Estrogen-dependent Production of Erythropoietin in Uterus and Its Implication in Uterine Angiogenesis*

Although erythropoietin (Epo) has been shown to possess in vitro angiogenic activity, its physiological significance has not been demonstrated. Normally angiogenesis does not occur actively in adults but an exception is the female reproductive organ. In the uterine endometrium, angiogenesis takes place actively for supporting the endometrial growth that occurs during transition from the diestrus to estrous stage. This transition is under control of 17β-estradiol (E2), an ovarian hormone, and can be mimicked by injection of E2 to ovariectomized (OVX) mouse. Thus, the uterus is a pertinent site to examine the Epo function in angiogenesis. We found that Epo protein and its mRNA were produced in an E2-dependent manner, when the uterus from OVX mouse was cultured in vitro. The de novoprotein synthesis was not needed for E2 induction of Epo mRNA. Administration of E2 to OVX mouse induced a rapid and transient increase in Epo mRNA in the uterus. Injection of Epo into the OVX mouse uterine cavity promoted blood vessel formation in the endometrium. Furthermore, injection of the soluble Epo receptor capable of binding with Epo into the uterine cavity of non-OVX mouse in diestrus stage inhibited the endometrial transition to proestrus stage, whereas heat-inactivated soluble Epo receptor allowed the transition to occur. These results, combined with our finding that the endothelial cells in uterine endometrium express Epo receptor, strongly suggest that Epo is an important factor for the E2-dependent cyclical angiogenesis in uterus.

Angiogenesis is the formation of new blood vessels by the extension of pre-existing vessels into avascular area and involves the proteolytic degradation of the vascular basal membrane, proliferation and migration of endothelial cells, and alignment of the migrating cells for tubular formation. Angiogenesis occurs very actively in embryogenesis, but it is downregulated in the healthy adult. Active neovascularization in adults takes place in certain pathological conditions such as arthritis, diabetic retinopathy, wound healing, and tumor growth (reviewed in Ref. 1). An exception in adults is the female reproductive organ, where active angiogenesis is demanded to support the cyclic remodeling of tissues. In every estrus cycle, capillary networks in the ovaries are formed for supporting development of follicles and corpora lutea. In the uterus, cyclic formation of blood vessels in the functional endometrium occurs to compensate for the lost vessels. In response to embryonic implantation, decidual transformation of the endometrium is accompanied by neovascularization, which ultimately leads to formation of maternal vessels in the placenta.
Cyclic development of the uterine endometrium is under the control of E 2 , 1 which is produced by ovarian follicles (2). This endometrial development can be mimicked by the administration of E 2 to the OVX immature or adult animals (3)(4)(5). A number of growth factors including fibroblast growth factor, tumor growth factor, and VEGF have been implicated in angiogenesis (Refs. 6 -8 and references therein). One of the early events caused by the E 2 administration to the OVX rats is the increased vascular permeability in the endometrium (9). Based on the temporal pattern of mRNA expression after the E 2 administration (10) and capability of increasing vascular permeability as well as the mitogenic activity for vascular endothelial cells (11)(12)(13), VEGF has been proposed to be a critical factor in the early phase of E 2 -induced angiogenesis (10,14).
Epo is a key factor for regulating erythropoiesis by stimulating proliferation and differentiation of late erythroid precursor cells (15)(16)(17)(18)(19). Epo involved in erythropoiesis is produced by the kidney in adults and the liver in fetuses. In addition to the erythropoietic function of Epo, we and others (20 -26) have recently shown that the brain has a paracrine Epo/EpoR system, which is independent of erythropoietic system; neurons express EpoR (20,22) and astrocytes produce Epo (24 -26). We have shown (22,27) that brain Epo contributes to neuron survival by protecting neurons from ischemic damage.
Angiogenic activity of Epo has been studied by the use of in vitro cultured endothelial cells. EpoR mRNA is expressed in endothelial cells from human umbilical vein, bovine adrenal capillary, and rat brain capillary (28,29). Epo stimulates proliferation and migration of human and bovine endothelial cells (30) and also angiogenesis of the rat thoracic aorta (31). Recent studies of human umbilical vein endothelial cells indicate that Epo signaling in endothelial cells is conducted via tyrosine phosphorylation of proteins including phosphorylation of transcription factor STAT-5, which is similar to that in erythroid cells (32). However, it is unknown whether endothelial EpoR is physiologically functional or is only a vestige reflecting a common developmental lineage between endothelial cells and hematopoietic cells (8). The uterus where the active angiogenesis takes place in an E 2 -dependent manner may be a target pertinent to examine the physiological significance of Epo in angiogenesis. We found E 2 -dependent Epo production in uterus, suggesting the estrous cycle-dependent fluctuation of Epo concentration in the uterine tissues. This finding prompted us to examine the role of Epo in the E 2 -induced endometrial regeneration with an expectation that Epo acts as a uterine angiogenic factor.

EXPERIMENTAL PROCEDURES
Epo, Epo Assay, and sEpoR-Recombinant human Epo was produced and isolated as described previously (33,34). Epo was measured with a sandwich-type enzyme-linked immunoassay using two monoclonal antibodies that bind Epo at different epitopes (35). This assay measures Epo as low as 1 pg/ml. Recombinant human Epo was used as a standard. Recombinant murine soluble EpoR (sEpoR), an extracellular domain of EpoR capable of binding with Epo, was produced and isolated as described in the previous report (36).
RT-PCR-Total RNA was prepared by the use of RNA Isolation System kit (Promega). The RT reaction was performed using a random nonamer primer and 1 g of RNA in a volume of 20 l. PCR primers of Epo and ␤-actin were those described previously (22,26), and those of VEGF were sense primer (mVEGF130F, 5Ј-TGCTGTACCTCCACCAT-GCCAA-3Ј) and antisense primer (mVEGF657R, 5Ј-ACCGCCTTGGCT-TGTCACATCT-3Ј) (GenBank TM accession number M95200; Ref. 48). PCR cycles and conditions for denaturation, annealing, and elongation were 40 cycles, 1 min at 94°C, 2 min at 63°C, and 3 min at 72°C of Epo; 25 cycles, 1 min at 94°C, 2 min at 64°C, and 3 min at 72°C of VEGF; 25 cycles, 1 min at 94°C, 2 min at 61°C, and 3 min at 72°C of ␤-actin. The amplified DNA was fractionated by electrophoresis and stained with ethidium bromide. Band intensity was quantified using a Macintosh computer using the public domain NIH image program. 2 The PCR cycle used here was set within the range so that the cycle number was approximately proportional to the band intensity of the amplified DNA.
Mouse-Animals were maintained and handled in accordance with the guidelines for the care and use of laboratory animals at Kyoto University. Four-week-old outbred mice of the ICR strain (Clea) were ovariectomized and used for experiments at 4 -6 weeks after OVX. Ten-week-old mice in the diestrus stage, which was determined by the vaginal smear, were used for experiments of non-OVX mice.
FIG. 1. In vitro cultured uterus from OVX mouse produces Epo protein and its mRNA in an E 2 -dependent manner. As described under "Experimental Procedures," one horn of the uterus from an OVX mouse was cultured in the medium containing E 2 for 6 h, and the contralateral horn was cultured without E 2 as a control. Epo secreted in the culture media was measured, and the tissues were used for detec- Culture of the Uterus from OVX Mouse-Bilateral horns of the uterus from OVX mouse were cut into two separate horns. One horn was cultured in the medium containing E 2 for 6 h in a humid 5% CO 2 atmosphere at 37°C in phenol red-free Dulbecco's modified Eagle's medium supplemented with 20% charcoal-treated fetal calf serum, and the contralateral horn was cultured without E 2 as a control. Epo protein secreted in the culture media was measured. Tissues were used for detection of Epo mRNA by RT-PCR.
Administration of E 2 and Epo to OVX Mouse-E 2 (0.5 mg/kg) or vehicle (olive oil) was given intraperitoneally. Epo was injected into the uterine cavity of the OVX mouse as follows. The abdominal wall of the mouse under deep anesthesia was incised, and the entire uterus was pulled out. Then the uterus was ligated at the oviduct and vaginal ends to form three uterine cavities, two lateral and one central. Epo in 100 l of saline was injected into one of the lateral cavities through a microsyringe with a 32-gauge needle (Hamilton) and saline into the contralateral cavity as a control. Then the whole uterus was returned to the abdominal cavity, and the abdominal incision was sutured. At 24 h after injection, the uterus was excised and fixed in Zamboni solution.
Administration of sEpoR to Non-OVX Mouse-sEpoR or heat-inactivated sEpoR in 100 l of saline was injected into uterine cavity of 10-week-old non-OVX mice in the diestrus stage with the above-described procedures for Epo injection. At 24 h after injection, the uterus was excised and fixed in Zamboni solution.

E 2 -dependent Production of Epo and Its mRNA by the in Vitro Cultured
Uterus-If we assume that Epo plays an important role in E 2 -dependent angiogenesis in the uterine endometrium, there would be two possibilities for Epo production. One possibility is that uterine target cells gain responsiveness to Epo by the action of E 2 , and the serum Epo derived from the kidney acts on the E 2 -sensitive cells. The other would be that a local site for Epo production exists in the uterine tissues, and the production is induced by E 2 , resulting in an E 2 -dependent increase of Epo concentration in uterine tissues sufficient for activating angiogenesis. To test the latter possibility, we first examined if the in vitro cultured uterus from OVX mouse produced Epo protein and its mRNA in an E 2 -dependent manner. One of the bilateral horns of the uterus prepared from the OVX mouse was cultured in the medium containing E 2 , and the other horn was cultured in the absence of E 2 as a control. Epo production into the culture medium was almost undetectable in the absence of E 2 , whereas the culture with E 2 produced Epo in an E 2 -dependent manner, and the increase was evident at physiological concentrations of E 2 (10 Ϫ8 M) (Fig. 1a). Cycloheximide and actinomycin D completely inhibited E 2 -induced production of Epo by the cultured uterus (Fig. 1b), demonstrating that Epo secreted into the culture media was newly synthesized. Epo mRNA was detected with RT-PCR, and restriction mapping confirmed that the amplified DNA band was the specific product derived from Epo mRNA (data not shown). Epo mRNA was definitely expressed in the uterus cultured with E 2 , but it was undetectable in that cultured without E 2 (Fig. 1b). Actinomycin D completely blocked expression of Epo mRNA, whereas the expression was superinduced by cycloheximide, indicating that the de novo protein synthesis is not needed for induction of Epo mRNA. Epo that supports erythropoiesis is produced by the kidney in a hypoxia-inducible manner (17,18). We cultured the uterus in the absence or presence of 10 Ϫ7 M E 2 in 21, 5, and 2% O 2 , but Epo production was not activated by the low oxygen concentrations (data not shown).
In Vivo E 2 -induced Expression of Epo mRNA in the Uterus-We examined whether or not E 2 injection to the OVX mouse induced Epo mRNA in the uterus. E 2 was given intraperitoneally, and the uterus was removed at intervals to extract RNA. Epo and ␤-actin mRNAs were measured semiquantitatively by setting PCR cycles to be approximately proportional to the band intensity of the amplified DNA. The two upper panels in Fig. 2a show Epo mRNA-and ␤-actin mRNA-derived products from two individual mice, and the lowest panel in Fig. 2a indicates band intensities of Epo mRNA-derived product (n ϭ 5 mice). There was a clear increase of Epo mRNA at 1 h after E 2 injection, and the increase continued for at least 4 h, but at 8 h its level decreased to that of E 2 -uninjected mice. VEGF mRNA was also increased at 1 h after E 2 injection, and thereafter its level was gradually reduced (the lowest panel in Fig. 2b), which was in agreement with the previous finding (10). Three amplified DNA bands (the most upper panel in Fig. 2b) are derived from VEGF mRNAs produced by alternative splicing of the primary transcript (39,40). The major band is derived from VEGF 164 mRNA.
In Vivo Effects of Epo on Uterine Endometrium-To examine the effects of Epo on uterine tissues in vivo and compare with those of E 2 , Epo was injected into one of the uterine bilateral cavities of OVX mouse and saline into the contralateral cavity as a control. E 2 or olive oil (solvent for E 2 ) was given intraperitoneally. At 24 h after injection, we inspected uterine tissue sections under light microscopy. Administration of E 2 to OVX mice caused development of uterine tissues including uterine hypertrophy and endometrial growth (compare Fig. 3c with a  and b). Similar development was found in the uterus after injection of Epo (Fig. 3e). Little enlargement of the uterus occurred in OVX mouse that received saline (compare Fig. 3, a  and d).
A clear difference between E 2 -or Epo-injected mouse and oilor saline-injected mouse was found when endometrial sections were stained with the antiserum against factor VIII, an endothelial cell-specific marker; blood vessel formation was stimulated upon administration of Epo (Fig. 3j) as well as E 2 (Fig.  3h). To demonstrate this stimulation by a more quantitative manner, we counted blood vessels in the sections. Stimulation by Epo was similar to that by E 2 , when vessel density was expressed per field; both increased the density by 210% (Fig.  4a). However, endometrial sections from E 2 -injected mouse were significantly lower in cell density than those from Epoinjected mouse (see Fig. 3, h and j). Thus, the vessel number per endometrial cell increased by 440% upon E 2 injection, whereas the increase by Epo injection was 230% (Fig. 4b). E 2 causes a rapid increase in microvascular permeability, resulting in a lower cell density of the endometrium (10). Moreover, the action of Epo somewhat differs from that of E 2 in the morphology of vessels; most of the vessels induced by Epo were smaller in diameter than those induced by E 2 . Neither development of uterine tissues nor stimulation of blood vessel formation was observed when Epo was pretreated with an excess amount of sEpoR (data not shown).
In Vivo Effects of sEpoR on Uterine Endometrium-The cyclic remodeling of uterine tissues in the murine estrus cycle takes place every 3-5 days. The above-mentioned results suggest that Epo is involved in the estrus cycle-dependent endometrial growth through stimulation of angiogenesis. To demonstrate this possibility, sEpoR or the heat (56°C, 30 min)inactivated sEpoR was injected into one of the uterine bilateral cavities of the non-OVX mouse in diestrus stage and saline into the contralateral cavity. At 24 h (proestrus) after injection, we inspected uterine tissue sections under light microscopy. Transition from diestrus (Fig. 5a) to proestrus (Fig. 5d) caused endometrial growth. This endometrial growth was severely inhibited by the injection of sEpoR (Fig. 5b) but not by the inactivated sEpoR (Fig. 5c). To quantify the effect of sEpoR, the areas of myometrium and endometrium layers were calculated. The endometrium/myometrium ratios in the sEpoR-injected uterus were significantly smaller than those in the inactivated sEpoR-or saline-injected uterus (Table I), indicating that sE-poR was detrimental to the endometrial growth in the transition from diestrus to proestrus. Further detailed inspection of FIG. 4. Epo-and E 2 -induced endometrial angiogenesis in OVX mouse. Sections stained with anti-factor VIII antiserum (Fig. 3, g-j) were used to count vessel density in the endometrium. a, total vessels per mm 2 ; b, total vessels per 10 2 cells. Shaded columns indicate the results of control mice that received injection of olive oil or saline, and black columns indicate those of mice that received injection of E 2 or Epo. Individual vessels were counted on a ϫ 200 field (i.e. ϫ 20 objective lens and ϫ 10 ocular lens; 0.0564-mm 2 per field) that was displayed on a 19 ϫ 19 cm monitor by color video camera. Four to six mice were used for each experiment, i.e. olive oil, E 2 , saline, and Epo. More than 50 sections per uterus and at least 2 fields per section were selected at random for counting. The total number of cells was also counted per section. Results are expressed as mean Ϯ S.E. Asterisks indicate p Ͻ 0.05, significantly different from the values of controls (oil or saline). The statistical significance of differences was determined with Student's t test. magnified sections stained with the anti-factor VIII indicated that the formation of blood vessels in the endometrium is inhibited by sEpoR but not by the inactivated sEpoR (Fig. 5,  e-g).
Expression of EpoR in the Endometrial Endothelial Cells-Immunochemical staining of the uterine section from non-OVX mice in the estrous stage using the antiserum against the extracellular domain of EpoR (Fig. 6, a and b) and anti-factor VIII antiserum (Fig. 6, c and d) showed that the uterine microvascular endothelial cells express EpoR as well as factor VIII. DISCUSSION We found that the uterus is a novel site for Epo production. E 2 has been reported to influence Epo production (41, 42), but the production site responding to E 2 was not known. The major regulator in the uterus appears to be different from that in other Epo production sites. Oxygen is a primary signal for regulation of Epo biosynthesis in the kidney and liver; hypoxia induces transcriptional activation of the Epo gene (reviewed in Ref. 43). This induction requires de novo protein synthesis (43). In contrast, the E 2 -induced increase of Epo mRNA in the

TABLE I sEpoR but not heat-inactivated sEpoR inhibits endometrial growth
The values were calculated from the weight of the myometrial and endometrial areas cut out from the micrographs in Fig. 5, b-d. Ten sections per uterus of mouse (n ϭ 5-7). Each value is the mean Ϯ S.D.  uterus is very rapid and does not require de novo protein synthesis. Furthermore, Epo production in uterus is not induced by hypoxia. E 2 does not induce Epo production in the kidney and brain. 3 This unique regulation of uterine Epo production may ensure the appropriate local function of Epo in the uterus with small, if any, distortion on erythropoiesis and the central nervous system. A data base search of the DNA sequences of Epo genes (44) indicated that the 5Ј-flanking regions of human and mouse genes contain sequences highly homologous to the sequence responsible for E 2 receptor binding. A functional analysis of these sequences with respect to E 2 response is in progress.
Identification of the cells responsible for production of uterine Epo has not been completed. Collagenase digestion of the uterus from OVX mouse and the subsequent fractionation of the dispersed cells suggest that the uterine Epo is produced in an E 2 -dependent manner by the fraction rich in endometrial stromal cells. 4 Injection of Epo into the uterine cavity of OVX mouse caused endometrial alterations, which differed somewhat from those induced by E 2 . The cell density in the endometrium of the mouse given E 2 is apparently lower than that of the mouse given Epo. Presumably the low cell density results from an increased vascular permeability due to stimulation of VEGF production by E 2 (9). A high density of vessels with a larger diameter, which is seen in mice given an E 2 injection, may reflect more extensive proliferation and maturation of the endothelial cells. Taken together with the inhibitory effect of sEpoR on the endometrial transition from the diestrus to proestrus stage, these results indicate that Epo is one of the E 2 -regulated signal molecules that is required for execution of the cyclic angiogenesis in estrus cycle; Epo is essential, although it is not sufficient, for completion of this process. The interplay of Epo with other angiogenic factors, particularly with VEGF, must be studied for better understanding of uterine angiogenesis.
Vascular endothelial cells in the adult still express a large number of hematopoietic cell differentiation markers (8,45), which has been thought to indicate a common developmental lineage between hematopoietic cells and endothelial cells (46). In agreement with this finding, endothelial cells from some sources express EpoR (28,29) and Epo exhibits the in vitro angiogenic activity on these cells (30,31). These findings and the immunochemical detection of EpoR in uterine endothelial cells suggest that Epo directly acts on these cells.
It is hard to speculate that systemically distributable blood Epo is involved in physiological angiogenesis including that in the uterus. This possibility may be excluded by our finding that the uterine tissue produces Epo transiently in an E 2 -dependent manner, and the low ligand affinity to EpoR expressed in endothelial cells (29,30) may at least in part support the hypothesis that blood Epo is not involved in physiological angiogenesis. The ligand affinity to EpoR in endothelial cells is much lower than that in erythroid precursor cells; the K d value of EpoR in endothelial cells is ϳ1 nM (29,30), whereas the K d value of the high affinity site in erythroid cells is ϳ50 pM (38). The reason for the expression of only the low affinity site in endothelial cells is unclear. Since the normal concentration of serum Epo is ϳ15 milliunits/ml (ϳ5 pM) (34), EpoR in endothelial cells appears to be nonfunctional physiologically and only to be a remnant reflecting a memory of a common origin of hematopoietic cells and endothelial cells (8). Although the Epo-producing cells in uterus remain to be identified, probably a paracrine Epo/EpoR system would exist in the uterus so that the E 2 -induced increase in the uterine Epo concentration may be locally high enough to act on endothelial cells through the low affinity site. Presumably expression of only the low affinity site is crucial for preventing unregulated angiogenesis by the action of kidney-derived Epo in the circulation. A similar Epo/ EpoR system may also operate in the ovary where active angiogenesis takes place for follicle maturation and growth of corpus luteum (47), because the ovary produces Epo. 4 Homozygous mice carrying a null mutation in Epo or EpoR gene died around embryonic day 13 due to deficiency of the fetal liver erythropoiesis (15,16). No defect was found in endothelial cells in the EpoR null mouse embryos (15). Thus, Epo is unimportant in vasculogenesis, which occurs in the embryonic stage accompanying differentiation of precursors into endothelial cells, and its importance in blood vessel formation seems to be restricted in the female reproductive organ in adults.