SEARCH

SEARCH BY CITATION

Keywords:

  • internal ribosomal entry site;
  • KIF5;
  • kinesin;
  • postsynaptic density;
  • staufen;
  • synaptic plasticity;
  • mRNPs

Abstract

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Dendritic mRNA transport coupled with local regulation of translation enables neurons to selectively alter the protein composition of individual postsynaptic sites. We have analyzed dendritic localization of shank1 mRNAs; shank proteins (shank1–3) are scaffolding molecules of the postsynaptic density (PSD) of excitatory synapses, which are crucial for PSD assembly and the formation of dendritic spines. Live cell imaging demonstrates saltatory movements of shank1 mRNA containing granules along microtubules in both anterograde and retrograde directions. A population of brain messenger ribonucleoprotein particles (mRNPs) containing shank1 mRNAs associates with the cargo-binding domain of the motor protein KIF5C. Through expression of dominant negative proteins, we show that dendritic targeting of shank1 mRNA granules involves KIF5C and the KIF5-associated RNA-binding protein staufen1. While transport of shank1 mRNAs follows principles previously outlined for other dendritic transcripts, shank1 mRNAs are distinguished by their translational regulation. Translation is strongly inhibited by a GC-rich 5untranslated region; in addition, internal ribosomal entry sites previously detected in other dendritic transcripts are absent in the shank1 mRNA. A concept emerges from our data in which dendritic transport of different mRNAs occurs collectively via a staufen1- and KIF5-dependent pathway, whereas their local translation is controlled individually by unique cis-acting elements.

Intracellular transport and local translation of messenger RNAs (mRNAs) allow cells to supply proteins to specific subcellular sites. In neurons dendritic transport of specific mRNAs is considered as a mechanism to modify individual synapses by selectively changing the protein content of distal postsynaptic specializations (1,2). Local protein synthesis is believed to contribute to synaptic plasticity that requires a rapid supply of new proteins to specific synaptic sites in response to appropriate stimuli. A limited number of mRNAs is readily detected by in situ hybridization in dendritic fields of the mammalian brain (i.e. in molecular layers of the hippocampus and cerebellum). Among these, the mRNAs coding for the α-subunit of the calcium/calmodulin dependent protein kinase II (αCaMKII) and the arg3.1/arc protein have been investigated in detail. Dendritic transport depends on specific dendritic targeting elements (DTEs), usually located in 3 untranslated regions (UTRs; refs (3–6)). Live imaging studies using the MS2/green fluorescent protein (GFP) system revealed that mRNA particles are translocated in discrete transport packets or clusters (7,8). Whereas short distance movements (e.g. from dendritic shafts into spines or toward activated synapses) may involve actin filaments (9), the majority of long distance translocation of mRNAs into dendrites depends on microtubules. Hirokawa and coworkers (10) presented evidence that conventional kinesins of the KIF5 subfamily are responsible for dendritic transport of both the arg3.1/arc and αCaMKII mRNAs. By purifying a specific subpopulation of messenger ribonucleoprotein particles (mRNPs) associated with the cargo-binding domain of KIF5, numerous mRNA binding proteins were identified which are associated with these messages, giving rise to large mRNP particles of about 1000 S. Proteins such as staufen1 and Purα which have so far been suggested to act as trans-acting factors for the dendritic translocation of mRNAs (11–13) were found to be associated with these granules. Interestingly, there is no apparent sequence similarity among the DTEs that have been identified so far. This raises the question whether additional motor proteins and associated mRNPs exist, which are involved in the dendritic transport of mRNAs other than the arg3.1/arc and αCaMKII mRNAs.

Translation of dendritic mRNAs is believed to be regulated locally upon appropriate stimuli such as brain-derived neurotrophic factor (BDNF), metabotropic glutamate receptor (mGluR) activation or by miniature EPSPs (14,15). Transported mRNAs are assumed to be translationally repressed during transport by repressor proteins such as the fragile X mental retardation protein (FMRP) known to be present in transport granules (10,16). Several mechanisms have been proposed to relieve translational repression, including signaling via CPEB (cytoplasmic polyadenylation element binding protein; ref. (17)) and the mammalian target of rapamycin (mTOR). mTOR activation, in particular, leads to a general increase in the translational capacity of dendrites (18). However, it remains relatively unclear whether overall levels of dendritic protein synthesis are regulated by synaptic signals, or whether individual mRNAs are selected for translation by specific postsynaptic signaling pathways (19). Furthermore, cis-acting elements present in the UTRs of dendritic mRNAs have remained largely uncharacterized with respect to their role in directing translation efficiency. Interestingly, one study suggested that local translation of most dendritic mRNAs is not initiated via the conventional 5-cap dependent ribosomal scanning mechanism, but rather by internal initiation using internal ribosomal entry sites (IRESs; (20)).

Here we have analyzed the dendritic transport mechanisms of shank1 mRNAs which are localized to distal dendrites of neurons in the hippocampus and cerebellum due to a dendritic targeting element in their 3UTR (5). Shank family members (shank1–3) are scaffold proteins which connect different types of glutamate receptors in the postsynaptic density (PSD) of excitatory synapses (21) and significantly contribute to the establishment of dendritic spines (22–24). Haploinsufficiency for the Shank3 gene is associated with mental retardation or autism in human patients (25,26). Local synthesis of shank1 may be relevant to synaptogenesis and structural alterations of the PSD during synaptic plasticity. Interestingly, the PSD content of shank1 is reduced upon strong synaptic activity and increased during blockade of action-potential-dependent transmission (27). In contrast, several reports have indicated that other dendritically localized mRNAs are preferentially translated during periods of strong synaptic activity (28,29). Our analysis indicates that the transport of shank1 mRNAs employs similar cellular mechanisms as known for αCaMKII and arg3.1/arc messages. However, translation of shank1 mRNAs is not driven by an IRES element but is inhibited through a strong translational repressor in its 5UTR. This indicates that local translation of shank1 mRNAs is subject to a unique regulatory mechanism distinct from other dendritically localized messages.

Results

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Kanai et al. recently described a glutathione S-transferase (GST)-pulldown technique for the purification of a population of large dendritically transported mRNPs, which employs the cargo-binding domain of KIF5 proteins (10). We used the region of KIF5C described as the minimal binding domain for these mRNPs for pulldown experiments from mouse brain extracts. Purified samples were analyzed for their protein and mRNA content by western blotting or quantitative real time reverse transcription-polymerase chain reaction (RT-PCR), respectively. We observed an enrichment of candidate RNA-binding proteins, including FMRP, DDX1 and staufen1, as well as the translation factor eIF2α (Figure 1A), all of which were reported previously to be present in KIF5-associated mRNPs (10). Staufen proteins in particular, have been reported to play a role in the dendritic transport of mRNAs (11,12). The RT-PCR analysis showed that both the αCaMKII and arg3.1/arc positive control mRNAs and the shank1 mRNA were strongly enriched in the KIF5C-purified sample when compared to GST controls. Enrichment of shank1 transcripts (7.95 ± 2.4 fold) slightly exceeded values for αCaMKII (6.2 ± 3.6 fold) and arg3.1/arc (5.15 ± 1.03 fold) messages, whereas the glycerol aldehyde phosphate dehydrogenase (GAPDH) mRNA which was used as negative control was only slightly enriched (2.14 ± 1.5 fold) in the KIF5C samples (Figure 1B). In further control experiments, we used a part of the cargo-binding region of KIF3C, which was recently implied in intracellular transport of FMRP (30); here we observed only a slight enrichment of the shank1 mRNA (compared to the GST control; Figure 1B). These data show that shank1-containing mRNPs specifically associate with the KIF5C cargo-binding domain, thus identifying KIF5 as a possible motor protein involved in dendritic transport of shank1 mRNAs.

image

Figure 1. Shank1 mRNA containing mRNPs associate with the cargo binding domain of KIF5C. A) Cleared mouse brain lysates were used as input for affinity chromatography using either GST or GST-KIF5C coupled to glutathione sepharose. Samples were analyzed for the presence of individual proteins by western blotting using the antibodies indicated. B) Cleared mouse brain lysate was used as input (in) for affinity chromatography as in A. Samples purified on GST-KIF5C or GST alone were analyzed by quantitative RT-PCR using specific primers directed at the mRNAs indicated. In a separate set of experiments, KIF3C-GST was used for affinity chromatography. Data are presented as the enrichment obtained with GST-KIF5C (or GST-KIF3C) purified samples over the GST-purified samples (significantly different from GAPDH control experiment; *, p < 0.05; **, p < 0.01; n = 4 for KIF5C; n = 3 for KIF3C).

Download figure to PowerPoint

For further analysis of the role of KIF5 and associated proteins in dendritic mRNA transport, we examined their subcellular distribution in relation to shank1 mRNAs in cultured hippocampal neurons. Endogenous shank1 mRNAs are targeted to dendrites in these cells, as demonstrated by nonradioactive in situ hybridization using a full-length probe directed against the rat shank1a sequence (Figure 2). Starting on day 5 of neuronal differentiation in vitro (DIV5), shank1 mRNA particles were detectable in neuronal cell bodies but rarely extended into dendrites. From DIV8 onwards, shank1 mRNA became readily detectable in dendrites. Signals were distributed in a punctate pattern throughout the neurons. Shank1 mRNA was clearly detectable in dendrites at later time points (14 DIV, as well as 21 DIV), indicating that shank1 mRNAs are present in dendrites during the period of intense synaptogenesis known to peak between DIV8 and 14.

image

Figure 2. Developmental appearance of dendritic Shank1 mRNA particles. Neurons were cultured for the times indicated (DIV; days in vitro). After fixation, cells were analyzed by combined fluorescent in situ hybrization/immunohistochemistry using a DIG-labeled rat shank1 RNA antisense probe (left), and anti-MAP2 antibodies (center) to outline dendrites. Cells were also viewed by DIC microscopy (right panels). For DIV21, an enlargement of a dendritic segment is shown, demonstrating strongly clustered labeling for the shank1 mRNA. Note the absence of Shank1 mRNA in non-neuronal (MAP2-negative) cells, e.g. at DIV5. The specificity of the in situ labeling was further verified by using a shank1 sense probe in the bottom panel. Bar, 10μm.

Download figure to PowerPoint

By combining FISH detection of the shank1 mRNA with immunocytochemical detection of KIF5C, we determined that a significant proportion (27 ± 11%) of particles containing the shank1 mRNA are colocalized with KIF5C (Figure 3A). The number of KIF5C-containing particles by far exceeded that of shank1-mRNA containing particles, consistent with the notion that KIF5C also transports cargoes other than the shank1 mRNA. In addition, other KIF5 family members (KIF5A and KIF5B; ref. (10)) appear to play redundant roles with KIF5C in neurons and may participate in the transport of the shank1 mRNA.

image

Figure 3. Association of shank1 mRNA granules with KIF5 and staufen1. A) Colocalization of the endogenous shank1 mRNA and KIF5C by combined ICC/FISH. Neurons at DIV14 were fixed and stained using a DIG-labeled riboprobe directed against full-length shank1a (left) and anti-KIF5C (center). Both signals appear in dendrites and cell bodies in a punctate pattern. The merged pictures (right) show that about 27% (±11%; n = 10) of dendritic shank1 mRNA granules (red fluorescence) are also decorated with KIF5C specific green fluorescence (see arrows). Background levels of colocalization in this assay were determined to be 7.2 ± 5.7% (n = 10). Bar is 20μm. B) Colocalization of the endogenous shank1 mRNA and staufen1 by combined ICC/FISH. Neurons at DIV14 were fixed and stained using a DIG-labeled riboprobe directed against full-length shank1a (red signal, left) and anti-staufen1 (green, center). Both signals appear in dendrites and cell bodies in a punctate pattern; the merged pictures (right) show that about 28% (±10%; n = 10) of dendritic shank1 mRNA granules are also decorated with staufen1 specific fluorescence (see arrows in enlarged picture at the bottom). Single fluorescence pictures were inverted for clarity. Bar is 20μm.

Download figure to PowerPoint

As staufen1 co-purified with the shank1 mRNA in KIF5 pulldowns, we immunostained also for staufen1 together with FISH detection of shank1 mRNAs. These experiments indicated that again about one quarter (28 ± 10%) of shank1 mRNA containing particles were positive for staufen1 immunoreactivity (Figure 3B). Specificity of the primary antibodies in the above experiments (anti-KIF5C and anti-Staufen1) was ascertained in western blotting and immunohistochemistry experiments, where blocking with the appropriate antigens eliminated immunoreactivity (Figure S1, Supporting Information).

For a functional analysis of factors contributing to shank1 mRNA transport, we made use of the GFP-MS2 reporter system (7,31). Here a recombinant mRNA containing the coding sequence for the monomeric red fluorescent protein (mRFP; (32)), fused to eight repeats of the MS2 recognition element (MS2re), was expressed in neurons (mRFP/MS2re mRNA). The 3DTE of the shank1 mRNA and the 5UTR of the human shank1 mRNA were included in the reporter RNA as detailed in Figure 4A (5-mRFP/MS2re−3 mRNA; see Supporting Information Figure S2 for sequence analysis of shank1 5UTRs). For visualization of recombinant mRNAs, we coexpressed GFP fused to both a nuclear localization signal (NLS) and the MS2 phage capsid protein (GFP-MS2), the latter of which recognizes the MS2re on reporter mRNAs.

image

Figure 4. Control of dendritic transport and translation through the 3 and 5 UTRs of the shank1 mRNA. A) Scheme of the recombinant protein and mRNAs; features are labeled as indicated in the text. The table refers to the results obtained in (B) with respect to efficient dendritic targeting and translation of the mRFP reporter mRNA. B) Neurons were transfected with the indicated constructs; fluorescence of recombinant proteins was detected 24 h later by fluorescence microscopy. Note the complete absence of red fluorescence when a construct containing the shank1 5UTR was used. No fluorescence above background could be detected in 100% of neurons (five independent transfections; no mRFP fluorescence detectable in at least 20 GFP-positive cells per experiment). C) Neurons were transfected as before with GFP-MS2 and mRFP expression constructs as indicated. Analysis by combined immunocytochemistry (using anti-GFP; left) and fluorescent in situ hybridization (using an mRFP antisense probe; middle) indicated that the GFP-MS2 protein faithfully decorates mRFP-mRNA containing particles (overlay, right). After evaluating all dendrites from 30 neurons out of three separate experiments, 85% of GFP-labeled granules were decorated by red mRNA granules. All RNA granules were also positive for GFP (> 200 granules analyzed). D) Quantification of data shown in the left panels of (C), indicating the importance of the shank1 3UTR for dendritic localization. E) Neurons expressing the GFP-MS2 fusion protein together with the 5-mRFP-MS2re−3 mRNA were stained for GFP (green) and the dendritic marker protein MAP2 (red). The merged picture indicates that all GFP puncta are localized in MAP2 positive processes; Bar, 20μm.

Download figure to PowerPoint

Upon coexpression of mRFP/MS2re mRNA with the GFP-MS2 fusion protein in hippocampal neurons, GFP fluorescence was mainly detected in neuronal somata including the nuclei, whereas mRFP (indicative of translation of the recombinant mRNAs) was evenly distributed throughout cell bodies and dendrites (Figure 4B, top). After coexpression of GFP-MS2 with the 5-mRFP/MS2re−3 mRNA, GFP fluorescence (visualizing the recombinant mRNA) was present in cell bodies and extended far into dendrites (Figure 4B, middle). Although the GFP-MS2 fusion protein was clearly detectable by the GFP fluorescence in dendrites, we did not observe any mRFP fluorescence above the low level of background autofluorescence which can also be observed in untransfected cells (five independent transfections; no mRFP fluorescence detectable in at least 20 GFP-positive cells per experiment). mRFP could only be detected via immunostaining using an mRFP-specific antibody (not shown). To assess whether translation of the recombinant mRNA might be inhibited by the shank1 5UTR (which is extremely GC-rich and contains several upstream open reading frames; Figure S2) we deleted the corresponding part from our expression construct. While the dendritic localization of the reporter mRNA/GFP-MS2 protein was not affected, we were indeed able to observe mRFP fluoresence in a diffuse manner throughout the cell bodies and dendrites (Figure 4B, bottom), indicating efficient translation of the reporter mRNAs. Taken together, these experiments suggest that the 5UTR of the shank1 mRNA acts as a translation repressor, while the 3UTR is sufficient to mediate dendritic mRNA localization.

To verify that the GFP-MS2 clusters observed in these and the following experiments truthfully reflect the localization of the recombinant mRNAs, we performed combined fluorescent in situ hybridization/immunocytochemistry analysis using an mRFP-directed riboprobe and anti-GFP antibodies. Red fluorescence derived from mRFP was negligible when constructs containing the shank1 5UTR were used (see 5-RFP-MS2re−3 in Figure 4B). Therefore we could use red fluorescent Cy3-labeled antibodies for the detection of the FISH signal. Here we observed a near-complete colocalization of both fluorescent signals in dendrites (Figure 4C,D), demonstrating that the GFP reporter protein indeed labels mRNPs containing the 5-mRFP-MS2re−3 fusion mRNA. Elimination of the shank1 3 region from constructs completely abolished dendritic localization of recombinant transcripts (5-RFP-MS2re in Figure 4C,D). In parallel, expressing the 5-mRFP-MS2re−3 fusion mRNA together with GFP-MS2, we also confirmed that the GFP-labeled mRNP complexes were indeed in dendrites and not in axons, as all (100%) of GFP particles were present in MAP2-positive, i.e. dendritic, processes emanating from transfected neurons (four independent experiments; at least 10 neurons evaluated per experiment; see Figure 4E for an example).

To analyze whether the observed dendritic localization reflects active shank1 mRNP transport, time-lapse video microscopy was employed to track the movement of individual mRNA/GFP-MS2 particles within dendrites of living neurons (Figure 5A; see videos S1 and S2 in Supporting Information). Initial experiments showed that many particles in dendrites were stationary throughout a 10 min observation phase. About 1–5% of GFP particles (2.95 ± 2%) were mobile during this interval. Out of the mobile clusters analyzed, 80% exhibited oscillatory or saltatory movements; the remaining clusters moved unidirectionally, consistent with other described transport complexes (6,9). Anterograde transport occurred at rates of 4.14 ± 3.98μm/min (Figure 5B) with maximal velocities of up to 20μm/min in rapid unidirectional movements. Retrograde movements occur at similar velocities of 5.08 ± 4.26μm/min (Figure 5B; see Video S2 for a rapidly moving particle). The observation that only a small fraction of GFP-MS2 containing mRNPs is moving at any given time may reflect that only a limited number of shank1 mRNA particles are associated with the KIF5 motor protein in fixed cells (see Figure 3A).

image

Figure 5. Time-lapse video microscopy of shank1 mRNPs in neuronal dendrites. A) Left: Overview of dendritic ramifications emanating from a single neuron expressing GFP-MS2 protein and mRFP-MS2re−3 mRNA. The cell body is located below the imaged area. The time-lapse series on the right demonstrates movement of a single GFP-labeled particle (arrow) moving to the left toward a branch site in a retrograde manner (40–400 seconds). At the branch point it remains stationary and later on (840/880 seconds) it starts to move rather slowly in an anterograde manner. Fluorescence pictures were inverted for clarity. Bar, 20μm. B) Frequency distribution of velocities observed for individual GFP-labeled particles. 36 anterograde and 30 retrograde moving particles were evaluated.

Download figure to PowerPoint

Movement of mRNA particles is thought to depend on the cytoskeleton; we therefore used drugs that depolymerize microtubules, as well as the actin cytoskeleton, in order to determine which of these structures participates in dendritic localization of shank1 mRNAs. Whereas treatment with cytochalasin D, a blocker of actin filament elongation, did not affect the presence of mRNA/GFP-MS2 particles in dendrites, the microtubule depolymerizing agent nocodazole significantly reduced the number of granules in dendrites (Figure 6A). The effectiveness of drug treatments was in each case verified by staining with anti-tubulin in the case of nocodazole, and phalloidin in the case of cytochalasin D (not shown). Taken together our data imply that the dendritic localization of the shank1 mRNA involves active microtubule-dependent transport.

image

Figure 6. Shank1 mRNA transport depends on microtubules and KIF5 motor proteins. A) Neurons expressing 5-mRFP-MS2re−3 and GFP-MS2 were treated with DMSO only (left); 6μg/mL nocodazole (center); or 5μg/mL cytochalasin D (right) for 4 h and stained for the GFP-MS2 fusion proteins. In each case the boxed area was enlarged at the bottom of the figure. Bar, 20μm. For quantitative evaluation, the number of GFP-positive granules in 50μm segments of the most intensely stained dendrite for 30 neurons was determined. *, significantly different from DMSO control experiment; p < 0.002; Student's t-test (n = 3). B) Neurons expressing 5-mRFP-MS2re−3 together with GFP (left), GFP-KIF5CDN (center) or GFP-dynamitin (right) were subjected to combined ICC/FISH analysis using anti-GFP (green, upper panels) and a riboprobe directed against the mRFP-mRNA (red, middle and lower panels). The lower panels depict an enlargement of the boxed areas. Bar, 20μm. Quantitative analysis was performed as in (A). *, significantly different from GFP control experiment; p < 0.002, Student's t-test (n = 3). C) Further dominant negative constructs were cotransfected with the 5-mRFP-MS2re−3 construct; dendritic localization of the recombinant mRNA was evaluated as in (B). D) For several key constructs (GFP/control; KIF5CDN2; KIF17DN), analysis as in (B) and (C) was repeated and the percentage of reporter RNA present in proximal and distal dendrites relative to the total fluorescence signal (cell body + dendrite) was determined as described in Figure S5. *p < 0.05; **p < 0.01; significantly different from GFP-KIF5CDN2 as determined by Student's t-test (n = 9).

Download figure to PowerPoint

To verify a role of KIF5 motor proteins in shank1 mRNA transport, we used several dominant negative (DN) constructs for KIF5C. GFP-KIF5C-DN1 contains the microtubule-binding domain which is considered to act in a DN manner as it binds also to the cargo-binding motif of KIF5C, and therefore prevents association of cargoes with the endogenous motor protein (33). GFP-KIF5C-DN2 and -DN3 lack the motor domain but do contain the cargo-binding domain (DN2) and in addition the stalk motif (DN3). To disrupt dynein motor activity we also overexpressed dynamitin known to specifically interfere with dynein function (34). As recombinant dynamitin and KIF variants were expressed as GFP fusion proteins, we omitted the GFP-MS2 plasmid in these experiments but used FISH to detect the 5-RFP-MS2re−3 reporter mRNA instead. Neither expression of GFP alone nor GFP-dynamitin interfered with the dendritic localization of shank1 mRNA clusters in neurons; in contrast, all three GFP-KIF5C-DN variants virtually eliminated the dendritic localization of mRFP/shank1 reporter mRNAs; this was evident as a reduction in the number of dendritic mRNA clusters (Figure 6B,C), as well as a reduction in the proportion of total cellular reporter mRNA, which could be detected in dendrites (Figure 6D). The efficiency and specificity of these assays were verified in further control experiments. Dynein is required for Golgi formation. Consistently, overexpressed GFP-dynamitin interfered with proper positioning of the Golgi apparatus, as evidenced by staining against the Golgi-resident protein GOPC/PIST (ref. (35–37); Supporting Information, Figure S3). Furthermore, coexpression of the KIF5C-DN1 construct (motor domain only) did not affect dendritic localization of a dsRed-fusion of the NMDAR1 subunit of the NMDA receptor, indicating that microtubule-dependent transport in general is not affected by the KIF5C-DN1 construct (Supporting Information, Figure S4). Finally, DN constructs for KIF21 and KIF17 were not efficient in blocking the dendritic targeting of the 5-mRFP-MS2re−3 reporter mRNA (Figure 6C).

The function of RNA-binding proteins for dendritic localization of shank1 mRNAs was assessed by expressing either full-length variants, or truncated, putative DN versions in hippocampal neurons. DN proteins were designed to contain individual RNA-binding domains, however, lacking other domains which would enable proteins to link bound RNAs to cytoskeletal motor proteins. Coexpression of full-length staufen1 did not interfere with the dendritic localization of recombinant 5-mRFP-MS2re−3 mRNAs, whereas DN-staufen1 significantly interfered with the dendritic localization of mRNA particles. The mutant protein itself was also absent from dendrites (Figure 7A,B). On the other hand, a DN construct for the nonrelated RNA-binding protein hnRNP-K did not interfere with the dendritic localization of the shank1 mRNA (Figure 7B and data not shown).

image

Figure 7. Shank1 mRNA transport depends on staufen1. A) Neurons expressing 5-mRFP-MS2re−3 together with Flag-tagged staufen1 (full length; upper panel) or a Flag/DN-staufenl construct (stauDN; lower panel) as indicated were subjected to combined ICC/FISH analysis using a riboprobe directed against the mRFP-mRNA (red, left) and anti-Flag (blue, center). The lower panels depict an enlargement of the boxed areas. Bar, 20μm. B) Quantitative analysis of the data shown in (A). In this case, wt and dominant negative constructs for the nonrelated RNA-binding protein hnRNP-K were included. **, significantly different from control experiment (no additional construct); p < 0.002, Student's t-test (n = 3).

Download figure to PowerPoint

In contrast to arg3.1/arc and αCaMKII mRNAs, a remarkable feature of shank1 mRNAs is the strong translational block exerted by the 5UTR, which became apparent in neurons expressing the mRFP reporter constructs with or without the 5UTR (Figure 4). The 5UTR also blocks translation when linked to a luciferase reporter and analyzed in vitro in the rabbit reticulocyte lysate (Figure 8A). In contrast, the arg3.1/arc 5UTR had almost no effect on translation efficiency. The human shank1 5UTR used in these experiments is highly similar to corresponding sequences in mouse and rat; in particular, the high GC content and three upstream open reading frames are conserved (see Supporting Information Figure S2 for a sequence comparison). In addition, a 5UTR sequence variant has been reported to be expressed in rat brain, which is likely to be generated by usage of an alternative promoter located in intron 1 of the rat shank1 gene (38). This alternative 5UTR was also tested in translational control experiments in rabbit reticulocytes, and proved to be similarly repressive for translation of the luciferase reporter mRNA ( Figure 8A). As evidence for expression of this variant in other species (based on entries to the EST database) is missing, we did not include this alternative 5UTR in additional experiments.

image

Figure 8. The shank1 5UTR represses translation. A) Capped transcripts coding for Photinus luciferase (Pluc) were generated from empty pBL vector, or vector containing human shank1 5UTR, rat alternative 5UTR, or arg3.1/arc 5UTR as indicated. Transcripts were used for in vitro translation using rabbit reticulocyte lysates, followed by luciferase assays. Data are presented relative to control (no shank1 sequence included; n = 3) B) Bicistronic expression vectors coding for Pluc, followed by the encephalomyocarditis virus (EMCV) IRES sequence and Renilla luciferase coding sequence (Rluc) were transfected into HEK cells or cortical neurons. The shank1 5UTR was included as indicated. 24 h after transfection cells were lyzed and analyzed sequentially for activities of the two luciferases expressed. Data are presented as Pluc/Rluc ratio relative to control (no shank1 sequence included; n = 5). C,D). Neurons were transfected with bicistronic vectors as in (B); on the next day cells were treated with bicuculline (bic; 40μM; 2 h; C) or tetrodotoxin (ttx; 2μM; 10 h; D) and then assayed as in (B). E) Bicistronic vectors containing a short polylinker sequence (top), the shank1 5UTR (middle) or the arg3.1/arc 5UTR flanked by the open reading frames for Pluc and Rluc were transfected into cortical neurons. 24 hours after transfection cells were lyzed and analyzed sequentially for activities of the two luciferases expressed. Data are presented as Pluc/Rluc ratio relative to control (polylinker only; n = 4).

Download figure to PowerPoint

To analyze translational control by the shank1 5UTR in HEK cells and cortical neurons (Figure 8B) we used bicistronic vectors. Here the Photinus luciferase (Pluc; 5 coding sequence) was used to measure the effect of the shank1 5UTR on translation efficiency, whereas translation of Renilla luciferase (Rluc; coding sequence in 3 part of the mRNA) driven by a viral IRES element served to normalize for the amount of mRNA present in transfected cells. Again, a strong translational inhibition, evident as reduced Pluc/Rluc activity ratios, was detected both in HEK cells and cortical neurons. In neurons, this inhibition was not rescued by prolonged increases or decreases in synaptic activity. Treatment of transfected cells with bicuculline (which increases neuronal firing through blockade of GABAA receptors, and decreases postsynaptic shank1 content; ref. (27)) or tetrodotoxin (which inhibits action potentials and increases postsynaptic shank1 content; ref. (27)) did not specifically increase translation efficiency of the reporter mRNA carrying the shank1 5UTR (Figure 8C,D).

Whereas a detailed analysis of the role of 5UTRs in translational control of other dendritically localized messages has not been performed, one recent study suggested that several of these mRNAs were translated by an IRES-dependent mechanism (20). We therefore analyzed the possibility that the shank1 mRNA is also translated by this mechanism. For this purpose, we again used bicistronic expression vectors, in this case containing coding sequences for Photinus and Renilla firefly luciferases, separated by a multiple cloning region for the insertion of candidate IRES sequences (40). Introduction of the basic vector into primary cortical neurons, and determination of the ratio of Renilla to Photinus (Rluc/Pluc) luciferase activity showed that the intervening sequence between both coding regions in the basic vector allowed for a weak but consistent translation of the second cistron, leading to Rluc/Pluc ratios of 0.027 ± 0.007 (set as 100% in Figure 8). Introduction of the shank1 5UTR significantly suppressed the expression of Renilla luciferase, suggesting that the 5UTR does not function as an IRES. In contrast, when the 5UTR of the arg3.1/arc mRNA, one of the messages which was shown to contain an IRES element (20), was introduced into the bicistronic reporter mRNA, it did enhance the production of Renilla luciferase (Figure 8E). These data strongly indicate that translation of the arg3.1/arc mRNA is enhanced by an IRES element, whereas translation of the shank1 mRNA is not.

Discussion

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Dendritically localized mRNAs are thought to provide neurons with the means to locally regulate the levels of individual proteins at synapses, thus enabling cellular phenomena such as synaptogenesis or synaptic plasticity, which occur locally in dendrites in a very spatially restricted manner. Since most studies on dendritic transcripts have so far focused on αCaMKII and to a lesser extent on the arg3.1/arc and MAP2 messages (3; 4; 6; 14), it is yet unknown whether all dendritically localized mRNAs are incorporated into similar mRNPs and transported by similar cellular mechanisms. In addition, it is unclear if and how neurons differentially initiate translation of individual mRNAs which may be present in dendrites at the same time, possibly also in the same transport particle.

Our biochemical analysis identifies a number of similarities between shank1 mRNAs and other dendritically localized transcripts. This includes the association of shank1 mRNAs with KIF5 motor proteins; the dependence of dendritic transport on the RNA-binding protein staufen (10,11); the dependence of dendritic mRNA localization on intact microtubules; and the inhibition of dendritic localization by disrupting the function of KIF5C. Furthermore, the velocity of movement of shank1 mRNA particles reported here (4 − 5μm/min) is in line with previous measurements on the αCaMKII mRNA (2.4−3 μm/min; ref. (7)) and KIF5-containing mRNPs (2μm/min; ref. (10)). Dynein-based motors on the other hand, which have also been implicated in dendritic mRNA transport (41), do not appear to play a decisive role for the localization of the shank1 mRNA. This may be surprising because anterograde and retrograde movements were observed in our live imaging experiments (Figure 5). Nevertheless, it is known that microtubules show mixed polarities in dendrites (42), so that kinesins could also mediate retrograde transport.

Taken together with previous work these data suggest that a common transport pathway exists for several (if not all) mRNAs which are localized to the distal segments of neuronal dendrites. The protein complement of shank1 mRNPs may therefore be similar to that of mRNPs containing other dendritically localized transcripts. This is in fact somewhat surprising, as the dendritic targeting elements of shank1, arg3.1/arc and αCaMKII mRNAs do not exhibit any apparent sequence similarity (4–6). So far, it is unclear as to which RNA-binding proteins link transported mRNAs to their respective motor units. KIF5C-associated mRNPs contain a number of RNA-binding proteins (such as staufen, Purα, DDX1 or FMRP) in complex with dendritically transported mRNAs (ref. (10), and data presented here). Staufen isoforms have been repeatedly implicated (11,43); however, staufen1 recognizes stretches of double-stranded RNAs independent of their sequence and is therefore not likely to selectively recruit particular mRNA targets into transport RNPs. In fact we have currently no indication that staufen interacts directly with shank1 mRNAs. An important question is which of the additional RNA-binding proteins present in KIF5C-associated particles recognize the shank1 3UTR (or other DTEs) more specifically and mark shank1 mRNAs for transport into dendrites.

Transported mRNAs are believed to be translationally silent during transport, and translation is initiated in proximity to postsynaptic sites by an appropriate (synaptic) stimulus. So far, it is unclear as to how translational silencing during transport is maintained. We observed a massive translational block which is imposed on the shank1 mRNA by its rather long (422 nt) and extremely GC-rich (83% GC content) 5UTR. This inhibition occurs regardless of the cellular environment, i.e. in cell somata, in dendrites, as well as in non-neuronal cells or cell free systems (Figure 8). A similar translational inhibition has not been observed for other dendritic mRNAs (14); consistently we can not detect any sequence similarity between the shank1 5UTR and other 5UTRs. Interestingly, the 5UTRs of αCaMKII (length of 5UTR: 149 nt; 67% GC content) or arg3.1 mRNAs (216 nt; 67% GC content) have been reported to enhance translation efficiencies when linked to reporter mRNAs (20). We observed here only a slight reduction by the arg3.1 5UTR in the reticulocyte lysate system (see Figure 8A). Thus, despite the similarities we see here between different dendritically localized mRNAs with respect to the mechanism of transport, the regulation of translation appears to distinguish the shank1 mRNAs from other dendritic transcripts. This difference is underlined by the observation that the shank1 5UTR does not support cap-independent initiation via an IRES, whereas both αCaMKII and arg3.1/arc mRNAs have been shown to do so (20; also confirmed here for arg3.1/arc). In fact, it is likely that shank1 synthesis should be controlled differently, as neuronal activity decreases the amount of shank1 at synapses (27). We could not overcome the translational inhibition by the shank1 5UTR by increasing or decreasing neuronal activity, suggesting that the activity-dependent changes in postsynaptic shank1 levels which have been observed under these conditions (27) are not mediated on the translational level. In contrast, dendritic synthesis of arg3.1/arc and αCaMKII is strongly induced by local electrical activity (28,44). Unlike αCaMKII and arg3.1, shank1 does not contribute to synaptic plasticity in experimental models such as long-term potentiation (24). It has been suggested that local synthesis of shank1 may be particularly important for spine formation and synapse assembly (22). Consistently, dendritic shank1 mRNA particles appear at DIV8 in cultured neurons (Figure 2), at a time when synaptogenesis starts in this model system. As shank proteins are required for synapse formation, a putative signaling pathway overcoming the translational inhibition is likely to be initiated during early phases of synaptogenesis, possibly by synaptic cell-adhesion molecules. In the absence of its postsynaptic interaction partners GKAP and PSD-95, shank1 forms aggregates which are prone to degradation (45). This situation could occur if the protein was synthesized in large quantities in the cell soma before transport to postsynaptic sites. Thus the ability of shank to self-associate (46) and induce spines (22,23) may require a dendritic synthesis of the protein. In addition, it should be noted that shank protein levels apparently need to be controlled quite accurately, as loss of only one copy of the gene coding for shank3 is associated with mental retardation and autism in humans (25,26).

Currently we can only speculate which cellular mechanisms enable translation of the shank1 mRNAs in dendrites. Further work will focus on proteins that travel together with the shank1 mRNA in dendritic mRNPs, such as FMRP or DDX1. Proteins carrying an RNA helicase activity such as DDX or related proteins are of interest here, as they may help in breaking up the strong secondary structure of the 5UTR. Consistently, it was shown recently that the RNA helicase DHX29 is required for efficient translation of mRNAs with highly structured 5 regions (47).

Taken together, our data indicate that dendritic mRNAs share a common mechanism of transport into dendrites, but their translation is regulated by divergent extracellular stimuli, as well as divergent signaling pathways. Further work will be required to identify the stimulus which actually allows for efficient translation of the shank1 mRNA in dendrites.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Expression constructs and antibodies

A cDNA coding for the monomeric red fluorescent protein (32) was kindly provided by Roger Tsien (Univ. of California, San Diego, CA) and was cloned into pEGFP-C1 such that the EGFP coding sequence was replaced, thereby generating pmRFP-C1. Constructs containing eight copies of the MS2 recognition element and coding sequence for an EGFP-MS2 fusion protein (29) were obtained from Robert Singer (Albert Einstein College of Medicine; New York). The 8xMS2 cassette was inserted into pmRFP-C1 (3 to the mRFP stop codon), followed by the first 588 nucleotides of the rat shank1a 3UTR (nt 6593–7180; NM_031751). A cDNA clone encoding the human shank1 5 untranslated region (5UTR; accession EU872208) was obtained by screening a human brain cDNA library (Clontech; see ref. (48) and Figure S2). In mRFP-based vectors, it was inserted at a position 5 to the mRFP coding region such that the first in frame AUG start codon (start1 in Supplemental Figure S2) is in frame with the mRFP open reading frame. cDNA corresponding to an alternative 5UTR of rat shank1 (38) was amplified by RT-PCR from rat brain RNA. cDNA encoding full-length rat staufen1 or only the second and third RNA-binding domains of Staufen1 isoform rStau-I6(+) (AF227200, nt 512–1047; ref. (34)) were cloned into pCMV2B (Stratagene, La Jolla, CA) to create vectors for expression of flag-tagged full-length Staufen1 and the DN protein Stau1-RBD2/3(+). A DN-hnRNP-K construct was generated by cloning the cDNA coding for K-homology domains 1 and 2 (nt 91–729; NM_057141) into pCMV2B. Three KIF5C-DN constructs were generated: GFP-KIF5C-DN1 was obtained by cloning the first 990 nucleotides of the KIF5C coding region (NM_001107730; kindly supplied by A. Schepis; EMBL Heidelberg, Germany) into pEGFP-N1. GFP-KIF5C-DN2 and -DN3 were generated by cloning bp 2383–3219 (DN2; corresponding to amino acids 678–955) into pEGFP-C1, and bp 1357–3219 (DN3, residues 335–955) into pEGFP-C3. GFP-KIF21b-DN contains bp 4310–5183 (NM_001105990; corresponding to the C-terminal cargo-binding domain of rat KIF21b), in pEGFP-C1. GFP-KIF17DN contains bp 2355–2747 of XR_008709, corresponding to residues 848–979 of KIF17. The dynamitin-EGFP expression vector has been described ((34); kindly provided by Dr Richard Vallee).

Antibodies were obtained from the following sources: Cy3-labeled Mouse anti-DIG, rabbit anti-GFP, anti-DDX1 and anti-eIF2a were from Abcam; rabbit anti-Flag from Sigma; rabbit anti-KIF5C from Affinity Bioreagents; rabbit anti-staufen1 has been described before (43,49); rabbit anti-FMRP was custom made by Pineda Antikörper Service) using a glutathione-S-transferase (GST) fusion protein containing residues 1–459 of rat FMRP as antigen; rabbit anti-MAP2 was obtained from Prof. Craig Garner (Stanford, CA); guinea-pig anti-PIST has been described (35).

Neuronal cell culture and time-lapse video microscopy

Hippocampal neurons were prepared from E19 rat embryos, and cultured on glass coverslips as described (50). Neurons were transfected at day 7 in vitro (DIV), and cultured for an additional 24 h before imaging on an Axiovert 200. Plasmids coding for GFP-MS2 fusion protein (0.5μg/35 mm dish) and shank1-UTR/MS2re chimeric RNAs (2.5μg/dish) were transfected using the calcium phosphate method. The ratio of 5:1 (RNA:protein expression vector) was chosen to make sure that most or all of the GFP fusion proteins are associated with RNA, as described in (8). Initial experiments with higher amounts of the GFP-MS2 vector, or higher total amounts of DNA, failed as large, immobile aggregates of GFP fusion protein were observed (data not shown). For time-lapse video microscopy, neurons were plated on glass-bottom dishes (MatTek) in order to provide an interface to the microscope objective. Microscope and image acquisition were controlled by MetaVue6.2r6 software (Universal Imaging).

In situ hybridization and immunocytochemistry

For generating sense and antisense riboprobes, the full-length coding region of rat shank1a (kindly provided by Carlo Sala, Milano, Italy) as well as the mRFP coding sequence were cloned into pBluescript. After linearization using appropriate enzymes, in vitro transcription was performed using T7 or T3 RNA polymerases in the presence of digoxygenin-labeled UTP. Transcripts were subjected to base hydrolysis to obtain an average length of 200–300 nt. Fixed neurons were prehybridized in 50% formamide; 5× SSC; 5× Denhardts solution; 0.2% SDS; 50μg/mL heparin; 250μg/mL yeast tRNA; 0.25 mg/mL herring sperm DNA at 50°C for 2 h. Hybridization was performed in the same solution using 200–400 ng/mL specific RNA probe, at 50°C overnight. After washing, neurons were blocked and incubated with Cy3-labeled anti-DIG in blocking reagent (Roche). At this stage other primary antibodies were included, eventually followed by the appropriate fluorescently labeled secondary antibodies. Cells were viewed using an Axiovert135 microscope, equipped with a Hamamatsu camera. Openlab 2.2.5 software (Improvision) was used for image acquisition. In addition, an Axio Observer.Z1/Cellobserver in combination with Axiovision-Software was used. For quantitative evaluation, the number of GFP-positive granules in 50μm segments of the most intensely stained dendrite was determined, at a distance of at least one cell body diameter from the cell body, in at least three independent experiments. The number of colocalized particles was determined by overlaying micrographs from the in situ hybridization and immunocytochemical analyses. To determine background levels of colocalization, images were shifted relative to each other by the pixel equivalent of 0.5μm. This procedure yielded a colocalization for 7.2 ± 5.7% of shank1 mRNA particles with KIF5C particles.

Purification of mRNPs associating with KIF cargo binding domains

cDNA fragments coding for residues 826–920 of mouse KIF5C (database accession # NM_008449) and residues 689–796 of KIF3C (NM_008445) were cloned into pGEX4T2 (GE Biotech), allowing for bacterial expression of KIF5C/GST and KIF3C/GST fusion proteins. Fusion proteins were expressed in Escherichia coli and purified using GSH-sepharose (GE Biotech). Proteins were left on sepharose beads and used as affinity matrices for purification of KIF-associated cargoes, as described (10). Briefly, mouse brains were homogenized in buffer containing 20 mM HEPES (pH 7.4), 140 mM potassium acetate, 1 mM magnesium acetate, 1 mM EGTA, supplemented with protease inhibitors (Roche) and RNAse inhibitor. After centrifugation at 1000×g, the supernatant was further centrifuged at 10000×g. Supernatants were cleared with glutathione sepharose and then used for affinity chromatography. After washing, RNPs were eluted from the matrix with homogenizing buffer including 0.5 M NaCl. Samples were analyzed by western blot, as well as RNA isolation (RNeasy Mini Kit; Qiagen) followed by quantitative RT-PCR. RNA analysis was performed on a Rotor-Gene 3000 (Corbett) using the QuantiTect SYBR Green RT-PCR Kit (Qiagen). The RT step was performed at 50°C for 30 min followed by gene-specific PCR (cycle conditions: 95, 58, and 72°C for 15 seconds respectively; 40 repeats). The following gene-specific primers were used to detect co-precipitated mRNA molecules: GAPDH (Fw TGGCAAAGTGGAGATTGTTGCC; Rev AAGATGGTGATGGGCTTCCCG), αCamkII (Fw ACCTGCACCCGATTCACAG; Rev TGGCAGCATACTCCTGACCA); and shank1 (Fw AGCCTGCAGCAGTGCCCAGCA; Rev ATGCGAGGCCGCCAGGCCCA). For the arg3.1/arc mRNA the appropriate QuantiTect primer assay was obtained from Qiagen.

Luciferase assays

The coding region for Photinus luciferase was cloned into pBluescript to generate pBL which then contains a multiple cloning site for insertion of sequences of interest. Shank1 5UTRs derived from human and rat, as well as the arg3.1 5UTR were introduced upstream of the luciferase coding region. After linearization using BamHI, capped RNA was in vitro transcribed using T7 RNA polymerase (mMessage mMachine kit, Ambion). 100 ng of RNA was translated in vitro using rabbit reticulocyte lysate (Promega); luciferase activity was determined by the luciferase assay system (Promega), and a Berthold luminometer (Bad Wildbach). Bicistronic vectors were generated on the basis of pBicFire (40) and transfected into HEK cells or cortical neurons using the calcium phosphate method. Cell lysates were sequentially assayed for Photinus and Renilla luciferase activities using the dual luciferase assay system from Promega.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

We wish to thank Hans-Hinrich Hönck for excellent technical assistance; Robert Singer (New York), Roger Tsien (San Diego), Dietmar Kuhl (Berlin), Carlo Sala (Milano), Richard Vallee (New York), Antonino Schepis (Heidelberg), Philip Washbourne (Eugene, OR) and Jan Christiansen (Kopenhagen) for plasmids. Financial support by Deutsche Forschungsgemeinschaft (FOR885 to H.-J.K., S.K. and M.K.; Bo1718-4 to H.-J.K.) and the Fritz-Thyssen-Stiftung (to D.R. and S.K) is acknowledged.

References

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Materials and Methods
  6. Acknowledgments
  7. References
  8. Supporting Information

Supporting Information

Additional Supporting Information may be found in the online version of this article:

Video S1: Time-lapse microscopy of dendrites emanating from a single neuron expressing the GFP-MS2 and mRFP-MS2re−3 constructs. Images were acquired every 20 seconds for a total of 23 min.

Video S2: Section of a dendrite from a hippocampal neuron transfected as in Video S1. Images were acquired every 20 seconds for a total of 5 min.

Figure S1: Antibody specificity. A) Lysates of cultured rat hippocampal neurons (DIV15) were analyzed by western blotting using anti-Staufen1 and anti-KIF5C antibodies. To determine the specificity of the antibodies, the respective antigens (a GST-fusion of Staufen1; a synthetic peptide for KIF5C) were included as competitors (comp) as indicated. Equal loading was ascertained by a second exposition of the blots using anti-tubulin antibody. B) Cultured neurons were stained with anti-MAP2 (blue), anti-KIF5c in the absence or presence of the peptide antigen (green; left), or anti-Staufen1 (Stau1) in the absence or presence of the GST fusion which was used as the antigen (green, right) as indicated. Micrographs are shown as merged pictures with the MAP2 fluorescence (top), to indicate the position of dendrites, or as single fluorescence (bottom). Bar, 2μm

Figure S2: Structure of the shank1 5untranslated region. A) Structure of the genomic region encompassing exons 1 and 2 of the shank1 gene in human, rat and mouse. Whereas the human genome is fully sequenced, both rat and mouse genome sequences contain gaps in exon1, likely due to the high GC content. Two alternative transcripts observed in rat are indicated (38). B) cDNA alignment of human, rat and mouse shank1 5UTRs. Human sequence was used for most experiments in this study. Mouse data are based on five entries from the EST database; four of these (acc. BE957245; BE952753; AW123204; AW123209) contain sequence 3 to the gap at base pair (bp) 313–332; one of them (acc. CJ117711) is located 5 to this gap. As most of EST clones were generated using the restriction enzyme NotI, the data are consistent with the interpretation that base 326 is a (C) in the mouse sequence, generating a NotI site which disrupted EST clones. Rat data are based on the published 5UTR (38; acc. AF141902) and genomic sequences. Identity is indicated by dots, gaps in the alignment by ’–’ and unknown sequences (due to gaps in the genomic sequences or lack of EST clones) by asterisks. Two alternative start codons leading to synthesis of two alternative shank1 isoforms (differing by 70 amino acids) are underlined. Upstream open reading frames (uORFs) are indicated by bold print. uORF3 is not in frame with the shank1 open reading frame. C) cDNA sequence of the alternative rat 5UTR, which arises from a putative transcriptional start site in intron 1, thus including Exon2a. Sequences involved in overlapping uORFs1–3 are again indicated. uORF3 is identical to uORF3 in the spliced version shown in (B).

Figure S3: Specificity of dynamitin. The efficiency of the GFP-dynamitin construct in blocking dynein-based transport was assessed by staining transfected neurons for the Golgi-resident protein GOPC/PIST (35,36). Staining for PIST occurs in a typical Golgi-like fashion close to the nucleus with some extensions into the dendrites in cells which are either not transfected or expressing GFP only. In contrast, staining is fragmented and more diffuse in cells expressing GFP dynamitin, as has been described before (37)

Figure S4: Specificity of KIF DN constructs. Neurons were cotransfected with expression constructs coding for GFP-KIF5CDN1 and the NMDA receptor subunit NMDAR1 in fusion with the red fluorescent protein dsRed (39). Cells were stained using anti-MAP2 (blue staining). Fluorescent proteins were directly visualized as indicated

Figure S5: Quantitative analysis of dendritic mRNA localization. To determine the relative levels of RNA in cell bodies and dendrites, neurons expressing 5’-mRFP-MS2re-3&apos;together with GFP/control, GFP-KIF5CDN2 or GFP-KIF17DN as indicated were subjected to combined ICC/FISH analysis using anti-GFP (not shown) and a riboprobe directed against the mRFP-mRNA (left; shown in black and white for clarity). All neurons were photographed so that no saturation of the fluorescence signal was observed in the cell body, while attempting to make maximum use of the intensity spectrum of the microscope camera (0–255 gray scale). Using ImageJ software, the major dendrite was traced with the Freehand line tool, as indicated in the lower KIF17DN picture. Following the trajectory of this dendrite, the line was extended into the cell body, avoiding the nuclear region. Using the &apos;Plot Profile’ tool in ImageJ, the fluorescence intensity under this line was plotted as a 2D graph (intensity/gray value versus distance/dendrite length). The length of the cell body segment was measured and dendritic segments were defined as &apos;proximal’ for the first equivalent of a cell body length, and as &apos;distal’ for the next equivalent (see lower KIF17DN fluorescent picture). The area under the curve in each segment was integrated and taken as a measure of the amount of RNA present in these segments. In the Figure, these areas are identified as light grey for cell bodies; dark grey for proximal dendrites and black for distal dendrites. White areas, i.e. signal beyond the &apos;distal’ segment, were not included in the calculations. The ratio of the signal either in the &apos;proximal’ or the &apos;distal’ segments relative to the total signal (cell body+proximal+distal) is represented in Figure 6D. Marker for fluorescent images: 10μm

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
TRA_912_sm_FigureS1.tif2149KSupporting info item
TRA_912_sm_FigureS2.pdf76KSupporting info item
TRA_912_sm_FigureS3.tif3465KSupporting info item
TRA_912_sm_FigureS4.tif3750KSupporting info item
TRA_912_sm_FigureS5.tif3246KSupporting info item
TRA_912_sm_MovieS1.wmv1362KSupporting info item
TRA_912_sm_MovieS2.wmv136KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.