol extraction. Biotin-labeled cRNA was synthesized from cDNA using the BioArray HighYield RNA Transcript Labeling kit (Enzo, Farmingdale, NY) and cleaned by RNeasy spin columns. Clean cRNA was fragmented by incubation at 94°C in the presence of 40 mM Tris-acetate (pH 8.1), 100 mM potassium acetate, and 30 mM magnesium acetate for 35 min. Ten micrograms of fragmented cRNA were hybridized to rat genome U34A arrays for 16 h at 45°C with rotation (60 rpm). Each microarray was washed and stained using an Affymetrix Fluidics Station 400 and scanned in an Affymetrix Confocal GeneArray Scanner. Affymetrix MAS4.0 software was used to scale data to a target intensity of 1500 and to calculate transcript abundance.
Statistical Analysis
The microarray study was statistically designed to guarantee high confidence in the findings (study 1, A and B, as described in Table 1). Five animals were used in each treatment group to control for biological variation. Each biological sample was then hybridized to duplicate chips to account for the variation caused by chip performance, hybridization quality, and other differences.
To test whether a gene was differentially expressed, a mixed-effects model was fitted on each of the 8799 probe sets on the chip. The intensity value (Affymetrix MAS4 signal) of a particular gene was modeled as
 |
i(k) is the animal variation (random effect) having distribution N(0, 
); with 
A representing the standard deviation of the variation due to animal; and 
kij, independent of i(k), is the chip-to-chip variation following distribution N(0, 
) with 
representing the standard deviation of the chip measurement error. Because thousands of hypotheses were tested simultaneously, the issue of multiplicity was addressed statistically. To eliminate the false positives, the Benjamini and Hochberg [23] false-discovery rate (FDR) was used to adjust the P values derived from the above-described mixed model. The significance level of a change in expression of a probe set was based on the resulted FDR.
The ANOVA tests were done with treatment groups being the fixed effect and animals being the random effect. The P values for the pairwise comparisons were derived based on the corresponding Student t-statistics.
Bioinformatics Analysis
Principal component analysis [24] was employed to reduce the dimensionality of the data and to observe any animal-to-animal variability by taking advantage of coregulation among a large number of genes while retaining as much of the variation as possible. The reduction is achieved by transforming the data into a new set of independent variables, the principal components. The principal components are ordered in such a way that the first few retain most of the variation present in all original genes. Data was standardized with zero means and unit variance, and the analysis was performed in R programming environment [25]. The first three principal components with at least 60% variation were exported into Spotfire (Sommerville, MA) to generate a sample scatter plot to visually examine the data structure and outlier detection. This analysis revealed that duplicate chips corresponding to two E-treated animals and one Ral-treated animal were very different from the remaining animals in their respective treatment group. The expression data from these outlier animals were subsequently removed from further analysis.
Differentially expressed genes were identified based on the following criterion: FDR less than 0.05, median signals greater than 500 for each animal within one or more treatment groups, and at least a twofold change from a comparison group based on median signals. A few genes which missed our cutoff (procollagen enhancer protein [Pcpe], insulin-like growth factor-binding protein 6 [Igfbp6], tissue inhibitor of metalloproteinase 3 [Timp3], matrix metalloproteinase (MMP) 14 [Mmp14], elastase 1, and kallikrein-3) that are known to be biologically relevant also were selected for further analysis.
Because genes with a similar biological function tend to be coexpressed and to cluster together when global gene expression data are analyzed using hierarchical clustering analysis (HCA) [26], the expression of selected genes for each treatment group was clustered by HCA in Spotfire. Clustering was done in euclidean space by the complete linkage method. Heat map visualization of clusters formed was generated using range-scaled expression values and is shown in Figure 1. To compensate for local minima in accidental gene expression in HCA, self-organizing maps (SOMs) [27] was also employed to cluster genes. The SOM analysis was carried out in Spotfire with a grid size of 6 x 6 with default parameters. Clusters from SOM analysis were compared with those from HCA, and similar clusters were merged into one.
|
Gene Annotation
Target sequences for each chip were downloaded from the Affymetrix website and then compared to the National Center for Biotechnology Information genome builds, to UniGene, and to RefSeq transcripts with BLAT [28]. Annotations from these sources were used to map the probe sets both to LocusLink IDs and to full-length sequence IDs. The LocusLink IDs were mapped into the HumanPSD database [29] via indices provided by that database. Note that this database aspires to contain the full protein complement of mouse and rat as well as human. For probe sets without a LocusLink-based map, the full-length sequences were compared to the protein sequences in HumanPSD. Alignments with at least 100 amino acids of 100% identity and/or a BLAST E-value of better than 1E-20 were recorded. Multiple identifications at the same reliability level were suppressed as potential conflicts. Functional information including Gene Ontology [30] classification, gene names, and descriptions were retrieved from HumanPSD.
Uterine Explants
Half the uterus from each rat was stored in ice-cold medium (Dulbecco modified Eagle medium/Ham F-12 [3:1] phenol-free with 20 mM Hepes, 1% antibiotic/antimycotic, and 0.1% BSA fraction V [Invitrogen]) until prepared for explant culture. Explants were prepared as described by Too et al. [31] with some modifications. Specifically, the uterine halves were carefully laid open using the tip of sharp-sharp microscissors to expose the lumen and were extensively washed in culturing medium. To rid the exposed lumen of any debris, a filled, 10-ml syringe was used to gently force media over the surface. Once sufficiently cleaned, four sections (thickness, 
5 mm) of the uterine horn (excluding the two outermost sections of the horn) were cultured with 5% CO2 at 37°C for 24 h in serum-free media. Supernatants were collected and rapidly spun (14 000 x g for 1 min) at 4°C, and the cleared supernatants were frozen at –80°C until they were evaluated for MMP2 activity. Amersham MMP2 Biotrak Activity Assay System kits were used to analyze the endogenous levels of active MMP2 in the supernatants from explant cultures according to the manufacturer's instructions. The levels of active MMP2 were then normalized to the wet weight of uterine tissue and expressed as nanograms per milliliter of MMP2 per milligram of explant.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
We evaluated both short- and long-term effects of ovariectomy and treatment with high-affinity ER ligands, E and Ral, on mRNA levels in the uterus of ovariectomized rats. As expected, ovariectomy induced dramatic gene expression changes in the rat uterus relative to ovary-intact (sham) controls both at 13 days or at 5 wk postsurgery, with 1930 or 2908 probe set identifiers (PIDs) changed, respectively. Treatment of ovariectomized rats with 0.1 mg kg–1 day–1 of 17-ethinyl estradiol induced changes in 2389 PIDs following 1 day of treatment and in 2990 PIDs following 5 wk of treatment. In contrast, Ral had a more modest effect on the ovariectomized rat uterus, with 50% fewer RNA changes as compared to E at 1 day of treatment (1,213 PIDs) and 25% fewer RNA changes at 5 wk (2258 PIDs).
Gene Expression Profiles Following Prolonged Treatment with E and Ral
To understand the molecular profile of the uterus resulting from prolonged treatment with Ral compared to that with E, initial analysis and validation efforts focused on the 5-wk treatment microarray data. These data were reduced to 560 differentially expressed probes sets that were changed by 2-fold or more from respective controls (Sham vs. Ovx, E or Ral vs. Ovx). Visualizing large and complex data sets of gene changes requires the use of algorithms such as SOMs and HCA to identify trends in gene expression across treatment groups. The present study used both HCA (Fig. 1) and SOM (data not shown) analyses to find genes that had been changed by ovariectomy and subsequently were returned to Sham levels by both drugs or were uniquely regulated by either treatment. Of the 560 probe sets significantly altered, 19% were uniquely regulated back to Sham levels by E and 5% by Ral, and 75% were commonly regulated in the same direction by E and Ral. (Fig. 2A). Forty-four percent of the "common" genes had equivalent levels of change from Ovx controls following treatment with E and Ral (the majority of which were suppressed from Ovx levels). The remainder of the common genes were regulated to Sham levels by one of the treatments but only partially returned to Sham levels (achieving 30% or more of the Sham value) by the other drug (Fig. 2B).
|
Figure 1 visualized the HCA as a heat map. Heat maps display the clustered gene expression data as a matrix of cells where each cell is colored according to its expression level. Green cells represent low expression values, red high values, and black the intermediate level of expression for a given probe set. The resulting view is a colored picture that displays regions of like color grouped together that are indicative of similar expression profiles. This heat map (Fig. 1) was used to identify clusters of coregulated genes by each drug. Both Ral and E stimulated a subset of genes (shown in red) that were not significantly or were only minimally altered by Ovx and were termed E or Ral unique (listed in Table 2). In addition, some genes had been significantly changed by Ovx but were returned to Sham levels by only one of the drug treatments. These genes were termed E or Ral recoverable (listed in Tables 3 and 4). Three clusters containing 78 genes were identified that E restored to Sham levels whereas Ral had little to no effect on their expression. These genes were termed E recoverable A–C Table 4). Conversely, Ral normalized the expression of 11 genes to Sham levels (termed Ral recoverable A–C) whereas E treatment did not (Table 3). Table 5 lists six genes identified in a cluster where E and Ral had opposite effects on gene expression. Treatment with Ral further increased the expression of these genes beyond the Ovx-induced levels, but treatment with E suppressed them to or below Sham levels.
|
|
|
|
|
Confirmation of Gene Changes
In additional animal studies, we varied the age of the rat, duration of dosing, and duration of ovariectomy before treatment to confirm the RNA changes observed in the microarray (Table 1). One or more genes were selected within each cluster identified by SOMs and HCA as a representative gene to be validated in subsequent experiments and are noted in bold in Tables 2–5. Additionally, we evaluated gene expression in the vagina to determine what differences exist in the regulation of these genes from those of the uterus.
The gene Igfbp6 was chosen from the Ral unique cluster (Table 2) to be validated, and within 1 wk of treatment, a modest up-regulation by Ral was observed (Fig. 3A). Three of the genes shown to be oppositely regulated by the drugs (Table 5)—namely, aortic carboxypeptidase-like protein (Aclp), thymus cell surface antigen (Thy1), and glypican-3 (Gpc3)—were consistently found to be differentially regulated between E and Ral in the additional animal studies. The original microarray measured the changes at 5 wk of treatment, but we were able to observe differential regulation of Thy1 and Gpc3 within 2 wk of treatment (Fig. 3B) and that of Aclp by 1 wk (Fig. 3A). A similar pattern of E suppression of these three genes was observed in vaginal tissue, but Ral treatment did not alter expression of these genes (Fig. 4B).
|
|
The expression level of Pcpe was measured in confirmation studies as a representative gene found in the cluster of Ral recoverable genes. Ovariectomy did not appear to suppress Pcpe expression as observed on the microarray, and the stimulation of Pcpe by Ral from Ovx was more modest. However, the differential regulation between Ral and E seen on the microarray was validated in both the uterus (Fig. 3) and vagina (Fig. 4B), because the suppression of Pcpe by E was consistently observed.
Aquaporin 3 (Aqp3) was selected from the E recoverable cluster for validation, and its pattern of regulation from the microarray was confirmed in the uterus (Ovx, –3.4-fold change; E, +4.7-fold change). Interestingly, when Aqp3 was evaluated in the vagina, greater levels of expression were observed than in the uterus, and its increase with E treatment was more than 12-fold that of Ovx (Fig. 4A).
Estrogen uniquely induced the expression of 17 genes (Table 2). One of these genes, Mmp7, was confirmed to be highly stimulated by E in all validation studies and was observed as early as 1 wk of treatment (Fig. 5A). Vaginal expression of Mmp7 also was increased within 2 wk of E treatment, but to a lesser extent (twofold) than that observed in uterine tissue (Fig. 4A). Treatment with Ral slightly stimulated Mmp7 compared to Ovx in the microarray and confirmation studies, although the increase was markedly less than those observed with E treatment (Fig. 5). No stimulation of Mmp7 by Ral was found in the vagina (data not shown).
|
Other Matrix Genes of Interest Confirmed
Because Ral and E differentially regulated many genes associated with matrix turnover, we chose to validate Timp3 and Mmp2 in follow-up experiments. Expression of Mmp2 was down-regulated with Ovx and then slightly induced with both E and Ral treatment in the microarray experiment. Treatment with E did not increase Mmp2 in confirmation studies. Rather, E slightly decreased (10%) or did not change the expression of Mmp2. The induction of Mmp2 mRNA by Ral was confirmed in the 1- and 2-wk validation studies shown in Figure 5. The microarray indicated that Timp3 expression was enhanced following Ovx and that both E and Ral treatment reduced its expression, but to different degrees. Validation studies confirmed reduction of Timp3 by E in both uterus and vagina, but no effects of Ral were observed (Fig. 6).
|
MMP2 Activity Studies
Because genes associated with collagen synthesis and catalysis (Mmp7, Mmp14, and Mmp2) and some inhibitors of collagen turnover (Pcpe and Timp3) were differentially regulated by E and Ral treatment, we determined what biological consequence might result from such RNA changes. Because both MMP7 and MMP14 have been shown to activate pro-MMP2 to its active form [32–34], and because differential regulation of Mmp2 RNA levels had been noted, we evaluated the amount of active MMP2 enzyme in uterine tissue explants following treatment with E or Ral. Figure 7 shows that following 5 wk of treatment, Ral did not significantly alter MMP2 enzyme activity in the uterus, whereas E significantly increased MMP2 activity 7-fold that of Ovx.
|
Figure 8 illustrates that at an earlier time point (2 wk of treatment), MMP2 activity following E administration was elevated 1.5-fold over that of Ovx, but this increase did not achieve statistical significance. Administration of Ral had no effect on MMP2 activity compared to Ovx. We also evaluated the effects of two other SERMs, lasofoxifene and levormeloxifene, on MMP2 enzyme activity in uterine tissue explants. Lasofoxifene stimulated MMP2 activity to a higher level than that observed with E treatment (1.8-fold over Ovx) and nearly achieved significance (P < 0.07) at the dose of 3.0 mg kg–1 day–1. Levormeloxifene explants exhibited significant dose-dependent increases (P < 0.02) in MMP2 activity, up to nearly sevenfold that of Ovx.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
At each time point (1 day and 5 wk), E regulated a larger number of genes in the ovariectomized rat uterus than did Sham versus Ovx, suggesting that pharmacologic administration of E has additional effects in the uterus beyond those observed with cyclical, physiologic levels of systemic E (Sham). Conversely, Ral regulated nearly 50% fewer genes than did E relative to Ovx at 24 h and approximately 25% fewer at 5 wk.
In an effort to understand the chronic effects of Ral treatment, which has been demonstrated clinically to lack negative effects on pelvic floor relaxation [7] and UI [8, 9], we focused on understanding and validating the 5-wk microarray data. Only genes that changed by twofold or more from their respective control (Sham vs. Ovx, Ral or E vs. Ovx) were investigated further with multiple bioinformatics tools as described in Materials and Methods. Both E and Ral were found to similarly regulate a third of the selected genes to the same magnitude and an additional 42% in a common direction. However, distinctive clusters of genes also were observed for each treatment. Estrogen restored sixfold more genes to Sham level than did Ral (termed E recoverable), but Ral recovered genes to Sham that E did not (termed Ral recoverable). Both treatments also elevated some genes that either had not or had only been slightly altered by Ovx (termed E or Ral unique). These recoverable and unique gene changes are detailed in Tables 2– 5.
Mmp7 Uniquely Induced with E Treatment
The increased expression of Mmp7 with E treatment has been described in the involuting rat uterus [35]; however, its sustained elevation with E treatment concomitant with E's unique down-regulation of MMP inhibitors (i.e., Timp3 and Pcpe) has not been previously shown. Treatment with E uniquely up-regulated Mmp7 by microarray analysis, in which it was found to be elevated over all treatments in validation studies in uterus and vaginal tissues. A small increase in Mmp7 message was found with Ral treatment in some follow-up studies (Fig. 5); however, E consistently stimulated this to a level 3- to 10-fold that of Ral, depending on the length of treatment. It also should be noted that Ral did not suppress the expression of the MMP-inhibitors Timp3 (Fig. 6) and Pcpe (Figs. 3 and 4), as was observed with E treatment.
Genes Induced with Ral Treatment
A third of the genes up-regulated by Ral after 5 wk of treatment (Tables 2 and 5) encode molecules that have been shown either to inhibit cell growth or to induce apoptosis in other cell systems. Sequestosome 1 [36], Thy1 [37], Gpc3 [38, 39], Ras p21 protein activator [40], and apoptosis-associated tyrosine kinase [41] can induce apoptosis, whereas neuroblastoma candidate region suppression tumorigenicity 1 (Nbl1 or Dan) [42], WAP four-disulfide core domain 1 (Wfdc1 or PS-20) [43], and Igfbp6 [44–46] can act as negative regulators of cell growth. That these negative regulators of cell growth were induced by Ral suggest that the lack of uterine stimulation by Ral in the uterus as compared with E could result from the up-regulation of these negative growth regulatory genes.
Additionally, Ral restored expression levels of several genes associated with extracellular matrix (ECM) proteins: biglycan (Bgn), and procollagen C-proteinase enhancer protein (Pcpe). Bgn is a ubiquitously expressed proteoglycan that binds to collagen, and its mRNA regulation in the human cervix is increased threefold during involution [47]. In mice, the presence of BGN protein in the endometrium is associated with thicker collagen fibrils [48], whereas the loss of BGN results in morphologically abnormal and smaller collagen fibrils [49]. In the present study, Ovx lowered Bgn expression, and Ral was able to stimulate its expression back to Sham levels. PCPE likely has dual roles in tissue remodeling: as a glycoprotein that potentiates the enzymatic cleavage of type I procollagen, yielding a mature collagen molecule capable of forming fibrils [50]; and as an inhibitor of MMPs [51]. It also is interesting to note that loss of Pcpe expression has been implicated in the etiology of leiomyomata [52, 53]. The loss of Bgn and Pcpe expression together with the increase in Mmp7 with E treatment observed in the present study could have a negative impact on the structural integrity of the uterine collagenous matrix.
Genes Oppositely Regulated by E and Ral
In the present study, Aclp, Thy1, and Gpc3 were confirmed to be elevated by Ral and suppressed by E in the uterus (Fig. 3), and the suppression of all three genes by E was additionally noted in the vagina (Fig. 4). Aclp has been identified in collagen-rich tissues during embryogenesis (i.e., skeleton, vasculature, and dermis) and in the skin and adipose tissue of the adult, and it also is called the adipocyte-enhancing binding protein 1 [54]. The ACLP protein contains a discoidin domain that associates with the ECM and is strongly up-regulated during vascular smooth muscle differentiation [55]. The Aclp knockout mouse has an anterior abdominal wall defect that results in herniation of the abdominal organs and impaired dermal wound healing [54]. We describe here, to our knowledge for the first time, that Aclp is expressed in the rat uterus (itself a collagen-rich, smooth muscle organ) and that its mRNA levels can be modulated positively by Ral but suppressed by E treatment. Whether the loss of Aclp expression by E in the uterus adversely affects uterine ECM has not been established and requires further study. Similarly, the finding that E inhibited Gpc3 expression also may have negative implications for the uterine matrix. Loss of function mutations in Gpc3 in humans leads to an overgrowth syndrome called Simpson-Golabi-Behmel syndrome, and in mice, Gpc3 deficiency also causes overgrowth [39]. Thus, the loss of Gpc3 in the uterus resulting from E treatment could contribute to its large increase in size, whereas the maintenance or increase in Gpc3 message by Ral treatment allows for this molecule to be available for interaction with the ECM.
Implications of MMP2 Activity on Matrix of the Uterus
One hypothesis for pelvic floor relaxation in humans is that it results from defects in the collagen matrix of the pelvic floor region, which leads to mechanical weakness [56]. Recent reports of biochemical defects in both collagen synthesis and organization, deficits in total collagen content, and misassembled collagen fibrils have been described in patients with POP and UI [57–67]. Whereas the mechanism by which this occurs is not clear, the resulting loss and disorganization of collagen content seems to be a largely consistent finding, regardless of the site of tissue studied (vaginal, periurethral, or cervical). Estrogen has been shown to increase collagen content in the skin [68]. Thus, E was thought to be beneficial for pelvic tissues. However, reports of increased collagen degradation also have been reported with E use [58], and the use of hormone replacement therapy has been noted as a highly significant risk factor for POP recurrence [13] and for exacerbating stress UI [12, 14, 5, 69]. The rat model has not been validated for studying the mechanism of prolapse, but it has been used to study the role of MMPs in the hormonally regulated activities of the uterus, such as involution, menstruation, pregnancy, cervical dilation, and implantation. Many similarities and localization of the mRNA and protein of various MMPs have been noted between the rat, mouse, primate, and human [35, 70]. The uterus undergoes radical shifts in proliferation and degradation throughout these processes, and in all species, MMPs and their inhibitors have been shown to play key roles.
The present microarray and confirmation studies indicate that various genes involved in matrix remodeling were differentially affected by Ral and E treatment. We found that Ral slightly increased Mmp2 mRNA but that E elevated Mmp7 and Mmp14 expression in the microarray. However, MMP activity is not solely regulated at the mRNA level. The MMPs also are regulated at the protein level and are synthesized in a proform. This inactive form must be proteolytically cleaved to its active form by other proteins, which serves as an additional layer of control over MMP activity. The transcripts for two enzymes known to activate latent MMP2 to its active form, MMP7 and MMP14, were elevated by E treatment in the microarray. Furthermore, inhibitors, such as the tissue inhibitor of metalloproteinases, can regulate the cleavage of the pro-MMP. Timp3, which inhibits MMP14, was severely down-regulated by E, thus possibly leaving more MMP14 available for enzymatic activity. The possible heightened activity of MMP2 could be profound, because MMP2 it is a key effector of ECM remodeling. Increased MMP2 activity could result in the proteolytic cleavage of ECM molecules, such as type I and type IV collagen. In addition to its direct activity on collagen, MMP2 is capable of activating interstitial collagenase that also cleaves type I and II collagens [33]. It also has been shown to be an important mediator of cervical dilation and uterine involution [70].
The observation that many genes encoding proteins involved in the remodeling of matrix (i.e., Timp3, Mmp7, Pcpe, Gpc3, and Aclp) were differentially regulated by treatment, coupled with the importance of matrix turnover in prolapse, led to bioassays for MMP2 activity. Supernatants from uterine explants removed from rats that had been treated with E for 5 wk showed significantly elevated MMP2 activity, nearly threefold that found with Ral treatment and sevenfold that found with Ovx (Fig. 7). Because MMP2 activity had been shown to be elevated in vaginal epithelium of premenopausal patients with genitourinary prolapse [66], we performed an additional experiment with another SERM, levormeloxifene, which was associated with increased uterine prolapse in clinical trials (the primary reason that its clinical development was halted) [16, 17]. Treatment for only 2 wk with levormeloxifene resulted in significantly elevated levels of MMP2 activity in a dose-dependent fashion, and the levels reached were more than fourfold those seen with E treatment. The effect of lasofoxifene, a SERM currently in development for osteoporosis prevention and treatment, on human uterus is largely undescribed, but in the present rat study, lasofoxifene treatment induced levels of MMP2 activity slightly above those induced by E treatment (Fig. 8).
In summary, the present study describes the effects on RNA regulation by ovariectomy and sustained treatment with Ral and E in the rat uterus. Overall, E regulated twice as many genes as Ral. Treatment with E consistently and significantly induced gene changes from controls that are associated with matrix turnover. Ral treatment specifically increased mRNAs associated with cell death and negative cell growth while maintaining the expression of genes whose products are important for matrix integrity (Gpc3, Aclp, Timp3, and Pcpe). One biological consequence of the mRNA changes was the level of enzymatic activity of MMP2 in uterine explants from treated animals, in which it was demonstrated that E treatment significantly elevated MMP2 activity (Fig. 7). Whereas Ral slightly elevated levels of uterine Mmp2 mRNA, treatment with Ral did not increase MMP2 enzymatic activity, highlighting the importance of other regulators of MMP2 activation in the tissue environment (i.e., the presence of inhibitors such as TIMP3 and activators of MMP2 such as MMP7 and MMP14). It was further observed that SERMs differentially affect MMP2 activity and may serve as a possible mechanism whereby SERMs can be differentiated from each other in the uterus (Fig. 8). Taken together, these data suggest that genes and proteins associated with matrix integrity in the rat uterus show significantly different patterns of regulation between E and Ral treatment.
|
FOOTNOTES |
|---|
Received: 13 September 2004.
First decision: 11 October 2004.
Accepted: 10 November 2004.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
B activation. J Biol Chem 2002 277:14641-14646
| ||||||||||
