Subcellular Rna Compartmentalization

PII: S0301-0082(98)00066-5

SUBCELLULAR RNA COMPARTMENTALIZATION

EVITA MOHR* University of Hamburg, Institut fuÈr Zellbiochemie und klinische Neurobiologie, Martinistraûe 52, 20246 Hamburg, Germany

(Received 15 May 1998)

AbstractÐThe phenomenon of mRNA sorting to de®ned subcellular domains is observed in diverse organisms such as yeast and man. It is now becoming increasingly clear that speci®c transport of mRNAs to extrasomal locations in nerve cells of the central and peripheral nervous system may play an important role in nerve cell development and synaptic plasticity. Although the majority of mRNAs that are expressed in a given neuron are con®ned to the cell somata, some transcript species are speci®cally delivered to dendrites and/or, albeit less frequently, to the axonal domain. The physiological role and the molecular mechanisms of mRNA compartmentalization is now being investigated extensively. Even though most of the fundamental aspects await to be fully characterized, a few interesting data are emerging. In particular, there are a number of dierent subcellular distribution patterns of dierent RNA species in a given neuronal cell type and RNA compartmentalization may dif- fer depending on the electrical activity of nerve cells. Furthermore, RNA transport is dierent in neurons of dierent developmental stages. Considerable evidence is now accumulating that mRNA sorting, at least to dendrites and the initial axonal segment, enables local synthesis of key proteins that are detrimental for synaptic function, nerve cell development and the establishment and maintenance of nerve cell polarity. The molecular determinants specifying mRNA compartmentalization to de®ned microdomains of nerve cells are just beginning to be unravelled. Targeting appears to be determined by sequence elements residing in the mRNA molecule to which proteins bind in a manner to direct these transcripts along cytoskeletal components to their site of function where they may be anchored to await transcriptional activation upon demand. # 1998 Elsevier Science Ltd. All rights reserved

CONTENTS

1. Introduction 508 2. RNA targeting to dendrites 508 2.1. Dierent classes of mRNAs are targeted to dendrites 508 2.2. RNA transport takes place along the cytoskeletal network 511 2.3. Velocity of dendritic RNA transport 511 2.4. mRNA transport in immature and mature neurons 512 2.5. Components of the translational machinery in dendrites 513 2.6. Targeting signals necessary for dendritic mRNA transport 514 2.7. Local translation of dendritic mRNAs 517 2.8. Translation is not required for mRNA sorting to dendrites 518 3. mRNAs located in axons 519 3.1. Axonal transcripts in vertebrates 519 3.2. Axonal mRNAs in invertebrates 522 4. Conclusion and perspectives 522 Acknowledgements 523 References 523

* Corresponding author. Tel.: 49-40-4717 4553; Fax: 49-40-4717 4541; e-mail: [email protected].

507 508 E. Mohr

ABBREVIATIONS AMP Adenosine monophosphate mRNA Messenger RNA Arc Activity-regulated cytoskeleton-associated NMDA N-Methyl-D-aspartate protein NT-3 Neurotrophin-3 Arg3.1 Activity-regulated gene 3.1 oligo(dT) Oligo-deoxythymidine BB Brattleboro OML Outer molecular layer BDNF Brain-derived neurotrophic factor OMP Olfactory marker protein CaMKIIaa-Subunit of the Ca2+/calmodulin-dependent OT Oxytocin protein kinase II pcp-2 Purkinje cell protein-2 CDCH Caudodorsal cell hormone poly(A) Polyadenylated CREB Cyclic AMP response element binding protein RER Rough endoplasmic reticulum DNA Deoxyribonucleic acid RNA Ribonucleic acid eEF Eukaryotic elongation factor RNP Ribonucleoprotein eIF Eukaryotic initiation factor rRNA ribosomal RNA ELH Egg-laying hormone SCG Superior cervical ganglion GA Golgi-apparatus SRP Signal recognition particle GAP Growth-associated protein SSR Signal sequence receptor GlyR Glycine receptor SSTR1 Somatostatin receptor subtype 1 Golf a-Subunit of the olfactory system-speci®c TGN Trans-Golgi network heterotrimeric G-protein tRNA Transfer RNA Insp3r 1 Inositol-1,4,5-Trisphosphate receptor type 1 UTR Untranslated region kDa kiloDalton Vg 1 Vegetal 1 LTP Long term potentiation VP Vasopressin MAP Microtubule-associated protein MBP Myelin basic protein

1. INTRODUCTION are particularly interesting models because mRNA transport is strictly controlled in a spatial and tem- The highly polarized nature of neuronal cells of the poral manner and it is absolutely required to allow central and peripheral nervous system requires elab- for correct body pattern formation [for review see orate and accurate sorting mechanisms of their St. Johnston (1995); Bassell and Singer (1997); macromolecular constituents. Intracellular transport Gavis (1997)]. While the molecular determinants of is detrimental for the generation and maintenance of mRNA targeting in nerve cells are still largely the polarized morphology and ultimately for cell unknown, studies performed in non-neuronal sys- communication within the neuronal network. tems indicate the involvement of cis-acting signals Continuous redistribution of macromolecules is inherent to the mRNA molecules to be transported probably required upon formation of new synapses and trans-acting protein factors which bind to these as well as remodelling of pre-existing ones, for signals either directly or indirectly via protein/pro- instance during the course of learning and memory tein interactions to guide the RNAs to their ultimate consolidation. How a neuron achieves the equip- intracellular destinations [for review see St. ment of distinct microdomains with a de®ned assort- Johnston (1995); Bassell and Singer (1997); Gavis ment of proteins that are needed at particular sites (1997)]. There is circumstantial evidence for similar for a cell to function as it does such as membrane- mechanisms to exist in neurons. The present review associated receptors and other factors involved in will summarize our current knowledge of individual synaptic plasticity is still being investigated. Initially, components of the subcellular mRNA transport ma- it has been assumed that proteins are generally syn- chinery in nerve cells and the question concerning thesized in the cell body and are subsequently deliv- the functional signi®cance of this process will be ered to sites that may be located at considerable addressed. distances from the cell somata, for instance in axons and dendrites. In recent years, however, a variety of mRNA species have been detected in neuronal processes indicating that a decentralized translation ma- 2. RNA TARGETING TO DENDRITES chinery might also be operative, at least in dendrites 2.1. Dierent Classes of mRNAs are Targeted to and in the initial axonal segment both of which pos- Dendrites sess protein synthesizing capacity (Steward and Levy, 1982; Steward and Ribak, 1986). Some RNAs Apart from BC1 RNA, a non-coding RNA poly- are delivered to distal axonal segments [for review merase III transcript (Tiedge et al., 1991) all of the see Mohr and Richter (1995)] which are believed to few RNA species residing in dendrites of various lack components necessary for translation, at least nerve cells are mRNAs (Table 1). BC1 RNA forms in mammals (Lasek and Brady, 1981). part of a ribonucleoprotein particle and its function Consequently, the physiological meaning of these has not yet been determined (Kobayashi et al., transcripts has remained obscure. mRNA transport 1992). This RNA is detectable in dendrites of var- to distinct locations within the cell is not restricted ious nerve cell types both in the rat central nervous to nerve cells but has been observed in various non- system as well as in primary cultured neurons neuronal systems. Developing systems such as (Tiedge et al., 1991). Transcripts encoding the Xenopus and Drosophila oocytes and early embryos microtubule-associated protein (MAP) 2, a marker Subcellular RNA Compartmentalization 509

Table 1. RNAs located in dendrites of various nerve cell tissues

Dendritic RNAs Species/tissue Reference

Arc/Arg3.1 Rat/hippocampus Lyford et al. (1995); Link et al. (1995) BC1 Rat/hippocampus Tiedge et al. (1991) BDNF Rat/hippocampus Dugich-Djordjevic et al. (1992) CaMKIIa Rat/hippocampus Burgin et al. (1990) Dendrin Rat/forebrain Herb et al. (1997) GAP 43 Rat/brain Landry et al. (1994) Glycine receptor a-subunit Rat/spinal cord Racca et al. (1997) Glutamate receptors Rat/hippocampus Miyashiro et al. (1994) InsP3 receptor Mouse/cerebellum Furuichi et al. (1993) MAP2 Rat/hippocampus Garner et al. (1988) Neurogranin/RC3 Rat/brain Landry et al. (1994) Oxytocin Rat/hypothalamus Mohr et al. (1995) Pcp-2/L7 Mouse/cerebellum Bian et al. (1996) SSTR 1 Rat/cortex Mohr, unpublished results Vasopressin Rat/hypothalamus Mohr et al. (1995)

Arc, activity-regulated cytoskeleton-associated protein; arg, activity-regulated gene; BC, brain cytosolic; BDNF, brain-derived neurotrophic factor; CaMKIIa, a-subunit of the Ca2+/calmodulin-dependent protein kinase II; GAP, growth-associated protein; InsP3, inositol trisphosphate; MAP, microtubule-associated protein; pcp, Purkinje cell protein; SSTR, somatostatin receptor. of the dendritic cytoskeleton, was the ®rst mRNA mal part. In contrast, the mRNA for the a-subunit shown to be located in dendrites of neurons of the of Ca2+/calmodulin dependent protein kinase II rat central nervous system (Garner et al., 1988). The (CaMKIIa), a protein involved signal transduction dendritic compartmentalization is particularly well mechanisms and highly enriched at postsynaptic observed in the neuropil layer of the hippocampus sites, is detectable even in the distal dendritic seg- where the mRNA extends into approximately one to ments of hippocampal and cortical neurons (Fig. 1), two thirds of the dendrites, referred to as the proxi- while dendritic targeting of the same RNA is not

Fig. 1. Dark-®eld micrographs showing the subcellular localization of mRNAs encoding glyceraldehyde phosphate dehydrogenase (GAPDH), microtubule-associated protein 2 (MAP2) and the a-subunit of the CaMKIIa in the rat dentate gyrus which is schematically depicted in (D). Coronal brain sections have been hybridized in situ with radiolabelled anti-sense oligodeoxynucleotides speci®c for the respective mRNAs. While GAPDH mRNA is con®ned to (A) the cell body layer, (B) MAP2 and (C) CaMKIIa transcripts extend into the proximal and distal dendritic ®elds (m, molecular layer), respectively. Courtesy of Dr Stefan Kindler, University of Hamburg, Germany. 510 E. Mohr observed in cerebellar Purkinje cells (Burgin et al., tein (arc) mRNA [also named arg3.1, Link et al. 1990). Recently, the prominent dendritic localization (1995)] is expressed by an immediate early gene of dendrin mRNA, encoding a protein of unknown induced by synaptic activity (Lyford et al., 1995). A function, in rat forebrain structures has been massive dendritic compartmentalization is observed reported (Herb et al., 1997). In an in situ hybridiz- particularly in the dentate gyrus of the hippocampal ation study aimed at characterizing the subcellular formation (Fig. 2) and, to a lesser extent, in cortical distribution of several mRNAs in the rat brain, the neurons. Arc/arg3.1 protein is enriched in cell mRNAs encoding neurogranin (also termed RC3) somata and in dendrites and is assumed to interact and the growth-associated protein (GAP) 43 have with the neuronal cytoskeleton. Hence, it might play been found in proximal dendrites of some cortical a role in the activity-regulated response by interact- neurons (Landry et al., 1994). While neurogranin/ ing with and modifying the function of preexisting RC3 is located at postsynaptic sites (Watson et al., proteins (Link et al., 1995; Lyford et al., 1995). 1992) GAP 43 is rather thought to be a presynaptic Kainic acid-induced seizure activity in rats of post- constituent, and a dendritic compartmentalization natal day 13 appears to be accompanied by the has not been observed in cultured sympathetic neur- transport of brain-derived neurotrophic factor ons (Bruckenstein et al., 1990; Prakash et al., 1997). (BDNF) mRNA to the proximal dendrites of hippo- However, the dendritic localization of GAP 43 campal neurons (Dugich-Djordjevic et al., 1992). mRNA is also suggested by RNA blot analyses of Activity-dependent transport of this RNA to den- synaptosome preparations enriched in mossy ®ber drites has also been observed in cultured neurons terminals of the hippocampal CA3 region (Chicurel (Tongiorgi et al., 1997). Finally, extrasomal target- et al., 1993). In one study, dendritic mRNAs have ing of transcripts is not restricted to cells of the cen- been characterized by reverse transcription/polymer- tral nervous system. Recently, expression of the ase chain reaction ampli®cation of material isolated genes encoding glycine receptor (GlyR) a- and b- from individual dendritic segments of cultured hip- subunits has been studied in neurons of the rat pocampal neurons. By employing these dicult and spinal cord. While mRNAs for the b-subunits are elaborate techniques, mRNAs encoding various glu- exclusively located in the cell bodies, those encoding tamate receptors have been ampli®ed (Miyashiro et various a-subunits show a somato-dendritic distri- al., 1994). Earlier in situ hybridization studies were, bution (Racca et al., 1997). however, unable to detect the corresponding tran- It may be concluded from these data that dendri- scripts in the dendritic domain (Craig et al., 1993). tic mRNAs can be classi®ed, at least to a certain This may be indicative for the existence of many extent, according to the functions of the proteins more mRNA species in dendrites of nerve cells in concentrations below the detection limit when conventional techniques are employed. Surprisingly, these studies have revealed the dendritic localization of glutamate receptor subtype 6 mRNA even though it was not detectable in the cell body (Miyashiro et al., 1994). This may re¯ect the actual distribution of the receptor mRNA within one particular neuron but could also be due to technical limitations of the procedure. In another brain region with a laminated structure, the cerebellum, dendritic mRNA transport to proximal and distal segments has been docu- mented for transcripts encoding inositol 1,4,5- trisphosphate receptor type 1 (Insp3r 1) and L7 (also termed Purkinje cell protein-2; pcp-2), a cerebellum-speci®c protein of unknown function, respectively (Furuichi et al., 1993; Bian et al., 1996). A biochemical approach was used to investigate the dendritic compartmentalization of the peptide hormone mRNAs encoding the vasopressin (VP) and oxytocin (OT) precursors in rat hypothalamic magnocellular neurons (Mohr et al., 1995), which was con®rmed by in situ hybridization analyses at the Fig. 2. Dendritic compartmentalization of arg3.1 mRNA light microscopical (Bloch et al., 1990) and ultra- in the dentate gyrus of the rat hippocampus after seizure. structural level (Prakash et al., 1997). Likewise, non- The dark ®eld micrographs demonstrate the expression of radioactive in situ hybridization studies have been arg3.1 mRNA in the (A) rat hippocampus in control ani- successfully employed to detect transcripts for the mals and (B) animals subjected to seizure activity. The cor- rat somatostatin receptor type 1 (SSTR 1) in proxi- onal rat brain sections have been hybridized to an arg3.1- mal dendrites of cortical neurons (Mohr, unpub- speci®c radiolabelled anti-sense riboprobe. The hybridiz- lished results). ation signal in the molecular layer of the dentate gyrus is While the aforementioned mRNAs appear to be uniformly distributed and is not associated with Nissl- stained cell somata. GL, granular cell body layer; ML, localized to dendrites constitutively there are other molecular layer containing the dendrites of granule cells. examples of dendritic compartmentalization being The arrowheads show the border of the of the CA1 sub- dependent on the electrical activity of nerve cells. ®eld of the hippocampal formation. Reproduced courtesy The activity-regulated cytoskeleton-associated proof Link et al. (1995). Subcellular RNA Compartmentalization 511 they encode. MAP2 and probably arc/arg3.1 are cultures was directed anterogradely only, a bidirec- cytoskeleton-associated proteins. Many proteins tional transport was observed in dendrites of older such as receptors and CaMKIIa are involved in sig- neurons. Transport along microtubules was nal transduction pathways. Others (BDNF, VP, OT) suggested because the neuritic granules delocalized belong to the large group of growth factors and neu- following treatment with microtubule-depolymeriz- rotransmitters, respectively. All of these proteins/ ing drugs. These particles may in fact represent the peptides, however, appear to exert important roles transport units in transit from the perikaryon to in synaptic functions. de®ned subcellular destinations in the neurites. It should be stressed and will be discussed in detail in 2.2. RNA Transport Takes Place along the Section 2.8, however, that ongoing translation is not Cytoskeletal Network required for dendritic mRNA transport to take place. It is interesting to note that similar obser- It has long been noted that RNA is not free to vations have been reported concerning the subcellu- diuse within the cell. The majority of translatable lar targeting of myelin basic protein (MBP) mRNA mRNAs is associated with the cytoskeleton, while in cultured mouse oligodendrocytes. Fluorescently untranslated mRNAs and individual ribosomes are labelled MBP transcripts rapidly formed granules soluble [for review see Hesketh and Pryme (1991); soon after cytoplasmic microinjection (Ainger et al., Singer (1992)]. Presumably, the process of speci®c 1993). Recent data show the co-localization of argi- mRNA localization is a multi-step process. Upon nyl-tRNA synthetase, eEF 1a and rRNA in addition entering the cytoplasm the nucleic acids have to be to MBP mRNA in the majority of particles transported within the cell to de®ned subcellular (Barbarese et al., 1995). The authors suggest these sites and they must be anchored at their ®nal desti- macromolecular assemblies might increase the over- nations. Evidence is accumulating that transport of all eciency of MBP synthesis in the cell periphery mRNAs to the dendritic domain is mediated by and might at the same time provide a vehicle for cytoskeletal elements, in particular by microtubules transporting speci®c mRNAs to distinct subcellular which appear to play an important role (see below). locations. Recently published data are in line with Alternatively, mRNAs could initially diuse within the view of the fundamental importance of an intact the cell and could subsequently be localized by bind- cytoskeleton for dendritic transport of mRNA in ing to previously transported receptor or anchoring the human brain. No dendritic compartmentaliza- proteins. Due to the physico-chemical properties of tion of neurogranin/RC3 transcripts was observed the cytoplasm diusion of large macromolecular in brain sections derived from Alzheimer disease components (see below) appears to be highly unli- patients in which the cytoskeleton is severely kely (Luby-Phelps et al., 1987; Luby-Phelps, 1994). aected. In contrast, dendritic targeting of this Again, primary cultured neurons have been used as RNA species was preserved in central nervous sys- a tool for investigating and characterizing the den- tem neurons of age-matched fronto-temporal dritic transport of polyadenylated RNA [poly(A) dementia patients in which the neuronal cytoskele- RNA]. Cultures of cortical neurons strongly hybri- ton appears to be unaected (Chang et al., 1997). dized with radiolabelled oligo(dT) both in the cyto- Given that mRNAs are transported along microtu- plasm and in proximal dendrites following detergent bules the involvement of motor proteins is strongly extraction methods which leave the cytoskeleton suggested. While in the axon microtubules are intact while soluble macromolecules are extracted. oriented with their plus ends towards the distal part These data have been interpreted to indicate that at (Heidemann et al., 1981), a mixed orientation is least a fraction of poly(A) RNA is physically associ- observed in dendrites (Baas et al., 1988). Yet, ated with the cytoskeleton (Bassell et al., 1994). whether or not MAPs belonging to the dynein or Cells cultured in the presence of drugs dissociating kinesin family are involved in dendritic mRNA either micro®laments or microtubules, respectively, transport is currently an interesting but unresolved suggest an interaction of poly(A) RNA with micro- question. tubules in somata, dendrites and the initial axonal segment (Bassell et al., 1994). Sorting of RNA to 2.3. Velocity of Dendritic RNA Transport dendrites has also been analysed in living neurons by using a membrane permeable nucleic acid stain By employing a mathematical computer modelling that binds predominantly to RNA (Knowles et al., approach the velocity of RNA transport to den- 1996). Label was distributed non-randomly and was drites has been calculated in hippocampal cell cul- organized in a granular pattern in dendrites. The tures in pulse/chase experiments (Davis et al., 1990). granules apparently exhibited a complex compo- Following exposure of neurons to media containing sition since they contained, in addition to poly(A) tritiated uridine for 1 hr (pulse) cells were returned RNA, the large ribosomal subunit and eukaryotic to unlabelled media (chase) for various intervals, elongation factor (eEF) 1a. Further investigations ®xed and subjected to autoradiography. The dis- will be necessary to elucidate whether these granules tance of labelling in dendrites relative to the harbour the complete protein biosynthesis machin- cell nucleus was determined, statistically evaluated ery and, thus, represent translationally competent and the mean distance of transport as a function of supramolecular structures as discussed by the time was calculated by assuming dierent velocities authors. Interestingly, a subpopulation of these of transport. Apparently, movement of newly granules was motile moving with an average velocity synthesized RNA was dierent in short of 0.1 mm sec1 (for a detailed discussion of mRNA (0.26 mm day1) and long (0.5 mm day1) dendrites motility see Section 2.3). While movement in young (Davis et al., 1990). RNA was transported at similar 512 E. Mohr velocities in nerve cells of dierent stages of devel- on the other hand, appears to be similar to that opment, but dierent transport velocities among the observed in mature nerve cells: while transcripts for dendrites of an individual neuron have been MAP2 and CaMKIIa show a somato-dendritic dis- observed. The results suggest that dendritic RNA tribution, GAP 43, tubulin and actin mRNAs are transport might be dierentially regulated in the con®ned to the cell somata at all developmental processes of a single neuron. Even though the stages and are never detected in neurites (Kleiman et underlying mechanisms remain to be de®ned the al., 1994). Somewhat dierent results have been data imply that dendritic RNA transport might be reported recently in a study investigating the subcel- more complex than initially anticipated. Possibly, lular distribution of mRNAs encoding the b- and g- dendritic transport involves a whole array of com- actin isoforms in immature (4 days in vitro) embryo- ponents as it is the case in the axon of nerve cells. nic rat cortical neurons (Bassell et al., 1998). While Recently, the transport velocity of an individual transcripts encoding the g-isoform are detectable mRNA, arc/arg3.1 mRNA, has been investigated only in the cell body b-actin mRNA is located in which moved rapidly to dendrites with an apparent both axons and dendrites with particularly high con- rate of ca 300 mmhr1 [corresponding to centrations in growth cones. The b-actin protein is 7.2 mm day1; Wallace et al. (1998)] which is very highly enriched in growth cones, and it is virtually similar to the movement of RNA granules in living undetectable in the cell body, suggesting that this neurons (Knowles et al., 1996). Estimates of the rate mRNA is not translated until it leaves the cell body. of dendritic transport of radiolabelled BC1 RNA (a Local translation may be required for a constant non-translated RNA) injected into the cytoplasm of replenishment of the actin monomer pool during the cultured neurons derived from embryonic rat su- period of neurite outgrowth. perior cervical ganglion (SCG) gave similar velocity In another study the composition of mRNAs in rates of ca 5±6 mm day1 (Muslimov et al., 1997). individual growth cones of hippocampal neurons These migration rates are faster by approximately kept in culture for 6±72 hr has been investigated by one order of magnitude than those of newly syn- a technique called reverse Northern blot analysis thesized RNAs measured by pulse/chase labelling (Miyashiro et al., 1994). After microdissection of experiments. This may be attributable to the fact growth cones from the cell bodies their contents was that incorporation of [3H]uridine labels a hetero- soaked into a pipette and subjected to reverse tran- geneous population of RNA that is predominantly scription/polymerase chain reaction ampli®cation composed of ribosomal and transfer RNAs. techniques resulting in the generation of radiolabelled ampli®ed RNA. This was subsequently 2.4. mRNA Transport in Immature and Mature hybridized to DNA probes immobilized on nylon Neurons membranes, a procedure termed reverse Northern blotting (Crino and Eberwine, 1996). Strikingly, 22 In the majority of cases RNA compartmentaliza- out of 31 candidate mRNAs investigated in that tion is being investigated in mature nerve cells either study have been detected in growth cones of neurons in brain sections or in neuronal cells in cultures. cultured for 72 hr, and nine of these were detectable From these data it is clear that most mRNAs are as early as 6 hr following plating including mRNAs located in the cell body, while only a few are tar- for neurotrophin receptors and ion channels (Crino geted to dendrites but not to axons (exceptions will and Eberwine, 1996). The identi®cation of so many be discussed in Section 3). There is increasing evi- dierent mRNA species in dendritic growth cones dence to suggest that somewhat dierent mechan- could in fact indicate that adequate mRNA sorting isms might be operative in very young neurons that mechanisms to dendrites are not fully developed in do not yet show the mature morphological appear- young neurons. If true, one would expect many of ance of fully dierentiated nerve cells. In hippocam- the mRNAs expressed by the nerve cell at a particu- pal neurons kept in culture for only a few days lar time point of development in vitro to be present newly synthesized poly(A) RNA is not only targeted in dendritic growth cones. Also, since the target to the dendritic processes but, at early time points molecules have been subjected to ampli®cation these after plating, substantial amounts of mRNA and data are hard to interpret in terms of the relative rRNA are detectable in the axonal compartment abundance of growth cone mRNAs in comparison with particularly high concentrations in growth to their concentrations in the cell somata. Even if cones. Upon further maturation, dierentiation and mRNA transport to dendrites might in principle be ongoing synapse formation the concentration of a strictly controlled mechanism, a few molecules are mRNA and rRNA gradually declines until they likely to escape the sorting machinery. become undetectable in axons of fully matured Consequently, very low concentrations of mRNA nerve cells (Kleiman et al., 1994). These data may molecules in dendrites that become only identi®able be interpreted in two ways. RNA targeting to imma- by employing powerful ampli®cation protocols ture axons may re¯ect the fact that speci®c sorting could in fact lack any physiological relevance. It mechanisms are not yet fully developed in young should also be considered that these and other ana- nerve cells. Alternatively, sorting to axons may be lyses have been done with neurons in culture. These speci®c and may be restricted to a narrow time win- cells have been introduced into an extremely arti®- dow during the period of dierentiation. These axo- cial surrounding and might not behave as they nal mRNAs may thus provide the opportunity to would do in their natural environment, at least not locally synthesize key proteins that are detrimental for a few hours after plating. It appears to be an im- for axonal development and/or axon guidance. portant issue to verify mRNA heterogeneity in den- mRNA sorting to the dendrites of young neurons, dritic growth cones of young neurons in vivo and to Subcellular RNA Compartmentalization 513 elucidate their functional role the more so since never detected in dendritic growth cones. Again, the rough endoplasmic reticulum (RER) which is dendritic distribution of signal recognition particle required for membrane protein synthesis appears to (SRP), as assessed by in situ hybridization with a be absent in growth cones of cultured hippocampal probe speci®c for 7SL RNA, the nucleic acid com- neurons [Tiedge and Brosius (1996); see Section 2.5]. ponent of the SRP, was very similar to that of SSR. These data are important, because they show that 2.5. Components of the Translational Machinery in membrane-associated and secretory proteins may Dendrites indeed be synthesized in certain dendritic segments. Reports on whether or not dendrites harbour orga- Several years before mRNAs have been noticed to nelles identical with or similar to the GA located in reside in dendrites of various nerve cell types the the cell body are, at least in part, contradictory. principal protein synthesizing capacity of this sub- This question has been addressed in a number of cellular domain has been suggested since ribosomes studies during the past years, because many of the have been detected by ultrastructural analyses key proteins at synaptic sites are membrane-associ- (Steward and Levy, 1982; Steward and Reeves, ated. Hence, it is tempting to speculate that these 1988; Chicurel and Harris, 1992). Support for this might be among the likely candidates for local on- hypothesis came from experiments reported by site synthesis in dendrites to allow for rapid modu- Torre and Steward (1992). These authors have lation and modi®cation of the macromolecular com- developed a cell culture system allowing incorpor- position of de®ned synapses upon demand. This ation of isotopically labelled amino acids into neur- view is, furthermore, strengthened since a number of ites after dissection from their cell bodies. Label was RNAs undergoing dendritic targeting encode mem- incorporated into nerve cell processes immuno- brane-bound receptors and secretory proteins reactive for MAP2, a marker of the dendritic cytos- (Table 1). Indeed, evidence for glycosylation of keleton. Protein synthesis was inhibited by newly synthesized proteins in dendrites is now ac- cycloheximide but only to a minor degree by chlor- cumulating. The issue was addressed by employing amphenicol indicating that in this system the mito- the above-mentioned cell culture system in which chondrial protein synthesis is negligibly low. living neurites have been separated from their cell Subsequent investigations of translation in bio- bodies (Torre and Steward, 1992). Dierent radio- chemically puri®ed synapto-dendrosome prep- labelled sugars were added to the culture media to arations are less informative. Even though assess protein glycosylation in either the RER (man- numerous proteins are synthesized in these prep- nose) or the Golgi compartment (galactose and arations (Rao and Steward, 1991), it must be kept fucose) in pulse-chase experiments (Torre and in mind that synapto-dendrosomes are considerably Steward, 1996). Incorporation of [3H]mannose was contaminated by non-neuronal elements such as observed in ca 30% of MAP2-positive neurites with glial cell processes (Chicurel et al., 1993). decreasing concentrations towards the distal tips. While all of the initially detected dendritic Mannose-incorporation was drastically diminished mRNAs encode proteins such as MAP2 and in the presence of tunicamycin, an inhibitor of N- CaMKIIa that are translated on free ribosomes or glycosylation in the RER. These data are in agree- polysomes, subsequently identi®ed mRNA species ment with the results reported by Tiedge and encoding membrane-associated or secretory proteins Brosius (1996). [3H]galactose and [3H]fucose were (for examples see Table 1) would require the pre- incorporated into 30±40% of isolated dendrites. sence of RER and the Golgi apparatus (GA) in den- Label was highest over the proximal (thicker) dendrites for protein sorting and post-translational dritic segments, while thin (presumably distal) seg- modi®cation. Extensive studies aimed at characteriz- ments were often not labelled. Inhibition of protein ing the complexity of individual components of the synthesis by cycloheximide reduced the incorpor- translational machinery in dendrites of primary cul- ation of sugars underscoring that it is mediated by tured hippocampal nerve cells has been performed glycosyl transferases in the RER and GA, respect- by two groups (Tiedge and Brosius, 1996; Torre and ively. The studies have been extended to pulse/chase Steward, 1996). In the ®rst study (Tiedge and labelling of intact nerve cells held under conditions Brosius, 1996) immunocytochemistry with anti- that inhibit vesicular tracking from the RER to bodies directed against a ribosomal protein con- the GA (addition of brefeldin A to the culture ®rmed the presence of ribosomes within dendrites. media) or between the GA and the cell membrane Staining was particulate, heterogeneous and some- (culturing cells at 208C), respectively. The dendritic times clustered with the distal tips often but not localization of both the RER and the GA was con- always stained. Very similar if not identical patterns ®rmed since either treatment did not aect incorpor- were observed with antisera directed against arginyl- ation of radiolabelled sugars into newly synthesized tRNA synthetase, eukaryotic initiation factor (eIF) proteins in dendrites. Immunocytochemical staining 2 and eEF 2, respectively, which likewise labelled with antibodies directed against marker proteins of cell bodies and dendrites but not axons. Of particu- the cis-, intermediate- and trans-Golgi compart- lar interest is the identi®cation of constituents of the ments was consistent with autoradiographic label- RER. Immunostaining of the a subunit of the signal ling pattern in isolated neurites and intact nerve sequence receptor (SSR) revealed that this com- cells. Golgi markers were detected in ca 70% of cells ponent of the translocation complex in RER mem- in the cell bodies only, one (usually the major) den- branes is located in some but not all dendritic drite showed immunoreactivity in ca 20% of the segments. Staining was often restricted to major cells while all dendrites were stained in 10% of neur- dendrites or major proximal segments but it was ons (Torre and Steward, 1996). Detection of Golgi 514 E. Mohr marker proteins in one dendrite of cortical neurons transgenic mice underscore the involvement of the has also been reported by Lowenstein et al. (1994). 3'-untranslated region (UTR) in dendritic mRNA Moreover, in peripheral nerve cells ¯at continuous compartmentalization. Transgenic lines lacking cisternae oriented in parallel to the dendritic plasma these sequences fail to sort the corresponding membrane that are immunoreactive for a protein mRNA to dendrites (Mayford et al., 1996). residing in the cis-Golgi network have been Dendritic localizer elements appear to be organized observed (Racca et al., 1997; Gardiol et al., 1997). in a complex way in transcripts encoded by the rat Immunocytochemical studies using rat hippocampal VP gene which is highly expressed in hypothalamic cells in culture, on the other hand, have failed to magnocellular neurons. VP and the closely related detect a trans-Golgi network protein (TGN38) in OT mRNAs are exceptional since they belong to a the dendritic domain (Krijnse-Locker et al., 1995). small group of RNAs that are located both in axons Whether or not these contradicting results are at- and in dendrites (see Section 3.1). In order to delin- tributable to a dierent ®xation protocol employed eate cis-acting sequences that may be responsible for in the latter study (Krijnse-Locker et al., 1995) VP and OT mRNA targeting to the neuronal pro- remains to be seen. cesses, eukaryotic expression vector constructs have been microinjected into the nuclei of primary cultured SCG neurons (Prakash et al., 1997). After 2.6. Targeting Signals Necessary for Dendritic injection of constructs expressing VP and OT mRNA Transport mRNA these were distributed throughout both cell Our understanding of the mechanisms of mRNA bodies and dendrites [Fig. 4(B and D)]. The mRNA sorting to dendrites of nerve cells is still at its residing in the dendritic compartment of SCG neur- infancy. It is hypothesized that, in analogy to the ons appeared to be organized in a granular form. As components required for subcellular sorting of tran- discussed in Section 2.2, similar observations have scripts in non-neuronal systems (which will be been made in other nerve cell types. In contrast, brie¯y discussed below), dendritic mRNA transport nuclear injection of constructs expressing the anti- is mediated by cis-acting signals residing in the sense RNAs failed to give rise to dendritic targeting RNA itself as well as trans-acting factors, that is, despite of an abundant expression in the cell somata proteins that bind to these signals to direct the [Fig. 4(C and E)]. These data are consistent with the nucleic acid molecules to their ultimate destination current view that the nucleotide sequence or a sec- along the cytoskeletal network (Fig. 3). Trans-acting ondary structure formed by a particular sequence factors involved in dendritic mRNA transport have (cis-acting signals) of the mRNA contains the sort- not been identi®ed so far. However, recent data ing signals. The dendritic targeting capacity of var- obtained by introducing CaMKIIa constructs into ious segments spanning the VP RNA has been analysed in more detail. The corresponding cDNA segments have been ligated individually to the 3'-end of a-tubulin cDNA such that they form part of the 3'-UTR. a-tubulin RNA has been chosen since it is not subjected to dendritic compartmentalization but remains con®ned to the cell somata (Prakash et al., 1997). As schematically summarized in Fig. 5, several parts of VP mRNA were able to confer dendritic targeting to the chimeric transcript. Consequently, the dendritic localizer elements within VP mRNA are redundant and they are also located in the coding region, even though the 3'-UTR alone is moderately able to direct the recombinant RNAs to the proximal dendritic domain [Mohr et al., unpublished results; Prakash et al. (1997); for review see Mohr and Richter (1997)]. It is interesting to note that targeting elements are located in the coding region of yemanuclein-alpha mRNA which shows a complex sorting pattern in the Drosophila oocyte (Capri et al., 1997) as well as in transcripts encoding a transcription factor called Ash-1 which is speci®cally localized to the presumptive daughter cell in budding yeast (Long et al., 1997). Thus, primary cultured SCG neurons are obviously equipped Fig. 3. Model of mRNA targeting to the dendrites of nerve with the machinery necessary for dendritic compart- cells: upon entering the cytosol (1) mRNAs destined to be mentalization of mRNAs even if these, like VP and transported to dendrites are believed to bind to trans-act- OT transcripts, are not endogenously expressed. ing factors, the RNA-binding proteins, to give rise to RNP Hence, molecules acting on dendritic mRNA trans- particles (2). These are subsequently recruited to the micro- port do not appear to be cell-speci®c, at least in tubular network (3) along which the protein/RNA complexes are guided to extrasomal locations. Eventually, part, a view which is in line with observations made these mRNAs become translationally activated on-site to in non-neuronal systems. mRNAs for the various enable local synthesis of proteins involved in synaptic func- actin isoforms, for example, are localized to dierent tions (4). subcellular sites in several cell types. As demon- Subcellular RNA Compartmentalization 515

Fig. 4. Subcellular distribution of VP and OT sense and anti-sense RNAs following injection of eukaryotic expression vectors schematically shown in (A) into the cell nuclei of primary cultured neurons derived from embryonic rat superior cervical ganglia. Expression of the inserted cDNA is driven by the cytomegalovirus (CMV) promoter. A short oligonucleotide derived from the bacterial b-galactosidase gene (b- gal) was inserted such that it forms part of the 5'-UTR of vector-expressed mRNAs. This allows subsequent detection of vector-expressed RNAs with a universal b-gal-speci®c anti-sense oligonucleotide. (B±E) VP and OT cDNAs were ligated individually into the expression vector to give rise to either sense or anti-sense RNA expression. The bovine growth hormone (BGH) poly(A) signal allows for addition of a poly(A) tail to the resulting RNAs. The subcellular distribution of vector-expressed RNAs was analysed by non-radioactive in situ hybridization procedures. VP and OT sense transcripts are located in the cell bodies and in dendrites while the corresponding anti-sense RNAs are con®ned to the cell somata. For experimental details see Prakash et al. (1997). Scale bar, 20 mm. 516 E. Mohr

Fig. 5. Schematic representation of dendritic localizer elements within VP mRNA. To de®ne the sequences mediating dendritic transport of VP mRNA eukaryotic expression vectors were designed and subsequently microinjected into cultured neurons. The expression vector is schematically shown in (A). Expression of any inserted cDNA is driven by the cytomegalovirus (CMV) promoter. A short oligonucleotide derived from the bacterial b-galactosidase gene (b-gal) was inserted such that it forms part of the 5'-UTR of vector-expressed mRNAs. It allows subsequent detection of vector-expressed RNAs with a universal b-gal-speci®c anti-sense oligonucleotide. Individual parts of the rat VP cDNA (nucleotide positions are indicated by numbers), schematically depicted in (B), were ligated to the 3'-end of rat a tubulin cDNA. The VP sequence forms part of the 3'-UTR of the chimeric mRNAs. The bovine growth hormone (BGH) poly(A) signal allows for the addition of a poly(A) tail to resulting transcripts. The subcellular distribution of vector-expressed RNAs was analysed by non-radioactive in situ hybridization procedures. Dendritic localizer elements within VP RNA, as summarized on the right of (B), are redundant and they are located in the coding region as well as within the 3'-UTR. It is obvious that individual subfragments from the region spanning nucleotide positions 201±595 were only able to direct the chimeric mRNAs to the proximal parts of the dendites. The full extent of dendritic targeting capacity to distal locations was mediated by a fragment spanning the complete nucleotide sequence from position 201±595 indicating a synergistic action of the various localizer elements. For experimental details see Prakash et al. (1997). strated by transfection studies correct sorting of the cells that cannot easily be manipulated in vivo and a-, b- and g-actin transcripts occurred even if a par- in cell culture. ticular isoform was not endogenously expressed in The non-random distribution of de®ned mRNA that cell type (Kislauskis et al., 1995). species has been described for many non-neuronal As mentioned above, VP and OT mRNAs are cell types and is particularly well characterized in localized to dendrites and to axons in hypothalamic oocytes and early embryos of Drosophila. In the magnocellular neurons. In SCG neurons, however, Drosophila oocyte, various maternally synthesized axonal compartmentalization of these transcripts mRNAs, among them bicoid and oskar transcripts, has not been observed (Prakash et al., 1997). It is are speci®cally targeted to the anterior and posterior conceivable that the machinery necessary for sorting pole, respectively, in a strictly controlled spatial and of distinct mRNA species to the axonal domain temporal manner. Impaired sorting of these mRNAs may be cell-speci®c or may be present in only a severely interferes with correct body pattern for- restricted number of nerve cell types. Alternatively, mation. Drosophila is an excellent model system for the mRNA levels in this compartment might be too studying subcellular RNA transport mechanisms low to allow for detection by in situ hybridization since molecular biological investigations can be with non-isotopically or even radioactively labelled complemented by detailed genetic analyses. probes. Combined data indicate that sequence elements From the above-mentioned data it is clear that residing in the 3'-UTR of targeted mRNAs and mRNA sorting to dendrites of nerve cells is depen- de®ned proteins are indispensable for subcellular dent on cis-acting signals residing in the RNA mol- RNA sorting to take place. A speci®c interaction of ecules to be transported. However, many steps of protein factors with the localizer elements has been the presumably complex temporal and spatial trans- demonstrated in some instances [reviewed in St. port pathway have still to be elucidated. This is lar- Johnston (1995)]. For example, the RNA-binding gely due to the fact that neurons are postmitotic protein Staufen which is involved in distinct steps of Subcellular RNA Compartmentalization 517 bicoid mRNA transport binds to a well character- at de®ned sites, for instance during synapse for- ized segment within the bicoid 3'-UTR containing mation and/or remodelling. Indeed, there are recent the localizer element (Ferrandon et al., 1994). reports in favour of this idea. In guinea pig hippo- Importantly, subcellular RNA transport apparently campal slices the eect of aerent pathway stimu- includes a component of translational control, since lation on protein synthesis in dendrites of CA1 the bicoid mRNA is not translated until it reaches pyramidal cells was investigated by very brief its ®nal destination within the embryo. Local syn- (3 min) exposure to radiolabelled amino acids (Feig thesis of Bicoid protein, a transcriptional activator, and Lipton, 1993). In unstimulated cells, label incor- at the anterior pole then allows formation of a pro- poration into dendrites was ca 10% of that measur- tein gradient with decreasing concentrations towards able in the cell body layer. Addition of the posterior pole, a prerequisite for induction of a cycloheximide reduced the label over the cell bodies balanced gene expression along the axis of the by 90% but it had no eect on incorporation of embryo [reviewed in Whittaker and Lipshitz (1995)]. amino acids in the molecular layer indicating negli- Consequently, bicoid mRNA transport is detrimen- gibly low levels of extramitochondrial protein syn- tal because a morphogen gradient is indispensable thesis in dendrites. Patterned Schaer collateral for proper body pattern formation. Parenthetically, stimulation in the presence of the cholinergic agonist it is interesting that Bicoid has a dual role in that it carbachol increased the amino acid incorporation is not only a transcription factor but has also turned speci®cally into dendritic ®elds (almost three-fold) in out to act as an RNA-binding protein involved in a cycloheximide-sensitive manner, while neither car- translational repression. Bicoid binds to the 3'-UTR bachol nor Schaer collateral stimulation when of caudal mRNA, a transcript evenly distributed in given alone had any eect. The data strongly the Drosophila oocyte, and inhibits the initiation of suggest that de novo protein synthesis in dendrites translation. Thus, a posterior-to-anterior gradient of contributes to synaptic plasticity mediated by the the Caudal morphogen is generated (Rivera-Pomar combined activation of N-methyl-D-aspartate et al., 1996; Dubnau and Struhl, 1996). (NMDA) and cholinergic receptors (Feig and Lipton, 1993). 2.7. Local Translation of Dendritic mRNAs Neurotrophin-induced synaptic plasticity is another model that apparently requires immediate The presence of mRNAs and components of the protein synthesis in dendrites (Kang and Schuman, translational machinery in dendrites implies that 1996). Application of BDNF or neurotropin-3 (NT- proteins encoded by these transcripts may be syn- 3) to rat hippocampal brain slices enhanced synaptic thesized on-site. As pointed out in Section 2.5, there transmission at the Schaer collateral/CA1 pyrami- is indeed experimental evidence for protein synthesis dal neuron synapse which was markedly attenuated in the dendritic compartment. Yet, little is known after preincubation with eukaryotic but not prokar- about the spatial and temporal mode of translation yotic protein synthesis inhibitors. The requirement of de®ned mRNA species in dendrites. Experiments of immediate local de novo synthesis of proteins in supporting the view of local translation in dendrites the dendritic compartment rather than in the cell have been reported recently using transgenic mice body layer is strongly suggested since the eect was technologies (Mayford et al., 1996). Mice were gen- still observed after microdissection of the pre- and/ erated bearing a reporter gene (the bacterial b-galac- or postsynaptic cell bodies. It is hypothesized that tosidase gene) with an additional nuclear BDNF and NT-3 trigger local translation of preex- localization signal. Cell-speci®c expression was dri- isting mRNAs and that their translational products ven by the CaMKIIa gene promoter and dendritic are required for the induction of synaptic enhance- targeting of transgene mRNA was mediated by the ment (Kang and Schuman, 1996). Indeed, the eects CaMKIIa 3'-UTR. Histochemical b-galactosidase described above are observed within minutes and staining revealed, as expected, strong staining in the thus, they are too fast to allow for mRNA transport cell nuclei of the hippocampal formation but, in ad- from the perikaryon to dendrites prior to sub- dition, in the distal layer of the dendritic ®eld while sequent local translation (see Section 2.3). It is con- proximal dendrites showed little staining. Obviously, ceivable that they are attributable to preexisting the sorting machinery is capable to rapidly sequester mRNAs rendered translationally inactive until exter- protein synthesized in the cell somata and proximal nal signals trigger their rapid unmasking and sub- dendrites to the cell nucleus. The authors suggest sequent translation by mechanisms initiated by that the b-galactosidase in the distal tips of dendrites synaptic input. Their translational products may may have arisen by local translation of the chimeric belong to the key components required for the rapid mRNA. remodelling of synaptic connections. It will be extre- A currently unresolved question is whether the mely interesting to determine the identity of these subcellular distribution of particular proteins such proteins. as MAP2 and CaMKIIa, for instance, is exclusively Another potentially interesting facet of local pro- dependent on local translation of their mRNAs in tein synthesis in dendrites and its consequence for dendrites. In this very simple scenario the cell would synaptic plasticity has been addressed most recently lack the molecular machinery for sorting the corre- by Eberwine and his colleagues (Crino et al., 1998). sponding proteins (synthesized in the cell bodies) to They postulated that synaptic input might be con- the dendritic compartment. On the other hand, den- veyed directly from dendrites to the cell nucleus in dritic transcripts could ful®l more subtle functions order to modulate or alter the genetic activity within by supplying a pool of RNAs initially translation a particular neuron. To evaluate this hypothesis repressed but ready to be translated upon demand reverse Northern blotting techniques with radio- 518 E. Mohr labelled ampli®ed RNA extracted from individual with non-isotopically labelled probes will be dendritic segments (see Section 2.4) were performed required for unambiguous identi®cation of the cell with immobilized cDNAs encoding various tran- type harbouring NMDA receptor 1 transcripts in scription factors. This revealed the dendritic localiz- the lesioned animals. Given the dendritic transport ation of mRNA encoding CREB (cyclic AMP of this mRNA and its local translation during response element binding protein). Thus, it is tempt- synaptogenesis another interesting issue emerges. ing to speculate that synaptic input triggers local The dierent spatial distribution of NMDA receptor translation of dendritic CREB mRNA. Subsequent protein (in the OML) and the corresponding mRNA retrograde transport of the transcription factor pro- (throughout the dendritic ®eld) implies local intra- tein to the cell nucleus and the initiation of tran- dendritic control of translation. scription of responsive genes may then have an It may be envisioned that strong aerent acti- impact on the dendritic arborization and synaptic vation may also exert some in¯uence on the relative connections. abundance or spacial distribution of dendritic tran- With local protein synthesis in dendrites in mind scripts. However, induction of long term poten- it is also an interesting question whether synaptic tiation (LTP) by high frequency electrical input modulates the composition, the relative abun- stimulation of the perforant path did not result in dance and/or the subcellular distribution of mRNAs major changes in the distribution of MAP2 and in this compartment. The issue has been addressed CaMKIIa mRNAs although there may be a slight by several groups. For instance, the distribution of increase in the level of CaMKIIa transcripts in indi- MAP2 and CaMKIIa mRNAs in dendrites, particu- vidual animals (Steward and Wallace, 1995). Other larly in those of the granule cells of the dentate paradigms employed for induction of LTP have led gyrus, is quite dierent. Whereas CaMKIIa tran- to dierent results. Thus, perforant path stimulation scripts are distributed rather evenly within the entire in freely moving rats resulted in an increase in length of dendrites, MAP2 mRNA is con®ned the CaMKIIa mRNA which accumulated in the cell proximal part (see Fig. 1). These dierences in the body layer within the ®rst 12±24 hr and later on in distribution are reminiscent of the pattern of aer- the dendritic ®elds with maximum values at 48 hr ent innervation. The perforant path innervation post stimulation (Thomas et al., 1994). Recurrent (originating from the ipsilateral entorhinal cortex) limbic seizures, on the other hand, led to a reduction terminates on the outer two-thirds of the molecular of the same mRNA in cell bodies and dendrites with layer (OML, outer molecular layer) while the inner maximum reduction after 24 hr, especially in the cell third (IML, inner molecular layer) is innervated by body layer and the dendritic ®elds of the granule the commissural/associational system. Unilateral cell of the dentate gyrus, indicating that neuronal perforant path lesions, however, apparently had no activation has an impact on the levels of particular eect on the intradendritic distribution of these mRNAs in dendrites (Murray et al., 1995). A bipha- mRNAs when compared to the unlesioned contral- sic regulation of CaMKIIa mRNA has been ateral side, at least not within 2 and 4 days post observed following NMDA infusion into the mol- lesion (Steward and Wallace, 1995). Studies aimed ecular layer of the dentate gyrus. While an accumu- at characterizing the dynamic regulation of NMDA lation in somata and dendrites was detected within receptor 1 in the rat dentate gyrus after synaptic the ®rst 6 hr post infusion, a reduction was observed reorganization induced by unilateral perforant path after 24 hr (Johnston and Morris, 1995). With transection suggest dierential subcellular targeting regard to the mRNA encoding the immediate early of NMDA receptor 1 mRNA and protein (Gazzaley gene product arc/arg 3.1 it is clear that dendritic et al., 1997). During the phase of denervation, no mRNA levels are regulated by synaptic activity dierences in the relative abundance of the receptor (Link et al., 1995; Lyford et al., 1995). protein in the OML ipsi- and contralateral to the Electroconvulsive seizures induced the expression of lesion have been observed. Upon synapse reorgani- this gene particularly in the dentate gyrus with a zation, however, there was an increase in NMDA rapid transport of the corresponding transcripts to receptor 1 immunostaining that was restricted to the dendrites whereas other immediate early gene dendritic segments of the denervated OML of the mRNAs remained con®ned to the cell somata granule cell body layer. Interestingly, NMDA recep- (Wallace et al., 1998) indicating mRNA transport tor 1 mRNA levels at 5 and 9 days post lesion speci®city. increased concomitantly, but throughout the full extent of the molecular layer. The dendritic com- 2.8. Translation is not Required for mRNA Sorting partmentalization of this particular mRNA has not to Dendrites been observed so far (Laurie and Seeburg, 1994). This may re¯ect a very low basal level of expression Recruitment of mRNAs encoding secretory and that becomes suprathreshold during synapse reorga- membrane proteins to the RER requires signals in- nization. Since the NMDA receptor 1 gene is also herent to the growing polypeptide chain. A similar expressed in astrocytes the increase in hybridization mechanism could also be operative for transporting signal intensity in the dendritic ®eld of granule cells mRNAs to dendrites. However, evidence is now ac- might be due to an elevated gene expression in glial cumulating that ongoing translation is not necessary cells rather than to the dendritic compartmentaliza- for sorting the corresponding transcripts to den- tion of the corresponding RNA. However, at least drites. Kleiman et al. (1993) have made use of fully in primary cultured hippocampal neurons, NMDA dierentiated cultured hippocampal neurons in receptor 1 mRNA shows a somato-dendritic distri- order to assess dendritic transport of newly syn- bution (Gazzaley et al., 1997). In situ hybridization thesized RNAs in the presence of dierent transla- Subcellular RNA Compartmentalization 519 tional inhibitors. Neither cycloheximide nor puro- mRNAs located in the dendrites of nerve cells, the mycin treatment altered the subcellular distribution role of transcripts residing in the axonal compart- of newly synthesized RNA or of those transcripts ment is less clear. Notwithstanding, as summarized that are localized to dendrites in untreated cells even in Table 2, a variety of RNAs, often in substantial though the antibiotics interfere with dierent steps amounts, are clearly detectable in axons of various in translation. While cycloheximide blocks translo- nerve cell types from vertebrates including mammals cation and therefore traps mRNA molecules to ribo- and invertebrates. Invertebrate neurons, however, somes, the aminoacyl-tRNA analogue puromycin dier considerably from those of vertebrates, dissociates ribosomes bound to mRNAs. Hence, a because they usually do not show dierentiation of signal sequence-like targeting mechanism does not morphologically and functionally distinct dendrites appear to be involved in dendritic RNA transport. and axons. Instead, they rather develop only one It should be noted that translational inhibition lead type of process which is usually termed the axon but to the dendritic compartmentalization of a few should better be designated as neurite because it mRNAs, for instance those encoding GAP 43, actin harbours characteristics of both axons and dendrites and tubulin, that are restricted to the perikarya in [for review see Steward et al. (1995)]. For this non-treated cells. This may be indicative of the exist- reason, these systems will be discussed in separate ence of mechanisms that keep mRNAs in the cell sections. somata. However, recent data rather argue that this phenomenon may only be observed in cultured neurons, since in vivo administration of cyclohexi- 3.1. Axonal Transcripts in Vertebrates mide did not result in the dendritic localization of at least tubulin mRNA in the rat hippocampus while mRNA targeting of VP and OT mRNAs to the dendritic localization of arc/arg 3.1 mRNA in re- axons of hypothalamic magnocellular neurons has sponse to limbic seizures was unaected by anti- been extensively studied in many laboratories. Due biotic treatment (Wallace et al., 1998). to the lack of protein synthesizing ability of mam- Con®rmation for dendritic mRNA transport in malian axons (Lasek and Brady, 1981) the func- the absence of translation has been obtained by tional role of mRNAs directed towards this employing a dierent experimental approach. Using compartment has remained obscure until today. transgenic animal techniques the subcellular distri- Since hypothalamic magnocellular neurons exhibit a bution of pcp-2/L7 mRNA lacking any possible high secretory activity it is formally possible that translational start codons in mouse cerebellar some mRNA molecules might gain access to this Purkinje cells has been analysed (Bian et al., 1996). compartment by a passive, rather than an active The mutant transcript exhibited an intraneuronal transport process. The main constituents of antero- distribution indistinguishable from that of its wild- grade axonal ¯ow to the nerve terminals are the type counterpart. Thus, mRNAs appear to be trans- peptide hormone containing secretory granules. ported as pretranslational complexes that await mRNAs could, for instance, unspeci®cally adhere to translational activation at their ®nal destinations these organelles. Ultrastructural in situ hybridization within the cell. studies (Trembleau et al., 1994) have shown that VP mRNA is located in nerve swellings, the so called herring bodies, but never in undilated parts of the 3. mRNAs LOCATED IN AXONS axon. Moreover, in the posterior pituitary VP mRNA is located in nerve swellings and/or term- While a clear physiological relevance, namely inals the majority of which lack the peptide hor- local protein biosynthesis, can be ascribed to mone and vice versa (Trembleau et al., 1996).

Table 2. RNAs located in axons of various nerve cell tissues

Axonal RNAs Species/tissue Reference b-Actin Chick/sympathetic neurons Olink-Coux and Hollenbeck (1996) BC1 Rat/hypothalamus Tiedge et al. (1993) CaBP Mouse/cerebellum Bian et al. (1996) CDCH Snail/CNS van Minnen (1994) Dynorphin Rat/hypothalamus Mohr and Richter (1992) ELH Snail/CNS Dirks et al. (1989) MAP tau Rat/cortex Litman et al. (1993) MIP Snail/CNS van Minnen (1994) NF Squid/stellate ganglion Giuditta et al. (1991) NF-L Rat/hypothalamus Mohr and Richter (1992) Odorant receptors Mouse/olfactory bulb Ressler et al. (1994) OMP Mouse/olfactory bulb Vassar et al. (1994) Oxytocin Rat/hypothalamus Mohr et al. (1991) Pcp-2/L7 Mouse/cerebellum Wanner et al. (1997) Vasopressin Rat/hypothalamus Mohr et al. (1991)

BC, brain cytosolic; CaBP, calbindin; CDCH, caudo-dorsal cell hormone; ELH, egg-laying hormone; MAP, microtubule-associated protein; MIP, molluscan insulin-related peptide; NF, neuro®lament; NF-L, low molecular weight neuro®lament; OMP, olfactory marker protein; pcp, Purkinje cell protein. 520 E. Mohr

Consequently, secretory granules and VP-encoding highly unlikely (but cannot completely be disre- transcripts are not co-localized in any case. garded) that deletion of one nucleotide would inter- Furthermore, the axonal VP and OT mRNAs pos- fere with axonal mRNA targeting. However, the sess poly(A) tails that are shorter by ca 100 adenine mutant RNA lacks an in-frame stop-codon as a nucleotide residues when compared to those located consequence of the mutation and thus, may not be in the cell bodies. Even during experimentally released eciently from ribosomes. Subsequent local induced osmotic stress, which induces poly(A) tail degradation might render the RNA unavailable for elongation of both VP and OT mRNAs in the peri- axonal targeting. In heterozygous BB animals which karya, no change in the length of the poly(A) tracts express both the wild-type and the mutant allele, is apparent for transcripts residing in the axon wild-type but not mutant VP transcripts are sub- which should be expected if non-speci®c targeting jected to axonal compartmentalization consistent was the case. In addition, osmotic challenge is ac- with the view of transport taking place following companied by an approximately three-fold increase translation. In contrast, VP mRNA targeting to in both mRNA species in the cell bodies of magno- dendrites appears to occur prior to translation since cellular neurons. The axonal transcripts, however, mutant VP mRNA is readily detectable in dendrites accumulate in a very dierent manner. While the of homozygous animals [Mohr et al. (1995) and increase in OT mRNA levels is similar to that references therein]. observed in the cell bodies, VP transcript levels rise As far as the `why' is concerned, the purpose of at least 17-fold, therefore exceeding the mRNA mRNA compartmentalization to axons is comple- increase in the perikaryon (Mohr et al., 1991). A tely unknown so far. Almost certainly VP and OT dierential rate of accumulation in axons has also mRNAs are not locally translated in axons of hypo- been reported for BC1 RNA which is subject to den- thalamic magnocellular neurons, since they are dritic targeting in hippocampal neurons in vivo and clearly not associated with ribosomes or polysomes in cell culture (see Section 2.1). This RNA is not (Mohr et al., 1995). Thus, it remains a matter of detectable in axons of hippocampal cells by in situ speculation whether axonal transcripts are directed hybridization (Tiedge et al., 1991) but it is targeted out of the perikaryon in order to get degraded out- to axons of rat hypothalamic magnocellular neurons side the cell body or in order to be stored there for (Tiedge et al., 1993). Like VP and OT transcripts later use. It has been shown in a variety of systems axonal BC1 RNA levels increase (2.5-fold) during that mRNAs might transiently be rendered transla- osmotic stress with a maximum being observed after tionally inactive by association with proteins to three days of salt loading. At this time VP tran- form ribonucleoprotein (RNP) complexes also scripts still continue to accumulate. Within one day known as informosomes (Spirin, 1969, 1994). By after release of the osmotic stimulus, BC1 RNA employing electron spectroscopic imaging techniques levels have returned to control values whereas a sub- particles with a diameter of ca 20 nm that are remi- stantially slower decline is observed for VP mRNA niscent of RNPs have been detected in axons of rat which reached control levels seven days after release hypothalamic magnocellular neurons in the median of the osmotic stimulus (Trembleau et al., 1995). A eminence and the posterior pituitary [for review see mutant rat strain, the Brattleboro (BB) rat, has been Mohr et al. (1993) and references therein]. helpful in characterizing the temporal mode of VP Interestingly, these particles increase in number in mRNA sorting to axons. Assuming that axonal animals subjected to osmotic challenge, a condition mRNA transport is a speci®c process, the question known to result in a substantial elevation of axonal is when (and why) mRNA molecules are directed VP and OT mRNA levels (see above). Regardless of towards the axon. mRNAs actively translated at the the function that might be considered for axonal RER are probably not available for targeting. mRNAs, the combined data suggest completely Those molecules that have just left the cell nucleus dierent roles of the dendritic and axonal VP and but are not associated with ribosomes, or molecules OT mRNAs, respectively, as schematically outlined that are no longer translated but not yet fully in Fig. 6. degraded may be subject to axonal transport. Magnocellular neurons of the rat hypothalamo- Investigating the intracellular tracking of VP neurohypophyseal tract are not the only example mRNA in homozygous and heterozygous BB rats for cells that contain mRNAs within the axonal has provided evidence consistent with a targeting compartment. High levels of transcripts encoding mechanism that takes place following translation. the olfactory marker protein (OMP) as well as sev- The disease known as diabetes insipidus in the BB eral odorant receptors have been detected in the rat is caused by a frameshift mutation (deletion of a axon terminals of sensory neurons projecting to the single guanine nucleotide residue) in the second olfactory bulb in rats and mice (Ressler et al., 1994; exon of the VP gene. Although the mutant gene is Vassar et al., 1994; Wensley et al., 1995). Speci®city transcribed at an almost normal level, synthesis of is indicated by the complete absence of an equally the VP precursor is highly impaired. Due to the abundant mRNA, namely that encoding the a-subu- altered C-terminal portion the mutant protein is nit of the olfactory system-speci®c heterotrimeric G- unable to completely cross the membrane of the protein Golf, in the axonal compartment. RER and might eventually be degraded by the cha- Parenthetically, these studies have allowed to map peron system [Schmale et al. (1989) and references very precisely the topographic organization of sen- therein]. In the homozygous BB rat VP mRNA is sory projections to the olfactory bulb. However, undetectable in axons of magnocellular neurons they have also been unable to assess the functional when conventional Northern blot hybridization and signi®cance of axonally targeted OMP- and odorant ribonuclease protection analyses are employed. It is receptor mRNAs. It may in fact be argued that Subcellular RNA Compartmentalization 521

Fig. 6. Schematic view of the distribution of VP and OT precursors and of the corresponding mRNA in hypothalamic magnocellular neurons. The genes encoding the peptide hormone precursors VP and OT are expressed in two dierent populations of magnocellular neurons located in the hypothalamic para- ventricular and supraoptic nuclei. Their axons project via the median eminence and the hypophyseal stalk to the posterior pituitary. Following synthesis at the RER and passage through the GA the peptide hormone precursors are packaged into neurosecretory vesicles (NSV) that are transported down the axons to the nerve terminals. The hormones are stored there and eventually released into the systemic circulation upon appropriate physiological stimulation. NSV are also detectable in the dendrites of magnocellular neurons and both VP and OT are released into the CNS where they are believed to act as neu- rotransmitters and/or neuromodulators. The peptide hormone mRNAs are not con®ned to the cell body but are targeted to axons as well as to dendrites. Since dendrites contain ribosomes and small cisternae of RER it is conceivable that the dendritic mRNAs are locally translated. The axon, on the other hand, lacks components required for protein biosynthesis. Hence, axonal mRNAs may serve a function unre- lated to translation. They may, for instance, be stored for later use or they may be directed towards the axon in order to get degraded outside the cell body. Taken together, the axonal and dendritic peptide hormone mRNAs appear to exhibit completely dierent functions in hypothalamic magnocellular neurons. neurons harbouring mRNAs in the axonal compart- (Behar et al., 1995). Various parts of tau cDNA ment are obviously specialized neuronal cell types. have been fused to a reporter, the bacterial b-galac- Hypothalamic magnocellular neurons are excep- tosidase gene. Unfortunately, instead of determining tional because they are at the same time secretory the subcellular distribution of the chimeric tran- cells. Likewise, sensory odorant receptor neurons scripts directly by in situ hybridization, the authors dier from the majority of nerve cells in that they have taken the subcellular distribution of the b- have the capability to regenerate throughout the life- galactosidase protein to indicate indirectly the distri- time of the organism. bution of the corresponding mRNA. An intracellu- The situation may be dierent for mRNAs lar distribution of b-galactosidase immunoreactivity located in the initial axonal segment (as well as in resembling that of the endogenously expressed tau the cell body) such as MAP tau transcripts (Litman mRNA was observed with a construct containing et al., 1993). In contrast to the distal axonal segment the distal part of tau mRNA 3'-UTR while ligation which is generally believed to lack components of the proximal part to the reporter gene lead to a required for translation, at least in mammals (Lasek ubiquitous localization of b-galactosidase (implicat- and Brady, 1981), the initial segment is capable of ing a ubiquitous tau mRNA distribution). From protein biosynthesis (Steward and Ribak, 1986). these data it was concluded that the axonal targeting Consequently, mRNAs localized to this domain sequences are located in the distal part of tau 3'- may have functions comparable to those transcripts UTR. Proteins with an apparent molecular weight that are compartmentalized towards dendrites. of 38 and 43 kDa, respectively, extracted from Transfection studies have been done in order to de- whole brain cultured neurons or PC12 cells bound ®ne cis-acting sequences involved in tau mRNA tar- to a subregion within the distal 3'-UTR. Whether or geting to the axon of cultured whole brain neurons not they are the mediators of axonal tau mRNA tar- 522 E. Mohr geting remains to be seen (Behar et al., 1995). types of the snail central nervous system. There is Interestingly, tau mRNA transport to de®ned sub- now substantial experimental evidence for the local cellular sites was observed in a heterologous system, synthesis of proteins in the axonal compartment of namely in Xenopus oocytes which are well known these animals, especially in the snail [for reviews see for speci®c transport of a variety of maternal van Minnen (1994) and Steward et al. (1995)]. mRNAs to either the vegetal or animal pole [for Poly(A) RNA and components of the transla- review see Gavis (1997)]. Injection of the in vitro tional machinery such as ribosomes and tRNAs transcribed tau mRNA 3'-UTR led to sorting to the have been detected early on in the squid giant axon vegetal hemisphere whereas the coding region was (Giuditta et al., 1990; Giuditta et al., 1991). Several uniformly distributed within the oocyte (Litman et members of the axonal mRNA population have al., 1996). Thus, subcellular RNA sorting mechan- meanwhile been identi®ed, for instance those encod- isms appear to be conserved even in heterologous ing b-actin, tubulin, kinesin heavy chain, neuro®la- systems. Furthermore, RNA/protein interaction stu- ment protein or even proteins of unknown functions dies revealed binding of a 69 kDa protein to the tau [Chun et al. (1997) and references therein]. Direct 3'-UTR, a factor that might be implicated in trans- proof for local translation of mRNAs in the squid port of the endogenously expressed Vg1 mRNA to giant axon has not been obtained so far, but it the vegetal pole of Xenopus oocytes [for review see appears to be very likely in view of the above-men- Yisraeli et al. (1995)]. tioned data. Targeting to the proximal axonal segment of cer- In the pond snail, Lymnaea stagnalis, the intra- ebellar Purkinje cells of calbindin mRNA is dis- cellular distribution of various peptide hormone pre- cussed controversially. While a somato-axonal cursor mRNAs has been analysed in neurons that distribution of this transcript species was reported express more than one peptide hormone gene at the by one group (Bian et al., 1996) no such targeting same time. Strikingly, there is no correlation was observed in a recent study. Rather than being between the concentration of mRNA located in the localized into the proximal axonal segment calbindin cell body and in the axon, respectively. For instance, mRNA appeared to be concentrated towards the in caudo-dorsal cells of the snail central nervous sys- axonal pole (Wanner et al., 1997). It is interesting to tem expressing the caudo-dorsal cell hormone note that, by using highly sensitive in situ hybridiz- (CDCH) I and II genes the CDCH I mRNA levels ation techniques performed with ¯uorescently are high in the cell body while this transcript is tagged probes and subsequent confocal laser scan- hardly detectable in neurites. Vice versa, the CDCH ning microscopy, a proximal and even distal axonal II mRNA concentration is low in the perikaryon localization of pcp-2/L7 mRNA, which is abundant but the RNA is quite abundant in the axonal in the dendritic arbor of Purkinje cells (see Section domain (van Minnen, 1994). Recently, protein 2.1), has been demonstrated (Wanner et al., 1997). synthesizing capacity of neurites has been demon- This adds another example of RNAs localized to strated in cultured neurons from Lymnaea central both axons and dendrites of one and the same nerve nervous system by mRNA injection into neurites cell. isolated from their corresponding cell bodies. Selective RNA transport to the axonal compart- Synthetic egg-laying hormone (ELH) RNA has been ment takes also place in cultured chicken sympath- introduced by microinjection into individual neurites etic neurons. However, these cells have been of pedal A cluster neurons which do no express the maintained under conditions allowing outgrowth of corresponding gene endogenously. ELH-like immu- axons but not dendrites. Nevertheless, poly(A) noreactivity was readily detectable a few hours later RNA is detectable as a punctate pattern. RNA clus- strongly suggesting local translation in neurites of ters are clearly visible at axonal branch points, var- the injected RNA (van Minnen et al., 1997). The icosities and in some growth cones being general protein synthesizing capacity was further reminiscent of the distributions of RNAs in the den- con®rmed. Neurites isolated from their cell bodies dritic compartment. Speci®city of axonal RNA contain ribosomes and they incorporate radio- transport is suggested because axonal RNAs consti- labelled amino acids. Denaturing polyacrylamide gel tute only a subpopulation of the total mRNAs electrophoretic separation of newly synthesized pro- expressed in these neurons. Reverse transcription teins have, moreover, shown quantitative and quali- and subsequent ampli®cation by polymerase chain tative dierences in the patterns of proteins newly reaction techniques has, for instance, shown axonal synthesized in the cell somata and in isolated neur- targeting of b-actin, but not tubulin transcripts. ites, respectively (Bergman et al., 1997). Transport of RNAs to axons of chicken sympathetic neurons depends on intact microtubules. The presence of a translational machinery has not been 4. CONCLUSION AND PERSPECTIVES investigated. Thus, determination of the functional role of mRNA transport to axons of chicken sym- The last few years have shown that speci®c pathetic neurons must await further analyses (Olink- mRNA sorting is observed in various eukaryotic cell Coux and Hollenbeck, 1996). types throughout the animal kingdom. It appears to be one of the fundamental mechanisms operative in 3.2. Axonal mRNAs in Invertebrates cells in order to create and maintain polarity. Studying these phenomena in nerve cells is particu- Extrasomal mRNA transport has been investi- larly interesting because neurons certainly represent gated in detail in two invertebrate systems, namely one of the most complex cell types. They communi- in the squid giant axon and in various nerve cell cate via thousands of synapses with other nerve cells Subcellular RNA Compartmentalization 523 and their protein repertoire is extremely complex. It Bergman, J. J., Syed, N. I., van Kesteren, E. R., Smit, A. B., may be envisioned that the molecular composition Geraerts, W. P. M. and van Minnen, J. (1997) Modulation of local protein synthesis in neurites of identi®ed Lymaea neurons. of individual synapses and their rapid remodelling is Soc. Neurosci. Abstr. 23, 596. at least in part achieved by subcellular mRNA Bian, F., Chu, T., Schilling, K. and Oberdick, J. (1996) transport and local synthesis of synaptic constitu- Dierential mRNA transport and the regulation of protein syn- ents upon demand. Currently, we are only able to thesis: selective sensitivity of Purkinje cell dendritic mRNAs to translational inhibition. Mol. Cell. Neurosci. 7, 116±133. acknowledge the facts without a deeper understand- Bloch, B., Guitteny, A. F., Normand, E. and Chouham, S. (1990) ing of the molecular determinants leading to the Presence of neuropeptide messenger RNA in neuronal pro- regulation of the spatial and temporal mode of in- cesses. Neurosci. Lett. 109, 259±264. tracellular mRNA tracking in nerve cells. In fact, Bruckenstein, D. A., Lein, P. J., Higgins, D. and Fremeau, R. T. in some instances we even do not understand the (1990) Distinct spatial localization of speci®c mRNAs in cultured sympathetic neurons. Neuron 5, 809±819. physiological role which is true for mRNA transport Burgin, K. E., Waxham, M. N., Rickling, S., Westgate, S. A., to distal axonal segments of mammalian neurons. Mobley, W. C. and Kelly, P. T. (1990) In situ hybridization his- Yet, the available evidence suggests that even this tochemistry of Ca2+/calmodulin-dependent protein kinase in phenomenon is not a merely unspeci®c process. developing rat brain. J. Neurosci. 10, 1788±1798. Capri, M., Santoni, M.-J., Thomas-Delaage, T. and Ait-Ahmed, Forthcoming years will most likely ®ll the gaps and O. (1997) Implication of 5' coding sequences in targeting ma- provide the clues for understanding the key com- ternal mRNA to the Drosophila oocyte. Mechan. Dev. 68, 91± ponents involved in mRNA compartmentalization 100. to the neurites of nerve cells and its physiological Chang, J. W., Schumacher, E., Coulter, P. M., Vinters, H. V. and meaning provided adequate techniques will be at Watson, J. B. (1997) Dendritic translocation of RC3/neurogranin mRNA in normal aging, Alzheimer disease and fronto-tem- hand to address these questions. In this respect a poral dementia. J. Neuropathol. Exp. Neurol. 56, 1105±1118. recently developed technique, the yeast tri-hybrid Chicurel, M. E. and Harris, K. M. (1992) Three-dimensional system, deserves mentioning (Putz et al., 1996; analysis of the structure and composition of CA3 branched den- SenGupta et al., 1996). This modi®cation of the dritic spines and their synaptic relationships with mossy ®ber boutons in the rat hippocampus. J. Comp. Neurol. 325, 169± yeast two-hybrid system allows in vivo characteriz- 182. ation of proteins, expressed from a library of Chicurel, M. E., Terrian, D. M. and Potter, H. (1993) mRNA at cDNAs, speci®cally binding to target RNAs, for the synapse: analysis of a synaptosomal preparation enriched in instance to localizer elements mediating dendritic hippocampal dendritic spines. J. Neurosci. 13, 4054±4063. (or axonal) mRNA transport. As a major advan- Chun, J.-T., Gioio, A. E., Crispino, M., Eyman, M., Giuditta, A. and Kaplan, B. B. (1997) Molecualr cloning and characteriz- tage, the system allows identi®cation even of low ation of a novel mRNA present in the squid giant axon. J. abundant proteins and their cloning at the same Neurosci. Res. 49, 144±153. time. A detailed knowledge of the individual entities Craig, A. M., Blackstone, C. D., Huganir, R. L. and Banker, G. will pave the way to understand the underlying mol- (1993) The distribution of glutamate receptors in cultured rat hippocampal neurons: postsynaptic clustering of AMPA-selec- ecular mechanisms of subcellular RNA transport tive subunits. Neuron 10, 1055±1068. both in neurons and in non-neuronal cells. Crino, P. B. and Eberwine, J. (1996) Molecular characterization of the dendritic growth cone: regulated mRNA transport and local AcknowledgementsÐThe author thanks Dr Dietmar protein synthesis. Neuron 17, 1173±1187. Richter (University of Hamburg) for helpful discussion Crino, P., Khodakhah, K., Becker, K., Ginsberg, S., Hemby, S. and Eberwine, J. (1998) Presence and phosphorylation of tran- and critical reading of the manuscript, Drs Dietmar Kuhl scription factors in developing dendrites. Proc. natl Acad. Sci. and Stefan Kindler (University of Hamburg) for their U.S.A. 95, 2313±2318. contribution of Figs 1 and 2, and the Deutsche Davis, L., Burger, B., Banker, G. A. and Steward, O. (1990) Forschungsgemeinschaft for ®nancial support. Dendritic transport: quantitative analysis of the time course of somatodendritic transport of recently synthesized RNA. J. Neurosci. 10, 3056±3068. Dirks, R. W., Raap, A. K., van Minnen, J., Vreugdenhil, E., Smit, A. B. and van der Ploeg, M. (1989) Detection of mRNA mol- REFERENCES ecules coding for neuropeptide hormones in the pond snail Ainger, K., Avossa, D., Morgan, F., Hill, S. J., Barry, C., Lymnaea stagnalis by radioactive and non-radioactive in situ hy- Barbarese, E. and Carson, J. H. (1993) Transport and localiz- bridization: a model study for mRNA detection. J. Histochem. ation of exogenous myelin basic protein mRNA microinjected Cytochem. 37, 7±14. into oligodendrocytes. J. Cell Biol. 123, 431±441. Dubnau, J. and Struhl, G. (1996) RNA recognition and transla- Baas, P. W., Deitch, J. S., Black, M. M. and Banker, G. A. (1988) tional regulation by a homeodomain protein. Nature 379, 694± Polarity orientation of microtubules in cultured hippocampal 699. neurons: uniformity in the axon and not in the dendrite. Proc. Dugich-Djordjevic, M. M., Tocco, G., Willoughby, D. A., Najm, natl Acad. Sci. U.S.A. 85, 8335±8338. I., Pasinetti, G., Thompsen, R. F., Baudry, M., Lapchak, P. A. Barbarese, E., Koppel, D. E., Deutscher, M. P., Smith, C. L., and Hefti, F. (1992) BDNF mRNA expression in the developing Ainger, K., Morgan, F. and Carson, J. H. (1995) Protein trans- rat brain following kainic acid-induced seizure activity. Neuron lation components are colocalized in granules in oligodendro- 8, 1127±1138. cytes. J. Cell Sci. 108, 2781±2790. Feig, S. and Lipton, P. (1993) Pairing the cholinergic agonist car- Bassell, G. J. and Singer, R. H. (1997) mRNA and cytoskeletal bachol with patterned Schaer collateral stimulation initiates ®laments. Curr. Opin. Cell Biol. 9, 109±115. protein synthesis in hippocampal CA1 pyramidal cell dendrites Bassell, G. J., Singer, R. H. and Kosik, K. S. (1994) Association via a muscarinic. NMDA-dependent mechanism. J. Neurosci. of poly(A) mRNA with microtubules in cultured neurons. 13, 1010±1021. Neuron 12, 571±582. Ferrandon, D., Elphick, L., NuÈ sslein-Volhard, C. and St. Bassell, G. J., Zhang, H., Byrd, A. L., Femino, A. M., Singer, R. Johnston, D. (1994) Staufen protein associates with the 3' UTR H., Taneja, K. L., Lifshitz, L. M., Herman, J. M. and Kosik, of bicoid mRNA to form particles that move in a microtubule- K. S. (1998) Sorting of b-actin mRNA and proteins to neurites dependent manner. Cell 79, 1221±1232. and growth cones. J. Neurosci. 18, 251±265. Furuichi, T., Simon-Chazottes, D., Fujino, I., Yamada, N., Behar, L., Marx, R., Sadot, E., Barg, J. and Ginzburg, I. (1995) Hasegawa, M., Miyawaki, A., Yoshikawa, S., Guenet, J.-L. and Cis-acting signals and trans-acting proteins are involved in tau Mokoshiba, K. (1993) Widespread expression of inositol 1,4,5- mRNA targeting into neurites of dierentiating neuronal cells. trisphosphate receptor type 1 gene (Insp3r1) in the mouse cen- Int. J. Dev. Neurosci. 13, 113±127. tral nervous system. Receptors and Channels 1, 11±24. 524 E. Mohr

Gardiol, A., Racca, C., Rios, R. M., Bornens, M. and Triller, A. dence for evolutionary conservation of localization mechanisms. (1997) Dendritic and subsynaptic organelles involved in syn- Dev. Biol. 176, 86±94. thesis of transmembrane proteins. Soc. Neurosci. Abstr. 23, 649. Long, R. M., Singer, R. H., Meng, X., Gonzalez, I., Nasmyth, K. Garner, C. C., Tucker, P. R. and Matus, A. (1988) Selective local- and Jansen, R.-P. (1997) Mating type switching in yeast is con- ization of messenger RNA for cytoskeletal protein MAP2 in trolled by asymmetric localization of ASH1 mRNA. Science dendrites. Nature 336, 674±677. 277, 383±387. Gavis, E. R. (1997) Expeditions to the pole: RNA localization in Lowenstein, P. R., Morrison, E. E., Bain, D., Shering, A. F., Xenopus and Drosophila. Trends Cell Biol. 7, 485±492. Banting, G., Douglas, P. and Castro, M. G. (1994) Polarized Gazzaley, A. H., Benson, D. L., Huntley, G. W. and Morrison, J. distribution of the trans-Golgi network marker TGN38 during H. (1997) Dierential subcellular regulation of NMDAR1 pro- the in vitro development of neocortical neurons: eects of noco- tein and mRNA in dendrites of dentate gyrus granule cells after dazole and brefeldin A. J. Neurosci. 6, 1453±1465. perforant path transection. J. Neurosci. 17, 2006±2017. Luby-Phelps, K. (1994) Physical properties of cytoplasm. Curr. Giuditta, A., Menichini, E., Castigli, E. and Perrone Capano, C. Opin. Cell Biol. 6, 3±9. (1990) Protein synthesis in the axonal territory. In: Neurology Luby-Phelps, K., Castle, P. E., Taylor, L. and Lanni, F. (1987) and Neurobiology, vol. 59, pp. 205±218. Eds. A. M. G. Stella, J. Hindered diusion of inert tracer particles in the cytoplasm of de Vellis and J. R. Perez-Polo. Wiley±Liss, New York. mouse 3T3 cells. Proc. natl Acad. Sci. U.S.A. 84, 4910±4913. Giuditta, A., Menichini, E., Perrone Capano, C., Langella, M., Lyford, G. L., Yamagata, K., Kaufmann, W. E., Barnes, C. A., Martin, M., Castigli, E. and Kaplan, B. B. (1991) Active poly- Sanders, L. K., Copeland, N. G., Gilbert, D. J., Jenkins, N. A., somes in the axoplasm of the squid giant axon. J. Neurosci. Lanahan, A. A. and Worly, P. F. (1995) Arc, a growth factor Res. 28, 18±28. and activity-regulated gene, encodes a novel cytoskeleton-associ- Heidemann, S. R., Landers, J. M. and Hamborg, M. A. (1981) ated protein that is enriched in neuronal dendrites. Neuron 14, Polarity orientation of axonal microtubules. J. Cell Biol. 91, 433±445. 661±665. Mayford, M., Baranes, D., Podsypanina, K. and Kandel, E. Herb, A., Wisden, W., Catania, M. V., MareÂ chal, D., Dresse, A. (1996) The 3'-untranslated region of CaMKIIa is a cis-acting and Seeburg, P. H. (1997) Prominent dendritic localization in signal for the localization and translation of mRNA in den- forebrain neurons of a novel mRNA and its product, dendrin. drites. Proc. natl Acad. Sci. U.S.A. 93, 13250±13255. Mol. Cell. Neurosci. 8, 367±374. van Minnen, J. (1994) RNA in the axonal domain: a new dimen- Hesketh, J. E. and Pryme, I. F. (1991) Interaction between sion in neuronal functioning?. Histochem. J. 26, 377±391. mRNA, ribosomes and the cytoskeleton. Biochem. J. 277, 1±10. van Minnen, J., Bergman, J. J., van Kesteren, E. R., Smit, A. B., Johnston, H. M. and Morris, B. J. (1995) N-Methyl-D-aspartate Geraerts, W. P., Lukowiak, K., Hasan, S. U. and Syed, N. I. and nitric oxide regulate the expression of calcium/calmodulin- (1997) De novo protein synthesis in isolated axons of identi®ed dependent kinase II in the hippocampal dentate gyrus. Mol. neurons. Neuroscience 80, 1±7. Brain Res. 31, 141±150. Miyashiro, K., Dichter, M. and Eberwine, J. (1994) On the nature Kang, H. and Schuman, E. M. (1996) A requirement for local pro- and dierential distribution of mRNAs in hippocampal neurites: tein synthesis in neurotrophin-induced hippocampal synaptic implications for neuronal functioning. Proc. natl Acad. Sci. plasticity. Science 273, 1402±1406. U.S.A. 91, 10800±10804. Kislauskis, E. H., Ross, A., Latham, V. M., Zhu, X., Bassell, G., Mohr, E. and Richter, D. (1992) Diversity of mRNAs in the axo- Taneja, K. L. and Singer, R. H. (1995) The mechanism of nal compartment of peptidergic neurons in the rat. Eur. J. mRNA localization: its eect on cell polarity. In: Localized Neurosci. 4, 870±876. RNAs, pp. 185±195. Ed. H. D. Lipshitz. Springer-Verlag, Mohr, E. and Richter, D. (1995) mRNAs in extrasomal domains Heidelberg. of rat hypothalamic peptidergic neurons. In: Localized RNAs, Kleiman, R., Banker, G. and Steward, O. (1993) Inhibition of pro- pp. 275±287. Ed. H. D. Lipshitz. Springer-Verlag, Heidelberg. tein synthesis alters the subcellular distribution of mRNA in Mohr, E. and Richter, D. (1997) Neuroendocrine cells revisited: a neurons but does not prevent dendritic transport of RNA. Proc. system for studying subcellular mRNA compartmentalization. natl Acad. Sci. U.S.A. 90, 11192±11196. In: Neuroendocrinology. Retrospects and Perspectives, pp. 55±70. Kleiman, R., Banker, G. and Steward, O. (1994) Development of Eds. H.-W. Korf and K.-H. Usadel. Springer-Verlag, Berlin. subcellular mRNA compartmentation in hippocampal neurons Mohr, E., Fehr, S. and Richter, D. (1991) Axonal transport of in culture. J. Neurosci. 14, 1130±1140. neuropeptide encoding mRNAs within the hypothalamo-hypo- Knowles, R. B., Sabry, I. H., Martone, M. E., Deerinck, T. J., physeal tract of rats. EMBO J. 10, 2419±2424. Ellisman, M. H., Bassell, G. J. and Kosik, K. S. (1996) Mohr, E., Terjung, D., Martin, R. and Richter, D. (1993) Translocation of RNA granules in living neurons. J. Neurosci. Vasopressin mRNA in the hypothalamo-hypophyseal tract. In: 16, 7812±7820. Vasopressin, pp. 119±129. Eds. P. Gross, D. Richter and G. L. Kobayashi, S., Higashi, N., Suzuki, K., Goto, S., Yamoto, K. and Robertson. John Libbey Eurotext, Paris. Anzai, K. (1992) The 10S BC-1 ribonucleoprotein particle con- Mohr, E., Morris, J. F. and Richter, D. (1995) Dierential subcel- tains identi®er sequence-binding proteins that interact with an lular mRNA targeting: deletion of a single nucleotide prevents array of GCAAG/CTTGC motifs between split promoter the transport to axons but not to dendrites of rat hypothalamic sequences for RNA polymerase III. J. Biol. Chem. 267, 18291± magnocellular neurons. Proc. natl Acad. Sci. U.S.A. 92, 4377± 18297. 4381. Krijnse-Locker, J., Parton, R. G., Fuller, S. D., Griths, G. and Murray, K. D., Gall, C. M., Benson, D. L., Jones, E. G. and Dotti, C. G. (1995) The organization of the endoplasmic reticu- Isackson, P. J. (1995) Decreased expression of the alpha subunit lum and the intermediate compartment in cultured rat hippo- of the Ca2+/calmodulin-dependent protein kinase type II campal neurons. Mol. Biol.Cell 6, 1315±1332. mRNA in the adult rat CNS following recurrent limbic seizures. Landry, C. F., Watson, J. B., Kashima, T. and Campagnoni, A. Mol. Brain Res. 32, 221±232. T. (1994) Cellular in¯uences on RNA sorting in neurons and Muslimov, I. A., Santi, E., Homel, P., Perini, S., Higgins, D. and glia: an in situ hybridization histochemical study. Mol. Brain Tiedge, H. (1997) RNA transport in dendrites: a cis-acting tar- Res. 27, 1±11. geting element is contained within neuronal BC1 RNA. J. Lasek, R. J. and Brady, S. T. (1981) The axon: a prototype for Neurosci. 17, 4722±4733. studying expressional cytoplasm. Cold Spring Harb. Sym. Olink-Coux, M. and Hollenbeck, P. J. (1996) Localization and Quant. Biol. 46, 113±124. active transport of mRNA in axons of sympathetic neurons in Laurie, D. J. and Seeburg, P. H. (1994) Regional and developmen- culture. J. Neurosci. 16, 1346±1358. tal heterogeneity in splicing of the rat brain NMDAR1 mRNA. Prakash, N., Fehr, E., Mohr, E. and Richter, D. (1997) Dendritic J. Neurosci. 14, 3180±3194. localization of rat vasopressin mRNA: ultrastructural analysis Link, W., Konietzko, U., Kauselmann, G., Krug, M., Schwanke, and mapping of targeting elements. Eur. J. Neurosci. 9, 523± B., Frey, U. and Kuhl, D. (1995) Somatodendritic expression of 532. an immediate early gene is regulated by synaptic activity. Proc. Putz, U., Skehel, P. and Kuhl, D. (1996) A tri-hybrid system for natl Acad. Sci. U.S.A. 92, 5734±5738. the analysis and detection of RNA±protein interactions. Nucl. Litman, P., Barg, J., Rindzoonski, L. and Ginzburg, I. (1993) Acids Res. 24, 4838±4840. Subcellular localization of tau mRNA in dierentiating neur- Racca, C., Gardiol, A. and Triller, A. (1997) Dendritic and post- onal cell culture: implications for neuronal polarity. Neuron 10, synaptic localization of glycine receptor a subunit mRNAs. J. 627±638. Neurosci. 17, 1691±1700. Litman, P., Behar, L., Elisha, Z., Yisraeli, J. K. and Ginzburg, I. Rao, A. and Steward, O. (1991) Evidence that protein constituents (1996) Exogenous tau RNA is localized in oocytes: possible evi- of postsynaptic membrane specializations are locally syn- Subcellular RNA Compartmentalization 525

thesized: analysis of proteins synthesized within synaptosomes. Tiedge, H., Zhou, A., Thorn, N. A. and Brosius, J. (1993) J. Neurosci. 11, 2881±2895. Transport of BC1 RNA in hypothalamo-neurohypophyseal Ressler, K. J., Sullivan, S. L. and Buck, L. B. (1994) Information axons. J. Neurosci. 13, 4214±4219. coding in the olfactory system: evidence for a sterotyped and Tongiorgi, E., Righi, M. and Cattaneo, A. (1997) Activity-depen- highly organized epitope map in the olfactory bulb. Cell 79, dent dendritic targeting of BDNF and TrkB mRNAs in hippo- 1245±1255. campal neurons. J. Neurosci. 17, 9492±9505. Rivera-Pomar, R., Niessing, D., Schmidt-Ott, U., Gehring, W. J. Torre, E. R. and Steward, O. (1992) Demonstration of local pro- and JSÏ ckle, H. (1996) RNA binding and translational suppres- tein synthesis within dendrites using a new cell culture system sion by bicoid. Nature 379, 746±749. that permits the isolation of living axons and dendrites from Schmale, H., Borowiak, B., Holtgreve-Grez, H. and Richter, D. their cell bodies. J. Neurosci. 12, 762±772. (1989) Impact of altered protein structures on the intracellular Torre, E. R. and Steward, O. (1996) Protein synthesis within den- trac of mutated vasopressin precursor from Brattleboro rats. drites: glycosylation of newly synthesized proteins in dendrites Eur. J. Biochem. 182, 621±627. of hippocampal neurons in culture. J. Neurosci. 16, 5967±5978. SenGupta, D. J., Zhang, B., Kraemer, B., Pochart, P., Fields, S. Trembleau, A., Morales, M. and Bloom, F. E. (1994) Aggregation and Wickens, M. (1996) A three-hybrid system to detect RNA± of vasopressin mRNA in a subset of axonal swellings in the me- protein interactions in vivo. Proc. natl Acad. Sci. U.S.A. 93, dian eminence and the posterior pituitary: light and electron 8496±8501. microscopic evidence. J. Neurosci. 14, 39±53. Singer, R. H. (1992) The cytoskeleton and mRNA localization. Trembleau, A., Melia, K. R. and Bloom, F. E. (1995) BC1 RNA Curr. Opin. Cell Biol. 4, 15±19. and vasopressin mRNA in rat neurohypophysis: axonal com- Spirin, A. S. (1969) The second Sir Hans Krebs lecture: informo- partmentalization and dierential regulation during dehydration somes. Eur. J. Biochem. 10, 20±35. and rehydration. Eur. J. Neurosci. 7, 2249±2260. Spirin, A. S. (1994) Storage of messenger RNA in eukaryotes: Trembleau, A., Morales, M. and Bloom, F. E. (1996) Dierential envelopment with protein, translational barrier at 5' side, or compartmentalization of vasopressin mRNA and neuropeptide conformational masking by 3' side?. Mol. Reprod. Dev. 38, 107± within the rat hypothalamo-neurohypophysial axonal tracts: 117. light and electron microscopic evidence. Neurosci. 70, 113±125. Steward, O. and Levy, W. B. (1982) Preferential localization of Vassar, R., Chao, S. K., Sitcheran, R., NunÄ ez, J. M., Vosshall, L. polyribosomes under the base of dendritic spines in the granule B. and Axel, R. (1994) Topographic organization of sensory cells of the dentate gyrus. J. Neurosci. 2, 284±291. projections to the olfactory bulb. Cell 79, 981±991. Steward, O. and Reeves, T. M. (1988) Protein±synthetic machinery Wallace, C. S., Lyford, G. L., Worley, P. F. and Steward, O. beneath postsynaptic sites on CNS neurons: association between (1998) Dierential intracellular sorting of immediate early gene polyribosomes and other organelles at the synaptic site. J. mRNAs depends on signals in the mRNA sequence. J. Neurosci. 8, 176±184. Neurosci. 18, 26±35. Steward, O. and Ribak, C. E. (1986) Polyribosomes associated Wanner, I., Baader, S. L., Brich, M., Oberdick, J. and Schilling, with synaptic specializations on axon initial segments: localiz- K. (1997) Subcellular localization of speci®c mRNAs and their ation of protein±synthetic machinery at inhibitory synapses. J. protein products in Purkinje cells by combined ¯uorescence in Neurosci. 6, 3079±3085. situ hybridization and immunocytochemistry. Histochem Cell Steward, O. and Wallace, C. S. (1995) mRNA distribution within Biol. 108, 345±357. dendrites: relationship to aerent innervation. J. Neurobiol. 26, Watson, J. B., Sutclie, J. G. and Fischer, R. S. (1992) 447±458. Localization of the protein kinase C phosphorylation/calmodu- Steward, O., Kleiman, R. and Banker, G. (1995) Subcellular local- lin-binding substrate RC3 in dendritic spines of neostriatal neur- ization of mRNA in neurons. In: Localized RNAs, pp. 235±255. ons. Proc. natl Acad. Sci. U.S.A. 89, 8581±8585. Ed. H. D. Lipshitz. Springer-Verlag, Heidelberg. Wensley, C. H., Stone, D. M., Baker, H., Kauer, J. S., Margolis, St. Johnston, D. (1995) The intracellular localization of messenger F. L. and Chikaraishi, D. M. (1995) Olfactory marker protein RNAs. Cell 81, 161±170. mRNA is found in axons of olfactory receptor neurons. J. Thomas, K. L., Laroche, S., Errington, M. L., Bliss, T. V. P. and Neurosci. 15, 4827±4837. Hunt, S. P. (1994) Spacial and temporal changes in signal trans- Whittaker, K. L. and Lipshitz, H. D. (1995) Mechanisms and duction pathways during LTP. Neuron 13, 737±745. functions of RNA localization to the anterior pole of the Tiedge, H. and Brosius, J. (1996) Translational machinery in den- Drosophila oocyte and early embryo. In: Localized RNAs,pp. drites of hippocampal neurons in culture. J. Neurosci. 15, 7171± 41±61. Ed. H. D. Lipshitz. Springer-Verlag, Heidelberg. 7181. Yisraeli, J. K., Oberman, F., Schwartz, S. P., Havin, L. and Tiedge, H., Fremeau, R. T., Weinstock, P. H., Arancio, O. and Elisha, Z. (1995) Vg1 mRNA localization in Xenopus oocytes: a Brosius, J. (1991) Dendritic location of neural BC1 RNA. Proc. paradigm for interior decorating. In: Localized RNAs, pp. 157± natl Acad. Sci. U.S.A. 88, 2093±2097. 171. Ed. H. D. Lipshitz. Springer-Verlag, Heidelberg.