The LIM Domain Protein FHL2 Interacts with the NR5A Family of Nuclear Receptors and CREB to Activate the Inhibin-α Subunit Gene in Ovarian Granulosa Cells

Abstract

Nuclear receptor transcriptional activity is enhanced by interaction with coactivators. The highly related nuclear receptor 5A (NR5A) subfamily members liver receptor homolog 1 and steroidogenic factor 1 bind to and activate several of the same genes, many of which are important for reproductive function. To better understand transcriptional activation by these nuclear receptors, we sought to identify interacting proteins that might function as coactivators. The LIM domain protein four and a half LIM domain 2 (FHL2) was identified as interacting with the NR5A receptors in a yeast two-hybrid screen of a human ovary cDNA library. FHL2, and the closely related FHL1, are both expressed in the rodent ovary and in granulosa cells. Small interfering RNA-mediated knockdown of FHL1 and FHL2 in primary mouse granulosa cells reduced expression of the NR5A target genes encoding inhibin-α and P450scc. In vitro assays confirmed the interaction between the FHL and NR5A proteins and revealed that a single LIM domain of FHL2 is sufficient for this interaction, whereas determinants in both the ligand binding domain and DNA binding domain of NR5A proteins are important. FHL2 enhances the ability of both liver receptor homolog 1 and steroidogenic factor 1 to activate the inhibin-α subunit gene promoter in granulosa cells and thus functions as a transcriptional coactivator. FHL2 also interacts with cAMP response element-binding protein and substantially augments activation of inhibin gene expression by the combination of NR5A receptors and forskolin, suggesting that FHL2 may facilitate integration of these two signals. Collectively these results identify FHL2 as a novel coactivator of NR5A nuclear receptors in ovarian granulosa cells and suggest its involvement in regulating target genes important for mammalian reproduction.

Nuclear receptors (NR) comprise a large class of DNA binding proteins (48 known human NR) that regulate gene transcription (1, 2). NR have a conserved structure that includes a DNA binding domain (DBD) composed of two zinc fingers and a carboxy-terminal ligand binding domain (LBD) composed of 12 alpha helices that form a hydrophobic pocket that in many cases binds a cognate ligand. NR are classified based on their binding to their DNA response elements: homodimers bind to inverted repeat response elements [i.e. estrogen receptor (ER), progesterone receptor, androgen receptor (AR)], heterodimers with retinoid X receptor as a partner bind to direct repeat response elements (i.e. retinoic acid receptor, peroxisomal proliferator-activated receptor), and NR monomers bind to half-site response elements (3). The transcriptional activation potential of many NR is dependent on the binding of small, lipophilic molecules to the LBD pocket. A small number of NR, most of which bind as monomers, are constitutively active and regulate gene transcription in the absence of exogenous ligand. These receptors are often referred to as orphan nuclear receptors (4).

Liver receptor homolog 1 (LRH-1) and steroidogenic factor 1 (SF-1) are members of the NR5A subfamily of NR (5, 6). The NR5A subfamily is also known as the FTZ-F1 family because the fushi tarazu segmentation gene in Drosophila melanogaster was the founding member of this gene family (7). Although LRH-1 and SF-1 are often characterized as orphan nuclear receptors, phospholipids have been found in the binding pockets of several NR5A LBD crystal structures, and these may serve as ligands that stabilize the folded structure of the receptor (8–10). SF-1 (NR5A1) is a transcriptional regulator of steroidogenic gene expression in endocrine tissues including the gonads, adrenal gland, and ventral medial hypothalamus. SF-1 knockout mice lack gonads and adrenal glands, highlighting the essential role that SF-1 plays in development (11). LRH-1 (NR5A2) is expressed in endoderm-derived tissues including the gonads, liver, pancreas, and intestine and is involved in processes including development, cell proliferation, and tumorogenesis (5, 12, 13). Initial work with LRH-1 was performed largely in the liver, in which it is a key regulator of the conversion of cholesterol to bile acid salts through the transcriptional control of the cytochrome P450 7A gene (CYP7A1) (5, 14, 15). Subsequently, LRH-1 expression was detected in ovarian granulosa cells in which SF-1 was already known to be expressed (16–18).

LRH-1 and SF-1 are highly related proteins with 96% amino acid similarity in the DBD and 75% amino acid similarity in the LBD. Consistent with the highly conserved DBD, LRH-1 and SF-1 regulate several of the same target genes including many that are involved in reproductive function. These include inhibin-α (19, 20) and cholesterol side-chain cleavage (CYP11A1, P450scc) (21, 22). Although they share homology and regulate many common target genes, LRH-1 and SF-1 have unique patterns of expression and regulation in the ovary (18, 23), and ovarian conditional knockout mice indicate that each is critically important for normal reproductive function and fertility (24, 25).

Coactivators interact with NR and enhance their activity by recruiting/interacting with components of the RNA polymerase II preinitiation complex, altering chromatin structure through histone modification or adding posttranslational modifications directly to the receptor (26). Coactivators such as steroid receptor coactivator 1 (SRC-1) and cAMP response element-binding protein (CREB) binding protein, which increase the activation potential of many nuclear receptors and are broadly expressed, coactivate LRH-1 and SF-1 on multiple target gene promoters including the inhibin-α gene (20, 27–30). Other, more specific NR5A coregulators have been reported, including multibridging factor 1 (31, 32), β-catenin (33, 34), peroxisome proliferator-activated receptor-γ coactivator-1α (35, 36), and proline-rich nuclear receptor coregulatory protein (37). In addition, the corepressors prospero-related homeobox 1 (38) and zinc finger protein 67 kDa (39) have been shown to modulate the activity of the NR5A nuclear receptors.

To better understand the transcriptional activation of the NR5A receptors within the reproductive axis, we performed a yeast two-hybrid screen of a human ovarian cDNA library to identify NR5A interacting proteins and thus possible coregulatory proteins. We identify the LIM domain proteins of the four and a half LIM (FHL) family, FHL1 and FHL2, as being expressed in the ovary and as interacting with LRH-1 and SF-1. The FHL domain family of multifunctional proteins is composed of five members (FHL1–4 and activator of cAMP response element modulator in the testes), with each gene displaying tissue specific expression. FHL2 is expressed in the heart, gonads, prostate, adrenal gland, and intestines (40, 41), whereas FHL1 is more broadly expressed in tissues including in the heart, gonads, muscle, kidney, lung, and brain (41). FHL3 expression is more restricted with high expression in skeletal muscle and lower expression in the heart (41, 42), whereas FHL4 and activator of cAMP response element modulator in the testes are exclusively expressed in germ cells of the testes (41, 43, 44). Each LIM domain is composed of two zinc fingers; however, the FHL proteins do not bind to DNA. Instead the FHL proteins are often characterized as protein interaction partners and are reported to act as tissue-specific coactivators [e.g. FHL2 is a known coactivator of the AR (45)]. Although FHL1 and FHL2 expression has been detected in the ovary (41, 46), a function in reproduction has not yet been established. We demonstrate in this report that knockdown of FHL1 and FHL2 decreases the expression of genes involved in mammalian reproduction and that FHL2 functions as a coactivator of the NR5A proteins in activating the inhibin-α subunit gene in ovarian granulosa cells. We also show that FHL2 interacts with CREB and may integrate multiple signaling pathways controlling inhibin gene expression.

Materials and Methods

Animals

CD1 mice (Harlan, Indianapolis, IN) were maintained in accordance with the policies of Northwestern University's Animal Care and Use Committee. Mice were housed at a constant temperature and humidity in a 12 h light, 12-h dark cycle with food and water available ad libitum. Mice were fed soy-free chow to limit exogenous phytoestrogen intake through food.

Plasmids

cDNA plasmids were kindly provided by the following people: mouse LRH-1, Dr. Alan Tall (Columbia University, New York, NY) (47); mouse SF-1, Dr. Larry Jameson (Northwestern University, Feinberg School of Medicine, Chicago, IL); human ERβ and human glucocorticoid receptor (GR), Dr. Peter Kopp (Northwestern University, Feinberg School of Medicine, Chicago, IL); mouse FHL1, Dr. Ju Chen (University of California, San Diego, La Jolla, CA); and rat CREB, Dr. Jon Kornhauser (Children's Hospital, Boston, MA). GAL4 activation assay fusion constructs were generated in the pBIND vector (CheckMate mammalian two-hybrid system; Promega, Madison, WI). Yeast two-hybrid bait constructs were generated in the pGBKT7-DBD vector (MatchMaker GAL4 Yeast Two-Hybrid Kit System 3; CLONTECH, Mountain View, CA). A human FHL2 construct was generated by digesting an FHL2 cDNA identified in the yeast two-hybrid assay followed by ligation into pcDNA3.1 (Invitrogen, Carlsbad, CA). Mouse FHL2 was generated by PCR using mouse ovary cDNA as a template. The PCR product was cloned into to the pCR2.1 TA cloning vector (Invitrogen) and then subcloned into pcDNA3.1. Glutathione-S-transferase (GST) expression vectors were generated by cloning inserts into the pGEX2T vector (GE Life Sciences, Pittsburgh, PA). Mutant constructs were generated with the QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA).

Yeast two-hybrid screens and assay

The yeast two-hybrid screen was performed using the MatchMaker GAL4 Yeast Two-Hybrid System 3 (CLONTECH) according to the manufacturer's protocol. A human ovary cDNA library (in pACT2 vector; CLONTECH) was screened with the LRH-1 K289R LBD/pGBKT7-DBD construct with a transformation efficiency of 3.2 × 10⁴ CFU/μg DNA. Interaction colonies were isolated from SD-HIS/-TRP/-LEU containing 10 mm 3-amino-1,2,4-triazole. β-Galactosidase assays (o-nitrophenyl β-galactosidase substrate) were performed with 10 μg of mated yeast protein lysate in a 96-well plate format. Reactions were performed at 30 C for 60 min. Plates were read on a SpectraMax 384 Plus plate reader (Molecular Devices, Sunnyvale, CA) at 420 nm.

Cell culture, transfection, and luciferase assay

GRMO2 cells (48) were maintained and transfected as previously described (49). Cells were transfected with 500 ng of the rat inhibin-α promoter (547 bp) fused to a luciferase reporter construct (inhibin-α luc) (50), 50 ng of LRH-1 or SF-1, and 50 ng of human FHL2 when necessary. DNA quantities were held constant with empty vector. GAL4 activation assays were performed with 500 ng of pG5-luc and 50 ng of LRH-1 or SF-1. Samples were assayed for luciferase activity using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA). Luciferase values were normalized to total protein content per well as assayed by a Bradford reagent (Bio-Rad Laboratories, Hercules, CA) per the manufacturer's protocol.

GST protein expression, purification, and pulldowns

GST fusion proteins were expressed in BL21 CodonPlus cells (Stratagene). Cultures were induced with 0.2 mm isopropyl β-d-1-thiogalactopyranoside and grown at 30 C for 1–2 h. Pelleted cells were resuspended in lysis buffer (50 mm Tris-Cl, pH 8.0; 150 mm NaCl; 1 mg/ml lysozyme; 50 μm ZnCl) and subjected to three rounds of sonication on ice: 1 × 20 sec, 2 × 10 sec. After sonication, the cells were lysed with Triton X-100 (1% final) and digested with deoxyribonuclease (50 μg/ml final) at 4 C with agitation for 30 min. Cell debris was harvested by centrifugation and the resulting supernatant was bound to 50% glutathione 4B sepharose (Pierce, Rockford, IL) overnight. Bound proteins were washed twice in a low-salt buffer (50 mm Tris-Cl, pH 8.0; 150 mm NaCl) and once in a high-salt buffer (50 mm Tris-Cl, pH 8.0; 500 mm NaCl). Proteins were stored at 4 C on the beads in low-salt buffer containing protease inhibitors. GST pulldowns were performed in a buffer containing 20 mm HEPES (pH 7.5), 50 mm KCl, 25 mm MgCl₂, 1 mm dithiothreitol (DTT), 0.1 mm EDTA, and 0.15% Nonidet P-40. In vitro transcribed and translated proteins were generated with transcribed and translated T7 coupled reticulocyte lysate system (Promega) incorporating ³⁵S-methionine (PerkinElmer, Norwalk, CT). A defined amount of each GST fusion protein bound to glutathione 4B sepharose was used in each pulldown, and the proteins were coincubated at 4 C for 2 h. Two microliters of each ³⁵S-labeled protein were used in the pulldowns. The interacting protein complexes bound to the beads were washed three times with pulldown buffer and then boiled in a sodium dodecyl sulfate (SDS)-containing buffer. Eluted proteins were resolved on a 10% SDS-PAGE gel. Gels were fixed in isopropanol/water/acetic acid (25:65:10), impregnated in Amplify solution (GE Life Sciences), dried, and exposed to autoradiographic film at −80 C with intensifying screens.

Immunohistochemistry and immunocytochemistry

Ovaries were harvested from 18-d-old mice. Immunohistochemistry was performed as previously described (51) using an anti-FHL2 polyclonal antibody, diluted 1:2000 (kindly provided by Dr. Beat Schäfer, University Children's Hospital, Zurich, Switzerland) (40) or an anti-FHL1 polyclonal antibody diluted 1:1000 (no. 49241, Abcam, Cambridge, MA). For immunocytochemistry of endogenous or overexpressed proteins in GRMO2 cells, cells were plated overnight on glass coverslips. For detection of overexpressed proteins, cells were transfected with 500 ng of FLAG-tagged constructs for 24 h before immunocytochemistry detection. For endogenous and overexpressed proteins, cells were fixed in 1% paraformaldehyde and permeabilized in PBS/0.25% Triton X-100. Blocking was performed in PBS/0.25% Triton X-100 with 5% serum followed by incubation with primary antibody overnight at 4 C: anti-FHL2 1:600 (kindly provided by Dr. Roland Schüle, University of Freiburg, Freiburg, Germany) (52), anti-FLAG 1:250 (no. F3165; Sigma, St. Louis, MO). After an incubation with fluorophore-conjugated secondary antibodies (1:200; Sigma), the coverslips were incubated with fluorescent phallotoxins [phalloidin-TRITC (Sigma) or Alexa Fluor 488 phalloidin (Invitrogen)] to stain the actin filaments. Coverslips were mounted on slides with Vectastain [4′,6′-diamino-2-phenylindole (DAPI); Vector Laboratories, Burlingame, CA] for visualization of nuclei. Images were captured on a Leica DM5000B microscope (Leica Microsystems, Wetzlar, Germany).

Primary granulosa cell isolation and small interfering RNA (siRNA) knockdown

Primary granulosa cells were extracted from the ovaries of d 21–24 immature CD1 mice. Ovaries were trimmed of fat and collected into 4F media (1× DMEM/F12; 15 mm HEPES, pH 7.4; 5μg/m; apo-transferrin; 2μg/m; insulin; 40 ng/m; hydrocortisone; 2.43 g/liter sodium bicarbonate; 0.5× penicillin-streptomycin) containing 0.5 m sucrose and 10 mm EGTA, incubated at 37 C for 20 min to loosen cell junctions, and washed three times in cold 4F media. Cells were isolated by follicle puncture in 4F media using 26-gauge needles under a Reichert Stereo Star ZOOM dissection microscope (Reichert, Heidelberg, Germany). The extruded granulosa cells were collected via centrifugation in a swinging bucket rotor (500 × g, 10 min, Accuspin 3R; Fisher Scientific, Fair Lawn, NJ) and resuspended in 4F media containing 10% fetal bovine serum (FBS). The resuspended cells were filtered through a 40-μm nylon mesh strainer (Millipore, Bedford, MA) to remove oocytes and tissue pieces. The granulosa cells were plated in 4F media/10% FBS at an approximate density of 2 × 10⁵ cells/well in a 12-well plate. Cells were allowed to adhere for 48 h and then transfected with 10 nm of each siRNA duplex pool (20 nm per well total with the amount held constant with control duplexes when necessary) (On-Target Plus Smartpool; Dharmacon, Lafayette, CO) using Lipofectamine RNAiMax reagent (Invitrogen) for 6 h. Cells were recovered in 4F media/10% FBS, and RNA or protein was extracted 72 h after transfection.

RNA extraction and PCR

RNA was extracted from primary mouse granulosa cells or from whole tissues using the QIAGEN RNAeasy kit (Valencia, CA). cDNA was generated using avian myeloblastosis virus-reverse transcriptase and random hexamer primers. Semiquantitative PCR was performed in the presence of ³²P-deoxy-CTP radiolabeled nucleotides (Amersham Pharmacia Biotech, Piscataway, NJ). The resulting PCR products were electrophoresed on a 5% acrylamide gel. The gels were dried and exposed to autoradiographic film. Quantitative PCR (qPCR) was performed with SYBR green master mix (Applied Biosystems, Carlsbad, CA) with 100 ng of reverse-transcribed cDNA per well. Reactions were performed using a 7300 Realtime Machine (Applied Biosystems). The primers used in the semiquantitative PCR and qPCR assays are listed in Table 1.

Protein extraction and Western blotting

Whole-cell protein lysates were generated from primary granulosa cells as previously described (49). Protein concentrations were determined using a Bradford protein dye (Bio-Rad Laboratories). Then 13.5% SDS-PAGE gels were run with 100 μg of each protein lysate. Immunoblotting was performed as previously described with antibodies specific to each protein of interest [anti-FHL2, 1:8000 (kindly provided by Dr. R. Schüle, University of Freiburg, Freiburg, Germany)] (52), antiinhibin-α 1:1000 (kindly provided by Dr. W. Vale, Salk Institute, La Jolla, CA), and anti-actin 1:1500, no. A2066; Sigma). The nitrocellulose membranes were stripped with Re-Blot strong stripping solution (Millipore) after blotting with each primary antibody.

Chromatin immunoprecipitation (ChIP) assay

ChIP assays were performed with GRMO2 cells grown in 10-cm dishes in normal growth media. The cells were stimulated with 10 μm forskolin (FSK) or dimethylsulfoxide (1:1000) in growth media for 1 h at 37 C. The cells were washed with PBS and then cross-linked with 1% formaldehyde (in PBS) for 10 min at 37 C. The cross-linking was quenched with 125 mm glycine, on ice, for 5 min. The cells were washed and collected in PBS, pelleted by centrifugation, resuspended in lysis buffer (50 mm Tris-Cl, pH 7.9; 10 mm EDTA; 1% SDS; 1 mm DTT; protease inhibitors) and lysed for 10 min on ice. The cell debris was removed by centrifugation. The cross-linked chromatin was sheared via sonication to an average size of 500 bp or less (10 × 20 pulses on ice, 30% output). The sheared chromatin was diluted 1:10 in dilution buffer (20 mm Tris-HCl, pH 7.9; 150 mm NaCl; 2 mm EDTA; 0.5% Trion X-100; 1 mm DTT; 2 μg/ml sonicated salmon sperm; and protease inhibitors) and then precleared with 20 μl of a 50% protein A slurry (2 h at 4 C, nutating). The chromatin-protein complexes were incubated overnight at 4 C with the following antibodies: 2 μl FHL2 (no. 66399; Abcam), 2 μl LRH-1 (53), 2 μl SF-1 (no. PA1-800; Affinity BioReagents, Golden, CO), 10 μl CREB (no. sc-186; Santa Cruz Biotechnologies, Santa Cruz, CA) The antibody-bound complexes were pulled down (2 h, 4 C, nutating) with 40 μl of 50% protein A beads blocked with 100 μg of sonicated salmon sperm DNA and 100 μg BSA per 100 μl of 50% slurry beads. The beads were repeatedly washed to remove unbound chromatin-protein complexes (3 × 5min with 20 mm Tris-HCl, pH 8.0; 250 mm NaCl; 2 mm EDTA; 0.25% Nonidet P-40; 0.05% SDS; protease inhibitors; and 1× instant wash with Tris/EDTA). The chromatin-protein complexes were simultaneously eluted from the beads and reverse cross-linked at 65 C overnight followed by a proteinase K digestion for 30 min at 42 C. The DNA was extracted with phenol/chloroform/isoamyl alcohol followed by a sodium acetate and ethanol precipitation. The DNA was resuspended in 50 μl of water and used in subsequent qPCR reactions using ABI SYBR Green Mastermix (7300 Realtime Machine; Applied Biosystems) (for qPCR primer sequences, see Supplemental Table 1, published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org).

Statistical analysis

Statistics were performed using GraphPad Prism 4.02 for Windows (GraphPad Software, La Jolla, CA). The data were analyzed using a two-tailed, paired Student's t test. P < 0.05 was considered statistically significant.

Results

NR5A proteins potentially interact with FHL2

Proteins that might modulate NR5A function were identified in a yeast two-hybrid interaction assay. A human ovary library was screened with an LRH-1 carboxy terminus construct containing a mutation of a known small ubiquitin-related modifier acceptor site (SUMOylation site) (LRH-1 K289R LBD, Fig. 1A) (54). This construct was used in the yeast two-hybrid screen because it had robust activity in a GAL4 activation assay in granulosa cells compared with a wild-type (WT) construct (Supplemental Fig. 1). Six of 120 isolated and sequenced cDNA clones were identified as the multifunctional protein FHL2 (Fig. 1B, yeast growth assay) (40). The entire FHL2 open reading frame was present in the clones isolated.

Fig. 1.

FHL2 interacts with the NR5A proteins in yeast. A, Schematic representation of the LRH-1 and SF-1 constructs used in the yeast two-hybrid assays. B, Yeast two-hybrid growth assay in which the GAL4-DBD vector expresses tryptophan (TRP⁺), GAL4-AD vector expresses leucine (LEU⁺), with a positive interaction activating an endogenous histidine (HIS⁺) reporter in the yeast. A positive interaction is indicated by yeast growth on a synthetic medium lacking HIS, TRP, and LEU. 3-Amino-1,2,4-triazole was included in the medium to block leaky HIS expression. C, β-Gal assays performed with yeast cell lysates. The data represent the mean ± sem of three independent experiments performed in duplicate.

Because the yeast two-hybrid screen was performed with the LRH-1K289R LBD construct, we asked whether FHL2 could also interact with WT LRH-1 LBD. Yeast β-galactosidase (β-gal) assays demonstrated that the strength of interaction between FHL2 and LRH-1 WT LBD or LRH-1 K289R LBD was similar (Fig. 1C, 49.8 ± 5.3 vs. 52.5 ± 5.0 β-gal specific activity units, respectively). Therefore, the presence of the LRH-1 small ubiquitin-related modifier (SUMO) acceptor site does not affect the interaction with FHL2 in yeast cells. Because the LRH-1 construct used had 75% amino acid similarity (55% amino acid identity) with the equivalent region of SF-1, we asked whether FHL2 could also interact with SF-1. The β-gal-specific activity generated between the interaction of FHL2 and WT SF-1 LBD or SF-1 K194R LBD (mutation of a known SF-1 SUMO acceptor site) (54, 55) was 45.2 ± 7.0 and 51.3 ± 1.7, respectfully, demonstrating that FHL2 also interacts with SF-1 in yeast cells (Fig. 1C).

Many coregulatory proteins are known to interact with the activation function 1 and/or activation function 2 (AF2) region of nuclear receptors. For example, mutation of the second pair of hydrophobic residues in the AF2 hexamer (Fig. 1A) has previously been shown to abolish interaction between SF-1 and SRC-1 (27). Mutation of the LRH-1 or SF-1 AF2 region (LLIEML to LLIEAA) substantially decreased the interaction with FHL2 (Fig. 1C). FHL2 did not interact with the ERβ or GR LBD in the yeast β-gal assays (Fig. 1C), demonstrating specificity of the NR5A interaction. However, FHL2 has been previously reported to interact with the AR (45).

FHL proteins are expressed in granulosa cells

As noted above, we isolated full-length FHL2 from the human ovary cDNA library. Previous reports have demonstrated that FHL2 is expressed in the ovary, but these studies did not examine the localization of FHL2 in the ovary. To confirm the expression of FHL2 in the rodent ovary, we performed semiquantitative RT-PCR using FHL2-specific primers and mouse ovary or liver cDNA. FHL2 was expressed in the ovary but not in the liver, as previously reported in mouse tissues (Fig. 2A) (41). In addition, FHL2 was expressed in GRMO2 cells, which are an immortalized mouse ovarian granulosa cell line (Fig. 2A, right panel). Using immunohistochemistry, we demonstrated that FHL2 protein is expressed in the ovary predominantly in the granulosa cells and oocytes of follicles of all sizes, with lesser expression in thecal cells (Fig. 2B, with the negative control located in Supplemental Fig. 2A).

Fig. 2.

FHL family proteins are expressed in granulosa cells. A, Semiquantitative RT-PCR to detect FHL2 or RPL19 mRNA expression in mouse tissues and GRMO2 cells. B, Immunohistochemistry with an antibody against FHL2 (40) on a paraffin-embedded d 18 mouse ovary section. Closed arrow indicates granulosa cells. C, Immunocytochemistry on GRMO2 cells using an antibody against FHL2 (green) (52). The nuclei were detected with DAPI (blue) and actin filaments were detected with phalloidin (red). D, Immunocytochemistry on GRMO2 cells using an antibody against FLAG to detect overexpressed FHL2-FLAG (red). The nuclei were detected with DAPI (blue) and actin filaments were detected with phalloidin (green). E, Quantitative RT-PCR analyzing the relative mRNA expression of genes in primary mouse granulosa cells. The data represent the mean ± sem of three independent experiments. F, Immunohistochemistry with an antibody against FHL1 on a paraffin-embedded d 18 mouse ovary section. Closed arrow indicates granulosa cells. Negative controls are found in Supplemental Fig. 2.

FHL2 is known to be present and have functions both in the cell nucleus and cytoplasm (52). Immunofluorescence localized endogenous FHL2 protein to both the nucleus and cytoplasm of GRMO2 cells (Fig. 2C). The extranuclear FHL2 expression overlaps with actin filaments stained with phalloidin, with condensed punctuate staining at the end of stress fibers. FHL2 is known to interact with α- and β-integrin chains (56) as well as focal adhesion kinase (57); thus, these punctuate structures are likely focal adhesion complexes. Overexpressed FHL2-FLAG is expressed predominantly in the nucleus of GRMO2 cells (Fig. 2D).

FHL2 belongs to a family of proteins (FHL1–5) that display tissue specific but, in some cases, overlapping expression. The FHL2 family member FHL1 is also reported to be expressed in the ovary (41). We demonstrated that FHL1 is expressed in primary mouse granulosa cells, albeit at a much lower level than FHL2 (Fig. 2E). Similar to FHL2, FHL1 protein is expressed predominantly in the granulosa cells and oocytes of follicles of all sizes, with lesser expression in thecal cells (Fig. 2F, with the negative control located in Supplemental Fig. 2B).

Knockdown of FHL proteins attenuates ovarian gene expression

To determine whether FHL1 and FHL2 have a functional role in gene regulation in the ovary through their interaction with NR5A receptors, we performed siRNA-mediated knockdown experiments in primary mouse granulosa cells and examined the effect on the expression of several NR5A target genes. Both inhibin-α (Inha) and p450scc (Cyp11a1) are known to have NR5A binding sites and be regulated by NR5A proteins (19, 22). Upon cotransfection of the FHL1 and FHL2 siRNA smart pools, endogenous FHL1 mRNA expression was reduced by 62% (Fig. 3A), whereas FHL2 mRNA levels were reduced by 49% (Fig. 3B) compared with control siRNA. Each siRNA smart pool was independently verified for the ability to specifically reduce the mRNA level of the targeted gene: Fhl1 = 63% and Fhl2 = 47% (data not shown). The reduction in Fhl2 mRNA levels resulted in a 40% decrease in protein level (Fig. 3C). The combined knockdown of Fhl1 and Fhl2 resulted in a small but significant 18% reduction in Inha mRNA expression (Fig. 3D). The inhibin-α protein level (pro-inhibin-α) was significantly decreased by 29% when FHL1 and FHL2 were simultaneously knocked down (Fig. 3E). The NR5A target gene Cyp11a1 mRNA expression was also reduced by 23% after cotransfection of FHL1 and FHL2 siRNA smart pools compared with control siRNA (Fig. 3F). Thus, FHL1 and FHL2 appear to be important for the normal expression of these target genes in ovarian granulosa cells.

Fig. 3.

Knockdown of FHL1/2 decreases the expression of NR5A target genes in granulosa cells. Fhl1 (A) and Fhl2 (B) mRNA expression in primary mouse granulosa cells were quantified by quantitative RT-PCR after siRNA-mediated knockdown of FHL1 and FHL2. The mRNA level of each sample was normalized to the internal control RPL19 and plotted as fold over control siRNA. The data represent the mean ± sem of three independent experiments. C, Densinometric quantification of FHL2 protein levels normalized to actin protein levels, as analyzed by immunoblotting (not shown), after siRNA-mediated knockdown of FHL1 and FHL2. The data are plotted as fold over control siRNA and represents the mean ± sem of at least three independent experiments. Inha mRNA (D), inhibin-α protein (E), and Cyp11a1 (F) mRNA was analyzed and plotted as described in A–C. Bars marked by asterisk are significantly different (P < 0.05) from the corresponding control in the same graph.

FHL family proteins interact with NR5A proteins through specific domains

To ascertain whether the FHL proteins interact with LRH-1 or SF-1 in vitro, GST pulldown assays were performed. Recombinant FHL1 and FHL2 were purified as GST fusion proteins from Escherichia coli cells. GST-FHL1 and GST-FHL2 interacted with in vitro-transcribed and -translated ³⁵S-methionine-labeled LRH-1 and SF-1 (Fig. 4). GST-FHL2 did not interact with ³⁵S-methionine GR (Fig. 4, bottom, left panel), indicating that the interaction with the NR5A proteins was specific, as was observed in the yeast interaction studies (the GST-FHL1/GR interaction was not examined). Because granulosa cells express far more FHL2 than FHL1 (Fig. 2E) and because only FHL2 was identified in the initial yeast two-hybrid screen, we focused our remaining studies on FHL2.

Fig. 4.

FHL and NR5A proteins interact in vitro. GST pulldown assays were performed with full-length GST-FHL2 (left panels) or GST-FHL1 (right panels). Ten percent of the input sample was run as a positive control, whereas GST alone was used as a negative control.

FHL2 is composed of four and a half highly homologous LIM domains, with each consisting of two zinc fingers (Fig. 5A). The FHL2 amino terminal LIM domains (LIM 0–2) and carboxy terminal LIM domains (LIM 3–4) were purified as GST fusion proteins, and we found that LRH-1 and SF-1 interacted with both the amino and carboxy terminal LIM domains of FHL2 (Fig. 5A, lanes 1–5). The LIM domains were next purified as individual GST fusion proteins (GST-LIM 0, GST-LIM 1, GST-LIM 2, GST-LIM 3, and GST-LIM 4) and used in pulldown assays. LRH-1 and SF-1 interacted with the single zinc finger, half LIM domain (Fig. 5A, lane 9) as well as the individual full LIM domains (Fig. 5A, lanes 6–13; SF-1 was not tested with GST-LIM 3 or GST-LIM 4). Thus, each of the individual LIM domains is capable of independent interaction with the NR5A receptors.

Fig. 5.

Mapping of the interaction domains between FHL2 and NR5A proteins. GST pulldown assays were performed with GST fused to different LIM regions of FHL2 and incubated with full-length SF-1 or LRH-1 (A) or GST fused to full-length FHL2 and incubated with the indicated domains of LRH-1 (B and C). Ten percent of the input sample was run as a positive control, whereas GST alone was used as a negative control. Images taken from different gels are indicated by a break between the panels.

To explore which region of the NR5A proteins interacts with FHL2, deletion constructs were generated that spanned LRH-1 (Fig. 5B). A very weak interaction was present between LRH-1 AF1 and FHL2 in yeast assays (Fig. 1C), but this interaction was not detected in the GST pulldowns (Fig. 5B, lanes 1–3). However, GST-FHL2 did interact with the LRH-1 DBD when it was expressed as an independent domain (Fig. 5B, lanes 7–9) or as part of the amino terminal (Fig. 5B, lanes 4–6) or carboxy terminal (Fig. 5B, lanes 10–12) domains of the protein. In agreement with the yeast β-gal data (Fig. 1C), the LRH-1 LBD also interacted with GST-FHL2 in the absence of the LRH-1 DBD (Fig. 5B, lanes 13–18). Thus, both the DBD and LBD of LRH-1 appear to facilitate interaction with FHL2. Although the LBD association is typical of many coactivators, the DBD interaction is rather unique and was explored further.

The NR5A DBD is composed of a core DBD containing two zinc fingers followed by a carboxy terminal α-helix (FTZ-F1 helix) that is necessary for protein-DNA complex stability (49, 58). We previously hypothesized that the FTZ-F1 helix may be an interaction surface for protein association (49). Consistent with this hypothesis, GST-FHL2 did not interact with the LRH-1 DBD when the FTZ-F1 helix was deleted (LRH-1 DBD ΔC) (Fig. 5C, lanes 4–6). When the FTZ-F1 helix was disrupted by a mutation introducing two proline residues (LRH-1 R195P, D196P), the interaction with GST-FHL2 was decreased compared with WT LRH-1 DBD (Fig. 5C, lanes 7–9). This same mutation is known to reduce the transcriptional potential of SF-1 (49).

FHL2 coactivates the NR5A family of NR in granulosa cells

Because we determined that NR5A target gene expression was reduced after FHL2 knockdown in granulosa cells and we identified in vitro interactions between FHL2 and the NR5A proteins, we asked whether FHL2 had a functional effect on the transactivation potential of LRH-1 or SF-1. GRMO2 cells were transfected with a 547-bp proximal promoter construct of the rat inhibin-α subunit gene fused upstream of a luciferase reporter cassette (inhibin-α luc) (50). GRMO2 cells were cotransfected with LRH-1 alone or in combination with FHL2. FHL2 alone did not activate inhibin-α luc (Fig. 6A). LRH-1 transactivated inhibin-α luc 4.8-fold over basal levels and the activation was increased to 7.4-fold over basal when FHL2 was coexpressed. Similar coactivation results were obtained with SF-1 (Fig. 6B). SF-1 activated inhibin-α luc 2.3-fold over basal levels, and the addition of FHL2 increased the activation to 3.7-fold over basal. Therefore, FHL2 coactivates the NR5A nuclear receptors on the proximal promoter of the inhibin-α subunit gene in granulosa cells.

Fig. 6.

FHL2 coactivates the NR5A proteins in granulosa cells. GRMO2 cells were cotransfected with inhibin-α luc, FHL2, NR5A receptor, or the combination. A, Results for LRH-1. B, Results for SF-1. The data represent the mean ± sem of at least 11 experiments performed in triplicate. Bars with different letters are significantly different from each other by a value of P ≤ 0.05.

FHL2 interacts with CREB and enhances synergism between NR5A and cAMP regulation of the inhibin-α promoter

FHL2 has been reported to interact with and coactivate CREB (41). We confirmed the interaction of FHL2 and CREB in yeast cell lysates using β-gal assays and in vitro using GST pulldown assays (Supplemental Fig. 3). The NR5A proteins bind to the inhibin-α promoter at a conserved site (the SF-1 binding site that is 8 bp 5′ to a conserved cAMP response element that binds CREB (19, 20, 50) (Fig. 7A). Previous work indicated that NR5A proteins and cAMP signaling exhibit synergism in activating the inhibin-α promoter and that the NR5A proteins do not activate the promoter if the CRE site is mutated (20), leading to the conclusion that there is an integration between the NR5A and CREB responses. Such integration could occur in part through a bridging protein such as FHL2. LRH-1 and CREB individually interact with GST-FHL2, and they both interact with GST-FHL2 when present in the same pulldown (Fig. 7B). The simultaneous, direct interaction of these factors with GST-FHL2 indicates that they do not compete for binding to the same interaction surface of FHL2. This is likely because the multiple LIM domains each have the ability to bind to both proteins. To test whether these factors are simultaneously recruited to the inhibin-α promoter in granulosa cells, we performed ChIP experiments in GRMO2 cells (Fig. 7C). We found that FHL2 is recruited to the promoter along with both SF-1/LRH-1 and CREB and that this recruitment increases after the activation of the protein kinase A (PKA) pathway with the adenylate cyclase activator FSK.

Fig. 7.

FHL2 interacts with both NR5A receptors and CREB and is recruited to the inhibin promoter with these factors after FSK stimulation. A, Gene structure schematic of the mouse inhibin-α promoter. The NR5A and CREB response elements and the promoter region included in the luciferase reporter construct are indicated. The ChIP qPCR primer position is also indicated. B, LRH-1 and CREB do not compete for interaction with FHL2. GST pulldown was performed with full-length GST-FHL2. Ten percent of the input sample was run as a control. C, ChIP assay showing the relative enrichment of FHL2, NR5A receptors, and CREB recruited to the inhibin-α promoter after FSK stimulation in GRMO2 cells. The data are plotted with the IgG immunoprecipitations subtracted as background. The data represent the mean ± sem of at least three experiments. The asterisk marks the factors with significantly different relative enrichment (P < 0.05) between dimethylsulfoxide (DMSO) and FSK treatment.

We next tested the functional effect of this corecruitment by asking whether FHL2 could coactivate LRH-1 after the activation of the PKA pathway. Stimulation of the PKA pathway with FSK, which is known to increase CREB phosphorylation, increased the activation of inhibin-α luc. We did not observe a significant increase in this response in the presence of cotransfected FHL2 (Fig. 8A). However, the combination of LRH-1 and FSK increased the activation of inhibin-α luc to 15.7-fold over basal activity, which was increased to 31.5-fold over basal activity upon the addition of FHL2 (Fig. 8A). Similar results were generated with FSK stimulation in the presence of both SF-1 and FHL2 (Fig. 8B). Thus, FHL2 augments the combined response to NR5A receptors and the cAMP-PKA-CREB pathway and may serve to integrate the activity of these two important regulators of reproductive hormone gene expression.

Fig. 8.

FHL2 augments inhibin-α promoter activity in the presence of NR5A receptors and FSK. GRMO2 cells were cotransfected with inhibin-α luc and LRH-1 and FHL2 (A) or SF-1 and FHL2 (B). Twenty-four hours after transfection, the cells were stimulated with 10 μm FSK for 6 h. The data represent the mean ± sem of at least eight experiments performed in triplicate. Bars with different letters are significantly different from each other by a value of P ≤ 0.05.

Discussion

We have identified FHL2 as an interactor of the NR5A nuclear receptors LRH-1 and SF-1, shown that FHL2 coactivates the NR5A receptors on the inhibin-α subunit promoter, and demonstrated that FHL2 interacts with CREB and enhances the combined effects of NR5A receptors and cAMP/CREB signaling on the inhibin-α subunit gene. FHL2 is a multifunctional protein that can shuttle between the cytoplasm and nucleus. FHL2 is a major component of cytoskeletal architecture, interacting with actin filaments (59) and the cytoplasmic tails of multiple integrin chains (56, 60). Stimulation of cells with serum or activation of Rho GTPases causes translocation of FHL2 from the cytoplasm to the nucleus (52, 61). Once in the nucleus, FHL2 stimulates the transcription of genes involved in cell proliferation and tumorogenesis (40, 52, 62), bone formation (63, 64), and cardiac function. Transcription factors reported to be coactivated by FHL2 include the AR (45), CREB (41), the aryl hydrocarbon receptor (65), activator protein 1 (61), and Wilm's tumor 1 (WT-1) (46).

Two groups have generated FHL2 knockout mice, and both strains are viable (66, 67). Although FHL2 is expressed in the myocardium of the developing murine heart, these mice have largely normal cardiac development (66, 67). Slight functional abnormalities were reported in the FHL2 knockout mice including exaggerated cardiac hypertrophy in response to a sustained β-adrenergic stimulus (67), osteopenia due to reduced activity of osteoblasts (63), and impaired wound healing due to a slow cell migration resulting from decreased expression of α-smooth muscle actin and decreased effector functions of focal adhesion kinase including reduced expression of p130cas and decreased activity of Rac GTPase (68). The FHL2 knockout mice are fertile, but reproductive function in these mice has not been carefully examined. It is postulated that the lack of a severe phenotype in these mice may be due to functional redundancy with other FHL proteins, including FHL1, which are expressed in the same tissues.

Partial knockdown of FHL1 and FHL2 in primary granulosa cells resulted in a significant decrease in NR5A target gene expression. We were unable to achieve a full knockdown of FHL1 and FHL2, perhaps because of the specific duplexes used or perhaps because of the inefficiency in transfecting these primary cells, and this likely contributes to the modest impact on target gene expression. In addition, we know that the promoters of these genes are under complex and multifactorial regulation (20), so it is perhaps not surprising that knockdown of this single factor did not have a larger effect. Nonetheless, these studies suggest that FHL family proteins are necessary for full ovarian expression of genes that play key roles in reproductive function.

FHL2 interacts with the NR5A nuclear receptors both in vivo and in vitro. We determined that LRH-1 and SF-1 interact with full-length FHL2 and that a single LIM domain is sufficient for the interaction. The FHL2 LIM domains required for protein interaction vary, depending on the partner protein. The AR was shown to interact with FHL2 LIM 0–2 and LIM 3–4, but individual LIM domain interactions were not analyzed (45). The interaction between FHL2 and β-catenin requires all of the FHL2 LIM domains (69, 70), but the interaction between FHL2 and WT-1 requires only LIM 0–2 (46). The interaction between FHL2 and ERK-2 is strongest with LIM 1–4, but an interaction was also detected with only LIM 2 (71). There are 58 known human LIM domain proteins, which are classified into three groups: LIM only proteins, LIM homeodomain proteins, and LIM kinase proteins (72). The FHL family members are LIM-only proteins, and their effects on gene regulation are likely indirect and are thought to be mediated through protein-protein interactions, such as those demonstrated here for LRH-1 and SF-1.

The NR5A nuclear receptors, like other nuclear receptors, have an AF2 hexamer located near the carboxy terminal end of the LBD (27, 73), and this forms a docking site for coactivators. Yeast β-gal assays demonstrate that this AF2 hexamer is required for the full interaction of both LRH-1 and SF-1 with FHL2. However, we determined that FHL2 also interacts with the FTZ-F1 helix in the NR5A DBD. The FTZ-F1 helix forms a hydrophobic patch on the surface of the DBD that was previously hypothesized to act as a docking site for interacting factors (49). Other coregulatory proteins are reported to interact with the NR5A DBD including WT-1 (74) and multibridging factor 1 (31, 32). ER and GR both have a highly conserved AF2 hexamer but lack the FTZ-1 helix, and neither interacts with FHL2. Thus, the FTZ-F1 helix may provide an additional surface contributing important specificity to the interaction between NR5A receptors and FHL2.

Well-characterized coactivators of LRH-1 and SF-1 include SRC-1 and CREB binding protein (20, 27–30). The transcriptional effects of these coactivators are mediated, in part, by their histone acetyltransferase activity (75). FHL2 is different from these types of coactivators because it possesses transcriptional coactivation capability but has no known inherent enzymatic activity. A well-documented characteristic of LIM domains is that they facilitate protein-protein interactions (72, 76), and FHL2 is often referred to as an adaptor or scaffolding protein (77, 78). It is through this scaffolding mechanism that FHL2 likely integrates transcriptional responses on the inhibin-α promoter through interactions with proteins such as the NR5A receptors and CREB. Interestingly, FHL2 has recently been shown to interact with and corepress the NR4A orphan receptor Nur77 (79), which is implicated in the negative regulation of inhibin gene expression after ovulation (80). Therefore, FHL2 may be a central interaction partner for the complex of proteins that regulate inhibin-α subunit gene expression in granulosa cells.

Collectively the results presented in this paper identify FHL2 as a novel coactivator of the NR5A family of nuclear receptors in ovarian granulosa cells and suggest roles for FHL2 and related LIM domain proteins in regulating mammalian reproduction.