BACKGROUND: We recently showed that vascular endothelial growth factor (VEGF) expression by endometrial glandular epithelial and stromal cells, and endometrial microvascular endothelial cell permeability, an early step in angiogenesis, were rapidly increased by estradiol (E2) administration to ovariectomized baboons. We proposed that estrogen promotes endometrial angiogenesis by regulating VEGF expression by glandular epithelial and stromal cells. In the present study, we developed a co‐culture of human endometrial cells and microvascular endothelial cells to determine whether the regulatory role shown for estrogen on endometrial angiogenesis in vivo in the non‐human primate would be demonstrable in vitro in the human. METHODS AND RESULTS: Human endometrial glandular epithelial and stromal cells were co‐cultured with human myometrial microvascular endothelial cells (HMMECs) and E2. HMMEC tube formation (means ± SEM, % endothelial tube area/total endothelial cell area), an index of angiogenesis, was 65% (P < 0.05) and 2‐fold (P < 0.01) greater in cells co‐cultured with human glandular epithelial cells (54 ± 7%) and glandular epithelial cells plus E2 (66 ± 11%), respectively, compared with medium (33 ± 4%). In contrast, endothelial tube formation was not altered in HMMECs incubated with endometrial stromal cells (32 ± 4%), stromal cells plus E2 (36 ± 2%) or E2 (39 ± 3%). CONCLUSIONS: We propose that estrogen, by regulating expression and secretion of angiogenic factors such as VEGF by glandular epithelial cells of the endometrium, regulates endometrial angiogenesis.
Estrogen has an important role in developing the vascular system within the uterine endometrium during each menstrual cycle (reviewed in Brenner and Slayden, 1994); however, the mechanisms underlying this process are not well understood. Vascular endothelial growth factor (VEGF) has a pivotal role in new blood vessel formation, i.e. angiogenesis, by promoting microvascular endothelial cell proliferation, migration and assembly into new vessels (reviewed in Ferrara and Davis‐Smyth, 1997). We and others recently showed that VEGF mRNA levels were markedly suppressed in endometrial glandular epithelial and stromal cells after ovariectomy of baboons (Niklaus et al., 2002, 2003) and rhesus monkeys (Nayak and Brenner, 2002) and restored by chronic administration of estradiol (E2). Moreover, in ovariectomized baboons, acute administration of E2 rapidly increased endometrial VEGF mRNA expression and endometrial microvascular paracellular cleft width integral to microvascular permeability (Albrecht et al., 2003), an early step in angiogenesis (Dvorak et al., 1995, 1999). We proposed, therefore, that estrogen promotes angiogenesis in the primate endometrium by regulating the expression of VEGF by glandular epithelial and stromal cells (Albrecht and Pepe, 2003).
Although comparable in vivo studies designed to show a cause and effect relationship between steroid hormones and endometrial angiogenesis have not been conducted for ethical reasons in humans, VEGF mRNA and protein are expressed by human endometrial glandular epithelial and stromal cells (Charnock‐Jones et al., 1993; Torry et al., 1996; Shifren et al., 1996; Gargett and Rogers, 2001; Möller et al., 2001), and estrogen stimulated VEGF expression in cultures of human endometrial cells (Charnock‐Jones et al., 1993; Shifren et al., 1996; Huang et al., 1998), through a functional variant estrogen response element (Mueller et al., 2000). It is well established that VEGF increases angiogenic responses, e.g. cell proliferation, in cultures of purified human microvascular endothelial cells (Ferrara et al., 1992). However, because microvascular endothelial cells express the estrogen receptor (Critchley et al., 2001) and estrogen stimulates endothelial cell proliferation (Morales et al., 1995), estrogen may also promote new blood vessel development by acting directly on microvascular endothelial cells. Thus, it is not clear in the human endometrium whether estrogen promotes angiogenesis directly, and/or indirectly via expression of angiogenic factors by particular endometrial cells. Therefore, in the present study, we developed a co‐culture of human endometrial cells and microvascular endothelial cells, as an in vitro model of the morphological and functional interaction that exists between these cell types in vivo, to determine whether the regulatory role shown for estrogen on endometrial VEGF formation and angiogenesis in vivo in the non‐human primate would be demonstrable in vitro in the human.
Materials and methods
Human endometrial tissue collection and cell dispersion
Human endometrial tissue was collected from 21‐ to 45‐year‐old women after hysterectomy because of benign gynaecological conditions, e.g. pelvic pain or uterine fibroids. All patients exhibited menstrual cyclicity and were not on birth control pill, GnRH analogue or ovulation induction medication. The stage of the menstrual cycle was determined by record of menses and endometrial histology. Procedural approval for tissue collection was obtained from the Institutional Review Board of the University of Maryland School of Medicine.
Endometrial specimens were either immediately frozen in liquid nitrogen for VEGF mRNA analysis by RT–PCR, fixed in 10% neutral‐buffered formalin for 24 h and embedded in paraffin for VEGF immunocytochemistry, or enzyme‐dispersed and separated into enriched glandular epithelial and stromal cell populations by a modification of the method of Osteen et al. (1989). Endometrial tissue was minced into 2–3 mm pieces in Dulbecco’s modified Eagle’s medium (DMEM)/F12 (Sigma Chemical Co., St Louis, MO) and digested with 5 ml of a solution containing 4 mg/ml collagenase type 4 (Worthington, Biochemical Corp., Lakewood, NJ), 1 mg/ml hyaluronidase (Sigma), 1 mg/ml protease (Sigma), 200 U/ml DNase I type II (Sigma) and 2% chicken serum (Sigma) in magnesium‐free Hank’s balanced salt solution (HBSS; Gibco, Invitrogen Corp., Grand Island, NY) in an orbital shaking water bath at 37°C for 45–60 min. The resultant cell suspension was then washed in calcium‐ and magnesium‐free HBSS and filtered through 85 µm nylon mesh (Small Parts Inc., Miami Lakes, FL). The retentate containing glandular epithelial fragments was washed from the filter with HBSS, resuspended in DMEM/F12 and stored overnight at 4°C. The stromal cell‐enriched filtrate was treated with DNase for 10 min, layered over a 66% Percoll (Sigma) gradient to remove erythrocytes, washed in DMEM/F12, DNase‐treated and filtered through 20 µm mesh. Cells were then washed in DMEM/F12 and resuspended in DMEM/F12 containing 4% fetal bovine serum (FBS; Hy Clone Laboratories Inc, Logan, UT). Glandular epithelial fragments were digested with collagenase–hyaluronidase–protease–DNase in HBSS at 37°C for 15 min, washed in HBSS, and filtered through 20 µm mesh. The retentate containing gland fragments was digested further for 30–45 min, then washed and resuspended in DMEM/F12 containing 4% FBS. Isolated glandular epithelial and stromal endometrial cells (2–4 × 105 cells/ml) were plated on 15 mm culture wells (Nalge Nunc International, Naperville, IL) in DMEM/F12 supplemented with 4% FBS and incubated at 37°C in a humidified atmosphere of 95% air:5% CO2 for 7 days, with medium change every 2 days.
Culture of human myometrial microvascular endothelial cells
Cryopreserved human uterine myometrial microvascular endothelial cells (HMMECs; UtMVEC‐Myo, Clonetics, Walkersville, MD) were obtained at passage 3 and plated on T 75 flasks at 5 × 103 cells/cm2 in a specialized microvascular endothelial cell growth medium (EGM‐2‐MV, Clonetics) containing 5% FBS, hydrocortisone, ascorbic acid, gentamicin, amphotericin‐B, human epidermal growth factor (hEGF), human fibroblast growth factor (hFGF) and human insulin‐like growth factor‐1 (hIGF‐1), but without hVEGF. The attached HMMECs were grown for a pre‐confluent period of 5–7 days at 37°C in a humidified atmosphere of 95% air:5% CO2 with medium change every 2 days, trypsinized and either cryopreserved or subcultured on T 75 flasks. In subsequent experiments, HMMECs from passages 4–10 were grown for 5–6 days in EGM‐2‐MV with medium change to endothelial cell basal medium (EBM, Clonetics) supplemented with 4% FBS for 24 h.
Co‐culture of endometrial cells and HMMECs: endothelial cell tube formation
Media from glandular epithelial and stromal cell cultures were replaced after 7 days with 0.5 ml of EBM supplemented with 4% FBS and cells incubated for 24 h in the absence or presence of E2 (10–6 or 10–8 mol/l, Sigma). Cell culture inserts (0.2 µm pore size, Anopore, Nalge Nunc International) were coated with 0.15 ml of Matrigel (growth factor reduced without phenol red, BD Biosciences, Bedford, MA) per insert and allowed to polymerize at 37°C for 1 h. HMMECs (25 × 103 cells/0.5 ml) in EBM supplemented with 4% FBS were then added to Matrigel‐coated inserts and co‐incubated in triplicate in wells with endometrial cells for 24 h at 37°C in the absence or presence of E2 and 50 ng/ml VEGF. Incubation of human endometrial cells with either 10–6 or 10–8 mol/l E2 yielded similar results for HMMEC tube formation in our laboratory. Therefore, in the present study with enriched glandular epithelial and stromal cell incubations, 10–6 mol/l E2 was employed.
HMMEC tube formation was assessed by image analysis in 2–3 representative fields of view in each culture well. HMMEC images were captured at a final linear magnification of 40× using an Olympus Inverted Research Microscope (Olympus, Tokyo, Japan) coupled to a Polaroid Digital Microscope camera (Polaroid, Cambridge, MA). Images were prepared in Adobe Photoshop 5.0 and exported to an image analysis software package (IP Lab Scientific Image Processing, Scanalytics, Fairfax, VA) for identification of endothelial cell tube‐like networks and quantification of total endothelial tube and endothelial cell area. The mean percentage tube formation from each treatment group was calculated as the ratio of the endothelial tube area/total endothelial cell area × 100.
Endothelial cell proliferation assay
The Cell Titer 96 Aqueous One Solution Cell Proliferation Assay (Promega, Madison, WI) was used to confirm the responsiveness of HMMECs to the mitogen VEGF. HMMECs (5 × 103 cells/0.1 ml) were plated in triplicate in 96‐well plates in EBM supplemented with 4% FBS and incubated at 37°C in a humidified atmosphere of 95% air:5% CO2 for 24 h. After 24 h, medium was replaced with EBM supplemented with 4% FBS in the absence or presence of increasing concentrations (0.1–10 ng/ml) of hVEGF (BioSource Int., Camarillo, CA) and/or 1 µg/ml anti‐hVEGF antibody (R&D Systems, Inc., Minneapolis, MN). HMMECs were incubated at 37°C in a humidified atmosphere of 95% air:5% CO2 for 4 days with medium change after 2 days. At the end of the 4‐day period, medium was replaced with EBM supplemented with 4% FBS and 20 µl of One Solution Reagent and incubated for 2 h at 37°C. Absorbance was measured at 490 nm using a plate reader. The One Solution Reagent contains MTS tetrazolium (3‐[4,5‐dimethylthiazol‐2‐yl]‐5‐[3‐carboxymethoxyphenyl]‐2‐[4‐sulfophenyl]‐2H‐tetrazolium), which is bioreduced by cells into a formazan product. The absorbance of the formazan product is directly proportional to the number of cells.
Enzyme‐linked immunoassay (ELISA) of VEGF protein
VEGF protein levels in the conditioned medium at the end of co‐culture of glandular epithelial or stromal cells and HMMECs were determined by ELISA using a commercially available kit (R&D Systems), according to the manufacturer’s instructions. Briefly, 0.20 ml of conditioned medium or recombinant human VEGF standard (0–1000 pg/ml) were incubated for 2 h at room temperature in wells coated with monoclonal mouse anti‐human VEGF capture antibody, washed and incubated for 2 h with 0.2 ml of polyclonal goat anti‐human VEGF antibody conjugated to horseradish peroxidase. Wells were then washed and samples incubated for 20 min with 0.2 ml of a 1:1 mixture of colour reagent A (H2O2) and colour reagent B (tetramethylbenzidine) and the reaction stopped by addition of 0.05 ml of 2 mol/l H2SO4. Optical density was then determined at 450 nm (with wavelength correction at 570 nm) using a microplate reader. The limit of sensitivity of the assay was <5 pg/ml, and intra‐ and inter‐assay coefficients of variation were each <10%.
Endometrium. Immunocytochemistry of VEGF was performed as described previously (Hildebrandt et al., 2001). Paraffin blocks of uterine tissue were serially sectioned (4 µm), deparaffinized and rehydrated in graded alcohols. Tissue sections were boiled in 0.01 mol/l Na citrate for 10 min, incubated in 1% H2O2, and blocked in 10% normal goat serum (Protein Block Serum, Dako Corp, Carpinteria, CA). Tissues were incubated overnight at 4°C with goat anti‐human primary antibody to VEGF (AF‐293‐NA, diluted 1:25 in 5% goat serum, specific for the 121, 165 and 189 splice variants; R&D Systems). Following incubation with biotinylated anti‐goat imunoglobulin (Vector Laboratories, Inc., Burlingame, CA), sections were immersed in an avidin–biotin complex solution (Elite Vectastain ABC Kit, Vector Laboratories, Inc.), and incubated with 3,3′‐diaminobenzidine (DAB; 0.2 mg/ml, Sigma) to produce a brown reaction product. Negative controls included omission of the primary antibody or pre‐absorption of primary antibody with 10‐fold excess of human recombinant VEGF peptide (R&D Systems). Sections were counterstained with Harris haematoxylin.
Cell cultures. Glandular epithelial and stromal cells (4 × 105 cells/0.9 ml) were plated on Lab‐Tek II chamber slides (Nalge Nunc International) in DMEM/F12 supplemented with 4% FBS and incubated at 37°C in a humidified atmosphere of 95% air:5% CO2 for 7 days. HMMECs (25 × 103 cells/0.9 ml) were plated on chamber slides in EBM supplemented with 4% FBS and incubated for 1–2 days. Cells were then fixed for 90 min in 10% buffered formalin, rinsed in three changes of 1× phosphate‐buffered saline (PBS), incubated in H2O2 for 10 min to inhibit endogenous peroxidase, and blocked with 10% normal goat serum for 1 h. Glandular epithelial cells, stromal cells or HMMECs were incubated overnight at 4°C in a humidified chamber with mono‐ or polyclonal antibodies to either cytokeratin 7 (Cytokeratin‐7‐Monoclonal DAKO Clone OV‐TL), vimentin (Vimentin‐Monoclonal DAKO Clone VIM 3B4) (1:4000 and 1:6000, respectively, DAKO), flt‐1 (C‐17, 1:500) or KDR/flk‐1 (C‐1158, 1:750, both from Santa Cruz Biotechnology Inc., Santa Cruz, CA). Cells were incubated for 1 h at room temperature with biotinylated anti‐mouse or anti‐rabbit immunoglobulins (Vector), followed by avidin–biotin–peroxidase complex (ABC Elite, Vector) for 1 h at room temperature using DAB as chromagen. Cells were lightly counterstained with Harris haematoxylin for 30 s, clarified in acid alcohol and blued using lithium carbonate. Slides were dehydrated in graded concentrations of ethanol and cleared in xylene. Negative controls included omission of the primary antibody, substitution of goat IgG for primary antibody, and pre‐absorption of primary antibody with 10‐fold excess of control peptide for flt‐1 (Santa Cruz) or F(ab′)2 fragment‐specific IgG (Jackson ImmunoResearch Labs Inc., West Grove, PA).
Competitive RT–PCR of VEGF mRNA
RNA isolation and oligonucleotide primers. Total RNA was isolated from endometrium by guanidine isothiocyanate–caesium chloride and quantified by UV absorption spectrophotometry to permit normalization of VEGF mRNA levels. Oligonucleotide primers were synthesized by Invitrogen Life Technologies, Inc. and based on the hVEGF cDNA sequence (Tischer et al., 1991), as detailed previously (Hildebrandt et al., 2001).
Competitive reference standard (CRS). Homologous RNA fragments containing the same primer‐binding regions but shortened internal sequence with respect to the target RNA for VEGF were prepared as described previously (Riedy et al., 1995; Babischkin et al., 1997). Reverse transcription of total RNA (0.5–3.0 µg) from baboon placenta was performed at 42°C for 60 min in a reaction containing 1 mmol/l each of dNTPs (Invitrogen Life Technologies, Inc.), 200 U of SUPERSCRIPT RNase H‐reverse transcriptase (RT) or Molony murine leukaemia virus RT (Invitrogen Life Technologies, Inc.), 1× RT buffer, 40 U of RNAguard (Amersham Pharmacia Biotech, Piscataway, NJ) and 250 ng of random primers (Invitrogen Life Technologies, Inc). A 5 µl aliquot of the RT reaction was added to separate PCRs containing 0.2 mmol/l dNTPs, 1.25 U of cloned Thermus aquaticus DNA polymerase (Amplitaq, Perkin‐Elmer Corp/Cetus, Norwalk, CT), 1× PCR buffer and 10 pmol of paired primers to generate cDNA template for VEGF. PCR was performed in a programmable thermal cycler (MJ Research, Inc., Cambridge, MA) for 25 sequential cycles. An aliquot of each reaction was subjected to 2.0% agarose gel electrophoresis, amplified products gel purified, and the CRS synthesized using the MEGAscript T7 transcription kit (Ambion, Inc., Austin, TX). The cDNA templates were extracted with chloroform:isoamyl alcohol, and aliquots of CRS quantitated spectrophotometrically.
RT–PCR assay. VEGF was quantified by competitive RT–PCR assay (Babischkin et al., 1997; Niklaus et al., 2002). A constant amount of RNA (10 ng) was added to an RT mixture containing serial dilutions of VEGF‐CRS (25–3 attomol). In all experiments, the presence of possible pseudogene or genomic DNA contamination was checked by control reactions in which either the RT enzyme or RNA was omitted.
A 5 µl aliquot of the RT mixture was added to separate PCR mixtures containing 10 pmol of the primers for VEGF. Endometrial samples were amplified for 34 sequential cycles, and PCR products gel fractionated, visualized with ethidium bromide and photographed using type 665 positive/negative film (Polaroid Crop, Cambridge, MA).
Negatives were scanned using a Gel Doc 1000 imaging system and Multi‐Analyst software program (Bio‐Rad Laboratories, Hercules, CA). The intensity of amplified target and CRS cDNA products was represented as the relative area under each product band. The logarithm (log) of the ratio of CRS to target area was plotted as a function of the log concentration of VEGF added to each PCR. The concentration of VEGF target mRNA was determined where the ratio of the log of CRS and target area was equal to 0 (i.e. the equivalence point).
Data were expressed as the means ± SEM and analysed either by Student’s t‐test or by ANOVA with post hoc comparisons of means by Newman–Keuls multiple comparisons test.
Endometrial VEGF mRNA expression
Using primers which spanned the alternative splice site to generate the multiple mRNA transcripts, two VEGF mRNA species that encode the 121 (434 bp) and 165 (566 bp) amino acid isoforms were expressed in high level and the 145 (506 bp) and 189 (638 bp) isoforms in low level within the human endometrium (Figure 1A), as reported previously (Torry et al., 1996).
VEGF mRNA levels were quantified by competitive RT–PCR, using primers upstream from the alternative splice site, to yield a 323 bp product that reflected collective expression of all the VEGF isoforms, in whole endometrial tissue obtained in the proliferative and secretory phases of the menstrual cycle. The slopes of the log of CRS and target areas plotted as a function of the log of increasing CRS concentrations were similar in tissue obtained during the proliferative (correlation coefficient, r2, determined by linear regression was 0.90, P < 0.01) and secretory (r2 = 0.95, P < 0.01) phases of the menstrual cycle. Levels of the 323 bp VEGF mRNA product were similar in endometrial tissue obtained during the proliferative (702 ± 197 attomol/µg total RNA) and secretory (714 ± 92 attomol/µg total RNA) phases of the menstrual cycle (Figure 1B).
Endometrial VEGF immunocytochemistry
VEGF protein expression was abundant in glandular epithelial cells of the human endometrium (Figure 2A). VEGF protein was also present in the stroma, although the intensity appeared lower than in glandular epithelial cells. Specificity of VEGF immunocytochemistry was evident by the absence of staining when primary antibody was pre‐absorbed with recombinant VEGF (Figure 2B).
HMMEC flt‐1 and KDR/flk‐1 immunocytochemistry
The VEGF flt‐1 (Figure 3A) and KDR/flk‐1 (Figure 3B) receptors were expressed by HMMECs as assessed by immunocytochemistry. Specificity was demonstrated by absence of staining when the primary antibodies were pre‐absorbed with recombinant flt‐1 or KDR/flk‐1 or when goat immunoglobulins replaced primary antisera (Figure 3C).
Recombinant hVEGF stimulated the proliferation of HMMECs in a dose‐dependent manner from a mean (±SEM) of 0.28 ± 0.04 absorbance units in the absence of VEGF to 0.76 ± 0.10 in the presence of 10 ng/ml VEGF (P < 0.01, Figure 4). The increase in proliferation of endothelial cells observed with VEGF was blocked by addition of neutralizing VEGF antibody (Figure 4). Thus, HMMECs responded to VEGF with respect to cell proliferation under the culture conditions employed in this study.
Effect of E2 and human endometrial cells on HMMEC tube formation
Figure 5 is a photomicrograph illustrating tube formation by HMMECs incubated in the presence of 50 ng/ml hVEGF. HMMECs placed on Matrigel‐coated inserts migrated, aligned and formed elongated tubular structures, as an index of angiogenesis under the present culture conditions.
Figure 6 shows cytokeratin and vimentin immunocytochemistry in glandular epithelial and stromal cells isolated by filtration from human endometrium. An enriched population of glandular epithelial cells (∼80%) that exhibited cytoplasmic immunostaining for epithelial cell‐specific cytokeratin (Figure 6A) and little immunoreactivity for mesenchymal cell‐specific vimentin (Figure 6B) was obtained. Moreover, a relatively homogenous population of stromal cells (∼99%) that showed considerable immunoreactivity with the antibody for vimentin (Figure 6D) and little staining with the antibody for cytokeratin (Figure 6E) was obtained. Specificity was confirmed by absence of immunocytochemical reactivity in glandular epithelial (Figure 6C) or stromal (Figure 6F) cells incubated with goat immunoglobulins.
The effect of isolated human endometrial cells and estrogen on tube formation by microvascular endothelial cells, expressed as mean ± SEM percentage tube area/total endothelial cell area, is shown in Figure 7. Human recombinant VEGF alone (50 ng/ml) increased HMMEC tube formation by ∼65% (P < 0.03, t‐test) to 50 ± 6%, indicating the responsivity of these cells under the present culture conditions (not shown in Figure 7). Endothelial cell tube formation was similar when HMMECs were cultured with medium alone (33 ± 4%), E2 (39 ± 3%), human endometrial stromal cells (32 ± 4%) or stromal cells and E2 (36 ± 2%). However, tube formation by endothelial cells co‐cultured with human glandular epithelial cells alone was ∼65% greater (54 ± 7%, P < 0.05) and with glandular epithelial cells and E2 2‐fold greater (66 ± 11%, P < 0.01) than with medium alone.
VEGF protein levels in co‐cultures of human endometrial cells and HMMECs
The levels of VEGF protein quantified by ELISA in conditioned medium from the co‐cultures of human endometrial cells and HMMECs are shown in Figure 8. VEGF protein was not detectable in cultures of HMMECs and medium alone or HMMECs and E2. Mean (±SEM) VEGF protein levels were 276 ± 76 and 366 ± 110 pg/106 cells, respectively, in the presence of human stromal cells or stromal cells plus E2. VEGF protein levels were 491 ± 231 in the medium of HMMECs cultured with glandular epithelial cells. The highest level of VEGF protein was observed in HMMECs and glandular epithelial cells cultured with E2 (693 ± 377 pg/106 cells).
The results of the present study show that E2 significantly promoted tube formation by microvascular endothelial cells cultured with an enriched population of human glandular epithelial cells. In contrast, microvessel tube formation was not significantly altered in endothelial cells incubated with E2 alone. This suggests that estrogen stimulated the secretion of a product(s) by glandular epithelial cells that in turn promoted tube formation by the endothelial cells. VEGF protein was abundantly expressed by glandular epithelial cells in the endometrium of women in the present study and was present at relatively high levels in the medium of co‐cultures of human glandular epithelial cells and HMMECs incubated with E2. Moreover, we (Albrecht et al., 2003) recently showed that E2 administration to ovariectomized baboons rapidly stimulated VEGF mRNA expression by glandular epithelial cells and endometrial microvessel paracellular cleft width/vascular permeability, a process integral to angiogenesis. Estrogen also enhanced VEGF mRNA and/or protein expression in vivo in the rhesus monkey (Nayak and Brenner, 2002), rat (Cullinan‐Bove and Koos, 1993; Hyder et al., 2000) and sheep (Reynolds et al., 1998) uterus and in vitro in human endometrial cells (Charnock‐Jones et al., 1993; Shifren et al., 1996). Collectively, on the basis of the results of these in vivo and in vitro studies and the present co‐culture study, we suggest, as illustrated in Figure 9, that estrogen, by stimulating expression and secretion of angiogenic factors such as VEGF by glandular epithelial cells of the endometrium, regulates endometrial microvascular tube formation and thus angiogenesis. Furthermore, we suggest that co‐culture of endometrial and microvascular endothelial cells provides a sound in vitro system to determine whether the regulatory role shown for steroid hormone on endometrial angiogenesis in vivo in the non‐human primate is operable in the human. Because other factors, e.g. EGF, of endometrial cell origin appear to mediate the estrogen‐induced increase in microvessel cell migration (Sandberg et al., 2001), additional studies with the present co‐culture system, e.g. using an antagonist of VEGF or its receptors, are needed to show definitively that the stimulatory effect observed for E2 on microvascular endothelial cell tube formation requires and thus is linked to endometrial VEGF.
In contrast to the increase in vascular endothelial tube formation observed with E2 and human glandular epithelial cells in the current study, endometrial stromal cells in the presence or absence of E2 did not significantly alter microvascular tube formation. The absolute level of and relative increase in VEGF mRNA expression observed with acute E2 treatment of ovariectomized baboons were also most pronounced in endometrial glands (Albrecht et al., 2003). The absolute level and relative responsivity to hypoxia of VEGF protein secretion in vitro was also greater in human glandular epithelial than in stromal cells (Sharkey et al., 2000). Therefore, estrogen‐dependent expression of glandular epithelial VEGF may have a major role in regulating angiogenesis, e.g. in the immediately subjacent region of the endometrium where a rich subepithelial microvascular network develops. However, VEGF protein was expressed by endometrial stromal cells of women in the present study and was present in the medium of and thus apparently secreted by human stromal cells co‐cultured with HMMECs. Moreover, E2 significantly increased stromal cell VEGF mRNA expression in vivo in the baboon (Albrecht et al., 2003; Niklaus et al., 2003) and in vitro in the human (Shifren et al., 1996) endometrium. Thus, estrogen‐dependent synthesis of VEGF or other factors locally within the stroma may also play a role in mediating angiogenesis in the uterus. The reason(s) for the lack of stimulation of endothelial tube formation despite the presence of VEGF protein in cultures of human stromal cells is unknown, but may reflect the lower level of VEGF expression displayed by stromal cells compared with glandular epithelial cells and/or the concomitant secretion of angio‐inhibitory factors by stromal cells. Additional study is needed to assess these possibilities.
Although E2 alone did not significantly increase myometrial microvessel endothelial cell tube formation in the present study, estrogen receptors are expressed in endothelial cells (Critchley et al., 2001), and estrogen promoted proliferation of and tube formation by human umbilical vein endothelial cells in culture (Morales et al., 1995). These apparently disparate results may reflect the very different endothelial cell types and cell culture conditions that were employed in the two studies. Under certain conditions, VEGF may be expressed by endothelial cells (Concina et al., 2000), or by neutrophils focally in association with microvessel endothelial cells (Gargett and Rogers, 2001). Therefore, it has been proposed that cells within the vasculature are in vivo sources of angiogenic factors for non‐sprouting angiogenesis, i.e. intussusception and elongation, within the endometrium (Gargett and Rogers, 2001).
As shown in the present study and previously by others (Li et al., 1994; Lau et al., 1998; Charnock‐Jones et al., 2000), VEGF mRNA levels were similar in endometrial tissue collected during the proliferative and secretory phases of the human menstrual cycle, despite cyclical surges in E2. Thus, it has been suggested that estrogen has no role in endometrial VEGF expression or angiogenesis (Smith, 1998; Gargett and Rogers, 2001), although the protein levels should be taken into account because VEGF mRNA stability may be altered in response to estrogen. We suggest that the levels of estrogen, although low, immediately preceding and following the late proliferative–midcycle surge in E2, are nevertheless sufficient and necessary to maintain endometrial VEGF expression to progressively promote angiogenesis. Thus, it may only be when estrogen is very low, e.g. after ovariectomy of non‐human primates (Niklaus et al., 2002, 2003; Albrecht et al., 2003), early in the proliferative phase or after removal of endometrial tissue from in situ for cell culture as in the present study, that VEGF synthesis declines and endometrial cells become responsive to exogenous estrogen. The less marked, but nevertheless significant, increase in HMMEC tube formation elicited by human glandular epithelial cells alone in cultures of the present study may reflect continued VEGF synthesis by endometrial cells previously upregulated by the estrogen environment in situ.
In the present study, we focused on the regulatory role of estrogen, because progesterone was not effective in stimulating endometrial VEGF expression or microvessel permeability in baboons (Albrecht et al., 2003). However, progesterone increased VEGF mRNA expression invivo in the rat uterus (Cullinan‐Bove and Koos, 1993) and in vitro in human endometrial cells (Shifren et al., 1996). Thus, additional study with the co‐culture of human endometrial and microvascular endothelial cells is needed to determine the potential regulatory role of progesterone in angiogenesis in this system.
In summary, we established a co‐culture of human endometrial cells and microvascular endothelial cells to determine whether the regulatory role shown for estrogen on endometrial angiogenesis in vivo in the non‐human primate is operable in vitro in the human endometrium. The results of this study show that E2 significantly promoted tube formation by microvascular endothelial cells co‐cultured with human glandular epithelial cells. On the basis of the results of the current study and of other in vitro and in vivo studies, we suggest that estrogen, by regulating expression and secretion of angiogenic factors such as VEGF by glandular epithelial cells of the endometrium, regulates endometrial microvascular endothelial cell tube formation and thus angiogenesis.
The authors gratefully acknowledge the assistance of Dr Antonino Passaniti with assessing microtubule formation in vascular endothelial cells, and Ms Kathleen Daugherty and Dr Dennis Putney for the endometrial cultures and analysis of VEGF mRNA. We sincerely appreciate the secretarial assistance of Mrs Wanda James with the manuscript. This study was supported by NIH U54 HD36207 as part of the NICHD Specialized Cooperative Centers Program in Reproduction Research.