Transcriptional Regulation and Expression of the Dominant Drusen Gene FBLN3 (EFEMP1) in Mammalian Retina

Abstract

purpose. To determine the important transcriptional control elements and sites of expression of fibulin-3 in mammalian retina.

methods. Sequencing and 5′ rapid amplification of cDNA ends (RACE) were undertaken to characterize the genomic sequence upstream of the FBLN3 coding sequence. Reporter deletion–mutation constructs were used in luciferase transfection assays to determine the important regulatory motifs. Fibulin-3 expression in mouse and human retina was studied by in situ hybridization and RT-PCR. The effect of 17β-estradiol on fibulin-3 production was studied in COS-7 and ARPE-19 cells.

results. Two untranslated exons were fully sequenced completing the characterization of FBLN3 that comprises 12 exons. Reporter assays suggest that the FBLN3 proximal promoter is contained within 425 bp upstream of exon 1. Important regulatory elements include three Sp1-binding sites, a Tant motif (trans-activating) and an estrogen response element (ERE) binding site (trans-repressing). No TATA or CAAT regulatory boxes were identified. RT-PCR suggests that the fibulin-3 gene is expressed in murine and human RPE, and in situ studies confirm that Fbln3 is expressed in the outer and inner nuclear layers, but strikingly not in the ganglion cell layer. Fibulin-3 expression in ARPE-19 cells could be modified by varying the amount of estrogen in the cell culture medium.

conclusions. The 5′ end of the FBLN3 gene has been characterized, and the important upstream motifs regulating its transcription have been identified. Fibulin-3 is expressed in adult retina and at early stages in human and mouse development. Estrogen may be an important regulator of fibulin-3 expression, and this highlights a novel mechanism by which circulating estrogen may control the composition of the retinal extracellular matrix.

The FBLN3 gene (also called EFEMP1 or S1-5) encodes fibulin-3, a very recent addition to the list of fibulin proteins. Fibulins are a family of extracellular matrix proteins that have characteristic calcium-binding epidermal growth factor (EGF) repeat domains, a homologous C-terminal globular domain, and divergent N-terminal domains (Fig. 1) . ¹ During embryogenesis, they are expressed in basement membranes and at sites undergoing epithelial–mesenchymal transformation. In mature tissue they are found in the extracellular matrix where they serve a structural role and are involved in cell signaling. Fibulins as a group are involved in many important biological mechanisms, including organogenesis, hemostasis, and tissue homeostasis. Through interaction with cell surface receptors, fibulins have a key role in cell growth, adhesion, and motility. In human disease, fibulins have an established role in cancer and cardiovascular disease. ² In mature tissue, fibulin-1 isoforms are found in the extracellular matrix of most tissues and are involved in hemostasis, growth factor signaling and ovarian cancer. ³ Fibulin-2 is also a basement membrane component and is particularly important in the integrity of the cardiovascular system. Fibulin-4 (also called MBP1, H411, or EFEMP2) is widely expressed and has recently been associated with colon cancer. ² Fibulin-5 (also called EVEC or DANCE) is important in establishing the elastic properties of distensible tissues. It is upregulated in atherosclerosis, ⁴ and FBLN5 mutation has been causally associated with autosomal recessive cutis laxa. ⁵ Very little is yet known about fibulin-6, also called hemicentin ⁶ (FIBL-6 or him-4) which may have a role in dominantly inherited age-related macular degeneration (Schultz DW, et al. IOVS 2003;44:ARVO E-Abstract 2017). 

Much less is known about fibulin-3. It is highly expressed in the brain, placenta, heart, medium-sized vessels, and eye. ¹ Fibulin-3 has approximately 60% identity at the protein level with fibulin-4. ¹ Recently, a R345W missense mutation in FBLN3 has been reported. ^{7 8 9 10 11} This mutation occurs in the fifth EGF-repeat domain of fibulin-3 and is causally linked to the human disease–dominant drusen (also called Doyne honeycomb retinal dystrophy or Malattia Leventinese). The molecular mechanism by which FBLN3 R345W-mutation leads to disease is unknown. Recent studies in a rat-derived retinal pigment epithelium cell line (RPE-J) show that this mutation causes the protein to misfold and accumulate within RPE cells and Bruch’s membrane. ¹² In addition, immunohistochemical studies have shown that in normal human retina, fibulin-3 is predominantly expressed in the nerve fiber layer and photoreceptors. However, in diseased retina, fibulin-3 accumulates beneath the RPE overlaying drusen, a pattern of expression that is absent in normal retina. ¹² Potentially therefore, fibulin-3 is an important constituent of the retinal extracellular matrix and is implicated in retinal disease. To gain better understanding of the function of this gene, we proposed to characterize the proximal promoter controlling its expression in human cells. 

Methods

Tissues Used for Studies

Human fetal tissue was obtained from the MRC Tissue Bank, Hammersmith Hospital (London, UK), with ethical approval and in accordance with the World Medical Association Declaration of Helsinki (Ethical Principles for Medical Research Involving Human Subjects). Fetal embryos were staged by last menstrual period and crown–rump length. Specimens were transferred less than 4 hours after surgery to liquid N₂ for RNA extraction. ¹³ Mouse embryos were obtained from matings of C57BL/6 x CBA mice. The day on which the vaginal plug was detected was designated as embryonic day (E)0.5. Human adult retina was obtained from surgical specimens with informed consent, and mouse eyes were obtained from adult C57BL/6 mice. Animal studies were conducted in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

RNA Isolation and RT-PCR in Human and Mouse Tissue and Cells

Total RNA was extracted from human eyes with a kit (RNAzol; Biogenesis, Poole, UK), whereas a second extraction process (TRIzol; Invitrogen, San Diego, CA) was used to isolate total RNA from mouse eyes and ARPE-19 cells according to manufacturers’ instructions. Total RNA was treated with RNase-free DNaseI (Invitrogen) before cDNA synthesis, using a preamplification system (SuperScript; Invitrogen). Oligo(dT) primer was used to generate first-strand cDNA for PCR from 1 μg of total RNA. Identical aliquots of cDNA were used for amplification using gene-specific primers. Human FBLN3 exon 11 forward primer (5′-TGCCATCAGACATCTTCCAG-3′) and exon 12 reverse primer (5′-TGCCTGTGGTTGACTCTTAGAA-3′) were designed to span intron 11 and amplify a product of 292 bp. In mouse, the human exon 11 forward primer was used with mouse exon 12 reverse primer (5′-AACACAGAGCTTGTGCGGAA-3′), generating a product of 229 bp. The PCR profile incorporated an initial denaturing step at 94°C for 3 minutes followed by 30 cycles of denaturation at 94°C for 45 seconds, annealing at 55°C for 45 seconds, and extension at 72°C for 45 seconds, with a final extension at 72°C for 5 minutes. Time of expression was tested on three independent tissue samples at each time point. 

For multiplex RT-PCR of FBLN3 and HPRT the following primers were used: FBLN3-forward (5′-CTACAAATACATGAGCATCCG-3′) and FBLN3-reverse (5′-CACAGAGCTTGTGCGGAAGG-3′) spanning intron 11 and amplifying a 261-bp product. Primers spanning intron 1 of the HPRT gene generated a 354-bp product; HPRT-forward (5′-CCTGCTGGATTACATTAAAGCACT-3′) and HPRT-reverse (5′-GTCAAGCGCATATCCAACAACAAA-3′). 

5′ Rapid Amplification of cDNA Ends

To extend the 5′ end of the human FBLN3 transcript 5′ RACE was performed on human retinal mRNA. Gene-specific primers were used in conjunction with a 5′ RACE system (Invitrogen-Life Technologies, Gaithersburg, MD): Ex6R, 5′-GCACTCGTCTATGTCTTGGC-3′ (exon 6); Ex5R, 5′-TGCACACTGGATACGGTGG-3′ (exon 5); and 5AR, 5′-ATGCTGCTGGCAGCTACAACC-3′ (exon 5). PCR products were cloned into pGEM-T vector (Promega, Southampton, UK), and several clones were sequenced. Nucleotide sequence determination was performed on a DNA sequencer (Prism 377; Applied Biosystems, Foster City, CA) using a cycle sequencing kit (BigDye Terminator; Applied Biosystems) according to the manufacturer’s instructions. 

Isolation of Genomic Clones Containing the Upstream Region of the FBLN3 Gene

IMAGE ¹⁴ clone 361052 (GenBank AA017386; http://www.ncbi.nlm.nih.gov/Genbank; provided in the public domain by the National Center for Biotechnology Information, Bethesda, MD), a retinal cDNA clone containing exons 3 to 12 of the FBLN3 gene was obtained from the U.K. Human Genome Mapping Project (HGMP) Resource Centre (Cambridge, UK). The insert of IMAGE clone 361052 was then used to probe the RPCI1 PAC library ¹⁵ and the LLNL chromosome 2 library ¹⁶ obtained on filter grids (UK HGMP Resource Centre). Two PAC clones were isolated, and DNA from the positive clones was restriction digested (PstI and EcoRI) and subcloned into pUC18 cloning vector. FBLN3 exon 1 PCR product was labeled with [³²P]-α-dCTP (Amersham Pharmacia, Amersham, UK) using an oligo-labeling kit (Amersham Pharmacia) and hybridized against colony lift filters in sodium phosphate buffer and SDS overnight at 65°C. The filters were washed in SDS and SSC at 65°C before exposure to x-ray film. Positive subclones containing the 5′ flanking region were sequenced. DNA sequences were analyzed using the BLAST program available through the National Center for Biotechnology (NCBI, Bethesda, MD; http://www.ncbi.nlm.nih.gov/gci-bin/BLAST/). ¹⁷ The Omiga1.1.3 program (Oxford Molecular, Oxford, UK) was used to predict transcription factor sites in the upstream sequence. The PROSCAN program ¹⁸ (available at http://bimas.dcrt.nih.gov/molbio/proscan/ provided in the public domain by the BioInformatics Molecular Analysis Section, Center for Information Technology, National Institutes of Health, Bethesda, MD) was also used to analyze the putative promoter sequence. 

Cell Culture

COS-7 cells were cultured in Dulbecco’s modified Eagle’s Medium (DMEM). ARPE-19 cells (ATCC, Manassas, VA) were cultured in DMEM/Ham’s F12. For both cell lines the culture medium was supplemented with 10% fetal bovine serum (FBS), 200 mM l-glutamine, and 100 U of penicillin/streptomycin (all Invitrogen) in 5% CO₂. 

To determine whether withdrawal of culture medium constituents (such as estrogen) affected FBLN3 expression, ARPE-19 cells were incubated in the absence of FBS for up to 24 hours. In separate experiments, ARPE-19 cells were then incubated with various concentrations of 17β-estradiol (Sigma-Aldrich, Poole, UK), in the presence of serum, for up to 24 hours. Fibulin-3 protein levels were then determined at regular intervals. 

In addition, we looked at how this effect of 17β-estradiol on fibulin-3 expression was influenced by tamoxifen (Sigma-Aldrich), a drug that competitively interferes with estrogen binding to cytoplasmic receptors. RT-PCR and Western blot experiments were undertaken using 100 pM 17β-estradiol and 1 μM tamoxifen. In each individual experiment, cells were incubated for 4-hour periods. 

Reporter Vector Construction and Luciferase Transfection Assays

An 843-bp fragment of the upstream region of FBLN3 was directionally cloned into the luciferase reporter assay plasmid pGL3 (Promega). This product, proposed to contain the transcriptional regulatory elements of the FBLN3 gene, was designated c-808. This construct was made with forward primer c-808F (5′-ATGGCTGGTACCAGAAATGAGATTGCTG-3′) and reverse primer c-808R (5′-AGCCATGCTAGCGAGGGGAGTGCGCAGG-3′). Shorter deletion constructs were then obtained from c-808 by the PCR method, followed by cloning into pGL3. These other constructs were generated by using the following primers: c-425F (5′-ATGGCTGGTACCTAGCATTCACTTATTG-3′); c-243F(5′-ATGGCTGGTACCGCGAGTCTGGGAAACG-3′); and c-182F (5′-ATGGCTGGTACCTGGAGCAGGGGGCGCG-3′). Each was used in turn with reverse primer c-808R. These pGL3 plasmid constructs were purified with a standard protocol (Midiprep; Qiagen, Crawley, UK). 

Standard liposome-mediated transfections were performed in COS-7 cells with 6 μL transfection reagent (LipofectAMINE; Invitrogen) and 1 μg of each construct, according to manufacturer’s instructions. Renilla luciferase (pRL-CMV) was cotransfected as a control for efficiency of transfection. Five hours after transfection, the medium was removed from the cells and replaced with DMEM supplemented with 10% FBS. Twenty-three hours after this period, the medium was again removed from the cells, and two washes in PBS buffer (Invitrogen) were performed. 

The luciferase assays were performed on the transfected cell lysates using the standard protocol for a luciferase assay (Promega). Each pGL3 construct was transfected at least three to six times, and mean values taken from at least two independent transfections. Total protein concentration of the lysates was determined using the Coomassie blue standard protocol (Pierce, Rockford, IL). Relative light values were normalized per microgram of total protein in each sample. 

To determine the important cis-regulatory elements in the putative promoter region, in vitro mutagenesis (Quikchange; Stratagene, La Jolla, CA) was used to knockout specific regulatory motifs. Three bases were mutated within each putative cis-regulatory motif. The ERE motif was mutated at base-pairs −266 (G to A), −268 (A to G), and −270 (C to T). The most distal Sp1 site was mutated at base-pairs −195 (G to A), −196 (C to T), and −197 (G to A). The next Sp1 site was mutated at −167 (G to A), −168 (C to T), and −169 (G to A). The Sp1 site most proximal to exon 1 was mutated at −83 (C to T), −84 (G to A), and −85 (C to T). The Tant site was mutated at −102 (C to A), −103 (C to T), and −104 (G to A). Mutagenesis was confirmed by sequencing. 

In Situ Hybridization

Freshly enucleated mouse eyes were fixed overnight in 4% paraformaldehyde in phosphate-buffered saline (PBS) and incubated overnight in 20% sucrose in PBS at 4°C. The eyes were oriented and flash frozen in optimal cutting temperature (OCT) compound (BDH, Poole, UK) using a chamber of dry ice and isopentane. In situ hybridization analysis was performed with digoxigenin-labeled riboprobes on 10-μm cryostat sagittal sections through the central retina. The probes were generated from a pGEM-T plasmid (Promega) into which a 447-bp fragment of the mouse 3′ Fbln3 cDNA (forward primer, 5′-CTGGTCTTCTTCAAGAGAGC-3′, and reverse primer, 5′-GGGTATCTGGTTCATTTTGG-3′) had been cloned. Plasmids were linearized with either NotI (antisense probe) or NcoI (sense probe). Digoxigenin-labeled riboprobes were generated using either SP6 RNA polymerase (antisense) or T7 RNA polymerase (sense) in the presence of digoxigenin RNA labeling mix (Roche Diagnostics, Lews, UK) in accordance with the manufacturers’ instructions. Immediately before use, the probes were diluted to a concentration of 1:50 in hybridization buffer (10 mM Tris-HCl [pH 7.5], 200 mM NaCl, 5 mM NaH₂PO₄, 5 mM Na₂HPO₄, 5 mM EDTA, 1 mg/mL tRNA, 50% formamide, 10% dextran sulfate, 1× Denhardt’s). The probes were denatured for 5 minutes at 70°C and quenched on ice. Hybridization was performed overnight at 65°C in a humidified chamber of 50% formamide and 2× SSC. Immunodetection was performed using a 1:1000 dilution of anti-digoxigenin alkaline phosphatase-conjugated Fab fragments (Roche Diagnostics). Hybrids were visualized with the BCIP/NBT substrate (Roche Diagnostics). Images were viewed by microscope (Aristoplan; Leitz, Wetzlar, Germany) with digital image capture (DP10 camera; Olympus, Tokyo, Japan). In situ experiments were repeated on three different eyes. 

Antibodies

A peptide was synthesized commercially using Fmoc chemistry (Sigma-Genosys) corresponding to amino acids 37-52 (PADPQRIPSNPSHRIQ) of fibulin-3 protein, ^{19 20 21} avoiding the signal peptide sequence. The peptide was conjugated to keyhole limpet hemocyanin and then used to generate a rabbit polyclonal antibody by standard protocols. Serum samples were desalted before IgG fraction by diethylaminoethyl (DEAE) gel chromatography (Affi-gel Chromatography; Bio-Rad, Herts, UK). Protein concentrations were estimated with a Bradford assay (Pierce). 

Western Blot Analysis

Fibulin-3 protein was generated in vitro from a full-length FBLN3 clone (IMAGE clone 360091) using a coupled reticulolysate system (TNT T7/T3 Quick; Promega). Native fibulin-3 was obtained from either mouse retina or ARPE-19 cells. Mouse retinal extract was prepared as previously described. ¹³ ARPE-19 cells were washed two times in ice-cold PBS and lysed in 1 mL solubilization buffer (150 mM NaCl, 1% Tergitol [detergent: poly(oxy-1,2-ethanediyl),α-(4-nonylphenyl)-ω-hydroxy; Baker, Phillipsburg, NJ], 0.5% sodium deoxycholate, 0.1% SDS, and 50 mM Tris-HCl [pH 8.0]) containing protease inhibitor cocktail (Sigma-Aldrich) and kept on ice for 15 minutes. Cells were centrifuged for 5 minutes at 1000 rpm, resuspended in 400 μL of solubilization buffer, and physically fragmented by homogenization (Dounce; Bellco Glass, Vineland, NJ). Cells were then centrifuged at 13,000 rpm for 20 minutes. The supernatant was then mixed 4:1 with Laemmli sample buffer and denatured for 2 minutes at 99°C, and the proteins were separated by 12% SDS-PAGE relative to low molecular weight standards (Bio-Rad). Each sample loaded on a gel contained 10 μg of protein. Proteins were electrophoretically transferred onto polyvinylidene difluoride (PVDF) membrane (Immun-Blot; Bio-Rad) using a Tris-glycine buffer system (48 mM Tris, 39 mM glycine, 0.037% SDS, 20% methanol). The membrane was preincubated in blocking buffer (5% nonfat milk, 0.1% Tween-20 in PBS) at 4°C overnight followed by incubation with primary fibulin-3 antibody (diluted 1:150 in blocking buffer) for 1 hour at room temperature (RT). After three washes in PBS, 0.1% Tween-20, the membrane was incubated for 1 hour at RT with a 1:3000 dilution of goat anti-rabbit conjugated-alkaline phosphatase (AP) or -horseradish peroxidase (HRP) in blocking buffer. After three final PBS washes, the membrane was processed using the alkaline phosphatase color development or enhanced chemiluminescence (Amersham Pharmacia Biotech). Bands were quantitated by computer (Multianalyst, ver. 1; Bio-Rad). 

Statistical Analysis

Statistical analyses were performed on computer(SPSS, ver. 10.0; SPSS Science, Inc., Chicago, IL). Data were analyzed with one-way analysis of variance (ANOVA) and the Bonferroni post hoc test. All results are expressed as the mean ± SEM. P ≤ 0.05 was considered statistically significant. 

Results

Genomic Characterization of the Human FBLN3 Gene and Proximal Promoter

5′-RACE was performed on human retinal mRNA, and a product of approximately 400 bp was obtained, which was subsequently cloned and sequenced. Immediately upstream of exon 3 an exon of 68 bp was identified, (GenBank: AY027910). To confirm the presence of this exon in the FBLN3 transcript RT-PCR was performed on total RNA isolated from human retina and from the Y-79 and ARPE-19 cell lines. Two products were obtained in all samples tested, and these were found to differ in size by 41 bp. The larger product contained the 68-bp exon, followed by an additional exon of 41 bp, (GenBank: AY027911), then exons 3 to 12 of the FBLN3 transcript. The smaller product was the expected, alternatively spliced variant of the FBLN3 transcript that lacked the 41-bp exon. Thus, there is a simple alternative splicing pattern in which either exon 2 is present (exon 1 spliced to exon 2) or absent (exon 1 spliced to exon 3). The relevant exon/intron boundaries are consistent with the splice site consensus sequences (Fig. 2) . The 68- and 41-bp exons are hence confirmed as exons 1 and 2, respectively. 

Two genomic clones were then identified that contained the entire FBLN3 gene: one from the RPCI PAC library (310118) and one from the LLNL chromosome 2 library (AI8-B18). For isolation of the promoter region of human FBLN3, the exon 1 PCR product was radiolabeled and used to screen the EcoRI, HindIII, and PstI sublibraries produced from the genomic clones. An EcoRI subclone containing 5 kb of DNA including exons 1 to 3 of the FBLN3 gene and a PstI subclone containing 4 kb of DNA, including exons 1 and 2, were sequenced to obtain approximately 1.8 kb of sequence upstream of exon 1. This new sequence was then analyzed for putative transcription factor-binding motifs with the OMIGA and PROSCAN programs. Both programs identified several motifs and defined the minimal promoter as a region between −22 and −318 bp, numbering the putative transcription start site +1 as the start of exon 1. Position −812 corresponds to position 56109434 on chromosome 2 and position +1 of the sequence corresponds to position 56108622 (April 2003 freeze, UCSC BLAT server; http://genome.ucsc.edu/ hosted by the University of California Santa Cruz). 

No consensus TATA or CAAT boxes are present in this sequence. Instead, several Sp1 sites are present at positions −81 to −86, −164 to −172, and −194 to −199 (Fig. 2) . Other recognition sites present include UCE-2 (−25 to −29) and Tant (T-protein binding; −102 to −106). Of interest was the estrogen response element (ERE) also identified at position −261 to −273 (5′-TGTCAACGTGTCC-3′), which closely resembles the canonical ERE sequence (5′-GGTCANNNTGACC-3′). ²² There are two single nucleotide polymorphisms (SNPs) in the 5′ untranslated region (UTR) at position −80 bp (C/T; SNP 3762514) and position +50 bp (A/G; SNP 3762515). A multiple alignment of the 5′UTR from −172 to +43 bp in human, mouse, and rat show a high level of sequence homology, with the first Sp1 site and Tant site being highly conserved (Fig. 3A) . Homology at the nucleotide level further upstream is weaker; however, the rodent sequences have similar putative cis-acting elements (Fig. 3B) . Both the mouse and rat have a 5′ ERE half site and an NFE-1 site at about the same positions. All three sequences contain a (CA)n polymorphic repeat; however, there appears to be an inversion in the rodent genomes from the (CA)n to NFY-CBF motif. 

Functional Analysis of the Promoter

Deletion constructs were generated to examine the 5′ upstream sequence of the FBLN3 gene. Specific putative cis-acting motifs were mutated as illustrated in Figure 4 . Removal of the region −808 to −425 bp upstream of exon 1 had no effect on transcriptional activation, suggesting that the NFE-1 and NFY-CBF motifs are not major factors in high-level FBLN3 transcription. Removal of the next 182-bp (including the ERE site) however resulted in a marked increase (doubling) in luciferase activity. Removal of a further 61 bp caused a dramatic decline in relative luciferase activity. This suggests that the most important positive regulatory elements were contained within the c-243 construct. In addition, the difference in relative luciferase activity between constructs c-425 and -243 suggests that a significant negative regulatory element exists between base pairs −243 and −425. 

Specific mutations within the c-425, -243, and -808 constructs highlighted the relative roles of particular motifs (Fig. 4) . Mutagenesis at the ERE site resulted in marked upregulation of luciferase activity equivalent to that with construct c-243. This suggests that the ERE site is the major inhibitory element between base-pairs −243 and −425. Sequential mutagenesis of the three proximal Sp1-binding sites suggests equal contribution to luciferase activation. Most activation was seen when all three were present. Some luciferase activity was retained if all Sp1-binding sites were mutagenized, but this was abolished if the Tant site was also removed. This suggests that this regulatory motif is also important, but that the UCE-2 motif is not important in FBLN3 transcriptional regulation. 

Expression in Murine and Human Retina

In situ studies suggest that Fbln3 is expressed in the murine retina in the outer nuclear layer (ONL), the inner nuclear layer (INL), and possibly the RPE but not ganglion cell layer (GCL; Fig. 5A ) Because dominant drusen is considered an RPE disease, ^{7 12} we confirmed that it is expressed in human RPE (Fig. 5B) . RT-PCR of human eye mRNA showed expression as early as 8.6 weeks after conception (Fig. 6A) , and in fetal murine eyes expression was seen as early as E9.5 (Fig. 6B) . This suggests a role for FBLN3/Fbln3 in the developing mammalian eye. 

Regulation of Fibulin-3 in Human-Derived Cultured RPE Cells

We first generated a rabbit polyclonal antibody against a human fibulin-3 peptide that exhibited high specificity for both native and in vitro–generated fibulin-3 protein by Western blot and immunoprecipitation analyses. It specifically detected fibulin-3 from ARPE-19 cells (Fig. 7A) and culture medium (Fig. 7C) , from mouse retinal extracts (Fig. 7B) , and from COS-7 cells (Fig. 7E) demonstrating its cross-species reactivity. It detected a 55-kDa band from fibulin-3 protein generated in vitro (Fig. 7D) as would be predicted from its sequence, yet from cells and tissue, it detected a 43-kDa band. This is in contrast to other studies that have suggested that fibulin-3 is a 55-kDa protein. ^{12 19} However, there is an alternative signaling peptide cleavage site after alanine in position 124 that generates a 43-kDa protein, as previously shown. ¹⁹ Thus, it is likely that we are detecting this fibulin-3 variant. 

Using this characterized fibulin-3 antibody, we studied the production of fibulin-3 protein in human-derived ARPE-19 cells to explore further the possible role of estrogen in the transcriptional regulation of FBLN3. We measured intracellular rather than extracellular levels to quantify more directly and sensitively the changes in protein production in response to promoter modulation. Serum withdrawal resulted in a rapid increase in fibulin-3 production (Fig. 8A) . Although this was a nonspecific response, we hypothesized that it may be due to withdrawal of the estrogen from the culture medium. Adding 17β-estradiol to the culture medium resulted in a rapid decrease in fibulin-3 production (Fig. 8B) . This response was dose dependent, with a more rapid decrease in fibulin-3 production using 100 pM compared with 10 pM 17β-estradiol (Fig. 8B) . Similar results were obtained when COS-7 rather than ARPE-19 cells were used (Fig. 8) . 

A study of the effect of tamoxifen on fibulin-3 mRNA and protein production was also performed. Adding 1 μM tamoxifen alone had little effect on basal FBLN3 mRNA (Fig. 9A , comparing lanes 1 and 4) or fibulin-3 protein production (Fig. 9B , comparing lanes 1 and 4). It was most striking that adding tamoxifen blocked the effect of estrogen on both FBLN3 mRNA (Fig 9A , compare lanes 2 and 3) and fibulin-3 protein production (Fig. 9B , compare lanes 2 and 3). Densitometry results from the Western blot experiment (Fig. 9B) confirm the effects of tamoxifen and 17β-estradiol on fibulin-3 protein production (Fig. 9C) . 

Discussion

The detailed description of the two untranslated exons completes the genomic characterization of the whole FBLN3 gene that comprises 12 exons spanning 18 kb of genomic DNA. ^{19 20 21} FBLN3 mRNA is alternatively spliced where exons 1 and 2 are spliced to exon 3, or exon 1 is spliced directly to exon 3. Northern blot analysis in mice have suggested only one transcript of 2.2 kb, ⁷ but in human tissues two transcripts are described. ¹⁹ These two transcripts in humans are 2.2 and 3 kb in size, and these could be explained by the human transcript having two different length 3′ UTRs. The function of these two transcripts is unclear. Alternatively spliced transcripts are often associated with regulation of spatiotemporal transcriptional expression. ²³ Alternatively, they may play a role in regulating fibulin-3 translation. Exon 1 is relatively GC rich (exon 1, 81% GC; exon 2, 46% GC) and such exons have been shown to hinder RNA polymerase, thus regulating protein translation. ²³  

Analysis of the 5′ end of the FBLN3 gene has identified important transcriptional regulatory elements. Transient transfection assays demonstrated that the minimum sequence needed for high-level transcription of the human gene is approximately 106 bp upstream of untranslated exon 1, which contained a Tant site ²⁴ (binding T-protein) and one Sp1 element that seems to be conserved in mouse and rat. However, deletion of the ERE element from the −425-bp construct demonstrated that it is an active inhibitory regulator of FBLN3 transcription. We therefore conclude that the key FBLN3 proximal promoter elements are contained within 425 bp upstream of the untranslated exon 1. This proximal promoter is also GC-rich. Specifically, 200 bp upstream of exon 1 is 75% GC and contains 23 CpG dinucleotides. It is interesting that this is similar to the FBLN1 proximal promoter. ²⁵ The C residues in the CpG dinucleotides are major sites of DNA methylation, and the methylation state of the 5′ end of genes is linked to the control of gene transcription. ²⁶ The three cis-acting Sp1 sites and the ERE identified in this study may be active in transcription initiation, because it has been suggested that Sp1 has a role in transcription initiation in GC-rich and ERE-containing gene promoters. ²⁷ DNA-protein interactions in the human FBLN3 promoter sequence must be verified by DNase I footprinting and electrophoretic mobility shift assays, as well as cotransfection studies of FBLN3 constructs with recombinant estrogen receptor α, to show direct action of 17β-estradiol on the promoter. 

Results of in situ hybridization and RT-PCR showed that FBLN3/Fbln3 mRNA is expressed in mammalian neurosensory retina and the retinal pigment epithelium (RPE). Most striking was that Fbln3 mRNA was not detected in the GCL in mouse retina. This is particularly interesting, because it has been shown that in human retina, fibulin-3 protein is predominantly found in the nerve fiber layer, ¹² but the source is unclear. It may be that although fibulin-3 is not produced in mouse ganglion cells, it is in humans, suggesting an important difference between mouse and human retina. Alternatively, it could be that fibulin-3 in the nerve fiber layer in humans is not derived from ganglion cells. 

Fbln1, Fbln2, and Fbln5 and have been shown to be expressed during development. ^{28 29 30} In this study, we establish that Fbln3 is also expressed during mammalian development, at least in the eye. In particular we also show that FBLN3 is expressed very early in the developing human eye. ³¹ The role of fibulin-3 in the developing retina is unknown, however. Fbln1 ^−/− mice die soon after birth, ³² suggesting that fibulins as a class play a crucial role in mammalian development. 

The identification of the ERE motif as an inhibitory transcriptional regulator was a particularly striking conclusion from the reporter assays. It is most likely that in dominant drusen, the target cell for disease is the RPE. It is therefore important that we have shown that extracellular estrogen could be a factor that modulates fibulin-3 production in cells derived from the RPE. To some extent, these results do not establish a direct effect of estrogen on the FBLN3 proximal promoter, although two findings show that it probably has an effect. First, the 17β-estradiol downregulation of fibulin-3 was blocked by tamoxifen, an estrogen antagonist that works by competing with estrogen for binding with cytoplasmic estrogen receptors that in turn would bind to the ERE. Second, the rapidity of the response and the fact that it was dose dependent also suggest a direct, classic effect on fibulin-3 production. ³³  

Long-term tamoxifen treatment itself has been shown to cause a retinopathy with both intracellular and extracellular deposits, ³⁴ and tamoxifen has direct effects on RPE cells. ³⁵ Our study suggests that estrogen plays a role in the expression of retinal extracellular matrix components, and it may be that matrix constituents such as fibulin-3 may play a role in tamoxifen-induced retinopathy. 

Very little is yet known about the role of estrogen in the retina. Estrogen receptors have been found in rodent, ³⁶ bovine, ³⁶ fish, ³⁷ and human retina. ^{38 39} Among their many actions, estrogens regulate the expression of genes important for extracellular matrix turnover including tissue inhibitors of metalloproteinase (TIMPs) and matrix metalloproteinases such as collagenase and stromelysin. ⁴⁰ Specifically, in the retina it has been shown that estrogen can indirectly upregulate expression of gelatinase (matrix metalloprotease-2) in human RPE cells. ⁴¹ This role for estrogen in retinal extracellular matrix may be relevant to retinal disease. It is beyond the scope of this work to determine whether estrogen status influences dominant drusen severity. A number of large population-based studies, however, have suggested that estrogen deficiency, as in early menopause, may be associated with age-related macular degeneration, ^{42 43 44 45} although this is still not certain. ⁴⁶ This may relate to the fact that fibulin-3 has been found as a constituent (albeit minor) of the retinal deposits in age-related macular degeneration. ¹² It is also known that estrogen upregulates fibulin-1 expression ⁴⁷ and that there are estrogen receptor half sites in the proximal promoter of fibulin-2. ⁴⁸ This may imply that estrogen control of gene transcription is a common theme in the expression of fibulins. 

Our work therefore enhances findings in previous studies and support the role of fibulin-3 in the developing and mature mammalian retina and, at least in humans, a role in retinal disease. As the expression of retinal fibulin-3 can be modified by estrogen, this may suggest a novel role by which estrogen modulates the extracellular matrix in the outer retina. Further work is needed to gain better understanding of the role of fibulins in the retina and the effects on fibulins of circulating estrogen. 

 Supported by British Retinitis Pigmentosa Society Project Grant GR614.

 Submitted for publication February 3, 2003; revised July 8, 2003; accepted July 29, 2003.

 Disclosure: J. Blackburn, None; E.E. Tarttelin, None; C.Y. Gregory-Evans, None; M. Moosajee, None; K. Gregory-Evans, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Schematic of comparative protein structure of the fibulin gene family. Each protein has an N-terminal and domains I through III. HUGO gene name is shown in italic; checkered box: signal peptide; open circles, anaphylatoxin motif; open squares, Ca²⁺-binding EGF repeats; open ovals, globular domain; TSP, thrombospondin element; IgG/MHC, immunoglobulin-major histocompatibility complex region.

Figure 2.

 

Sequence at the 5′-end of the human FBLN3 gene. (A) Exons 1 and 2 (GenBank: AY027910 and AY027911) in bold uppercase. FBLN3 transcription start site is numbered +1 (position 56108622, chromosome 2; April 2003 freeze, UCSC BLAT server available at http://genome.ucsc.edu/). Sp1 sites present between nucleotides −81 to −86, −164 to −172, and −194 to −199. An estrogen response element (ERE: −261 to −273) is highlighted. Deletion construct 5′ ends indicated with forward arrow and the common 3′ end with a reverse arrow. Vertical arrowheads: SNPs.

Figure 3.

 

Comparison of human (h), mouse (m), and rat (r) FBLN3/Fbln3 proximal promoters. (A) Human sequence −172 to +43 bp, compared with mouse and rat. Cis-elements are bold underscored; Exon 1 is in uppercase; 5′UTR/promoter is in lowercase; vertical lines show homology; dotted gaps in mouse/rat represent sequence matching human sequence. (B) Schematic comparison of promoter organization. Filled boxes: conserved elements; open boxes: elements not conserved between species; hatched boxes: dinucleotide repeat; checkered boxes: 5′-ERE half-site. Numbers above boxes are the start of each motif. Human sequence +1 corresponds to position 56108622 on chromosome 2; in mouse +1 corresponds to position 29012191 on chromosome 11; in rat +1 corresponds to position 105560285 on chromosome 14.

Figure 4.

 

Functional analysis of human FBLN3 putative promoter. Top: schematic representation of region upstream of exon 1. Symbols represent cis-acting elements and the hatched box is exon 1. Bottom left: schematic representation of FBLN3 deletion (c) and mutation (M) constructs transfected into COS-7 cells. M1 to M10 indicate which specific cis-element was ablated in specific deletion constructs. L: firefly luciferase reporter. Bottom right: luciferase activities of the deletion and mutation constructs relative to the full-length construct (c-808 = 100%). Results are the means ± SEM from at least three separate transfections performed in triplicate. Empty vector was used as the negative control. *P < 0.001, relative to c-808 activity.

Figure 5.

 

Expression pattern of Fbln3. (A) In situ mRNA expression in adult murine retina showing expression in the ONL and INL (arrows) and absence from the GCL. (B) RT-PCR showing FBLN3 expression in a human-derived RPE cell line (ARPE19) and from a primate COS-7 cell line. The control was without reverse transcriptase (−RT).

Figure 6.

 

FBLN3/Fbln3 RT-PCR in developing mammalian eye. (A) Human fetal eye mRNA showing FBLN3 expression as early as 8.6 weeks after conception. (B) Murine fetal eye mRNA showing Fbln3 expression from E9.5 and corresponding Hprt expression as loading control (C). M, φX174/HaeIII DNA ladder; ±, with/without reverse transcriptase; AR, adult retina.

Figure 7.

 

Characterization of fibulin-3 peptide antibody. (A) Immunoblots of ARPE-19 cell extracts showing high specificity of the antibody for a single 43-kDa band. PI, preimmune serum negative reactivity; PEP1, preblocking with peptide used to raise the antibody; PEP2, preblocking with nonspecific peptide. (B) Immunoblots of mouse retinal tissue extract. Lane 1: preimmune serum; lane 2, fibulin-3 antibody. (C) Immunoblots of ARPE-19 cell extract (lane 1) and immunoprecipitated 46-kDa protein from ARPE-19 cell culture medium (lane 2). H, heavy chain IgG; L, light chain IgG. (D) Immunoblot of the in vitro–translated full-length FBLN3 cDNA clone: (i) autoradiogram of 55-kDa protein; (ii) corresponding immunoblot. (E) Immunoblot of COS-7 cell extract (lane 1) showing presence of fibulin-3 protein. M, molecular weight standards in kilodaltons.

Figure 8.

 

Fibulin-3 expression in ARPE-19 and COS-7 cells in response-changing culture media. (A) Serum withdrawal increased fibulin-3 protein levels in ARPE-19 (○) and COS-7 cells (▪). (B) Adding 100 pM 17β-estradiol reduced fibulin-3 protein levels in both ARPE-19 (○) and COS-7 cells (▪). The effect of 10 pM 17β-estradiol in ARPE-19 cells is also shown (▵). Each time point is the mean result of at least three independent experiments ± SEM. Data points that are significantly different from resting levels are denoted (*P < 0.01, **P < 0.05). (C) Representative Western blots used for quantitation of fibulin-3 levels. Lane 1: 0 hours; lane 2, 6 hours after serum withdrawal; lane 3, 0 hours; lane 4, 10 pM 17β-estradiol; lane 5, 0 hours; lane 6, 100 pM 17β-estradiol.

Figure 9.

 

Effect of tamoxifen on 17β-estradiol suppression of intracellular fibulin-3 level in ARPE-19 cells. 17β-Estradiol at 100 pM and tamoxifen at 1 μM were used. (A) RT-PCR of FBLN3 in the presence (lanes 1 to 4) or absence (lanes 5 to 8) of reverse transcriptase. Top band HPRT; bottom band FBLN3. Lane M: 100 kb DNA ladder; lanes 1 and 5: untreated cells; lanes 2 and 6: 17β-estradiol alone; lanes 3 and 7: 17β-estradiol+tamoxifen; lanes 4 and 8: tamoxifen alone. (B) Fibulin-3 Western blot. Lane M: molecular weight marker; lane 1: untreated cells; lane 2: 17β-estradiol alone; lane 3: 17β-estradiol+tamoxifen; and lane 4: tamoxifen alone. (C) Densitometric analysis of Western blot in (B). Lane I: untreated cells; lane II: 17β-estradiol alone; lane III: 17β-estradiol+tamoxifen; lane IV: tamoxifen alone. Each measurement is an average result of at least three experiments ± SEM. Data points that are significantly different to resting levels are denoted (*P < 0.05).