Domain necessary for Drosophila ELAV nuclear localization: function requires nuclear ELAV

DrosophilaELAV is one of a large number of RNA recognition motif (RRM)-containing proteins characterized by a structurally conserved 80-90 amino acid domain, the RRM, which has been shown to function in mediating the protein-RNA interactions of a number of family members (Burd and Dreyfuss, 1994). Many of these RRM-containing proteins function in posttranscriptional RNA processes such as splicing (Ruskin et al., 1988), stability (Bernstein et al., 1989), and transport (Michael et al., 1995). ELAV is the first identified member of a sub-group of RRM containing proteins, the ‘ELAV family’, which have been grouped together based on the number, arrangement, and extensive homology within their RRMs.

ELAV is expressed exclusively in neurons appearing at the onset of neuronal differentiation (Robinow and White, 1991), and is required for both viability and proper nervous system development (Campos et al., 1985). The ELAV family genes, however, now identified in a variety of vertebrates and invertebrates, express either exclusively within the nervous system, or both in the nervous system and in other tissues as well (Antic and Keene, 1997; Good, 1995).

The ELAV family proteins are characterized by an N-terminal auxiliary domain followed by three RRMs, the second and third of which are separated by a ~54-80 amino acid hinge region (Fig. 1b). Amino acid comparison shows that all family members share a significant amount of identity within their respective RRMs. In contrast, the amino terminal auxiliary domains share little similarity. Within the hinge region there is considerable divergence between the two Drosophilafamily members and between Drosophilaand vertebrates. The four vertebrate ELAV family genes differ in tissue distribution and developmental expression (Antic and Keene, 1997). The vertebrate family members share more homology, in the hinge, than the invertebrates do, however, it is less than the extensive homology shared within their RRMs. This is due, in part, to alternative splicing which is restricted primarily to the hinge (Abe et al., 1996; Liu et al., 1995; Okano and Darnell, 1997). The significance of these various hinge isoforms is not known, but potentially they may modulate the localization and/or function of some of these proteins. Despite the divergent character of the hinge there are two conserved motifs, which we term the pentapeptide and the octapeptide, which are present in all ELAV family members, except for DrosophilaRBP9 which does not have the pentapeptide (Kim and Baker, 1993), thereby providing the only real hinge homology across species (see below).

ELAV, which is predominantly nuclear and has a distinctive subnuclear distribution (Yannoni and White, 1997), is necessary for the generation of the neural specific isoform of the cell adhesion molecule Neuroglian (Koushika et al., 1996), and has been recently implicated in autoregulating its mRNA (Samson, 1998). In contrast, some vertebrate ELAV family members have been localized to the cytoplasm (Perron et al., 1997), both the cytoplasm and the nucleus (Antic and Keene, 1998; Atasoy et al., 1998; Gao and Keene, 1996; Wakamatsu and Weston, 1997), or found to shuttle between the nucleus and cytoplasm (Atasoy et al., 1998; Fan and Steitz, 1998b; Peng et al., 1998). Moreover, the vertebrate proteins have been shown to regulate polyadenylation (Wu et al., 1997), mRNA stability (Fan and Steitz, 1998b; Myer et al., 1997; Park et al., 1998; Peng et al., 1998) and mRNA translatability (Antic et al., 1999; Jain et al., 1997). It seems likely then, that the functional and localization properties of these proteins perhaps evolved differentially and that these properties might be encoded at the amino acid level. This hypothesis is suggested by the localization of the XenopusELAV family member, elrB, and ELAV when expressed in neural tube cells of the developing Xenopusembryo. Here, elrB localizes to the cytoplasm while ELAV localizes, as it does in Drosophilaneurons, to the nucleus (Perron et al., 1997). This observation emphasizes that amino acid sequence, as opposed to the cellular environment alone, may be responsible for mediating the differential distribution of these two proteins and, potentially, of other ELAV family members within a given cell type. The divergent hinge and N-terminal auxiliary domains are likely candidates for determinants important in mediating the localization and function of these proteins.

In this study we have expressed transgenes in flies that encode ELAV mutant proteins with deletions in the hinge and N-terminal domain, and analyzed the steady state localization of ELAV mutant proteins in third instar larval neurons, their natural cellular environment, using immunohistochemistry. We conclude that amino acids necessary for localizing ELAV to the nucleus reside in the hinge which separates the second and third RRMs and that an ELAV N-terminal domain deletion-mutant protein, ELAV^RBD60(Yao et al., 1993), which is partially impaired in ELAV function, has a mild localization phenotype. Additionally, we also analyzed in vivo function of the mutant ELAV proteins using genetic complementation. These functional assays reveal that (1) nuclear concentration of ELAV is important for ELAV function, (2) although the highly conserved octapeptide motif, RFSPMAGD, is necessary for localization, it is dispensable for function and, (3) that hinge sequence N-terminal to the octapeptide has both localization and functional components. This study provides insights into the role of ELAV hinge in localization and the relevance of nuclear localization to the protein’s in vivo function.

Hinge deletion-mutants were created using oligonucleotide mediated site directed mutagenesis. p2.6SEtub (Yao and White, 1991), which contains the elavgenomic fragment, SmaI-EcoRI, was used as a template to generate each deletion-mutant: p2.6del. Oligonucleotides used in the mutagenesis were: CGC CAT TGG TGA AAA TCG CGA GCC GGG CGT ATT GGA G to generate ΔNh, CAG CAT TAC GTC CAG CAT CGC GAG ACC CTT GTT AAC to generate ΔOh, GAT GAA AAT GGG ATA CGC GTC GCC CGC CAT GGT TG to generate ΔCh, and GAT GAA AAT GGG ATA CGC GTT GGA GAA TTT TAC CAC to generate ΔEh. XbaI sites were used to reverse the orientation of the deletions in p2.6del creating p2.6del-rev. The elavpromoter from p6.6BS (Yao and White, 1991) was cloned into p2.6del-rev, KpnI-XmaI, creating p6.6del. The 5′ end of the elavpromoter from p6.6BS (Yao and White, 1991) was cloned into CaSpeR1 (Pirrotta, 1988), PstI-XbaI, generating pCaSpP-X. The final injection constructs, pCaSpdel, were created by inserting the elavpromoter and coding region from p6.6del into pCaspP-X using XbaI. The hinge GFP fusion proteins were created by cloning GFP S65T from pRSET B (a gift from R. Y. Tsien, University of California at San Diego) into pBluescript II SK+ (Stratagene Ltd, Cambridge), EcoRI- KpnI, to form pSKGFP. Site directed mutagenesis was used to insert the ribosome-binding site and AUG initiation codon of the Drosophila Adhgene (Mottes and Iverson, 1995) 5′ to the elavhinge in p2.6SEtub using the oligonucleotide: CTG GAT GAT CTT ACT GGT CGA CAT GGT GAC TTC TTT TTT CAA TTC TGC GGC CGC GCT GCC GGG CGT ATT GGA G, thereby generating p2.6rbs. The ribosome- binding site, AUG initiation codon, and hinge were then PCR amplified from p2.6rbs using the oligonucleotides: CCC GGC AGC GCG GCC GCA G which contains a NotI site, and GCG CGG CCG CGA ATT CGC CGC CAG GCC CAC TGG C which contains both EcoRI and NotI sites, and cloned into pSKGFP, NotI-EcoRI, generating pSK-h- GFP. After Klenow filling the BamHI site in pSK-h-GFP the ribosome- binding site, AUG initiation codon, hinge, and GFP were cloned into a pUAST derivative (Brand and Perrimon, 1993), pUASTf, which has a Klenow filled EcoRI site, NotI-KpnI, generating pUAS-h-GFP.

The hingeRRM3-GFP fusion protein was generated by PCR amplifying the ribosome-binding site, AUG initiation codon of the Drosophila Adhgene, hinge and RRM3 from p2.6rbs using the oligonucleotides: CCC GGC AGC GCG GCC GCA G which contains a NotI site and CCG GAA TTC CTT GGC TTT GTT GGT CTT G which contains an EcoRI site. The amplified fragment was cloned into pUAS-h-GFP, EcoRI-NotI, creating pUAS-hRRM3-GFP.

To create the full length ELAV-GFP fusion protein the EcoRI sites in both p2.6SEtub and pBluescript II SK+, were cut and Klenow filled generating pSK. XbaI sites were used to reverse the orientation of this modified elavfragment into pSK fill creating p2.6-rev. Oligonucleotide mediated site directed mutagenesis using the oligonucleotide: CAA GAC CAA CAA AGC CAA GGA ATT CCG GGG AAG ATC TTA GAG CGG CCC AAA TGG encoding EcoRI and BglII sites directly 5′ to the elavstop codon, were used to create p2.6EB. GFP S65T from pRSET B was inserted into p2.6EB, EcoRI BglII, creating p2.6-GFP. The final injection construct: pCaSpELAV- GFP, was generated using p2.6-GFP, p6.6BS, and pCaSpPX as described for pCaSpdel.

Site directed mutagenesis was used to insert the nuclear localization sequence (NLS) from polyoma large T antigen (Richardson et al., 1986) using the oligonucleotides CAC TCA GTG GAC TTT ATT TGC GGC TAC GGC GTG AGC CGC AAG CGC CCC CGC CCC GGC GCA AAT ACC GGA GCT GGC for insertion at the amino terminal end of elavin p2.6E-h-rev creating p2.6NLS-nE-h or p2.6NLS-cE-h. p2.6NLS-nE-h and p2.6NLS-cE-h and were used to assemble the injection constructs, pCaSpNLS-nE-h and pCaSpNLS- cE-h, as described for pCaSpdel. The NLS and flanking sequence was excised from pCaSpNLS-nE-h and inserted into the remaining deletion-mutants, pCaSpdel, with SapI generating pCaSpNLS-ndel.

Drosophila melanogasterwere raised on standard corn meal-molasses- agar-yeast medium. Germ line transformants were created by injecting CasPeR constructs into embryos (Rubin and Spradling, 1982). F1 progeny embryo of the cross Df(1)wfemales X yw; KiΔ2-3males were used for injections; Δ2-3provides the transposase activity (Robertson et al., 1988). Transformants were selected for red eye color and were mapped and balanced using standard genetic techniques.

The localization of the ELAV deletion-mutants was made possible by an elavtransgene P{elav⁻¹³}2 that encodes an ELAV protein which is fully functional and is deleted for a 13 amino acid epitope recognized by the monoclonal antibody (mAb) 5D3C5 (M. Lisbin and K. White, unpublished data). The localization of these proteins was carried out as described by Yannoni and White (1997). The crosses were elav^e5/elav^e5; elav⁻¹³X w/Y; Tg/Tgor Balancerwhere Tgdenotes the transgene. Male larvae (elav^e5/Y; Tg/+) were selected and subjected to immunocytochemical analysis.

For pan-neural expression, flies with GAL4 drivers, elav-GAL4or a GAL4 enhancer trap C155were crossed to flies carrying transgenes expressing fusion proteins under the UAS transcriptional control. The list and source of transgenes used in this study is provided in Table 1.

To analyze the function provided by the hinge deletion-mutants in the elav^e5, elav^l1and elav^ts1genetic backgrounds, females of the genotype elav^xwsn/Balancerwere crossed to males of the genotype w/Y;Tg/Balancer; all flies were raised at 25^oC. Rescued progeny are able to eclose as adults carrying the transgene and the mutant elavallele. Percentage rescue was calculated as # of rescued flies {(elav^xwsn/Y;Tg/+) / # of control flies (elav^xwsn/w;Tg/+)}×100 for e5and l1alleles. In case of ts1, rescue was determined by subtracting the number of escaper males (carrying elav^ts1as their only copy of elav) from the number ts1males with the transgene and dividing this number by the number of sibling females carrying the transgene {# (elav^ts1wsn/Y;Tg/+)−#(elav^ts1wsn/Y;Balancer/+) /# (elav^ts1wsn/w;Tg/+)}×100.

Procedures used for collection of wandering third instar larvae, preparation of whole mounts and squashes, and immunostaining are described in (Yannoni and White, 1997). Visualization of nuclei in whole mount tissue was achieved using RNAse A treatment, 100 µg/ml, for 30 minutes at 25°C and followed by staining with propidium iodide.

The following antibodies were used: a mouse anti-ELAV mAb 5D3C5, at a dilution of 1:5 or 1:10 (hybridoma line courtesy of Dr G. Rubin) and a mouse anti-HuD mAb 16A11, at a dilution of 1:1000 (gift of Dr H. Furneaux). Secondary antibodies conjugated to FITC (Jackson ImmunoResearch Laboratories, West Grove, PA) were used at a dilution of 1:50.

For fluorescence and phase contrast microscopy a Zeiss Axiophot microscope equipped with a ×100 objective numerical aperture: 1.25 was used to obtain phase contrast images. Fluorescence images were obtained using a reflector slider 3 Fl to select for excitation at 488 nm. Photographs were taken using Kodak TMAX 400 film. Confocal images were collected on a Bio-Rad MRC-600 confocal microscope.

Heads from 2-4 day old flies of the genotype w/Y; Tg/Balancerwere homogenized in protein extraction buffer as described by Samson et al. (1995). 10 µg per lane were loaded on 10% gels. ELAV proteins were recognized with a rat anti-ELAV polyclonal serum (1:750) and GFP proteins were detected with a rabbit anti-GFP polyclonal antibody (Clonetech, Palo Alto, CA) (1:1000). For loading controls β-tubulin immunoreactivity was assessed with chemiluminescence using a tubulin- specific mouse primary antibody (Sigma, St Louis, MO). As secondary antibody, Peroxidase-conjugated anti-mouse antibody was used as (Amersham Life Sciences Inc., Arlington Heights, IL) at 1:1000.

ELAV and the human ELAV family protein HuD (Szabo et al., 1991), share a significant amount of identity, 63-74% in their RRM domains (Fig. 1b), however, they localize differently. ELAV is predominantly nuclear and is enriched in discrete subnuclear domains in Drosophilaneurons (Yannoni and White, 1997). NIH3T3 cells transfected with a FLAG tagged HuD construct show HuD localized predominantly within the cytoplasm, using mAb 16A11, although some nuclear staining is reported as well (Wakamatsu and Weston, 1997). Similarly, FLAG-HuD localized with the FLAG specific antibody also shows predominantly cytoplasmic signal (Wakamatsu and Weston, 1997). To determine if the difference between ELAV and HuD localization is due to cellular environment, rather than inherent to the protein, we expressed HuD in Drosophilaneurons. We used the GAL4/UAS targeted gene expression system (Brand and Perrimon, 1993) to drive the expression of a UAS-HuDtransgene pan- neurally using two different neural specific GAL4 drivers, elav-GAL4 (Luo et al., 1994) and C155 (Lin and Goodman, 1994). Central nervous system (CNS) neurons of larvae expressing HuD were stained using mAb 16A11 (Marusich et al., 1994) which does not cross react with endogenous ELAV. Results obtained from three independent UAS-HuDtransgenes showed that in Drosophilaneurons, like in vertebrate cells (Wakamatsu and Weston, 1997) HuD is predominantly cytoplasmic (Fig. 1a). ELAV, when expressed under a pan-neural GAL4 driver, localizes similar to endogenous ELAV (see below).

Given the significant amount of identity in the ELAV and HuD RRM domains, we reasoned that the determinants responsible for their differential localization reside in the hinge, which has only 16% identity (Fig. 1b). Although the N-terminal domains of these proteins are distinct, we consider the possibility unlikely that the ELAV N-terminal domain would have a critical role in mediating nuclear localization, because the ELAV N-terminal domain deletion-mutant protein, RBD60 (Yao et al., 1993), is localized predominantly to the nucleus as visualized in whole mounts of the third instar larval CNS (Yannoni and White, 1997). In this study we used higher resolution CNS squash preparations to analyze RBD60 localization. We observed a mild RBD60 localization phenotype as ELAV-immunoreactivity was also observed in the cytoplasm, but still the nuclear staining was quite extensive (Fig. 1a).

To determine whether the hinge (amino acids 333-400), is necessary for ELAV nuclear localization, four hinge deletion mutants were constructed and introduced into the Drosophilagermline (Fig. 2b). The template used in generating these deletions consisted of a 9.2 kb genomic fragment containing the 6.6 kb elavpromoter and a 2.6 kb fragment containing the ELAV coding region and 3′ UTR, which completely suppresses the lethality associated with the elavnull allele, elav^e5(Yao and White, 1991). We deleted the entire hinge region (ΔEh). Additionally, the conserved octapeptide motif (ΔOh) and the hinge sequences N-terminal (ΔNh) and C-terminal (ΔCh) to the octapeptide were deleted. Subdividing the hinge allowed us to determine the contribution of the octapeptide motif itself, as well as the flanking hinge amino acids to ELAV localization. The coding sequence and size of each deletion-mutant protein was confirmed using both sequence and immunoblot analysis (Fig. 3a,b).

To circumvent the possibility that the hinge deletion-mutant proteins might not provide vital ELAV function, a second elavtransgene, elav⁻¹³, was used. The elav⁻¹³transgene encodes a fully functional ELAV protein that is missing the 13 amino acids recognized by the ELAV mAb 5D3C5. Combining elav⁻¹³and a hinge deletion-mutant transgene in the elav^e5null genetic background allowed us to generate viable organisms in which to localize the deletion-mutant proteins, regardless of whether these mutant proteins provided function. In these animals only the deletion-mutant protein is recognized by mAb 5D3C5.

Analysis of ELAV hinge deletion-mutant protein localization in neurons shows that hinge sequence is required for nuclear localization of ELAV. Amino acids necessary for nuclear translocation are located between residues 333 and 374, and include the conserved octapeptide motif. Hinge- mutants missing amino acids 333-400 (ΔEh), 333-366 directly N-terminal to the octapeptide motif (ΔNh), or the octapeptide amino acids 367-374 (ΔOh), localize to the cytoplasm (Fig. 2a). In contrast, amino acids 375-400 were not required for nuclear localization since ΔCh localizes to the nucleus and has a wild-type subnuclear distribution (Fig. 2a). The wild-type distribution of ΔCh illustrates that reducing the spacing between RRMs 2 and 3 by 26 amino acids does not per se disrupt the localization of ELAV, even at the subnuclear level. Since amino acids 375-400 are not necessary for nuclear localization, the cytoplasmic localization of ΔEh is likely to result from deletion of amino acids 333-374. Also, the cytoplasmic localization of ΔNh and ΔOh is likely to result from a requirement for determinants between amino acids 333-374 in nuclear localization rather than from the reduced spacing between RRMs 2 and 3. For each mutant with cytoplasmic localization, in some neurons a subnuclear ELAV-dot, a characteristic feature of ELAV′s subnuclear distribution was discerned (Fig. 2a; Yannoni and White, 1997). These ELAV-dots indicate that a residual amount of protein is able to enter the nucleus for each of these mutants.

Many nuclear proteins have ‘classical’ nuclear localization sequences (NLSs) which can be classified as either the SV40 (Chelsky et al., 1989; Kalderon et al., 1984) or bipartite type (Gorlich and Mattaj, 1996; Nigg, 1997). These sequences, as well as shuttling sequences which have both nuclear import and export activity, function to mediate the transport of NLS- or shuttling sequence-containing proteins into the nucleus in a well characterized process (Gorlich and Mattaj, 1996; Myer et al., 1997). They contain all the information, both necessary and sufficient, for providing nuclear import. We examined whether the ELAV hinge can provide all the determinants required for nuclear localization by testing its ability to target the green fluorescent protein (GFP) to the nucleus. GFP is a well characterized reporter used in both nuclear localization and transport assays (Afonina et al., 1998; Chen et al., 1998; Prieve et al., 1996; Stade et al., 1997). Initially, we looked at the localization of GFP alone using a UAS-GFPtransgene expressed in neurons with pan-neural GAL4 drivers. Fig. 4ashows that GFP localizes to both the nucleus and cytoplasm. Next, we tested if full length ELAV could localize GFP to the nucleus by expressing an ELAV-GFP fusion protein using the same parent construct used to generate the hinge deletion- mutants. Animals transgenic for UAS-ELAV-GFPproduce ELAV-GFP fusion protein which, like wild-type ELAV, localizes to the nucleus (Fig. 4a). Thus, ELAV has all the determinants required to translocate GFP into the nucleus. In addition we show that GFP is localized efficiently to the nucleus in Drosophilaneurons by the SV40 NLS, as expected from a previous report (Davis et al., 1995).

To test the ability of the hinge to mediate the nuclear localization of GFP, a hinge-GFP (h-GFP) fusion was constructed and introduced downstream of the UAS sequence. When the UAS-h-GFPtransgene is driven by pan-neural GAL4 drivers the h-GFP fusion protein localizes predominantly to the cytoplasm (Fig. 4a). As controls, neurons from flies expressing UAS-elavunder the pan-neural GAL4 driver were analyzed for localization of ELAV. These animals express ELAV protein which, like endogenous ELAV, localizes to the nucleus (Fig. 4b). The GAL4/UAS system then, is not responsible for the predominantly cytoplasmic distribution of the h-GFP fusion protein. Based on these data, we conclude that the ELAV hinge amino acids 333-400, do not contain all the ELAV residues necessary to localize GFP to the nucleus.

We investigated whether RNA binding, provided by the addition of an RRM, would facilitate the nuclear entry of GFP. We generated a construct encoding the hingeRRM3 region in frame with GFP, downstream of the UAS sequence (UAS- hRRM3-GFP). Larval CNS neurons, from transgenic animals in which the UAS-hRRM3-GFPtransgene was driven by the pan-neural GAL4 drivers, expressed the hRRM3-GFP fusion protein which, like the h-GFP fusion, localized predominantly to the cytoplasm (Fig. 4a). Therefore, neither the hinge alone, nor in combination with RRM3, provides the determinants required to translocate GFP into the nucleus.

The localization data obtained using the GFP fusion proteins suggested that sequence N-terminal to the hinge might be important in localizing ELAV. This possibility is further suggested by the localization of RBD60, which shows some cytoplasmic signal in addition to extensive nuclear signal (Fig. 1a). However, we have noted that the expression level of the h-GFP and hRRM3-GFP proteins, by GFP-fluorescence, is low in comparison with ELAV-GFP, possibly indicating partial degradation or modification of the fusion proteins. The fusion proteins, therefore, may be disabled in terms of their ability to mediate nuclear entry of GFP.

We determined whether, in addition to providing determinants required for localization, the hinge had a functional component as well. First, western blots were used to assess the level of steady state expression for each deletion-mutant protein, in adult head homogenates, using ELAV polyclonal antiserum. Homogenates were isolated from independent transgenic lines which express one copy of a deletion-mutant and one copy of the elav^RBD60allele which provides ELAV function. Western blot analysis shows that ΔEh and ΔNh, which are localized to the cytoplasm, consistently display less protein than ΔCh which is localized to the nucleus (Fig. 3a). ΔOh, which is also localized to the cytoplasm, however, shows protein equivalent in level to the nuclear protein ΔCh (Fig. 3b). ΔEh and ΔNh lines with highest amounts of protein were chosen for functional analysis. The hinge mutant-protein levels in the chosen lines is at least comparable to RBD60 (Fig. 3a).

To assess function, we assayed the viability of flies that carry one copy of a transgene encoding a hinge deletion-mutant (see Materials and Methods) using the elav^e5background. Since ELAV is essential for embryonic viability, rescue to adulthood demonstrates that the mutant protein alone is able to provide the viability function associated with elavlocus. For comparison we also determined the amount of rescue provided by full length ELAV, when produced from the transgene elav^DmORF(Yao et al., 1993) at 74% (Table 2). Of the deletion-mutant proteins which localize to the cytoplasm, only ΔOh provided some rescue. Four independent transgenic lines expressing ΔOh were tested. Two lines provided some rescue: 12% and 3.2% (Table 2); these corresponded to the high expressers by the western signal (Fig. 3b, lanes 3 and 4). ΔNh and ΔEh, which also localize predominantly to the cytoplasm, did not provide ELAV vital function. In contrast, ΔCh, which localizes to the nucleus, provided a substantial amount of rescue: 64% or 55%, which approaches the rescue provided by the control DmORF. The ability of ΔCh to provide ELAV vital function, demonstrates that similar to localization, the deleted 25 C-terminal hinge amino acids and the reduced spacing between RRMs 2 and 3 does not affect the ability of ELAV to function. Taken together, the functional properties of the four hinge deletion-mutants suggest that nuclear localization is important to the function provided by ELAV.

Rescue of the elavnull, elav^e5, is a stringent test that requires the hinge deletion-mutants to provide ELAV vital function. To determine whether these mutants were able to provide function in a less restrictive environment, we assayed their ability to provide functional complementation when combined with each of two mutant ELAV proteins which are encoded by the alleles elav¹and elav^ts1.

elav¹mutants, similar to elav^e5mutants, die during embryogenesis; however, unlike elav^e5, the elav¹allele produces a full length ELAV protein. As with elav^e5, we observed functional complementation with ΔOh and ΔCh, but not with ΔNh and ΔEh. In this environment, all four of the ΔOh lines tested provided some rescue, between 12% to 0.2%. As expected, the rescue provided by ΔCh is comparable with that provided by DmORF (Table 2).

elav^ts1is a temperature sensitive hypomorphic allele in which the tryptophan (TGG) at amino acid 419 in RRM3 is changed to the stop codon, TAG. This allele

encodes a truncated protein of molecular mass ~45 kDa and an additional full length protein of 50 kDa which is thought to be generated through a translational read-through mechanism. The phenotype associated with elav^ts1likely arises from an insufficient amount of the 50-kDa protein at a critical stage in development or, a change in protein structure brought about by an amino acid substitution at the site of suppression (Samson et al., 1995). elav^ts1associated lethality is mild at 18°C; at 25°C, however, the lethality is severe as hemizygous males rarely survive (Campos et al., 1985). At the restrictive temperature of 25°C, both ΔOh and ΔCh provided significantly more rescue in combination with elav^ts1than with either elav^e5or elav¹. The levels of rescue provided by ΔCh are comparable to the rescue provided by DmORF. ΔNh, which showed no functional complementation in combination with either elav^e5or elav¹, provided a modest 5.3% and 18% rescue in each of the two lines tested, but ΔEh was not able to provide any rescue (Table 2).

Since wild-type ELAV is predominantly nuclear we asked whether these deletion-mutant proteins could provide increased function if translocated into the nucleus using a heterologous NLS. An NLS peptide derived from polyomavirus large-T antigen (Richardson et al., 1986), which has been used to direct nuclear targeting of photoactivatible reagents in Drosophila(O’Farrell and Girdham, 1994), was inserted at the N-terminal end of ΔEh, ΔOh, ΔNh, and ΔCh creating ΔEh[NLS], ΔOh[NLS], ΔNh[NLS], and ΔCh[NLS] (Fig. 5b).

The distribution of the NLS tagged hinge deletion-mutant proteins was analyzed in larval CNS neurons. Each NLS- tagged hinge deletion-mutant that had previously localized to the cytoplasm, showed a significant amount of nuclear staining (Fig. 5a). Having ascertained that these NLS hinge mutants do localize to the nucleus, we assayed the functional properties of each ELAV hinge deletion-mutant protein after nuclear translocation. Immunoblot analysis was used to confirm the production of the NLS containing deletion-mutant proteins. Lines producing the most similar amounts of protein were selected for this analysis (Fig. 3c) and assayed for genetic complementation (Table 2). The full rescue provided by ΔCh[NLS] shows that the NLS does not have an adverse affect on the function of ELAV. For all other deletion-mutants carrying the NLS, an increase in functional complementation is observed compared to the deletion-mutants alone (Table 2). Virtually full function was provided by ΔOh[NLS] when combined with all three elavalleles (Table 2). ΔEh[NLS] and ΔNh[NLS] provide a limited amount of rescue in the most restrictive genetic backgrounds, elav^e5and elav¹(Table 2). The increased ability of these two mutants to provide function was further revealed in combination with elav^ts1. ΔNh[NLS] provided 23% and 40% rescue in the two lines tested and ΔEh[NLS] provided a comparable 25% and 22% rescue in the two lines tested (Table 2).

Our analysis demonstrates that ELAV (50 kDa) requires specific sequences for nuclear localization, although it is within the predicted molecular mass range of 40-60 kDa which would allow for diffusion into the nucleus (Silver, 1991). Hinge residues 333-374, N-terminal to and including the octapeptide motif, are critical for ELAV nuclear localization, but the 25 amino acids C-terminal to the octapeptide are not. Therefore, reducing hinge length by 25 amino acids has no effect on the localization of ELAV. Hinge sequence necessary for ELAV localization contains 5 basic amino acids, the hallmark of the classical SV40 type or bipartite type NLSs (Gorlich and Mattaj, 1996; Fig. 6b). Aside from these basic amino acids this region has no significant homology to known classical NLSs, however, there is limited similarity to the hnRNP A1 38 amino acid nucleocytoplasmic shuttling sequence, M9 (Michael et al., 1995; Fig. 6a). Recently, hinge sequence within the vertebrate ELAV family member HuR, designated ‘HNS’ for HuR nucleocytoplasmic shuttling sequence, has been shown to be both necessary and sufficient for the nuclear localization and shuttling activity of HuR. Amino acid comparison shows that the HuR hinge also has limited similarity with the hnRNP A1 sequence, M9 (Fan and Steitz, 1998a). A comparison of the hinge region from ELAV and the four vertebrate ELAV family members, including their splice variants, is shown in Fig. 6b. This amino acid comparison high lights both the pentapeptide, RRXYG, and the octapeptide, RFSPXYZD, found in all ELAV family members except for DrosophilaRBP9, which does not have a conserved pentapeptide sequence. Note that both the HuR HNS and the ELAV hinge sequence delineated in this study include these two conserved motifs. The shuttling activity provided by the HNS has been linked to the mRNA stabilizing function of HuR (Fan and Steitz, 1998a) Although shuttling activity has not yet been ascertained for ELAV our data predicts that, unlike the HNS, a putative ELAV shuttling sequence would not include the octapeptide, since full ELAV function is provided without this motif.

The GFP reporter studies presented here suggest that the hinge, either alone or in combination with RRM3, does not facilitate the nuclear localization of GFP, while the entire ELAV protein does. Therefore, the ELAV hinge although necessary may not be sufficient for nuclear entry. This conclusion is in part, based on the assumption that the hinge- GFP and hRRM3-GFP fusion proteins remain intact in Drosophilaneurons. Immunoblot analysis showed the presence of both GFP alone and ELAV-GFP using antibodies against GFP and ELAV, respectively. However, none of the other GFP fusion proteins were detected using a GFP antibody on immunoblots (data not shown). Therefore, although cytoplasmic fluorescence shows that GFP in some form is made, it is possible that the h-GFP and hRRM3-GFP fusion proteins are degraded or modified, in vivo, such that their ability to provide determinants for localization might be impaired.

Specific protein-protein interactions have been shown to mediate the localization of a number of nuclear and shuttling proteins. A requirement for the two conserved hinge motifs, in the localization of both ELAV and HuR then, may be indicative of a general role for these motifs in mediating the localization of ELAV family members through interactions with proteins found in both vertebrates and invertebrates. Interestingly, alternative splicing, which occurs in three of the four vertebrate genes HuB, Huc, and HuD, is seen in between these two regions (Fig. 6b) (Okano and Darnell, 1997). The splice variants, generated from these three genes, encode protein isoforms in which a conserved thirteen amino acid sequence, RLDNLLNMAYGVK, is differentially regulated. This 13 amino acid sequence, as well as other residues located between the pentapeptide and the octapeptide, may regulate interactions with tissue or species specific proteins to modulate the localization or function of these proteins (Fan and Steitz, 1998a; Keene, 1999).

Viability assays under stringent conditions in which hinge deletion mutant proteins were the sole source of ELAV function revealed that those mutant proteins that localize primarily in the cytoplasm were impaired in providing ELAV vital function but, nuclear ΔCh, provided full viability. Interestingly, ΔOh provided a low amount of survival (3.2 to 12%), which was dependent on the transgene line tested. Since in these lines, a small amount of ΔOh is still nuclear as evidenced by the subnuclear ELAV-dot, we suspect that nuclear ΔOh is responsible for the residual function. This partial rescue provided the initial evidence that, although the octapeptide was essential for localization, it is not required for function.

In the less stringent functional assay, where the mutant protein had to provide function supplemental to elav^ts1, ΔEh was still a functional null, but now supplemental functional activity was revealed for ΔNh (5.3- 18%) and ΔOh (30-50%). We propose that as with ΔOh in combination with both elav^e5and elav¹, that the rescue provided by both ΔNh and ΔOh, in combination with elav^ts1results from the small amount of nuclear deletion-mutant protein which is discernible in some cells as the nuclear ELAV-dot. Since nuclear ΔCh was fully functional in all three complementation tests, reducing hinge length by 25 amino acids does not affect the function of ELAV.

One hypothesis which would explain the results of the viability studies is that the hinge deletion-mutant proteins are still capable of at least some biological function, but are unable to provide this function due to their mislocalization. This hypothesis was tested by translocating each hinge deletion- mutant into the nucleus using a heterologous NLS. Since the activity of the NLS is restricted to transport, any increase in function provided by the NLS-tagged mutants can be attributed to the ELAV mutant protein itself rather than the NLS.

ΔOh[NLS] provided complete rescue in all three functional

complementation tests showing that the activity provided by the octapeptide is restricted to transport and not to ELAV function. Neither ΔEh[NLS] nor ΔNh[NLS], showed rescue with the two most restrictive elavalleles. Instead, the elav^ts1environment was necessary in order to reveal the function for both these mutants even after translocation into the nucleus.

ΔEh[NLS] provided a level of complementation once in the nucleus, 22% and 25%, comparable in level to that provided by ΔNh[NLS] at, 23% and 29%. Since ΔNh[NLS] is still only partially functional, even after nuclear translocation, N- terminal hinge amino acids, in addition to providing residues required for localization, provide determinants required for ELAV function as well. The impaired function of ΔNh[NLS] might result from reducing the spacing between RRMs 2 and 3 by 35 amino acids. However, this seems unlikely since the 25 amino acid deletion in ΔCh has no effect on the function of ELAV. The functional component of ΔNh[NLS] and ΔEh[NLS] is similar as judged by levels of rescue provided by these mutants in the elav^ts1background. Thus, we suggest that it is likely that the amino acids encompassed in ΔNh[NLS] are primarily responsible for the functional impairment and the juxtaposition of RRMs 2 and 3 contribute only slightly.

Our data supports an emerging picture of the role of hinge sequence in modulating the localization of the ELAV family members, a property that is likely to influence the diverse functions of these proteins.

We thank D. Bordne, J.-R. Neptune, and P. Parmenter for technical assistance as well as E. Dougherty for assistance with microscopy. We are grateful to C. Yannoni for generating the viability rescue data and to S. Koushika for assembling the HuD injection construct. We thank Dr H. Furneaux for the HuD antibody, Dr R. Tsien for GFP plasmids and Dr P. O’Farrell and Bloomington DrosophilaStock Center for fly strains. We thank Dr R.Sen and Dr L. Davis for helpful discussions. This work was supported by NIH grant NS36179 and a NIH Shared Instrumentation grant RR05615.