Makorin Ring Zinc Finger Protein 1 (MKRN1), a Novel Poly(A)-binding Protein-interacting Protein, Stimulates Translation in Nerve Cells*

Background: Synaptic activity induces translation of mRNAs in dendrites of neurons. Results: Makorin 1 (MKRN1) interacts with poly(A)-binding protein, stimulates translation, accumulates in neuronal dendrites after plasticity-inducing stimuli, and is associated with dendritic mRNAs. Conclusion: MKRN1 has the potential to locally control the translation of dendritic mRNAs at synapses. Significance: MKRN1 is a novel positive regulator of translation. The poly(A)-binding protein (PABP), a key component of different ribonucleoprotein complexes, plays a crucial role in the control of mRNA translation rates, stability, and subcellular targeting. In this study we identify RING zinc finger protein Makorin 1 (MKRN1), a bona fide RNA-binding protein, as a binding partner of PABP that interacts with PABP in an RNA-independent manner. In rat brain, a so far uncharacterized short MKRN1 isoform, MKRN1-short, predominates and is detected in forebrain nerve cells. In neuronal dendrites, MKRN1-short co-localizes with PABP in granule-like structures, which are morphological correlates of sites of mRNA metabolism. Moreover, in primary rat neurons MKRN1-short associates with dendritically localized mRNAs. When tethered to a reporter mRNA, MKRN1-short significantly enhances reporter protein synthesis. Furthermore, after induction of synaptic plasticity via electrical stimulation of the perforant path in vivo, MKRN1-short specifically accumulates in the activated dendritic lamina, the middle molecular layer of the hippocampal dentate gyrus. Collectively, these data indicate that in mammalian neurons MKRN1-short interacts with PABP to locally control the translation of dendritic mRNAs at synapses.

Neurons are considered the most complex cell type of higher organisms. To ensure appropriate function within an intricate network, their molecular constituents have to be synthesized, modified, sorted, and degraded in due time. Synapses in partic-ular are constantly modified in an activity-dependent manner (1), and synaptic function requires a well regulated interplay of mechanisms to maintain accurately balanced protein levels (2). In part, this is achieved by local translation of mRNAs that are transported to dendrites to await translational activation (3)(4)(5)(6). RNA transport is specified by sequences called "zip codes," and proteins that collectively build up messenger ribonucleoprotein complexes referred to as RNA granules (7). These complexes are most likely remodeled to allow for translational activation, translational silencing, or degradation of mRNA molecules. Different types of granules exist with distinct contents of proteins and RNAs (8 -12), yet further work is needed to entirely reveal the diversity and functional significance of RNA granules in neurons.
The cytoplasmic poly(A)-binding protein (PABP) 2 associates with the poly(A) tail of mRNAs and thus promotes translation (13). Yet the protein also interacts with various non-poly(A) sequences including the zip code of the dendritically localized vasopressin mRNA (14,15). As a part of distinct multiprotein complexes, PABP is involved in translational stimulation (16 -18) and silencing (19,20), RNA localization (21), and mRNA turnover (22). In neurons, PABP localizes to RNA granules (8,11,23), indicating that it is involved in dendritic mRNA metabolism.
To better understand the molecular action of PABP in regulating mRNA fate in nerve cells, we have employed yeast twohybrid techniques to identify binding partners required in sup-port of its function. Here, we report the characterization of a novel PABP-interacting protein, RING zinc finger protein Makorin-1 (MKRN1) (24). The mkrn1 gene belongs to family members that encode putative RNA-binding proteins. MKRN1 is a modular protein with distinct arrays of C 3 H zinc finger (ZF) motifs, a ZF structure with unusual cysteine/histidine spacing, and a RING domain typically found in E3 ubiquitin ligases (25). Apparently, MKRN1 exhibits divergent functions both in the cell nucleus and the cytoplasm. As an E3 ubiquitin ligase it acts on itself and the catalytic subunit of human telomerase reverse transcriptase (26), p53 and p21 (27). Furthermore, MKRN1 modulates RNA polymerase II-mediated transcription (28) and may play a role in mRNA decay (29).
In our yeast two-hybrid screen with PABP bait, we have exclusively isolated a shorter isoform (called MKRN1-short) of hitherto unknown function encoded by exons 1-5 of the mkrn1 gene. We show that this protein is the major isoform in rat brain. MKRN1-short expression in forebrain neurons is more abundant than elsewhere in the brain, and the protein resides in both the nucleus as well as the cell body and dendrites. MKRN1-short contains a PAM2 (PCI/PINT associated module 2)-like motif that mediates its interaction with PABP in an RNA-independent manner. PAM2 motifs are found in several PABP-interacting proteins, for example the PABP-interacting protein 1 (PAIP1) and PAIP2 (30) that affect translation in a positive and negative manner, respectively (31,32). MKRN1short exerts a strong positive effect on translation when it is tethered to a reporter mRNA in primary neurons. In vivo, MKRN1-short accumulates in dendrites after induction of long term potentiation (LTP), a cellular process that requires de novo protein synthesis (33,34). Taken together, these findings suggest that in mammalian brain neurons MKRN1-short functions as a modulator of local protein synthesis in dendrites.

Experimental Animals
Wistar-or Sprague-Dawley rats were used. Animals were bred and handled in accordance with national guidelines for animal welfare.

Electrophysiological Manipulation and Brain Tissue Preparation
Adult male Sprague-Dawley rats (250 -500 g; Charles River) were deeply anesthetized with urethane (1.25 g/kg body weight, subcutaneously initially, and additional injections as needed). Surgery and stimulation procedures were performed as described (35). Briefly, stimulating electrodes were placed in the angular bundle of the medial perforant path. Recording microelectrodes were placed in the dorsal blade of the granule cell layer. High frequency stimulation was applied for 2 h to maximally evoke population spikes and induce robust LTP in granule cells as has been described (36). One train consisted of 8 pulses (500 A, 0.1-ms pulse duration) of 400 Hz once per 10 s. Immediately after the end of the stimulation, rats were transcardially perfused with 4% paraformaldehyde.

Cloning Procedures
DNAs encoding PABP, MKRN1, DDX6, and Shank3 were either amplified by PCR techniques, or constructs were generated by subcloning procedures. Constructs generated by PCR were subjected to DNA sequencing. The clones employed in this study are summarized in supplemental Table 1. The following vectors were used: pGEX-6P-3 (GE Healthcare), pGBKT7 (Clontech), pcDNA6/myc-His (Invitrogen), pEGFP-C (Clontech). pN22-C1 and pN22-FLAG3-C1 are derivatives of pEGFP-C1 (Clontech), in which the EGFP cDNA has been replaced by regions encoding 22 amino acid residues from the N protein of the phage (N22; 37) and a fusion protein consisting of N22 and three consecutive FLAG epitopes, respectively. The eukaryotic expression vector pinFiRein-boxB16B is based on the previously described plasmid pFiRe-basic (38). It contains two recombinant genes, both of which are controlled by independent CMV immediate-early promoters, contain a chimeric intron from pFN21 (Promega) upstream of the coding region, and encode Photinus (PhoLuc) and Renilla luciferase (RenLuc), respectively. In their 3Ј-UTRs, PhoLuc transcripts include 16 consecutive copies of the 15-nucleotide RNA hairpin termed box B that specifically interacts with the N22 domain (37). The 3Ј-UTR was chosen for box B insertions because this part of mRNAs often regulates translation (39). pcDNA-T7 is a pcDNA3 derivative (Invitrogen) containing a T7 tag-encoding sequence (kindly provided by Dr. Hans-Jürgen Kreienkamp, University Medical Center Hamburg-Eppendorf, Hamburg, Germany).
Rat hippocampal and cortical neurons were prepared from rat embryos (E19) according to published procedures (40) with the following modifications; neurons were grown in Neurobasal A medium containing B27 supplement (Invitrogen) without a glial feeder layer, and transfections were performed at day 7 in vitro (7 DIV) using calcium phosphate co-precipitation techniques (40). For luciferase assays and quantitative PCR analyses (see below), respective constructs were transfected in at least five independent series of transfections.

Yeast Two-hybrid Technique
Rat PABP cDNA was cloned into pGBKT7 (see "Cloning Procedures"). PABP-interacting proteins were identified by using the matchmaker yeast-pretransformed human brain cDNA library (Clontech) as outlined in the manufacturer's instruction manual.

RNA Extraction and Purification
Total RNA from cultured cells and tissues was prepared with PeqGold RNAPure (PeqLab) according to the manufacturer's instructions followed by a cleanup using RNeasy (Qiagen). RNA purity and concentration was determined spectrophotometrically.
cDNA Synthesis 2-5 g of total RNA were reverse-transcribed using Superscript II RT (Invitrogen) or RevertAid Premium Reverse Transcriptase (Fermentas) and 0.5 g of oligo(dT) according to manufacturer instructions. Subsequently, RNA was digested with 2 units of RNase H (Invitrogen or Fermentas) for 20 min at 37°C. For quantitative PCR analyses (see below) RNAs were subjected to two independent cDNA syntheses.

Real-time PCR with TaqManா Probes
A TaqMan PCR reaction (20 l) included 1.4 l of cDNA, 1ϫ TaqMan Universal PCR Master Mix (Applied Biosystems), and either the TaqMan Gene Expression Assays Ren-Luc (Applied Biosystems custom-made part no./assay ID 4332079/AIGJPCV) or PhoLuc (Applied Biosystems custommade part no./assay ID 4332079/AIFAQ6N), respectively. Taq-Man PCR was carried out in a Sequence Detection System Applied Biosystems PRISM 7900 HT thermocycler using the standard program (2 min at 50°C, 10 min at 95°C, 40 cycles of 15 s at 95°C and 1 min at 60°C) and analyzed with Applied Biosystems PRISM 7900 HT software. Data were evaluated using the comparative C T (threshold cycle) method for the relative quantification of the amount of PhoLuc cDNA normalized to RenLuc cDNA as a reference, given by 2 Ϫ⌬⌬C T . The C T parameter is defined as the fractional cycle number at which the fluorescence exceeds a fixed threshold value. Each cDNA was analyzed in triplicate with both TaqMan Gene Expression Assays. Non-template and no reverse transcriptase controls were included in each run.

SYBRா Green Real-time RT-PCR
Semiquantitative determination of mRNA levels in immunoprecipitated messenger ribonucleoproteins (see below) was performed with the QuantiTect TM SYBR Green RT-PCR kit (Qiagen) and QuantiTect TM Primer assays Arc (QT00373086, Qiagen), MAP2 (QT01084244), and ␤-tubulin 3 (QT00188818) according to the manufacturer's protocol on a Rotor-Gene 3000 cycler (Corbett/Qiagen). PCR reactions (20 l each) were done in duplicate and included 1 l of RNA solution. Nontemplate and no reverse transcriptase controls were included in each run. Data analysis was performed using REST (relative expression software tool) 2008 software for group-wise comparison and statistical analysis of relative expression results in real-time PCR (41).

Fractionation of Polysomes
Separation of polysomes from freshly prepared rat hippocampal homogenates by sucrose gradient ultracentrifugation (15-45% sucrose) was done as described (42). Individual 1-ml gradient fractions were collected, immediately frozen in liquid nitrogen, and stored at Ϫ80°C until use. For RNA precipitation, 1 ⁄ 5 of each fraction was incubated in 70% ethanol, 10 mM EDTA, and 0.044 mg/ml yeast tRNA for 2 h at 4°C and subsequently centrifuged for 15 min, 12,000 ϫ g, and 4°C. RNA pellets were dissolved in RNase-free H 2 O and further purified as described above.
Protein Extracts from Transfected Primary Neurons for Messenger Ribonucleoprotein Immunoprecipitations-Neurons were washed twice with ice-cold Hanks' balanced salt solution. Per 10-cm dish, cells were lysed in 0.425 ml of lysis buffer (120 mM NaCl, 50 mM HEPES, pH 7.5, 5 mM EDTA, 0.5% (w:v) Triton X-100) containing protease inhibitors (Complete, Roche Applied Science and 200 units/ml RiboLock RNase inhibitor (Fermentas). Lysates were incubated on ice for 15 min and centrifuged for 15 min at 4°C and 13,000 ϫ g. Aliquots for RNA extraction (10%; input RNA) and Western blot analyses (2.5%, input protein) were withdrawn. The remainder was subjected to immunoprecipitation (see below).
Protein concentration was determined using the Protein Assay Reagent (Bio-Rad) with bovine serum albumin as standard. Proteins were snap-frozen in liquid nitrogen and stored at Ϫ80°C until use.

Immunoprecipitation
Immunoprecipitations of proteins dissolved in radioimmunoprecipitation assay buffer (see below) were usually performed with 700 -1000 g of proteins at a concentration of ϳ1 g/l.
For FLAG-agarose, T7-agarose, and GFP-Trap_A (Chromotek), 20 l of monoclonal mouse anti-FLAG M2-or anti-T7agarose or 10 l of GFP-Trap_A/reaction were washed 5ϫ with radioimmunoprecipitation assay buffer. Protein lysate was added and rotated overnight at 4°C. Supernatant was removed, and the agarose was washed 3ϫ with 1 ml of radioimmunoprecipitation assay buffer. Bound proteins were eluted with 1ϫ concentrated Laemmli sample buffer at 95-99°C for 5 min. In some cases immunoprecipitations were performed in the presence of RNase A as described (8).
Immunoprecipitations of recombinant proteins from transfected primary neurons were performed with 40 l of GFP-Trap_A/reaction. Lysates from two 10-cm dishes were combined, NaCl was added (final concentration, 200 mM), and lysates were incubated overnight at 4°C with constant rotation. Beads were washed 6ϫ with 500 l each of lysis buffer containing 200 mM NaCl. Aliquots of beads (ϳ5%) were withdrawn for Western blot analysis. From the remaining beads RNA was eluted with 100 l of elution buffer (200 mM sodium acetate, pH 5.5, 1 mM EDTA, 1% (w:v) SDS) for 10 min at 70°C. RNAs were purified using PeqGold RNAPure as described above.

Luciferase Assays
The Dual-Luciferase Reporter Assay System (Promega) was used according to the manufacturer's recommendations with cortical neuron extracts prepared about 24 h after transfection.

Immunocytochemistry
Cells grown on coverslips and 15-m rat brain tissue cryosections were fixed with 1 or 4% paraformaldehyde in PBS (150 mM NaCl, 10 mM sodium phosphate, pH 7.2) for 15 min at room temperature. After 2 washes in PBS (5 min each), cells and tissues were permeabilized with 0.3% Triton X-100 in PBS (5 min), washed twice as above, and blocked for 1 h in blocking solution (10% normal goat serum, 0.05% Triton X-100 in PBS). Primary and secondary antibodies diluted in blocking solution were incubated overnight at 4°C and 2-3 h at room temperature, respectively. Cells and tissue sections were washed 3 times (5 min each) in PBS. Finally, they were mounted in Permafluor (PerkinElmer Life Sciences) or ProLong Gold antifade reagent containing DAPI (Invitrogen). Images were acquired using either a Zeiss Axiovert 135 microscope (10/0.3, 25/0.8, 40/1.3, 63/1.25 objectives) equipped with a Hamamatsu camera using Openlab software (Improvision) or a confocal microscope (Leica SP2, 40/1.25 objective) using Leica SP2 software. Images were exported in TIFF.

MKRN1 Is Novel PAPB-interacting Protein-To better
understand the molecular action of PABP in mammalian neurons, we employed a yeast two-hybrid screen utilizing fulllength PABP bait and a human brain cDNA library. Among 69 positive clones, 20 independent cDNAs encoded MKRN1. Interestingly, all of the isolated clones encoded a short isoform, referred to as MKRN1-short, that is a C-terminal-truncated variant of full-size MKRN1 (MKRN1-long). As schematically shown in Fig. 1A, these isoforms are encoded by a single mkrn1 gene and arise by alternative splicing and differential polyadenylation (Ref. 24 and this study). MKRN1-long consists of four C 3 H-type ZFs, a MKRN-type ZF (MTZF), and a highly conserved C 3 HC 4 -type RING domain (Ref. 24, schematically outlined in Fig. 1B). C 3 H-type ZFs are RNA binding motifs (43,44), whereas the RING finger domain is a protein/protein interaction module characteristic for RING-class E3 ubiquitin ligases (45). MKRN1-short lacks the C-terminal ZF and the last six amino acids of the RING finger domain (RF⌬CC) essential for binding the second zinc ion (Fig. 1C). Human MKRN1short consists of 329 amino acids and exhibits a calculated molecular mass of 35.2 kDa, whereas the 53.3-kDa MKRN1long protein described by Gray et al. (24) is composed of 482 amino acids. To verify the MKRN1-short/PABP interaction, epitope-tagged proteins (T7-MKRN1-short and FLAG-PABP) were transiently expressed in HEK-293 cells and immunoprecipitated with appropriate antibodies, and precipitates were analyzed by Western blotting. As shown in Fig. 1D, immunoprecipitation of FLAG-PABP co-precipitated T7-MKRN1short (Fig. 1D, lanes 1-3). This interaction persisted after digestion with RNase A (Fig. 1D, lane 3), indicating that it is not mediated by RNA. Essentially the same results were obtained by using RNase ONE (data not shown) that cleaves phosphodiester bonds between any two ribonucleotides. Consequently, a possible interaction bridged by the poly(A) tail of mRNAs can be ruled out. In HEK-293 cells neither of the MKRN1 isoforms is expressed at detectable levels (Fig. 1D, lane 6). Thus, an interaction between MKRN1-short and PABP could not be investigated for the endogenous components. However, immunoprecipitation of T7-MKRN1-short pulled down endogenous PABP (Fig. 1D, lanes 4 and 5). Taken together, these data show that in living cells MKRN1-short associates with PABP in an RNAindependent manner.
Interaction of MKRN1-short with PABP Is Mediated by PAM2-like Motif-To determine which part of MKRN1-short mediates binding to PABP, full-length EGFP-MKRN1-short (M1, schematically depicted in Fig. 2A) as well as different EGFP-MKRN1-short deletion mutants (M2-M8, Fig. 2A) were expressed in HEK-293 cells along with FLAG-tagged PABP. Cell lysates were subjected to immunoprecipitation using FLAG-agarose beads. The Western blots shown in Fig. 2, B-F, demonstrate that the interaction of MKRN1-short with PABP depended on the N-terminal amino acids 1-235, whereas the C-terminal part of the molecule did not bind (Fig. 2, B-D, compare M2 and M3). Further analysis revealed that amino acids 161-176 were required for MKRN1-short/PABP interaction (Fig. 2, B-D, M7). However, they were not sufficient as the interaction was highly impaired upon deletion of adjacent amino acids 177-193 (Fig. 2, B-D, M8), indicating that these contribute substantially to PABP binding. A closer inspection (Fig. 3A) revealed that the sequence spanning amino acids 161-176 is very similar to PAM2 motifs within PAIP1 and -2 and Ataxin-2 (ATXN2), respectively (46). Yet in MKRN1-short a highly conserved leucine (Leu 128 in PAIP1, Leu 111 in PAIP2, and Leu 913 in ATXN2) is replaced by aspartate (Asp 165 ). The following region containing amino acid residues 177-193 exhibits some similarity to the C-terminal part of so-called domains of unknown function (DUF). In PABP-interacting proteins belonging to the GW182 family (such as human trinucleotide repeat containing gene protein TNRC6 and Drosophila GW182) DUF serve as PABP binding domains and are equivalents of PAM2 motifs (30). Obviously, the MKRN1-short PABP-interacting domain (henceforth referred to as PAM2- like domain) exhibits characteristics of both PAM2 motifs and DUF. To assess whether the PAM2-like domain binds to PABP by itself, MKRN1-short amino acids 161-193 were expressed as EGFP fusion protein (EGFP-PAM2 MKRN ) in HEK-293 cells. Immunoprecipitation using GFP-Trap beads efficiently pulled down endogenous PABP (Fig. 3B). Other domains of MKRN1-short do not contribute to this interaction as deletion of various parts of the molecule, including ZF2 and -3, did not impair PABP binding (supplemental Fig. 1). With the exception of Drosophila GW182 DUF (47), all of the PAM2 motifs and  Our results indicate that MKRN1-short contacts several sites within PABP but with different efficiencies (Fig. 4, B-F). The main interaction was mediated by RNA recognition motifs 1 and 2 (RRM1 and -2), being almost as strong as that seen with the full-size protein (Fig. 4E, compare P1 and P2). Yet other parts of the molecule such as RRMs 2 and 3, RRMs 3 and 4, and the C terminus bound to MKRN1-short as well (Fig. 4E,  P3-P5). Hence, the PABP binding properties of the MKRN1short PAM2-like domain and PAM2/DUF domains of the PABP-interacting proteins mentioned above are clearly distinct. This difference was further underscored by a direct comparison of the PABP/ATXN2 interaction in the same extracts. Distinct from MKRN1-short ATXN2 bound only to the MLLEcontaining C terminus of PABP but not to any of the RRMcontaining regions (supplemental Fig. 2). Moreover, the PAM2-like domain in MKRN1-short is by itself sufficient to confer binding to both the RRMs and C terminus of PABP in a manner largely identical to that seen with full-length MKRN1short (supplemental Fig. 3). This raises the question of whether MKRN1-short may interact with further RRM-containing proteins involved in posttranscriptional regulation of gene expression. Insulin-like growth factor II mRNA-binding protein 1 (IMP1), for example, is a translational regulator that harbors two N-terminal RRMs (48). Co-immunoprecipitation experiments performed with EGFP-tagged full-length IMP1 and truncated protein revealed an RNA-independent binding of T7-MKRN1-short to the RRMs of IMP1 (supplemental Fig. 4).
Characterization of mkrn1 Gene Expression in Rat Tissues-In human and mouse, MKRN1 mRNA is present in diverse non-neuronal and neuronal tissues (24). Our Northern blot analysis with RNA isolated from different rat tissues confirmed and extended these observations (Fig. 5A). Thus, highest transcript levels were detected in brain and testis. The 3.2-and 2.0-kb transcripts most likely encode rat MKRN1-long and -short, respectively. In testis, an additional smaller transcript of 0.75 kb was detected and presumably represents a further alternatively spliced MKRN1 mRNA encompassing exons 1-3 (49). PCR analysis with primers specific for MKRN1-long and -short cDNA, respectively, confirmed the presence of both variants in rat brain (Fig. 5B).
To address the question of MKRN1 expression at the protein level in rat brain, Western blot analyses were performed with a polyclonal antiserum raised against recombinant GST-MKRN1-short. We have observed that the forebrain contained the highest level of MKRN1 (see below). Hence, Western blot analyses were performed with proteins extracted from the hippocampal formation. The antiserum detected a protein with an apparent molecular mass of ϳ35 kDa (Fig. 5C, lane 1) that appears to represent MKRN1-short as it migrated just slightly faster than epitope-tagged recombinant human T7-MKRN1short (Fig. 5C, lane 2). In addition, a very faint band of about 55 kDa may correspond to MKRN1-long as it co-migrates with human T7-MKRN1-long (Fig. 5C, compare lanes 1 and 3). Recombinant rat MKRN1-short was recognized by the antibody raised against the human protein (Fig. 5C, lane 5; lane 4 contains a protein extract from non-transfected cells). We have noted that in HEK-293 (and HeLa) cells recombinant MKRN1long expression was very low unless cell cultures were treated with the proteasome inhibitor MG-132 (Fig. 5D, lanes 1-4). This is consistent with data reported by Kim et al. (26) showing that human MKRN1-long is subject to autoubiquitination, which leads to its very low abundance in cell lysates. Bands with lower electrophoretic mobility detected after 4 h of MG-132treatment (Fig. 5D, lane 4, marked by asterisks) probably represent ubiquitinated forms of MKRN1-long. Thus, extremely low levels of MKRN1-long in rat brain (Fig. 5C, lane 1) despite the relative high abundance of its mRNA (Fig. 5A) may be explained by its autoubiquitination properties. In contrast, MKRN1-short does not appear to catalyze its own ubiquitination and proteasomal degradation (Fig. 5D, lanes 5-8).
To determine the subcellular localization of both MKRN1 isoforms, we immunocytochemically analyzed HeLa cells and primary rat hippocampal neurons as well as rat brain sections. Despite the lack of any of the well characterized canonical nuclear localization signals, Myc-tagged MKRN1-short is present in both HeLa cell nuclei and cytoplasm (Fig. 6, A and B). In MG-132-treated HeLa cells, recombinant MKRN1-long essen- tially exhibited an identical distribution (data not shown). In primary neurons, Myc-tagged MKRN1-short was observed in both cell bodies and microtubule-associated protein 2 (MAP2)positive dendrites as well as putative axons (MAP2-negative neurites; Fig. 6, C and D). In immunohistochemical investigations of rat brain sections, MKRN1 was detected in most major brain areas (data not shown). Highest immunoreactivity was observed in neurons of the hippocampal formation (Fig. 6, E-H) and in the neocortex (data not shown). In these neurons, MKRN1 is present in nuclei, cell bodies, and proximal dendrites (Fig. 6, G and H). Because MKRN1-long was hardly detectable in rat brain lysates by Western blotting (see Fig. 5C, lane 1) the staining pattern is likely to primarily reflect the brain distribution of MKRN1-short. Moreover, in transfected primary neurons recombinant MKRN1-short partially co-localized with endogenous PABP within dendrites (Fig. 6, I-L). However, many PABP-positive puncta did not include MKRN1-short, indicating that only part of the cellular PABP interacts with MKRN1-short or that the contact is rather transient in nature. In addition, MKRN1-short partially resided in cellular subareas devoid of PABP.

MKRN1-short Stimulates Translation in Nerve Cells and Is Associated with mRNAs Localized to Dendrites-To investigate
if MKRN1-short, as a PABP-interacting protein, regulates translation efficiency and/or mRNA stability, we performed a so-called N22-tethering experiment followed by either a dual luciferase reporter assay or a TaqMan probe-based semiquantitative RT-PCR analysis (for a schematic representation of these assays, see supplemental Fig. 5). For this purpose we constructed a eukaryotic expression vector (pinFiRein-boxB16B) that contains two separate genes, each comprising its own CMV promoter, driving transcription from opposite DNA strands. Both genes possess an intron upstream of the region encoding either PhoLuc or RenLuc, respectively. In its 3Ј-UTR, the PhoLuc mRNA contains 16 copies of a stem-loop structure referred to as box B, which is specifically bound by a small protein domain (N22; Ref. 50). Notably, RenLuc transcripts lack these binding sites (internal tethering control). Thereby, any given N22-tagged protein is selectively tethered to PhoLuc but not RenLuc mRNAs. We co-transfected 7-day-old primary rat cortical neurons with both pinFiRein-boxB16B and an additional vector encoding one of several N22-tagged proteins. Two days later the neurons were harvested, and the activity of both luciferases in cell lysates was determined. To compensate for possible differences in the transfection rate, PhoLuc activity was normalized against RenLuc (henceforth referred to as "nPhoLuc activity"). The nPhoLuc activity observed in cells expressing a recombinant protein consisting of N22 and three tandem FLAG epitopes (N22-FLAG3) served as a reference and was arbitrarily set to 1. Selective tethering of N22-FLAG3 to PhoLuc mRNAs is thought to not significantly alter the metabolism of bound reporter transcripts. As compared with this control, tethering of both full-length MKRN1-short (N22-FLAG3-hMKRN1-short) and a mutant version lacking PABP binding activity (N22-FLAG3-hMKRN1-⌬aa161-176; see Fig.  2, M7) lead to a more than 4-fold increased nPhoLuc activity (Fig. 7A). In contrast, N22-tagged human DDX6 (N22-hDDX6), a DEAD-box helicase promoting recruitment and degradation of mRNAs in P-bodies (processing bodies; Ref. 51) reduced nPhoLuc activity to about 50%. Finally, the fusion protein N22-rShank Shank3-1-290 containing an N-terminal part of the postsynaptic scaffold protein that is not known to play any role in RNA metabolism (52) did not significantly alter nPhoLuc activity as compared with the control value. To assess whether MKRN1-short tethered to reporter transcript control mRNA translation and/or stability, we performed real-time RT-PCR analysis with total RNA isolated from transfected neurons and TaqMan probes hybridizing to sequences derived from two neighboring exons of RenLuc-and PhoLuc genes, respectively. Similar to the dual luciferase assay, PhoLuc mRNA levels were normalized to the concentration of RenLuc tran-scripts (nPhoLuc mRNA levels), and nPhoLuc transcript levels observed in N22-FLAG3-expressing cells again served as reference control that was arbitrarily set to 1. As shown in Fig. 7B, recombinant MKRN1-short (and, as expected, N22-FLAG3-Shank3-1-290) did not affect nPhoLuc mRNA levels as compared with the N22-FLAG3 control. However, consistent with its reported role in mRNA decay, N22-hDDX6 significantly reduced nPhoLuc mRNA abundance. As tethering of MKRN1short significantly increased nPhoLuc activity without altering PhoLuc mRNA levels, this effect can only be accounted for by an increased translation rate and not by enhanced stability or nuclear export of reporter transcripts. Taken together, our findings show that MKRN1-short significantly enhances translation efficiency of bound reporter transcripts via a molecular mechanism that does not require its direct interaction with PABP. In hippocampal homogenates separated on a sucrose gradient MKRN1-short resides only in the light messenger ribonucleoprotein fraction (Fig. 7, C-E), suggesting that it might stimulate protein synthesis at the level of initiation. Dendrites harbor a number of mRNA species that are locally translated (3)(4)(5)(6), among them transcripts encoding the MAP2 and activity-regulated cytoskeleton-associated protein (Arc/ Arg3.1). To assess whether these mRNAs are targets of MKRN1-short we transfected cultured cortical neurons with vectors encoding EGFP-MKRN1-short, EGFP-PABP (positive control), and EGFP alone (empty vector control), respectively. Recombinant proteins were immunoprecipitated with GFP-Trap (Fig. 8A), and individual associated mRNAs were detected by semiquantitative real-time RT-PCR analysis. As shown in Fig. 8B, immunoprecipitation of recombinant MKRN1-short led to an 11-and 8-fold enrichment of MAP2 and Arc/Arg3.1 mRNAs compared with the empty vector control. Enrichment of ␤-tubulin 3 transcripts was somewhat lower (6-fold). Similar data were obtained in experiments performed with recombinant PABP. Hence, in living neurons both proteins are associated with dendritic and somatically restricted mRNAs.
MKRN1-short Accumulates in Middle Molecular Layer of Dentate Gyrus after Plasticity-inducing Stimulation-Specific forms of LTP, a model system for mechanisms underlying long  A, shown are protein lysates from rat primary cortical neurons expressing recombinant EGFP-MKRN1-short, EGFP-PABP, or EGFP alone that were subjected to immunoprecipitation with GFP Trapா_A beads followed by SDS-PAGE and Western blot analyses using anti-GFP antibodies for detection of EGFP-MKRN1 (lanes 1 and 2), EGFP-PABP (lanes 3 and 4), and EGFP (lanes 5 and 6), respectively. I, input proteins (lanes 1, 3, and 5); E, immunoprecipitated proteins eluted from the beads (lanes 2, 4, and 6). The positions of molecular size marker proteins (in kDa) are indicated on the right. B, RNAs extracted from whole cell lysates (input) and immunoprecipitated protein fractions (eluate) were subjected to semiquantitative real-time RT-PCR using primers for MAP2-, Arc/Arg3.1-, and neuron-specific ␤-tubulin 3 cDNAs. The graph depicts the enrichment of individual transcripts in inputs and in eluates obtained by immunoprecipitation of EGFP-MKRN1-short (open bars) and EGFP-PABP (closed bars), respectively, compared with the empty vector control (EGFP-IP). Data were obtained using REST (relative expression software tool) 2008 software for group-wise comparison and statistical analysis of relative expression results in real-time PCR (41). S.E., vertical lines, enrichment/ eluates, p Ͻ 0.001; enrichment/inputs, not significant. Arc/Arg3.1, activityregulated cytoskeleton-associated.
term memory formation, require activity-induced local synthesis of proteins at synapses (33,53). Because MKRN1-short is located in dendrites and involved in translation control (see above), we asked whether synaptic activity might alter MKRN1short abundance at activated synapses in vivo. To answer this question we took advantage of the possibility to selectively stimulate the perforant path in the brain of living rats. These projections from the medial entorhinal cortex terminate in the middle molecular layer of the dentate gyrus (54). High frequency stimulation of this path leads to activity-dependent translation of dendritic mRNAs encoding the Ca 2ϩ /calmodulin-dependent kinase II ␣ subunit (55), MAP2 (55), and Arc within the activated dendritic layer (36). Two hours after high frequency stimulation of the perforant path, MKRN1-short levels strongly increased in a sharply defined dendritic lamina of the dentate gyrus ( Fig. 9, upper panel, left). In contrast, on the contralateral side this regional accumulation of the protein was not evident (Fig. 9, upper panel, right), indicating that the observed local increase in protein levels is triggered by the activity of nearby synapses. As in brain neurons MKRN1 mRNA is not detected in dendrites (supplemental Fig. 6) the accumulation of MKRN1-short in the stimulated dendritic lamina appears to reflect an activity-dependent redistribution of the protein rather than a local de novo protein synthesis. Distinct from MKRN1-short, the dendritic distribution of PABP did not change after stimulation of the perforant path ( Fig. 9, lower panel). The experimental design used here regularly induced local accumulation of F-actin in the activated dendritic lamina as well as robust c-fos gene expression in dentate gyrus granule cells of the stimulated side (supplemental Fig. 7), one of the hallmarks of LTP (55)(56)(57). Taken together, the data underscore that the MKRN1-short accumulation in the middle molecular layer after LTP induction is specific and strongly suggest that the protein likely plays a role in synaptic plasticity.

DISCUSSION
Our studies provide the first extensive characterization of MKRN1 expression at the protein level in the central nervous system. In rat brain a short isoform, MKRN1-short, predominates and is detectable in somata, dendrites, and nuclei of nerve cells. As this isoform lacks two of the cysteine residues essential for the formation of a functional C 3 HC 4 -type RING domain (58), it is unlikely to act as an E3 ubiquitin ligase such as MKRN1-long. The loss of autoubiquitination activity of the short MKRN1 isoform (see Fig. 5D) supports this assumption.
MKRN1-short interacts with PABP, a protein with numerous functions in mRNA metabolic pathways (15). Binding to PABP is mediated by a PAM2-like motif within MKRN1-short. A variety of PABP-interacting proteins possess such motifs, for example PAIP1, PAIP2, and ATXN2 (30). However, in contrast to PAM2 domains of those proteins, which exclusively associate with the PABP C-terminal MLLE domain (Refs. 59 and 60; see also supplemental Fig. 2), the MKRN1-short PAM2-like domain is peculiar as it preferentially interacts with RRMs 1 and 2 and to a lesser extent with RRMs 2 and 3, RRMs 3 and 4, and the C-terminal part of PABP. Whether MKRN1-short binding to these different parts of PABP occurs cooperatively or independently from each other remains to be seen. All in all, this binding behavior of MKRN1-short to PABP is somewhat reminiscent of that of PAIP1 and -2 (59, 60). The major difference is that a single motif in MKRN1-short is able to contact several sites within PABP, whereas binding of PAIP1 and -2 to the RRMs in PABP is mediated by a second motif called PAM1, which in fact represents the high affinity PABP interaction site (59,60). It is this interaction with PABP that is responsible for translational stimulation by PAIP1 (31) and translational silencing by PAIP2 (32), whereas the functional consequence of the PAM2/PABP C-terminal MLLE domain interaction is less clear.
In this study we have demonstrated that MKRN1-short significantly stimulates protein synthesis when tethered to a reporter mRNA while not influencing mRNA stability, thus implying an interaction of MKRN1-short with the translation machinery. As a MKRN1-short mutant that lacks the PABP binding site still enhances reporter mRNA translation similar to the full-length protein, it seems likely that a molecule distinct from PABP mediates the functional link of MKRN1-short to the translation machinery. Yet, in vivo, PABP may provide a platform for MKRN1-short to tie it to an mRNA molecule, a function that is not required in the described N22-box B tethering assay. Such a role of PABP has been described for recruitment of the transducer of ErbB2 (TOB), a protein implicated in cytoplasmic mRNA deadenylation, to target mRNAs (61).
Plasticity producing high frequency stimulation of the perforant path in vivo leads to the accumulation MKRN1-short in the activated dendritic lamina of dentate gyrus granule cells. The same stimulus promotes translation of mRNAs encoding MAP2 and Ca 2ϩ /calmodulin-dependent kinase II ␣ subunit in dendrites (55). Presumably, many more mRNA species are subject to activity-dependent translation activation given that dendrites harbor at least 150 different mRNA species (62) and that synaptic activation, which induces major changes in synapse composition, most likely requires local synthesis of many distinct protein species. Yet, there is no evidence for an increase in overall dendritic protein synthesis subsequent to neuronal activation (53). Furthermore, the number of polysomes at synapses is quite low (53). Hence, the available data suggest that dendritic mRNAs compete for a limited translational machinery and that synaptic activity triggers translation of selected mRNAs at the expense of other transcript species. The translation regulatory capacity of MKRN1-short, its association with dendritically localized mRNAs in living neurons, and its in vivo accumulation at activated synapses suggest that MKRN1-short controls activity-dependent reprogramming of the protein synthesis profile in dendrites. Due to the ability to bind to different parts of PABP, MKRN1-short could modulate recruitment, displacement, and/or binding properties of factors both at the N-terminal end harboring RRM1-4 and the C-terminal MLLE domain of PABP, each of which interacts with distinct sets of proteins (63). The ability of MKRN1-short to interact with IMP1 (see supplemental Fig. 4), a PABP-binding protein involved in translational regulation (64), is consistent with this view. Why does dendritic PABP distribution not change after synaptic stimulation? The most plausible interpretation is that PABP is already bound to all polyadenylated dendritic mRNAs. Unless there was gross redistribution of those transcripts as a result of the stimulus, an accumulation of PABP in the activated lamina is not to be expected.