The RNA-binding protein MARTA2 regulates dendritic targeting of MAP2 mRNAs in rat neurons

Authors


  • Data deposition: The MARTA2 cDNA sequence is available under GenBank accession number DQ144645.

Address correspondence and reprint requests to Stefan Kindler, Institute for Human Genetics, University Medical Center Hamburg-Eppendorf, Campus Forschung N27, Martinistr. 52, D-20246 Hamburg, Germany. E-mail: kindler@uke.de

Abstract

Dendritic targeting of mRNAs encoding the microtubule-associated protein 2 (MAP2) in neurons involves a cis-acting dendritic targeting element. Two rat brain proteins, MAP2-RNA trans-acting protein (MARTA)1 and MARTA2, bind to the cis-element with both high affinity and specificity. In this study, affinity-purified MARTA2 was identified as orthologue of human far-upstream element binding protein 3. In neurons, it resides in somatodendritic granules and dendritic spines and associates with MAP2 mRNAs. Expression of a dominant-negative variant of MARTA2 disrupts dendritic targeting of endogenous MAP2 mRNAs, while not noticeably altering the level and subcellular distribution of polyadenylated mRNAs as a whole. Finally, MAP2 transcripts associate with the microtubule-based motor KIF5 and inhibition of KIF5, but not cytoplasmic dynein function disrupts extrasomatic trafficking of MAP2 mRNA granules. Thus, in neurons MARTA2 appears to represent a key trans-acting factor involved in KIF5-mediated dendritic targeting of MAP2 mRNAs.

Abbreviations used
αCamKII

α-subunit of calcium/calmodulin-dependent protein kinase II

CPEB

cytoplasmic polyadenylation element binding protein

dsRBD

double-stranded RNA-binding domain

DTE

dendritic targeting element

EGFP

enhanced green fluorescent protein

FBP

far-upstream element binding protein

FMRP

fragile X mental retardation protein

FUSE

far-upstream element

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

GST

glutathione S-transferase

HPRT

hypoxanthine-guanine phosphoribosyltransferase

KH domain

heterogeneous ribonucleoprotein K homology domain

KIF

kinesin family protein

MAP2

microtubule associated protein 2

MARTA

MAP2-RNA trans-acting protein

RNP

ribonucleoprotein particle

RSW

ribosomal salt wash

RT-PCR

reverse transcriptase polymerase chain reaction

SDS

sodium dodecyl sulfate

UV CLA

ultraviolet cross-linking assay

The locally regulated protein synthesis in restricted cytoplasmic areas of different cells contributes to the establishment and maintenance of cell polarity (Bullock 2011; Jansen and Niessing 2012; Kindler and Kreienkamp 2012; Vollmeister et al. 2012). In neurons, trafficking and activity-dependent translation of select transcripts in dendrites plays a role in long-term synaptic plasticity (Bramham and Wells 2007; Dahm et al. 2007; Cajigas et al. 2010; Kindler and Kreienkamp 2012). Transport of mRNAs into dendrites involves an interaction of cis-acting dendritic targeting elements (DTEs) with trans-acting proteins and thus the formation of ribonucleoprotein (RNP) particles. Different RNP particles utilize particular motor proteins to move along filaments of the cytoskeleton (Bullock 2011; Baumann et al. 2012; Jansen and Niessing 2012; Kindler and Kreienkamp 2012; Vollmeister et al. 2012). In transcripts encoding microtubule-associated protein 2 (MAP2), a DTE is situated in the 3′ untranslated region (3′UTR) (Blichenberg et al. 1999). In vitro, two rat brain MAP2-RNA trans-acting proteins, MARTA1 and MARTA2, specifically bind to the DTE with high affinity (Rehbein et al. 2000). MARTA1 is a nucleic acid binding protein that primarily resides in the nucleus but is also present in the somatodendritic cytoplasm of neurons (Rehbein et al. 2002). It may thus mediate both nuclear export and extrasomatic trafficking of MAP2 mRNAs. Here, we affinity-purified and characterized MARTA2. Like MARTA1, MARTA2 belongs to the family of far-upstream element (FUSE) binding proteins (FBPs) (Davis-Smyth et al. 1996). All three family members (FBP1-3) use four central KH domains for single-stranded nucleic acid binding and have been implicated in gene expression and RNA metabolism (Rehbein et al. 2000, 2002; Chen et al. 2001; Gu et al. 2002; Kolev and Huber 2003; Gherzi et al. 2004; Briata et al. 2005, 2011; Linker et al. 2005; Bolognani and Perrone-Bizzozero 2008; Trabucchi et al. 2009; Danckwardt et al. 2011). We show that MARTA2 associates with MAP2 mRNAs in neurons and resides in somatodendritic regions including dendritic spines. Functional experiments strongly suggest that MARTA2 plays a central role in KIF5c-dependent dendritic targeting of endogenous MAP2 mRNAs.

Materials and methods

Brain fractionation, MARTA2 affinity purification and mass spectrometry

Rat brain fractions were prepared as described (Rehbein et al. 2000). From RSW fractions, MARTA2 was affinity-purified as reported (Rehbein et al. 2002) and analysed by mass spectrometry (Brendel et al. 2004). Wistar rats were bred in the central animal facility of the University Medical Center Hamburg-Eppendorf, Germany, and handled in accordance with national and international guidelines for animal welfare. Protocols utilized for anaesthetization of pregnant female rats, preparation of rat embryos and isolation of embryonic brains were approved by the animal committee of the University Medical Center Hamburg-Eppendorf and the city state Hamburg, Germany (permit number ORG 371).

cDNA cloning and expression vectors

Two primer pairs (forward1: 5′-ATGGCGGAGCTGGTGGTGCAGGGGCA-3′, reverse1: 5′-GTAAGTACACAGGGCCTCTC-3′, forward2: 5′-CAGAAGAATTCAAAGTGCCTG-3′, reverse2: 5′-CTAAATATAGTGTTTCACAGAAC-3′) derived from mouse FBP3 cDNA (GenBank accession number XM_130112.1) were used to PCR amplify two overlapping MARTA2 cDNA regions from adult rat brain cDNA (Monshausen et al. 2001). cDNA regions encoding rat MARTA2 (DQ144645, nt 211-1281), Staufen 2 (NM_134466, nt 1-1140), cytoplasmic polyadenylation elementbinding protein (CPEB) (NM_001106276, nt 30-1712) and KIF5c (NM_001107730, nt 352-1341) were cloned into pEGFP-N3 (Clontech, Palo Alto, CA, USA) to create pKH-EGFP, pStau2RBD-EGFP, pCPEB-EGFP and pdnKIF5c-EGFP respectively. EGFP-tagged fusion proteins contain amino acid residues #72-427 of rat MARTA2 (ABA18725; includes all four KH domains), #1-380 of rat Staufen 2 (NP_604461; contains all four dsRBDs) (Tang et al. 2001), #1-561 of rat CPEB (NP_001099746; full-length protein) and #1-330 of rat KIF5c (NP_001101200) (Schepis et al. 2007). Dynamitin-EGFP expression vector has been described (NM_006400, nt 133-1353) (Burkhardt et al. 1997).

Cell culture, antibodies and immunochemistry

Primary hippocampal neurons were essentially prepared and transfected as described (Blichenberg et al. 1999). However, cells were grown in NEUROBASAL medium (Life Technologies, Darmstadt, Germany) without glial feeder layer and transfected seven days after plating. Two pGEX-6P-3 derived vectors (GE Healthcare, Freiburg, Germany) containing two different MARTA2 cDNA parts (GenBank accession number DQ144645, nucleotides 1-310 and 1270-1710 respectively) were introduced in E.coli BL21-CodonPlus-RP cells (Stratagene, Heidelberg, Germany). Corresponding glutathione S-transferase (GST) fusion proteins (N-MARTA2-GST and C-MARTA2-GST respectively) were affinity-purified from bacterial lysates according to the manufacturer's protocol. Polyclonal antisera directed against N- (antiserum N54; 1 : 100 dilution for immunochemistry) and C-terminal (C98 and C99; 1 : 100 dilution) domains of MARTA2 were produced in rabbits and affinity-purified (BioGenes, Berlin, Germany). Monoclonal rat antibodies were raised against N-MARTA2-GST (1 : 3 dilution for immunochemistry). A6 and A7 antisera recognizing MARTA1 have been described (Rehbein et al. 2002). Antibodies for immunocytochemistry directed against MAP2 (HM2; Sigma, St. Louis, USA; 1 : 1000 dilution) and GFP (ab6556; Abcam, Cambridge, UK; 1 : 1000), and for western blots directed against α-tubulin (ab7291; Abcam; 1 : 2000), ribosomal protein L7A (RPL7A, #2403, Cell Signaling Technology, Danvers, MA, USA; 1 : 1000), ribosomal protein S3 (RPS3, #2579, Cell Signaling Technology; 1 : 1000) and histone H3 (A300-823Al; Bethyl laboratories, Montgomery, TX, USA; 1 : 2000) are commercially available. For immunocytochemistry, neurons were either fixed for 15 min at 20°C in 4% paraformaldehyde (w/v), 4% sucrose (w/v) in phosphate-buffered saline (PBS; 100 mM sodium phosphate buffer pH 7.4, 0.9% NaCl (w/v)) or ice-cold methanol for 4 min at −20°C. Immunocytochemistry was performed as described (Blichenberg et al. 1999) with AlexaFluor488, AlexaFluor546, AlexaFluor635, Marina Blue and Cy5 coupled secondary antibodies (Life Technologies, 1 : 500 dilution). Cells were mounted on slides with Permafluor (Beckman Coulter, Miami, FL, USA). For blocking experiments, affinity-purified N54, C89 and C99 antisera were pre-incubated with MARTA2-GST fusion proteins (2 μg/mL) at 4°C overnight. Images were captured with a Leitz Aristoplan microscope (Leica Mikrosysteme, Wetzlar, Germany) and Openlab software (Improvision, Tübingen, Germany) or a laser-scanning microscope (TCS-SP2, Leica Mikrosysteme) and mounted using Adobe Photoshop CS and FreeHand software (Adobe Systems GmbH, Unterschleißheim, Germany). Fluorescent intensities and sizes of selected areas were measured with ImageJ software (NIH, Bethesda, Maryland, USA). Light and electron microscopic analysis of brain tissue sections was essentially performed as described (Monshausen et al. 2001; Blässe et al. 2006) utilizing 1 : 100 diluted affinity-purified MARTA2 antisera. To determine the percentage of dendritic spines containing MARTA2, pictures were taken at 20 000x magnification from randomly chosen areas of hippocampal CA1 stratum radiatum. Dendritic spines possessing an obvious post-synaptic density (PSD) were evaluated (n = 650).

RNA synthesis, cross-linking assays and immunoprecipitation

In vitro RNA synthesis and UV CLAs are described (Rehbein et al. 2002). Immunoprecipitation of MARTA1 and MARTA2 after UV cross-linking was performed essentially as described (Rehbein et al. 2002). Briefly, for each 50 μL reaction 10 μg protein from a RSW fraction were incubated with 20 fmoles MAP2-DTE probe, cross-linked and digested with RNase. Proteins were precipitated at 4°C overnight with 20 μL protein A agarose suspension (Santa Cruz, Heidelberg, Germany) and 2.5 μg affinity-purified antibodies or none-specific rabbit immunoglobulin G (IgG), respectively, in 1 mL IP buffer (120 mM Tris-HCl pH 8.0, 120 mM NaCl, 1 mM EDTA, 0.5% NP40 (v/v), proteinase inhibitor cocktail Complete; Roche, Mannheim, Germany), washed three times with 1 ml IP buffer each and eluted by boiling in sodium dodecyl sulfate (SDS) sample buffer.

Glutathione-S-transferase pull-down, immunoprecipitation and real-time reverse transcriptase polymerase chain reaction (RT-PCR)

GST and the fusion protein GST-KIF5c containing the cargo domain of mouse KIF5c (amino acid residues 826-920 of NP_032475) were expressed in E.coli and purified using GSH-sepharose (GE Healthcare). GST-KIF5c-coated sepharose beads were used to affinity purify KIF5c associated cargoes from mouse brain homogenates (Kanai et al. 2004). Immunoprecipitation experiments were performed with affinity-purified rabbit MARTA2 or KIF5c antisera (Thermo Scientific, Rockford, IL, USA) and irrelevant rabbit IgGs and C57BL/6J mouse brain extracts (Iacoangeli et al. 2008; Kindler et al. 2009; Schütt et al. 2009). RNA was extracted using RNeasy Mini Kit (Qiagen, Hilden, Germany) and analysed by RT-PCR and quantitative real-time RT-PCR as described (Iacoangeli et al. 2008; Kindler et al. 2009; Schütt et al. 2009). PCR conditions were according to the manufacturer's standard protocol with an annealing temperature of 58°C (40 cycles) and primers specific for MAP2 (Johnson et al. 2006), Arc/Arg3.1 (Kindler et al. 2009), the α-subunit of calcium/calmodulin-dependent protein kinase II (αCamKII; Qiagen, QuantiTect primer QT00129164), β-actin (QT01136772), Shank1 (Shank1-fw: 5′-AGCCTGCAGCAGTGCCCAGCA-3′, Shank1-rev: 5′-ATGCGAGGCCGCCAGGCCCA-3′), HPRT (Kurth et al. 2007) and Pgk2 transcripts (QT02242471). RT-PCR products were analysed on 2% agarose gels.

In situ hybridization

Preparation of digoxygenin-labelled RNA probes and in situ hybridizations are described (Blichenberg et al. 1999). MAP2 sense and antisense probes transcribed from appropriately digested pBluescript derivatives (Stratagene) contain nucleotides 60-5552 of the rat MAP2 cDNA (accession number X51842). For detection of poly(A) RNA, we used digoxygenin-labelled oligo(dT)50 (Eurofins MWG Operon, Ebersberg, Germany). Control in situ hybridizations were performed in the presence of a 100-fold excess of unlabelled oligo(dT)50. A Cy3-coupled mouse monoclonal anti-digoxigenin antibody (1 : 100 dilution in blocking buffer; Roche) followed by a Cy3 coupled sheep anti-mouse antibody (1 : 100 dilution; Sigma) was used to detect digoxigenin.

Statistical data analysis

Data were analysed with a one-sided paired student′s t-tests (level of significance α = 0.05). Results from clustered experiments were analysed with a linear mixed models application considering the clustered structure of tests (SPSS 15.0.1; SPSS Inc., Chicago, IL, USA, 2006; level of significance α = 0,05).

Results

In vivo association of MARTA2 with MAP2 mRNAs

We have previously shown that MARTA2, a specific trans-acting factor of the MAP2-DTE, is strongly enriched in ribosomal salt wash (RSW) fractions obtained from adult rat brain (Rehbein et al. 2000). To affinity purify MARTA2, RSW proteins were incubated with magnetic streptavidin beads coated with biotinylated MAP2-DTE. After washing, RNA-bound proteins were eluted and different fractions obtained during affinity purification were tested in ultraviolet cross-linking assays (UV CLAs) with a radioactively labelled MAP2-DTE probe (Fig. 1a). Strong MARTA2 and MARTA1 binding activity was present in RSW (Fig. 1a, RSW) and eluate fractions (Fig. 1a, E1& E2), but was drastically reduced or not detectable in the supernatant fraction of the batch purification (Fig. 1a, S) and different wash fractions (Fig. 1a, W1–3) respectively. In a silver-stained SDS-polyacrylamide gel, a 63 kDa band was highly enriched in eluate fractions (Fig. 1b, E1 & E2) and was not present in three wash fractions (Fig. 1b, W1–3). Staining intensity of the 63 kDa band in different fractions correlated well with MARTA2 binding activity observed in corresponding UV CLAs (Fig. 1a), indicating that the eluted band indeed represents MARTA2. This band was clearly detected in a Coomassie-blue stained SDS-polyacrylamide gel loaded with concentrated material from several pooled eluate fractions (Fig. 1c). Proteins around 63 kDa were excised from the gel and digested, and resulting peptides were analysed by mass spectrometry. Five identified peptides (Table 1) revealed MARTA2 as the rat orthologue of human FBP3 (Davis-Smyth et al. 1996). Co-purified proteins of about 71 kDa and 84 kDa (Fig. 1c) are rat orthologues of human FBP1 and human FBP2 (Davis-Smyth et al. 1996) (vertebrate FBP2 and FBP3 forms are herein collectively referred to as MARTA1 and MARTA2 orthologues respectively). Four oligonucleotides derived from the mouse MARTA2 cDNA sequence were used to PCR amplify a cDNA region containing the entire rat MARTA2 coding region and part of the adjacent 3′UTR (GenBank accession number DQ144645). The 1710 nucleotide spanning open reading frame encodes a 569 amino acid residue protein (Figure S1) that has a molecular mass of 61.4 kDa and an isoelectric point of 8.27. All peptides identified by mass spectrometry are present in the deduced amino acid sequence. In its central part, MARTA2 contains four KH domains known to mediate nucleic acid binding (Braddock et al. 2002b). MARTA2 orthologues found in many vertebrate species, including H. sapiens, P. troglodytes, B. taurus, E. caballus, C. familiaris, R. norvegicus, M. musculus, G. gallus, X. laevis, D. rerio and S. salar are highly conserved (Figure S2).

Table 1. MARTA2 peptides identified by mass spectrometry
Peptide #Amino acid sequenceAmino acid position in MARTA2
  1. All peptides were analysed by tandem mass spectrometry. MARTA2, MAP2-RNA trans-acting protein 2.

1RLLGQIVDR139-147
2MVMIQDGPLPTGADKPLR196-213
3IQFKPDDGISPER287-299
4IINELILTAQER316-327
5GAPQQIEVAR409-418
Figure 1.

Purification of MAP2-RNA trans-acting protein (MARTA)2 from adult rat brain. (a) Autoradiograph of UV cross-linked samples separated on an SDS-polyacrylamide gel. Ultraviolet cross-linking assays (UV CLAs) were performed with 5 fmoles of P32-labelled dendritic targeting element (DTE) probe and 1 μg ribosomal salt wash (RSW) protein (lane 1) and 2 μL each of the different protein fractions from affinity purification (B, starting material diluted in binding buffer before addition to affinity-beads; S, supernatant fraction of affinity-beads; W1-3, wash fractions 1-3 respectively; E1 & E2, eluate fractions 1 and 2). Protein amounts used for assays shown in E1 and E2 represent 1/200 and 1/100 of the respective eluate fractions. MARTA2 is present in both eluate fractions, but not in the wash fractions. (b) Samples (1 μg protein in RSW, B and S; 20 μl each of fractions used in A, lanes W1-3 and E1-2) were analysed in a silver-stained SDS-polyacrylamide gel. The relative intensity of a 63 kDa band present in eluate samples correlates well with the observed MARTA2 binding activity as shown in (a). For a description of lanes see a, c) Concentrated proteins of the pooled eluate fractions were separated by SDS PAGE and stained with Coomassie-blue. Prominent bands of about 63, 70 and 80 kDa, which were excised and analysed by mass spectrometry, represent rat orthologues of MARTA2/far-upstream element binding protein (FBP)3, FBP1 and MARTA1/FBP2 respectively.

Three different polyclonal antisera (N54, C98 and C99), which were raised against N- or C-terminal parts of MARTA2, selectively recognize a 63-kDa band in a rat brain polysome fraction (Fig. 2a, arrow head). To confirm the identity of the purified 63 kDa protein with 65 kDa MARTA2 detected in UV CLAs, these antibodies were used to immunoprecipitate proteins from UV CLAs performed with ribosomal salt wash proteins and a radioactively labelled MAP2-DTE probe. Immunoprecipitates were analysed on an SDS gel (Fig. 2b). For all three antibodies, only the 65 kDa cross-linked MARTA2 protein was found in the precipitate (Fig. 2b; N54, C98 and C99), whereas the 90 kDa MARTA1 remained in the supernatant. Conversely, MARTA1 but not MARTA2 was precipitated with two different MARTA1-specific antibodies (Fig. 2b, A6 and A7). Neither MARTA2 nor MARTA1 is precipitated with irrelevant IgGs (Fig. 2b, IgG). These findings confirm that the affinity-purified 63 kDa MAP2-DTE binding protein indeed represents MARTA2 observed in UV CLAs (Fig. 2b, ‘CLA’) (Rehbein et al. 2000). In addition, MARTA2 antisera do not cross-react with MARTA1. To determine whether MARTA2 associates with MAP2 mRNAs in vivo, we probed for an association of both components in the mouse brain. Utilizing brain homogenates, we performed immunoprecipitations with MARTA2 specific antibodies (M-IP) and irrelevant rabbit IgGs (IgG-IP). A Western blot analysis confirmed that the precipitate from M-IP but not IgG-IP contains MARTA2 (Fig. 2c). RNA extracted from immunoprecipitates and input material was then probed for MAP2 and Shank1 mRNAs by reverse transcriptase polymerase chain reaction (RT-PCR). Analysis of amplification products revealed that while both transcripts were present in the input material, neither MAP2 nor Shank1 transcripts were detected in the water control and IgG-IP precipitate, thus excluding unspecific RNA precipitation (Fig. 2d). Notably, MAP2 but not Shank1 mRNAs were precipitated in the M-IP reaction. Also, utilizing quantitative real-time RT-PCR analysis, MAP2 mRNAs were found to be about eight-fold enriched in M-IP versus IgG-IP precipitates. However, M-IP levels of dendritic mRNAs encoding the α-subunit of calcium/calmodulin-dependent protein kinase II (αCamKII), β-actin or Shank1 were only slightly above background levels observed in IgG-IP samples (Fig. 2e). Taken together, these data show that in the mouse brain MARTA2 selectively associates with dendritic MAP2, but not αCamKII, β-actin or Shank1 transcripts.

Figure 2.

Immunoprecipitation of cross-linked MAP2-RNA trans-acting protein (MARTA) proteins and association of MARTA2 with MAP2 mRNAs in the mouse brain. (a) Western blot performed with affinity-purified MARTA2 antibodies C89, C99 and N54, which were raised against C- or N-terminal parts of MARTA2. Forty microgram protein from a rat brain polysome fraction was loaded. All antibodies selectively recognize MARTA2 at 63 kDa. Pre-incubation of N54 with N-MARTA2-GST fusion protein disrupts immunodetection of the 63 kDa protein (N54block). (b) Autoradiograph of UV cross-linked proteins separated on an SDS-polyacrylamide gel. Cross-linked material as shown in the first lane (CLA) was used as input for immunoprecipitation experiments. For each immunoprecipitation, 10 μg protein from a ribosomal salt wash (RSW) fraction of adult rat brain was incubated with 20 fmoles of P32-labelled MAP2-DTE probe, cross-linked with UV and digested with RNase. As shown in the input, the RNA probe is cross-linked to the 65 kDa and 90 kDa proteins MARTA2 and MARTA1. Immunoprecipitates performed with either MARTA2 or MARTA1 specific antibodies, selectively contain the respective RNA-labelled proteins. (c) Immunoprecipitations were performed with brain homogenates obtained from adult mice and antibodies directed against MARTA2 (C98, M-IP) as well as irrelevant IgGs (IgG-IP). In a western blot analysis, precipitates were probed for MARTA2 using the C99 antiserum. (d) RNA extracted from input material and immunoprecipitates was used as a template for RT-PCRs performed with primers specific for MAP2 and Shank1 mRNAs. MAP2 transcripts are detected in both input and M-IP, but not in the IgG-IP and water controls. In contrast, Shank1 transcripts are only detected in the input material. (e) Bar graph indicating the relative amounts of MAP2, αCamKII, β-actin and Shank1 mRNAs in M-IP versus IgG-IP as evaluated by real-time RT-PCR. Enrichment of MAP2 mRNA concentrations in M-IP versus IgG-IP (8.3-fold, 1.8 SEM, n = 12) is significantly different from minor increases of αCamKII (2.0-fold, 0.2 SEM, n = 5, p < 0.01, one-sided t-test), β-actin (2.4-fold, 1.1 SEM, n = 4, p < 0.05) and Shank1 (1.9-fold, 0.6 SEM, n = 2, p < 0.01) transcript levels in M-IP versus IgG-IP precipitates. Simple vertical lines specify standard error of mean. (f) Western blots performed with 40 μg protein each from different rat brain fractions, a nuclear fraction (nucl), crude lysate (CL), S90 cytosolic fraction (cytos) and polysome fraction (polys) and 20 μg from the RSW fraction (RSW). Utilized antibodies specifically recognize MARTA2 (C99, arrow head, upper part) or different marker proteins (indicated to the right by arrow heads) enriched in individual cell fractions: histone H3 (nucl), α-tubulin (cytos), ribosomal proteins L7A (RPL7A, polys) and S3 (RPS3, polys). The MARTA2 antiserum specifically detects a 63 kDa protein strongly enriched in the RSW fraction. Note that in the lane containing the RSW fraction only half the protein amount loaded in the other lanes is used. Under these conditions, the MARTA2 specific signal in the RSW fraction already saturates, whereas the protein is not visible in cytoplasmic fractions and only weakly detected in polysome and nuclear fractions.

Subcellular localization of MARTA2 in neurons

The subcellular distribution of MARTA2 in the adult rat brain was initially investigated by Western blot analysis of samples obtained by differential cell fractionation. Under conditions where MARTA2 was only weakly detected in the crude lysate (Fig. 2f, CL) and not yet visible in the cytosolic fraction (cytos), it was already clearly identified in nuclear (nucl) and polysome fractions (polys) and particularly strongly recognized in the RSW fraction (RSW; note that the RSW lane contains only half the protein amount of the cell fractions analysed in the other lanes). These findings are in agreement with the previously reported enrichment of a 65 kDa MAP2-DTE binding protein in the RSW fraction (Rehbein et al. 2000). Thus, cytosolic MARTA2 appears to preferentially associate with polysomes, from which it is efficiently released by high salt treatment. MARTA2 specific antisera were also used for immunocytochemistry. In 2-week-old primary hippocampal neurons, affinity-purified N54 serum recognizes MARTA2 in the somatodendritic cytoplasm (Fig. 3a). Along MAP2 protein-positive dendritic shafts, MARTA2 exhibits a granular distribution (Fig. 3a, insets). MARTA2 immunoreactivity was eliminated both in western blots (Fig. 2a, N54 block) and in primary neurons when the antiserum was pre-incubated with N-MARTA2-GST fusion protein (Fig. 3c). Similar results were obtained with polyclonal rabbit (C-98 and C-99) and monoclonal rat antibodies (data not shown). Semi-quantitative immunocytochemistry further revealed that in one-month-old primary hippocampal neurons about 84 ± 5.9% (n = 15) of MARTA2 resides in the somatodendritic region, whereas the rest of the protein is found in the nucleus (data not shown). By immunohistochemistry, MARTA2 was clearly identified throughout the soma of pyramidal neurons in the adult mouse hippocampus (Fig. 4a). Electron microscopy on neocortical regions not only confirmed the existence of MARTA2 in somata and dendrites of pyramidal cells (Fig. 4c, e) but also revealed its presence in dendritic spines of excitatory synapses (Fig. 4g), where it appeared to accumulate shortly underneath the post-synaptic density (PSD) (Fig. 4i). From a total of 650 analysed dendritic spines in the stratum radiatum of the hippocampal CA1 region that contained a clearly visible PSD, 348 spines (54%) showed a strong MARTA2 specific signal within the spine head and at the PSD, whereas the remaining 302 spines exhibited no or only a faint staining.

Figure 3.

MAP2-RNA trans-acting protein (MARTA)2 resides in the somatodendritic cytoplasm of primary neurons. (a) Two-week-old cultured hippocampal neurons were immunostained with polyclonal MARTA2 antiserum N54 and a monoclonal antibody against endogenous MAP2 proteins. MARTA2 (red) is present in the MAP2-positive (green) somatodendritic cytoplasm of primary neurons. Insets are enlargements of the boxed dendritic region to better visualize MARTA2 granules distributed along MAP2-positive dendritic shafts. (b) In neuronal somata, MARTA2 (red) resides in both the nucleoplasm (encircled by a white stippled line) and cytoplasm. MAP2 (green) is restricted to the somatodendritic cytoplasm. (c) When the MARTA2 antiserum N54 is pre-incubated with recombinant N-MARTA2- glutathione S-transferase (GST) antigen, the respective immunocytochemical signal on primary neurons is entirely lost (left panel), whereas the endogenous MAP2 protein in primary neurons is still detected (middle panel). MARTA2 and MAP2 proteins are visualized with AlexaFluor546- and AlexaFluor488-coupled goat secondary antibodies respectively. Micrographs were captured with a laser-scanning microscope. Scale bars: 20 μm.

Figure 4.

Somatodendritic and post-synaptic localization of MAP2-RNA trans-acting protein (MARTA)2 in the mouse brain. Subcellular localization of MARTA2 in pyramidal neurons of the adult mouse hippocampus (a) and neocortex (c, e, g and i) was determined via immunomicroscopy with peroxidase staining (a, c, e and g) and gold-labelling (i). (a) Light micrograph showing MARTA2 immunoreactivity particularly in somata of CA1 neurons. (b) Light micrograph of a negative control image taken from the CA1 region of a tissue section that was incubated with an antigen-blocked antiserum. (c, e, g) Electron microscopy revealed a patchy MARTA2 staining in the narrow cytoplasmic region surrounding the pyramidal cell nucleus (n in c), along dendritic shafts (arrows in e) and in dendritic spines (s in g) of excitatory synapses characterized by a thick post-synaptic density (PSD). (d, f, h) Electron micrographs from a negative control section treated with secondary antibodies only display regions corresponding to areas shown in c, e, g respectively. (i) Immunogold labelling indicates that MARTA2 accumulates in dendritic spines (s) shortly underneath the PSD. Note that pre-synaptic boutons (b in g and i) that contain multiple synaptic vesicles are not labelled. Scale bars: 20 μm (a and b), 2 μm (c and d), 1 μm (e and f) and 100 nm (g–i).

MARTA2 and KIF5c regulate dendritic MAP2 mRNA targeting

To examine a potential function of MARTA2 in dendritic mRNA trafficking, a truncated version of the RNA-binding protein was expressed in primary hippocampal neurons. The enhanced green fluorescent protein (EGFP)-tagged recombinant MARTA2 isoform contains only the four central KH domains (KH-EGFP). We assumed that while the truncated protein may still bind to the MAP2-DTE, it should no longer interact productively with other components of the mRNA localization machinery and therefore dominantly inhibit extrasomatic MAP2 mRNA targeting in vivo. In parallel experiments, we expressed EGFP alone or two other EGFP-tagged cytoplasmic RNA-binding proteins. Stau2RBD-EGFP contains the N-terminal 380 amino acid residues of rat Staufen 2 including four double-strand RNA binding domains (dsRBDs), but lacks the C-terminal portion of the protein. Over-expression of this fusion protein in primary neurons had been shown before to strongly reduce dendritic RNA levels (Tang et al. 2001). CPEB-EGFP contains full-length cytoplasmic polyadenylation element binding protein (CPEB). In cultured neurons, over-expression of full-length CPEB increased the dendritic levels of endogenous MAP2 transcripts (Huang et al. 2003). Neither Staufen 2 nor CPEB is structurally related to MARTA2. Vectors were transfected into one-week-old hippocampal neurons. Two days after transfection, the subcellular distribution of recombinant proteins and endogenous MAP2 transcripts was investigated via combined immunocytochemistry and in situ hybridization. Although the distribution of endogenous MAP2 transcripts was not strongly altered in the presence of high EGFP levels (Fig. 5a, b), over-expression of truncated MARTA2 entirely disrupted dendritic targeting of MAP2 mRNAs (Fig. 5c, d). In contrast, Stau2RBD-EGFP (Fig. 5e, f) or CPEB-EGFP (Fig. 5g, h) had no noticeable effect on MAP2 mRNA distribution. Levels of Stau2RBD-EGFP and CPEB-EGFP determined by semi-quantitative immunocytochemistry were 128% (± 35%, n = 38) and 101% (± 24%, n = 32) of the concentration of KH-EGFP (100 ± 25%, n = 29) respectively (data not shown). Because of the limited dynamic range of the laser-scanning microscope levels of EGFP alone were not exactly quantified, but are at least 200% (n = 37) of the concentration of KH-EGFP (data not shown). Compared with non-transfected neurons, the average density of MAP2 mRNA granules observed in the proximal 60 μm of the main dendritic branch was slightly reduced in cells over-expressing EGFP, Stau2RBD-EGFP or CPEB-EGFP (Fig. 5i). However, these granules were essentially absent in dendrites of neurons containing KH-EGFP. In comparison to non-transfected cells, the average density of somatic MAP2 transcript granules was moderately reduced in neurons containing EGFP or Stau2RBD-EGFP and somewhat more significantly diminished in cells over-expressing KH-EGFP (Fig. 5j). Although the lowered particle number in somata of KH-EGFP expressing neurons is likely to also result in a reduced mRNA granule density in dendrites, it cannot by itself explain the observed severe loss of dendritic transcript particles. Taken together, these data suggest that MARTA2 plays a critical role in both the assembly/stabilization and the dendritic trafficking of MAP2 mRNA granules. We also investigated whether loss of dendritic targeting of MAP2 mRNAs altered dendrite morphology, possibly through an effect on the distribution and/or levels of MAP2 proteins. A Sholl analysis of primary neurons, which were fixed two days after transfection did not reveal significant morphological differences between the dendritic arbors of KH-EGFP and EGFP transfected neurons respectively (Figure S3). This finding may at least in part be explained by the rather long half-life of MAP2 in the somatodendritic compartment of neurons (Okabe and Hirokawa 1989). Finally, over-expression of neither KH-EGFP (Fig. 6a, b), EGFP alone (Fig. 6c, d) nor Stau2RBD-EGFP (Fig. 6e, f) altered the subcellular distribution and concentration of the entire pool of polyadenylated mRNAs, thus excluding a general effect of KH-EGFP on neuronal mRNA processing and transport. Taken together, the above data imply that MARTA2 plays a key role in selective dendritic targeting of MAP2 mRNAs.

Figure 5.

Truncated MAP2-RNA trans-acting protein (MARTA)2 disrupts dendritic targeting of endogenous MAP2 mRNA granules. Enhanced green fluorescent protein (EGFP) (a, b), KH-EGFP (c, d), Stau2RBD-EGFP (e, f) and CPEB-EGFP (g, h) were expressed in 1-week-old hippocampal neurons. Two days after transfection, endogenous MAP2 transcripts were visualized by fluorescent in situ hybridization (FISH) with using a digoxygenin-labelled antisense RNA probe, followed by Cy3-coupled anti-digoxygenin antibodies (red). Endogenous MAP2 and various fusion proteins were detected via immunocytochemistry with anti-MAP2 and anti-EGFP antibodies and secondary antibodies coupled to Marina Blue (blue) and AlexaFluor488 (green) respectively. Panels a, c, e and g are triple-stained images, in which transfected (turquoise) and non-transfected neurons (blue) are easily distinguished. Panels b, d, f and h are corresponding overlays of red and blue channels only, to allow clearer visibility of MAP2 mRNA particles (magenta) along dendrites (blue) of transfected neurons. Higher magnification images of boxed areas are shown as insets to better visualize RNA granules distributed along MAP2-positive dendritic shafts. In non-transfected neurons, endogenous MAP2 mRNA granules are dispersed along dendritic shafts (e.g. arrowheads in lower left inset in d). However, upon the over-expression of truncated MARTA2, these transcript particles totally disappear from dendrites (arrows in upper right inset in d). In contrast, in neurons over-expressing EGFP (a, b), Stau2RBD-EGFP (e, f) or CPEB-EGFP (g, h), MAP2 mRNA granules are normally distributed along dendrites. Fluorescence micrographs were captured with a conventional fluorescence microscope. Scale bars: 20 μm. i and j) Bar graphs summarizing the mean number (plus SEM) of endogenous MAP2 mRNA granules in the proximal 60 μm of the main dendritic shaft (i) and in 100 μm2 of somata (j) of non-transfected hippocampal neurons and cells over-expressing different fusion proteins (as indicated left of the respective bar). Compared with non-transfected neurons, dendritic granule density was slightly reduced in cells over-expressing EGFP (81.2 ± 23.8% of non-transfected cells, n = 50), Stau2RBD-EGFP (74.4 ± 26,6%, n = 50) or CPEB-EGFP (70.3 ± 22,0%, n = 50). However, these granules were essentially absent in dendrites of neurons containing KH-EGFP (0 ± 0%, n = 50). The density of somatic MAP2 transcript particles was moderately reduced in neurons containing EGFP (82.8 ± 17.8% of non-transfected cells, n = 31) and Stau2RBD-EGFP (83.4% ± 16.1%, n = 27) and somewhat more significantly diminished in cells over-expressing KH-EGFP (76.5 ± 19.8%, n = 31). One-sided t-tests (level of significance α = 0,05; p < 0,001 is marked as ***). Simple vertical lines specify SEM. For abbreviations of protein names see text. Over-expression of truncated MARTA2 leads to a loss of dendritic MAP2 mRNA granules.

Figure 6.

Truncated MAP2-RNA trans-acting protein (MARTA)2 does not affect subcellular distribution of the bulk of polyadenylated mRNAs. KH-EGFP (a, b), enhanced green fluorescent protein (EGFP) (c, d) and Stau2RBD-EGFP (e, f) were expressed in 1-week-old hippocampal neurons. After 2 days, polyadenylated transcripts were visualized by FISH using an digoxygenin-labelled oligo(dT)50 probe, followed by Cy3-coupled anti-digoxygenin antibodies (red). Endogenous MAP2 and recombinant proteins were immunocytochemically detected with anti-MAP2 and anti-EGFP antibodies and secondary antibodies coupled to Cy5 (blue) and AlexaFluor488 (green) respectively. Panels a, c and e are triple-stained images, whereas panels b, d and f represent corresponding overlays of red and blue channels only. None of the recombinant proteins alters the concentration and distribution of polyadenylated mRNAs. Fluorescence micrographs were captured with a laser-scanning microscope. Scale bars: 20 μm.

Dendritic mRNA trafficking is thought to occur along microtubules with the help of microtubule-based motor proteins (Hirokawa 2006; Kindler and Kreienkamp 2012). In agreement with this finding, the cargo-binding domain of the microtubule-based kinesin family motor KIF5 associates with ribonucleoprotein (RNP) complexes, which contain at least four different dendritic mRNAs (Kanai et al. 2004; Falley et al. 2009; Kindler et al. 2009). To determine whether MAP2 transcripts may be transported as part of a KIF5-RNP complex, we used the minimal RNP-binding domain of KIF5c (Kanai et al. 2004) for GST-pull-down experiments with mouse brain extracts (Fig. 7a). RNA extracted from purified samples was analysed by real-time RT-PCR (Fig. 7b). In contrast to hypoxanthine-guanine phosphoribosyltransferase (HPRT) transcripts that were only slightly enriched in KIF5c versus GST samples, MAP2 mRNAs were significantly enriched in the kinesin cargo sample. To verify these results, we performed immunoprecipitations with KIF5 specific antibodies (K-IP) and irrelevant rabbit IgGs (IgG-IP) utilizing brain homogenates. RNA extracted from immunoprecipitates was analysed by real-time RT-PCR with oligonucleotide pairs specific for MAP2 and phosphoglycerate kinase 2 (Pgk2) mRNAs respectively. While MAP2 transcripts were more than seven-fold enriched in K-IP versus IgG-IP, Pgk2 mRNA concentrations in K-IP were close to background levels observed in IgG-IP samples (Fig. 7c). Thus, KIF5c associated RNP complexes contain MAP2 mRNAs. To further investigate a possible role of KIF5c in dendritic MAP2 mRNA targeting, we over-expressed a dominant-negative version of the motor protein (dnKIF5c-EGFP) (Schepis et al. 2007) in primary hippocampal neurons. Indeed, dnKIF5c-EGFP severely disrupted dendritic MAP2 mRNA targeting (Fig. 8a, b). In contrast, over-expression of the dynein-associated protein dynamitin (Dynamitin-EGFP; 109 ± 31.1% of dnKIF5c-EGFP levels, n = 26, data not shown), which is known to interfere with the cellular function of cytoplasmic dynein (Burkhardt et al. 1997), or EGFP (> 250% of dnKIF5c-EGFP levels, n = 37, data not shown) alone did not interfere with extrasomatic transcript trafficking (Fig. 8c, d and e, f respectively). Compared with non-transfected cells, the average density of MAP2 mRNA granules in the proximal 60 μm of the main dendritic branch was only slightly diminished in neurons synthesizing EGFP or Dynamitin-EGFP, while they essentially disappeared from dendrites of neurons expressing dnKIF5c-EGFP (Fig. 8g). Somatic granule density was moderately decreased in neurons synthesizing EGFP, but somewhat more notably diminished in cells over-expressing dnKIF5c-EGFP (Fig. 8h). Taken together, these data suggest that the microtubule-based motor protein KIF5c is involved in both the formation/stabilization and the dendritic translocation of MAP2 mRNA granules.

Figure 7.

MAP2 mRNAs are part of KIF5c associated mRNA transport complexes. (a) Schematic diagram of the motor protein KIF5c (upper part) indicating the motor domain (grey box), hinge region (thick vertical line) and minimal binding site for the association with RNP complexes (striped box), as well as the glutathione S-transferase (GST)-KIF5c fusion protein (lower part, oval indicates GST domain). (b) Affinity chromatography was performed with cleared mouse brain lysates and GST or GST-KIF5c coupled glutathione sepharose respectively. RNA extracted from purified samples was analysed by real-time RT-PCR using specific primers for MAP2 and hypoxanthine-guanine phosphoribosyltransferase (HPRT) mRNAs. The bar graph indicates the enrichment of a given transcript in GST-KIF5c versus GST purified samples. Simple vertical lines specify standard deviations. The strong enrichment of MAP2 mRNAs (7.4 ± 3.9-fold enrichment, n = 6) compared with the HPRT control (2.3 ± 3.9-fold, n = 6) is statistically significant (p = 0.0226, one-sided t-test). (c) Immunoprecipitations were performed with KIF5c specific antibodies (K-IP) and irrelevant rabbit IgGs (IgG-IP) and mouse brain homogenates. Total RNA extracted from immunoprecipitates was analysed by real-time RT-PCR to determine the relative concentration of MAP2 and phosphoglycerate kinase 2 (Pgk2) mRNAs in K-IP relative to IgG-IP precipitates respectively. MAP2 transcripts are more than 7-fold enriched in K-IP versus IgG-IP (7.4 ± 1.7-fold, n = 4; p = 0.0065, one-sided t-test). In contrast, Pgk2 transcript concentrations in K-IP are close to background levels observed in IgG-IP samples (2.4 ± 2.3-fold enrichment, n = 3).

Figure 8.

KIF5c mediates extrasomatic trafficking of endogenous MAP2 mRNA granules. dnKIF5c-enhanced green fluorescent protein (EGFP) (a, b), Dynamitin-EGFP (c, d) and EGFP (e, f) were expressed in 1-week-old primary neurons. Two days after transfection, endogenous MAP2 transcripts were visualized by FISH using a digoxygenin-labelled antisense probe, followed by Cy3-coupled anti-digoxygenin antibodies (red). Endogenous MAP2 and recombinant fusion proteins were immunostained with anti-MAP2 and anti-EGFP antibodies and secondary antibodies coupled to Marina Blue (blue) and AlexaFluor488 (green) respectively. Panels a, c and e are triple-stained images, whereas panels b, d and f are corresponding merged red and blue channels only. In the latter, MAP2 mRNA particles (magenta) along dendrites (blue) are clearly visible. Higher magnification images of boxed areas are shown as insets. In non-transfected neurons, endogenous MAP2 mRNA granules are scattered along dendritic shafts (b, arrowheads). Over-expression of dnKIF5c-EGFP disrupts dendritic targeting of these transcripts (b arrows). In contrast, over-expression of Dynamitin-EGFP (c, d) or EGFP (e, f) does not interfere with extrasomatic trafficking of mRNA granules. Fluorescence micrographs were captured with a conventional fluorescence microscope. Scale bars: 20 μm. g and h) Bar graphs showing the mean number (plus SEM) of endogenous MAP2 mRNA particles in the proximal 60 μm of the main dendritic shaft (g) and in a 100 μm2 somatic region (h) of non-transfected primary neurons and cells over-expressing different fusion proteins (as indicated left of the respective bar). Compared with non-transfected cells, the average density of MAP2 mRNA granules in dendrites was only slightly diminished in neurons synthesizing EGFP (81.2 ± 23.8% of non-transfected cells, n = 50) or Dynamitin-EGFP (97.1 ± 27.9%, n = 50), while these RNP particles disappeared from dendrites of neurons expressing dnKIF5c-EGFP (0 ± 0%, n = 50). Somatic granule density was only moderately decreased in neurons synthesizing EGFP (82.8 ± 17.8% of non-transfected cells, n = 31), but somewhat more notably diminished in cells over-expressing dnKIF5c-EGFP (63.5 ± 20.4%, n = 25). One-sided t-tests (p < 0,001 is marked as ***). Simple vertical lines specify SEM. For abbreviations of protein names see text. Over-expression of dnKIF5c-EGFP results in a loss of MAP2 mRNA granules along dendrites.

Discussion

Cytoplasmic mRNA trafficking involves a specific interaction of cis-acting targeting elements with trans-acting proteins (Cajigas et al. 2010; Bullock 2011; Jansen and Niessing 2012; Kindler and Kreienkamp 2012). Herein, we report the purification of MARTA2, a rat brain RNA-binding protein that similar to MARTA1 exhibits highly selective in vitro RNA-binding activity for the MAP2-DTE (Rehbein et al. 2002), a cis-acting signal sequence of MAP2 mRNAs mediating selective targeting into dendrites (Blichenberg et al. 1999). The identity of the purified 63 kDa protein with MARTA2 previously characterized by UV CLAs (Rehbein et al. 2000) was confirmed by immunoprecipitation assays. Antibodies raised against the purified 63 kDa protein specifically precipitated MARTA2 from UV CLAs, but did not react with MARTA1. MARTA2 is homologous to human FBP3 (Davis-Smyth et al. 1996). Together with rat MARTA1 (Rehbein et al. 2000, 2002) (also referred to as FBP2/KSRP (Davis-Smyth et al. 1996; Briata et al. 2011), ZBP2 (Gu et al. 2002) and VgRBP71 (Kolev and Huber 2003) in human, chicken and frog respectively) and human FBP1 (Davis-Smyth et al. 1996) it constitutes the FBP family (for an overview of synonyms used for vertebrate FBP family members see Table S2). Its three principal members are characterized by a conserved central region that contains four KH domains (Davis-Smyth et al. 1996) utilized for both single-stranded DNA (Braddock et al. 2002a,b; Chung et al. 2006) as well as RNA-binding (Rehbein et al. 2000, 2002; Chen et al. 2001; Gu et al. 2002; Kolev and Huber 2003; Gherzi et al. 2004; Briata et al. 2005, 2011; Linker et al. 2005; Bolognani and Perrone-Bizzozero 2008; Trabucchi et al. 2009; Danckwardt et al. 2011). The high sequence similarity between MARTA2 and MARTA1 in the KH domain region most likely accounts for the essentially identical RNA-binding specificity that both proteins exhibit in UV CLAs (Rehbein et al. 2000). Similarly, all three human FBP family members bind to the single-stranded FUSE, a cis-acting genomic DNA element regulating c-myc expression (Davis-Smyth et al. 1996; Chung et al. 2006). Consequently, FBP family members are multifunctional, playing roles in transcription (Chung and Levens 2005; Chung et al. 2006; Levens 2008), splicing (Briata et al. 2011), RNA processing (Trabucchi et al. 2009; Danckwardt et al. 2011), RNA editing (Briata et al. 2011), translational control (Kolev and Huber 2003), mRNA turnover (Chen et al. 2001; Gherzi et al. 2004; Briata et al. 2005; Linker et al. 2005; Bolognani and Perrone-Bizzozero 2008) and cytoplasmic mRNA targeting as shown here (Rehbein et al. 2000, 2002; Gu et al. 2002).

Here, we show that distinct from FBP1 and MARTA1 orthologues, which are highly concentrated in the nucleus (Gu et al. 2002; Kroll et al. 2002; Rehbein et al. 2002; Chung and Levens 2005; Chung et al. 2006), MARTA2 resides in both cytoplasm and nucleus. This is in agreement with the subcellular distribution of its human orthologue (Chung et al. 2006) and the finding that both FBP1 and MARTA1 orthologues contain a classical nuclear localization signal, which is disrupted in MARTA2 orthologues (Davis-Smyth et al. 1996; Rehbein et al. 2002; Chung et al. 2006) (Figure S1). In different neuronal cell types of the adult mouse brain and primary rat hippocampal neurons, we detected MARTA2 in granules, which are dispersed throughout the somatodendritic cytoplasm including dendritic spines. Noteworthy, several dendritically localized mRNAs, including MAP2 transcripts as is shown in this study, as well as trans-acting factors involved in the mRNA localization process have been shown to reside in cytoplasmic granules (Bramham and Wells 2007; Dahm et al. 2007; Cajigas et al. 2010; Kindler and Kreienkamp 2012). Taken together with the specific in vitro binding of MARTA2 to the MAP2-DTE (Rehbein et al. 2000, 2002) and the association of MARTA2 and MAP2 mRNAs in the mouse brain in vivo as shown herein, our findings are in agreement with a function of MARTA2 in dendritic MAP2 mRNA targeting. Most importantly, a strong dominant-negative effect was evoked by the isolated central KH domains of MARTA2. In primary hippocampal neurons over-expressing KH-EGFP, the density of MAP2 mRNA granules slightly decreased in somata, while these mRNA particles essentially disappeared from dendrites. These data imply that MARTA2 has a slight effect on the assembly/stability of MAP2 mRNA granules. Interestingly, MARTA1 orthologues have been shown to regulate the stability of a particular set of mRNAs (Gherzi et al. 2004). More importantly, our data suggest that MARTA2 strongly promotes dendritic translocation of MAP2 transcripts. In agreement with this finding, closely related MARTA1 orthologues bind to cis-acting targeting elements of MAP2 and β-actin transcripts and appear to play a role in the subcellular targeting of these mRNAs (Rehbein et al. 2000, 2002; Gu et al. 2002). Similar to the experiments presented herein, over-expression of truncated chicken MARTA1 has a dominant-negative effect on β-actin mRNA localization in neurons and fibroblasts (Gu et al. 2002). Thus, both MARTA2 and MARTA1 orthologues appear to be involved in cytoplasmic mRNA trafficking. Chicken MARTA1 was recently shown to facilitate nuclear binding of the trans-factor ZBP1 to β-actin transcripts (Pan et al. 2007). It is therefore tempting to speculate that MARTA1 may also support the nuclear recognition of MAP2 mRNAs by MARTA2. At least two independent observations support this hypothesis. First, in Xenopus oocytes Vg1 mRNA localization involves transcript recognition in the nucleus and the formation of a nuclear RNP complex, which is exported and remodelled in the cytoplasm via the exchange of transport factors (Kress et al. 2004). Second, the nuclear processing history of oskar transcripts in Drosophila oocytes has been shown to determine subsequent cytoplasmic mRNA trafficking events (Hachet and Ephrussi 2004). Taken together, the currently available data imply that similar to other mRNA transport events dendritic targeting of MAP2 transcripts is a multi-step process, in which the localized mRNA utilizes several trans-factors in a sequentially orchestrated manner.

The dominant-negative effect of KH-EGFP on dendritic mRNA targeting described herein is specific as under the same conditions neither EGFP alone nor Stau2RBD-EGFP and CPEB-EGFP disrupted extrasomatic MAP2 mRNA trafficking in primary neurons. Furthermore, an unspecific inhibitory effect of truncated MARTA2 on mRNA distribution and stability can be excluded, as it did not significantly alter the subcellular distribution or concentration of the entire pool of polyadenylated mRNAs. The latter finding also indicates that a large set of dendritic transcripts does not require MARTA2 for its extrasomatic targeting. In agreement with this hypothesis, MARTA2 exhibited no in vitro binding activity to several other dendritic mRNAs (Rehbein et al. 2000) and in vivo did not associate with αCamKII, β-actin or Shank1 transcripts as shown herein. In this context, it is interesting to note that among all dendritic transcripts, MAP2 mRNAs are unique inasmuch as their levels continuously decrease from proximal dendrites towards dendritic tips, while all other known dendritic transcripts appear to be more evenly distributed throughout the entire dendritic arbor (Kindler and Kreienkamp 2012). Nonetheless, in future experiments, it will be interesting to identify further transcripts that are associated with MARTA2 in vivo and determine whether MARTA2 plays a role in subcellular trafficking of these transcripts.

In cultured neurons, over-expression of full-length CPEB increased the dendritic levels of endogenous MAP2 transcripts (Huang et al. 2003). Yet, in a similar approach used herein recombinant CPEB-EGFP resulted in a slightly decreased density of dendritic MAP2 mRNA granules. Analysis of existing CPEB deficient mice may be necessary to clearly determine the putative role of CPEB in dendritic MAP2 mRNA trafficking. Over-expression of Stau2RBD-EGFP, which contains all four dsRBDs of Stau2 while lacking the C-terminal part of the protein, strongly reduced dendritic RNA levels in primary neurons (Tang et al. 2001). However, our findings presented herein show that Stau2RBD-EGFP does not disrupt extrasomatic targeting of endogenous MAP2 transcripts and does not strongly alter dendritic targeting of the entire pool of polyadenylated mRNAs. Identification of in vivo RNA targets and analysis of Stau2 deficient mice will be helpful to further analyse the significance of this RNA-binding protein for subcellular mRNA trafficking.

Recently, brain RNP complexes associated with the cargo-binding domain of microtubule-based kinesin motor KIF5 have been identified (Kanai et al. 2004). Along with several RNA-binding proteins, such as Staufen 1 and fragile X mental retardation protein (FMRP), the particles contain αCamKII, Arc, Shank1 and Jacob transcripts (Kanai et al. 2004; Falley et al. 2009; Kindler et al. 2009), four prominent dendritically localized mRNAs. These and additional functional data imply that the identified RNP granules are used to transport neuronal mRNAs into dendrites (Kanai et al. 2004). In agreement with this hypothesis, we show here that KIF5 associated transport complexes also contain MAP2 transcripts and that inhibition of KIF5 function in primary neurons disrupts extrasomatic targeting of endogenous MAP2 mRNAs. Bidirectional movement of RNP granules in neuronal dendrites observed by Kanai et al. (2004) further suggests that KIF5 transports dendritic mRNAs in coordination with minus-end motors such as cytoplasmic dynein. Similarly, different RNA-binding proteins such as CPEB, FMRP and Staufen, which have been implicated in dendritic mRNA targeting appear to be associated with both KIF motors and cytoplasmic dynein (Huang et al. 2003; Brendel et al. 2004; Kanai et al. 2004; Ling et al. 2004; Villace et al. 2004; Davidovic et al. 2007; Jeong et al. 2007; Dictenberg et al. 2008; Bianco et al. 2010). Also, in several non-neuronal systems cytoplasmic RNP targeting appears to involve both plus-and minus-end directed microtubule motors (Bullock 2011; Baumann et al. 2012; Jansen and Niessing 2012; Vollmeister et al. 2012). However, in our study disruption of dynein function through the over-expression of recombinant dynamitin did not reveal any evidence that this minus-end directed motor is involved in dendritic targeting of endogenous MAP2 transcripts in neurons. This may indicate that MAP2 mRNA granules described herein represent a particular subfraction of KIF5 bound transport complexes that are not associated with cytoplasmic dynein. Alternatively, a longer inhibition of dynein function in neurons might be necessary to significantly impair extrasomatic MAP2 mRNA trafficking. Further experiments are also required to determine the molecular composition and biogenesis of RNP complexes containing MAP2 mRNAs and MARTA2, define signalling pathways controlling the interaction of individual RNP components and investigate if the association of MARTA2 with motor proteins involves particular adapters as it has been shown for FMRP (Bianco et al. 2010).

Acknowledgements

We thank Richard B. Vallee (Columbia University Medical Center) for generously providing the dynamitin expression vector. Financial support from Deutsche Forschungsgemeinschaft (Ki488/2-6 to M.R./S.K.; KR1321/4-1 to H.-J.K./S.K) and Fritz Thyssen Stiftung (Az. 10.05.2.185 to S.K./D.R.) is acknowledged. This article is in part based on doctoral studies in the Faculty of Biology (K.H.Z./J.S) and Medicine (C.S.), University of Hamburg. The authors have no conflicts of interest to declare.

Ancillary