Poly(A) Tail Length-dependent Stabilization of GAP-43 mRNA by the RNA-binding Protein HuD*

The neuronal ELAV-like RNA-binding protein HuD binds to a regulatory element in the 3′-untranslated region of the growth-associated protein-43 (GAP-43) mRNA. Here we report that overexpression of HuD protein in PC12 cells stabilizes the GAP-43 mRNA by delaying the onset of mRNA degradation and that this process depends on the size of the poly(A) tail. Using a polysome-basedin vitro mRNA decay assay, we found that addition of recombinant HuD protein to the system increased the half-life of full-length, capped, and polyadenylated GAP-43 mRNA and that this effect was caused in part by a decrease in the rate of deadenylation of the mRNA. This stabilization was specific for GAP-43 mRNA containing the HuD binding element in the 3′-untranslated region and a poly(A) tail of at least 150 A nucleotides. In correlation with the effect of HuD on GAP-43 mRNA stability, we found that HuD binds GAP-43 mRNAs with long tails (A150) with 10-fold higher affinity than to those with short tails (A30). We conclude that HuD stabilizes the GAP-43 mRNA through a mechanism that is dependent on the length of the poly(A) tail and involves changes in its affinity for the mRNA.

The neuronal ELAV-like RNA-binding protein HuD binds to a regulatory element in the 3-untranslated region of the growth-associated protein-43 (GAP-43) mRNA. Here we report that overexpression of HuD protein in PC12 cells stabilizes the GAP-43 mRNA by delaying the onset of mRNA degradation and that this process depends on the size of the poly(A) tail. Using a polysomebased in vitro mRNA decay assay, we found that addition of recombinant HuD protein to the system increased the half-life of full-length, capped, and polyadenylated GAP-43 mRNA and that this effect was caused in part by a decrease in the rate of deadenylation of the mRNA. This stabilization was specific for GAP-43 mRNA containing the HuD binding element in the 3untranslated region and a poly(A) tail of at least 150 A nucleotides. In correlation with the effect of HuD on GAP-43 mRNA stability, we found that HuD binds GAP-43 mRNAs with long tails (A150) with 10-fold higher affinity than to those with short tails (A30). We conclude that HuD stabilizes the GAP-43 mRNA through a mechanism that is dependent on the length of the poly(A) tail and involves changes in its affinity for the mRNA.
The growth-associated protein GAP-43 is a neuronal-specific phosphoprotein that is expressed primarily in neurons during both the initial axonal outgrowth and remodeling of synaptic connections for reviews (1)(2)(3). The expression of GAP-43 is complex and features both transcriptional and post-transcriptional components (4 -11). Transcriptional factors of the basic helix-loop-helix family are known to control the neural specific expression of the GAP-43 gene (12,13). Yet, and most surprisingly, in several instances, the levels of gene transcription do not correlate well with the accumulation of the mature GAP-43 mRNA (8,9,11,14,15). In PC12 cells induced to differentiate by nerve growth factor, GAP-43 mRNA levels are regulated primarily through selective changes in the rate of degradation of the mRNA (9). This process depends on the activation of protein kinase C and is mediated by the interaction of highly conserved sequences in the 3Ј-untranslated region (3Ј-UTR) 1 of the mRNA with neuronal-specific RNA-binding proteins (16). One of these proteins was identified as the ELAV-like protein HuD (10,17). Blocking endogenous HuD expression using antisense RNA was found to decrease the levels of the GAP-43 mRNA and protein in PC12 cells and to limit process outgrowth (18). In contrast, overexpression of HuD in PC12 cells causes an increase in GAP-43 protein levels and cellular differentiation, even in the absence of nerve growth factor (19). In light of these observations, we proposed that HuD is the main regulatory protein involved in the post-transcriptional regulation of the GAP-43 gene.
HuD belongs to the highly conserved elav (embryonic lethal abnormal vision) family of RNA-binding proteins (20,21). HuD is one of four mammalian ELAV-like proteins that have been identified, three of which are exclusively expressed in the nervous system. These proteins contain three RNA recognition motifs (RRMs); RRMs I and II bind to AU-rich motifs (22), whereas RRM III binds to long stretches of poly(A) (23,24). HuD has been shown to bind to AU-rich elements (AREs) found in the 3Ј-UTRs of several mRNAs including c-Fos (25), tau (26), p21 waf1 (27), and GAP-43 (17).
In this study, we examined the mechanism by which HuD stabilizes the GAP-43 mRNA using a combination of in vivo and cell-free mRNA decay assays. We found that HuD increases the stability of this mRNA in both systems. This process depends on the presence of the HuD binding element in the GAP-43 3Ј-UTR and on the size of the poly(A) tail. GAP-43 mRNAs with short tails (A30) were found not only to be less stable than their A150 counterparts, they were also unable to be stabilized by HuD. Consistent with these findings, we observed that the affinity of HuD for the GAP-43 mRNA increased ϳ10-fold when the RNA contained long versus short tails. Altogether, our results suggest that HuD controls GAP-43 mRNA stability by binding to its recognition element in the 3Ј-UTR and decreasing the initial rate of deadenylation of mRNAs with long poly(A) tails.

EXPERIMENTAL PROCEDURES
Construction of DNA Clones-The construction of the inducible expression vectors used for in the in vivo mRNA decay assays has been described previously (10,16). Briefly, the rat GAP-43 cDNA was cloned into the pMEP4 vector (Invitrogen), containing the human metallothionein-IIA promoter. In the presence of cadmium, this vector increases the expression of the transgene by ϳ8 -10-fold (10). To co-express HuD in PC12 cells, we used a pcDNA3-derived construct containing the human HuD cDNA (a gift from H. Furneaux) downstream of the CMV promoter (pcHuD). As shown by Mobarak et al. (18), this construct directs the constitutive expression of HuD in PC12 cells increasing the levels of HuD protein in the cells 2-3-fold. In addition to the generation of mammalian expression vectors, additional constructs were generated for the synthesis of radiolabeled RNAs by in vitro transcription. A full-length rat GAP-43 cDNA was generated by RT-PCR using specific 5Ј-and 3Ј-UTR flanking primers (5Ј-GGAATAAGGATCCGAGGAG-GAAAGGAG-3Ј and 5Ј-GACGTCGACGCTAATTGGCACATTTGC-3Ј, respectively). This fragment was cloned into the BamHI site of the SP64-poly(A) (Promega, Madison, WI) to yield pSP64-GAP. When this plasmid is linearized with SmaI, the resulting transcript contains no poly(A) tail, but when it is digested with EcoRI, the RNA contains a tail of exactly 30 As. To delete the HuD binding site from the GAP-43 3Ј-UTR, we used a trans-PCR protocol described by Neve and Neve (28). Briefly, PCR reactions used the full-length GAP-43 cDNA as template and the following primers, 5Ј-ACACACTTGGAACTCCCACAGGGCCA-CACGCACCAG-3Ј and 5Ј-GCCCTGTGGGAGTTCCAAGTGTGTGTGT-GCAATGTT-3Ј to loop out the internal 27-nucleotide sequence. A second round of PCR used the flanking 5Ј-and 3Ј-UTR primers described above to yield the full-length product containing a single 3Ј-UTR deletion. This PCR product was then cloned into pCR-Script (Stratagene, La Jolla, CA), and the resulting plasmid was called p⌬HuD. To verify the specificity of the in vitro decay assay, the entire GAP-43 3Ј-UTR was replaced by that of ␤-globin and cloned into pMEP4 as described by Tsai et al. (10). This plasmid (called C2) was digested with HindIII and XhoI and the resulting insert, BG3ЈUTR, was cloned into pBlueScript II (Stratagene).
In Vivo mRNA Decay Assays-In vivo mRNA decay studies were performed in pMEP-4-GAP-43-transfected PC12 cells essentially as described by Tsai et al. (10). Briefly, cells were pre-incubated with 5 M CdCl 2 for 16 h to induce GAP-43 transcription from the human metallothionein IIA promoter. To start the decay analysis, the medium was washed out and replaced with fresh medium without cadmium (time 0). Cultures were incubated for various time periods, cells were then harvested, and RNA isolated using Tri Reagent (Sigma). For H-mapping studies, the RNA was further purified using the RNeasy minikit (Qiagen, Valencia, CA).
Northern Blot Analysis-Fifteen g of total RNA extracted from cells was run on 1.1% formaldehyde-agarose gel as previously described (9). Membranes were probed for GAP-43 mRNA or glyceraldehyde-3-phosphate dehydrogenase mRNA using 32 P-radiolabeled cDNA probes generated by random priming (Prime-a-Gene, Promega). Membranes were scanned on a PhosphorImager (Storm 860, Invitrogen) and analyzed using the ImageQuant software package. The density of the GAP-43 bands was normalized to that of glyceraldehyde-3-phosphate dehydrogenase mRNA, and results were expressed as percentage of the mRNA remaining relative to time 0 as previously described (9).
H-mapping-H-mapping was performed essentially as described by Brewer and Ross (29) with a few modifications. For in vivo decay assays, 30 g of total RNA extracted from transfected cells were dissolved in 1 mM EDTA and annealed to an antisense oligonucleotide (5Ј-CTTA-AAGTTCAGGCATGTTCTTGGT-3Ј) that is complementary to a sequence overlapping the beginning of the 3Ј-UTR. After digestion with RNase H (United States Biochemical Corp.) (0.75 units/reaction), the resulting fragments were run on a 1.35% NuSieve ® 3:1 agarose gel (FMC, Rockland, ME) and probed as described above. For in vitro decay assays, radiolabeled RNAs were purified using RNeasy minicolumns. Purified RNAs were annealed to the antisense oligonucleotide and treated with RNase H as described above. Fragments were analyzed using 10% polyacrylamide-urea gels.
In Vitro Transcription of Capped and Polyadenylated RNAs-Capped and polyadenylated RNAs were synthesized by a combination of in vitro transcription and subsequent polyadenylation. To radiolabel the body of the RNA, RNAs were transcribed in vitro in the presence of 1 mM cap analog 7m G(5Ј)ppp(5Ј)G (Invitrogen) and of [␣-32 P]UTP (PerkinElmer Life Sciences, 80 Ci/mmol). Wild type full-length GAP-43 RNA was synthesized in vitro using SP6 RNA polymerase (New England Biolabs, Beverly, MA) and pSP64 GAP. Constructs were linearized either with SmaI to prepare RNAs without tails or EcoRI to prepare RNAs with tails of 30 As. GAP-43 RNAs without the HuD binding site were synthesized using EcoRI-digested p⌬HuD and T3 RNA polymerase. To prepare chimeric GAP-43 RNAs with the ␤-globin 3Ј-UTR, BG3ЈUTR was linearized with XhoI and plasmids transcribed with T3 RNA polymerase. For studies measuring the decay of the poly(A) tail, unlabeled RNAs were transcribed in the presence of cold nucleotides and poly(A) tails were labeled in vitro as described below.
In Vitro Polyadenylation-Radiolabeled RNAs were polyadenylated in vitro using 1.5-2 units/fmol RNA of yeast poly(A) polymerase (United States Biochemical Corp.) and 500 M cold ATP according to the manufacturer's protocol. To prepare radioactive poly(A) tails, polyadenylation reactions were performed in the presence of 50 M ATP plus 25 Ci of [␣-32 P]ATP. Comparison of the sizes of mRNA with and without tails against an RNA ladder (Invitrogen) revealed that under these conditions the length of the poly(A) tail was between 150 and 200 As (data not shown).
In Vitro mRNA Decay Analysis-In vitro decay reactions were performed as described by Brewer and Ross (29) with minor modifications. Radiolabeled RNAs (2 fmol/reaction) were preincubated in 25 l of decay buffer (100 mM potassium acetate, 2 mM magnesium acetate, 2 mM dithiothreitol (DTE). 10 mM creatine phosphate, 1 g of creatine phosphokinase, 1 mM ATP, 0.4 mM GTP, 0.1 mM spermine, 10 mM Tris-HCl (pH 7.6), 80 units/ml RNasin) in the absence (control) or presence of 1 g of recombinant GST-HuD fusion protein. Recombinant proteins were prepared using pGEX-2T-HuD (a gift from Henry Furneaux) as described by Chung et al. (25). To control for the specificity of GST-HuD, some control reactions used 1 g of GST protein instead of the fusion protein. No differences were observed in the presence or absence of GST protein (data not shown). To begin the decay, isolated polysomes (0.1 A 600 unit/fmol of RNA) were added and samples incubated for various times. Decay reactions were terminated by addition of an equal volume of 2ϫ PK buffer (200 mM Tris-HCl (pH 7.5), 440 mM NaCl, 2% SDS, and 25 mM EDTA). Following proteinase K digestion, RNAs were isolated by phenol/chloroform extraction and ethanol precipitation and samples were analyzed in denaturing 10% polyacrylamide TBE, 8.3 M urea gels. All gels were dried and analyzed using a PC12-N36 cells transfected with either pMEP4-GAP alone or together with pcHuD were induced with CdCl 2 to express GAP-43. After removal of cadmium, cells were harvested at various times and total RNA was isolated and analyzed as described under "Experimental Procedures." A, representative Northern blots of RNA extracted from control and pcHuD-transfected PC12 cells. B, analysis of GAP-43 mRNA stability in transfected PC12 cells. GAP-43 mRNA levels were normalized to those for the glyceraldehyde-3-phosphate dehydrogenase mRNA and expressed as percentage of remaining mRNA relative to time 0. Regression analysis used a single rate exponential decay (Control) or a two-rate exponential decay (ϩHuD).
Personal Molecular Imager FX (Bio-Rad) and the accompanying Quantity One software.
Nitrocellulose Filter Binding Assays-Filter binding studies were performed as described by Chung et al. (25). Briefly, 2 fmol of [␣-32 P]UTP-radiolabeled full-length GAP-43 RNAs, with tails of 30 or 150 As, were incubated in the presence of recombinant HuD protein in binding buffer (50 mM Tris (pH 7.4), 1 mM MgCl 2 , 1 mM EDTA, 150 mM NaCl, and 1 mM DTE) at 37°C for 10 min. Protein concentrations for these studies ranged from of 30 to 600 nM. Samples were then transferred to nitrocellulose filters, and counts were analyzed by scintillation counting.

RESULTS
We have previously shown that HuD stabilizes the GAP-43 mRNA in transfected PC12 cells and neurons, leading to increased protein expression and neuronal differentiation (18,19,30). In this study, we sought to characterize the mechanism by which HuD exerts these effects using a combination of in vivo and in vitro mRNA decay assays. Initial studies examined the decay of the GAP-43 mRNA in transfected GAP-43-deficient PC12 cells using an inducible expression system. The pMEP4-GAP vector allows the transient expression of the GAP-43 mRNA by cadmium induction; its subsequent decay can then be measured by Northern blot analysis in the absence of any transcriptional inhibitors (10,16). In addition, to assess the effect of HuD in GAP-43 mRNA stability, some cultures were co-transfected with a constitutive HuD expression vector (pcHuD) that increases HuD protein level in the cell by 2-3-fold (18).
HuD Delays the Onset of Degradation of the GAP-43 mRNA-As shown in Fig. 1, in the absence of exogenous HuD, the GAP-43 mRNA decayed with a half-life (t1 ⁄2 ) of ϳ5 h, similar to that previously calculated using radioisotopic mRNA decay methods (9). Overexpression of HuD decreased the overall rate of decay of the mRNA, and the resulting t1 ⁄2 was calculated as 8.5 h. Regression analysis also indicated that the best fit for mRNA turnover in the presence of HuD was obtained using a two-rate exponential decay instead of a single rate as in the case of the GAP-43 mRNA alone. The biphasic decay profile consisted of an initial 3-h delay in the onset of degradation of the mRNA, followed a by similar second rate decay as in the absence of exogenous HuD. Delays in the onset of mRNA decay similar to this have been observed for mRNAs containing type I AREs in the 3Ј-UTR, such as c-Fos (31). In that case, the lag in decay reflected the time required to degrade the poly(A) tail. To test whether a similar mechanism was operating on the GAP-43 mRNA, subsequent experiments used H-mapping to monitor the decay of the 3Ј end of the mRNA.
After in vivo decay assays, RNAs were annealed to an antisense oligonucleotide against the 5Ј end of the GAP-43 3Ј-UTR and these RNA-DNA hybrids were digested with RNase H. The resulting fragments contained either the 5Ј-UTR and coding region or the entire GAP-43 3Ј-UTR and the poly(A) tail. As shown in Fig. 2, in the absence of excess HuD, both fragments decayed simultaneously with a rate similar to that of the entire mRNA. While the 5Ј end consisted of a band of unique size, the 3Ј end fragments showed greater heterogeneity, possibly because they contain poly(A) tails of varying sizes. This heterogeneity did not change during the decay, as 3Ј fragments disappeared without an apparent change in the relative abundance of the different size species. Therefore, the broad 3Ј end bands are more likely caused by the imprecise nature of the polyadenylation reaction than to a distributive deadenylation. The results in Fig. 2 also show that overexpression of HuD protein resulted in the stabilization of both fragments. The finding of similar decay rates for the 3Ј and 5Ј end fragments without the formation of intermediate size products could be the result of one of two alternative decay mechanisms. The first possibility is that the decay reaction is very processive. Supporting this possibility, processive deadenylation was shown to occur during the degradation of other ARE-containing mRNAs such as that for GM-CSF (31). Alternatively, the mRNA may be the target of endonucleolytic attack. To address these issues, subsequent studies utilized a cell-free mRNA decay system where individual components could be controlled more easily.
HuD Stabilizes the GAP-43 mRNA in Cell-free Decay Systems-We used a polysome-based mRNA decay system (29) to examine the effect of HuD on the half-life of in vitro synthesized GAP-43 mRNA. Isolated polysomes are known to contain all the enzymatic activities required for mRNA degradation. Although mRNAs decay at much faster rates in vitro than in vivo, the system faithfully reproduces the effect of cis-and trans-acting factors on mRNA half-life (32). In vitro decay reactions used a radiolabeled capped and polyadenylated fulllength GAP-43 mRNA as the substrate. Given that other RNAbinding proteins interact with the GAP-43 mRNA (16), it was important to preserve the structure of the mRNA as closely as that in the cells. Radiolabeled mRNAs were added to decay reactions containing purified polysomes from adult rat brain, in the presence or absence of recombinant HuD protein. Control reactions used recombinant GST at the same molar concentration, whereas others omitted recombinant protein altogether. Both controls yielded identical results (data not shown).
Initial studies evaluated the specificity of the in vitro decay system by examining the structural requirements for GAP-43 mRNA decay. As shown in Fig. 3, the same structural elements that are necessary for effective translation, the 5Ј cap and poly(A) tail, were also required for effective mRNA degradation. In addition, swapping the GAP-43 3Ј-UTR for that of ␤-globin dramatically increased the stability of the mRNA. This is consistent with the known stabilizing effect of the ␤-globin 3Ј-UTR when attached downstream of any unstable mRNA, including that of GAP-43, and is comparable with previous observations in cultured PC12 cells (10).
Once the specificity of the in vitro system was established, we examined the effect of HuD on the half-life of full-length GAP-43 mRNAs. Similar to the results obtained in PC12 cells (Fig. 1), addition of recombinant HuD protein increased the half-life of the GAP-43 mRNA by ϳ2-3-fold (Fig. 4). However, we found virtually no delay in the onset of mRNA decay in vitro, and the GAP-43 mRNA began to decay 2 min after starting the reaction. The lack of an apparent delay in the cell-free system may be caused by the increased speed of mRNA degradation, which would make it difficult to detect small delays in the decay. Alternatively, additional factors not present in the in vitro system may contribute to the delay. In any case, it is clear that the presence of excess HuD protein increases GAP-43 mRNA stability both in vivo and in vitro. There are two possible explanations for this stabilization. First, HuD could be affecting the turnover of the mRNA by binding to both the 3Ј-UTR and the poly(A) tail. Alternatively, the third RNA recognition motif (RRM III) could be binding to the poly(A) tail nonspecifically, preventing degradation of the messenger RNA. To test the requirement of the interaction of HuD with the GAP-43 3Ј-UTR in mRNA stabilization, we synthesized a mRNA in which the single 27-nucleotide HuD binding site was deleted. This mRNA (⌬HuD) was then used in decay experiments in the presence or absence of GST-HuD protein. As shown in Fig. 5, HuD did not affect the decay of GAP-43 mRNAs that lack the HuD binding site, indicating that this protein must specifically bind to its recognition site in the 3Ј-UTR to stabilize the GAP-43 mRNA. This is in agreement with a recent study demonstrating that mutations of the ARE of tumor necrosis factor-␣ abolished the capacity of HuR to stabilize the mRNA (33).
A separate question about the in vitro reaction was whether the decay of GAP-43 mRNA proceeded in a processive manner as was found to occur in vivo. To address this question, we used H-mapping to measure the decay of the 3Ј and 5Ј ends of the mRNA. As shown in Fig. 6, the fragments corresponding to both ends of the mRNA decayed rapidly in a processive manner, similar to that found in vivo (Fig. 2). Given that we were unable to determine the rate of deadenylation from these reactions, we utilized a different strategy to measure the rate of poly(A) removal, using mRNAs with radiolabeled tails (see below).
HuD Decreases the Rate of Poly(A) Tail Removal-To measure the effect of HuD on the rate of poly(A) removal, we prepared cold GAP-43 mRNAs that were radiolabeled in the tail by in vitro polyadenylation and used these as substrates

FIG. 5. HuD does not stabilize ⌬HuD, a mRNA with a deletion in the HuD binding site in the 3-UTR. A, p⌬HuD was constructed and used for in vitro transcription as indicated under "Experimental
Procedures." A, the ⌬HuD mRNA was analyzed by in vitro decay assays as described in Fig. 4. B, graph showing that the rate of degradation of the GAP-43 mRNA lacking the HuD binding site did not change in presence or absence of recombinant HuD. for the decay reaction. As shown in Fig. 7, addition of HuD protein caused a significant decrease in the rate of deadenylation of the RNA (t1 ⁄2(ϪHuD) ϭ 2 min, t1 ⁄2(ϩHuD) ϭ 4 min). The finding that the t1 ⁄2 observed for the decay of the tails is significantly shorter than that found for the whole mRNA (t1 ⁄2(ϪHuD) ϭ 8 min, Fig. 4) suggests that the decay of the poly(A) tail precedes the decay of the body.
HuD Preferentially Stabilizes GAP-43 mRNAs with Long Poly(A) Tails-To further understand the role of HuD in the stabilization of GAP-43 mRNA, we wanted to explore whether the size of the poly(A) tail, which shows different affinities for the third RRM in HuD, also played a role in the stabilizing effect of HuD. To this end, we synthesized and tested GAP-43 mRNAs with either short (30 As) or long (150 As) tails in our cell-free decay system. As shown in Fig. 8, in the absence of added HuD, GAP-43 mRNAs with short tails decayed at a significantly faster rate than those containing long tails (t1 ⁄2 ϭ 4 min A30 versus t1 ⁄2 ϭ 8 min A150). These results are also consistent with a 3Ј to 5Ј direction in the decay of the mRNA. Addition of recombinant HuD protein was found to stabilize preferentially the mRNAs with long tails and did not affect the decay of those with short tails (t1 ⁄2 ϭ 5 min A30 versus t1 ⁄2 ϭ 17 min A150).

HuD Binds with Higher Affinity to GAP-43 Molecules with Long Poly(A) Tails-The previous experiments suggested that
HuD has an effect on GAP-43 mRNA turnover only when poly(A) tails are long, which corresponds to the initial phase of decay relative to the life of an individual mRNA molecule. To test whether HuD binds with varying affinities to GAP-43 mRNA molecules with either long or short tails, we used nitrocellulose filter binding assays (Fig. 9). Consistent with its effects on mRNA stability, HuD was found to bind with an 8-fold higher affinity to mRNAs with long tails of (K d(A150) ϭ 5 nM) compared with those with short tails (K d(A30) ϭ 41 nM). We believe that this change in binding affinity, which is dependent on the length of the poly(A) tail, is a key feature in the function of HuD. It allows HuD to bind to GAP-43 mRNAs with long poly(A) tails with high affinity impacting the stability of the mRNA during the initial phase of degradation.  (18,19,30). Furthermore, we have recently found overlapping distributions of the GAP-43 and HuD mRNAs in many regions of the mature nervous system (38). Despite compelling biological evidence for the role of HuD in GAP-43 expression, the mechanism by which HuD exerts its effects is poorly understood. Here, we show that HuD stabilizes the GAP-43 mRNA in transfected PC12 cells by delaying the onset of decay. This is accomplished by the binding of HuD to the 3Ј-UTR of GAP-43 mRNAs with long poly(A) tails. As the size of the poly(A) tail becomes shorter, the affinity of HuD for the ARE in the GAP-43 3Ј-UTR decreases and HuD is no longer able to stabilize the mRNA. Because long poly(A) tails are also required for effective translation, these results suggest that HuD selectively protects translation-competent mRNAs.
It is widely accepted that degradation of most eukaryotic mRNAs studied to date begins at the 3Ј end, with deadenylation as the first step (for reviews, see Refs. 39 and 40). Results from both our in vivo and in vitro decay assays suggest that the GAP-43 mRNA also decays in a 3Ј to 5Ј direction. First, we found that the half-life for the decay of the poly(A) tail was significantly shorter than that of the body. Second and most importantly, we showed that mRNAs with short tails (A30) decayed significantly faster than those containing long tails (A150). Further studies on the mechanisms of degradation of the GAP-43 mRNA revealed that the 3Ј and 5Ј ends of the mRNA simultaneously decayed without producing intermediates with increasingly shorter tails. Decay kinetics similar to these are seen in class II AREs like GM-CSF, tumor necrosis factor-␣, and interleukin-3 mRNAs, which are characterized by processive deadenylation activity and asynchronous poly(A) tail removal (41). Given the processivity of the reaction, we were not able to completely rule out the involvement of endonucleases in GAP-43 mRNA degradation. In this regard, it is interesting to note that none of the ARE elements studied to date have been associated with any endonucleolytic activity.
Using a polysome-based cell-free decay system, we observed that the elements involved in GAP-43 mRNA degradation are the same as those required for translation. This is in agreement with previous studies demonstrating a coupling between translation and degradation of mRNAs (42)(43)(44). Furthermore, we found that HuD decreased the rate of poly(A) tail removal of GAP-43 mRNAs with long tails. In contrast to our results, two recent studies found that other ELAV-like proteins did not affect the rate of deadenylation of different mRNAs (45,46). There are several experimental differences between these studies and ours. First, the study by Peng et al. (45) examined the effect of HuR on the decay of a chimeric RNA containing ␤-globin sequences linked to multiple type I ARE sequences, whereas we used a full-length GAP-43 mRNA with a single type III ARE. It is known that mRNAs containing multiple adjacent AREs decay at faster rates than those with a single or dispersed AREs (41). Additionally, deadenylation rates resulting from type III AREs like c-Jun are slower than those of type I AREs (47,48). It is therefore likely that the differences between Peng et al. (45) and our study are the result of the type of ARE sequence present in the mRNA. Likewise, there are several differences in the experimental system used by Ford et al. (46), who found that the ELAV-like proteins HuR and Hel-N1 did not affect the rate of deadenylation of GM-CSF. These investigators used S100 extracts as the source for the mRNA decay activity and synthetic mRNAs with a tail of 60 As at the 3Ј end as the substrate for the in vitro decay reactions. They also used mRNAs containing a different type of ARE, the type II ARE in GM-CSF. Thus, the observed differences could be caused by the in vitro decay system and substrate used for the decay reaction. Additionally, as our data indicated, the impact of HuD on deadenylation is only observed with longer poly(A) tails and mRNA with tails of 60 As may not be long enough to see an effect. Alternatively, the apparent discrepancies might be the result of functional differences between HuR, Hel-N1, and HuD and the decay pathways or protein complexes with which they are associated.
In agreement with the requirement of long tails for the stabilizing effect of HuD on the GAP-43 mRNA, we found that this protein has an 8-fold higher affinity for its recognition element in the GAP-43 3Ј-UTR when the mRNA has long tails. One possible explanation for this increased affinity is that the structure of mRNAs with long tails makes them a better target for the binding of HuD. Alternatively, the third RRM of ELAVlike proteins, which binds to long stretches of poly(A) (23,24), may facilitate the binding of RRMs I and II of HuD to the ARE in the GAP-43 3Ј-UTR. Along these lines, it is worth noting that the affinity of HuR for poly(A) RNA increases with the length of the RNA from A100 to A300, whereas this protein shows little binding to A30 on its own. Furthermore, HuR was found to interact simultaneously with ARE sequences in the 3Ј-UTR and with the poly(A) tail (23). In this study we only measured the binding of HuD to the site in the 3Ј-UTR as a function of tail length; however, our data are consistent with the interaction of HuD with long poly(A) extensions.
Based upon our results, we propose a biphasic model for the effect of HuD on GAP-43 mRNA stability. At the initial stages of decay, GAP-43 mRNAs contain long poly(A) tails and HuD binds with high affinity to its recognition element in the 3Ј-UTR and to the tail. This RNA-protein interaction decreases the rate of deadenylation of the poly(A) tail, which in turn delays the onset of degradation of the body of the mRNA. Once poly(A) tails become shorter, the affinity of HuD for its recognition element decreases and the protein is no longer effective at protecting the mRNA from nuclease attack. Although approximate, this model predicts that the length of the poly(A) tail is the primary element in determining the ability of HuD for protecting the mRNA. Given the role of the poly(A) tail in protein synthesis, it is likely that the effects of HuD may be restricted to translation-competent mRNAs. As the ultimate goal of gene expression control is achieved at the level of protein synthesis, it is not surprising that the structural elements for GAP-43 mRNA stabilization and translation are interrelated. In fact, in all cases examined, there is a strict correlation between GAP-43 mRNA and protein levels in the cell (1, 2, 8, 18, 19). It is becoming apparent that deadenylation is the first and rate-limiting step in mRNA degradation (49). Deadenylation rates depend on sequences at the 5Ј and 3Ј ends of the mRNAs and on the interactions of many RNA-binding proteins, including the poly(A)-binding protein, the cap-binding protein (eIF4E), the initiation factor eIF4G, and the poly(A)-specific RNA nuclease PARN (43,50,51). Besides the ELAV-like proteins, different ARE-binding proteins such as AUF1/hnRNPD (52,53) and other types of regulatory factors (e.g. EDEN-BP (Ref. 48)) can regulate mRNA stability by influencing poly(A) tail removal. The mechanisms by which neurons control the size of poly(A) tail in the GAP-43 mRNA are not well understood. Given the number of proteins involved in this process, there is potential for complex regulation. With the recent generation of transgenic mice overexpressing HuD and other RNAbinding proteins, it will be soon possible to address these issues in vivo.