A Family of Human RNA-binding Proteins Related to the Drosophila Bruno Translational Regulator*

The post-transcriptional regulation of gene expression by RNA-binding proteins is an important element in controlling both normal cell functions and animal development. The diverse roles are demonstrated by the Elav family of RNA-binding proteins, where various members have been shown to regulate several processes involving mRNA. We have identified another family of RNA-binding proteins distantly related to the Elav family but closely related to Bruno, a translational regulator in Drosophila melanogaster . In humans, six Bruno-like genes have been identified, whereas other species such as Drosophila , Xenopus laevis , and Caenorhabditis elegans have at least two members of this family, and related genes have also been detected in plants and ascidians. The human BRUNOL2 and BRUNOL3 are 92% identical in the RNA-binding domains, although the BRUNOL2 gene is expressed ubiquitously whereas BRUNOL3 is expressed predominantly in the heart, muscle, and nervous system. Both of these proteins bind the same target RNA, the Bruno response element. The RNA-binding domain that recognizes the Bruno response element is composed of two consecutive RNA recognition motifs at the amino terminus of vertebrate Bruno protein. The possible involvement of the Bruno family of proteins in the CUG repeat expansion disease myotonic dystrophy is discussed.

defined that contains a conserved domain, called the RNA recognition motif (RRM), an 80 -90 amino acid domain (2)(3)(4). The most highly conserved sequences within the RRM are the ribonucleoprotein 1 (RNP1) and RNP2 motifs that are signature sequences for the RRM and have been shown to specifically interact with RNA (5-7). Often, consecutive copies of the RRM combine to form a single RNA-binding domain (8 -12), although a single RRM can function to specifically bind RNA (3)(4)(5). In addition, the RRM can serve to mediate proteinprotein interactions (13)(14)(15). RRM-containing proteins (RRM proteins) usually have multiple functional domains, both an RNA-binding domain comprised of the RRM(s) and other domains that provide other functions to the corresponding fulllength protein (16). For example, the human heterogeneous nuclear ribonucleoprotein (hnRNP) A1 has two consecutive RRMs at the amino terminus that bind RNA or single-stranded DNA and then a carboxyl-terminal domain that contains a nuclear localization signal (17), mediates protein-protein interactions (18), and encodes an RNA annealing activity (19). Thus, RRM-containing proteins function as components of RNP complexes that mediate many post-transcriptional regulatory events.
The importance of RBPs in development is underscored by the isolation of mutants with interesting developmental phenotypes where the defective gene encodes an RBP. In Drosophila melanogaster (20), Caenorhabditis elegans (21), mouse (22), and Arabidopsis thaliana (23), mutants with defects in RBPs are defective in cell growth and differentiation. An example of an RBP that regulates development is provided by the Bruno protein and its role as a translational repressor of oskar mRNA. In Drosophila, oskar is required for formation of germ cells and positioning of the posterior of the embryo (24). Both oskar mRNA and the encoded protein must be properly localized to the posterior pole of the oocyte for correct development (25,26). Localized expression of Oskar protein is determined in part by translational silencing of the oskar mRNA outside of the posterior of the oocyte. This repression is mediated by cis-acting sequences in the 3Ј-untranslated region (UTR) of oskar mRNA called Bruno response elements (BREs) and a corresponding trans-acting factor, the Bruno protein. Deletion of these BREs results in inappropriate translation of Oskar protein in the anterior end of the oocyte leading to embryos with two posterior poles. The Bruno protein is an RRM-containing protein present in oocytes. Extracts prepared from Drosophila ovaries recapitulate this Bruno-dependent translational repression of oskar mRNA in vitro (27). By regulating the localized expression of Oskar, Bruno has a key role in germ cell formation and early embryogenesis.
The Bruno protein is similar in domain structure to the Elav family of proteins (28). Members of this family provide exam-ples of the diversity in RBP functions and regulatory mechanisms. The original member of this family, the elav gene, was identified in Drosophila through mutants with an embryonic lethal, abnormal visual system phenotype (20). One mechanism of Elav protein function is to regulate the alternative splicing of a cell adhesion molecule (29). A related gene in Drosophila, rbp9, functions in the cytoplasm during oogenesis and appears to control the accumulation of the Bag-of-Marbles protein, a regulator of oocyte differentiation (30). In vertebrates, four genes have been identified by similarity to the Drosophila elav gene (31). The corresponding proteins, which are detected both in the nucleus and the cytoplasm, bind to AU-rich elements in mRNA and can regulate the stability, translation, and localization of target mRNAs (28,32). Members of the Elav family play roles in regulating differentiation because overexpression of different family members enhances the differentiation of 1) 3T3-L1 cells into adipocytes (33), 2) the teratocarcinoma cell line N-Tera2 into neurons (34,35), 3) chicken neural crest stem cells into neurons (36), and 4) the PC12 pheochromocytoma cells into neurons (37). In embryos, overexpression of Elav-like proteins results in altered neural differentiation in both frogs (38) and mice (39). Thus, the Elav family has diverse roles in regulating development through several different mechanisms.
The Xenopus laevis etr-1 gene previously was identified as a marker of the developing nervous system and is distantly related to the elav gene (40). Subsequently, the Etr-1 protein was shown to be related to the Drosophila Bruno protein (41). Here we describe a family of human genes related to both Etr-1 and Bruno. The corresponding proteins have three RRMs and share a domain structure with the Elav family of proteins. We have characterized in detail two members of this family, the BRU-NOL2 gene, which is ubiquitously expressed, and the BRU-NOL3 gene, which is expressed preferentially in muscle, heart, and the nervous system. The BRUNOL2 and BRUNOL3 proteins bind to the same RNA sequence as the Drosophila Bruno protein, demonstrating a conservation of both protein sequence and RNA binding specificity. This binding occurs through the first two consecutive RRMs. The BRUNOL2 protein is identical to the CUGBP1, an RBP that binds to CUG repeats and is implicated in the etiology of the triplet repeat expansion disease myotonic dystrophy (42). Thus, members of this gene family may be involved in human disease as well as differentiation of specific cell types.

EXPERIMENTAL PROCEDURES
Nomenclature-The nomenclature for the genes in this family identified to date varies according to how the original clone was isolated. The original name for the family, Etr, elav-type ribonucleoprotein, is easily confused with the Elav family of genes. Other names, CUGBP, CAGH4, and neuroblastoma apoptosis-related RNA-binding protein (NAPOR) (see text) define limited properties of only one or two genes. Because the genes in this family are highly related by sequence, we propose a standard nomenclature to clearly identify this relationship. Similar to the Elav family, we propose that this family be named for the Drosophila member of the family, Bruno. According to the HUGO nomenclature (43), the human genes are named Bruno-like (BRUNOL), and each distinct gene receives a different number. The relationship of this nomenclature to that of previously isolated genes is as follows: BRUNOL1 ϭ CAGH4 (44); BRUNOL2 ϭ CUGBP1 (42); and BRU-NOL3 ϭ ETR-3 (45) or NAPOR1, -2, and -3 (46). The name of the genes in other species uses the nomenclature appropriate to that species.
Identification of Clones: Human Genes-Sequences of the human Bruno proteins were identified by searching the GenBank TM expressed sequence tag (EST) and high throughput genomic sequence (HTGS) data bases with the frog BrunoL-1 protein sequence using TBLASTN. Unless otherwise noted, EST cDNA clones were obtained either directly from the I.M.A.G.E. consortium or purchased from Research Genetics, American Type Culture Collection, or Genome Systems. Further sequencing of the cDNA clones was performed manually by Sequenase 2.0 dideoxy-sequencing reactions (Amersham Pharmacia Biotech). The sequence was obtained either from templates generated by transposon insertion using vector oligonucleotides (47) or from the original insert with synthetic oligonucleotide primers (Life Technologies, Inc.). The cDNA clone names sequenced to identify different Bruno contigs and the corresponding accession numbers for the sequences are identified in Table I. The clone F2607 was obtained from C.C. Liew (University of Toronto). The EST corresponding to HFBCC22 had two EcoRI inserts from two different genes. The 1.0-kilobase pair fragment corresponding to BRUNOL2 cDNA was subcloned and used for sequencing. For BRU-NOL2, two cDNA clones that had an overlap of 440 base pairs were combined to obtain the sequence for the complete open reading frame. These two clones had a single difference, the addition of 12 base pairs in the linker region of the corresponding protein in clone HIBBL93 that was absent in HFBCC22.
For BRUNOL4 and BRUNOL5, the cDNA sequences do not encode the complete open reading frame so additional sequence information was extracted from the HTGS data base to identify the missing sequences. First, the corresponding cDNA sequences were used to identify genomic sequence corresponding to the respective cDNA. The accession numbers for the genomic sequences are in Table I. Then exons identical to the cDNA sequence were assembled to confirm the cDNA sequence. Finally, the missing exons were identified from TFASTA searches of the genomic sequence with the Xenopus BrunoL-1 protein sequence. The conservation of sequence was sufficient to identify the missing exons, whereas a manual inspection of the sequence identified the precise exon boundaries. These missing exons were assembled to complete a fulllength predicted open reading frame. For Fig. 1B Chromosome location was obtained from public data bases of radiation hybrid maps using sequence-tagged sites linked to either cDNA or genomic clones. The data were obtained from the UniGene data base and GeneMap'99 web site.
Genes in Other Species-A similar strategy of data base searching was used to identify genes in other species. In C. elegans, the cDNA clone yk109f3, encoding the Etr-1 protein, was obtained from Yuji Kohara and sequenced (Accession no. U53931). This cDNA corresponds to gene T01D1.2. In Drosophila, the same genomic DNA clone that encodes the 3Ј-end of the arrest (bruno) gene also contains exons that  3 Recently, the Drosophila Genome Sequencing Consortium has identified these new genes as CG6319 and CG12478, respectively (50). An EST from Halocynthia roretzi encodes a putative protein corresponding to the first two RRMs of a Bruno protein. 4 Mouse EST clones I.M.A.G.E.:467796 and I.M.A.G.E.:474835 were partially sequenced and shown to encode the mouse orthologs of BRUNOL2 and BRUNOL4, respectively. 5 Suzuki et al. (52) recently reported the identification of the zebrafish etr-1 (brunol-1) and brul (brunol-2) genes. Other Bruno genes in rat, mouse, and zebrafish identified by data base searching were detected but are not reported because of their orthologous relationship to the genes described in this report.
Sequence Analysis-Sequence analysis was performed with webbased searches of data bases through the National Center for Biotechnology Information (NCBI) and using the GCG suite of programs (Wisconsin Package Version 8.0, Genetics Computer Group (GCG), Madison, WI). The multiple sequence alignment was produced with the GCG Pileup program, whereas the tree dendrogram was produced with ClustalW and then displayed with TreeView.
Northern Blots-The cDNA inserts for BRUNOL2 and BRUNOL3 were labeled by random-primed synthesis in the presence of [ 32 P]dCTP (Amersham Pharmacia Biotech) and used to probe a human tissue Northern blot (CLONTECH; Catalog 7760 -1) with hybridization and washing conditions as described previously (53). Following washing, the filter was exposed to a phosphorstorage screen for 24 h, and the signal was detected with a PhosphorImager (Molecular Dynamics).
Plasmids for Protein Expression-General molecular biology methods were as described in Sambrook et al. (54). Primers used to amplify cDNA inserts are listed in Table II. To express BrunoL proteins in bacteria, cDNA inserts where inserted into the pET30 expression vector (Novagen). For human BRUNOL1, mouse BrunoL-2, and mouse Bru-noL-4, inserts from an EST clone, clones HIBBM44, I.M.A.G.E.:467796, and I.M.A.G.E.:474835, respectively, were ligated into the appropriate pET30 expression vector to produce an in-frame His-tagged protein. For Xenopus BrunoL-1, the third RRM was PCR-amplified with the XBru-noL-1 RRM3 AUG primer and a vector oligo using a plasmid DNA template (40) and inserted into pET30 between the NcoI and XhoI sites. For Xenopus BrunoL-3 and human BRUNOL3, the full-length open reading frame was amplified from a plasmid template with a primer encoding the putative AUG start codon and a vector primer and inserted into pET30 between the NcoI and SmaI sites. For human BRU-NOL2, primers encoding the putative AUG start codon and an antisense primer in the 3Ј-UTR were used to amplify the cDNA from human brain cDNA (Invitrogen). The resulting PCR product was subcloned into pUC18 and sequenced to verify the identity of the clone. This cDNA was excised from pUC18 with BspLU11I and SalI and inserted into pET30 between the NcoI and XhoI sites. The expression plasmid for the Xenopus ElrC will be described elsewhere. 6 Deletion Constructs of Xenopus BrunoL-3-Using the nomenclature from the legend to Fig. 6, plasmids encoding R1-1, R1-2, R1R2, and R1R2link were made by digesting the full-length pET30/XBrunoL-3 with EcoRI, NcoI, PvuII, and StuI, respectively, further digesting with XhoI at the 3Ј-end of the insert, filling in overhanging ends with the Klenow fragment of DNA polymerase I, and religation with T4 DNA ligase. Following transformation in bacteria, the correct plasmids were selected by size and restriction enzyme pattern. For the plasmid encoding LinkR3, the original Xenopus BrunoL-3 cDNA insert was digested with NcoI and XhoI and inserted into pET30 between the NcoI and XhoI sites. For the plasmid encoding R2, the RRM2 sequences were amplified with PCR primers from the original Xenopus BrunoL-3 cDNA insert and subcloned into pGEM-T Easy according to the instructions from the supplier (Promega). This insert was further subcloned into pET30 between the NcoI and XhoI sites.
Point Mutations of Xenopus BrunoL-3-Mutations in the conserved RNP1 motif of the first two RRMs were introduced using PCR-based mutagenesis. For both the first and second RRMs, the conserved aromatic amino acids in the RNP1 motif were changed to leucine residues. For RRM1, 101 KGCCFVTF 108 was changed to 101 KGCCLVTL 108 , whereas for RRM2, 190 RGCAFVTF 197 was changed to 190 RGCALVTL 197 . The entire open reading frame for Xenopus BrunoL-3 was amplified with a PCR primer encoding the putative AUG codon (XBrunoL-3 AUG) and a T7 vector primer from the original cDNA insert 6 and inserted into pGEM-T Easy. This parent plasmid was mutagenized using synthetic oligonucleotides using the ExSite mutatgenesis kit according to the supplier's instructions (Stratagene). The subsequent insert was sequenced to confirm the mutation before inserting into the BamHI site of pGEX3X. To make mutations in both the first and second RRMs, a plasmid with a mutation in the first RRM was subjected to a second round of mutagenesis to change the second RRM.
UV-Cross-linking Assay-The UV-cross-linking assay was performed as described by Webster et al. (41). The plasmids p116 (BRE) and p120 (BREM) (Phillipa Webster, Department of Genetics, University of Washington) were linearized with BglII and transcribed with T7 RNA polymerase in the presence of [ 32 P]UTP (Amersham Pharmacia Biotech) to make radiolabeled RNA probe. Reaction mixes (10 l) containing bacterial proteins, 1 mg/ml yeast tRNA, and 1 mg/ml heparin were assembled in microtiter plates. After incubation for 5 min at 24°C 1 l of probe RNA was added and the incubation continued for another 10 min. The samples were cross-linked using a Gene Linker UV-crosslinking apparatus (Bio-Rad) for 15 min at 4°C. The reaction mixture was treated with RNase A (10 g; Sigma) and incubated at 37°C for 15 min, and then the digestion was stopped by the addition of SDS sample buffer. Following RNase A treatment, samples were electrophoresed on a 10% SDS-polyacylamide gel. For competition assays, the unlabeled RNA was synthesized from the same template, quantified by UV absorbance, and added to the reaction for 5 min prior to the addition of the The underline indicates the start codon used for the protein coding region, whereas the lower case letters indicate nucleotides that are changed to introduce a mutation into the protein sequence.
The Bruno Family of RNA-binding Proteins radiolabeled probe. The amount of cold probe excess over radiolabeled probe is indicated in the legend to Fig. 6.
Western Blots-Equal amounts of total soluble protein from induced bacterial extracts were loaded onto each lane and electrophoresed on 10% SDS-polyacrylamide gels. Immunoblots were transferred using a semidry procedure and processed as described by Harlow and Lane (55). To normalize for different induced protein levels, the relative levels of the fusion proteins in the extract were determined by immunoblotting with an antibody to the poly-His tag (Sigma) used at a 1:1000 dilution. Similar levels of His-tagged proteins were displayed on a second filter and probed with the 3B1 antibody used at 1:1000 (42). These blots were developed with alkaline phosphatase-conjugated secondary antibodies and nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl-phosphate color reaction.
Yeast Three-hybrid Assay-General methods for yeast culture, transformation, and manipulation were as described by Rose et al. (56). The yeast three-hybrid assay was performed as described by Zhang et al. (57). The Xenopus BrunoL-3 cDNA was inserted into the activation domain vector pACTII at the NcoI and SmaI sites using enzyme sites engineered into the BrunoL-3 cDNA by PCR (Table II). RNA expression plasmids with the MS2 recognition element and the BRE or BREM were constructed by insertion of synthetic oligonucleotides that encode two copies of either BRE or BREM into SmaI-digested pIII-MS2-1 (Table II). The yeast strain L40 coat was transformed by the lithium acetate procedure with the pACTII/BrunoL-3 plasmid plus the pIII-MS2-1 RNA expression plasmid and the yeast cells containing the two plasmids selected on minimal medium agar plates that lack uracil and leucine. The interaction between the RNA and BrunoL-3 was tested by a filter assay for ␤-galactosidase and confirmed by a soluble assay for ␤-galactosidase from yeast extracts and growth on medium lacking histidine (57). The ␤-galactosidase activity was normalized to the total protein in the extract as determined using the BCA assay (Pierce).
Yeast cells expressing the full-length Xenopus BrunoL-3 protein as a fusion with the activation domain in the pACTII plasmid grew slowly and produced small colonies after transformation. With time, larger colonies arose that had the same activity in the ␤-galactosidase filter assay as the small colonies. Cells from one of these large colonies produced a truncated Bruno protein as indicated by immunoblotting with a monoclonal antibody to the hemagglutinin-epitope tag encoded by pACTII (12CA5 antibody; Roche Molecular Biochemicals). The plasmid encoding this BrunoL-3 variant was rescued from the yeast strain and transformed back into bacteria, and the insert was sequenced to identify the mutation. This BrunoL-3 variant has a nonsense mutation that results in termination after the aspartic acid residue at amino acid 466. The three-hybrid experiment described in the paper is performed with this naturally arising mutant of BrunoL-3.

RESULTS
Bruno-like Genes-The Xenopus brunol-1 (etr-1) gene was described as a marker of the embryonic nervous system and encodes a putative RNA-binding protein (40,58). When originally identified, this cDNA was most similar to the Elav family of RBPs, although this similarity was primarily because of highly conserved residues in the RRMs. Recently, the Drosophila Bruno protein, a protein that binds to the 3Ј-UTR of oskar mRNA and regulates its translation, was identified and shares a substantial similarity with Xenopus BrunoL-1 (41). These proteins share a similar domain structure with an aminoterminal domain followed by two consecutive RRMs, a linker region, and finally a third RRM. We refer to this new gene family as the Bruno family to avoid confusion with the Elav family of genes (see "Experimental Procedures"). In accordance with the rules for naming human genes, the human members are called Bruno-like genes (BRUNOL) with numbers to distinguish between different family members. Other names that have been assigned to members of this family when they have been isolated using various approaches are described below.
Human Genes-Searching the human EST sequence data base with the Xenopus BrunoL-1 sequence identified six distinct contigs that are derived from six different genes (Table  III). Representative human ESTs for some of these contigs were sequenced to determine the putative encoded proteins. Two full-length Bruno-related proteins, BRUNOL2 and BRU-NOL3 (Fig. 1) were identified which are 80% identical over the entire length of the protein. In particular, the sequences of the RRMs share over 92% identity, strongly suggesting that these two proteins bind to the same targets. For BRUNOL2, two cDNAs were sequenced that differed by 12 base pairs within the linker region of the protein. This insertion results in inclusion of four amino acids, LYLQ, after alanine 229 (Fig. 1). Comparison of the cDNA sequence to genomic sequence shows that this difference is because of the alternative use of different 3Ј-splice sites. 5 Both of these BRUNOL2 isoforms are encoded in EST sequences from mouse. 5 The human BRUNOL2 protein was previously identified as a protein that binds to CUG repeats in certain mRNAs and was named the CUG-binding protein 1 (CUGBP1) (42). The sequence for the BRUNOL2 cDNA differs from the CUGBP1 cDNA in the 3Ј-UTR region. Analysis of genomic sequence shows that this difference is because of alternative splicing such that the BRUNOL2 3Ј-UTR sequence from this paper contains an unspliced exon, whereas the CUGBP1 3Ј-UTR contains a downstream alternative exon. 5 Given that the BRUNOL2 (CUGBP1) protein is implicated in the etiology of a triplet repeat expansion disease, myotonic dystrophy, the conservation of sequence with BRU-NOL3 suggests that both proteins may be involved.
The cDNAs from other EST contigs encoded only partial copies of predicted Bruno proteins. The partial sequence of the predicted human BRUNOL1 protein shares 91% identity with the Xenopus BrunoL-1(Etr-1), 5 strongly suggesting that the corresponding genes represent orthologous genes from the two species. An additional cDNA encoding human BRUNOL1 was identified in a screen for brain cDNAs with CAG repeats (44). The BRUNOL1 cDNA encodes CAG repeats in the open reading frame that are translated into glutamine residues in the corresponding protein. Other partial sequences for three other proteins, BRUNOL4, BRUNOL5, and BRUNOL6, were identified. Recently, a search of the human HTGS data base allowed the identification of genomic sequences corresponding to these cDNAs. From this genomic sequence, the predicted protein sequences for BRUNOL4, BRUNOL5, and BRUNOL6 were determined (Fig. 1). ESTs corresponding to all six of these human BRUNOL genes have also been identified in the rat and mouse EST sequencing projects. 6 This conservation between mammalian species, along with the observation that these genes have similar exons, 5 suggests that these BRUNOL genes are functional and not pseudogenes. All human Bruno genes mapped to different human chromosomes (Table III).
Multiple isoforms of the BRUNOL4 protein are encoded by EST sequences in the data base. In particular, two forms of the protein are commonly seen that lack part of the third RRM.
Comparison to the genomic sequences shows that one form lacks an exon that encodes the first half of the third RRM (Fig.  1B, Isoform B), whereas the other form arises from the use of an alternative 5Ј-splice site to make a miniexon that removes a portion of this RRM (Fig. 1B, Isoform B). The resulting alternatively spliced mRNAs encode proteins with deletions of 48 and 20 amino acids, respectively, that remove essential sequences for RNA binding by the RRM. The possible function for these isoforms is not known. Bruno Proteins in Other Species-A C. elegans EST corresponding to the predicted gene T01D1.2 from the genome project was sequenced and encodes a full-length Bruno protein (Fig. 1). This gene has been called etr-1 according to the previous nomenclature and encodes a protein necessary for function of muscles (59). A further search of the C. elegans genomic sequence revealed an additional Bruno protein encoded by the predicted gene C17D12.2. Nothing yet is known about this gene.
An extensive search of the GenBank TM data base revealed several other genes encoding predicted proteins related to Bruno. In Drosophila, a gene encoding a protein related to Bruno is immediately adjacent to the arrest gene, which encodes the Bruno protein, on chromosome 2L. This Drosophila bruno-2 gene is conserved in sequence and exon structure with the arrest gene; however, it is not known whether this gene is expressed or what the phenotype of a mutant is. In addition, additional exons can be identified that encode a third Drosophila Bruno protein (Bruno-3) on chromosome 3L most similar to the Xenopus BrunoL-1. From the ascidian H. roretzi, a single EST has been identified that encodes the first two RRMs of a Bruno-like protein that shares 64% identity with the human BRUNOL3 protein. In Arabidopsis, the FCA gene, which controls flowering time, has been shown to encode a protein with two RRMs with sequence similarity to the Xenopus BrunoL-1 protein (39% identity over the first two RRMs of X BrunoL-1) (23). In addition, the Arabidopsis genome sequencing project has identified two related genes that encode Bruno-like proteins that share 41 and 40% sequence identity with the Xenopus BrunoL-1 protein (Fig. 2). Thus, members of the Bruno family are widely dispersed over different phyla.
Sequence Conservation of Bruno Proteins-Alignment of the putative human and C. elegans full-length Bruno family of proteins to the Drosophila Bruno and Xenopus BrunoL-1 is presented in Fig. 1. Most of the sequence conservation is within the RRMs; the amino-terminal and linker regions are highly variable between different proteins (except as described below). Of the three RRMs, the third RRM has the most sequence conservation between the different proteins. The sequence of the linker region contains no identifiable motifs. However, several of the Bruno proteins have homopolymeric amino acid The Bruno Family of Proteins-All members of the Bruno family share a common domain structure with the Elav family of proteins. A dendrogram based on a pairwise comparison of all members of the Bruno and vertebrate Elav family is presented in Fig. 2. The Elav proteins form a distinct group of proteins. A yeast protein, PUB1, has a domain structure similar to Bruno but is most similar to the Elav family of proteins. The vertebrate Bruno proteins fall into two subfamilies, one containing BRUNOL2 and -3 and the other containing BRU-NOL1, -4, -5, and -6. Comparison of the amino-terminal and linker regions of proteins within subfamilies reveals some sequence similarity in these regions. Orthologous proteins in Xenopus have been identified for BRUNOL1, -2, and -3; presumably the frog BrunoL-4, -5, and - 6 have not yet been identified. One of the C. elegans proteins, C17D12.2, is most similar to the BRUNOL1, -4, -5, and -6, whereas the other, Etr-1, is distantly related to BRUNOL2 and -3. The similarity of the domain structure but divergence of the primary sequence from the Elav proteins, proteins that are involved in many different RNA processing events both during development and in normal cell function (28,32), suggests that the Bruno proteins may have related roles in regulating target RNAs but bind to different subsets of mRNAs than the Elav-like proteins.
The 3B1 Antibody Detects All Bruno Proteins-The monoclonal antibody, 3B1, developed against the human BRUNOL2 (also called Nab50; CUGBP1), recognizes a nuclear protein found in many tissue culture cell types (42). Because Bruno proteins are highly conserved, we tested whether the 3B1 antibody recognizes different Bruno family members by immunoblotting. Different Bruno proteins were expressed in bacteria as His-tagged fusion proteins. The 3B1 antibody recognized all of these fusion proteins (Fig. 3). Because some of the partial cDNAs only encoded the C termini of the proteins, the 3B1 epitope must map to this end of the protein, probably to the third RRM, the most highly conserved region of the Bruno proteins. Deletion mapping of the third RRM of the Xenopus BrunoL-3 protein demonstrates that the 3B1 epitope is encoded within the C-terminal 30 amino acids of the protein. 5 Expression of Bruno Genes-Xenopus brunol-1 and Drosophila bruno (arrest) are expressed in specific tissues, the nervous system and germ cells, respectively. We examined the expression of BRUNOL2 and BRUNOL3 in adult human tissues by Northern blots (Fig. 4). The BRUNOL2 gene is expressed as three abundant mRNA species in all tissues tested. However, the relative levels of the different mRNA species vary with tissue. The identity of these different mRNA species is not known. However, because two BRUNOL2 cDNAs have been described that differ in their 3Ј-UTR, these RNAs likely have different lengths of 3Ј-UTR. BRUNOL3 encodes multiple mRNA species and is predominantly expressed in heart, brain, and muscle, with lower levels of expression in the pancreas, lung, and placenta. The different tissues express unique BRU-NOL3 mRNA species. The brain expresses only one very large mRNA species (Fig. 4, band A), the heart and muscle expresses two RNA species in common (Fig. 4, bands B and C), whereas the heart exclusively expresses an additional mRNA species (Fig. 4, band D). The sequence differences between these mRNA species are not known. The other members of the human Bruno family could not be specifically detected by Northern blots. 6 However, the ESTs for these genes have only been isolated from libraries made from brain, retina, or embryonic mRNA (with the exception of some cDNAs clones isolated from cancer cell line libraries).
The Human BRUNOL2 and BRUNOL3 Bind RNA-The Bruno protein in Drosophila binds to a BRE in the oskar mRNA and acts as a repressor of oskar translation. Given the conservation of sequence between Bruno and the vertebrate homologs, we asked whether the human BRUNOL2 and BRU-NOL3 also bind to a BRE. Using a UV-cross-linking assay, GST fusion proteins of both human BRUNOL2 and BRUNOL3 bind to an RNA containing a BRE but not to one containing a mutated form of the BRE (BREM) (Fig. 5B). The binding to the BRE was sensitive to competition with an excess of RNA containing a BRE but not a mutated form of the BRE (BREM, Fig.  5C). Similar results were obtained using an RNA gel mobility shift assay. 5 Consistent with the Bruno proteins binding to BRE elements, the Xenopus EDEN-BP (BrunoL-2) also has been shown to bind to an RNA element that is identical to the consensus BRE (26,60).
By analogy with the Elav proteins (61-63), we predict that the binding site in Bruno proteins for the BRE will be the amino-terminal half of the protein containing both the first and second RRMs. To test this hypothesis, various deletions of the Xenopus BrunoL-3 protein were constructed, the corresponding proteins were expressed in bacteria, and these proteins were tested for binding to RNA containing a BRE using the UVcross-linking assay (Fig. 6, A and B). We choose the frog Bru-  noL-3 protein, because it is highly conserved with corresponding human BRUNOL3, it binds an RNA containing a BRE, 6 and it has convenient restriction enzyme sites to manipulate the insert. As predicted, only proteins containing the first two RRMs intact bound to RNA containing the BRE. To further define the functional RNA-binding domain, we constructed mutants of the frog BrunoL-3 with point mutations in conserved aromatic residues in the RNP1 motif. The RNP1 sequences for Xenopus BrunoL-3, which are identical in the human BRU-NOL3 sequence, for the first and second RRMs are 101 KGC-CFVTF 108 and 190 RGCAFVTF 197 , respectively. The amino acid corresponding to the underlined phenylalanine has been shown to directly intereact with RNA for other RRM-containing proteins (7,64). To mutate this sequence, both phenylalanines in these sequences were changed to leucine. When expressed in bacteria as GST fusion proteins, mutations in the first RRM separately had little effect on binding to the BRE while mutations in the second RRM reduced binding. When mutations in the first and second RRM were combined, significant reduction in RNA binding was detected (Fig. 6C). Thus, both RRMs contribute to binding the BRE confirming the deletion analysis.
Next, we tested whether the Bruno proteins could bind to RNA containing a BRE in vivo using the yeast three-hybrid assay to detect RNA-protein interactions (65). The Xenopus BrunoL-3 protein was inserted into a yeast activation domain plasmid and potential RNA targets were inserted into a yeast RNA expression plasmid. If an interaction occurs between Bru-noL-3 and the RNA target, LacZ will be expressed in the yeast three-hybrid host cells containing both plasmids. Cells expressing the activation domain/BrunoL-3 protein and RNA containing the BRE target accumulated 15-fold more LacZ than cells expressing activation domain/BrunoL-3 protein and RNA containing a mutated form of the BRE or the activation domain alone and RNA containing a BRE. The expression of LacZ was 3-fold lower than for the positive control, an interaction between the iron response protein and the iron response element. DISCUSSION A Conserved Gene Family-The Bruno family of genes is conserved through evolution with members present in plants, worms, fruit flies, and vertebrates. The general domain structure of these proteins also is present in the Elav family although sequence identity is only seen in highly conserved residues of the RRM. A multiple sequence alignment identifies two subfamilies with unique patterns of gene expression for different genes. Because tissue-specific Bruno proteins are present in multicellular organisms, the Bruno proteins will be involved in regulating or maintaining cell differentiation. This prediction is clearly true for the Drosophila Bruno protein (encoded by the arrest gene), which regulates body axis forma- whereas ElrC is an Elav family member that binds both the BRE and mutated BRE probe, and Uninduced is a protein sample prepared without induction. The migration of molecular weight markers is indicated at the right, whereas the band labeled X on the left is a bacteria protein that cross-links to most RNA probes tested. The migration for the full-length bacterial proteins, as determined by immunoblots with a His tag antibody, for each construct is indicated with the dot at the right side of the lane. C, point mutations were introduced into the conserved RNP1 motif of the first two RRMs for the Xenopus BrunoL-3 protein that should inactivate the RNA-binding activity of the RRM. The corresponding proteins were expressed as fusions with GST and tested for binding the BRE by UV cross-linking. The lanes are labeled with the input protein extract where Ϫ is extract from bacteria expressing only GST, whereas XB3-1 and XB3-2 are two different preparations of wild type BrunoL-3. XB3 pmR1, XB3-pmR2, and XB3-pmR12 are mutations in the first RRM, the second RRM, or both RRMs. The positions of BrunoL-3 and a nonspecific bacterial RNA-binding protein (X) are labeled at the right. tion by controlling oskar expression and has a role in oogenesis (41). C. elegans etr-1 functions in muscle development because inactivation of etr-1 with RNA interference results in a phenotype similar to that seen in worms with defects in muscle function (59). The zebrafish brul gene (brunol-2) is expressed in the vegetal pole of early oocytes and embryos suggesting a possible role in embryonic patterning (52). Finally, the Xenopus EDEN-BP (BrunoL-2) binds to an RNA element that specifies deadenylation and subsequent translational silencing of mRNAs that are no longer needed during early cleavage stages of development (60).
Intriguingly, the human BRUNOL3 was identified as encoded by a gene induced during apoptosis of neuroblastoma cells and is called NAPOR (46). Consistent with a role in apoptosis, expression of mouse brunoL-3 during development is predominantly in the nervous system where it appears to colocalize with areas of apoptosis (66). Other RNA-binding proteins have been identified as possible effectors of apoptosis in lymphocytes (67).
The mouse homolog of BRUNOL3 was isolated by homology to CUGBP1 (BRUNOL2) (45) or NAPOR (66). In the adult mouse, brunoL-3 is expressed in all tissues, although abundant expression was detected in brain, lung, and skeletal muscle. As expected from the sequence similarity to the CUGBP1 (Bru-noL2), the mouse BrunoL-3 bound to CUG repeats in various mRNAs.
Members of the Bruno family are alternatively spliced to produce multiple protein isoforms. One example of these isoforms is the identification of two different BRUNOL2 cDNAs that encode proteins that differ by insertion/deletion of a 4-amino acid sequence in the linker region. A similar insertion/ deletion of a 6-amino acid sequence occurs in a different part of the linker region for human BRUNOL3 (NAPOR) proteins (46). Other examples include the identification of three separate cDNA that encode BRUNOL3 (NAPOR) that differ in the sequence at the 5Ј-end of the insert. These various cDNAs result in alternative amino-terminal domains (46). Comparison of these cDNA sequences to genomic sequence suggests that these alternative mRNAs are produced by alternative promoters that encode unique 5Ј-exons. Similar alternative 5Ј-end sequences are also seen for the Xenopus BrunoL-3 gene. 6 The functional significance of these different isoforms is not known. However, because these sequences are outside of the RNA-binding domains present in the first two RRMs, we propose that these different isoforms alter the function of the Bruno proteins either by changing a protein interaction domain or by altering another attribute of the protein such as cellular localization. Finally, the human BRUNOL4 protein has multiple isoforms that have deletions in the third RRM that may alter the RNA binding function of this domain (Fig. 1).
Bruno Proteins in Myotonic Muscular Dystrophy-Members of the Bruno family may have a role in the etiology of myotonic dystrophy (myotonia dystropha (DM)), a dominant genetic disorder with defects in multiple organ systems. The defect in this disease has been mapped to a gene, the myotonic dystrophy protein kinase (DMPK), where the defect is a triplet repeat expansion (CTG) in the 3Ј-UTR. In contrast to other triplet repeat expansion disorders, where the expansion in the coding region or 5Ј-UTR results in either a defective protein or inactivation of the promoter, the exact effect of the triplet repeat expansion in DM is not yet known (68). One of many hypotheses to explain the cause of this disease, the RNA interference hypothesis, is that the expanded CTG repeats are expressed in the DMPK mRNA and act to interfere in cis with expression of the DMPK protein and in trans with the expression of other genes. These repeats, expressed as CUG repeats in the mRNA, presumably bind an RBP whose function is important for the expression of DMPK and other genes. Clearly, the expression of the mutant DMPK allele mRNA is reduced in muscle cells from DM patients, consistent with this model (69 -72). Furthermore, expression of CUG repeats in a muscle cell line blocks expression in cis of a reporter gene (73) and in trans of genes needed for muscle differentiation (73,74). In myoblasts, overexpression of CUG repeats also results in altered splicing of the cardiac troponin T pre-mRNA, a prediction from the observation of CUG repeats in a splicing enhancer for the corresponding pre-mRNA (75). BRUNOL2 (CUGBP1) was identified as a candidate RBP for binding to the CUG repeats in the DMPK mRNA (42). The monoclonal antibody 3B1 identified this protein as being an hnRNP that binds mRNA in the nucleus. Interestingly, BRU-NOL2 (CUGBP1) is phosphorylated by DMPK, and this phosphorylation appears to regulate the nuclear localization of the protein (76). CUGBP1 regulates the alternative splicing of the cardiac troponin T pre-mRNA by interacting with the splicing enhancer containing CUG repeats (75). The BRUNOL3 protein shares extensive sequence similarity to CUGBP1 (BRUNOL2) (Fig. 1). As expected, both BRUNOL2 (CUGBP1) and BRU-NOL3 both can bind to the same RNA sequence, either the BRE (Fig. 5) or CUG repeats (42,45). Relevant to DM, BRUNOL3 is preferentially expressed in muscle, heart, and brain (three organ systems affected in DM patients), whereas BRUNOL2 (CUGBP1) is ubiquitously expressed. Thus, both BRUNOL2 (CUGBP1) and BRUNOL3 are RBPs that bind CUG repeats and may play a role in the etiology of DM.
We propose that the BRUNOL2 and -3 proteins act as mRNA-shuttling proteins to regulate the cytoplasmic accumulation of mRNAs containing the target sequence, either a CUG repeat or a BRE. Furthermore, the defect in DM is in part because of the expanded CUG repeats interfering with the mRNA shuttling function of the Bruno proteins. This RNA interference hypothesis includes a cis defect, the shuttling of DMPK mRNA with expanded repeats, and a trans defect, the shuttling of other mRNAs that are bound by Bruno proteins. Consistent with this hypothesis, the DMPK mRNA with expanded CUG repeats accumulates in the nucleus in foci and does not accumulate in the cytoplasm to the same levels as mRNA without expanded repeats (70,71,77). The mRNA retained in the nucleus is properly spliced and polyadenylated (70,72). Thus, DM patient cells have a defect in the transport of mature mRNA to the cytoplasm. The similarity of the Bruno proteins with the Elav proteins in the domain structure suggests these proteins may share similar functions. The human HuR (Xenopus ElrA) protein, an Elav family member, is an mRNA-shuttling protein that binds to mRNAs with AU-rich sequence elements (78 -80). By analogy to the HuR proteins, BRUNOL2 and -3 are potential mRNA shuttling proteins, and defects in mRNA shuttling caused by expression of CUG repeats in trans may block shuttling of other mRNAs and result in the symptoms observed in DM patients. Consistent with this idea, BRUNOL2 (CUGBP1) was isolated as a protein that interacts in a yeast two-hybrid assay with the yeast Nab2 protein, an essential mRNA-binding protein with a role in nucleocytoplasmic mRNA transport (81). In mammalian cells, BRUNOL2 (CUGBP1) was originally defined as an hnRNP protein, and several hnRNP proteins function as mRNA-shuttling proteins (82). Additional experiments are needed to test this hypothesis about Bruno protein function as an mRNA shuttling protein and to further examine its role in DM.
Model for Bruno Protein Function-Like many RBPs, a simple model for the function of the Bruno proteins involves the binding of a subset of mRNAs and forming RNP complexes, which regulate the expression of the corresponding gene product. However, it is not clear how these proteins can have different functions depending on the target mRNA or cell type and how different Bruno proteins that bind the same RNA target sequence might nevertheless have unique functions. To expand on this simple model, we propose that the Bruno proteins will bind to target mRNAs and interact with other proteins to determine the specificity of function. These other proteins might be RBPs that recognize additional sequence elements on the target mRNA or isoform-specific interacting proteins that modify the function of the RNP. This type of mechanism is observed in the translational silencing of hunchback by the Pumilo protein in Drosophila. In this case, Pumilio specifically binds to elements in the 3Ј-UTR of the hunchback mRNA and recruits a nonspecific RNA-binding protein, Nanos, to form a complex on the mRNA and silence its translation (83).
Consistent with this idea, recent work on the Bruno protein in Drosophila demonstrates that BRE does not function independently, and additional RNA elements in the 3Ј-UTR of the oskar mRNA are required for correct regulation (27). Presumably, these additional sequences bind other proteins that interact with Bruno. Several RNA-binding proteins have been identified that interact with Bruno. These include Apontic, encoded by a gene involved in heart and head development that binds sequences in the oskar 3Ј-UTR. Genetic experiments clearly show that at least apontic and arrest, the gene that encodes the Bruno protein, interact in the regulation of oskar translation (84). In addition, Vasa, an RNA helicase involved in regulating translation of oskar and other mRNAs during early development (41), and Squid, an hnRNP required for proper axis formation (51), also have been demonstrated to interact with the Bruno protein.
Additional RNA elements also contribute to the regulation of mRNA deadenylation by EDEN-BP (BrunoL-2) in frogs. AUrich elements either in the mos mRNA or in a synthetic RNA function to enhance the deadenylation of the corresponding mRNA (49). Similar, deletion analysis of the cdk2 mRNA demonstrates that multiple RNA elements contribute to the deadenylation (48). The conclusions from these experiments were that multiple proteins bind to the mRNA and contribute to the deadenylation reaction.