Basic fibroblast growth factor induces proliferation and collagen production by fibroblasts derived from the bovine corpus luteum

Abstract

Cyclic regression of the ovarian corpus luteum, the endocrine gland responsible for progesterone production, involves rapid matrix remodeling. Despite fibroblasts in other systems being known for producing and maintaining extracellular matrix, little is known about fibroblasts in the functional or regressing corpus luteum. Vast transcriptomic changes occur in the regressing corpus luteum, among which are reduced levels of vascular endothelial growth factor A (VEGFA) and increased expression of fibroblast growth factor 2 (FGF2) after 4 and 12 h of induced regression, when progesterone is declining and the microvasculature is destabilizing. We hypothesized that FGF2 activates luteal fibroblasts. Analysis of transcriptomic changes during induced luteal regression revealed elevations in markers of fibroblast activation and fibrosis, including fibroblast activation protein (FAP), serpin family E member 1 (SERPINE1), and secreted phosphoprotein 1 (SPP1). To test our hypothesis, we treated bovine luteal fibroblasts with FGF2 to measure downstream signaling, type 1 collagen production, and proliferation. We observed rapid and robust phosphorylation of various signaling pathways involved in proliferation, such as ERK, AKT, and STAT1. From our longer-term treatments, we determined that FGF2 has a concentration-dependent collagen-inducing effect, and that FGF2 acts as a mitogen for luteal fibroblasts. FGF2-induced proliferation was greatly blunted by inhibition of AKT or STAT1 signaling. Our results suggest that luteal fibroblasts are responsive to factors that are released by the regressing bovine corpus luteum, an insight into the contribution of fibroblasts to the microenvironment in the regressing corpus luteum.

Elevation of FGF2 expression during luteal regression may lead to alterations in the luteal microenvironment by stimulating proliferation of luteal fibroblasts and production of extracellular matrix.

Graphical Abstract

Graphical Abstract.

Introduction

The ovarian corpus luteum is a transient endocrine gland formed from the remaining cells of the dominant follicle following ovulation [1, 2]. Luteal formation is characterized by rapid angiogenesis driven by potent endothelial cell mitogens, vascular endothelial growth factor A (VEGFA), and fibroblast growth factor 2 (also known as basic fibroblast growth factor, hereon referred to as FGF2) [3]. Studies in the cow demonstrate that inhibition of FGF2 action at the time of ovulation resulted in non-functional corpora lutea with insufficient vasculature and impaired serum progesterone levels [4]. Luteal progesterone is imperative because it prepares and maintains the uterine lining for successful establishment and maintenance of pregnancy. When pregnancy does not occur, the corpus luteum undergoes rapid functional and structural regression, resulting in a reduction in luteal tissue and circulating progesterone levels, followed by rapid tissue remodeling into a fibrotic mass called the corpus albicans [1, 5]. In ruminants, this process is initiated by a uterine-derived hormone called prostaglandin F2α (PGF2α) [1, 5, 6].

FGF2 is a heparin-binding growth factor originally named from its ability to induce proliferation of NIH 3T3 fibroblasts and it is known to exert pleiotropic actions in many tissues [7, 8]. The FGF family of signaling peptides includes 18 secreted proteins that bind to four different tyrosine kinase FGF receptors (FGFRs 1–4), and FGF2 preferentially binds to FGFRs-1c, -2c, and -3c [8, 9]. FGF2 is also a potent angiogenic factor that promotes ovarian angiogenesis [10, 11]. In the corpus luteum, FGFRs have been shown to be localized to microvascular endothelial cells [12], and both FGF2 and FGFR have been observed in luteal steroidogenic cells [13].

As a growth factor, FGF2 stimulates downstream AKT Serine/Threonine Kinase 1 (AKT1), Mitogen Activated Protein Kinases MAPK1/3 (commonly referred to as ERK1/2), and Signal Transducer and Activator of Transcription 1 (STAT1) signaling to induce transcription and subsequent cellular proliferation [8, 9]. In vitro studies have revealed that FGF2 increases progesterone production in cultured luteinizing granulosa cells [14]; however, this response may be dependent on species and culture time because other studies found little effect of FGF2 on progesterone production in bovine and human luteinizing follicular cells [15, 16]. Results in primary luteal cells are more consistent, as FGF2 was shown to increase progesterone secretion of ovine and bubaline luteal cells, even at low doses [14, 17]. Similar effects were observed in micro-dialyzed bovine corpora lutea, whereby infusion with FGF2 resulted in increased progesterone secretion [18, 19].

Fibroblasts are mesenchymal-like cells that produce and maintain extracellular matrix, and they express alpha-smooth muscle actin (SMA, also called ACTA2) when they are activated to form myofibroblasts [2]. Although the presence of fibroblasts within the corpus luteum has been established [20], the role of fibroblasts in regulating both the maintenance and demise of the corpus luteum has remained largely unknown. Myofibroblast-like, SMA-positive, cells have been detected in late luteal phase and regressing bovine corpora lutea [21]. In addition, FGF2 increased the number of SMA-positive cells within bovine follicles, suggesting that FGF2 may be involved in regulating corpus luteum fate [15, 16]. Previous studies from our laboratory demonstrated that luteal fibroblasts increase collagen production and matrix contractility in response to transforming growth factor beta (TGFB1) [22]. It is noteworthy that TGFB1 is increased very early during luteal regression [23] and is considered an important mediator of fibrotic responses in many tissues [24–27]. A previous study demonstrated that TGFB1 did not stimulate proliferation of luteal fibroblasts, and treatment with factors known to be apoptotic to steroidogenic and endothelial cells did not diminish luteal fibroblast viability [22]. These findings hint at the fibroblasts being the last viable cells in luteal tissue during luteal regression. However, little is known about the contribution of fibroblasts to the microenvironment of the corpus luteum.

Despite FGF2 being discovered as a fibroblast growth factor [7], no data exist to our knowledge on the role of FGF2 on the residential fibroblasts of the corpus luteum. Transcriptomic studies of bovine luteal tissue revealed that mRNA for FGF2 rapidly increases after intramuscular (i.m.) administration of a luteolytic dose of PGF2α [28]. FGF2 was also shown to be elevated in late stage and naturally regressing corpora lutea [29, 30]. In fibroblasts derived from other organ systems, FGF2 increases proliferation, cell migration, and VEGFA production [31, 32]. FGF2 has also been shown to increase matrix production by cardiac [33] and flexor tendon [34] fibroblasts. Knowledge of the contributions of FGF2 to luteal fibroblast proliferation and collagen production could provide insight into how factors secreted in the early stages of luteal regression impact tissue remodeling and subsequent corpus albicans formation.

In the present study, we tested the hypothesis that bovine luteal fibroblasts respond to FGF2 with increased proliferation and collagen production. FGF2 is known to act through multiple signaling pathways (i.e., ERK, AKT, and STAT1) [35], so we also determined whether these pathways are required for proliferation of luteal fibroblasts.

Materials and methods

Reagents

Antibodies used in the study are listed in Supplementary Table S1. The Truseq mRNA library kit was acquired from Illumina (San Diego, CA, USA). FGF2 (Gibco, #PHG0367), TrypLE (Gibco, #12604021), western blot developing reagent, 10% neutral buffered formalin, Pierce RIPA buffer, Bicinchoninic acid (BCA) assay kit, and 100x protease/phosphatase inhibitor were purchased from ThermoFisher Scientific (Waltham, MA, USA). Dimethyl sulfoxide (DMSO), bovine serum albumin (BSA), and Permount mounting medium were acquired from Fisher Scientific (Hampton, NH, USA). Fludarabine was purchased from Tocris Bioscience (Bristol, UK). Phosphate buffered saline (PBS) and number 1 glass coverslips were purchased from Corning Life Sciences (Marlborough, MA, USA). Microscope slides, Mayer’s hematoxylin, and clear nail polish were acquired from Electron Microscopy Services (Hatfield, PA, USA). U0126 was purchased from Enzo Life Sciences (Farmingdale, NY, USA). Penicillin/streptomycin and trypan blue were from Invitrogen Corporation (Carlsbad, CA, USA). The 4 mm biopsy punches were acquired from Integra Biologics (Hudson, NH, USA). Fetal bovine serum (FBS) and TGFB1 were purchased from R&D Systems (Minneapolis, MN, USA). Sodium chloride, Tris–HCl, glycerol, sodium dodecyl sulfate (SDS), beta-mercaptoethanol, bromophenol blue, paraformaldehyde, and rapamycin were purchased from Sigma-Aldrich (St. Louis, MO, USA). AKT kinase inhibitor was purchased from Calbiochem (now Millipore-Sigma, Burlington, MA, USA). High-glucose Dulbecco’s modified essential medium (DMEM) was acquired from Cytiva Life Sciences (Marlborough, MA, USA). VECTASHIELD antifade mounting medium, antigen unmasking solution, rabbit IgG ABC kits, and 3,3′-diaminobenzidine (DAB) staining kits were purchased from Vector Labs (Newark, CA, USA).

Ethic statement: Animal studies were approved by the institutional animal care and use committee at the University of Nebraska–Lincoln, as previously published [23].

Induced regression of bovine corpora lutea in vivo

Post-pubertal female beef cows [25% MARC (1/4 Angus, 1/4 Hereford, 1/4 Pinzgauer, 1/4 Red Poll) and 75% Red Angus] ages 3–6 at the Eastern Nebraska Research, Extension, and Education Center were used for this study. Cows received two intramuscular (i.m.) injections of the PGF2α mimic, Lutalyse (25 mg) 11 days before the experiment, to synchronize their estrous cycles. All cows were monitored using ultrasound imaging to detect the presence of a corpus luteum prior to the study. At day 10 (midcycle), cows received an injection (i.m.) of either saline or PGF2α at 4 or 12 h prior to ovariectomy. Blood samples were collected before and after PGF2α treatment to measure plasma progesterone as a confirmation of luteal regression. Plasma progesterone was measured using the ImmuChem Coated Tube Progesterone ¹²⁵I radioimmunoassay kit from ICN Pharmaceuticals, Inc. (Costa Mesa, CA, USA). The intra-assay coefficient of variability was 4.25%, and the inter-assay variability coefficient was 4.90%. At each time point, bilateral ovariectomy was performed through the right flank while under local anesthesia, as previously reported [36]. Immediately following ovariectomy, the corpus luteum was dissected from the ovary and 0.5 g was snap-frozen for RNA sequencing. Samples were stored at −80°C until use.

RNA sequencing of bovine luteal tissue

RNA was isolated and processed by the Genomics Core at the University of Nebraska Medical Center (UNMC). Libraries were constructed using the Illumina Tru-seq mRNA library kit using 1 μg of total RNA in accordance with the manufacturer’s protocol. RNA was sequenced (100 bp single read) on an Illumina Novaseq 6000. The original fastq format reads were trimmed by fqtirm tool to remove adapters, terminal unknown bases (Ns), and low-quality 3′ regions (Phred score < 30). Trimmed fastq files were processed by FastQc [37]. The reads were mapped to ARS-UCD1.2 cow reference genome using Star 2.5 [38] and the transcripts were quantified using a strand-specific protocol in RSEM 1.3 [39]. Differentially expressed gene (DEG) analysis was performed using the R/Bioconductor package DESeq2 [40] with raw read counts from RNA-Seq. Only genes with at least 10 raw read counts were included in the analysis. Read count normalization was performed using the regularized logarithm (rlog) method provided in DESeq2. Negative binomial generalized linear models were fitted for each gene using DESeq2. P values were first calculated using the Wald test and then adjusted for multiple testing using the Benjamini–Hochberg method [37]. DEGs were determined by adjusted p value (Q < 0.05). RNA sequencing data from this study are deposited at NCBI as GSE217053.

Corpus luteum collection and fibroblast isolation

Bovine ovaries were collected from a local abattoir (Greater Omaha Packing Company, Omaha, NE, USA). Midcycle corpora lutea were determined by gross morphology, as described in [41]. Luteal fibroblasts were isolated by tissue outgrowth, a commonly utilized method of isolating fibroblasts [42–44]. Corpora lutea were dissected from the ovary and 4 mm punch biopsies were placed on a 6-well plate that was pre-coated with FBS. After 1 h of acclimation, high-glucose DMEM containing 10% FBS was added to the punch biopsies. The following day, medium was replaced with high-glucose DMEM containing 20% FBS and changed daily until fibroblasts reached confluency. Then, they were passaged to eliminate contaminating endothelial cells and cultured in high-glucose DMEM with 10% FBS. Passages between 2 and 6 were used for experiments. Experiments were replicated with fibroblasts isolated from different corpora lutea on different days.

Immunohistochemistry

Bovine luteal tissue collected from a local abattoir was rinsed in 1× phosphate buffered saline (1XPBS) and fixed in 10% formalin overnight and subsequently transferred to 70% ethanol. Samples were paraffinized and sectioned into 4 μm sections at the UNMC Histology Core. Sections were then deparaffinized and rehydrated in varying grades of ethanol. Antigen retrieval was performed using sodium citrate (pH 6) heated in a microwave on high for 20 min. After antigen retrieval and blocking of endogenous peroxidases with 3% H₂O₂ in methanol, sections were blocked in 10% normal donkey serum for 1 h before addition of antibodies (listed in Supplementary Table S1). Slides were incubated with primary antibodies (diluted in blocking buffer) overnight at 4°C. Slides were then rinsed in 1XPBS 3 times for 10 min and incubated in secondary biotinylated rabbit antibody for 1 h at room temperature, followed by a second series of rinses. Following the rinses, slides were incubated in avidin/biotin from an ABC kit for 30 min and rinsed 3 times for 5 min in 1XPBS. To visualize the stain, slides were stained with DAB and rinsed in deionized, distilled water to stop the reaction. Following DAB staining, slides were counterstained with Mayer’s hematoxylin for 1.5 min and rinsed under tap water for 20 min. Slides were dehydrated, mounted, and sealed with clear nail polish.

Fibroblast culture and measurement of downstream signaling

Luteal fibroblasts were grown and maintained in high-glucose DMEM with 10% FBS. To measure downstream signaling, fibroblasts were plated at a density of 5.8 × 10⁴ cells per well on a 12-well plate (70% confluent). The next day, cells were acclimated for 2 h in serum-free high-glucose DMEM and subsequently treated with FGF2 (1 ng/mL) from 2 to 120 min. Protein lysate was collected for western blot analysis.

Measurement of cell proliferation

To measure cell proliferation and collagen production, luteal fibroblasts were plated at a density of 4.2 × 10⁴ cells per well on a 12-well dish for western blotting, or 8.4 × 10⁴ cells per well on a 6-well dish (50% confluence) for live cell counting and EdU incorporation. To quantify the number of live cells, proliferation was measured by trypan blue exclusion. Bovine luteal fibroblasts were plated as mentioned previously. The following day, medium was changed to fresh high-glucose DMEM with 10% FBS, and the cells were acclimated for 2 h. Following acclimation, FGF2 (0–10 ng/mL) was added and incubations terminated after 24 h or 48 h. For cell counts, medium was removed, and cells were rinsed in 1XPBS and detached with TRYPLE solution and counted on a Countess Automated Cell Counter (ThermoFisher Scientific, Waltham, MA, USA).

To confirm FGF2-induced proliferation, an EdU (5-ethynyl-2′-deoxyuridine) incorporation assay was performed using the iClick EdU Andy Fluor 647 Imaging Kit (ABP Biosciences, catalog # A006; Beltsville, MD, USA). Luteal fibroblasts were plated as mentioned previously. The following day, cells were rinsed and acclimated in DMEM with or without 10% FBS for 2 h. Following acclimation, half of the medium was removed and replaced with medium containing EdU (final concentration 10 μM EdU, 5% FBS). Fibroblasts were subsequently treated with FGF2 (1 ng/mL) or control medium for 48 h. Cells were fixed in 4% paraformaldehyde in 1XPBS, stained, and rinsed according to the manufacturer’s directions. Slides were mounted in VECTASHIELD antifade mounting medium and imaged at the UNMC Advanced Microscopy Core on a ZEISS 800 confocal microscope. EdU-incorporated nuclei were detected at a wavelength of 647 nm and all nuclei (Hoechst) were detected at 405 nm. The numbers of EdU-positive and total nuclei were counted using the cell counter feature on ImageJ and quantified as the fold change of nuclei compared to control. Ten images were taken for each slide at 10× magnification (NA = 0.30), and each biological replicate is the average of its respective 10 images.

Effect of signaling pathway inhibitors on luteal fibroblast proliferation

In order to determine the effect of downstream signaling on FGF2 induced proliferation, luteal fibroblasts were plated in 12-well plates at 4.2 × 10⁴ cells per well (50% confluence). ERK phosphorylation was inhibited with the MEK1 (upstream kinase of ERK) inhibitor, U0126 (10 μM). AKT was inhibited with an AKT kinase inhibitor (10 μM). STAT1 activation was inhibited with fludarabine (100 μM), and MTOR/p70S6K signaling was inhibited with rapamycin (50 nM). Following 2 h acclimation in high-glucose DMEM with 10% FBS, luteal fibroblasts were preincubated with either inhibitor or DMSO for 1 h, and then treated with either medium or FGF2 (1 ng/mL) for 24, 48, or 72 h. At the end of the experiment, either protein lysate was prepared for western blot analysis or cells were counted as described previously. For cell counts, cells were treated for 48 h and 2 wells were averaged for each treatment in each experiment.

Western blot

Western blotting was performed as previously published [45]. Following the experiment, cells were immediately put on ice, rinsed with ice-cold 1XPBS, and lysed with Pierce RIPA buffer. Lysates were sonicated at 40% amplitude for 3 s and centrifuged (13 000 × g for 10 min). Supernatant containing protein was transferred to a clean 1.5 mL tube and its concentration was determined by BCA assay. Lysates (10 μg/lane) were separated on a 10% SDS-PAGE gel at 110 V and transferred onto nitrocellulose membranes (100 V for 2 h). Membranes were blocked in 5% BSA in Tris-buffered saline with 0.1% Tween-20 (1XTBST) for 30 min and incubated with primary antibodies (Supplementary Table S1) overnight at 4°C. Membranes were washed (three 10 min washes in 1XTBST), blocked in secondary horseradish peroxidase-conjugated anti-rabbit or anti-mouse antibodies for 1 h in 2.5% non-fat milk in 1XTBST, and washed again (three 10 min washes in 1XTBST). Blots were developed with chemiluminescence and imaged using the Invitrogen iBright 1500 (Thermo Fisher, Waltham, MA, USA) imaging system. Images were quantified using ImageJ.

Measurement of collagen production

Luteal fibroblasts were cultured as described previously and acclimated for 2 h in serum-free DMEM prior to the addition of FGF2 at various doses. After 24 or 48 h of FGF2 treatment, protein lysates were collected, and western blot analysis and quantification was performed as described previously.

Statistical analysis

Statistically significant differences from RNA sequencing data were calculated using DESeq2, and multiple testing was accounted for using Benjamini–Hochberg method, as mentioned earlier.

One-way analysis of variance (ANOVA) was used to compare the changes in plasma progesterone levels before and after PGF2α treatment. If the ANOVA yielded significant results (p < 0.05), then a Dunnett’s post hoc test was performed to identify changes in progesterone compared to the saline control.

For the in vitro studies, each assay was repeated 3–6 independent times, using fibroblasts isolated from different corpora lutea. Continuous outcomes per assay, such as normalized phosphorylation and protein content from western blot analysis, and normalized counts of viable cells, were summarized by descriptive statistics for each FGF2 treated time, FGF2 dose, or inhibitor treatment × FGF2 treated time combination. Comparisons between two paired groups were calculated using paired t-tests. Mixed-model analyses for repeated measures were applied to explore the differences in continuous outcomes either over at least three FGF2-treated doses, or over at least three FGF2-treated time points, or over inhibitor at least three treatments at each timepoint of FGF2 treatment, respectively, accounting for correlations between repeated measurements from the same corpus luteum. If the mixed model analysis yielded significant results, post hoc tests with Dunnett’s methods for multiple comparisons were conducted to compare each of the other doses against dose 0, or compare each of the other time points against time 0, or compare each of the other inhibitor treatments against vehicle control at a given time point of FGF2 treatment. Log2 transformations were applied to continuous outcomes to meet mixed model assumptions as necessary. Data analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA) and Graph Pad Prism v. 9.4.0 (La Jolla, CA, USA). A value of p <0.05 was considered statistically significant. P values shown for mixed model analyses are Dunnett’s corrected p values.

Results

Fibroblast localization and activation within the corpus luteum

Activated fibroblasts express the serine protease, fibroblast activation protein (FAP) [46, 47]. Immunohistochemistry of FAP shows that luteal fibroblasts are interspersed throughout the mid-luteal phase corpus luteum (Figure 1). Immunohistochemistry also revealed that regressed corpora lutea have more FAP-positive cells (Figure 1).

Figure 1.

Fibroblast activation protein (FAP) is increased in regressing corpora lutea. Localization within midcycle and regressed corpora lutea for fibroblast activation protein in corpora lutea obtained from a local abattoir. White arrowheads indicate luteal steroidogenic cells and black arrowheads indicate stained fibroblasts.

Luteal regression was confirmed by a decreased serum progesterone. At 4 h post-PGF2α, progesterone decreased to 59 ± 11% (mean ± SE, n = 4, p < 0.01) of basal, compared to a 129 ± 19% change in the saline controls. At 12 h post-PGF2α, progesterone decreased to 24 ± 3.6% of basal (n = 4, p < 0.001), compared to saline controls. In addition, RNA sequencing of mid-luteal phase (Day 10) and regressing corpora lutea revealed a 4-fold increase in the number of FAP transcripts 12 h post-PGF2α injection in vivo (p < 0.05) (Table 1).

Table 1.

Induction of pro-fibrotic mediators during PGF2α-induced luteal regression

Gene1	0 h (TPM)	SD 0 h	4 h (TPM)	SD 4 h	Adjusted p value	12 h (TPM)	SD 12 h	Adjusted p value
CCN2	110.20	53.53	167.44	44.10	0.151	417.24	232.93	3.23E−08
SPP1	95.92	46.29	682.79	320.91	5.60E−12	1775.60	343.58	8.41E−41
LGALS3	76.50	17.20	87.78	9.50	0.444	135.00	4.33	3.13E−09
TGFB1	35.23	1.27	85.77	22.88	8.16E−09	143.07	42.12	3.03E−21
S100A4  2	22.70	8.36	19.28	3.89	0.750	54.33	19.40	1.55E−06
SERPINE1	12.31	10.61	284.86	115.88	6.99E−10	280.75	122.30	7.69E−14
FAP  2	5.91	5.28	9.47	3.41	0.324	25.57	12.99	3.59E−06

¹ LGALS3 = galectin 3, CCN2 = cellular communication network factor/connective tissue growth factor, SPP1 = secreted phosphoprotein 1, TGFB1 = transforming growth factor-β1, SERPINE1 = serpin family E member 1, FAP = fibroblast activation protein, S100A4 = S100 calcium binding protein A4.

²Fibroblast marker.

RNA sequencing also revealed significantly increased mRNA levels of pro-fibrotic genes (Table 1) including galectin 3 (LSGALS3), connective tissue growth factor (CCN2), secreted phosphoprotein 1 (SPP1), TGFB1, and serpin family member E 1 (SERPINE1). In addition, mRNA levels of S100 calcium binding protein A4 (S100A4), also known as fibroblast specific protein, was increased 2.4-fold (adjusted p < 0.0001) (Table 1).

Endothelial cell dysfunction has been associated with tissue fibrosis [48]. Genes encoding angiogenic factors (e.g., VEGFA and KDR) were decreased at 4 and 12 h post-PGF2α (Supplementary Figure S1A), whereas genes encoding anti-angiogenic factors were increased (e.g., thrombospondins (THBS) -1 and -2, and TGFB1) (Supplementary Figure S1B).

FGF ligands and FGFR in the midcycle and regressing bovine corpus luteum

RNA sequencing of mid-luteal phase (Day 10) luteal tissue (GSE217053) (n = 4 per group) identified the presence of 12 members of the FGF family of ligands and all four members of the FGF family of receptors (Supplementary Table S2). The effects of a luteolytic injection of PGF2α on the most prominently expressed FGF ligands and receptors are shown in Figure 2. PGF2α treatment caused a transient increase in FGF1 expression after 4 h (2-fold, adjusted p < 0.0001) that returned to control levels within 12 h post-PGF2α (Figure 2A). Of note, treatment with PGF2α provoked a robust increase in the number of FGF2 transcripts at 4 h (8-fold, adjusted p < 0.0001), which remained elevated 12 h post-PGF2α (3.5-fold, adjusted p < 0.0001). PGF2α treatment did not affect levels of FGF7, and levels of FGF12 were decreased by 40% (adjusted p < 0.05) and 51% (p = 0.08) following PGF2α treatment for 4 and 12 h, respectively.

Figure 2.

Luteal fibroblast growth factors (FGFs) in vivo (A–B) and in vitro (C). Mid-luteal phase cows were injected with saline (Control) or prostaglandin F2α (PGF2α, 25 mg, i.m.) and ovariectomized after 4 and 12 h to collect corpora lutea. RNA sequencing of whole luteal tissue was performed. (A) Most abundant transcripts of FGF family members in the bovine corpus luteum at midcycle and 4- and 12 h post-PGF2α injection. (B) mRNA levels of FGF receptors (FGFRs) at midcycle and 4- and 12 h post-PGF2α injection. Data are presented as mean number of transcripts per million (TPM) ± SEM, n = 4, ^*p < 0.05, ^***p < 0.001, ^****p < 0.0001 compared to 0 h by DESeq2 analysis, Benjamini–Hochberg correction. P values shown are adjusted p values for multiple comparisons. (C) Fibroblasts were treated with or without FGF2 (1 ng/mL) for up to 60 min and cellular lysates prepared for western blot analysis. Shown is a representative experiment repeated in duplicate.

Figure 2B shows levels of FGFR transcripts at mid-luteal phase and following an injection of a luteolytic dose of PGF2α. At the mid-luteal phase, levels of FGFR1 mRNA were greater (approximately 7-fold) than levels of mRNA for FGFR2, FGFR3, and FGFR4 (Supplementary Table S2). Treatment with PGF2α did not affect levels of FGFR1, FGFR3, and FGFR4 transcripts. However, FGFR2 transcripts were reduced (adjusted p < 0.0008) 2-fold at 4 h and 1.4-fold (adjusted p < 0.05) at 12 h following PGF2α treatment (Figure 2B). All FGF and FGFR transcripts identified in control and PGF2α-treated luteal tissue are described in Supplementary Table S2.

FGF2 stimulates multiple signaling pathways in luteal fibroblasts

Western blot analysis was conducted to determine the rapid intracellular signaling responses of luteal fibroblasts to FGF2. Full western blots with the molecular weight ladders are provided in Supplementary Figure S2. Figure 2C shows that luteal fibroblasts respond directly to FGF2 (1 ng/mL) as evidenced by the rapid and robust increase in FGFR tyrosine phosphorylation within 2 min of treatment. Several FGFR signaling pathways were evoked within 2–5 min of treatment, including the phosphorylation of ERK, AKT, p70S6K, and STAT1 (Figure 2C).

Additional studies were performed to quantify the temporal nature of the cellular signaling responses over a 2 h period (n = 4–6) (Figure 3). Mixed model analysis for repeated measures indicated that the mean normalized phosphorylation of each signaling protein differed across four time points (p < 0.01). Post hoc comparisons between each of the other time points against time 0 using the Dunnett correction for each signaling protein showed that within 10 min of treatment, the mean normalized STAT1 phosphorylation (Y701) was increased over 11-fold (p < 0.001), gradually reducing but remaining elevated approximately 6-fold after 120 min (p < 0.05) (Figure 3A). The mean normalized phosphorylation of ERK (T202/Y204) was increased 3.7-fold within 10 min and remained elevated at least 3.4-fold throughout the 120 min time course (p < 0.005) (Figure 3B). In contrast, phosphorylation of AKT (S473) increased early in response to FGF2 (1.7-fold after 10 min, p < 0.001) and was increased 1.4-fold within 120 min post-FGF2 treatment but was not statistically significant (Figure 3C). FGF2 treatment also stimulated phosphorylation of molecules downstream of AKT. GSK3B phosphorylation (S9) was elevated 1.9-fold after 10 min (p < 0.001) and remained elevated at least 1.4-fold throughout the 120 min time course (each of other times vs. 0 min, p < 0.05), and phosphorylation of the MTOR substrate, p70S6K (T389), was elevated 2.8-fold after 10 min (p < 0.0001) and remained elevated after 120 min (2.3-fold, p < 0.01) (Figure 3D and E).

Figure 3.

Temporal response to FGF2 on STAT1, ERK, and AKT signaling in bovine luteal fibroblasts. Bovine luteal fibroblasts were treated with 1 ng/mL FGF2 for up to 120 min and protein extracts were processed for western blot analysis of signaling proteins. FGF2 rapidly induces STAT1 (A) and ERK (B) phosphorylation. FGF2 rapidly induces AKT phosphorylation (C) and its downstream targets, GSK3B (D) and p70S6K (E). Statistically significant differences in phosphorylation compared to timepoint 0 were determined by a mixed model analysis with Dunnett’s post hoc multiple comparisons. P values shown are Dunnett’s corrected p values. Data are presented as means ± SEM, n = 5–6 experiments with different fibroblast preparations, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001.

FGF2 stimulates proliferation of luteal fibroblasts

To determine the effectiveness of FGF2 on proliferation of bovine luteal fibroblasts, we performed a concentration response study on protein expression of proliferating cell nuclear antigen (PCNA, a known marker for cell proliferation), EdU incorporation, and number of live cells (Figure 4). FGF2 treatment of luteal fibroblasts for 24 h increased PCNA content (Figure 4A) but did not significantly increase cell numbers (n = 3) (Figure 4B). However, 48 h of treatment with FGF2 increased PCNA content of bovine luteal fibroblasts and stimulated proliferation of luteal fibroblasts at 1 and 3 ng/mL (n = 5, p < 0.01) (Figure 4C and D).

Figure 4.

FGF2 increases proliferation of bovine luteal fibroblasts. Cells were treated with varying concentrations of FGF2 for 24 h or 48 h. Proliferation was determined by western blot analysis for the cell proliferation marker PCNA and by direct counting of viable cells by trypan blue staining. (A) Representative PCNA western blot after 24 h of treatment with 0–10 ng/mL of FGF2. (B) Summary of cell counts after 24 h incubation with FGF2 (1 ng/mL). (C) Representative PCNA western blot after 48 h of treatment with FGF2 (0–10 ng/mL). (D) Summary of cell counts after treatment with 0–10 ng/mL of FGF2 for 48 h. Statistically significant differences were identified by mixed model analysis. EdU incorporation assay in the serum-free (E) or 5% serum (F) conditions. The assay was performed as described in the methods. Pink nuclei represent nuclei with EdU incorporation. Cell nuclei were stained with Hoechst (blue). Statistically significant differences between EdU+: total nuclei ratio of control and FGF-treated fibroblasts were calculated by using a paired t-test. Scale bar = 200 μm. Data are represented as means ± SEM, n = 3–5, ^*p < 0.05, ^***p < 0.001, compared to untreated control.

We also monitored cell proliferation by measuring incorporation of EdU into bovine luteal fibroblasts. To accomplish this, bovine luteal fibroblasts were incubated for 48 h without or with FGF2 (1 ng/mL) in serum-free medium or with medium containing 5% FBS and EdU incorporation into DNA was quantified. FGF2 increased EdU incorporation by 1.5-fold in serum-free conditions (p = 0.056) and 1.7-fold in medium containing 5% FBS (p < 0.05) (Figure 4E and F).

Signaling pathways required for FGF2-induced cell proliferation

To determine the signaling mechanisms required for FGF2-induced proliferation of luteal fibroblasts, small molecule inhibitors were used to inhibit activity of the downstream mediators that were identified in Figures 2 and 3. Luteal fibroblasts were pre-incubated with either vehicle (DMSO) or inhibitor for 1 h prior to FGF2 treatment (24–72 h) (n = 5). Proliferation was measured by western blot analysis of PCNA and by determining cell counts. Prior to the experiment, the effectiveness of the inhibitor was confirmed by western blot analysis of short-term (5–30 min) FGF2 treatment (1 ng/mL). Inhibition of ERK phosphorylation with the MEK inhibitor, U0126, and inhibition of AKT with an AKT kinase inhibitor were confirmed with western blot analysis (Figure 5A). Although the presence of the MEK inhibitor inhibited basal and FGF-induced ERK signaling and decreased basal PCNA, FGF2 remained able to induce temporal increases in PCNA at 48 h and 72 h (Figure 5A and B). Furthermore, inhibition of ERK slightly decreased FGF2-induced cell number (0.79-fold, p < 0.05) (Figure 5C). In contrast, AKT inhibition effectively reduced the ability of FGF2 to stimulate PCNA levels (p < 0.0001) and cell proliferation (p < 0.001) (Figure 5B and C).

Figure 5.

Inhibition of AKT attenuates FGF2-induced proliferation of bovine luteal fibroblasts. Bovine luteal fibroblasts were pretreated for 1 h with vehicle (DMSO) or ERK inhibitor (U0126, 10 μM) and AKT kinase inhibitor (AKTi, 10 μM) prior to treatment with FGF2 (1 ng/mL) for up to 30 min to examine cell signaling events and for up to 72 h for measurements of cell proliferation. (A) Western blot analysis and confirming inhibition of cell signaling. (B) Representative western blot analysis of PCNA and quantification. (C) Cell numbers were determined after 48 h. Data are quantified as a fold change compared to vehicle for its respective timepoint and are presented as mean ± SEM, n = 5, ^*p < 0.05, ^***p < 0.001, ^****p < 0.0001. Statistically significant differences between groups were calculated using a mixed model analysis with Dunnett’s post hoc analysis.

In order to identify other mechanisms of FGF2-induced proliferation, luteal fibroblasts were treated with fludarabine [49], which acts by down-regulating STAT1 by inhibiting its activation and nuclear localization, and the MTOR inhibitor, rapamycin, which inhibits p70S6K activity [50]. Rapamycin effectively diminished basal and FGF2-induced p70S6K phosphorylation (Figure 6A) and lowered PCNA content at baseline (n = 4, p < 0.001), but luteal fibroblasts remained responsive to FGF2, as indicated by western blot analysis for PCNA (Figure 6B). However, rapamycin treatment decreased the FGF2-induced increase in cell numbers (n = 4, p < 0.05) (Figure 6C). In contrast, fludarabine treatment abrogated both FGF2-stimulated PCNA levels (Figure 6B) and FGF2-induced proliferation of luteal fibroblasts (n = 4, p < 0.05) (Figure 6C).

Figure 6.

STAT1 activity is required for FGF2-stimulated proliferation of luteal fibroblasts. (A) Bovine luteal fibroblasts were pretreated for 1 h with the MTOR inhibitor rapamycin (Rap, 50 nM) or the STAT1 inhibitor fludarabine (Flu, 100 μM) and then incubated without or with FGF2 (1 ng/mL) for 24–72 h (B). The proliferation marker PCNA was measured by western blot analysis. β-Actin (ACTB) was used as a loading control, n = 4. (C) Proliferation was also measured by cell counting via trypan blue exclusion after treatment with FGF2 for 48 h. Data are quantified as a fold change compared to vehicle for its respective timepoint and are presented as mean ± SEM, n = 4. Dunnett’s corrected p values were determined by mixed model analysis, ^*p < 0.05, ^***p < 0.001, ^****p < 0.0001.

FGF2 stimulates collagen production by luteal fibroblasts

A major function of fibroblasts is to produce collagen; therefore, we performed experiments to determine whether FGF2 impacts collagen production (n = 3). In serum-free conditions, FGF2 stimulated increases in type I collagen content at concentrations greater than 0.3 ng/mL, by at least 1.8-fold (p < 0.05) after 48 h of treatment (Figure 7). As a positive control, we show that treatment with TGFB1 significantly increases luteal fibroblast collagen production (p < 0.001).

Figure 7.

FGF2 increases collagen production by bovine luteal fibroblasts. Bovine luteal fibroblasts were treated without or with increasing concentrations of FGF2 (0–10 ng/mL) for 48 h. Treatment with TGFB1 (1 ng/ml) was used as a positive control. (A) Representative western blot of type 1 collagen. β-Actin (ACTB) was used as a loading control. (B) Quantification of collagen production. Data are quantified as a fold change compared to control (no FGF2) and presented as mean ± SEM, n = 3. Statistical differences between groups were calculated by a mixed model analysis with Dunnett’s post hoc analysis. Dunnett’s corrected p values ^*p < 0.05, ^**p < 0.01, ^***p < 0.001.

Discussion

The mechanisms by which PGF2α induces structural regression of the corpus luteum are poorly understood. Previous studies document reductions in progesterone occurring within 1–2 h [23], decreases in blood flow between 4 and 8 h [51], and increases in collagen deposition occurring within 8 h [52] of in vivo PGF2α administration. Because fibroblasts in the corpus luteum are the only luteal cell type to be resistant to luteolytic ligands [22], we posit that the fibroblasts contribute to increased collagen deposition seen in the regressing corpus luteum. In this study, we observed increased FAP mRNA in luteal tissue during the later stage of luteal regression. We also observed rapid and sustained increases in FGF2 transcripts during PGF2α-induced luteal regression in vivo. We evaluated a role for FGF2 on luteal fibroblasts in processes that might impact the microenvironment on the corpus luteum. In vitro studies revealed that activation of the FGFR stimulates AKT and STAT1 signaling pathways that are coupled to FGF2-induced proliferation of luteal fibroblasts. We also documented that luteal fibroblasts produce type 1 collagen in response to FGF2 in vitro. These findings provide new information on the regulation of luteal fibroblasts and events orchestrated in response to PGF2α during structural regression of the corpus luteum.

RNA sequencing revealed increased FAP transcripts in bovine luteal tissue 12 h post-PGF2α. FAP is a type II transmembrane serine protease that is expressed in activated fibroblasts of many systems [46, 53]. It was originally thought to only be present in activated cancer associated fibroblasts, but it has also been shown to be present in physiological tissue remodeling processes such as mouse embryogenesis [53] and tadpole tail resorption, where its transcription is increased temporally [54]. FAP has also been identified in wound fibroblasts and activated stromal cells [46]. FAP belongs to the dipeptidyl protease family, which share the capability to cleave proline. Because proline is a major component of collagen fibers, one of the targets of dipeptidyl proteases is cleaved collagen types I and III [47]. FAP is also known to be activated by inflammatory cytokines tumor necrosis factor-α and interleukin-1β, as well as TGFB1, all of which are increased during luteal regression [55]. Therefore, it is possible that the increase in FAP transcripts in the regressing bovine corpus luteum is an indicator of fibroblast activation, which may enhance the tissue remodeling capacity of luteal fibroblasts during regression. Our immunohistochemistry also agrees with this hypothesis, as FAP deposition was increased in the regressed corpus luteum.

In further agreement with fibroblast activation and luteal fibrosis, RNA sequencing revealed increased mRNA levels of key pro-fibrosis genes, including rapid increases in TGFB1 transcripts. Of note, we observed increased SERPINE1, SPP1, and THBS1 mRNA (22-fold, 18-fold, and 3.4-fold, respectively), which is in agreement with our previous study [23] and others [56–59]. SERPINE1, which encodes plasminogen activation inhibitor type-1 (PAI-1), is a TGFB1 target and has a prominent role in tissue fibrosis, particularly by inhibiting plasminogen-induced matrix metalloproteinase activation [60, 61]. During luteal regression, SERPINE1 is induced by luteolytic factors, THBS1 and TGFB1 [57]. THBS1 is a luteolytic factor that acts in synergy with TGFB1 to mediate tissue remodeling and vascular instability in the regressing corpus luteum [57, 62]. SPP1, or osteopontin, is a pro-fibrotic factor that has been shown to induce collagen deposition of fibroblasts [63]. However, it is also known to be a immunomodulatory molecule, as SPP1^hi macrophages were highly proliferative and present in fibrotic lung tissue [64]; and in the corpus luteum, SPP1 was demonstrated to modulate T-lymphocyte chemotaxis [65]. Because immune cells are known modulators of fibrosis [66], these upregulated transcripts, even if not acting directly on fibroblasts, can regulate fibroblast activation and luteal tissue remodeling.

RNA sequencing studies also demonstrated that in response to a luteolytic dose of PGF2α in vivo, FGF2 transcription increases more so than other FGF family members within 4 h. The response to PGF2α on FGF2 transcripts was sustained while transcripts for FGF7 and FGF12 decreased 12 h post-PGF2α. The early increases in FGF2 transcription in vivo are consistent with previous reports [28, 67]. Despite reports demonstrating that exogenous FGF2 in the early luteal phase can elevate progesterone and stimulate angiogenesis [4, 68], the robust elevation in FGF2 mRNA following PGF2α injection was associated with reductions in progesterone production within 2 h [23]. Furthermore, direct intra-luteal injection of FGF2 at midcycle did not affect plasma progesterone in the cow [69]. Therefore, the elevation of FGF2 in the regressing corpus luteum does not appear to prevent the reduction in steroidogenesis or the disruption of the microvasculature.

A previous study determined that bovine luteal fibroblasts do not possess the receptor for PGF2α [70]. Our studies indicate that luteal fibroblasts are responsive to FGF2. FGF2 activated the FGFR on luteal fibroblasts as evidenced by rapid increases in the phosphorylation of the FGFR on tyrosine residues associated with recruitment of intracellular signaling pathways [35]. We observed that FGF2 robustly increased ERK, AKT, and STAT1 phosphorylation, which are known to promote cell proliferation and survival [8, 71, 72]. FGF2 also stimulated the phosphorylation of GSK3B and p70S6K, signaling intermediates downstream of AKT [73]. Thus, luteal fibroblasts are highly responsive targets of FGF2.

FGF2 was found to be a potent mitogen for bovine luteal fibroblasts. Our studies demonstrated that FGF2 provoked increases in PCNA content and EdU incorporation, which were associated increases in proliferation of luteal fibroblasts. The stimulatory effects of FGF2 on luteal fibroblast proliferation were enhanced in the presence of serum, indicating that other factors also contribute to growth of luteal fibroblasts. Our studies established that AKT and STAT1 signaling pathways were essential for proliferation of luteal fibroblasts. However, the ability of FGF2 to stimulate luteal fibroblast proliferation under conditions that effectively inhibited ERK signaling was surprising because of the importance of ERK in cell proliferation [74]. The explanation for the differential response is not apparent but may reflect tissue and cell type specificities to FGF2.

The tissue remodeling that occurs during luteal regression has similarities to a tissue fibrotic response, characterized by elevated TGFB1 and collagen production [23, 75]. RNA sequencing indicated that TGFB1 is elevated at 4 and 12 h post PGF2α-induced regression. This is in agreement with previous reports from our laboratory [23, 76] and others [77–79]. Also in accordance with a previous study [22], we observed that TGFB1 stimulates collagen production by luteal fibroblasts. Herein, we provide the initial observation that FGF2 stimulates concentration-dependent increases in collagen production by bovine luteal fibroblasts. The main function of fibroblasts is to produce matrix; thus, we posit that FGF2- and TGFB1-induced collagen production by the fibroblasts contributes to the matrix deposition observed in the regressing corpus luteum [52]. In addition, type 1 collagen has been shown to sensitize luteal microvascular endothelial cells [70], but not luteal fibroblasts [22] to the cytotoxic effects of luteolytic cytokines. Changes in matrix composition may also aid the influx of immune cells that occurs during PGF2α-induced luteal regression [80]. Therefore, fibroblast matrix deposition may serve multiple purposes to ensure luteal regression.

We observed a significant reduction in expression of FGF12 mRNA following PGF2α injection. FGF12 is a member of the family of intracellular non-secretory FGF forms, which lack the N-terminal secretion sequences, are localized to the nucleus, and function without activating FGF receptors [81]. Although little is known about the expression or role of FGF12 in the ovary, reports indicate that FGF12 is highly expressed in the nervous system and positively regulates voltage-gated sodium channels in neurons [81]. Recent studies also revealed that FGF12 is expressed in the heart and vascular smooth muscle cells, and lower FGF12 is associated with vascular wall thickening in atherosclerosis and pulmonary arterial hypertension [82–84]. Our studies also showed that PGF2α injection slightly elevated and then reduced the expression of FGF7 mRNA. Although little is known about FGF7 in the ovary, it was shown to suppress fibrosis and reduce scarring in a wound healing study [85]. Thus, elevations in FGF2 expression and reductions in FGF7 and FGF12 in response to PGF2α may serve to coordinate matrix remodeling during luteal regression. Further studies are needed to determine the roles of FGF7 and FGF12 in the corpus luteum.

Conclusion

The fine mechanics of rapid tissue turnover during luteal regression are unclear. While the roles of luteal steroidogenic, endothelial, and resident immune cells have been investigated and identified (reviewed in [1, 86, 87]), that of resident fibroblasts has not yet been established. In other tissues, fibroblasts serve as mediators of matrix breakdown and deposition [88], but no studies have shown how fibroblasts are regulated in the corpus luteum. The results of this study reveal a new role of both FGF2 and luteal fibroblasts during luteal regression. We demonstrate altered transcription of several FGF family members in response to the luteolytic ligand, PGF2α, including sustained elevation of transcripts for FGF2. Our in vitro studies demonstrate that bovine luteal fibroblasts are highly responsive to FGF2 by rapidly inducing downstream signaling pathways vital for fibroblast proliferation. The proliferative response to FGF2 is mediated largely by AKT and STAT1 signaling, as inhibition of these pathways reduces the response of luteal fibroblasts to FGF2. We also found that FGF2, like TGFB1, stimulates collagen production. Therefore, locally produced FGF2 could activate fibroblasts, thus facilitating remodeling of the luteal microenvironment during luteolysis.

Furthermore, the luteal phase is often overlooked in fertility research. In dairy cows, despite fertilization being more than 90% successful, nearly 30% of embryos are lost during the first trimester in pregnancy, a cause being inappropriate timing or amount of progesterone to optimize the conditions for embryo survival [89]. Insight into the mechanisms of luteal regression, including the actions of FGF2 and other growth factors on the matrix-producing cells of the corpus luteum, can aid in the development of targets for optimizing the length of the luteal phase for fertility in both dairy cows and women.

Author contributions

The manuscript was written by CFM and JSD. In vitro experiments were designed by CFM and JSD. Data analysis was performed by FQ, PX, CFM, JSD, and JWG. ASC and JRW were involved with experimental design for in vivo studies and obtaining mid cycle bovine corpora lutea. CFM, MRP, and EP assisted with collecting luteal tissue. All authors read and approved the manuscript prior to submission.

Data availability

All data are available upon reasonable request to the authors. RNA sequencing data are available at NCBI as GSE217053.