The Message and the Messenger: Delivering RNA in Neurons

See allHide authors and affiliations

Science's STKE  02 Apr 2002:
Vol. 2002, Issue 126, pp. pe16
DOI: 10.1126/stke.2002.126.pe16


Synaptic plasticity results in enduring changes in synaptic function. Localized protein synthesis is part of this process. Kosik and Krichevsky describe how a dynamic macromolecular structure, the RNA granule, may be a key element contributing to changes in protein production leading to synaptic plasticity.

With the confinement of DNA to the nucleus in the eukaryotic transition, RNA assumed the role of a spatial vector serving to separate translation from transcription by localizing protein synthesis to the cytoplasm. This evolutionarily ancient role of RNA cleared the way for the cell to establish independently regulated locales within which subsets of messenger RNAs (mRNAs) were segregated. Precise positioning of specific mRNAs along with regulatory elements and a translational apparatus was exploited in the oocyte and early embryo for creating developmental asymmetries (1). Synapses or clusters of synapses appear competent to control translation locally, not with the vastly disparate fates of the anterior and posterior poles in oocytes, but in a considerably more dynamic manner. Many synapses occur at structures called dendritic spines, which are morphologically distinct areas of synaptic activity. Indeed, thousands of synaptic connections on a single neuron can individually change their morphologies by forming and resorbing these spines, "retroconverting" spines to filopodia (2) or transforming spine shape while their receptor surfaces grow and shrink on a complex scaffold. The extent to which local translation contributes to this complex repertoire is an open question. Frequently cited in favor of the existence of local translation are the abundant single ribosomes and polysomes in dendritic shafts, which often extend to the base of spines and first led to the speculation that mRNA translation might occur in proximity to the postsynaptic compartment (3). Also frequently cited is the growing number of mRNAs that are distributed beyond the soma and into the dendrites. Some tallies place the number of dendritic mRNAs at over 400 (4) or about 5% of neuronal mRNAs. This pool of mRNAs encodes a multitude of postsynaptic proteins: scaffolding proteins (such as Shank) (5); protein kinases [such as calcium calmoduline kinase IIα (CaMKIIα] (6); receptors, including all ionotropic glutamate receptors; cytoskeletal proteins [such as microtubule-associated protein 2 (MAP2)]; transcription factors [such as cyclic AMP response element-binding protein (CREB)]; and neuropeptides (such as vasopressin), among others (4). Indeed, dendritically localized mRNAs can undergo local translation after synaptic activity (7, 8). Despite these "smoking guns" all through the dendrite, the mechanism is obscure by which local translation is coupled to synapse-specific changes that allow synapses or clusters of synapses to alter their thresholds enduringly in response to a presynaptic input.

A structure called the RNA granule, observed by light microscopy, contains some mRNAs. Most descriptions of RNA granules in neurons have been based on in situ hybridization and have suggested that many different mRNAs reside in these structures (9-13). RNA granules have been best visualized in dendrites; however, these granules can also be detected in neuronal somata, where they are difficult to resolve because of their own density. Granule formation can be seeded by an exogenously introduced mRNA (14, 15), and granules can act as discrete motile structures (9, 16) that carry specific mRNAs (15). RNA granules are often docked along microtubule tracks (17); however, a subset of granules was found to translocate at rates of ~4 μm to 6 μm per min (9, 18, 19), which is consistent with fast transport motors such as kinesin and dynein. Another subset of granules was found to oscillate near a cluster of synapses (15) in a motion referred to as "corralled vibration" (20).

Captured in situ, the RNA granule ultrastructure consisted of a nonmembrane-bound, densely packed cluster of ribosomes (9). RNA granules isolated from a discrete "heavy" peak by sedimentation gradients from cultured cerebrocortical neurons contained structures with a very similar ultrastructure to those observed in situ (21). This biochemically isolated fraction contained many mRNA transcripts but excluded essential elements of cap-dependent and cap-independent translation and failed to incorporate 35S-methionine, which suggests that RNA granules are translationally silent. When neurons were depolarized with KCl, RNA granules behaved as highly dynamic structures. By electron microscopy, they appeared to loosen the compact organization of their ribosomes and extend spirals of ribosomal beads from their periphery (Fig. 1) (20). They also released to translating polysomes mRNAs encoding proteins that may function in plasticity; specifically tested were those that encode the N-methyl-D-aspartate receptor (NMDAR)1 , the neurotrophin receptor TrkB, and CaMKIIα (20). A local burst of a specific set of mRNAs, ribosomes, and probably some translation factors may serve as a fast macromolecular mechanism to initiate translation locally at the site of synaptic activation. Thus, the granule could serve as a macromolecular structure capable of coupling RNA localization to activity-dependent translation [(Fig. 2); (Animation 1)]. This model is consistent with data that showed a rapid onset of local translation in isolated dendrites after glutamate stimulation, as detected by an exponential increase in green fluorescent protein signal that was strongest over sites of increased ribosomal density (4). The correlation between increased ribosomes and the thickness of postsynaptic densities in the stimulated rat neocortex (22) suggests that the site map of translational activation within a dendrite may be a reflection of its history of stimulation.

Fig. 1.

Electron micrographs of RNA granules from cultured cerebrocortical neurons. RNA granules isolated from the control culture appear as very densely packed clusters of ribosomes. RNA granules isolated from KCl-treated cells are less compact, with single ribosomes protruding from granules. Scale bar, 100 nm.

Fig. 2.

A model for local regulation of translation near synapses. RNA granules (GR) contain mRNA; ribosomes and ribosomal subunits; and other components. Some RNA granules are docked, whereas others are motile with oscillatory movements. A stimulus (red burst) induces reorganization and positional readjustment of granules in relation to a cluster of synapses and local release of stored mRNA/ribosomes/translation factors. P, preexisting polysomes; R, preexisting clusters of ribosomes; MT, microtubules. See (Animation 1).

It is tempting to speculate that RNA granules deliver a prepackaged collection of mRNAs in the correct stoichiometric ratios poised for rapid release like an "activation cartridge." By local release of mRNA, the granules would be capable of priming the local synthesis of those proteins needed for the more enduring changes that occur at synapses. Within a single neuron type, granules may harbor a stereotypical collection of mRNAs that undergo specific and coordinated regulation depending on local conditions. Once released, mRNAs may have different fates, depending on the local requirements for synthesis. Alternatively, a regulatory step at the point where granules are loaded may generate enough variety in the RNA composition of granules to account for different local circumstances and the finding of differences in mRNA composition between two branches from a single hippocampal neuron (23). Observed size diversity among RNA granules, as seen among electron micrographs of the rodent brain (, in which dense clusters of ribosomes that resemble RNA granules are present, may reflect mRNA compositional and functional differences in granule populations.

Selecting a population of mRNAs for local translation requires that they possess a delivery mechanism to reach the postsynaptic compartment. The signals that localize mRNAs to the postsynaptic compartment might also direct them to granules. Many, if not all, dendritic mRNAs carry cis-acting dendritic targeting signals within their 3′ untranslated regions (3′ UTRs). A precise identification of the minimal cis-acting element has not been determined for most dendritic mRNAs, and even when elements have been identified, there is inconsistency. For example, the 3′ UTR of CaMKIIα was competent to localize a reporter transcript to dendrites (24), and several cis-acting localization sequences have been identified in the 3′ UTR of CaMKIIα. A 30-nucleotide (nt) sequence (nts 26 through 58) in the 3′ UTR was sufficient for dendritic localization, whereas a more distal element in the 3′ UTR was postulated to inhibit dendritic localization in the resting state (25). A further study demonstrated that an 1100-base pair fragment (nts 2035 through 3129) of the 3′ UTR was also sufficient for dendritic localization (26). In addition to targeting signals, the 3′ UTR of CaMKIIα also contains elements that regulate its translation. The 3′ UTR of CaMKIIα contains two cytoplasmic polyadenylation elements (CPEs) and the hexanucleotide polyadenylation signal sequence, elements that are necessary for polyadenylation-induced translation. The CPEs of CaMKIIα bind cytoplasmic polyadenylation binding protein (CPEB), which interacts with the hexanucleotide sequence to initiate elongation of the polyA tail, a process that leads to the enhanced translation of CaMKIIα (27). In Xenopus, CPEB also has a localizing function in positioning the cyclin B1 mRNA in proximity to mitotic spindles (28). The CPEs and hexanucleotide sequence of CaMKIIα mRNA together can localize a transcript to the dendrite (29) and therefore have the potential to couple localization to translation. Multiple cis-acting localization elements in the CaMKIIα 3′ UTR may create a hierarchy of translocation signals, allowing multiple discrete pools of the transcript to exist and the transcript to be recruited to different locations in response to stimuli. In contrast, dendritic localization of the mRNA, which encodes the immediate-early gene product named Arc (activity-regulated cytoskeleton-associated protein) (30), is not dependent on stimulation, but instead the newly synthesized transcript is immediately translocated to the dendrite, and within the dendrite, precisely to the site of synaptic activation (3).

Directing mRNAs to their correct locale is only one facet of the problem. Once on site, an mRNA may undergo constitutive or regulated translation. Much is known about translational regulation in nonneural systems; however, most of these mechanisms act globally to coordinate the level of translation with the overall metabolism of the cell. In most physiological systems, global regulation is achieved by reversible modification of translation factors. Two steps of the initiation pathway have a central role in regulating translation. First is binding of the initiator methioine transfer RNA (Met-tRNAi) to the 40S ribosomal subunit, which is decreased by phosphorylation of eIF2α (31). Second is binding of the preinitiation complex (43S) to the 5′ terminus of mRNA, which is potentiated by the eIF4F complex, consisting of mRNA cap-binding protein, a scaffolding protein, and an RNA helicase (31). Selective regulation of some individual mRNA molecules may arise through the way in which a specific mRNA "senses" the general components of the translational machinery. For example, the mechanism underlying a depression in overall protein synthesis and a concomitant increase in CaMKIIα mRNA translation may involve NMDAR-mediated phosphorylation and inactivation of elongation factor EF2 (32). This modification should favor the up-regulation of translation of abundant but weakly initiated mRNAs such as CaMKIIα. Another way of regulating translation of a specific mRNA is through specific cis-acting sequences that bind trans-acting factors capable of interfering with the translational machinery. Translational regulation of local mRNAs probably bears a complex relationship to global translation. KCl depolarization of cultured rat cerebrocortical neurons resulted in a 15 to 30% decrease in global translation, as detected by a reduction in the amplitude of the polysome peak and the amount of labeled amino acid incorporated (21). At the same time, the translation of specific localized mRNAs increased. Neuronal stimulation appears to be capable of reprogramming translation in favor of mRNAs that are more spatially restricted, and an RNA granule may serve here as a local macromolecular regulator. However, the quantitative contribution of locally translated mRNAs to specific facets of synaptic strengthening is unknown.

Assembling all the components necessary to remodel a synapse is no trivial task. An initial signal, such as glutamate, will trigger rapid changes in local calcium concentrations that ramify over different distances and time scales to implement an orderly sequence by which proteins assemble into complexes and undergo posttranslational modifications leading to activation or inactivation. Delimiting calcium diffusion within the confines of the spine may represent a mechanism of activating local translation, whereas changes in somatic calcium may affect global translation. Whatever the contribution of local translation to plasticity, it is only a part of the story. A host of posttranslational modifications also occur, including phosphorylation, dephosphorylation, ubiquitination (33), proteolysis, assembly into polymers, membrane insertion (exocytosis), or endocytosis. The complex coordination of transcription (34), local translation (35), and posttranslational modifications has been well demonstrated in the single bifurcated Aplysia sensory neuron and its synaptic contact with spatially separated motor neurons.

The cast of protein players to implement these plastic changes is arriving onstage, and the elements of the choreography--the entrances, the exits, and who partners with whom--are just beginning to emerge. The independent ability of synapses or clusters of synapses to enduringly alter their thresholds and implement the many possible accoutrements of synaptic strengthening, such as spine formation and postsynaptic density perforation, requires mechanisms that each leave their historical trace on the dendrite. Calibrating translation based on local synaptic input will likely leave its own mark in the synaptic memory trace.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
View Abstract

Navigate This Article