Modulation of hepatocyte growth factor secretion in human female reproductive tract stromal fibroblasts by poly (I:C) and estradiol

Hepatocyte Growth Factor (HGF) was discovered as a mitogen of hepatocytes more than 25 years ago 1 , 2 . We now know that HGF is a pleiotropic agent produced by stromal cells that induces epithelial cell proliferation, motility, morphogenesis as well as angiogenesis in several organs 2 . HGF initiates its effects on epithelial cells through tyrosine phosphorylation of its receptor, c-Met. In addition to organ development, HGF has protective effects on cells and tissues through wound-healing as well as anti-apoptotic and anti-inflammatory actions. Epithelial-stromal cell interactions are essential in facilitating steroid hormone-induced growth and development in the endometrium 3 , 4 . Interestingly, estradiol-induced actions on epithelial cells may be indirect, and estradiol-induced HGF from underlying stromal cells may be the primary mediator of estradiol’s effects on the epithelial cells. Cooke et al used FRT tissues from estrogen receptor (ER) knockout mice to demonstrate that underlying estrogen receptor positive stromal cells regulate the differentiation of the adjacent epithelial cells 3 . These studies indicated that estradiol acts on stromal cells to release one or more paracrine factors that then modulate estradiol’s effects on epithelial cells for growth and differentiation. Many actions of HGF are mirrored by estradiol in the female reproductive tract (FRT). We have shown that some of estradiol’s effects on mouse uterine epithelial cells are mediated via HGF secreted by stromal cells in response to estradiol stimulation 5 . For example, HGF was identified as the key paracrine mediator of TNFα production by uterine epithelial cells since TNFα secretion was abrogated by the addition of blocking antibodies to either c-Met or HGF 6 . Sugawara et al 7 also have identified HGF as a critical factor in estradiol-mediated regulation of uterine epithelial cells. We recently showed that estradiol treatment of human primary uterine stromal fibroblasts results in increased secretion of HGF 8 , again implying a relationship between estradiol and HGF. In addition to similarities between endogenous HGF and estradiol in normal FRT physiology, several investigators have reported a parallel of HGF and estradiol in pathologies of the FRT. In reproductive tissue cancers and endometriosis, relatively high estradiol concentrations are correlated with increased HGF levels 9 – 17 . Uterine and Fallopian tube epithelial cells are known to express Toll-like Receptors (TLR) as does the lower FRT 18 . TLR are designed to recognize pathogen-associated molecular patterns (PAMP) that are characteristic of various pathogens, such as lipopolysaccharide (LPS), a component of gram-negative bacterial cell walls, and double-stranded RNA, associated with viral infection. Upon activation, TLR stimulate the secretion of cytokines and chemokines, as well as both extra- and intra-cellular anti-microbicidal agents, such as secretory leukocyte protease inhibitor (SLPI) 19 and 2’, 5’ OAS 20 , respectively. Whereas the response to TLR agonists has been reported in immune cells and FRT epithelial cells, relatively little research has been reported with TLR stimulation of FRT stromal fibroblasts in the presence of estradiol. The goal of this study was to determine if TLR stimulation of FRT stromal fibroblasts alters HGF secretion, which then may have effects on innate immune functions of FRT epithelial cells. We also wanted to determine whether the response, if any, was differentially modulated in the upper vs lower FRT (uterine, cervical, and ectocervical compartments). Finally, we wished to compare the magnitude of any TLR3-induced HGF increase to that induced by estradiol, as well as any modulation of HGF secretion of FRT stromal fibroblasts to the co-stimulation of FRT stromal fibroblasts with a TLR3 agonist and estradiol.

To determine if ectocervical, cervical and uterine stromal fibroblasts secrete HGF, fibroblasts from each tissue were isolated and grown in primary culture to confluence. Conditioned stromal media (CSM) were recovered from a minimum of 4 wells after 48 hours and individually assayed for HGF by ELISA. As shown in Figure 1 , stromal fibroblasts from all three tissues constitutively secrete HGF, although levels vary among individual donors tested. Values were assessed after stromal fibroblasts grew to confluence in culture for 1–2 weeks with media changes every 48 hours, so the effect of endogenous hormones on HGF secretion would be eliminated. Also, since the stromal fibroblasts were cultured in media supplemented with stripped FBS, there should be no effect from exogenous steroid hormones. There was no significant difference in the mean values of HGF secretion from stromal fibroblasts derived from the uterus, endocervix and ectocervix. In other studies, we observed that constitutive HGF secretion was consistent over time when uterine stromal fibroblasts were cultured for 10 days during which 48 hour conditioned media were obtained and tested for HGF. There was no difference in the levels of secreted HGF assessed at 2, 4, 6, 8 and 10 days (data not shown). We have previously reported that estradiol stimulates uterine stromal fibroblasts to secrete HGF from 10 donor patients 8 . In order to better understand these results, we explored the effect of two different estradiol doses on HGF secretion. As seen in Figure 2 , addition of estradiol for 48 hours more than doubled HGF secretion at concentrations of 10 −8 M and 10 −7 M, indicating that physiological non-saturating and saturating receptor doses elicit an HGF response. In other studies, to determine if uterine stromal fibroblasts can respond to repeated doses of estradiol, media containing 10 −8 M estradiol were completely removed from the cells after 48 hours and replaced with fresh media containing 10 −8 M estradiol. This was repeated after another 48 hours in culture. As shown in Figure 3 , HGF induced by successive estradiol treatment at 2, 4 and 6 days showed a trend toward increasing HGF production, perhaps related to receptor induction and saturation or other mechanisms. Since estradiol increased HGF secretion by uterine stromal fibroblasts, we determined if estradiol would have a similar effect on endo- and ecto-cervix cells. Figure 4 shows that estradiol increases HGF secretion from uterine stromal fibroblasts, but not from stromal fibroblasts derived from the endo- and ecto-cervix of the same patient. Similar results were obtained from 4 other patients. Since TLR agonists induce innate immune functions in epithelial cells, and HGF also regulates aspects of epithelial cell function, studies were undertaken to determine if treatment of uterine stromal fibroblasts with a panel of TLR agonists would alter the secretion of HGF. Uterine stromal fibroblasts were treated with a panel of specific TLR agonists, including LPS (TLR 4), Poly (I:C) (TLR 3), Pam3Cys (TLR 2), Flagellin (TLR 5), Zymosan (TLR 2), Peptidoglycan (TLR 2), Imiquimod (TLR 7,8) 24 , and heat-killed Listeria monocytoides (HKLM) (TLR 2) 25 . Most TLR agonists had no effect on HGF secretion ( Figure 5 ). However, treatment of human uterine stromal fibroblasts with the TLR 3 agonist Poly (I:C), which mimics viral double-stranded viral RNA, at 25 µg/ml 24 , resulted in significant HGF secretion compared to control at 48 hours. Since Poly (I:C) increased HGF secretion from uterine stromal fibroblasts, we selected Poly (I:C) for further studies with FRT stromal fibroblasts. Figure 6 shows the HGF secretion from stromal fibroblasts isolated from 3 FRT tissues of one donor following treatment with Poly (I:C). The TLR3 ligand significantly increased HGF secretion in UT, CX and ECX. This is in contrast to the effects of estradiol treatment, which increased HGF secretion by uterine stromal fibroblasts only (see Figure 2 ). Of the 3 FRT tissues studied, stromal fibroblasts from the ectocervix and uterus secreted the most HGF, both constitutively and in response to Poly (I:C). In other studies, stromal fibroblasts from all 3 FRT tissues expressed mRNA for TLR3, with the ectocervix expressing approximately twice that of the cervix and uterus (data not shown). The stromal fibroblasts from only 3 of 5 donors of cervical tissue responded to Poly (I:C) with an increase in HGF secretion, while stromal fibroblasts from the ectocervix and the uterus of all donors produced HGF (5/5). To compare estradiol-induced HGF secretion with that of Poly (I:C)-induced secretion, stromal fibroblasts from 3 FRT tissues were treated with 10 −8 M 17-β estradiol, a physiological dose of estradiol, or 25 µg/ml Poly (I:C) or both for 48 hours. As shown in Figures 7(A–C) and in previous Figures, both Poly (I:C) and estradiol individually enhanced HGF secretion only by uterine stromal fibroblasts. The Poly (I:C)-induced increase in HGF secretion from uterine stromal fibroblasts was significantly greater that that seen with estradiol. Co-treatment showed an additive HGF response compared with either estradiol or Poly (I:C) alone ( Figure 7A ). Since estradiol did not increase HGF secretion by either endo- or ecto-cervix stromal fibroblasts ( Figures 4 , 7(B), 7(C) , co-treatment with estradiol and Poly (I:C) showed only an equivalent response to Poly (I:C) alone ( Figures 7 (B) and (C) ).

Sex hormones play a key role in regulating immunity in the FRT 26 – 31 , and we have previously shown that estradiol increases HGF secretion by human uterine stromal fibroblasts 8 . Our results extend our prior observation to demonstrate that stimulation of primary human uterine stromal fibroblasts with the TLR 3 agonist Poly (I:C), the synthetic double stranded analog for viral RNA, significantly increases HGF secretion. The Poly (I:C)-induced HGF increase is more extensive in uterine stromal fibroblasts than that induced by estradiol; co-treatment of uterine stromal cells results in an additive effect in terms of HGF secretion. The additive response to combined estradiol and Poly (I:C) treatment in uterine stromal fibroblasts may reflect the cyclical relationship between estradiol levels and uterine morphology, as well as suggest that estradiol and Poly (I:C) increases HGF by different mechanisms. Poly (I:C) treatment of ECX stromal fibroblasts also results in increased HGF secretion, though combined estradiol and Poly (I:C) treatment does not differ from Poly (I:C) treatment alone. Poly (I:C) treatment of CX stromal fibroblasts resulted in variable HGF secretion; three of five donors increased HGF secretion while the other two were unresponsive. Co-treatment of these cells with Poly (I:C) and estradiol also had no effect on HGF secretion in cells from these two patients. Published reports have identified individual types of TLR on fibroblasts, though studies measuring a full panel of TLR-specific RNA and protein expression are lacking on stromal fibroblasts. There are few published reports concerning TLR expression in stromal fibroblasts of the FRT. Human Fallopian tube stromal fibroblasts have been reported to express TLR 4 32 . TLR 4 mRNA message and protein expression have been reported in human uterine stromal fibroblasts 33 . Hirata et al demonstrated that uterine stromal cells expressed TLR 4 mRNA via RT-PCR and in situ hybridization, as well as secreted IL-8 in response to LPS stimulation 34 . The stromal cells in that study were cultured to confluence for only two or three days and characterized only as CD14-positive; therefore, it is not clear what percentage of the stromal cells were fibroblasts and what leukocytes might be present. Uterine stromal fibroblasts were not LPS-responsive in our experiments, which may suggest TLR 4 inactivity or absence; perhaps the lack of an LPS-binding protein in our stripped FBS media prevented a response to LPS. We found that uterine and ectocervical stromal fibroblasts significantly increased HGF secretion in response to Poly (I:C) treatment, indicating that TLR 3 is both present and functional in uterine and ectocervical stromal fibroblasts. However, cervical stromal fibroblasts had variable responses to Poly (I:C) (3/5). This may be explained by patient variability, but conclusions cannot be drawn without a larger pool of donors. This finding does suggest the presence of functional TLR 3 in FRT stromal fibroblasts, at least from some patients. To the best of our knowledge, this study is the first to demonstrate TLR 3 mediated modulation of HGF in the human FRT. HGF has modulatory effects on migration, mitosis, and morphogenesis of epithelial cells (reviewed in 2 ). HGF may also be involved with the innate immune response to HIV. HGF has been shown to increase epithelial cell expression of the receptor CXCR4, an important cofactor in HIV infection eceptor 35 , 36 , in human breast cancer cell lines 37 . Increased HGF secretion has been associated with oncogenic-type HPV and HIV co-infection in the cervix 38 . It is also possible that HGF may link some viruses to development of cancer by the following mechanism: exposure of stromal fibroblasts to oncogenic viruses may increase HGF secretion, leading to metastatic cancer characterized by overexpression of epithelial c-Met or other epithelial receptors involved in viral infection. For example, a receptor (CXCR4) for a different factor, stromal derived factor-1 (SDF-1), is expressed on epithelial cells of the FRT 39 , 40 . Importantly, HGF has been demonstrated to regulate human cervical carcinoma, a pathology of viral origin, in conjunction with SDF-1 15 . Also, replication of SV40 in human mesothelial cells induces HGF/c-Met receptor activation 41 . Significantly, these studies link HGF with viral infection. The results of these studies demonstrate a tissue-specific Poly (I:C) response; uterine and ectocervical stromal fibroblasts increase HGF secretion when treated with Poly (I:C), though cervical stromal fibroblasts from two of five did not. TLR 3 mRNA is expressed in uterine, cervical and ectocervical stromal fibroblasts from at least one donor; the fact that all three tissues responded with an increase in HGF secretion to Poly (I:C) indicates that the TLR 3 is functional in stromal fibroblasts from this donor. Our studies suggest that a combination of estradiol secretion and viral infection on uterine stromal fibroblasts may enhance HGF production. These results suggest the potential that HGF may induce uterine epithelial cells to spread to other tissues and possibly enhance, under certain conditions, the development of endometrial cancer and endometriosis. Therapies to control HGF secretion and/or c-Met expression by stromal fibroblasts due to hormonal or viral causes have been suggested to control these serious diseases 42 .

FRT mucosal tissue was obtained following surgery from women who underwent hysterectomies at Dartmouth-Hitchcock Medical Center. Tissues used in this study, including uterine, cervical, and ectocervical tissues, were distal to the sites of pathology and were determined to be unaffected by disease upon inspection by a trained pathologist. Pathologists also determined the menstrual status, as well as the stage in the cycle of pre-menopausal donors. Tissues were transported from Pathology and procedures to prepare purified epithelial sheets and stromal cells began within 2 hours of surgery. Approval to use tissues was previously obtained from the IRB. Uterine, cervical and ectocervical cells were dispersed into epithelial and stromal cell fractions by enzymatic digestion as described previously 21 . Briefly, tissues were minced under sterile conditions into 1 – 2 mm fragments and then digested at 37°C for 1 hour using a "PHC" enzyme mixture that contains final concentrations of 3.4 mg/ml pancreatin (Gibco Life Technologies, Rockville, MD), 0.1 mg/ml hyaluronidase (Worthington Biochemical Corporation, Freehold, NJ), 1.6 mg/ml collagenase (Worthington) and 2 mg/ml D-glucose (Sigma, St Louis, MO) in Hank's balanced salt solution (HBSS, Gibco, Invitrogen). This procedure was chosen to maximize digestion of the extracellular matrix, as verified by microscopy of hematoxylin and eosin stained frozen sections after digestion, while minimizing digestion of cell surface antigens as determined by flow cytometry. After incubating in this enzyme cocktail, uterine, cervical, and ectocervical cells were collected, washed with media consisting of DMEM/F-12 without phenol red (Gibco), 2 mM fresh L-glutamine, 25 mM HEPES, 50 ng/ml primocin (InvivoGen, San Diego, CA) and separated from tissue aggregates by sedimentation for 5 minutes. Medium containing predominantly single cells and small epithelial cell glands was removed and dispersed through a 250µm nylon mesh screen (Small Parts, Miami Lake, FL). To further separate epithelial and stromal cells, tissue fragments were re-suspended in complete media. Dispersed uterine cells were then pooled and separated on a 40 mm mesh screen (Small Parts) into fractions containing epithelial cell glands and a mixed population of sub-epithelium stromal cells. Stromal cells passed through this fine mesh whereas small epithelial cell glands did not. Stromal cells were centrifuged at 500×g for 5 minutes in a Beckman-Coulter (Fullerton, CA) centrifuge and resuspended in complete media. Stromal cells were cultured in T75 flasks (Falcon) in an atmosphere of 95% air and 5% CO 2 at 37°C in an InCuSafe cell culture incubator (Sanyo Scientific, Wood Dale, IL). Stromal fibroblasts were grown to confluence in T75 flasks, cultured in complete media consisting of DMEM/F-12 without phenol red (Gibco) supplemented with 10% heat inactivated fetal calf serum (Hyclone, Logan UT), 2 mM fresh L-glutamine, 25 mM HEPES, 50 ng/ml primocin (InvivoGen, San Diego, CA). Once stromal fibroblasts reached confluence, they were trypsinized with 0.05% trypsin EDTA (Invitrogen), counted for viability via trypan blue exclusion, then seeded into 24 well plates at a density of 5 × 10 5 cells/well (1 × 10 6 cells/ml) and cultured in complete media. Twenty four hours after seeding, wells were washed with HBSS and agitated to remove remaining red blood cells and other non-fibroblastic stromal cells. After four days of culture, all cells had the characteristic fibroblast morphology, and the culture was devoid of other cell types 22 . Media was replaced at 48-hour intervals. Once the cells reached confluence, they were cultured in media containing 10% stripped FBS and allowed to acclimate to this media for 48 hours prior to treatment with estradiol. Charcoal-dextran stripped FBS was substituted for defined FBS, because there are very low, yet measurable levels of estradiol (10 −11 M) present in defined FBS (Hyclone). 17β-estradiol (CalBiochem, San Diego, CA) was weighed and diluted in ethanol to a concentration of 10 −3 M. 20µl of this stock solution was evaporated in a clean glass scintillation vial and resuspended in culture media to make a working stock solution of 10 −5 M; specific treatment concentrations were diluted from the working stock as needed. A control media was made using 20µl 100% ethanol evaporated in a clean glass scintillation vial and resuspended in culture media, parallel to the estradiol solution. Fresh steroid preparations were made for each experiment, and all treatment media needed for a specific culture/experiment was made at the time of estradiol preparation. Stromal fibroblasts were grown to confluence in 24 well plates, and then switched to media containing 10% stripped FBS. After 24 hours, the media were replaced and the fibroblasts were treated. Conditioned stromal medium (CSM) was collected at 48 hours, centrifuged and stored at −80° and assayed for HGF by ELISA. Quadruplicate wells of stromal fibroblasts were treated with estradiol at a dose of 10 −8 M, with a control treatment made with the ethanol control media. The secretion of HGF was tested over a period of eight days. This time course covered the collections that have yielded significant HGF secretion, while minimizing the risk of over-culturing the stromal fibroblasts and altering the primary phenotype. Unless otherwise indicated, stromal fibroblasts were treated with 10 −8 M estradiol. This is the standard estradiol treatment concentration used by our laboratory and is within the physiological range of estradiol concentration 23 . Since FRT cells are constantly exposed to estradiol throughout the menstrual cycle, in some cases estradiol was added for a second 48 hour treatment. To determine if estradiol treatment increased the number of uterine stromal fibroblasts, confluent cultures in 24 well plates treated with or without estradiol at a concentration of 10 −8 M were washed with HBSS, trypsinized with 0.05% trypsin/EDTA; individual wells were collected in individual microfuge tubes and washed with culture media. Cells were assessed for viability via trypan blue dye exclusion and counted manually with a hemocytometer. Our results indicated that estradiol did not increase cell numbers of the stromal fibroblasts in any experiment (data not shown). For these experiments, cultures were treated with a panel of TLR agonists. The TLR-agonists used included 1 µg/ml LPS (TLR 4) (InVivogen),, 1 µg/ml Pam3cys-Ser-(Lys)4 (TLR2) (EMC Microcollections, Tubingen, Germany), 25 µg/ml Poly I:C (TLR3) (InVivogen), 100 ng/ml flagellin isolated from E. coli (TLR5) (Inotek Pharmaceuticals, Beverly, MA), 10 µg/ml Zymosan isolated from Saccharomyces cerevisiae (TLR2) (InVivogen), 10 µg/ml peptidoglycan isolated from S. aureus (TLR2) (InVivogen), 5 µg/ml Imiquimod (TLR7, 8) (InVivogen), and 10 8 heat-killed Listeria monocytogenes (HKLM, TLR2) (InVivogen). We have shown these doses to be effective in stimulation of human uterine epithelial cell TLR 24 . In further experiments, uterine stromal fibroblasts were treated with both estradiol at 10 −8 M and Poly (I:C) at 25 µg/ml. Treatment groups consisted of four wells per treatment. After 48 hours, conditioned stromal media (CSM) were collected, spun at 10,000×g to remove cellular debris, and stored at −80° C. In other studies, the effect of both estradiol and TLR treatment were assessed. Treatment groups included control, estradiol-treatment (10 −8 M), TLR agonist treatment, and co-treatments of estradiol plus TLR agonist. Time-course experiments over six to eight days of treatment with estradiol and TLR agonists were performed. Supernatants from each well were collected at 48-hour intervals, centrifuged at 10,000× g in a microfuge (Eppendorf, Westbury, NY) to remove any cellular debris, and stored in a −80°C freezer (Revco Scientific, Asheville, NC) until assayed. Culture supernatants were analyzed for HGF via a commercially available ELISA duoset (Quantikine; R&D system, Minneapolis, MN). This kit measures the biologically active form of HGF. The concentrations of HGF in the supernatants were measured from quadruplicate wells. The limit of sensitivity for this assay was 20 pg/ml. The plate was read on an Elisa reader (Dynex, Chantilly, VA). Standard curves and sample concentrations were determined using the Revelation software package called (Dynex). The data for HGF secretion by the uterine, cervical, and ectocervical stromal fibroblasts are presented as the mean ± SEM. InSTAT software (GraphPad Software, San Diego, CA) was used to perform a one-way repeated-measures analysis of variance (ANOVA) for each individual donor. Thus, significance in a donor is relative to that donor’s control secretion. When ANOVA analysis indicated that significant differences existed among means, paired comparisons were made using the Tukey method to adjust p-values. A p-value of <0.05 was considered statistically significant.