Localization of tau mRNA to the axon requires the axonal localization cis signal (ALS), which is located within the 3′ untranslated region, and trans-acting binding proteins, which are part of the observed granular structures in neuronal cells. In this study, using both biochemical and morphological methods, we show that the granules contain tau mRNA, HuD RNA-binding protein, which stabilizes mRNA, and KIF3A, a member of the kinesin microtubule-associated motor protein family involved in anterograde transport. The granules are detected along the axon and accumulate in the growth cone. Inhibition of KIF3A expression caused neurite retraction and inhibited tau mRNA axonal targeting. Taken together, these results suggest that HuD and KIF3A proteins are present in the tau mRNA axonal granules and suggest an additional function for the kinesin motor family in the microtubule-dependent translocation of RNA granules. Localized tau-GFP expression was blocked by a protein synthesis inhibitor, and upon release from inhibition, nascent tau-GFP `hot spots' were directly observed in the axon and growth cones. These observations are consistent with local protein synthesis in the axon resulting from the transported tau mRNA.
Asymmetric localization of mRNAs as a regulatory mechanism for the determination of cellular asymmetry has been shown in yeast, Xenopus oocytes, Drosophila embryos and somatic cells ( Bassell et al., 1999; St Johnston, 1995). In neuronal cells, mRNA localization is involved in the development of neuronal polarity and neuronal function ( Kiebler and DesGroseillers, 2000; Steward, 1997; Tiedge et al., 1999). Restriction of mRNAs to a particular microdomain of the polarized cell and their local translation facilitates the rapid supply of proteins where they are needed ( Wilhelm and Vale, 1993). In neuronal cells, several localized mRNAs are upregulated by neuronal activity or following application of physiological stimuli ( Akalu et al., 2001; Kuhl and Skehel, 1998; Rook et al., 2000; Schuman, 1999; Zhang et al., 2001). As a result of specific neuronal stimulation, local translation of a green fluorescence protein (GFP) reporter in dendrites of hippocampal neurons, and even in isolated dendrites, has been observed ( Akalu et al., 2001; Eberwine et al., 2001; Job and Eberwine, 2001).
mRNA localization is initiated by association of the mRNA molecule, through a targeting signal most commonly located in the 3′ untranslated region (3′ UTR), with RNA-binding proteins (RBPs). Together they form ribonucleo-protein (RNP) granules that can translocate along the microtubules (MTs) ( Ainger et al., 1993; Aronov et al., 2001; Ferrandon et al., 1994; Knowles et al., 1996; Wang and Hazelrigg, 1994). Although the granules have been directly visualized in Drosophila embryos, oligodendrocytes and neurons, relatively little is known about their components or about their functional role in mRNA transport to the dendrites and axons ( Ainger et al., 1993; Ainger et al., 1997; Deshler et al., 1998; Hoek et al., 1998; Kiebler et al., 1999; Knowles et al., 1996; Kohrmann et al., 1999; Ross et al., 1997; Theurkauf and Hazelrigg, 1998). Biochemical analysis of granules in Drosophila oocytes has recently yielded information on the presence of at least seven proteins in the oskar mRNA granule ( Wilhelm et al., 2000). Granules in neuronal cells contain translational components including ribosomes that are packed in clusters and that are not translationally competent; instead, they serve as local storage sites for mRNA molecules ( Bassell et al., 1998; Krichevsky and Kosik, 2001). The identities and functions of the multiple proteins composing the granules, motor proteins and RNA-binding proteins are just emerging ( Gu et al., 2002; Hirokawa, 2000; Kikkawa et al., 2001; Ross et al., 1997; Tang et al., 2001).
Tau, a neuronal cytoskeletal protein, is a MT-associated protein (MAP) that stabilizes MTs and promotes their assembly. In neuronal cells, tau is found primarily in the cell body and axon. Axonal localization of tau mRNA to the proximal segment of the axon is dependent on 3′ UTR cis-acting signals, neuronal proteins and assembled MTs ( Aranda-Abreu et al., 1999; Aronov et al., 2001; Behar et al., 1995; Litman et al., 1993; Litman et al., 1994).
In a recent study to identify the minimal tau axonal localization signal (ALS), differentiating P19 embryonic carcinoma (EC) cell lines stably transfected with GFP-tagged tau constructs linked to fragments derived from the 3′ UTR region of tau mRNA were employed. The minimal tau ALS, which is required and sufficient for axonal localization, was identified. This region includes the stabilization sequence of tau mRNA, which binds to the HuD stabilization protein ( Aranda-Abreu et al., 1999; Good, 1997). We observed that tau mRNA was non-randomly distributed in the cells; instead it was localized as discrete granules along the axon and in the growth cone and it colocalized on the MTs with ribosomal proteins, which indicated the presence of protein synthesis machinery in the axon ( Aronov et al., 2001). RNA granules have been observed in fibroblasts, oligodendrocytes and primary neuronal cell cultures, suggesting that they may consist of RNA-protein complexes that contribute to the formation of cellular microdomains ( Bassell et al., 1999; St Johnston, 1995; Wilhelm and Vale, 1993).
In this study, we identify HuD and KIF3A as components of the tau RNP granules in the neuronal axon and growth cone. We suggest that these proteins are involved in the movement of the granules and the attachment of the granules to the MT tracks. Following targeting of tau mRNA to the axon, we observed local translation of GFP-tau protein as `hot spots' along the axon and in the growth cone, thus indicating simultaneous local translation, coinciding with tau mRNA localization in the distal processes. To our knowledge, this is the first axon-targeted mRNA to be shown to be locally translated in the axon.
Materials and Methods
P19 stable cell cultures
A P19 stable cell line expressing a GFP-tau coding-ALS construct (axonal localization signal of 240 bp from tau 3′ UTR 2529-2760) was used in this study ( Aronov et al., 2001). P19 cells were grown and differentiated as previously described ( Falconer et al., 1992) in MEM medium containing 10% fetal calf serum, 100 units/ml penicillin, 100 μg/ml streptomycin, and 2 mM L-glutamine in an incubator with 5% CO2.
In situ hybridization analysis, RNA probes and immunohistochemical staining
Differentiated P19 cells were fixed with 4% paraformaldehyde in the presence of 4% sucrose ( Litman et al., 1993). GFP (452 bp), tubulin (400 bp) and tau (400 bp) single-stranded RNA probes ( Aronov et al., 2001) were synthesized in the sense and antisense orientations, using the appropriate polymerase (T3 or T7 RNA polymerase) in the presence of digoxigenin UTP (RNA transcription kit, Boehringer Mannheim Biochemicals). In situ hybridization was performed as previously described ( Aranda-Abreu et al., 1999). For visualization of the in situ hybridization signals together with immunostaining, the slides were incubated overnight at 4°C with HRP-conjugated monoclonal anti-Dig conjugated to CY5 (1:500) (Jackson ImmunoResearch) together with the specific antibodies: monoclonal tubulin (1:200) (Sigma), KIF3A (1:30) monoclonal antibodies, KIF 1A(1:100) polyclonal [kindly provided by N. Hirokawa, Japan ( Kondo et al., 1994)], and HuD (1:1000) monoclonal antibodies 16A11 or human purified HuD antibodies [kindly provided by H. Furneaux, USA ( Marusich et al., 1994)]. The slides were washed with PBS and then incubated for 2 hours at room temperature with the appropriate secondary antibodies, that is, goat. anti-mouse (1:500) (Jackson), goat anti-human (Jackson), and goat anti-rabbit (Jackson) for the polyclonal and monoclonal antibodies, respectively. The coverslips were mounted with mowoil and visualized with an LSM confocal laser scanning imaging system equipped with a 40× objective, using the following laser wavelengths: for GFP, excitation 488 nm, emission 505-550 nm; for CY3-goat anti human, excitation 545 nm, emission 560-580 nm; and for tau mRNA labeled with CY5-digoxigenin, excitation 650 nm, emission 680 nm. For detection of endogenous tau mRNA, an RNA probe of tau repeats was used ( Fig. 3) ( Aranda-Abreu et al., 1999). KIF3A was detected by FITC-goat anti-mouse secondary antibodies, excitation 488 nm, emission 505-550 nm. Control experiments for in situ hybridization showed no background with the sense probe. In analysis of the fluorescence staining, no signal was observed in the absence of primary antibodies, and no penetration of signals between specific laser was detected.
Image analysis of in situ hybridization experiments was performed using a Zeiss LSM confocal microscope with LSM 510 software equipped with a 40× (1.0 numerical aperture) and 100× (1.4 numerical aperture) oil immersion lenses. The axon showing the in situ hybridization signal was divided into 20μ m segments, and the number of granules was counted and their size measured. For colocalization experiments, images of the same axon region observed by the GFP or rhodamine specific filter set and marked by the program coordinate system were analyzed. To determine whether a tau mRNA hybridization signal colocalized with HuD and KIF3A proteins, the coordinates of the particle and the proteins were compared. Only when the coordinates were identical were the particles considered to colocalize.
For quantitative analysis, 25-45 cells were analyzed on one coverslip for each experiment. Experiments were done with three coverslips for each variable, and each experiment was repeated at least three times. The data were analyzed by ANOVA and Student's t test, and statistical significance was determined for the experimental conditions.
Protein synthesis inhibition in P19 cells
Differentiated P19 cells (8-10 days) were treated with 50 μg/ml emetine (Sigma), a protein translation inhibitor, for 3 to 4 hours. At the end of the treatment, the medium was replaced with fresh medium, and the cells were left to recover for the specified time periods. Cells were visualized by confocal microscopy with GFP-adapted filters as described above. The quantitative analysis of mRNA and protein localization within the axon was measured by fluorescence (pixel) intensities. Regions of interest (ROIs) were marked, and the mean fluorescence intensity per pixel (±s.e.m.) was measured over time, relative to t0 ( Job and Lagnado, 1998). Care was taken to minimize `background pixels' within ROIs, and background fluorescence was subtracted from the mean fluorescence. The data were analyzed by ANOVA and Student's t-test to find the statistical significance under the various experimental conditions. The data were displayed by using Adobe Systems (Mountain View, CA).
To monitor tau protein synthesis during the time-lapse experiments, cells grown on coverslips were transferred to growth chambers maintained at 37°C on a heated stage. The cell medium was changed to Ham's F12 medium without phenol-red indicator. Fluorescence images were captured at the indicated time points with a cooled CCD camera and processed with LSM 510 software (Zeiss Germany). To minimize photobleaching and phototoxicity of the living cells, a computer-driven automatic shutter was used to achieve minimum illumination.
Antisense oligodeoxynucleotide (ODN) treatment of differentiated P19 cells
Differentiated P19 cells (8-10 days) were treated for 30 hours with 50μ M unmodified sense or antisense KIF3A ODN, essentially as previously described ( Aranda-Abreu et al., 1999). The medium of the treated cells was replaced every 10 hours, with fresh medium and antisense ODN. The sequence of the antisense KIF3A ODN was 5′-TCCTACTTATTGATCGGCAT-3′ (1-20) ( Kondo et al., 1994). No significant homology was found between the antisense KIF3A ODN and other sequences in the database. The morphological appearance of the treated P19 cells was observed by light microscopy. Following removal of the antisense ODN and replacement with fresh medium, the cells resumed normal morphology.
Immunoblot analysis of P19 protein extracts
Proteins were extracted from P19 cells (treated with KIF3A sense or antisense ODN) in 1 volume of lysis buffer consisting of 50 mM Tris pH 8.5, 1% Triton X-100, 5 mM EDTA, 0.15 M NaCl and 50 μg/ml PMSF. Cell debris was removed by centrifugation for 10 minutes at 16,000 g at 4°C. Protein samples (25 μg) were resolved by sodium dodecyl sulfate (SDS) polyacrylamide-gel electrophoresis, transferred to nitrocellulose filters and reacted with the specified monoclonal antibodies at 4°C for 16 hours. Following incubation for 1 hour at room temperature with HRP anti-mouse secondary antibodies (Jackson ImmunoResearch), the blots were developed using the ECL chemiluminescence technique.
Immunoprecipitation analysis of P19 cell extracts
The preparation of cell extracts and immunoprecipitations was performed as previously described ( Aranda-Abreu et al., 1999). Anti-HuD sera, anti-KIF3A and anti-β tubulin (Sigma) were added and the mixture incubated for hour at 4°C, followed by incubation with protein A-sepharose (Pharmacia, 10% final concentration). The immunocomplex was precipitated by centrifugation, washed, analyzed by SDS-PAGE and processed as already described.
Identification of HuD and kinesin proteins in the tau mRNA granules
In a previous study using differentiated PC12 cells, we detected an in vivo association between HuD RBP and tau mRNA ( Aranda-Abreu et al., 1999). This binding stabilizes the tau message and is dependent on an AU-rich element located in the tau ALS. The binding of the HuD protein is mediated by association with the MTs, which may anchor the message prior to its local translation ( Aronov et al., 2001).
To further characterize tau mRNA granules in differentiated P19 cells, we tested the colocalization of tau mRNA and HuD protein by in situ hybridization and immunostaining using confocal microscopy ( Fig. 1Aa). GFP-tau protein was seen in the cell body and axon ( Fig. 1Ab), which is similar to the location of transfected GFP-tau mRNA detected using the GFP probe in differentiated cell lines ( Fig. 1Ad). Immunohistochemical analysis using HuD-specific antibodies demonstrated that HuD protein is highly enriched in the neuronal cell body and in the axon of differentiated P19 cells, whereas only a faint staining was observed in the dendrites ( Fig. 1Ac). The merged image, presented at a higher magnification (×2500), of tau mRNA (red signal) and HuD protein (cyan signal) coincided with that of GFP-tagged tau protein (green signal), all of which overlapped to yield white-pink granules, which are distributed along the axon (filled arrowhead) ( Fig. 1Ae).
A merged image at high magnification of the axon and the growth cone is shown in Fig. 1B. The size of the small granules is between 0.3-0.5 μm (20 granules were measured in each axon of 18 different cells, P<0.01), and the size of the larger granules is between 0.7-1.5 μm (10 granules were measured in each axon of 20 different cells). The variation in the size of these granules, which include tau mRNA, suggests the presence of two size groups, which may represent the small granules and aggregates of granules attached to the MTs that were previously detected in neuronal and oligodendrocyte cells ( Knowles et al., 1996; Bassel et al., 1998; Ainger et al., 1997). The previously described small granules contain other mRNAs and exhibit dynamic behavior, whereas the larger granules were hypothesized to represent the translationally inhibited granules that were recently identified in neuronal primary cultures ( Krichevsky and Kosik, 2001).
Free HuD protein ( Fig. 1B), seen as cyan dots (marked by an asterisk), and not associated with MTs, may represent an oligomeric structure consisting of the protein alone or bound to other mRNAs that are not detectable by the GFP in situ hybridization probe ( Tenenbaum et al., 2000). Similar aggregates have been observed in embryonic cortical neuron cell cultures and HeLa cells stained with HelN1 and HuR antibodies, both of which are HuD homologues ( Brennan et al., 2000; Gao and Keene, 1996).
The kinesin motor family is involved in the directional movement of RNP granules along the MTs ( Carson et al., 1997; Severt et al., 1999). To determine whether kinesin expression affects axon outgrowth and tau mRNA sorting in differentiated P19 cells, we focused on one member of the kinesin motor family, KIF3A, which is localized primarily in the axons and is involved in anterograde movement ( Kondo et al., 1994). Our experimental data showed increased levels of KIF3A protein in the differentiated P19 cells, as detected by KIF3A-specific antibodies (data not shown). Immunostaining of differentiated P19 cells with KIF3A-specific antibodies showed intense staining in the cell body and axon and fainter staining in the dendrites. The staining of kinesin protein showed colocalization with the tau mRNA ( Fig. 2Ad), which was detected as pink granules in the axon of the enlarged confocal image, while no colocalization was seen in the dendrites ( Fig. 2Ae). A high magnification image of the axon and growth cone ( Fig. 2B) shows granules that contain tau mRNA, kinesin and GFP-tau protein distributed along the axon and in the growth cone (white-pink granules). Free kinesin protein, which did not colocalize with either tau mRNA or GFP-tau protein, is seen as cyan dots along the axon.
The above experiments indicated that tau RNP granules in the P19 cell line expressing GFP-tau ALS ( Aronov et al., 2001) included HuD and KIF3A proteins and prompted us to test the colocalization of both proteins in the tau mRNA granule. For that experiment, non-transfected cells were used (owing to technical limitations of visualizing more than three chromophores). Endogenous tau mRNA was visualized by in situ hybridization using an RNA probe derived from the region of tau 4 repeats ( Fig. 3b) together with HuD ( Fig. 3c) and KIF3A ( Fig. 3d) antibodies. Colocalization of KIF3A, HuD and endogenous tau mRNA is shown in Fig. 3e and is magnified twice compared with the magnifications shown in Fig. 1B or Fig. 2B for HuD and KIF3A, respectively. HuD and KIF3A proteins, which colocalize with tau mRNA, are present in the axon and growth cone. Quantitative analysis was performed on 10-12 fields from four separate experiments, analyzing the size and proportion of tau mRNA granules that colocalize with KIF3A or HuD proteins. The size of the granules ranged between 0.5-0.7 μm (P<0.01), and 57.5% of the granules exhibited tau mRNA and HuD (P<0.01), whereas 15-22% of the granules exhibited tau mRNA, HuD and KIF3A (P<0.001). These results demonstrated that only a fraction of tau mRNA granules colocalized with the KIF3A motor protein; this fraction may represent the granules that are capable of moving in the anterograde direction. Previous experiments using live cells imaging have shown that mRNA granules may move in both the retrograde and anterograde directions ( Rook et al., 2000; Zhang et al., 2001). The higher proportion of tau mRNA granules colocalized with HuD protein may indicate the multiplicity of functions that HuD is involved in; thus it is in the core of the granule. In the analysis obtained from this experiment, granules containing endogenous tau were visualized. These comprise about 15-20% of total axonal granules, as tested by uniform SYTO14 staining (S.A. and I.G., unpublished). Taken together, these results suggest that both HuD and KIF3A are components of the tau granules.
The effect of KIF3A antisense ODN treatment on differentiated P19 cells
The effect of KIF3A antisense ODN treatment on the distribution of tau mRNA was tested in differentiated P19 cells. Differentiated P19 cells were treated with ODN for a period of up to 30 hours, and the effect of the treatment on the axons is shown in the field view ( Fig. 4A) and at the single cell level ( Fig. 5). The treatment with kinesin antisense ODN caused axonal retraction, whereas in control sense ODN-treated cells no axonal retraction was observed, and both GFP-tau and kinesin were observed throughout the axons and reached the growth cones ( Fig. 4Aa,b). (Growth cones are seen in the upper right hand corners and marked by solid arrowheads; Fig. 4.) The average axon length of antisense-treated P19 cells dropped by 3.5-fold to 91.5μ m±13.91, whereas sense-treated cells exhibited axon lengths of 346.18 μm±38.15, P<0.05 (20-30 cells were measured per treatment, in four separate experiments).
The level of KIF3A protein was measured in total cell extracts prepared from untreated differentiated P19 cells and from cells treated with KIF3A antisense or sense ODN. The results of the western blot analysis ( Fig. 4B) demonstrate the effectiveness of the antisense ODN treatment. The level of KIF3A protein was 35% of control cells, whereas no effect was observed following treatment with KIF3A sense ODN. The level of KIF1A, neurofilament synaptophysin and MAP2 proteins remained unaffected, suggesting that the treatment is specific to KIF3A. The levels of endogenous tubulin and tau proteins dropped to 70% and 80%, respectively, which may reflect the retraction of neurites in the treated cells.
Closer analysis of a single cell treated with KIF3A antisense ( Fig. 5f-j) or sense ( Fig. 5k-o) ODN compared to the control cell ( Fig. 5a-e) is shown in Fig. 5. Antisense-treated cells had diminished GFP-tau fluorescent signals ( Fig. 5g) and tau mRNA levels ( Fig. 5h) as well as axonal retraction. The levels of GFP-tau protein and tau mRNA levels reached 50% and 55%, respectively, of sense-treated cells ( Fig. 5l,m), measured in pixels per unit area, as described in Materials and Methods. Measurements were taken from four different experiments, and 30 cells were analyzed. The in situ hybridization analysis with the GFP probe revealed that the tau mRNA signal is lower in the axon and in the cell body ( Fig. 5h) compared with control cells ( Fig. 5c). Moreover, its distribution is altered and is seen in the cell body concentrated close to the outer membrane ( Fig. 5h). The perturbed distribution of tau mRNA could have been caused by the inhibition of mRNA transport from the cell body to the axon. Previously, we demonstrated that when tau is not targeted to the axon, reduced levels of tau mRNA and protein are observed ( Aronov et al., 2001). Similar findings have been reported for myelin basic protein and α-CAMKII mRNAs, when translocation was disrupted by treatment with kinesin antisense ODNs in oligodendrocytes and hippocampal neurons, respectively ( Carson et al., 1997; Severt et al., 1999). We wondered whether reduced tau mRNA targeting results in lower tau protein levels and whether it affects MT stability and/or assembly, leading to reduced tubulin levels. To examine the specificity of the KIF3A ODN treatment, cells were stained with tubulin, and in situ hybridization was performed using the tubulin probe. The staining of tubulin shows the retraction of the axon in antisense-treated cell in comparison with sense ( Fig. 5n) and control untreated cells ( Fig. 5d), although the distribution pattern of tubulin mRNA is retained in the cell body ( Litman et al., 1993), and its intensity is not markedly reduced ( Fig. 5e,j,o). These results indicate that the reduced level of kinesin motor protein can disrupt axonal sorting of tau mRNA, which affects the distribution and levels of microtubule proteins involved in axonal growth.
To test for the physical association between HuD, kinesin and tubulin proteins, co-immunoprecipitation (IP) experiments were performed ( Fig. 6). Cell extracts were prepared from differentiated P19 cells and immunoprecipitated with HuD, kinesin and tubulin antibodies. The immunoprecipitates were separated on SDS gels and analyzed by immunoblotting with the specified antibodies. The results using the HuD antibodies ( Fig. 6A) demonstrate that kinesin coprecipitates with HuD protein (IP KIF lane), as indicated by the strong signal observed. When the presence of HuD protein was tested following immunoprecipitation with tubulin antibodies (IP Tub), a lighter HuD signal was observed, suggesting that HuD is associated with tubulin but at a lower affinity. This conclusion is supported by our previous studies, which showed an association between HuD protein and MTs, as tested by biochemical assays and by their colocalization as viewed by confocal microscopy ( Aranda-Abreu et al., 1999).
Fig. 6B shows the reciprocal assay in which the immunoprecipitated complex was immunoblotted with KIF3A antibodies. The results indicate co-immunoprecipitation of HuD protein with kinesin, suggesting an interaction between the two proteins. When the immunoprecipitation was performed with anti-tubulin antibodies (IP Tub) and tested for the presence of kinesin, a strong signal was observed, suggesting that a strong affinity exists between KIF3A and the tubulin. This supports previous studies that demonstrated kinesin movement on microtubules ( Hirokawa, 1998). The cumulative results suggest that an association exists between HuD and kinesin proteins. Furthermore, they indicate that the complex associates with MTs. Our previous results, demonstrating an association between tau mRNA, HuD protein and MTs, together with the results presented in this study, add KIF3A to the protein components present in the tau RNP granules.
Local translation of tau protein in the axons of living P19 neuronal cells
According to the multistep localization hypothesis, following the targeting of the mRNAs, local translation should occur, which may respond to external signals ( Wilhelm and Vale, 1993; Zhang et al., 2001). In a previous study we demonstrated the presence of GFP-tau mRNA, tau proteins and ribosomes in the axons of differentiated P19 cells ( Aronov et al., 2001). To examine the local translation of GFP-tau protein, emetine, an inhibitor of the translocation step in protein synthesis, was added to the differentiated P19 cells stably expressing a GFP-tau-coding construct, which included the ALS at its 3′ UTR ( Fig. 7). Application of emetine for 3-4 hours caused the disappearance of the GFP-tau fluorescent signal in the cell body and axon ( Fig. 7Ac, Fig. 7Ad). After a recovery period of 3 to 4 hours, GFP-tau protein fluorescence appeared as `hot spots' along the axon and in the growth cone ( Fig. 7Ae, Fig. 7Af). After 3 hours of recovery, the intensity ratio, expressed as pixels per unit area and measured in the cell body and the axon, was 1:2.5. The quantitative analysis was based on a total of 18 to 30 cells measured at each time point in four separate experiments.
To exclude the possibility that the GFP-tau protein observed in the axon originated from transport out of the cell body, time-course analysis of the local synthesis was visualized ( Fig. 8). GFP-tau protein was initially observed along the axon after 1.5 hours of recovery, whereas no signal was yet seen in the cell body. The signal was observed in the cell body only 2.5 hours later. (The cell bodies are circled in Fig. 8.) Moreover, the intensity of the GFP-tau `hot spots' was stronger in the axon than in the cell body. The increase in intensity of the axonal `hot spots', measured as fluorescence pixels per area unit, was highest (3.5 fold) during the first 30 minutes of observation. From 1.5-2 hours and 2-3 hours of incubation, the intensity increased by two- and 1.3-fold, respectively. Recently, using a similar approach in transiently transfected hippocampal neurons, local translation of dendritically targeted GFP-reporter protein was observed. It was suggested that the rate of GFP translation was exponential in the dendrites but linear in the cell bodies ( Job and Eberwine, 2001).
Since emetine affects protein synthesis, we checked whether tau mRNA was still present in cells after 3 hours of treatment with emetine ( Fig. 7B). Tau mRNA was observed in the cell body and axon ( Fig. 7Bc) when no GFP-tau signal was visible ( Fig. 7Bb), which is similar to the localization of tau mRNA in differentiated cells ( Fig. 1c).
On the basis of the time frame during which the GFP-tau `hot spots' were visualized along the axon and no signal was observed in the cell body, these observations are consistent with local protein synthesis in the axon being dependent on transported tau mRNA.
In this study, employing differentiated living P19 cells, we identified two proteins contained in tau mRNA granules that are highly concentrated in the cell body, axons and growth cones. The advantage of the P19 system is that it is not transiently transfected and therefore does not overexpress the studied mRNA. As such, it may mimic the regulation of endogenous tau mRNA ( Aronov et al., 2001). The identity of these granules was indicated by their specific in situ hybridization with tau probe, immunostaining with HuD and KIF3A antibodies and colocalization with MTs (Figs 1 and 2). These results are in agreement with our previous demonstration of in vivo binding of HuD to a specific sequence located in the 3′ UTR of tau mRNA and its colocalization with the MT system ( Aranda-Abreu et al., 1999; Litman et al., 1994).
The identification of HuD protein as a component of the tau RNP granule suggests multiple functions for this protein. It can act as an mRNA-stabilizing protein ( Aranda-Abreu et al., 1999) and can also serve as a linker protein to the MT tracks going down the axon. Interestingly, recent experiments have suggested that Hel-N1, a homologue of HuD, controls the translation of neurofilament M mRNA in human embryonic terato carcinoma cells through the recruitment of neurofilament mRNA into the heavy polysome fraction; this recruitment is dependent on sequences located in the 3′ UTR of neurofilament M mRNA ( Antic et al., 1999). This activity fits with the recent data demonstrating that RNA granules include mRNAs and clusters of ribosomes in a non-translated form, which upon activation move to the polysome fraction ( Krichevsky and Kosik, 2001). The Elav protein family was originally identified in Drosophila as being involved in neurogenesis and thus may function in an MT-dependent manner ( Antic and Keene, 1998; Aranda-Abreu et al., 1999; Aronov et al., 2001). As we demonstrated that HuD protein binds directly to a specific cis signal located in the tau 3′ UTR, this may suggest that HuD binding is among the initial events in mRNA granule assembly and is also involved in mRNA shuttling from the nucleus to the cytoplasm ( Campos et al., 1985; Ma et al., 1997). Moreover, recent data demonstrating that the Elav protein family binds to additional protein ligands, which do not bind directly to the targeted mRNA, adds additional insight into their function in diverse signaling cascades in vivo ( Brennan et al., 2000).
Studies on β-actin axonal localization in neuronal primary cultures have demonstrated that the accumulation of β-actin mRNA and protein in the axonal growth cones can be induced by dibutyryl cAMP treatment or stimulated by application of neurotrophins. This axonal localization depends on the formation of an RNP complex between β-actin mRNA with two proteins, ZBP1 and ZBP2, that have been identified through their specific binding to the axonal localization sequence within the 3′ UTR ofβ -actin mRNA. Both proteins have a role in β-actin mRNA localization, and ZPB2 most probably shuttles between the nucleus and the cytoplasm ( Bassell et al., 1998; Gu et al., 2002; Zhang et al., 1999; Zhang et al., 2001).
Deciphering which of the motor proteins, belonging to the kinesin or dynein family, translocate the granules along their specific tracks — either MTs or microfilaments — is a crucial and unresolved issue ( Hirokawa, 1998; Kikkawa et al., 2001; Schnapp, 1999). The specific interaction between these motor proteins and mRNAs may explain the asymmetric mRNA localization within the neuronal microdomains. These motor proteins use the cytoskeleton for active subcellular mRNA localization. Using HuD antibodies as bait for co-immunoprecipitation experiments, we found that KIF3A is present in the complex that associates with MTs ( Fig. 6). Moreover, as shown previously by RT-PCR analysis, tau mRNA was identified in the pellet immunoprecipated by HuD antibodies ( Aranda-Abreu et al., 1999). Similarly, tau-mRNA was identified in the KIF3A immunoprecipitate (data not shown). KIF3A belongs to the kinesin superfamily and has been characterized as an MT-based anterograde motor enriched in neuronal axons ( Hirokawa, 1998; Hirokawa, 2000; Kondo et al., 1994). A complex between the tau 3′ ALS region and a protein of a similar size to KIF3A was previously detected by a UV crosslinking assay using brain or neuronal cell extracts ( Behar et al., 1995).
To examine the possible functional association of tau with kinesin in vivo, we tested the effect of treatment with antisense ODN specific for KIF3A on neuronally differentiated P19 cells. The treatment caused retraction of neurites, with a more severe effect on the axons, where a higher concentration of KIF3A protein was detected. This treatment caused a specific decrease of KIF3A protein level in the cell, whereas the level of KIF1A protein was not affected. There was a concomitant decrease in tau mRNA levels, as tested by in situ hybridization, and a change in its distribution in the cell body, probably because of the inhibition of tau mRNA targeting. Previous studies have shown that translocation of myelin basic protein in oligodendrocytes requires MTs and kinesin ( Carson et al., 1997; Carson et al., 1998). In another study, inhibition of kinesin heavy chain expression by ODN treatment of neonatal rat hippocampal neurons inhibited dendritic localization of α-CAMKII. Although we cannot disprove completely the idea that the effects on tau mRNA and protein localization are caused by the secondary effect of axonal retraction, this possibility is less favorable since our preliminary results using SYTO14 staining of total granules ( Knowles et al., 1996) show that following anti-kinesin ODN treatment, the velocity of the granules is reduced, whereas their density per unit length remains the same (S.A. and I.G., unpublished). Therefore, the current results and the previous observations in oligodendrocytes and neurons indicate the function of kinesin motor proteins in targeting of mRNA granules to specific cellular microdomains ( Carson et al., 1997; Severt et al., 1999). The interaction of the kinesin proteins with MTs may link the granules to the MT-assisted ATP movement in the plus-end-oriented MTs present in the axons ( Kikkawa et al., 2001). Previous studies have shown that transport by the KIF3A complex exhibits a velocity of 0.3 μm/second and that its association with vesicles in rat axons is involved in fast axonal transport ( Hirokawa, 2000). This may imply its involvement in axonal sprouting events by supplying the required material ( Takeda et al., 2000).
We observed two major classes of granules in our studies, which may represent small granules and aggregates of granules, both of which are mostly concentrated toward the lower region of the axon and in the growth cones. The calculated size of the observed small granules was between 0.3 and 0.5 μm in diameter, similar to the size observed in neuronal cells and in lamellopodia of fibroblasts ( Kohrmann et al., 1999; Krichevsky and Kosik, 2001; Latham et al., 1994). These granules are smaller than the calculated size of localized particles in oligodendrocytes [0.7 μm in diameter ( Barbarese et al., 1995)], suggesting that the size may vary among different cells and that these particles can accommodate many proteins and mRNA molecules. The big granules, which may reach a size of 0.7-1.1 μm, may represent the densely packed RNA granules that have recently been characterized in primary neurons. It was suggested that they include components of the protein translational apparatus and clusters of ribosomes, which are in a translationally inhibited state and can be activated by external signals, leading to local protein synthesis ( Krichevsky and Kosik, 2001).
Following the targeting of mRNA granules to their specific microdomain, local translation can ensue, as predicted in the multistep localization pathway hypothesis ( Wilhelm and Vale, 1993). We showed the presence of ribosomal proteins in P19 axons, and the presence of additional protein synthesis components of the translational machinery has been demonstrated in the dendrites and axons of developing neuronal cells ( Bassell et al., 1998; Crino and Eberwine, 1996; Zhang et al., 1999). In addition, focal centers of local translational activity have recently been shown along mammalian axons ( Koenig and Giuditta, 1999; Koenig et al., 2000).
In this work, blockage of protein synthesis by emetine, an inhibitor of the translocation step of protein synthesis, caused the disappearance of GFP-tau protein from the axon, although tau mRNA was detected in the cell body and axon ( Fig. 7B). After releasing the cells from inhibition and following a recovery period, the presence of GFP-tau protein in the axon in the form of `hot spots' — indicating local protein synthesis — was observed. The initial presence of the GFP signal in the axon after 1.5 hours was later followed, after 2.5 hour interval, by its appearance in the cell body, demonstrating that GFP-tau local synthesis in the axon occurs independently from the cell body.
Local protein synthesis has recently been demonstrated in dendrites of neuronal cells and in isolated dendritic preparations, following recovery from inhibition of translation or after stimulation by pharmacological or neurotrophic factors. In all these experiments, the appearance of newly synthesized proteins has been on a similar time scale to the results observed in this study, which represent local mRNA translation ( Akalu et al., 2001; Eberwine et al., 2001; Job and Eberwine, 2001; Kacharmina et al., 2000). A recent study, which lends further support to the idea that growth cones exhibit a remarkable amount of cellular autonomy, showed that isolated Xenopus retinal neurons contain the protein synthesis machinery required for local protein synthesis in the growth cones in response to chemotrophic stimuli ( Campbell and Holt, 2001; Eberwine, 2001). The translocation of both tau and β-actin mRNAs to the axon, a movement which is MT dependent, suggests that growth factors can exert their effects by signaling through the MT system and further supports the involvement of the dynamic cytoskeleton in neuronal function.
Identification of the protein components present in the RNP granules may lead to a better understanding of their function in the movement of the granules and their regulation during neuronal differentiation and growth cone outgrowth. Some of the identified proteins may belong to conserved protein families or, alternatively, may share conserved elements in their structure that are involved in the localization of diverse mRNA species. The granules may contain additional trans-acting protein factors, which are less abundant, and may contribute to the specificity of movement in various cell systems, and/or at different stages of neuronal development. Further studies on the specificity of KIF3A motor protein in mRNP translocation will lead to identification of their multiple cargoes in anterograde microtubule-based function in the axon. Additional characterization of tau RNP granule components and the kinetics of their movement is currently underway in our laboratory.
This work was supported by grants from the Minerva Foundation, Germany, the Nella and Leon Benoziyo Center for Neuroscience, WIS, Grant No 1999149 from the United States-Israel Binational Science Foundation (BSF), Jerusalem, Israel. I.G. is the incumbent of the Sophie and Richard S. Richards Professorial Chair in Cancer Research. The authors wish to thank I. Nevo for editing the manuscript.
- Accepted July 17, 2002.
- © The Company of Biologists Limited 2002