Two rat brain Staufen isoforms differentially bind RNA

Authors


Address correspondence and reprint requests to Stefan Kindler, Institute for Cell Biochemistry and Clinical Neurobiology, University of Hamburg, D-20246 Hamburg, Germany. E-mail: kindler@uke.uni-hamburg.de

Abstract

In neurones, a limited number of mRNAs is found in dendrites, including transcripts encoding the microtubule-associated protein 2 (MAP2). Recently, we identified a cis-acting dendritic targeting element (DTE) in MAP2 mRNAs. Here we used the yeast tri-hybrid system to identify potential trans-acting RNA-binding factors of the DTE. A cDNA clone was isolated that encodes a member of a mammalian protein family that is highly homologous to the Drosophila RNA-binding protein Staufen. Mammalian Staufen appears to be expressed in most tissues and brain areas. Two distinct rat brain Staufen isoforms, rStau+I6 and rStau-I6, are encoded by alternatively spliced mRNAs. Both isoforms contain four double-stranded RNA-binding domains (dsRBD). In the larger rStau+I6 isoform, six additional amino acids are inserted in the second dsRBD. Although both isoforms interacted with the MAP2-DTE and various additional RNA fragments in an in vitro north-western assay, rStau-I6 exhibited a stronger signal of bound radioactively labelled RNAs as compared with rStau+I6. Using an antibody directed against mammalian Staufen, the protein was detected in somata and dendrites of neurones of the adult rat hippocampus and cerebral cortex. Ultrastructural studies revealed that in dendrites, rat Staufen accumulates along microtubules. Thus in neurones, rat Staufen may serve to link RNAs to the dendritic microtubular cytoskeleton and may thereby regulate their subcellular localization.

Abbreviations used
3AT

3-aminotriazole

dsRBD

double-stranded RNA-binding domain

DTE

dendritic targeting element

GST

glutathione S-transferase

MAP

microtubule-associated protein

MHR

MAP1B homology region

RRE

Rev response element

UTR

untranslated region

VLE

vegetal localization element

Cell polarity is thought to depend on both selective protein sorting and a locally restricted, regulated translation of specific mRNAs in subcellular regions (Wilhelm and Vale 1993; St Johnston 1995; Bassell and Singer 1997). In neurones (Kindler et al. 1997; Kuhl and Skehel 1998; Schuman 1999; Tiedge et al. 1999; Wells et al. 2000), oligodendrocytes (Carson et al. 1998), chicken myo-and fibroblasts (Kislauskis et al. 1993; Ross et al. 1997), Drosophila embryos (St Johnston and Nüsslein-Volhard 1992; Curtis et al. 1995; Lehmann 1995; Rongo and Lehmann 1996), and Xenopus oocytes (Schwartz et al. 1992; Elisha et al. 1995; Deshler et al. 1997; Havin et al. 1998), ordered interactions between cis-acting RNA sequences and trans-acting factors seem to govern cytoplasmic transport and site-specific translation of various mRNAs. Moreover, the cytoskeleton is thought to play an important role in regulating mRNA localization, translation and stability (Bassell and Singer 1997; Nakielny et al. 1997).

In mammalian neurones, dendritic targeting of RNAs is an energy-dependent process (Davis et al. 1987) that depends on cytoskeletal components (Bassell et al. 1994; Litman et al. 1994; Knowles et al. 1996). The ultrastructural detection of polysomes in dendrites (Steward and Levy 1982; Steward and Reeves 1988) initiated the hypothesis that extrasomatic mRNAs are locally translated. This idea was subsequently supported by the fact that isolated dendrites are capable of protein synthesis (Torre and Steward 1992; Crino and Eberwine 1996). In addition, specific forms of synaptic plasticity are likely to require dendritic translation (Kang and Schuman 1996; Casadio et al. 1999). Thus, an extrasomatic synthesis of specific proteins is thought to regulate the molecular composition of individual dendritic sections and thereby modulate synaptic functions (Kindler et al. 1997; Kuhl and Skehel 1998; Schuman 1999; Tiedge et al. 1999; Wells et al. 2000).

In contrast to cytoplasmic mRNA transport events in Drosophila embryos and Xenopus oocytes, the molecular mechanisms directing dendritic mRNA localization and local protein synthesis in neurones are poorly understood. We have recently identified a 640-nucleotide cis-acting dendritic targeting element (DTE) in the 3′ untranslated region (3′ UTR) of the mRNA encoding the microtubule-associated protein 2 (MAP2). The DTE mediates extrasomatic transcript localization in neurones (Blichenberg et al. 1999). Using the MAP2-DTE as the bait in a yeast tri-hybrid screen (Putz et al. 1996; Putz et al. 2000), a rat homologue of the Drosophila protein Staufen was found to interact with the DTE. In Drosophila, the double-stranded (ds) RNA-binding protein Staufen is essential for the subcellular localization of oskar (Ephrussi et al. 1991; Kim-Ha et al. 1991; St Johnston et al. 1991) and bicoid (St Johnston et al. 1989; St Johnston et al. 1991) mRNAs to the posterior and anterior poles of oocytes, respectively, and of prospero transcripts in neuroblasts (Li et al. 1997; Broadus et al. 1998). Cytoplasmic localization of these mRNAs is a multistep active transport mechanism that involves the cytoskeleton (Erdelyi et al. 1995; Pokrywka and Stephenson, 1995; St Johnston, 1995; Tetzlaff et al. 1996; Broadus et al. 1998). In the fly, the specific localization of oskar and bicoid transcripts plays an important role in the determination of the anteroposterior axis of the embryo (St Johnston 1995). Recently, mouse and human Staufen homologues were identified (Buchner et al. 1999; Marion et al. 1999; Wickham et al. 1999). They are present in many tissues and appear to colocalize with the rough endoplasmic reticulum. Here we show that rat Staufen binds a number of RNA fragments including the MAP2-DTE. Staufen transcripts can be detected in various rat tissues as well as in most mammalian brain regions. In the rat brain, two different Staufen isoforms are synthesized from alternatively spliced mRNAs. In vitro, these isoforms differentially bind RNA. In adult rat hippocampus and cerebral cortex, Staufen isoforms are present in somata and dendrites of neurones and colocalize with dendritic microtubules.

Materials and methods

Construction of vectors for the yeast tri-hybrid system

Construction of pPGKRRE, pDBRevM10, pRevRX, pRevR2 and pRevRGAP was previously described (Putz et al. 1996; Putz et al. 2000). pRevRDTE was constructed as follows: utilizing two oligonucleotides (5′-GGAAACGCGTGATCTAGCACTAAAATATCATT-3′ and 5′-GGAAGCGGCCGCTTACTGGACCTTCTTCTTTA-3′) and the plasmid pNEu (Blichenberg et al. 1999) as a template, the DTE of the MAP2 3′UTR was amplified by PCR, digested with NotI and MluI and subcloned into the vector pRevRX. For the construction of pRevRMAP2u1586−2435, a NotI fragment from pNEu1586−2435 (Blichenberg et al. 1999) was subcloned into the NotI digested pRevRX. A 941-bp AccI/BamHI fragment comprising nucleotides 677–1617 from the rat α-tubulin cDNA (GenBank/EMBL Data Bank accession number V01227) was subcloned into the MluI site in pPGKRRE via blunt end ligation to create pRtub. To generate pRevRtub, a XhoI/SalI fragment from pRtub containing the RRE-α-tubulin transcription unit cassette was transferred into the SalI site of pDBRevM10 (pRevRtub was kindly provided by E. Mohr, Institute for Cell Biochemistry and Clinical Neurobiology, University of Hamburg, Germany). For the construction of pRevRVLE, a primer pair (5′-AGCTACGCGTATTTCTACTTTATTTCTACA-3′, 5′-GATAGCGGCCGCTCAAGTCATATGGAC-3′) and a plasmid template containing the vegetal pole protein 1 (Vg1) cDNA (Elisha et al. 1995) were used to PCR amplify a 367 nucleotide fragment (accession number M18055, nucleotides 1440–1806). Subsequently, the PCR product was digested with MluI and NotI and subcloned into pRevRX. To construct pRevRαCaMK, PCR was performed with two oligonucleotides (5′-ACGCGTGGACTTTCATCGATTCTATT-3′ and 5′-GCGGCCGCGTAGCTATTTATTCCACTGA-3′) and a cDNA template containing the entire 3′ UTR of the rat αCaMKII mRNA (accession number AF237778). The PCR product was digested with MluI and NotI and subcloned into pRevRX. pACTII.73 was isolated from the library and contains the rat Staufen cDNA from nucleotide 671–3077 (accession number AF227200).

Tri-hybrid screen of a rat brain cDNA library

The bait plasmid pRevRDTE was used to screen a rat brain Matchmaker cDNA library (Clontech Laboratories, Palo Alto, CA, USA; product # RL4005AH) as described previously (Putz et al. 1996; Putz et al. 2000). In more detail, a yeast clone containing the bait plasmid was grown in 100 mL synthetic yeast medium lacking tryptophan at 30°C for 24 h. Subsequently, 10 mL of this culture were used to inoculate 250 ml of the same medium and cells were grown at 30°C for 13–15 h to reach an OD600 of 1. After the addition of 250 ml YPAD (50 g/L YPD powder; Bio101, Vista, CA; 40 mg/L adenine sulfate), cells were grown for 4 h and spun at 2000 × g for 5 min at room temperature. The pellet was washed with 20 mL sterile H2O, spun down and resuspended in 20 mL solution III (800 mm sorbitol, 100 mm lithium acetate, 10 mm Tris-HCl, 1 mm EDTA; pH 7.5). The cells were agitated for 30 min at 30°C and 5 g, pelleted and resuspended in 1.5 mL solution III. Subsequently, 1.5 mL DNA mix (500 µg rat brain cDNA library and 500 µg salmon sperm carrier DNA in 100 mm lithium acetate, 10 mm Tris-HCl, 1 mm EDTA; pH 7.5) and 13.1 mL solution II (40% polyethylene glycol, 100 mm lithium acetate, 10 mm Tris-HCl, 1 mm EDTA; pH 7.5) were added, and cells were mixed as before. After the addition of 1.4 mL dimethyl sulfoxide the suspension was incubated for 20 min at 42°C. Synthetic yeast medium lacking tryptophan, leucine and histidine (100 mL) was added, and cells were agitated for 2 h at 30°C and 30 g. Yeast cells were spun down, resuspended in 4–5 ml TE buffer (10 mm Tris-HCl, 1 mm EDTA; pH 7.5) and plated onto medium with 10 mm 3-aminotriazole (3AT) lacking tryptophan, leucine and histidine. For the determination of the number of transformants 2 µL and 10 µL of the suspension were plated onto medium lacking tryptophan and leucine.

Northern blots

Total RNA from rat brain was prepared with RNazol (Angewandte Gentechnologie Systeme GmbH, Heidelberg, Germany), separated on a formaldehyde gel, and blotted onto Hybond N membranes (Sambrook et al. 1989). The Prime it II kit (Stratagene GmbH, Heidelberg, Germany) was used to 32P-label cDNA fragments from pACTII.73 and pAS2.tub. Multiple tissue northern blots purchased from Clontech were hybridized with rStau and actin cDNAs according to the protocol of the manufacturer.

Polymerase chain reaction and 5′-end specific amplification

To isolate 5′-cDNA rStau fragments, PCR was performed with a rStau specific primer (5′-GTCTGCTAATAGGATTCATCCC-3′), the pACTII sequencing primer and 100 ng of a rat brain cDNA library (Clontech Laboratories, product # RL4005AH) as template. The PCR product was subcloned into pGEM (Promega, Madison, WI, USA) and sequenced. To investigate the existence of rStau mRNAs that are alternatively spliced in their 5′ regions, total RNA from P5, P7, P9 and P13 rat cerebral cortex and hippocampus as well as RNA from embryonic and adult total rat brain was reverse transcribed using superscript II reserve transcriptase (Gibco BRL, Life Technologies, Eggenstein, Germany). Using the single-stranded DNA as a template, PCR was performed with taq polymerase and two oligonucleotide primers Stau-s (5′-CTTCTCCGCAGCCACTGCCGTCTC-3′) and Stau-as (5′-CATTAAACTG- CTGTCCTCCCAC-3′). For the detection of rStau + I6 and rStau-I6 mRNAs, RT-PCR was carried out with two oligonucleotides 5-I6 (5′-TTGAGATTGCGCTGAAGCGG-3′) and 3-I6 (5′-CATGTGTGGCGGGCCACTCTC-3′) flanking the insert.

Construction of fusion proteins and Western blots

For the construction of prokaryotic vectors that express GST-rStau + I6 and GST-rStau-I6 two oligonucleotides (5′-CTCGGATCCATGTATAAGCCCGTGGAC-3′ and 5′-TCTGAATTC TCAGCACCTCCCGCACGCTG-3′) and cDNA that was reverse transcribed from total RNA of 5-day-old rat hippocampi was used to amplify the rStau coding region. The PCR products were digested with BamHI and EcoRI and subcloned into pGEX2T (Gibco BRL, Life Technologies). Construction of the vector expressing GST-SAP has been described previously (Müller et al. 1996). In XL1 blue cells, expression of fusion proteins was induced with 1 mm isopropyl-β-d-thiogalactopyranoside for 3–5 h at 37°C, cells were boiled for 5 min in Lämmli buffer, proteins were separated by SDS-PAGE (Sambrook et al. 1989) and transferred onto polyvinylidine fluoride membranes (Millipore Corporation, Bedford, MA, USA). For immunodetection, membranes were blocked for 1 h at room temperature with 1% bovine serum albumin in Tris-buffered saline (pH 7.4). GST-rStau isoforms and GST-SAP were detected with an anti-hStau antibody (Wickham et al. 1999) or anti-GST antibody (Müller et al. 1996). Blots were analysed with the ECL-System (Amersham Pharmacia Biotech, Uppsala, Sweden).

North-western assay

North-western assays were performed as described (Schumacher et al. 1995). Briefly, GST fusion proteins (1 µg each) were separated by SDS-PAGE and transferred onto polyvinylidine fluoride membranes (Millipore Corporation). Membranes were blocked for 1 h at room temperature with 5% milk powder in PBS (5 mm sodium phosphate buffer pH 7.4, 2.7 mm KCl, 137 mm NaCl) containing 1% Tween 20. For protein denaturation–renaturation membranes were incubated twice at 4°C in binding buffer (25 mm Hepes, pH 7.9, 25 mm NaCl, 5 mm MgCl2, 0.5 mm DTT) with 6 m guanidine hydrochloride for 5 min each. Subsequently, filters were incubated at 4°C in binding puffer containing decreasing concentrations of guanidine hydrochloride (3 m, 1.5 m, 0.75 m, and 0. 375 m, respectively) for 5 min each. Membranes were washed twice in binding buffer for 5 min each at room temperature and incubated with radioactively labelled RNA probes at a final concentration of 5 × 105 cpm in 5 mL binding buffer for 1 h at room temperature. Blots were washed five times with binding buffer as described above and exposed to an X-ray film at −70°C for several days. The radioactive RNA probes were in vitro transcribed with T3 RNA polymerase in the presence of [α32P]UTP using linearized pBluescript-SKII(–) plasmid templates containing the rat MAP2-DTE (GenBank/EMBL Data Bank accession number U30398, nucleotides 2432–3071), the rat α-tubulin cDNA (accession number V01227, nucleotides 1–1617), or the cDNA of the vegetal localization element (VLE) of the Xenopus Vg1 mRNA (accession number M18055, nucleotides 1440–1806).

Immunohistochemistry and electron microscopic investigation of rat brain sections

Adult Sprague–Dawley rats were perfused with 4% paraformaldehyde and 0.1% glutaraldehyde. Brains were removed and incubated in the perfusion fixative overnight. Twenty−35 µm vibratome sections were cut and washed four times in phosphate-buffered saline (PBS; 150 mm NaCl, 20 mm sodium phosphate; pH 7.4), incubated in a series of ethanol (10%, 20%, 40%, 20%, and 10%; 5 min each), washed in PBS, incubated in 0.3% H2O2 for 5 min, washed in PBS, and blocked with 10% (v/v) horse serum (HS), 0.2% (w/v) bovine serum albumin in PBS for 60 min. For immunostaining, sections were incubated overnight with a 1 : 200 dilution of the polyclonal anti-hStau antibody (Wickham et al. 1999) at 4°C in 0.1% (v/v) HS in PBS. After four washes in PBS, sections were incubated for 90 min at room temperature with the Avidin–Biotin Elite System (Vector Laboratories, Burlingame, CA, USA). Sections were washed once in PBS and three times in 50 mm Tris-HCl (pH 7.6). The final visualization of the bound antibody was accomplished by processing the sections in 50 mm Tris-HCl (pH 7.6), containing 0.05% (v/v) 3,3′-diaminobenzidine and 0.01% (v/v) hydrogen peroxide (H2O2) (SIGMA FAST; Sigma-Aldrich, Steinheim, Germany) for 10 min. The colour reaction was stopped by washing the sections three times in 50 mm Tris-HCl (pH 7.6). For light-microscopy sections were transferred onto glass slides, dried at room temperature overnight, dehydrated in an alcohol series (30%, 60%, 80%, 96%, and 100%; v/v) and xylol, and mounted with Entellan (Merck, Darmstadt, Germany). Pictures were taken with an Axiophot microscope (Zeiss, Oberkochen, Germany) equipped with a video camera. For electron microscopy, sections were postfixed in 1% osmium tetroxide in PBS for 30 min on ice. After several washes in PBS, the sections were dehydrated and mounted on slides with Epon resin between acetate foils and cured for 48 h at 60°C. Hippocampal regions containing labelled neurones were selected under the light microscope and mounted on Epon blocks. Ultrathin sections (60 nm) were cut with a Reichert-Jung ultramicrotome (Leica, Stuttgart, Germany), collected on copper grids, counterstained with 2.5% uranylacetate and lead citrate and examined with a Zeiss 902 electron microscope (LEO, Oberkochen, Germany).

Nucleotide sequence GenBank accession numbers

The rStau cDNA sequences were deposited in the GenBank database under accession numbers AF227200 and AF290989.

Results

Identification and structural characterization of distinct rat brain Staufen isoforms

We recently characterized a cis-acting DTE in MAP2 transcripts that mediates extrasomatic mRNA localization in neurones. Here we employed the yeast tri-hybrid system to identify putative trans-acting proteins of the DTE. By screening 1.5 × 106 clones of an adult rat brain cDNA library with the MAP2-DTE bait, 18 positive clones were identified and sequenced. Five clones contain two distinct previously unknown sequences. Ten cDNAs encode a splice factor, and one clone each encodes the transcription factor JunB and a member of the kinesin family of motor proteins. One additional positive clone encodes the carboxy-terminal part of a novel member of a mammalian protein family that is highly homologous to the Drosophila RNA-binding protein Staufen (St Johnston et al. 1991; Marion et al. 1999; Wickham et al. 1999; Roegiers and Jan 2000). To determine cDNA sequences encoding the missing amino-terminal segment of the protein, a 5′-end specific amplification protocol with total adult rat brain cDNA was performed. Amplified cDNA products contained a consensus initiation site for translation (Kozak 1987) with upstream stop codons located in all three reading frames and overlapped with the 5′ part of the previously isolated cDNA clone. The entire characterized cDNA sequence comprises 3059 nucleotides. A 1488 nucleotide open reading frame encodes a 54.8-kDa protein (rStau+I6) containing two short and two full-length double-stranded RNA-binding domains (dsRBDs) (Fig. 1). A second group of cDNAs generated by RT-PCR using two oligonucleotides flanking the coding region and hippocampal RNA of 5-day-old rats encodes a 54.2-kDa isoform, referred to as rStau-I6. It lacks a six amino acid insert (I6) located in the second dsRBD (Fig. 1). This indicates that mRNAs encoding two different rStau isoforms are created via alternative splicing. Both isoforms are collectively referred to as rStau. Performing RT-PCR with two oligonucleotides that flank the alternative splicing site, amplified products specific for both rStau+I6 and rStau-I6 mRNAs were detected in reactions with total RNA from rat cerebral cortex and hippocampus at different postnatal days (P5, P7, P9 and P13; Fig. 2) indicating that both rStau isoforms are expressed in postnatal brain. In human tissues, four alternatively spliced Staufen transcripts were described that differ in their 5′ regions and encode two distinct isoforms diverging in their amino-terminal domains (Wickham et al. 1999). To investigate whether corresponding rStau transcripts exist in the postnatal rat brain, RT-PCR was performed with the two oligonucleotide primers, Stau-s and Stau-as, that flank the putative splice area. With RNA from P5, P7, P9 and P13 cerebral cortex and hippocampus as well as total embryonic and adult rat brain, only PCR products were detected that are derived from the rStau cDNA sequence reported above (data not shown). Thus, alternatively spliced Staufen transcripts that are homologous to the four mRNAs described in human tissue may not exist in the rat brain.

Figure 1.

(a) Amino acid sequence of rStau. The double-stranded RNA-binding domains and the six amino acids that are different in rStau+I6 and rStau-I6 isoforms are highlighted in grey and black, respectively. Numbers on the left indicate amino acid positions. The rStau cDNA was submitted to the GenBank/EMBL Data Bank (accession number AF227200). (b) Schematic diagram of Staufen isoforms from rat and Drosophila melanogaster (accession no. M69111). Grey boxes represent short and full-length RNA-binding domains. The I6insert present in rStau+I6 is shown as a black box. The two MAP1B homology regions, MHR1 and 2, are depicted as hatched boxes. In comparison to rStau, Drosophila Staufen contains an additional amino-terminal full-length dsRBD.

Figure 2.

Developmental expression of rStau+I6 and rStau-I6 mRNAs in brain. PCR products were obtained with the oligonucleotide pair 5-I6/3-I6. For cDNA synthesis, total RNA from P5, P7, P9 and P13 rat cerebral cortex and hippocampus was reverse transcribed.

The rStau isoforms possess high sequence similarity to proteins that were recently characterized in mouse (mStau) (Wickham et al. 1999) and human (hStau) (Marion et al. 1999; Wickham et al. 1999) (97% and 90%, respectively). The number and position of the dsRBDs are conserved between these species homologues. The I6 insert present in rStau + I6 has not yet been described in mouse and human isoforms. In comparison to rStau, the two human isoforms contain 7 and 88 additional amino-terminal residues. The rStau isoforms are also related to human STAU2 that is derived from a different gene (Buchner et al. 1999). Both rStau isoforms possess two partially overlapping regions that exhibit sequence similarity to two distinct regions of the microtubule-associated protein 1B (MAP1B). We refer to the corresponding rStau domains as MAP1B homology regions (MHR). The first region, MHR1, starts carboxy-terminal of the third dsRBD in rStau and is 30% identical and 44% similar to a carboxy-terminal MAP1B domain (Fig. 3a) that has been described as a basic in vitro microtubule-binding domain (Zauner et al. 1992). MHR2 comprises about half of the second and the entire third dsRBD in rStau and shows 22% identity and 39% similarity to a region located proximal of a second amino-terminal microtubule-binding domain in MAP1B (Fig. 3B) (Noble et al. 1989; Zauner et al. 1992).

Figure 3.

Alignment of two homologous regions in rStau and MAP1B. Identical and similar (+) amino acids are listed in the line between the rStau and rat MAP1B sequence. Dashes indicate sequence gaps. MAP1B accession number: A56577

Staufen gene expression in mammalian tissues

To determine the regional distribution of Staufen mRNAs in different rat tissues and distinct regions of the mammalian CNS, Northern-blots were hybridized with an rStau specific cDNA probe (Fig. 4). rStau transcripts (each 3.1 kb) were detected in all tested adult rat tissues with the highest mRNA levels present in liver, heart, testis, brain and kidney (Fig. 4a). Low amounts of rStau transcripts were detected in lung, skeletal muscle, and spleen. A similar tissue specific distribution pattern of the Staufen mRNA has also been described in human (Marion et al. 1999; Wickham et al. 1999). In the rat cerebral cortex and hippocampus, rStau transcripts were detected in 3- and 13-day-old animals via Northern blots (Fig. 4b) and in 5-, 7-, 9- and 13-day-old rats via PCR (Fig. 2). Staufen transcripts of similar size were also present in all tested adult human CNS areas, including cerebellum, cerebral cortex, spinal cord and hippocampus (Fig. 4c). Taken together these data show that the mammalian Staufen gene is ubiquitously expressed in all tissues and CNS areas.

Figure 4.

Northern blot analysis of Staufen expression in different rat tissues and mammalian brain regions. (a) Rat multiple-tissue blot with poly (A)+ RNA (Clontech). (b) Northern blot of total RNA (10 µg per lane) prepared from cerebral cortex and hippocampus of postnatal day 3 (P3) and P13 rat brain. (c) Blot with poly (A)+ RNA from multiple human neuronal tissues (Clontech). Blots were probed with the α[32P]-labelled 2.5-kb rStau cDNA fragment from the clone pATCII.73 that was isolated in the initial yeast tri-hybrid screen. A 3.1-kb band is seen in all lanes. Control hybridizations were performed with probes against the β-actin or α-tubulin mRNAs. The size of molecular weight markers in kilobases and their positions are indicated.

rStau accumulates around dendritic microtubules

To determine the subcellular localization of rStau in rat brain neurones, cortical sections were immuno-stained using a polyclonal rabbit antiserum specific for mammalian Staufen (Wickham et al. 1999). As shown in Figs 5 (a) and (b), rStau is present in pyramidal cells of the CA1 area of the hippocampus and the cerebral cortex. In these neurones, the protein localizes to somata as well as dendritic processes. To investigate whether rStau is associated with specific cellular structures, we performed immunoelectron microscopy with adult rat hippocampal sections. In longitudinal and cross-sections of dendritic processes, rStau primarily accumulates along microtubules (Fig. 5c) suggesting an association with the cytoskeleton.

Figure 5.

Distribution of rStau in adult brain. Coronal rat brain sections were immunostained with anti-hStau. rStau exhibits a somatodendritic distribution in pyramidal neurones of the CA1 area in the hippocampus (a) and cerebral cortex (b). SP, stratum pyramidale; SR, stratum radiatum; SLM, stratum lacunosum-molecular. (c) Electron microscopic analysis of the subcellular distribution of rStau in dendritic processes of the rat hippocampus. In longitudinal (left side) and cross-sections (right side) of dendrites, rStau appears to be associated with microtubules (arrowheads). Synapses (arrows) and mitochondria are marked (m). Scale bars: 100 µm (a), 70 µm (b), 1 µm (c).

Distinct rStau isoforms differentially bind RNA

The I6 insert specific for rStau+I6 has not yet been described in Staufen proteins from other species. Although it does not possess a significant sequence similarity to any known protein domain, its position in the second dsRBD indicates that it may influence RNA-binding. To verify this hypothesis, we investigated the interaction of rStau isoforms with various RNAs in a North-western blot assay. In Escherichia coli, rStau+I6 and rStau-I6 were expressed as fusion proteins with glutathione S-transferase (GST) (Fig. 6). The GST-SAP fusion protein containing the first two PDZ domains of SAP102 (Müller et al. 1996) was used as a negative control. SAP102 is not known to exhibit any RNA-binding ability. Identical amounts of purified GST fusion proteins were separated on SDS polyacrylamide gels and transferred onto membranes. Immunodetection with antibodies directed against mammalian Staufen or GST confirmed that both full-length ∼78 kDa GST-rStau proteins and GST-SAP were blotted onto the membranes (Fig. 6, middle panel). Lower-molecular weight immunoreactive products are potentially generated by partial degradation. The blotted GST fusion proteins were incubated with identical amounts of three different radioactively labelled RNA probes. When the membranes were overlaid with a MAP2-DTE probe, the RNA was shown to interact with both GST-rStau isoforms, but not with GST-SAP (Fig. 6, right panel). Two other RNAs, α-tubulin and the vegetal localization element (VLE) of Vg1 transcripts (Elisha et al. 1995), also bound both rStau isoforms, but did not interact with GST-SAP. Thus, in vitro, both rStau proteins interact with multiple transcripts, implying that RNA-binding is at least in part sequence-independent. All three tested RNA probes led to a stronger labelling of rStau-I6 as compared to rStau+I6. The radioactive labelling of rStau-I6 by the VLE probe was slightly lower than the labelling by MAP2-DTE and α-tubulin mRNA probes. Taken together, these data show that both rStau isoforms exhibit differential RNA-binding properties.

Figure 6.

In vitro interaction of rStau with distinct RNAs. Recombinant GST-rStau isoforms and GST-SAP proteins (1 µg each) were separated in a polyacryamide gel. Left panel, a representative Coomassie blue stained protein gel with extracts from bacteria expressing GST-rStau+I6 (arrow). Proteins were transferred to a membrane, and incubated with antibodies against mammalian Staufen (anti-Stau) and GST (anti-GST) (middle panel) or with 32P-labelled RNA probes (right panel) comprising the MAP2-DTE, the vegetal localization element of the Vg1 mRNA (VLE) and α-tubulin mRNA sequences. In the Western blot, the bacterial expression of GST-rStau + I6 and GST-rStau-I6 (~75 kDa) as well as GST-SAP (~45 kDa) fusion proteins was confirmed. In the North-western blot, RNA–protein interactions were detected by autoradiography. All tested RNA fragments selectively interact with both GST-rStau fusion proteins, but not with GST-SAP. GST-rStau-I6 shows a stronger signal of bound radioactively labelled RNAs compared to GST-rStau + I6. The size of molecular weight markers in kilodaltons and their positions are indicated.

The yeast tri-hybrid system was used to further investigate the ability of the initially isolated partial rStau protein to interact with various mRNAs. In addition to the Rev response element (RRE) the chimaeric transcripts used in this assay contained the MAP2-DTE (vector pRevRDTE), MAP2 3′ UTR sequences located upstream from the DTE (pRevRMAP2u1586−2435), the 3′ UTR of transcripts encoding the alpha subunit of the Ca2+/calmodulin-dependent protein kinase II (pRevRαCaMK), α-tubulin (pRevRtub), or glycerolaldehyde phosphate dehydrogenase mRNA sequences (pRevRGAP), the VLE of Vg1 transcripts (pRevRVLE), or a second copy of the RRE (pRevR2). All yeast cells coexpressing rStau coupled to the GAL4 activation domain together with one of these recombinant transcripts were able to grow on His/Trp/Leu/3AT plates with the exception of those containing the double-RRE motif (data not shown). This finding suggests that in yeast cells rStau interacts with many RNAs in a rather sequence-independent manner. The inability of the double-RRE to interact with rStau may be explained by the low tendency of short RNA fragments to form a complex secondary structure which could be a prerequisite for binding to rStau.

Discussion

In neurones, dendritic targeting of specific transcripts seems to contribute to protein compartmentalization as well as synaptic plasticity (Kindler et al. 1997; Kuhl and Skehel, 1998; Schuman, 1999; Tiedge et al. 1999; Wells et al. 2000). It is thought that the molecular machinery underlying mRNA trafficking in dendrites involves cis-acting signal sequences on localized transcripts and trans-acting protein factors. In this paper, we describe the isolation of a putative trans-acting rat brain protein that exhibits strong sequence similarity to a group of proteins related to Drosophila Staufen (St Johnston et al. 1991; Buchner et al. 1999; Marion et al. 1999; Wickham et al. 1999; Roegiers and Jan 2000) and binds to the cis-acting DTE in MAP2 mRNAs.

Our data show that at least two different rStau isoforms are expressed in rat brain. In the longer rStau+I6 isoform an insert comprising six amino acids is integrated in front of the second dsRBD (dsRBD II). A similar isoform containing this insert has not yet been described in Drosophila, human or mouse (St Johnston et al. 1991; Buchner et al. 1999; Marion et al. 1999; Wickham et al. 1999). Although the six amino acid domain does not exhibit a strong sequence similarity to other proteins, its location in dsRBD II suggests that it may influence RNA-binding. This hypothesis is supported by our overlay data showing that this six amino acid insertion drastically reduces the amount of radioactively labelled RNAs that are bound to rStau. dsRBDs from different proteins fold into a compact αβββα structure (Bycroft et al. 1995; Kharrat et al. 1995; Nanduri et al. 1998). In Drosophila Staufen (dmStau) helix α1, loop2 and loop4 of dsRBD III mediate dsRNA-binding (Ramos et al. 2000). A sequence alignment between dsRBD II of rStau and dsRBD III of dmStau suggests that the I6 element in rStau+I6 is inserted into the first β-sheet. Thus, it seems unlikely that this insertion directly changes the surface of dsRBD II in rStau+I6 that interacts with the RNA stem-loop. In contrast, the insertion is flanked by a four amino acid stretch that forms a β-bulge within the first strand of the β-sheet of dsRBD III in dmStau. Mutations within this β-bulge region of dsRBD III in dmStau strongly reduce or abolish RNA-binding (Ramos et al. 2000). Taken together, these findings are compatible with the interpretation that dsRBD II in rStau+I6 does not function as RNA-binding domain. An association of rStau+I6 with RNA is most likely mediated by one or more of the remaining dsRBDs. In cells, differential RNA-binding of rStau may be established via a regulated six amino acid insertion into dsRBD II as a result of alternative precursor RNA splicing.

In the overlay system, both rStau isoforms were found to interact with three distinct RNAs. Similarly, rStau bound to a number of unrelated RNAs in the yeast tri-hybrid system. Thus, rStau appears to interact with multiple RNAs in a rather sequence-independent manner. Interestingly, neither dmStau (St Johnston et al. 1992) nor mouse and human homologues (Marion et al. 1999; Wickham et al. 1999) exhibit sequence-dependent RNA-binding in vitro. In contrast, in vivo dmStau appears to associate specifically with selected endogenous transcripts. Injection of 3′ UTR sequences of bicoid (Ferrandon et al. 1994) or prospero mRNAs (Schuldt et al. 1998), but not several control dsRNAs, into early Drosophila embryos was shown to recruit Staufen to form granules. These observations suggest that in vivo post-translational modifications of Staufen and/or additional cofactors are required for a selective association with specific transcripts. Although a number of proteins have been shown to interact with dmStau, cofactors that regulate its RNA-binding specificity have not yet been described (Roegiers and Jan 2000). Interestingly, in neurones, endogenous as well as recombinant Staufen form large RNA-containing granules that move along dendrites (Kiebler et al. 1999; Köhrmann et al. 1999). Moreover in dendritic shafts of transfected hippocampal neurones, chimaeric transcripts containing the MAP2-DTE have been shown to assemble into similar granular particles (Blichenberg et al. 1999). The characterization of endogenous Staufen complexes will be necessary to determine the nature of associated RNAs and putative protein cofactors.

We have identified two partially overlapping regions in rStau, MHR1 and MHR2, which are similar to MAP1B. MHR2 in rStau is homologous to a MAP1B region that has recently been functionally described as a domain interacting with GABAC receptor subunits (Hanley et al. 1999). This interaction is thought to mediate anchoring of GABAC receptors at synapses (Hanley et al. 1999). Currently, it is not known whether mammalian Staufen can also interact with GABAC receptor subunits. MHR1 is similar to a carboxy-terminal region in MAP1B that has been described in vitro as a basic microtubule-binding domain (Noble et al. 1989; Zauner et al. 1992). In an in vitro overlay assay, MHR1 in hStau was capable to interact with tubulin (Wickham et al. 1999). Our immunohistochemical data show that in the rat brain, rStau is found in somata and dendrites of cerebral and hippocampal neurones. In dendrites, rStau primarily localizes along microtubules. Similarly in cultured hippocampal neurones, rStau was found to exhibit a comparable somatodendritic distribution pattern as well as a microtubular association (Kiebler et al. 1999). Taken together, these observations suggest that in dendritic shafts, mammalian Staufen is bound to microtubule filaments. In the rat hippocampus, several mRNAs are found in the dendrites of pyramidal neurones, including transcripts encoding MAP2 (Garner et al. 1988; Tucker et al. 1989), the alpha subunit of the Ca2+/calmodulin-dependent protein kinase II (Burgin et al. 1990), arg3.1/arc (Link et al. 1995; Lyford et al. 1995) and dendrin (Herb et al. 1997). The ability of rStau to interact with the MAP2-DTE as well as other unrelated RNAs shown in this paper, suggests that in the mammalian brain, Staufen may serve to link dendritically localized mRNAs to microtubules. Three independent observations suggesting an attachment of Staufen and/or extrasomatic transcripts to cytoskeletal filaments corroborate this hypothesis. First, bidirectional movement of fluorescently tagged RNA-containing Staufen granules in dendritic shafts of transfected hippocampal neurones is microtubule-dependent (Köhrmann et al. 1999). Secondly, when neuronal cultures derived from embryonic mouse brain were used to biochemically separate polymerized microtubules from unassembled tubulin, the MAP2 mRNA concentration in the microtubule fraction was about five times higher than in the soluble fraction (Litman et al. 1994). Thirdly, using in situ hybridization, Bassell et al. (1994) have shown that in cultured cerebrocortical neurones, poly (A) mRNA in dendrites is bound to microtubules. The molecular characterization of neuronal Staufen complexes should help to understand how the interaction of Staufen with different RNAs and cytoskeletal filaments is regulated.

Acknowledgements

This research was supported by the Deutsche Forschungsgemeinschaft (Ri191–19–1, Ri192–21–1, Ri192–21–3, DFG FOR 296/2–1-3, DFG FOR 296/2–1-5). This article is based on a doctoral study by M. Monshausen in the Faculty of Biology, University of Hamburg.

Ancillary