Diamantina Institute for Cancer, Immunology and Metabolic Medicine, University of Queensland, Queensland, Australia
Address reprint requests to: Gethin Thomas, PhD, Diamantina Institute for Cancer, Immunology, and Metabolic Medicine, University of Queensland, Princess Alexandra Hospital, Woolloongabba, Queensland 4102, Australia
Parts of this work were previously published in abstract form in the Proceedings of the 2006 Annual Meeting of the American Society of Bone and Mineral Research, September 15–19, 2006, Philadelphia, PA, USA, and the 2006 Annual Meeting of the Australian and New Zealand Bone and Mineral Society, October 22–26, 2006, Port Douglas, Queensland, Australia.
Drs Salois, Sellin, Bessette, and Godin are employees of Phenogene Therapeutics. All other authors state that they have no conflicts of interest.
Published online on April 21, 2008
In the course of attempting to define the bone “secretome” using a signal-trap screening approach, we identified a gene encoding a small membrane protein novel to osteoblasts. Although previously identified in silico as ifitm5, no localization or functional studies had been undertaken on this gene. We characterized the expression patterns and localization of this gene in vitro and in vivo and assessed its role in matrix mineralization in vitro. The bone specificity and shown role in mineralization led us to rename the gene bone restricted ifitm-like protein (Bril). Bril encodes a 14.8-kDa 134 amino acid protein with two transmembrane domains. Northern blot analysis showed bone-specific expression with no expression in other embryonic or adult tissues. In situ hybridization and immunohistochemistry in mouse embryos showed expression localized on the developing bone. Screening of cell lines showed Bril expression to be highest in osteoblasts, associated with the onset of matrix maturation/mineralization, suggesting a role in bone formation. Functional evidence of a role in mineralization was shown by adenovirus-mediated Bril overexpression and lentivirus-mediated Bril shRNA knockdown in vitro. Elevated Bril resulted in dose-dependent increases in mineralization in UMR106 and rat primary osteoblasts. Conversely, knockdown of Bril in MC3T3 osteoblasts resulted in reduced mineralization. Thus, we identified Bril as a novel osteoblast protein and showed a role in mineralization, possibly identifying a new regulatory pathway in bone formation.
The skeleton is required to conform to the complex demands of calcium storage, locomotive efficiency, and structural integrity. Bone cell activity is regulated at the systemic, paracrine, and autocrine levels, and bone itself is a complex structure of collagenous and noncollagenous proteins and calcium and phosphate ions (hydroxyapatite).(1) Many different proteins have been identified as playing essential roles in bone physiology, and many of those are either secreted or membrane bound. For instance, the PTH/PTH-related protein receptor (PTH/PTHrP receptor) mediates the effects of systemic (PTH) and local (PTHrP) factors.(2–4) RANKL is expressed on the osteoblast cell surface and mediates cell–cell interactions with its receptor, RANK, on the osteoclast cell surface.(5) Another key osteoblastic membrane protein is alkaline phosphatase, an enzyme involved in the mineralization process.(6–8) The recent discovery of a role of the Wnt–Fzd pathway in bone(9) and an osteocyte-specific secreted component of this pathway, sclerostin,(10) suggests skeletal regulation is still incompletely understood. Secreted and membrane-bound proteins are of particular interest in drug-development because their extracellular nature renders them more accessible for therapeutic intervention.
In an attempt to discover new genes encoding secreted and membrane proteins expressed in bone, we developed a signal trap screening system.(11) A number of cDNA libraries were constructed from various skeletal tissue sources and screened for signal peptide-containing proteins. Screening a UMR106 rat osteosarcoma cell line library identified a small transmembrane protein previously identified as Ifitm5. Ifitm5 was named by automated gene prediction because of its location adjacent to the interferon inducible transmembrane protein family (Ifitm), clustered on mouse and human chromosomes 7 and 11, respectively. The Ifitm genes, including Ifitm5, have the same structure, two coding exons separated by a small intron. Despite the family name, only Ifitm1 and Ifitm3, the most studied and ubiquitous members, have been shown to be regulated by interferons.(12,13) Several different functions have been attributed to these genes notably in homotypic aggregation of lymphocytes,(14) interferon-mediated cell growth inhibition,(14) and more recently in specification of germ cells during embryonic development.(15–17) However no functional or in-depth localization studies have been undertaken on Ifitm5.
Ifitm5 expression in an osteoblastic cell line lead us to study it in bone. In this report, we describe the expression pattern, protein localization and in vitro functional characterization of Ifitm5 confirming its osteoblastic nature and showing a possible role in the mineralization process. To more accurately reflect these findings, Ifitm5 was renamed Bril (bone-restricted Ifitm-like).
MATERIALS AND METHODS
Cloning of mouse Bril cDNA and tagging with 3xFlag epitope
Bril was initially identified by screening a cDNA library derived from the rat UMR106 osteosarcoma cell line as described previously.(18–21)
The coding portion of mouse Bril was amplified from total RNA of MC3T3-E1 subclone 4 osteoblastic cells (hereafter designated MC3T3) by RT-PCR. Total RNA (2 μg) was reverse transcribed with Superscript III (Invitrogen) according to the manufacturer's instructions. cDNA was amplified with rTaq (New England Biosciences [NEB]) with the following conditions; forward primer, 5′-gccaccatggacacttcatatccccg-3′; reverse primer, 5′-ttagttatagtcctcctcatc-3′, annealing temperature = 57°C, 25 cycles, 413-bp product (GenBank accession number EU380257). The product was purified and ligated into a CMV-based expression plasmid. The mouse Bril cDNA was subcloned from the latter construct either downstream or upstream of 3× Flag epitope-containing plasmids generating an N- or C-terminally tagged version. The 3× Flag coding sequence is preceded by an exogenous methionine and ends with an extra leucine for the N-term fusion (mdykdhdgdykdhdidykddddklMDTSY …) and ends with a stop codon for the C-term fusion (… EEDYNdykdhdgdykdhdidykddddkSTOP; capitals indicate Bril coding sequence). Analysis using two computational algorithms (TMbase; TMHMM server(22)) indicated that addition of the 3× Flag tag to either end would not perturb the natural topology of the protein.
Production of an anti-Bril antibody
A synthetic peptide (2DTSYPREDPRAPSS16C) corresponding to the N-terminal portion of mouse Bril with an extra C-terminal cysteine was conjugated to activated keyhole limpet hemocyanin with maleimide. The peptide/carrier complex was mixed with complete Freund's adjuvant and injected into rabbits according to standard protocols (Affinity Bioreagents, Golden, CO, USA). The antibody was affinity-purified using the same peptide.
Cell culture and transfection
Primary calvarial osteoblasts were cultured from 19- to 20-day-old embryonic rats as previously described(21) and cultured with or without 5 mM β-glycerophosphate (βGP; Sigma) depending on experimental conditions. UMR106 osteosarcoma cells (American Type Culture Collection [ATCC]), were grown in DMEM supplemented with 10% (vol/vol) FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin. Cells were cultured with or without 50 μg/ml ascorbic acid and 10 mM βGP depending on experimental conditions. The mouse MC3T3 cells were obtained from ATCC and maintained in αMEM supplemented with 10% FBS and supplemented with 50 μg/ml ascorbic acid and 3 mM βGP where indicated. HEK293 cells were grown in DMEM supplemented with 10% FBS and transfected using Fugene-6 as described.(23) SaOS-2 human osteosarcoma cells (ATCC) were cultured in αMEM containing 10% FBS, 10−8 M dexamethasone, 1.8 mM KH2PO4, and 10 mM HEPES, as previously described.(24) Apposition of calcium into monolayers of SaOS-2 cells was performed using a colorimetric assay, as previously described.(24) Human primary osteoblasts were derived and cultured from cancellous bone obtained with informed patient consent at surgery for hip arthroplasty, as previously described.(25)
Immunofluorescence and Western analysis
Bril immunofluorescence and Western blotting were performed as described previously(18) using a 1/2000 dilution of affinity-purified rabbit anti-mouse N-terminal Bril. Double immunofluorescence was carried out as described previously.(26) Rabbit anti-mouse N-terminal Bril anti-sera was used at 1/1500 and mouse anti-rat BSP antibody was used at 1/200 dilution (Sigma).
In situ hybridization and immunohistochemistry
In situ hybridization was performed on mouse embryonic tissue sections from e14.5 and e16.5 embryos as described previously.(21,27) The cRNA antisense probe corresponded to the mouse coding sequence of Bril. For immunohistochemical detection of Bril, mouse embryos were fixed in 4% paraformaldehyde/0.1% glutaraldehyde and treated by microwave irradiation as previously described,(28) embedded in paraffin, and 4-μm sections were cut. For immunolabeling, sections were deparaffinized, blocked with 5% skimmed milk in PBS for 1 h, and incubated with 1/500 dilution of anti-Bril anti-serum or its corresponding preimmune serum for 3 h. Goat anti-rabbit-horseradish peroxidase (HRP) linked antibody (1/1000) was added for 1 h, and detection was carried out with diaminobenzidine using the DAkoCytomation Envison+ System kit (Dako). Sections were counterstained with methyl green. All steps were carried out at room temperature.
Northern and RT-PCR analyses
Northern blotting and RT-PCR were performed as described previously.(21) For the Northern blots, a rat Bril cDNA probe corresponding to the full coding sequence (Fig. 1A) was used. For mouse Bril RT-PCR, the conditions were as follows: forward primer, 5′-gctggaacccatggacacttcatat-3′; reverse primer, 5′-gtcctcctcatcaaacttggtgct-3′; annealing temperature 57°C; 25 cycles; 406-bp product. The mouse β actin primers used were 5′-tgggtatggaatcctgtggc-3′ and 5′-cagctcagtaacagtccg-3′, annealing temperature was 57°C, there were 22 cycles and a 348-bp product. The other probes and PCR conditions were as described previously.(21) For human Bril qRT-PCR, the specific forward primer was 5′-gagaccacttgatctggtcggt-3′ and the reverse primer was 5′-ccaccttctgatctcgggcctt-3′ (annealing temperature 64°C) generating a 105-bp fragment. Expression was quantified using SYBR-green incorporation and the comparative cycle threshold (ΔCT) method. Osteocalcin (OCN) qRT-PCR was as described previously,(24) and both osteocalcin and Bril expression were normalized to GAPDH.(24) To facilitate graphing both Bril and OCN expression on the same axis, Bril is expressed as 10−3 U and OCN as 10−6 U.
Generation of adenoviruses and Bril overexpression in primary osteoblasts cultures and UMR106 cells
To obtain a “shuttle” vector for insertion of transcription units into the adenoviral genome, the coding regions for mouse Bril or green fluorescent protein (GFP) from pQBI50-FC2 (qBiogene) were cloned downstream of the CMV promoter of plasmid pQBI-AdBN (qBiogene). The resulting vectors had the CMV transcriptional unit flanked by nucleotides 1–102 and nucleotides 3334–5779 of the adenovirus serotype 5 genome. Adenoviruses expressing Bril or GFP were generated by homologous recombination after co-transfection of the linearized plasmids together with the replication-defective genome AdCMVlacZΔE1/ΔE3 into HEK293 cells by calcium phosphate precipitation using standard protocols. Two days after transfection, cells were overlaid with medium containing 1.25% (wt/vol) low melting agarose. Isolated viruses producing plaques were picked after 14 days and tested for expression of Bril or GFP after infection of native HEK293 cells. Positive clones were plaque purified and amplified in HEK293 cells. Viral particles were purified on discontinuous and continuous cesium chloride gradients according to standard protocols (OD260).
Primary osteoblast cultures were infected with adenoviruses at confluence at the multiplicity of infection (MOI) indicated in normal primary cell media without βGP overnight. Media were changed, and cells were placed in normal primary culture media containing 5 mM βGP and cultured for a further 6 days before analysis. UMR106 cells were infected at confluence at the indicated MOI in normal medium without ascorbic acid or βGP. Media were changed after an overnight incubation with virus into fresh medium with or without 10 mM βGP and 50 μg/ml ascorbic acid. Cells were cultured for a further 6 days with medium changes every 2 days. Mineralization was assessed by 45Ca uptake into the cell monolayer as described previously.(21) von Kossa staining was used to visualize mineral deposition in UMR106 cultures. Briefly, cells were fixed in 4% paraformaldehyde stained with 1% silver nitrate (Sigma) under UV and the stain fixed in 5% sodium thiosulphate (Sigma).
Lentivirus-mediated small hairpin RNA knockdown of Bril in MC3T3 cells
Four different small hairpin RNAs (shRNAs) against mouse Bril were designed using commercial algorithms (Target Finder; Ambion). The regions chosen for the various shRNAs were those having least homology with other members of the Ifitm gene family. They were designated 1–4 and matched nucleotides 55–75, 290–310, 402–422, and 438–458 of the full-length mouse Bril GenBank clone NM_053088, as depicted in Fig. 7A. Oligonucleotide linkers were designed to have a 5′ AgeI and a 3′ EcoRI overhang. The sense and antisense 21-mer sequences were separated by an intervening XhoI loop and terminated with a polIII termination site. The following oligonucleotide sequences for shRNA 1 is given as an example: forward, 5′CCGGAACCCATGGACACTTCATATCCTCGAGGATATGAAGTGTCCATGGGTTTTTT-3′ and reverse, 5′AATTAAAAAACCCATGGACACTTCATATCCTCGAGGATATGAAGTGTCCATGGGTT-3′, where the XhoI loop is underlined, the AgeI–EcoRI overhangs are italicized, and the termination signal is bolded. After annealing, the double-stranded shRNA linkers were ligated downstream of the U6 promoter in the AgeI–EcoRI predigested pLKO.1-puro lentivirus backbone (Sigma). The latter plasmid also encodes the puromycin resistance gene driven by the phosphoglycerate kinase-promoter allowing selection of infected cells. All plasmids were confirmed by DNA sequencing. Lentiviruses were produced in HEK293 cells as previously described.(29) Lentivirus titers were determined in HEK293 and ranged from 1 × 107 to 1 × 108 infectious units/ml. Conditioned media containing the viral particles were used directly to infect MC3T3. As controls, two other lentivirus preparations were used carrying either the empty parental pLKO.1-puro plasmid (no shRNA insert) or a nontarget “nonsense” shRNA (pLKO.1-NT; Sigma). MC3T3 cells (50,000 cells/well) were cultured in a 6-well plate overnight and infected with HEK293 lentivirus-containing conditioned media in the presence of 8 μg/ml of polybrene (Sigma) for 8 h. Medium was changed, cells were grown until near confluence and passaged into 10-cm dishes, and puromycin was added (5 μg/ml) for selection. After 3 days, the cells were passaged again and seeded in 6-well plates for experiments. At confluence, cells were transferred into media supplemented with ascorbic acid and βGP. After 12 days, cells were collected and processed for RT-PCR and Western blotting analyses as detailed in the above sections. Mineralization was assessed by Alizarin red staining on ethanol-fixed cells with a 2% (wt/vol) solution (pH 4.2) for 10 min.
Identification and cloning of Bril
The rat Bril sequence was originally identified as part of a signal-trap screen performed in mammalian cells for cDNAs encoding secreted or membrane-bound proteins.(11) A cDNA fragment of 309 bp was retrieved from a library derived from confluent UMR106 cells (GenBank accession number EU380256). It contained 109 bp of 5′UTR sequence followed by an open reading frame of 66 residues. The translation product was predicted to contain a putative signal anchor as determined by the SignalP algorithm.(30) Further in silico analysis showed the retrieved fragment was homologous to a mouse full-length cDNA termed hemopoiesis-related membrane protein,(31) which was subsequently renamed interferon-induced transmembrane 5 (Ifitm 5) because of its close chromosomal localization, similar gene structure, and weak homology to the Ifitm family. However, we have renamed this gene Bril (bone-restricted Ifitm-like) to better reflect its osteoblastic localization and in vitro functional role described here.
Alignment of Bril protein sequences from public databases showed cross-species conservation as illustrated in Fig. 1A, sharing 88% similarity between mouse and human and 96% similarity between mouse and rat. In all species examined, the Bril gene is composed of two short exons, each encoding about one half of the protein (Fig. 1A). Mouse Bril has a predicted molecular mass of 14.6 kDa with no obvious motifs or domains common to other proteins. Bril possesses two transmembrane domains, with both N and C termini extracellular and an intracellular loop (see below, Fig. 4). The Bril gene is located on syntenic regions of chromosomes 7 and 11 in mice and humans, respectively, at the end of the cluster of the Ifitm genes (Fig. 1B). Although chromosomal localization and similar gene structure may suggest a common ancestry between Bril and the Ifitm family, protein sequence alignment (Fig. 1C) showed Bril is markedly dissimilar (Fig. 1D), most notably in the N and C termini extracellular regions.
Bril expression pattern in rats and mice
Northern blot hybridization on a panel of 20 different adult rat tissues indicated Bril expression is highly restricted (Fig. 2A). Signal was detected only in bone samples (mandible and calvaria) in addition to the two rat osteosarcoma cell lines UMR106 and ROS 17/2.8. The size of the transcript detected is in accordance with that of the predicted full-length cDNA sequence for the rat, roughly 750 bp (transcript ID ENSRNOT00000020023 at Ensembl). Northern blotting of embryonic and adult mouse tissues also confirmed the bone-restricted pattern of Bril expression (data not shown). To further establish the pattern of Bril expression in bone, expression was analyzed by Northern blot in long bones and calvaria in rats (Fig. 2B). Bril expression was detected in both long bones (femur and tibia) and calvaria from embryonic day 19 up to 8 mo of age, with expression decreasing in older bones (Fig. 2B). Similar patterns were also seen in mouse bone (data not shown). To assess the site-specific expression of Bril in developing bones, in situ hybridization was performed on embryonic mouse femora and tibias (Fig. 3). Bril was found to be highly expressed in the bone collar surrounding the developing diaphysis and at the metaphyseal/epiphyseal region in an e14.5 femur (Fig. 3A,C). A similar cortical pattern of expression was seen in an e16.5 tibia (Figs. 3B and 3D), with strong expression in the metaphyseal region possibly representing early trabecular bone. No significant expression was detected in the surrounding muscles, skin, and connective tissues. The sense Bril probe gave no significant specific signal (not shown). Immunohistochemical detection for Bril performed on mouse e15 embryos confirmed expression was confined to the developing bone collar in the humerus (Figs. 3E and 3F). Regions of vertebrae where endochondral ossification had started around the hypertrophic chondrocytes also showed strong Bril reactivity (Fig. 3H, arrowheads), whereas zones of nonhypertrophied chondrocytes were negative (Fig. 3H, arrows). Staining was also evident in intramembranous bony elements such as the orbital bone (Fig. 3I), mandible, and clavicle (data not shown). Serial sections incubated with preimmune antiserum were devoid of any specific signal (Fig. 3G).
Biochemical characterization of Bril in cultured cells
Western blotting using UMR106 cell extracts confirmed Bril migrated close to the expected 14.8-kDa size (Fig. 4A). The anti-mouse Bril antibody developed in rabbits was used and was expected to react equally well with the rat protein because only the last amino acid in the epitope differs (Fig. 1A). Both the total cell extract and NP-40 detergent-soluble cell fraction displayed a single immunoreactive band migrating close to the 16-kDa marker (Fig. 4A). The presence of three highly conserved cysteine residues within Bril could result in the formation of disulfide bridges. To test this, NP-40–soluble extracts from UMR106 cells were prepared and analyzed under reducing or nonreducing conditions by Western blotting (Fig. 4B). When probed with the N-terminal anti-Bril antibody, no discernible shift in the Bril protein band was observed under nonreducing conditions, suggesting no significant cysteine-mediated complexes were formed (Fig. 4B). Immunofluorescence localization on nonpermeabilized UMR106 cells gave a signal characteristic of membrane compartmentalization, highlighting the periphery of the cells (Fig. 4C). This result clearly indicated that Bril N terminus was extracellular.
To confirm Bril cell surface topology, the coding cDNA sequence of mouse Bril was N- or C-terminally tagged with a 3× Flag epitope and transfected into HEK293 cells, and the topology assessed by immunofluorescence with anti-Bril and anti-Flag M2 antibodies. Western blotting confirmed the anti-Bril antibody (Fig. 4D, top panel) detected the correct size wildtype mouse Bril (14.8 kDa) and the slightly higher molecular weight 3× Flag-Bril (17.6 kDa) and Bril-3× Flag (17.4 kDa). Only the 3× Flag versions were detected with the anti-Flag antibody (Fig. 4D, middle panel). Smaller molecular weight species were detected for the anti-Flag antibody, possibly resulting from internal proteolytic action or an abnormal internal cryptic translation site generating a shorter protein (Fig. 4D, middle panel). Immunofluorescence labeling performed on nonpermeabilized cells showed similar membrane localization for both the anti-Bril (Fig. 4E, a–c) and the anti-Flag antibodies (Fig. 4E, d–f). These data confirm the predicted topological features of Bril where both termini are extracellular and accessible to antibody reactivity. Interestingly, immunostaining performed on permeabilized HEK293 cells stably expressing mouse Bril showed the most intense membrane localization was at cell-to-cell contacts (Fig. 4E, h). Mock transfected cells (Fig. 4E, g) and cells stably expressing antisense Bril (Fig. 4E, i) gave only a faint background signal when probed with the anti-Bril antibody.
Bril expression in osteoblast cultures
The in situ hybridization and immunohistochemical detection, as well as UMR106 expression, suggested Bril maybe an osteoblast-specific gene. To study this further, MC3T3 osteoblasts were used as a well-established model that mimic normal osteoblast development; proliferation, differentiation, and mineralization.(32)Bril expression was absent from proliferating MC3T3s at 2 days after confluence, started on day 7 (differentiating cells), increased sharply until day 21, and declined at day 27 (Fig. 5A). Primary rat osteoblasts also lacked expression of Bril at confluence (day 5) but showed increasing expression through days 10 and 15 (Fig. 5B). In SaOS-2 osteosarcoma human cells, Bril expression also correlated tightly with differentiation and in vitro mineralization (Fig. 5C). The expression of Bril during matrix formation/maturation and a similar expression pattern to osteocalcin (a marker of osteoblast differentiation) suggests a role early in the mineralization process. Bril expression was also detectable in human primary osteoblasts derived from bone explant cultures (data not shown).
Bril expression was also assessed in other bone-associated cell types, with no expression being seen in the MLO-Y4 osteocytic cell line and in cultured mouse bone marrow–derived osteoclasts (data not shown). Further support for a role in osteoblast mineralization comes from immunohistochemical localization of Bril in mineralizing primary rat osteoblast cultures. Bril is clearly localized to the mineralizing nodules (Fig. 5D) and also co-localizes with bone sialoprotein (BSP), a mineral-associated osteoblast-specific protein.
Functional characterization of Bril in UMR106 and MC3T3 cells
To elucidate the function of Bril, an adenovirus system was used to overexpress Bril in UMR106 cells (Fig. 6) and lentivirus-mediated shRNA was used to decrease Bril levels in MC3T3 osteoblasts (Fig. 7). UMR106 can be induced to mineralize by culturing the cells in the presence βGP for 6 days after confluence.(33) Cultures were infected at confluence when cell proliferation had largely ceased to minimize dilution of viral effects by cell division. A dose response for Bril overexpression was carried out by infecting cells with MOIs from 1 to 100 with either GFP- or Bril-expressing adenovirus. At all levels of infection, GFP had no effect on mineralization, as measured by 45Ca uptake into the matrix (Fig. 6A). However, Bril-adenovirus–infected cells showed a dose-responsive increase in 45Ca incorporation, significantly so at MOIs of 20, 40, and 100 by 32%, 41%, and 42%, respectively (Fig. 6A; p < 0.01). The increased mineralization in Bril overexpressing cultures was visualized by enhanced von Kossa staining (MOI 20 shown; Fig. 6B). Bril overexpression in primary rat osteoblasts (MOI 40) also resulted in enhanced mineralization, exhibiting a 60% increase in 45Ca uptake (p < 0.01) relative to cultures infected with GFP-infected or noninfected cultures (Fig. 6C).
To study whether downregulation of Bril in osteoblasts would impair mineralization, MC3T3 osteoblasts were infected with lentiviruses expressing Bril-specific shRNAs. Four shRNA-expressing lentiviruses were generated (Fig. 7A), and three of these showed efficient Bril knockdown at both the RNA and protein level (Figs. 7B and 7C). MC3T3 cells infected with shRNAs 1–3 at day 3 showed decreased mineralization (measured at day 12) by Alizarin red staining (Fig. 7D). No effect was seen with either empty virus, virus expressing a nonspecific shRNA, or the nonfunctional shRNA 4 virus (Fig. 7D).
We identified a novel gene, Bril, in bone, which is highly enriched in osteoblasts and plays a role in the mineralization process.
Identification of novel genes acting in bone is a key approach to increasing the understanding of skeletal regulation and developing new therapeutic approaches for bone disease. This has already been shown for the development of anabolic therapies to replace lost bone with potential therapeutics being developed based on discoveries of novel genes or pathways involved in control of the skeleton, such as sclerostin and the Wnt-Fzd system.(34,35) To date, however, very few osteoblast-specific genes have been identified. In fact, currently, only five genes have been specifically associated with mineralizing tissue; two key transcription factors (Cbfa1(36) and osterix(37)), two structural proteins (BSP(38) and osteocalcin(39,40)), and the osteocyte-specific secreted protein sclerostin.(41,42) Our characterization of Bril has clearly shown that it is highly enriched in bone with both in vivo and in vitro data suggesting an osteoblastic origin. Our results concur with a recent array study identifying Bril (Ifitm5) in MC3T3 cells.(43) Further support for bone-specific expression of Bril came from in silico searches at the Mouse GeneAtlas GNF1M array dataset where Bril expression was only seen in bone on an array with 70 different cell and tissue samples.(44) The UniGene expressed sequence tag collection (Rn.82960, Mm.389989, Hs.443469) also shows Bril expression in bone. However, it cannot formally be excluded that other tissues express low level of Bril that would have gone undetected at the Northern level. Two previously published array studies have detected Bril in ATDC5 cells(45) and in the growth plate.(46) Both these sources are chondrocytic cells that mineralize. Bril was originally identified from a screen of bone marrow–derived hematopoietic stem cells.(31) However, the reported expression in brain and bone marrow was associated with a 2.1-kb mRNA signal, which clearly deviates from the 0.7-kb signal we observed (and predicted by GenBank) and may have been nonspecific. Macrophage and spleen expression of Bril has also been reported,(47) although no follow-up localization or functional studies have confirmed this, and we have not been able to reproduce these findings. Bril may be expressed at very low levels in blood cells, or the signal detected by Smith et al.(47) could have been caused by contaminating osteoblasts or nonspecific amplification caused by intraexonic primers or cross-amplification with other Ifitm members.
Our studies indicate both the N and C termini of Bril are extracellular. This seems to be in contrast to related Ifitm family members that have very short C termini (Fig. 1C), which has led to suggestions that only the N terminus is extracellular.(47) Interestingly, the N and C termini would be the most likely domains of Bril to interact with other proteins or cells, and it is these regions that display the least homology with the Ifitm proteins, further underlining its unique nature.
Bril expression increases with osteoblast differentiation, peaking with matrix production and mineralization, suggesting a role in the bone formation process. The in vivo embryonic localization shows Bril is associated with early bone formation. Colocalization of Bril with BSP is also suggestive of a role in the mineralization process. BSP has been hypothesized to be involved in mineralization(48–50) and is thought to complex with hydroxyapatite and the collagenous matrix.(49) In the previously cited study that identified Bril in MC3T3 cells,(43) it was found that culture conditions that were unfavorable for mineralization and expression of differentiated of osteoblasts markers (osteocalcin and BSP) also caused a drastic downregulation of Bril expression. Numerous reports have described altered levels of both secreted and membrane proteins in osteoblastic cells effecting mineralization, such as connexin 43,(51) BSP,(52) osteoactivin,(53) or Nell-1.(54) However, these effects are elicited through differing mechanisms, and none of the molecules studied bears resemblance to Bril. The data shown here does clearly show a role for Bril in the mature osteoblast, certainly at the stage of the mineralization of the matrix. Whether Bril is involved in the transition to an osteocytic phenotype is not clear.
The actual mechanism of action of Bril is unclear, and the lack of homology with other proteins (including the Ifitms) further complicates functional elucidation. Members of the Ifitm family have been associated with various functions including homotypic cell–cell interactions.(12,16,17) One speculative hypothesis would be that the membrane localization and the confirmed topology of Bril at the surface of mineralizing osteoblasts could promote the cohesion and aggregation of osteoblasts and favor the mineralization of newly formed osteoid. Whether this is mediated directly or indirectly by the extracellular portion of Bril or through interaction with companion proteins at the membrane or embedded in the extracellular matrix is still unknown. However, because of the sequence divergence of Bril relative to the other members, it could also be proposed that it has evolved for a more specialized function in osteoblasts, unrelated to the Ifitms.
Thus, we characterized Bril, an osteoblast-specific membrane protein that seems to play a role in the mineralization process or late stage osteoblast maturation. Identification of putative Bril-interacting factors is a priority and may provide, together with soluble forms of Bril, a promising potential pathway for bone therapeutics.
The authors thank A Wijenayaka and K Welldon for technical help and Drs C Lanctôt and M Crivelini for their contributions to the in situ hybridisation and immunohistochemistry data, respectively. Part of this research was funded by the Shriners of North America.