Human Uterine Leiomyomata Express Higher Levels of Peroxisome Proliferator-activated Receptor γ, Retinoid X Receptor α, and all-trans Retinoic Acid Than Myometrium

Uterine leiomyomata, benign tumors of smooth muscle cells, are the most common tumors in the human female pelvis and the leading indication for pelvic surgery (1, 2). Leiomyomata can cause anemia, pain, discomfort, menstrual disturbances, and reproductive failure. Therapy with gonadotropin-releasing hormone agonists induces a temporary hypoestrogenic state and reduces leiomyoma size by 50%. This is implemented as a prelude to either myomectomy or hysterectomy (3). However, the hypoestrogenism causes significant noncompliance because of climacteric-like symptoms, such as hot flashes, vaginal dryness, and decreased libido. Additionally, this therapy has the potential for bone loss and adverse cardiovascular changes when it exceeds 6 months duration (3).

The cause of leiomyomata is unknown, but estrogens produced by leiomyoma aromatase are suspected to promote their growth (4). Rodent models for uterine leiomyomata include the Eker rat (5), carrying a mutation in the tuberous sclerosis-2 gene, and two transgenic mice (6, 7). In our guinea pig model, exposure solely to E2⁷ silastic implants was associated with leiomyomata forming mainly on the abdominal wall, whereas exposure to E2 and retinoic acid implants “switched” leiomyoma formation to the uterus (8). This observation prompted us to test human leiomyomata for the expression of RARs and RXRs and for other members of the ligand-activated nuclear receptor superfamily (9). Encouraging immunoblot data focused our attention on PPARγ and RXRα and their potential relevance to the regulation of leiomyoma growth.

PPARγ, one of three mammalian PPAR isoforms (α, β/δ, and γ), is found in adipose tissues, where it mediates adipocyte differentiation (10), and other insulin-responsive tissues such as skeletal muscle and liver (11, 12, 13, 14, 15). RXR is regarded as the master nuclear receptor because it forms heterodimers with RAR, PPAR, vitamin D₃, or thyroid hormone receptors (9). In the presence of a PPARγ ligand, PPARγ-RXR heterodimers are formed and bind to peroxisome proliferation-responsive elements; the heterodimers are considered the functionally active receptor forms in vivo.

Retinoic acids are the biologically active form of vitamin A; atRA and 9cRA can be isomerized to each other both in vivo and in vitro. RARs (RARα, RARβ, and RARγ) bind atRA and 9cRA. The RXRα, RXRβ, and RXRγ bind 9cRA. RAR and RXR are encoded by different genes, and each subtype (α, β, and γ) differs mainly in their NH₂ terminus because of alternate mRNA splicing and use of different promoters.

Naturally occurring ligands for PPARγ are unsaturated fatty acids and prostaglandins (12, 14, 16, 17). PPARγ is found in the small intestines and colon, and as the receptor for fatty acids, PPARγ may have a primary role in colon cancer (18). Natural and synthetic PPARγ ligands, such as Tro, a thiazolidinedione, are sufficient to stimulate adipocyte differentiation in fibroblast-like preadipocytes (11, 14). The ability of Tro to bind and activate PPARγ correlates with its ability to improve insulin sensitivity in type II diabetic patients and in animal models of diabetes and obesity (11, 12, 13, 14).

Our in vivo guinea pig experiment with Tro (19) demonstrated that sustained exposure to E2, Tro, and atRA were optimal for uterine leiomyoma growth in this model. On the basis of comparable findings in human leiomyomata, we hypothesize that new therapeutic modalities for human leiomyomata could be based on antagonists for PPARγ and RXRα, given singly or in combination with SERMs.

Tissues were obtained according to protocols approved by the Institutional Review Boards for human and animal research. Uterine samples were obtained within 30 min after excision in a hysterectomy or myomectomy and were stored at −75°C. The human leiomyomata sampled were >2–3 cm and distant to necrotic, calcified, or infarcted regions; myometrial samples were free of endometrium. The stage of the menstrual cycle was determined by endometrial dating and the patient’s history.

Frozen human and animal tissues were finely cut and placed (at 150 mg wet tissue/ml) in a Tris-HCl (pH 8) buffer containing aprotinin, NP40, NaCl, NaF, phenylmethylsulfonyl fluoride, orthovanadate, and leupeptin (20) and were extracted overnight at 4°C with gentle tumbling. Recently, in a few homogenizations, we included 20 μm of the cathepsin-L/calpain inhibitor Z-Leu-Leu-Tyr-fluoromethyl ketone from Enzyme Systems Products (Livermore, CA). After 15 s of Tekmar homogenization and 2 h of tumbling, the homogenates were centrifuged at 14,000 × g for 30 min. In the supernatant, protein was measured with the BCA kit from Pierce (Rockford, IL) using BSA as a standard.

Dunkin-Hartley female guinea pigs (Cavia porcellus) were used in this study and received one or two silastic implants (21) of E2 (each contained 50 mg of powder), an implant of atRA from Sigma Chemical Co. (St. Louis, MO) or 9cRA (a gift from Hoffmann La-Roche, Nutley, NJ), each containing 40 mg of powder, and a small identification chip from AVID (Norco, CA) in the s.c. space of the intrascapular area after a bilateral oophorectomy via laparotomy (or sham operation) in the costovertebral angle. Anesthetics were administered as described before (22). Tro (Rezulin tablets) from Parke-Davis (Ann Arbor, MI) was suspended in an aqueous solution of 1% carboxymethylcellulose, 0.81% NaCl, and 0.1% Tween 80 and given daily p.o. at a dose of 10 mg/kg body weight in 1 ml, followed by 1 ml of fruit punch. All animals were weighed weekly.

NuPage Bis-Tris 4–12% gels from Novex (San Diego, CA) and recombinant S-tagged molecular weight markers from Novagen (Madison, WI) were used; 50–100 μg protein were loaded per lane. Polyclonal antibodies to RARα-β2-γ, RXRα (SC-552, epitope at NH₂ terminal), RXRβ-γ from Santa Cruz Biotechnology (Santa Cruz, CA) and Affinity Bioreagents (Golden, CO), monoclonal RXRα antibody 4RX3A2 (a gift from Dr. P. Chambon, IGBMC, Illkirch, France), and an antibody (15) to recombinant PPARγ2 (a gift from Dr. B. Spiegelman, Dana-Farber Cancer Institute, Boston, MA) were used. Equal lane loading was confirmed either by an α-desmin antibody from Sigma Chemical Co. (St. Louis, MO) or by protein staining of the nitrocellulose membrane with BLOT-FastStain from Geno Technology (St. Louis, MO), followed by destaining, blocking, and exposure to the primary antibody. Secondary antibodies were from Amersham (Arlington Heights, IL) and Novagen, and chemiluminescence kits were from Amersham.

Total RNA was isolated with the TRI reagent from MRC (Cincinnati, OH), and poly(A)⁺ RNA from total RNA using the Oligotex kit from Qiagen (Valencia, CA). Prehybridization and washings were carried out at room temperature with MRC solution WP-117, and hybridizations were performed at 65°C with MRC solution HS-114. cDNA probes (14) for PPARγ, RXR (gifts from Dr. B. Spiegelman), and α-desmin (a gift from Dr. Y. Capetanaki, Baylor College of Medicine, Houston, TX) were labeled with the random-primed DNA labeling kit from Boehringer-Mannheim (Indianapolis, IN) and [α-³²P]dCTP (6000 mCi/mol) from NEN (Boston, MA) and were purified with the DNA Clean-up kit from Promega (Madison, WI).

All procedures were performed under dim light to prevent photoisomerization of retinoids. Serum samples were extracted with 2-propanol, as described previously (23). Uterine tissues (200–400 mg) were cut frozen in 2-mm pieces and extracted for 5 min with three volumes of ice-cold 2-propanol, containing 13 μg of butylated hydroxytoluene/ml as antioxidant, followed by 10-min sonication at 4°C and shaking for 5 min at room temperature. Precipitated cell debris and proteins were removed at 6500 ×g, and the supernatant solutions from either serum or tissues were extracted on a C2 solid-phase cartridge to recover the retinoids and remove interfering substances (24). The cartridges were loaded into the HPLC system, and separation of retinoids was achieved on a C₁₈ column with gradient elution within 28 min; this method can resolve 17 retinoids (24, 25). Criteria for identification of HPLC peaks were the coincidence of retention time and ratio of peak areas at two wavelengths of detection with those of authentic retinoids. For identification of some retinoids, HPLC fractions were collected and rechromatographed, and UV spectra were recorded at 300–370 nm in the flow cell of the UV detector (26). An isocratic method (25, 27) was used for rechromatography of the putative atRA peaks from the uterine samples; this method can resolve 13cRA, 9,13-di-cRA, 9cRA, and atRA.

Unlike rodents, guinea pigs form uterine leiomyomata spontaneously as they age (21). The 1930s guinea pig model for leiomyomata was reexamined using E2 silastic implants and modern methods (21); uninterrupted administration of E2 caused large smooth muscle cell tumors (leiomyomata), primarily on the abdominal serosa and small leiomyomata on the uterine horns (Fig. 1, a and b). In women, however, leiomyomata are restricted to the uterus and rarely form in the abdominal cavity (leiomyomatosis peritonealis disseminata), lungs, digestive tract, or arteries. To establish the mechanism of leiomyoma growth and ultimately reverse the disease, we exposed (8) ovariectomized guinea pigs for 260–300 days to one or two E2 implants plus one atRA (n = 12) or one implant each of E2 and 9cRA (n = 5). The E2 implants, with or without 9cRA or atRA implants, maintained serum E2 levels in ovariectomized animals at 30–90 pg/ml (gestational levels in the guinea pig) for up to 400 days (21). Mean serum retinol (±SE, n = 8) was 270.5 ± 20.0 ng/ml, retinyl palmitate/oleate was 22.7 ± 4.2, and retinyl stearate was 35.2 ± 6.1 at 260–300 days of treatment; serum 13cRA was 4.7 ± 0.6 (n = 6), but atRA and 9cRA were detectableM only in six animals at ∼0.6 ng/ml.

Surprisingly, intramural and serosal uterine, but not abdominal, tumors were formed in 11 of 12 animals on E2 and atRA and 5 of 5 animals on E2 and 9cRA (Fig. 1, c and d). The tumors were leiomyomata, as verified by electron microscopy and desmin immunostaining (data not shown). In the absence of E2, atRA or 9cRA produced no leiomyomata. Animals on E2 and either atRA or 9cRA survived beyond 400 days, whereas treatment solely with E2 (21) for longer than 7 months produced extensive abdominal tumors, causing intestinal obstructions that forced the termination of the experiments.

Nuclear receptor RAR (α, β, and γ), RXR (α, β, and γ), and PPARγ protein levels were determined by immunoblotting of tissue extracts of human leiomyoma and myometrium and guinea pig leiomyoma and uterine horn cross-sections. PPARγ and RXRα expression were higher in human leiomyomata than myometrium, but only in the follicular phase; higher levels were noted in guinea pig leiomyomata relative to uterine horn cross-sections (Figs. 2 and 3).

In human tissues, no correlation was found between the leiomyoma:myometrium ratios for PPARγ and RXRα levels or between the PPARγ leiomyoma:myometrium ratios and body mass index. RARα, RARβ, or RARγ levels were not consistently different between leiomyomata and myometrium (data not shown). The RXRα immunoblots with an antibody to a RXRα NH₂-terminal segment (Fig. 3) suggested that the native RXRα form (M_r ∼50,000) was partially degraded; the degradation was more pronounced in the human than guinea pig tissue extracts. The M_r 50,000 and M_r 35,000 bands were included in RXRα quantitation (Fig. 3). Monoclonal antibody 4RX3A2 to human RXRα gave nearly identical immunoblots as those in Fig. 3 (data not shown), but it did not react with the guinea pig samples. Recently, it was reported that cathepsin-L partially degrades RXRα (28, 29). We tested in eight samples whether 20 μm Z-LLY-FMK, a new cathepsin-L/calpain inhibitor, added during human tissue homogenization would prevent RXRα degradation. Indeed, the main RXRα band was now at M_r ∼50,000 in both myometrial and leiomyoma immunoblots (data not shown), suggesting that the partial proteolysis of RXRα seen in earlier blots (Fig. 3) occurred during tissue homogenization by a protease(s) inhibited directly or indirectly by Z-LLY-FMK. Interestingly, Z-LLY-FMK reduced the α-desmin bands to two, at M_r ∼50,000.

In women, the 3- to 5-fold higher expression of PPARγ and RXRα in leiomyomata depended on the phase of the menstrual cycle (Fig. 4). Perhaps, in the follicular phase E2, unopposed by progesterone, is the primary stimulus for increased PPARγ and RXRα expression, leading to leiomyoma growth. It is possible that progesterone could decrease PPARγ or RXRα expression directly, not through decreased E2 receptor concentration.

To corroborate the Western blot data, Northern blots and reverse phase-PCR of PPARγ and RXRα were performed. Fig. 5 shows that PPARγ and RXRα mRNAs were equally expressed in human myometrium and leiomyomata throughout the menstrual cycle, except for PPARγ mRNA levels that were lower in leiomyomata than myometrium (P < 0.002) only in the luteal phase. Another Northern blot of RXRα using total RNA samples from three uteri in the luteal phase also showed equal mRNA levels between leiomyomata and myometrium (data not shown). Reverse phase-PCR showed that mRNA for total PPARγ (segment 146–620; Ref. 30) and RXRα (segment 1318–1430; Ref. 31) were expressed in both myometrium and leiomyoma throughout the cycle (data not shown).

HPLC analysis of retinoids in human tissue extracts (Fig. 6) showed 4.5-fold higher atRA levels in leiomyomata than myometrium in the follicular phase only. By contrast, the presence of atRA in the luteal phase was not confirmed because of interferences that coeluted with atRA; 9cRA was not detectable. Retinol (vitamin A) levels were 25% lower in follicular phase leiomyomata than myometrium (Fig. 6) and ∼35% lower in luteal phase leiomyomata than myometrium (data not shown). Retinyl ester levels were undetectable in most of these tissues, or when detected, did not exceed 15 ng/g, a much lower content than retinol. Lower retinyl ester levels than retinol are also found in guinea pig uterine tissues (data not shown) and in rodent embryos of 11–12 days gestational age (27).

Because PPARγ expression was higher in human leiomyomata, we tested in guinea pigs whether a PPARγ agonist, Tro, could cause leiomyomata. Tro was given daily for 75 days p.o. (10 mg/kg) to seven ovariectomized guinea pigs primed for 150 days with two E2 (n = 2) and either one atRA (n = 2) or 9cRA (n = 3) implant; two other animals received Tro only for 75 days. Only combined exposure to E2, atRA, and Tro produced uterine leiomyomata, 4-fold larger than seen previously in the guinea pig (Fig. 7,e; 4 cm at largest dimension). Tro alone had no effect on the atrophic uterus of the ovariectomized animals (Fig. 7,b) and even prevented abdominal leiomyoma formation in animals treated with E2 (Fig. 7,c). The apparent increase of abdominal fat attributable to Tro alone (Fig. 7,a) was not accompanied by weight gain during the 75-day treatment with Tro. Unexpectedly, the combination of 9cRA with E2 and Tro produced neither uterine or abdominal leiomyomata (Fig. 7,f), although 9cRA and E2 produced uterine leiomyomata (Fig. 1, c and d).

Immunoblots for PPARγ and RXRα showed (Fig. 8) that Tro given either alone or with E2 induced measurable PPARγ and RXRα levels both in the atrophic (Lanes 1a and 1a′) and the E2-stimulated uterine horn (Lanes 5a and 5a′); the distribution of these receptors between myometrium and endometrium has not yet been determined. It appears that relatively high expression of PPARγ and RXRα in the uterine horns (Lanes 1 and 5) does not support leiomyoma formation. Rather, a combination of additional stimuli, E2 and atRA, is needed for tumor formation.

Our hypothesis is that the rise of PPARγ and RXRα proteins in follicular phase leiomyomata reflects an increase of the transcriptionally active PPARγ:RXRα heterodimer, and that the latter is a key to leiomyoma growth. This heterodimer was suggested as a single-function complex and a molecular target for treatment of insulin resistance because of beneficial effects seen in diabetic patients treated with the PPARγ agonist Tro and in animal models treated with RXR agonists and, surprisingly, antagonists of RXR (32). In leiomyomata, we do not know the inducers of the putative PPARγ:RXRα complex or whether estrogens alone stimulate PPARγ and RXRα expression and higher atRA levels. Estrogens induce the formation of a prostaglandin D₂ metabolite that activates PPARγ in the duck uropygial gland (33). In the guinea pig model (Fig. 1), E2 stimulation is necessary but not sufficient for uterine leiomyoma growth. Estrogens are considered beneficial to insulin resistance, although they correlate with obesity in women. A modest increase in relative risk (confidence interval, 1.03–1.59) for uterine leiomyomata with elevated body mass index was reported recently in a large study of adult women (34).

Additional studies will determine whether “downstream” factors, such as nuclear receptor coactivators, corepressors, and histone acetyltransferases (35) are differentially expressed and important for leiomyoma growth, in addition to “upstream” inducers of retinol activation, 15-lipoxygenase and prostaglandin D and J biosynthesis that may act in leiomyomata. Undoubtedly, vitamin D₃ receptors participate in the development of the leiomyoma phenotype, such as calcification foci, common to human and guinea pig leiomyomata (Fig. 7 e; data not shown).

Fig. 9 summarizes our results on the pathways to leiomyoma formation. Three “inducers” are proposed as prerequisites of uterine leiomyoma development and growth, i.e., E2 stimulation in the absence of progesterone, higher atRA levels, and higher PPARγ-RXRα levels. Unlike “hits” leading to malignant transformations (36), this mechanism implies that leiomyoma growth is potentially reversible by lowering “inducer” levels, as shown in women who become hypoestrogenic by gonadotropin-releasing hormone agonist therapy (3).

In the presence of Tro, there is an apparent specificity for atRA, in terms of tumor growth promotion, because 9cRA seems to prevent growth. This apparent specificity is reminiscent of the therapeutic efficiency of atRA in cancers involving the bone marrow (acute promyelocytic leukemia). It is not known if the bone marrow is in the pathway of leiomyoma growth; Tro induces adipogenesis in cultures of bone marrow stromal cells (37).

A direct link at the molecular level between estrogens and the retinoid pathway (RAR, RXR, PPARγ) has not been established in leiomyomata. Two reports (38, 39), awaiting confirmation by other laboratories, suggested that in vitro heterodimers form between either RARα or RXRα and the ER α and β; no evidence was offered about PPARγ-ER dimers. One explanation why guinea pigs treated with E2-9cRA-Tro (Fig. 7,f) did not develop leiomyomata could be that increased 9cRA favors RXRα homodimerization, in effect, removing RXRα from transcriptionally active RXRα-PPARγ or other heterodimers. Perhaps 9cRA alone would be beneficial to patients with leiomyomata. In the model, the ability of 9cRA and E2 to produce small uterine leiomyomata in the absence of Tro (Figs. 1 and 9) points to a pathway that does not require further induction of PPARγ. The fact that Tro in the presence of E2 prevented even abdominal leiomyoma formation (Figs. 7,c and 9) may imply that leiomyoma formation relies on a narrow heterodimer stoichiometry, which can be disturbed by receptor agonists and antagonists.

In the follicular phase of the menstrual cycle, under the influence of elevated E2 and atRA, increased levels of the putative ER-RAR and ER-RXR heterodimers could form, along with RXRα-PPARγ heterodimers stimulated by endogenous PPARγ ligands, also under the control of estrogens (33). Anzano et al. (40) and Keller et al. (41) were first to suggest an interaction between estrogen action and RXRs/or 9cRA and/or PPARs, respectively. Nuňez et al. (42) also reported that RXRβ and PPARα are capable of activating estrogen-responsive genes by direct binding to estrogen response elements.

In the luteal phase, E2 levels are lower than those at the preovulatory stage, and rising progesterone could down-regulate essential heterodimers, causing arrest of leiomyoma cell growth (increase in cell size, hypertrophy) and initiation of mitosis (hyperplasia); more mitotic indices are found in human leiomyomata during the luteal than follicular phase (43).

If a mechanism of three principal “inducers” controls leiomyoma growth (Fig. 9), it is reasonable that new therapies should target these “inducers.” A synergistic interaction between the antagonists would lower each antagonist’s effective dose for tumor regression and would minimize side effects. For example, the SERM raloxifene (44) or LY383351 (45) given p.o. will undoubtedly cause leiomyoma regression. However, suboptimal SERM doses for leiomyoma monotherapy combined with PPARγ antagonists (under development) or RXRα antagonists and agonists (now in clinical trials for type II diabetes and breast cancer) may offer an advantage as new combination therapies for leiomyomata. Moreover, some of these heterodimers in leiomyomata may have redundant functions, as seen in other systems (46), and it may be necessary to target all three leiomyoma “inducers” to account for individual differences in endogenous receptor inducers and isoforms; even among separate leiomyomata from one uterus, the levels of PPARγ and RXRα proteins are not identical. Vaginal application or in situ (laparoscopic) delivery to the tumors may be preferable than oral administration of the antagonists. At present, there are no tissue-selective inhibitors of enzymes, e.g., retinol oxidases, to test the effect of lowering leiomyoma atRA levels.

We thank Drs. Yassemi Capetanaki, Pierre Chambon, Bruce Spiegelman, and Zhidan Wu for reagents, Dr. Santo Nicosia for assistance with leiomyoma pathology, and Drs. Ronald Chez and George Wilbanks for reading the manuscript.