NFI Transcription Factors Interact with FOXA1 to Regulate Prostate-Specific Gene Expression

Abstract

Androgen receptor (AR) action throughout prostate development and in maintenance of the prostatic epithelium is partly controlled by interactions between AR and forkhead box (FOX) transcription factors, particularly FOXA1. We sought to identity additional FOXA1 binding partners that may mediate prostate-specific gene expression. Here we identify the nuclear factor I (NFI) family of transcription factors as novel FOXA1 binding proteins. All four family members (NFIA, NFIB, NFIC, and NFIX) can interact with FOXA1, and knockdown studies in androgen-dependent LNCaP cells determined that modulating expression of NFI family members results in changes in AR target gene expression. This effect is probably mediated by binding of NFI family members to AR target gene promoters, because chromatin immunoprecipitation (ChIP) studies found that NFIB bound to the prostate-specific antigen enhancer. Förster resonance energy transfer studies revealed that FOXA1 is capable of bringing AR and NFIX into proximity, indicating that FOXA1 facilitates the AR and NFI interaction by bridging the complex. To determine the extent to which NFI family members regulate AR/FOXA1 target genes, motif analysis of publicly available data for ChIP followed by sequencing was undertaken. This analysis revealed that 34.4% of peaks bound by AR and FOXA1 contain NFI binding sites. Validation of 8 of these peaks by ChIP revealed that NFI family members can bind 6 of these predicted genomic elements, and 4 of the 8 associated genes undergo gene expression changes as a result of individual NFI knockdown. These observations suggest that NFI regulation of FOXA1/AR action is a frequent event, with individual family members playing distinct roles in AR target gene expression.

It is well recognized that signaling by the androgen receptor (AR) has important roles in normal prostate development, growth, and differentiation (1–3), as well as in benign and neoplastic conditions of the prostate (4). However, AR alone is not sufficient to mediate tissue-specific gene expression. Rather, it is the combinatorial control (5, 6) and activity of multiple factors that determine tissue-specific gene expression. Specifically, the ability of AR to engage other transcription factors (TFs) in a physical complex dictates tissue-specific gene expression in the prostate (7). In addition to the prostate, the AR is expressed in various tissues where it exhibits a distinct role for normal gene expression and physiology. For example, the AR in the skeletal muscle dictates anabolism of that tissue (8). Therefore, in addition to epigenetic mechanisms, it is the ability of AR to interact with other TFs that determines AR function in a given tissue.

Our interest in identifying factors that mediate tissue specificity of AR target gene expression led to identification of forkhead box (FOX) A1 (FOXA1) as an AR interacting protein (9, 10) and showed that this interaction is essential for the expression of AR-regulated, prostate-specific genes (for review, see Ref. 11). The FOXA family of proteins (FOXA1, FOXA2, and FOXA3) bind with differing affinity to the consensus DNA sequence [(A/C)AA(C/T)] and have been implicated in various developmental, homeostatic, and disease processes (12–14). Our focus has been on FOXA1 because FOXA2 is expressed only in neuroendocrine cells of the adult prostate and FOXA3 is not expressed in adult prostate (15). FOXA1 works as a “pioneer factor” and acts to increase TF accessibility to the DNA by displacing linker histones from nucleosomes, allowing for chromatin unfolding (16). Further studies by us and others have validated the importance of this AR/FOXA1 interaction in prostate cancer (14, 17–20) and demonstrated the interaction between FOXA1 and other steroid receptors (21–24).

The loss of FOXA1 in prostate cancer cell lines that express AR results in dramatic reprogramming of AR to different binding sites (20, 25). The ability of FOXA1 to interact with AR and specify binding to specific androgen response elements (AREs) suggests that other TFs involved with the AR/FOXA1 complex may further regulate tissue-specific gene expression. To identify novel TFs involved in the AR/FOXA1 transcription complex, we expressed a dual-tagged FOXA1 construct in an androgen-regulated prostatic cell line, LNCaP, and performed tandem affinity purification and mass spectrometry to identify a novel set of FOXA1 interacting proteins. Sixteen proteins were identified, only one of which, nuclear factor I X (NFIX), was a TF.

The NFI family of TFs contains 4 genes (NFIA, NFIB, NFIC, and NFIX) encoding proteins that bind to the consensus DNA sequence TTGGCN₅GCCAA (26). NFI family members can form either homodimers or heterodimers with each other, and these dimers have comparable affinity for DNA, stability, and specificity (27), suggesting that dimer combinations will be dictated by tissue-specific expression. Indeed, NFIX has been identified as a stromal-specific factor in the human prostate, whereas NFIB has been classified as basal-specific (28). Knockout studies of individual NFI genes in mice have revealed a variety of phenotypes, including corpus callosum agenesis (NFIA) (29), lung hypoplasia (NFIB) (30), tooth defects (NFIC) (31), and neurological and skeletal defects (NFIX) (32, 33). Thus, each NFI has nonredundant roles during development.

Although an unidentified NFI (34) and NFIX (17) have been identified as FOXA1 interacting partners, little is known about the role of individual family members in the prostate gland. Therefore, we set out to identify the specific NFI family members that interact with FOXA1 and determine the role of NFI family members on prostate-specific gene expression. Our studies have revealed that the NFI family provides an elaborate and well-balanced TF set that can provide AR, FOXA1, and the AR/FOXA1 complex precise control over tissue-specific gene expression.

Materials and Methods

Cell culture and establishment of LNCaP cells stably expressing FOXA1

All cell lines used were obtained from American Type Culture Collection and were maintained in RPMI 1640 medium (Invitrogen/Gibco), supplemented with 10% (v/v) fetal bovine serum (FBS) (Atlanta Biologicals). LNCaP cells express AR and FOXA1 and engage effectively in androgen-regulated prostate-specific gene expression. We therefore chose LNCaP cells to ectopically express dual affinity tagged FOXA1 for our binding partner studies. Tandem FLAG and 6xHis affinity tags were added to the N termini of FOXA1 via a standard PCR approach, followed by subsequent cloning into a pCR-TOPO 2.1 vector (Invitrogen) to generate FLAG-6xHis-FOXA1. After restriction enzyme digestion with XhoI and HindII (New England Biolabs), FLAG-6xHis-FOXA1 was cloned into the retroviral vector pLPCX (Clonetech), resulting in LNCaP-pLPCX-FOXA1. The sequence was confirmed by DNA sequencing. LNCaP cells were next infected with virus purified from Phoenix retroviral packaging cells transfected with either pLPCX-FOXA1 or pLPCX empty vector. Cells were selected and maintained in puromycin antibiotic (5 μg/mL).

Tandem affinity purification and mass spectrometry for the identification of FOXA1 binding partners

Nuclear extracts were prepared as described previously (35) from ∼10⁹ LNCaP-pLPCX-FOXA1 cells and LNCaP-pLPCX cells maintained in RPMI 1640 medium containing 10% FBS. Nuclear extracts were first subjected to TALON resin purification (BD Biosciences) according to the manufacturer's instructions to isolate FLAG-6xHis-FOXA1 via its His tags. This was followed by purification with anti-FLAG M2 affinity gel (Sigma-Aldrich) to purify FLAG-6xHis-FOXA1 by its FLAG tag, as well as FOXA1 binding partners. Fractions isolated from each step were subjected to Western blotting analysis for FOXA1 to verify purification success. In addition, Western blotting for the known FOXA1 binding partner, AR, was performed as a positive control after the use of anti-FLAG M2 affinity gel to verify the success of our purification approach. Purified protein was subjected to SDS-PAGE electrophoresis followed by band excision and tryptic digestion according to standard procedures. Peptide hydrolysate was then analyzed by C18 reverse-phase liquid chromatography-tandem mass spectrometry (LC-MS/MS) using a Thermo LTQ ion trap mass spectrometer equipped with a Thermo MicroAS autosampler and Thermo Surveyor HPLC pump system, nanospray source, and Xcalibur 2.0 instrument control using standard data-dependent methods. MS/MS data were analyzed with the SEQUEST algorithm against the International Protein Index (IPI) human database (135 674 entries, October 2007 release) including a concatenated reverse database for calculating the false-discovery rate. The presence of at least 2 peptides from 2 separate runs was used as the criteria for a positive hit.

Coimmunoprecipitation studies

JEG-3 cells are reported to be NFI deficient (36, 37) and do not express FOXA1, making them an ideal cell line to test the ability of these proteins to interact. After transfection with FOXA1 and hemagglutinin (HA)-tagged individual mouse NFI constructs or vector plasmid (PCH), JEG-3 cells were washed 3 times with cold PBS and lysed with 1 mL of nondenaturing lysis buffer (50 mM Tris, 150 mM NaCl, 10 mM EDTA, 0.02% NaN₃, 50 mM NaF, 1 mM Na₃VO₄, 1% NP-40, 1 mM phenylmethylsulfonylfluoride, 0.5 mM dithiothreitol, and 1× concentration of complete protease inhibitor cocktail [Roche]). After sonication, centrifugation, and preclearing, 1 mg of total cell lysate for each reaction was incubated at 4°C overnight with 20 μL (dry volume) of protein G-Sepharose beads (Amersham Biotech) conjugated with 1 μg of experimental antibody. Immunoprecipitation was performed in the presence of ethidium bromide (100 μg/mL; Sigma-Aldrich) to disrupt DNA-protein interactions, and BSA was added to reduce nonspecific binding. After overnight incubation, samples were centrifuged, and the pelleted protein G-Sepharose beads were washed 4 times with lysis buffer and once with PBS followed by protein dissociation and Western blotting analysis. The mouse monoclonal HA antibody (clone 12CA5; Roche) was used to immunoprecipitate HA-tagged NFI family members. Anti-HNF-3 (C-20; Santa Cruz Biotechnology) was used to immunoprecipitate FOXA1.

Glutathione S-transferase (GST)-fusion assays

Primer pairs containing EcoR1 and XhoI restriction enzyme sequences were used to amplify and subclone the NFIX coding sequence into the pGEX6p-1 vector, resulting in GST-NFIX for use in GST pull-down assays. T7 promoter-driven expression vectors encoding for FOXA1 deletion constructs were described previously (9) and were transcribed and translated in vitro using the TnT T7 Quick Coupled Transcription/Translation System (Promega). A standard reaction involved a 90-minute incubation at 30°C with 40 μL of TnT Quick Master Mix, 2 μL of cold 2 mM methionine, and 2 μg of plasmid DNA in a final volume of 50 μL. In vitro translated recombinant FOXA1 proteins were labeled with a C-terminal V5 epitope and were used immediately for in vitro binding reactions. For GST pull-down assays, 50 μL of swelled glutathione agarose beads (G-4510; Sigma-Aldrich) were incubated with 20 μg of GST or GST-NFIX fusion proteins for each reaction. GST-bound beads were equilibrated with PBS-T binding buffer (1× PBS [pH 7.4], 1% Tween 20, and protease inhibitors) and incubated for 2 hours at 4°C with 5 to 10 μL of products from the TnT reactions. Complexes were washed 4 times with 1.5 mL of cold binding buffer, heated for 10 minutes at 70°C in 1× SDS loading buffer, and separated by SDS-PAGE, after which V5-horseradish peroxidase antibody was used in a standard Western blot to determine which domain of FOXA1 is required for interactions with NFIX in vitro. Cell lysates and IP reactions were run in different orders on their respective gels. For clarity, input Western blot images have been reordered. A comparison of the original and reordered original can be found in http://press.endocrine.org/doi/suppl/10.1210/me.2013-12139/suppl_file/me-13-1213-1.pdfSupplemental Figure 1.

ARR₂PB and prostate-specific antigen (PSA) reporter gene assays

Transient transfection to determine the influence of NFI overexpression on androgen-regulated ARR₂PB and PSA reporter gene constructs was performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's recommendation and as reported previously (7). JEG-3 cells were plated at an initial density of 150 000 cells/well in 24-well tissue culture plates (BD Falcon) and allowed to attach overnight in RMPI medium 1640 supplemented with 10% FBS. On the following day, 0.25 μg of mouse NFIA, NFIB, NFIC, or NFIX plasmid, 0.25 μg of rat AR expression vector, and/or FOXA1 expression vector, 0.0125 μg of pRL-CMV (Promega) were mixed in serum-free Opti-Free MEM medium (Invitrogen) and subsequently combined with Opti-Free MEM containing Lipofectamine 2000 (Invitrogen) and incubated at room temperature per the manufacturer's instructions. Medium was aspirated from target cells for transfection, which were then incubated with the DNA-Lipofectamine mixture for 6 hours, followed by removal of the medium and addition of 0.5 mL/well of Opti-Free medium supplemented at a final concentration of 10 nM dihydrotestosterone (DHT) or ethanol vehicle control (Sigma). After DHT treatment for 24 hours, cells were harvested by removing the medium, washing the cells once with PBS, and incubating with 100 μL of passive lysis buffer (Promega) for 30 minutes at room temperature. Both firefly and Renilla luciferase activities were determined in a lumicounter (LUM/star; BMG LabTechnologies, Inc) by using the Dual-Luciferase reporter assay system (Promega) to control for transfection efficiency. Experiments were performed in triplicate and repeated at least twice.

Knockdown studies

LNCaP cells were transiently transfected with either ON-TARGETplus SMART pool human small interfering siRNA (siRNA) NFIA, NFIB, NFIC, or NFIX constructs or ON-TARGETplus Non-Targeting siRNA no. 2 (Dharmacon RNAi Technologies) as described previously (38) at a working concentration of 100 nM. After 48 hours of transfection in complete medium (10% FBS in RPMI medium 1640), quantitative real-time PCR (Q-RT-PCR) was used to determine the extent of individual NFI knockdown, and the influence of individual NFI knockdown on the prostate-specific genes PSA, TMPRSS2, FKBP5, and NKX3-1. Primer sequences and associated annealing temperatures used in this study are as follows: for NFIA, 55°C, forward 5′-CCTCTACGAGCTCCACAAAGC-3′ and reverse 5′-ATTGAGGAACCCCACCTGTCC-3′; for NFIB, 55°C, forward 5′-AGAGATCAAGATATGTCTTC-3′ and reverse 5′-CTGGCTGGTTTGTGGACTGGA-3′; for NFIC, 55°C, forward 5′-CCTGGACCGTTAAATGGA-3′ and reverse 5′-GATACCAGGACTGTGCCTG-3′; for NFIX, 58°C, forward 5′-CTGCCCAACGGGCACTTAA-3′ and reverseerse 5′-CTGTCATCGATGGACTTGGG-3′; for PSA, 58°C, forward 5′-GCAGTCTGCGGCGGTGTTCT-3′ and reverse 5′-GCGGGTGTGGGAAGCTGTGG-3′; for NKX3-1, 62°C, forward 5′-CCGAGACGCTGGCAGAGACC-3′ and reverse 5′-GCTTAGGGGTTTGGGGAAG-3′; for TMPRSS2, 58°C, forward 5′-GCACAGCCCACTGTGGTCCC-3′ and reverse 5′-CAGAGTAGGCCAGCGGCCAG-3′; for FKBP5, 60°C, forward 5′-CTGGAAGGCCGCTGTGGTGG-3′ and reverse 5′-TGCATAGGGACTCACACACCTTGA-3′; and for GAPDH, 58°C, forward 5′-GGCATGGACTGTGGTCATGAG-3′ and reverse 5′-TGCACCACCAACTGCTTAGC-3′.

Primers used to validate chromatin immunoprecipitation (ChIP)-sequencing (ChIP-Seq) data mining and ChIP studies are the following: for CLU, forward 5′ CCCACACTTCTGACTCGGAC-3′ and reverse 5′-ACTCCTCCCGGTGCTTTTTG-3′; for OR9A2, forward 5′-GTCTGCAGTCCCCCATGTAT-3′ and reverse 5′-CCATGGTCCCACAGGAAAAGT-3′; for GREB1, forward 5′-CGTGTGGTGACTGGAGTAGCTG-3′ and reverse 5′-TGGCATCTCAGATTCGGTGC-3′; for SOX6, forward 5′-CTGCGGAGAAGAATGTCTTCCAA-3′ and reverse 5′-TGCATTATGGGGTGCAGAGG-3′; for SMAD2, forward 5′-GCTCCCTCCGTCTTCCATAC-3′ and reverse 5′-CTTGTATCGAACCTCCCGGC-3′; for SYPL1, forward 5′-TGGCGCCCAACATCTACTTG-3′ and reverse 5′-AGAAGCAATCCACTCGAGGAC-3′; and for IL-8, forward 5′-CAGAGACAGCAGAGCACACA-3′ and reverse 5′-GGCAAAACTGCACCTTCACA-3′.

ChIP

LNCaP cells were used to determine the ability of NFI TFs to bind to AR target genes. ChIP assays were performed according to the manufacturer's instructions for the SimpleChIP Enzymatic Chromatin IP kit with magnetic beads (Cell Signaling Technology). Antibodies used for ChIP assays were as follows: ChIP-grade rabbit anti-AR (2 μg, 74272; Abcam), ChIP-grade rabbit anti-FOXA1 (2 μg, 23738; Abcam), rabbit anti-pan NF-1 N-20 (5 μg, sc-870; Santa Cruz Biotechnology), rabbit anti-NFIB (5 μg, HPA003956; Sigma-Aldrich), and normal rabbit IgG (5 μg, 2729; Cell Signaling Technology). LNCaP cells were treated with either ethanol (EtOH) control or 10 nM DHT for 2 hours, which represents the time required for peak AR recruitment to the PSA enhancer after DHT treatment (39). Cells were cross-linked for 9 minutes with formaldehyde and processed according to the manufacturer's instructions. ChIP DNA was analyzed by Q-RT-PCR. The primer set for ChIP analysis was hChIP PSA 206 (forward 5′-ACAGACCTACTCTGGAGGAA-3′ and reverse 5′-AAGACAGCAACACCTTTTTTTTTC-3′) used at an annealing temperature of 52°C. Results from PCR normalized to EtOH IgG are depicted. Primers used for ChIP validation are as follows: for TMPRSS2 ChIP, forward 5′-TGGTGTGTTAGGGATCTGGAG-3′ and reverse 5′-CACGCCCCGCTTTCTTTTTA-3′; for CLU ChIP, forward 5′-GCCTGGTTGTGCACTCATCTA-3′ and reverse 5′-TCCTGGTACACAGCAGTTCA-3′; for OR9A2 ChIP, forward 5′-CCCTAGCTGCTATGCTCCAA-3′ and reverse 5′-AGGTGGGAAGACTGAGTGGA-3′; for GREB1 ChIP, forward 5′-GTAGTCCTTCGGAGGCAAGC-3′ and reverse 5′-GTTTTGCTGGGTCACAGTGC-3′; for SOX6 ChIP, forward 5′-CAACATTACTGTGTCCCTGGC-3′ and reverse 5′-CTGTCTCCCTGAGTGGGTCT-3′; for SMAD2 ChIP, forward 5′-ACTGGAGTTCAGCGTGGAAG-3′ and reverse 5′-TGACTTTCCATCCAGTGGGAC-3′; and for IL-8 ChIP, forward 5′-TGCTCACCCAAATGGCAGAT-3′ and reverse 5′-ACATAGGAAAACGCTGTAGGTCA-3′.

Förster resonance energy transfer (FRET) construct development

The AR-Cerulean construct was created by amplifying the gene encoding AR with primers (forward 5′-AAAGCTAGCGCCACCATGGAAGTGCAGTTAGGGC-3′ and reverse 5′-AAAACCGGTCCACGCGTCTGGGTGTGGAAA-3′) and sequential digestion/ligation of the product and mCerulean3 (mCer3)-C1 vector using NheI and AgeI restriction enzymes. NFIX-Venus was created similarly with primers (forward 5′-AAAAGATCTATGTATAGCCCGTACTGCCTCACC-3′ and reverse 5′-AAAGGTACCTTCAGAAAGTTGCCGTCCC-3′) for the NFIX gene and an mVenus-C1 vector was used with NheI and AgeI restriction enzymes. Finally, the FOXA1-Venus construct was created with amplifying primers (forward 5′-AAAGCTAGCGCCACCATGTTAGGAACTGTGAAG-3′ and reverse 5′-AAAACCGGTCCGGAAGTGTTTAGGACGGG-3′) for FOXA1 and an mVenus-C1 vector using BglII and KpnI restriction enzymes. Sequences of the constructs were verified using the Vanderbilt Genome Sciences Sanger DNA sequencing laboratory.

Cellular sample preparation for FRET studies

HeLa cells were transiently transfected with plasmid DNA encoding mCer3-tagged AR, mVenus-tagged NFIX, and/or mVenus-tagged FOXA1; FRET8 (a dimer of enhanced cyan fluorescent protein and enhanced yellow fluorescent protein) or mCer3 and mVenus were used as controls. Transfection was accomplished using Lipofectamine 2000 (Invitrogen) transfection reagent according to the manufacturer's instructions. Cells were seeded onto no. 1 coverslip bottom dishes (MatTek) and cultured in DMEM (Invitrogen/Gibco) supplemented with 10% FBS (Invitrogen/Gibco), 100 U/mL penicillin (Mediatech), and 100 μg/mL streptomycin (Mediatech) at 37°C under 5% humidified CO₂. At 24 hours after transfection, samples were fixed with 4% paraformaldehyde, washed in Dulbecco's PBS (Invitrogen/Gibco), and mounted with Gelvatol. For samples containing the AR plasmid, 24 hours after transfection cells were incubated overnight with medium containing charcoal-stripped serum and then treated with DHT (Sigma-Aldrich) for 4 hours before fixing as stated above.

Fluorescence microscopy for FRET

FRET imaging was performed using a Zeiss LSM780 confocal microscope with excitation provided by a Coherent chameleon 2-photon laser at 800 nm with emission collected in spectral mode from 465 to 692 nm with 8.7-nm spectral resolution. In addition, images were collected with an argon laser at 514 nm to confirm cells expressing the mVenus-tagged constructs. Data were analyzed using ImageJ and GraphPad Prism software. FRET ratios were expressed as mVenus/mCer3 after linear unmixing and normalized to the vehicle-treated AR + FOXA1 control ratios. A total of 70 to 100 cells/dish were quantified for mean intensity, and the experiments were repeated in 3 cellular preparations. P values are the result of repeated-measures ANOVA compared with NFIX + AR.

Identification NFI binding sites in proximity to AR/FOXA1 binding sites

Previously published ChIP-Seq data (20, 40) was downloaded from the National Center for Biotechnology Information (NCBI) and converted to the FASTQ format using the Sequence Read Archive (SRA) toolkit (41), analyzed for quality using FastQC (42), aligned to the human genome using Bowtie (43), and interrogated using HOMER (44) for AR/FOXA1 adjacent peaks in LNCaP and VCaP cells. From this list, we searched for predicted NFI binding sites within 100 bp in either direction of the AR/FOXA1 region. Based on this list, several candidate genomic elements and their corresponding genes were arbitrarily selected and validated by ChIP and Q-RT-PCR. For programming specifics, please see the http://press.endocrine.org/doi/suppl/10.1210/me.2013-1213/suppl_file/me-13-1213-2.pdfSupplemental Methods.

Modeling of TF interactions

The model was constructed from the 3-dimensional structures of DNA-binding domains (DBDs) of 3 TFs bound to DNA. The atomic coordinates of the homodimeric DBD of AR were taken from crystal structure 1R4I (Protein Data Bank [PDB] ID; Ref. 45). The coordinates of the FOXA1 DBD are from the Nucleic Acid Database (NDB ID PDT013; Ref. 46), and the structure of the NFIX DBD was modeled on the dimeric Smad MH1 domain 1MHD (PDB ID; Ref. 47), using the 3D-Jury server at http://bioinfo.pl/ (48). These 3 DNA-bound protein structures were then aligned, using University of California, San Francisco (UCSF) Chimera (49), to a model of the probasin promoter region (−140 to −65) generated in idealized B-form by NAB (50) at the University of Southern California (USC) make-na server (51). Ovals representing the attached N- and C-terminal non-DBDs were added to illustrate potential interdomain contacts. The ovals are scaled roughly to the size of the non-DBDs. The locations of the ovals are consistent with their attachment points on the DBDs, but the orientations are speculative.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 6, and specific tests are identified in the figure legends.

Results

Tandem affinity purification and mass spectrometry identify a novel set of FOXA1-interacting proteins

In an effort to identify additional novel FOXA1 binding partners, we used a retroviral-based approach to establish LNCaP cells that stably overexpress dual affinity tagged FOXA1 (LNCaP-FOXA1) or empty vector (LNCaP-pLPCX). Although LNCaP cells express FOXA1, dual affinity tagged FOXA1 (FLAG-6xHis-FOXA1) was significantly overexpressed in these cells (Figure 1A), enabling the purification of FOXA1 and associated binding partners. To purify FLAG-6xHis-FOXA1 and associated binding partners, nuclear extracts from LNCaP-FOXA1 and LNCaP-pLPCX cells were subjected to purification via tandem TALON and FLAG purification. Western blotting analysis using antibodies specific for both FOXA1 and the His tag reveal significant enrichment for FOXA1 (Figure 1B) after purification. Importantly, Western blotting for AR revealed that this approach was capable of purifying AR, the known FOXA1 binding partner (Figure 1C). Total purified protein was subjected to SDS-PAGE and colloidal blue staining (Figure 1D), followed by tryptic digestion and LC-MS/MS analysis. Analysis of MS/MS data via the Sequest algorithm against the IPI human database identified a total of 16 proteins that appear to interact with FOXA1 (Table 1). Of the 16 proteins meeting the “hit” criteria, 6 have been reported previously to interact with AR by a direct or an indirect link (52–57). In addition to these known AR interacting proteins, we identified gene products that are generally important for the maintenance of protein stability, RNA processing, cell cycle control, and chromatin regulation. We chose to focus on NFIX and its family members because NFI family members are TFs, NFI motifs have been associated with AREs and NFI can bind to FOXA1, an AR coregulator (7, 17, 34, 58).

NFI family members are expressed in prostate cancer cell lines and physically interact with FOXA1

The NFI family of TF contains four genes (NFIA, NFIB, NFIC, and NFIX) encoding proteins that bind to the consensus DNA sequence TTGGCN₅GCCAA (26). The levels of NFI family member expression in commonly used cell lines was determined by Q-RT-PCR on LNCaP, PC3, and DU145 cells (Figure 2A) and on the JEG-3 choriocarcinoma cell line, which was previously reported to express low levels of NFI (36, 37). JEG-3 cells indeed expressed virtually undetectable levels of NFI family members. On the other hand, LNCaP, PC3, and DU145 cells expressed each NFI family member, with the highest levels of family member expression consisting of NFIA and NFIB (Figure 2A). Although NFIC and NFIX expression was detectible in these prostate cell lines, expression was relatively low compared with that for NFIA and NFIB.

Because JEG-3 choriocarcinoma cells express exceedingly low levels of NFI, we individually expressed HA-tagged NFIA, NFIB, NFIC, NFIX, or vector (PCH), along with FOXA1, in JEG-3 cells to identify the NFI family member(s) capable of interacting with FOXA1. Immunoprecipitation was performed using an anti-FOXA1 antibody, followed by Western blotting analysis with anti-HA monoclonal antibody. Immunoprecipitation results indicated that NFIA, NFIB, NFIC, and NFIX are capable of interacting with FOXA1 (Figure 2B, top panel). An unidentified protein that has previously reported to occur after ectopic expression (33), perhaps an NFIA degradation product, was also capable of interacting with FOXA1 in JEG-3 cells. This could also be an NFI family member (such as NFIC or NFIB, which are similar in size) that was purified with NFIA during the immunoprecipitation. Nonetheless, reciprocal immunoprecipitation analysis confirmed the ability of each NFI family member to interact with FOXA1 (Figure 2B, bottom panel).

To confirm the interaction between NFI family members and FOXA1 and also to identify the domain in FOXA1 required for mediating this interaction, GST-fusion assays using GST-tagged full-length NFIX, as well as an assortment of FOXA1 deletion constructs, were performed (Figure 2C). NFIX was selected because it was identified by LC-MS/MS analysis and coimmunoprecipitated strongly with FOXA1 (Figure 2B, bottom panel). Full-length FOXA1 indeed interacts with GST-tagged NFIX (Figure 2D), whereas deletion of the N terminus (ΔN) or more than half of the forkhead domain of FOXA1 (NT construct) abrogated this interaction (Figure 2D). Taken together, these data indicate that FOXA1 interacts with all members of the NFI family, and the interaction of NFIX and by extension other NFI family members with FOXA1 is mediated by the N-terminal transactivation and forkhead domains of FOXA1.

Overexpression of NFI family members represses androgenic induction of the probasin promoter and PSA enhancer

An NFI binding site has been identified within the probasin promoter (59), and a pan-NFI antibody ChIP analysis demonstrated that an NFI family member binds to the probasin promoter adjacent to the FOXA1 binding site (7). In addition, NFI binding sites exist in close apposition to AR and FOXA1 binding sites within the PSA enhancer and appear to be common in most AR-regulated genes (58). To determine the influence of individual NFI family members on the transcriptional activity of these prostate-specific regulatory sequences, we performed reporter gene assays using previously described probasin (ARR₂PB) and PSA luciferase reporters (7). Individual reporters were cotransfected with the AR as well as individual NFI family members either in the presence or absence of FOXA1 in JEG-3 cells (Figure 3). In the absence of FOXA1, individual NFI family members repressed androgen induction of the PSA (A) and probasin (C) reporter, and the ability of NFI family members to repress androgen-induced probasin reporter activity was maintained in the presence of FOXA1 (Figure 3, B and D). These results suggest that NFI family members can confer additional regulation to AREs.

NFI knockdown affects prostate-specific gene expression

Previous studies have examined the consequences of a pan-siRNA against NFI (siNFI) construct on AR target genes (PSA, KLK2, TGM2, TMPRSS2, and FKBP5) and observed gene-dependent changes (58), suggesting that NFI regulation of AR target genes may be NFI family member specific. To identify the role of individual NFI family members in prostate-specific gene expression, individual NFI family members were knocked down in LNCaP cells, and the cell line was examined by Q-RT-PCR for the known AR/FOXA1 target genes PSA, TMPRSS2, NKX3-1, and FKBP5 (Figure 4). Knockdown of individual NFI family members was specific and efficient (Figure 4A). Interestingly, NFIX knockdown resulted in up-regulation of NFIA and NFIB relative to that of nontargeting siRNA (Figure 4A, NFIA and NFIB). These observations suggest that in prostatic cells, NFI family members are capable of compensating for each other and may regulate each other as well. Adding complexity to the situation, NFI family members can be induced, repressed, or unregulated by DHT treatment (Supplemental Figure 2).

To simplify the analysis, we focused on examining the influence of individually knocking down NFI family members on AR target gene expression in LNCaP cells maintained in complete medium (Figure 4B). Knockdown of NFIA resulted in decreased TMPRSS2 and NKX3-1 expression, and increased FKBP5 expression. NFIB knockdown resulted in increased expression of NKX3-1 and FKBP5. NFIC knockdown resulted in decreased PSA and NKX3-1 expression. NFIX knockdown resulted in increased TMPRSS2 expression. In summary, although NFI expression levels affect the expression of androgen-regulated genes, this regulation is complex, and individual NFI family members can be redundant to each other.

NFI proteins bind to the PSA enhancer

To determine the ability of NFI proteins to bind to cis-regulatory regions of the human PSA enhancer and to determine whether NFI binding to this region is itself regulated by androgen, ChIP reactions were performed in LNCaP cells in the presence or absence of DHT. Because of the lack of specific ChIP-grade or ChIP-validated antibodies for NFIA, NFIC, and NFIX, only a pan-NFI and NFIB antibody were used. In addition, ChIP was performed using AR and FOXA1 antibodies. The Tess TF search program (http://www.cbil.upenn.edu/cgi-bin/tess/tess) revealed consensus NFI binding sites within AREIII, just downstream of AREIIIA, and adjacent to a FOXA1 binding site (9) within the PSA core enhancer (Figure 5A). Target sequence primers were designed to amplify a region within the PSA enhancer, encompassing AREIIIA and AREIIIB, as well as the 3 consensus NFI binding sites (9, 60). As expected, DHT treatment resulted in robust binding of AR to the PSA enhancer (Figure 5B). Antibodies for FOXA1, pan-NFI, and NFIB revealed that although AR is recruited to the PSA enhancer, FOXA1 and NFIB are already occupying these regions (Figure 5B). This finding is consistent with other reports suggesting that pan-NFI occupancy of target gene promoters and enhancers is less responsive to DHT (58).

NFIX interacts with AR in the presence of FOXA1

FRET, which is facilitated by tagging the proteins of interest with fluorescent proteins as reporters, is a phenomenon useful for studying protein-protein interactions (61). Because FRET occurs only when the fluorescent proteins are within 10 nm, it is commonly used to indicate the proximity of proteins of interest. Here, we tagged FOXA1, NFIX, and AR with either a donor (mCer3) or acceptor (mVenus) fluorescent protein and assayed the relative FRET ratios in HeLa cells expressing the constructs. The NFIX and FOXA1 constructs localized to the nucleus constitutively, whereas the AR translocated to the nucleus only after treatment with DHT (Figure 6, A and B).

Consistent with previous reports of AR and FOXA1 interactions (9, 10), cotransfected AR-mCer3 and FOXA1-mVenus exhibited a high FRET ratio of 0.93 ± 0.052 in cells treated with DHT (Figure 6C), compared with both positive (mCer3-mVenus dimer) and negative (cotransfection of single mCer3- and mVenus-expressing plasmids) controls (quantified in Figure 6G). To determine whether NFIX and FOXA1 are close enough to interact directly, cells were cotransfected with a NFIX-mCer3 donor and FOXA1-mVenus acceptor construct. This pair exhibited nuclear localization and a FRET ratio of 0.59 ± 0.026 (Figure 6, D and G). Importantly, when we performed this experiment with FOXA1-mCer3 as the donor and NFIX-mVenus as the acceptor, the FRET ratio was comparable to that of the negative control (data not shown). This result demonstrates that the folding and orientation of the proteins with fluorescent protein tags is critical for FRET to occur. Because of the constitutive localization of the NFIX-mCer3 donor and FOXA1-mVenus acceptor to the nucleus, a FRET signal of comparable strength can also be detected in HeLa cells cultured in charcoal-stripped serum medium without DHT treatment (Supplemental Figure 3).

The AR-mCer3 and NFIX-mVenus FRET ratio of 0.23 ± 0.021 was significantly lower (Figure 6, E and G), suggesting that these proteins are further away from each other than the AR and FOXA1. To test the hypothesis that overexpressing FOXA1 could bridge the interaction between NFIX and AR, we cotransfected cells with AR-mCer3, NFIX-mVenus, and untagged FOXA1. After treatment with DHT, the FRET ratio of AR-mCer3 and NFIX-mVenus was increased to 0.53 ± 0.058, a level intermediate between those for the FOXA1/AR and NFIX/AR samples (Figure 6, F and G). This result suggests that FOXA1 can either serve as a bridge to bring NFIX and AR together or alter the relative orientation of NFIX and AR, which may result in AR-mediated gene transcription.

NFI motif discovery and validation at AR and FOXA1 binding sites

ChIP-Seq has become a routine technique to investigate the spatial and temporal patterns of DNA occupancy by TFs and provides insight into the mechanisms of coregulation by TF complexes. Because the LNCaP cell line is one of the few AR-dependent prostate cancer cell lines, ChIP-Seq for AR and FOXA1 (Supplemental Figure 4) has been performed by several groups (20, 25, 40, 62). We took advantage of high-quality AR and FOXA1 ChIP-Seq data (20, 40) to explore AR/FOXA1 and NFI complex formation. From the ChIP-Seq data, we identified the genome-wide set of loci bound by AR and FOXA1 in LNCaP cells (Supplement 1). For these sets, we applied a computational algorithm, HOMER, to identify consensus NFI full-length and half-site motifs (Figure 7A) within 100 nucleotides of AR and FOXA1 binding sites. HOMER (44) uses a more stringent consensus sequence than the Tess TF search program (http://www.cbil.upenn.edu/cgi-bin/tess/tess). Our in silico analysis revealed that in the 5125 loci bound by AR and FOXA1, there are 1764 (34.4%) AR/FOXA1 peaks with at least 1 NFI motif. This finding suggests that about one third of the AR/FOXA1 target genes are potentially coregulated by NFI (Figure 7B).

The presence of NFI motifs at one third of AR/FOXA1 peaks raised the possibility that NFIs may also regulate FOXA1 (FOXA1-only) and AR (AR-only) binding separately. Further analysis of the data revealed that in the 3898 loci bound by AR only, there are 1495 (38.4%) AR-only peaks with an associated NFI motif (Figure 7B). In the 17 808 loci bound by FOXA1-only, there are 5848 (32.8%) FOXA1-only peaks with an NFI motif (Figure 7B). These observations suggest that roughly one third of AR alone-, FOXA1 alone-, and AR/FOXA1-occupied peaks have the potential to be regulated by NFI family members in LNCaP cells.

To obtain a more comprehensive picture of NFI binding beyond LNCaP cells, we also analyzed AR and FOXA1 ChIP-Seq data from VCaP cells (Supplement 1). At the 2016 loci bound by AR and FOXA1 in VCaP, we identified 372 (18.5%) AR/FOXA1 peaks with at least one NFI motif (Figure 7C) within 100 nucleotides. In the 2531 loci bound by AR only, there are 841 (33.2%) AR-only peaks with at least 1 NFI motif (Figure 7C). In the 11 905 loci bound by FOXA1 only, there are 3490 (29.3%) FOXA1-only peaks with at least 1 NFI motif (Figure 7C). The decrease in AR/FOXA1/NFI binding sites in VCaP cells is particularly interesting and requires further investigation. Nevertheless, in silico analysis of AR and FOXA1 binding sites in both LNCaP and VCaP cells reveals that NFIs may globally regulate AR-mediated gene expression.

AR/FOXA1/NFI co-occupied regions predict NFI binding and gene expression modulation

To validate our in silico predictions, 8 genomic loci/elements identified by in silico analysis were validated by ChIP and Q-RT-PCR. Validation targets were selected based on the proximity to genes of particular interest to prostate cancer, peaks conserved between LNCaP and VCaP cells, and peaks with high peak scores. Because of the limitations of specific ChIP-validated NFI antibodies, we again limited our analysis to NFIB and the pan-NFI antibody. ChIP for NFIB and pan-NFI followed by Q-RT-PCR to genomic loci with proximity to TMPRSS2, SYPL, CLU, and SMAD2 showed enrichment for pan-NFI, NFIB, or both at these loci compared with that for IgG (Figure 8Ai; AR and FOXA1 not shown). Q-RT-PCR analysis of these genes in response to individual NFI knockdown demonstrated NFI-specific and gene-specific changes (Figure 8Aii).

Although OR9A2 and SOX6 elements showed enrichment by ChIP (Figure 8Bi), their expression was unchanged by NFI knockdown (Figure 8Bii). Finally, IL-8 and GREB1 did not achieve significant enrichment or changes in gene expression (Figure 8C). Of our 8 candidate genomic elements/genes, 6 enriched at their predicted genomic element, and 4 had changes in gene expression in response to individual NFI knockdown. The 4 genes (IL-8, OR9A2, GREB1, and SOX6) that were unaffected by knockdown may either be unresponsive to NFI family members or they are not limited to specific homodimers or heterodimers in the presence of DHT. Regardless, validation of 6 of 8 binding sites by ChIP and gene changes in 4 of 8 of these proteins suggest that the NFI family members are broadly responsible for modulating AR/FOXA1 transcriptional activity.

Modeling the AR/FOXA1/NFI complex

Based on the significant effects of NFI knockdown on AR target gene expression, as well as the prediction that NFI may regulate approximately one third of the AR/FOXA1 target genes, a molecular model of this complex was developed using the probasin promoter. The homodimeric DBDs of AR (crystal structure 1R4I [PDB ID; Ref. 45]), FOXA1 (PDT013 [Nucleic Acid Database ID; Ref. 46]), and NFIX (modeled on the dimeric Smad MH1 domain 1MHD [PDB ID; Ref. 47]), were aligned to a model of the probasin promoter region ([140 to −65 base pairs). To illustrate potential interdomain contacts, ovals representing the attached N- and C-terminal non-DBDs were added and scaled roughly to an appropriate size. Although the locations of the ovals are consistent with their attachment points on the DBDs, their orientations are speculative. Nevertheless, the molecular model suggests that AR and FOXA1 bind directly adjacent to each other, with NFIX binding within 10 nm of the AR/FOXA1 complex (Figure 9). This model is consistent with the FRET results, which validated the AR/FOXA1 interaction, and suggests that FOXA1 can bridge the AR-NFIX interaction. Therefore, this model may be representative of the complex present on many if not most AR/FOXA1/NFI-regulated gene promoters.

Discussion

Several AR coregulators that both control and fine-tune AR activity, including FOXA1, which interacts with AR to control multiple prostate-specific genes in the epithelium, have been identified (9, 18). Recent reports have shown that AR and FOXA1 sites are commonly associated with androgen-regulated genes in prostate cancer cell lines (58, 63). In addition to AR, FOXA1 serves as a coregulator for various steroid receptors, including the glucocorticoid receptor (GR) (64) and estrogen receptor (ER) (65), suggesting that FOXA1 is a central mediator of endocrine responsive gene expression. To this end, multiple ER-responsive genes have been shown to contain FOXA1 binding sites (24) in paradigm similar to that for AR and FOXA1 sites, whereas GR binding sites are enriched for forkhead motifs (66). Unsurprisingly, both ER and GR have been reported to interact with FOXA1 (23, 24).

Much like FOXA proteins, NFI proteins are well known for their ability to regulate the activity of endocrine responsive cis-elements, such as the murine mammary tumor virus and the phosphoenolpyruvate carboxykinase promoters (67, 68). Furthermore, NFI activity is required to regulate expression of mammary gland–specific differentiation markers, including whey acidic protein and α-lactalbumin, among others (26). The frequent occurrence of NFI sites adjacent to AR and FOXA1 sites in multiple genes (58, 69) suggests that the NFI family is a central TF for androgen-regulated genes. Because NFI proteins bind to DNA as homodimers or heterodimers, the complexity for targeting specific genes regulated by the AR/FOXA1 complex is greatly increased. Further, preferential distribution of NFIX to prostatic stroma and NFIB to the basal cells (28), of which a subset contain AR and all lack FOXA1, suggests unique mechanisms to regulate AR action in different prostatic cell types. The central role of each NFI family member in organ development (29–33), as well as the fact that FOXA1 expression is restricted to specific organs, suggests a fundamental role for NFI and FOXA1 in combinatorial control (5, 6), resulting in tissue-specific gene expression.

These studies have demonstrated that all 4 family members are capable of interacting with FOXA1 via transient transfection of His-tagged NFI family members, which complements other studies that have shown an unidentified NFI protein (34) or that NFIX (17) can interact with FOXA1. The interaction between FOXA1 and NFIs modulates expression of AR target genes. When we determined the influence of knocking down individual NFI family members on the expression of PSA, TMPRSS2, FKBP5, and NKX3-1, we observed that NFIA and NFIC largely promote AR target gene expression, whereas NFIB and NFIX are mainly repressive. Subsequent ChIP analysis for the PSA enhancer, as well as 8 arbitrarily selected genomic loci identified by ChIP-Seq data mining, demonstrated that NFIs bind areas adjacent to the AR/FOXA1 peaks and can modulate the expression of genes associated with these loci. Our in silico analysis predicts that 34.4% of AR/FOXA1 target peaks have the potential to be regulated by NFIs. The NFIs, therefore, are potent modulators of AR-mediated gene expression in keeping with the observation they physically interact with FOXA1, which suggests that NFI family members play a critical, yet complex, role in prostate development and disease.

The frequent association of NFI binding sites with AR/FOXA1 binding sites (58, 69) suggests that this TF complex modeled on the probasin promoter (Figure 9) is not unique but represents a general model in which AR/FOXA1/NFI plays a fundamental role to regulate AR action. FRET was used to test the prediction that AR, FOXA1, and NFIX are in close proximity, resulting in the formation of a stable TF complex. The high FRET ratios shown by AR and FOXA1 were dependent on pretreatment of the cells with DHT to drive AR into the nucleus. Meanwhile, FOXA1 and NFIX FRET was independent of DHT treatment, further supporting the FOXA1 and NFI interaction. Although AR and NFIX did not exhibit a significant FRET ratio compared with that of negative controls, the additional transfection of an untagged FOXA1 significantly increased this ratio, suggesting that FOXA1 can bridge the AR-NFI interaction.

These observations are particularly interesting in the context of our data mining observations, which suggest that roughly one third of FOXA1-alone sites (ie, no AR) are associated with NFI sites, as are one third of AR-alone sites (ie, no FOXA1). For AR, this suggests that up 36% of AR binding sites in LNCaP cells may be regulated by NFIs, which may indicate that in the absence of FOXA1, such as in the stroma, an alternative AR cofactor bridges the NFI-AR interaction. Because FOXA1 is not expressed in the stroma (18), whereas NFIX expression is high in the stroma (28), androgen-regulated genes in the stroma may use NFIX and stromal forkhead genes, such as FOXF1, FOXF2, or FOXC1 (70), to mediate AR gene transcription. Indeed, AR interaction with FOXC1 has been recently reported (71). This alternate forkhead mechanism may also be true for the portion of the AR-NFI peaks that occur independently of FOXA1 in LNCaP cells.

Conversely, NFIs and FOXA1 can interact independently of AR, which suggests that FOXA1/NFI may play a role independent of AR, such as that in androgen-independent neuroendocrine (small cell) or castrate-resistant prostate cancer. Neuroendocrine prostate cancer is a rare and highly aggressive form of prostate cancer, which does not express AR (72), and is therefore considered to be androgen independent. NFIB has been identified as oncogenic in small cell lung cancer (73), and a mouse model of neuroendocrine prostate cancer (synaptophysin positive and androgen independent) has chromosomal gains at locus 4qC3, which contains NFIB, and results in increased expression of NFIB in the tumors (74). Whereas the status of FOXA1 in human neuroendocrine prostate cancer has not yet been reported, FOXA2 is expressed in human neuroendocrine prostate cancer, and mouse models of neuroendocrine prostate cancer have strong FOXA1 and FOXA2 expression (18). It is possible, therefore, that in the absence of AR, NFIB and FOXA1 interact to mediate gene transcription unique to neuroendocrine prostate cancer.

Unlike neuroendocrine prostate cancers that do not express AR, castrate-resistant prostate cancers continue to express AR and depend upon AR signaling, as demonstrated by their response to second-generation antiandrogens (75–77). Because FOXA1 expression increases as prostate cancer progresses (14), it is likely that the AR/FOXA/NFI complex will continue to mediate gene transcription. However, work by Sharma et al (69) has suggested that in castrate-resistant disease, the AR interaction shifts from FOXA1 and NFI to other cofactors such as STAT, MYC, E2F, AP-1, GATA, and NFKB. If AR is indeed being bound by other TFs or if antiandrogen therapy is interfering with AR/FOXA1/NFI complex formation, this would leave FOXA1 and NFI open to other interactions, perhaps with GR. Recent studies have indicated that FOXA1 can interact with GR (23). GR is particularly interesting because it can bind AREs and promote the expression of some AR target genes, and GR increases in castrate-resistant prostate cancer, supporting the failure of antiandrogens (78). In the presence of antiandrogens, therefore, the interaction of GR, FOXA1, and NFI may be uniquely important in driving castrate-resistant prostate cancer. Subsequent studies focusing on the NFI/FOXA1 complex in androgen-independent neuroendocrine prostate cancer, characterizing the AR/FOXA1/NFI complex in castrate-resistant prostate cancer, and identifying new coregulators that can bridge the AR-NFI divide will be particularly interesting. In summary, NFI family members interact with FOXA1, bind to AREs in an AR/FOXA1/NFI complex, and modulate AR-mediated gene transcription, implicating them as potent regulators of androgen-responsive, prostate-specific gene expression.

Acknowledgments

We thank Tom Case for his technical support, and Darlene Hancock for assistance with manuscript preparation.

R.J.M. was supported by the National Institutes of Health (Grant 2R01-DK055748-14) and the Joe C. Davis Foundation. R.M.G. was supported by the National Institutes of Health (Grant R01-HL080624-5) and NYSTEM (Grants C026429 and C026714). M.M.G. was supported by Vanderbilt University Medical Center (Integrated Biological Systems Training in Oncology Training Grant T32 CA119925). D.J.D. was supported by Vanderbilt University Medical Center (Multidisciplinary Training Grant in Molecular Endocrinology T32 DK007563-21 and Integrated Biological Systems Training in Oncology Training Grant T32 CA119925) and by a American Cancer Society Great Lakes Division-Michigan Cancer Research Fund Postdoctoral Fellowship. G.V.R. was supported by Department of Defense Institute of Defense Analyses (Grant W81XWH-12-1-0288).

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• AR

  androgen receptor

 
• ARE

  androgen response element

 
• ChIP

  chromatin immunoprecipitation

 
• ChIP-Seq

  chromatin immunoprecipitation-sequencing

 
• DHT

  dihydrotestosterone

 
• DBD

  DNA-binding domain

 
• ER

  estrogen receptor

 
• EtOH

  ethanol

 
• FBS

  fetal bovine serum

 
• FOX

  forkhead box

 
• FRET

  Förster resonance energy transfer

 
• GR

  glucocorticoid receptor

 
• GST

  glutathione S-transferase

 
• HA

  hemagglutinin

 
• LC-MS/MS

  liquid chromatography-tandem mass spectrometry

 
• mCer3

  mCerulean3

 
• NFI

  nuclear factor I

 
• PDB

  Protein Data Bank

 
• PSA

  prostate-specific antigen

 
• Q-RT-PCR

  quantitative real-time PCR

 
• siRNA

  small interfering RNA

 
• TF

  transcription factor.

References