p21waf1 mRNA Contains a Conserved Element in Its 3′-Untranslated Region That Is Bound by the Elav-like mRNA-stabilizing Proteins*

The Elav-like proteins are specific mRNA-binding proteins that regulate mRNA stability. The neuronal members of this family (HuD, HuC, and Hel-N1) are required for neuronal differentiation. In this report, using purified HuD protein we have localized a high affinity HuD binding site to a 42-nucleotide region within a U-rich tract in the 3′-untranslated region p21waf1 mRNA. The binding of HuD to this site is readily displaced by an RNA oligonucleotide encoding the HuD binding site of c-fos. The sequence of this binding site is well conserved in human, mouse, and rat p21waf1 mRNA. p21waf1 is an inhibitor of cyclin-dependent kinases and proliferating cell nuclear antigen and induces cell cycle arrest at G1/S, a requisite early step in cell differentiation. The identification of an Elav-like protein binding site in the 3′-untranslated region of p21waf1 provides a novel link between the induction of differentiation, mRNA stability, and the termination of the cell cycle.

The Elav-like proteins are tumor antigens that are the human homologues of the Drosophila protein Elav (1).There are four members of the human Elav-like family, namely HuD, HuC, Hel-N1, and HuR (1)(2)(3)(4)(5). HuD, HuC, and Hel-N1 are exclusively expressed in post-mitotic neurons and in neuroendocrine tumors. (4,6,7). Elav was originally defined in studies of mutants that were defective in neural function (8,9). In Elav mutant flies, neuroblasts fail to completely differentiate into neurons and result in embryonic lethality. It is thought that HuD, HuC, and Hel-N1 are also involved in neuronal differentiation, since they are expressed on termination of the neuroblast cell cycle (10 -12). Indeed, recent evidence has shown that HuD is necessary for neuronal differentiation. 1 Treatment of PC12 cells with nerve growth factor leads to cessation of the cell cycle and differentiation into neuron-like cells with neurite outgrowth (14). Treatment of PC12 cells with antisense oligonucleotides directed against HuD abrogated their response to nerve growth factor. 1 In complementary experiments, Wakamatsu and Weston (10) showed that transfection of HuD into chick neuroblasts accelerated their differentiation. Thus, the Elav-like proteins are required for neuronal differentiation, and it is important to establish their mechanism of action.
The Elav-like proteins are RNA-binding proteins and contain three RNA recognition motifs of the RNP2/RNP1 type (15). The first and second of these RNA recognition motifs are in tandem and are separated from the third by a segment rich in basic amino acids. Many mRNAs contain U-rich regulatory elements that direct their rapid turnover (16,17). The Elav-like proteins specifically bind to these elements, stabilize the mRNA, and thereby increase its steady state level. (3, 5, 18 -23). 1 The current hypothesis is that the Elav-like proteins promote neuronal differentiation by stabilizing specific mRNAs. We know that that the Elav-like proteins regulate tau and GAP-43 mRNAs, two gene products necessary for the later stages of neuronal differentiation (24). 1 However, the Elav-like proteins are turned on very soon after the last S phase of the precursor neuroblast (6,11). This event occurs before the documented changes in GAP-43 and tau expression. This suggested to us that the Elav-like proteins may regulate other mRNAs. In particular, we postulated that the Elav-like proteins may regulate mRNAs involved in the termination of the neuroblast cell cycle. In nerve growth factor-treated PC12 cells, it is thought that the cessation of the cell cycle is primarily mediated by the induction of p21 waf1 (25,26). p21 waf1 inhibits cyclin-dependent kinases and thereby induces cell cycle arrest at G 1 /S (27)(28)(29). It is known that p21 waf1 mRNA has a short half-life and can be regulated at the posttranscriptional level (30 -34). Thus p21 waf1 may be the connection between the Elav-like proteins and the termination of the neuroblast cell cycle. In this paper we show that HuD binds to a conserved U-rich element within the 3Ј-UTR 2 of p21 waf1 mRNA.
Purification of GST-HuD Proteins-An overnight culture of Escherichia coli BL 21 transformed with the GST-HuD construct (20) was diluted 1:50 in LB broth. At an A 600 of 0.4, the culture was induced with isopropyl-1-thio-␤-D-galactopyranoside (0.1 mM). After 4 h of further growth, cells were spun down and resuspended in 10 ml of buffer A (50 mM Tris (pH 8.0), 200 mM NaCl, 1 mM EDTA). The cells were lysed by adding lysozyme (0.2 mg/ml) and Triton X-100 (1%). The lysate was centrifuged at 12,000 ϫ g for 30 min. The resultant supernatant was loaded onto a glutathione-agarose affinity column (13 mg of protein/ml of resin). After washing the column with buffer B (50 mM Tris (pH 8.0), 200 mM NaCl, 1 mM EDTA, 1% Triton X-100), GST-HuD was eluted with 50 mM Tris (pH 8.0), 5 mM glutathione. The eluted material containing a predominant band of 67 kDa was pooled and stored at Ϫ70°C.
Assay of RNA⅐HuD Complex Formation-Reaction mixtures (20 l) contained 50 mM Tris (pH 7.0), 150 mM NaCl, 0.25 mg/ml tRNA, 0.25 mg/ml bovine serum albumin, 10 fmol of labeled RNA, and protein as indicated. Mixtures were incubated at 37°C for 10 min. After incubation, 4 l of a dye mixture (50% glycerol, 0.1% bromphenol blue, 0.1% xylene cyanol) was added. Five l of the reaction mixture was immediately loaded onto a 1% agarose gel in TAE buffer (40 mM Tris acetate, 1 mM EDTA) and electrophoresed at 40 V for 2.5 h. The gel was dried onto DE81 paper (Whatman) backed with gel drying paper and exposed to XAR5 film (Eastman Kodak Co.) for 4 -5 h at Ϫ70°C. All experiments were repeated at least twice with identical results.
RNase T1 Selection Assay-Reaction mixtures (20 l) contained 50 mM Tris (pH 7.0), 150 mM NaCl, 0.25 mg/ml bovine serum albumin, 0.25 mg/ml tRNA, 20 fmol of radiolabeled mRNA, purified HuD as indicated. After 10 min of incubation at 37°C, 0.5 units of RNase T1 was added to each reaction and incubated at 37°C for an additional 10 min. The mixtures were diluted 1:6 with buffer F (20 mM Tris (pH 7.0), 150 mM NaCl, 0.05 mg/ml tRNA) and filtered through nitrocellulose (BA85, Schleicher & Schuell). After washing the nitrocellulose twice with buffer F, bound HuD⅐RNA complex was extracted with phenol-chloroform and concentrated by ethanol precipitation. The RNA pellet was dissolved in formamide buffer and denatured at 65°C for 2 min. Samples were analyzed by electrophoresis on a 12% polyacrylamide, 50% urea gel. Gels were fixed in 1:1:8 acetic acid:methanol:water, dried, and exposed to the XAR5 film at Ϫ70°C overnight.
Nitrocellulose Filter Binding Assay-Reaction mixtures (20 l) contained 50 mM Tris (pH 7.0), 150 mM NaCl, 0.25 mg/ml bovine serum albumin, 0.25 mg/ml tRNA, 10 fmol of radiolabeled mRNA, and purified HuD as indicated. After incubation for 10 min at 37°C, the mixtures were diluted 1:6 with buffer F and filtered through nitrocellulose. Filters were washed twice with buffer F, and bound radioactivity was determined by Cerenkov counting. All data were reproducible within a fluctuation of 10%.

RESULTS
HuD Binds to p21 waf1 mRNA-The structure and sequence of the full-length p21 waf1 mRNA (p21/SmaI, 2121 nucleotides) is shown in Fig. 1. In the forthcoming text we will refer to the mRNA as only p21. This will simplify our references to the various subsegments. The transcript encoding this message is too large to assay using conventional gel retardation assays. Thus the RNase T1 selection assay (20) was used to ascertain whether the p21 message contained a binding site for HuD. Radiolabeled transcript was incubated with purified recombinant HuD protein and treated with RNase T1, a specific endonuclease that cleaves after G residues. This reaction was conducted with saturating RNase T1 levels, so that all G residues in the transcript are hydrolyzed under the reaction conditions used. The reaction mixture was then filtered through nitrocellulose. RNA/protein complexes are absorbed to nitrocellulose (36). These complexes were eluted from the nitrocellulose by phenol/chloroform extraction and analyzed by gel electrophoresis in order identify the segments of the transcript that bound HuD.
Three fragments of 24, 16, and 15 nucleotides were selected from the full-length transcript (p21/Sma, Fig. 1A) by HuD ( Fig.  2A, lane 2), whereas no fragments were selected by a GST protein (lane 1), used as a negative control in this experiment. To map these binding sites more precisely, we used this assay to examine two truncated transcripts, p21/Pst and p21/BstEII (Fig. 1A). The p21/BstEII transcript binds fragments of the same size as does the full-length transcript (p21/Sma), although the p21/PstI transcript fails to bind any fragments ( Fig.  2A, lanes 2, 5, and 8, respectively). These data suggest that the HuD binding site lies between the PstI and BstEII sites. To confirm this, we used the RNase T1 selection assay to analyze the three adjacent transcripts, p21/BsrBI, p21/BsrBI-BstEII, and p21 BstEII-Sma (Fig. 1A), that were synthesized to span the entire length of the p21 message. Since only the p21 BsrBI-BstEII transcript yielded the selected fragments (Fig. 2B, lanes  4 -6), the HuD binding site was localized to a region in the 3Ј-UTR between the BsrBI and BstEII sites.
Quantitative Determination of the Affinity of HuD for p21 waf1 mRNA-To quantitate the relative binding affinities of HuD to p21 waf1 mRNA, we employed the method originally used for the R17 coat protein (36). A low amount of labeled RNA (350 pM) was incubated with HuD protein under conditions of protein excess. The reactions were filtered through nitrocellulose, and the bound radioactivity was measured. Fig. 3A shows that the formation of p21/BsrBI-BstEII RNA⅐HuD complex is first detectable at a HuD concentration of 0.15 nM and plateaus at a concentration above 500 nM with about 72% of the input RNA bound. In contrast, complex formation with p21/BsrBI was not detectable even at 1000 nM HuD. A plot of the log of complex/ free RNA versus the log of HuD concentration is shown in Fig.  3B. From this plot, the K d of the p21/BsrBI-BstEII⅐HuD complex was determined to be 62.7 nM. Thus the affinity of HuD for p21 mRNA is similar to what we determined for c-fos mRNA (K d ϭ 19 nM) (20).
Mapping of the HuD Binding Site-We next examined the sequence of the p21/BsrBI-BstEII transcript (nucleotides 636 -1122, Fig. 1B) to delineate the likely location of the 15-, 16-, and 24-nucleotide binding sites. Based on this analysis, we divided p21/ BsrBI-BstEII segment into three contiguous transcripts: A, B, and C. Transcript A contains a 16-nucleotide RNase T1 fragment separated from a 24-nucleotide RNase T1 fragment by two nucleotides, whereas B contains two adjacent 15-nucleotide fragments, and C contains only a single 16-nucleotide fragment. Transcripts A, B, C, and p21 BsrBI-BstEII were then tested for HuD binding by gel retardation analysis (Fig. 4A). Purified recombinant HuD was incubated with labeled transcript and assayed for complex formation. HuD binds both p21/BsrBI-BstEII and A with similar affinity (lanes 3-5 and 8 -10, respectively) and with a considerably lower affinity to B (lanes [13][14][15]. No binding to C was observed within the range of HuD concentrations tested (lanes 18 -20). As expected, no complex formation was observed with transcripts A, B, or C when HuD was replaced by GST (lanes 2, 7, 12, and 17). These semiquantitative observations were confirmed by nitrocellu-  1-2), p21/BstEII (lanes 4 -5), and p21 Pst (lanes 7-8) transcripts (4 fmol, 20,000 cpm/pmol of UTP) were incubated with 50 nM HuD or GST as indicated at 37°C for 10 min. After treatment with RNase T1 (5 units/reaction) at 37°C for 10 min, the reaction mixtures were filtered through nitrocellulose. Bound RNA fragments were extracted and electrophoresed on a 12% acrylamide, 50% urea gel. Lanes 3, 6, and 9 show RNase T1 digestions of the indicated transcripts before nitrocellulose filtration. Lane M, size markers, X174/HinfI digest. B, same as A, except with 32 P-labeled p21/ BsrBI (lanes 1-3), p21/BsrBI-BstEII (lanes 4 -6), and p21/BstEII-Sma (lanes 7-9) transcripts (7 fmol, 45,000 cpm/pmol of UTP). lose binding assays by which the K d values of p21/BsrBI-BstEII, A, B, and C were determined to be 62.7, 85.6, 757, and 10,700 nM, respectively. On the basis of these data, it is likely that transcript A contains the major HuD binding site within the 3Ј-UTR of p21 waf1 mRNA. A minor HuD binding site appears to also be present in B, whereas no site with significant affinity for HuD is observed in C.
To confirm the location of the HuD binding sites within p21 waf1 mRNA, transcripts A and B were further analyzed by the RNase T1 selection assay (Fig. 4B). Fragments of 24 and 16 nucleotides in length were selected after incubation of transcript A with HuD (lane 5), whereas a 15-nucleotide fragment was the predominant species selected after the binding of tran-  1-5), A (lanes 6 -10), B (lanes 11-15), and C (lanes 16 -20) transcripts (7 fmol, 45,000 cpm/pmol of UTP) were incubated with the indicated concentrations of GST or HuD. After incubation at 37°C for 10 min, 20% of the reaction mixtures were resolved by gel electrophoresis on a 1% agarose gel. B, 32 P-labeled p21/BsrBI (lanes 1-2), A (lanes 4 -5), and B (lanes 7-8) transcripts (7 fmol, 45,000 cpm/pmol of UTP) were incubated with 50 nM HuD or GST as indicated at 37°C for 10 min. After treatment with RNase T1 (5 units/reaction) at 37°C for 10 min, the reaction mixtures were filtered through nitrocellulose. Bound RNA fragments were extracted and electrophoresed on a 12% acrylamide, 50% urea gel. Lanes 3, 6, and 9 show RNase T1 digestions of the indicated transcripts before nitrocellulose filtration. Lane M, size markers, X174/HinfI digest. C, the selected fragments (A24, A16, and B15) corresponding to the 24 and 16 nucleotide bands derived from transcript A and the 15 nucleotide band derived from transcript B, as FIG. 5. The effect of RNase T1 concentration on the selection of HuD binding sites. A, lanes 1-6, 32 P-labeled p21/BsrBI-BstEII (7 fmol, 45,000 cpm/pmol of UTP) was incubated with 100 nM HuD or GST as indicated at 37°C for 10 min. After treatment with the indicated amounts of RNase T1 at 37°C for 10 min, the reaction mixtures were filtered through nitrocellulose. Bound RNA fragments were extracted from the filter. RNA (lanes 7-12) was also incubated with the indicated amounts of RNase T1 at 37°C for 10 min. 10% of the reaction mixtures were electrophoresed on a 12% acrylamide, 50% urea gel. Lane M, size markers, X174/HinfI digest. B, same as for A, except with transcript p21/BsrBI, as a negative control. observed in Fig. 7, were eluted from a preparative gel and analyzed after further digestion with RNase T1, 0.5 units/reaction (lanes 2, 4, and  6, respectively) and with no further RNase T1 digestion (lanes 1, 3, 1, 4, and  7). In the RNase T1 selection assay, it is possible that a selected fragment could result from the protection by HuD of one or more G residues situated between two smaller fragments. The three selected fragments (24,16, and 15 nucleotides) were gel-purified and incubated with RNase T1 to determine whether they contain any protected G residues (Fig. 4C, lanes  2, 4, and 6 and lanes 1, 3, and 5, respectively). No new digestion products were observed, thus indicating that the observed fragments are as predicted 24, 16, (transcript A), and 15 nucleotides (transcript B) in length. To ensure that the Rnase T1 used in these experiments was active, the entire transcript B was digested by RNase T1 as a positive control (lane 8). Thus, based on the data obtained by gel retardation, RNase T1 digestion, and nitrocellulose binding assays, we conclude that the major binding site in p21 waf1 mRNA is localized to a 42-nucleotide region (nucleotides 657-698) within the 3Ј-UTR (Fig. 1B, boldface). This site is similar to other HuD binding sites in that it contains U-rich tracts (19,20,24). However, unlike other known HuD binding sites, two G residues are present within the site in p21 waf1 , and we obtained two fragments in the selection assay. If the localization of the 24-and 16-nucleotide fragments is correct, we should be able to select a 42-nucleotide fragment from a partial digest of p21 mRNA. Thus, we generated partial digests of p21/A using limiting concentrations of RNase T1 (Fig. 5A, lanes 7-12). As before, HuD selected fragments of 24 and 16 nucleotides from complete digests ( lanes  1-2). However, as predicted, HuD selected a 42-nucleotide fragment from the first partial digest and progressively larger fragments from less digested RNA (lanes [3][4][5]. Concomitant with the selection of the 42-nucleotide fragment was the disappearance of both the 24-and 16-nucleotide fragments (lanes  3-4). At all concentrations of RNase T1 tested, HuD binding to these fragments was specific. This was shown by the lack of selected fragments from similar partial digests of a transcript that lacks a HuD binding site (Fig. 5B, lanes 1-12). These data indicate that the minimal HuD binding site is 42 nucleotides in length.
The Site in p21 mRNA Is Similar in Structure to Other HuD Sites and Is Well Conserved-We have compared the sequence of the p21 mRNA binding site with those we have found in c-fos, c-myc, tau, interleukin-3, and GAP-43 mRNAs (Fig. 6A). The p21 binding site is similar to the others in that it contains U-rich tracts. We have shown that these U-rich tracts are essential for the binding of HuD to the c-fos element (20). The importance of the U tract sequences in the p21 site is further underscored by their strict conservation between human, mouse, and rat p21 mRNAs (Fig. 6B).
Next, we tested whether the HuD⅐p21mRNA complex would be displaced by the c-fos binding site. The sequence AUA-UUUAUAUUUUUAUUUUAUUUUUUU, termed the c-fos element, has been determined to be the minimal HuD binding site within c-fos mRNA (20). Another oligonucleotide, a mutated c-fos element with the sequence AUACGUAUAUCGC-UAUGCUAUUUUUUU that fails to bind HuD, served as the negative control in these experiments. As shown in Fig. 7, the addition of unlabeled c-fos element oligonucleotide in molar excess displaced the HuD⅐p21/BsrBI-BstEII complex in a gel retardation assay (lanes [3][4][5]. No complex displacement was observed when the same molar amounts of the mutated c-fos element was added (lanes 6 -8). We can therefore conclude that the binding site in p21 is similar to the one in c-fos mRNA and indeed to sites in c-myc, tau Il-3, and GAP-43 mRNAs.

DISCUSSION
The Elav-like proteins have emerged as an important class of proteins that are pleotropic effectors of gene expression in mammalian cells. There are four members of the human family HuD, HuC, Hel-N1, and HuR (1-4). These RNA-binding proteins regulate gene expression via a specific interaction with U-rich elements in mRNA (18, 20 -23).These U-rich elements  5 and lanes 6 -8, respectively). After incubation at 37°C for 10 min, 20% of the reaction mixtures were resolved on a 1% agarose gel. are found in a wide variety of mRNAs and usually target them for rapid degradation (17). The current model is that activation of the Elav-like proteins leads to the stabilization of a specific subset of mRNAs. It is likely that this mechanism is responsible for the induction of neuronal differentiation by the neuronalspecific Elav-like proteins. Some of the target mRNAs stabilized by the Elav-like proteins during neuronal differentiation have been identified. Two targets, GAP-43 and tau, are required for the later steps in neuronal differentiation (24). 1 The neuronal Elav-like proteins, however, are expressed before the up-regulation of GAP-43 and tau mRNA. Thus we have looked for new mRNAs targets that may directly affect the neuroblast cell cycle. In nerve growth factor-treated PC12 cells, there is compelling evidence to suggest that the induction of p21 waf1 , an inhibitor of cyclin-cdk complexes as well as proliferating cell nuclear antigen, plays a critical role in arresting these cells at G 1 /S, a requisite early step in cell differentiation (25,26). Thus we examined the interaction between p21 waf1 mRNA and HuD. In this report, we have localized a high affinity HuD binding site to a conserved 42-nucleotide region within a U-rich tract in the 3Ј-UTR of p21 waf1 mRNA. Thus, given that the Elav-like proteins regulate other mRNAs with similar sites, it is very likely that the Elav-like proteins will regulate p21 waf1 mRNA. With the detailed knowledge of the p21 waf1 binding site reported here, experiments to directly test this possibility can now be carried out.
Surprisingly, possible changes in the half-life of p21 waf1 mRNA during neurogenesis have not yet been examined. Recent studies, however, in other cell systems have shown that the half-life of p21 waf1 is regulated by diverse signals. In these systems, regulation cannot be due to binding of HuD or any of the other neuronal-specific members of the family. Rather, it is more likely to be mediated by the ubiquitously expressed member of the Elav-like protein family, HuR, which has also been shown to bind to this region of the p21 waf1 message (data not shown). p21 waf1 is up-regulated in p53-deficient promyelocytic HL-60 cells induced to differentiate along the monocytic lineage by phorbol ester or 1␣,25-dihydroxyvitamin D 3. (30). The marked increases of both p21 waf1 mRNA and protein expression observed in these cells is due to enhanced mRNA stability. Similar increases in p21 waf1 mRNA stability were observed in MCF-7 breast carcinoma cells (37) and in MDA-MB-468 cells (a p53-mutant derivative of MCF-7 cells) induced to differentiate with the retinoid CD437 (31). Zeng and El-Deiry (32) found that okadaic acid, a potent serine/threonine phosphatase inhibitor, induced p53-independent expression of p21 waf1 , both at the transcriptional and posttranscriptional levels in several human tumor cell lines. It has recently been reported (13) that UVC irradiation elevates p21 waf1 mRNA expression in mouse embryonal fibroblasts and human rectal carcinoma cells by significantly enhancing the stability of the mRNA encoding p21 waf1 . They provide experimental evidence indicating that this effect is mediated by a tyrosine kinase/phosphatase regulatory system (possibly one modulated by c-Jun NH 2 -terminal kinase) and, in contrast to the systems already discussed, requires the presence of functional p53. Thus it will be interesting to see if the binding of the Elav-like proteins to p21 waf1 mRNA is activated by any of these signals.
The connection drawn here between mRNA stability, differentiation and the cell cycle is a novel one. It has been tacitly assumed that transcriptional control is the rate-limiting pathway that defines commitment to differentiation. A distinctive feature of the mechanism postulated here is that the Elav-like proteins are capable of coordinately regulating a subclass of mRNAs. We have recently observed that the Elav-like proteins also bind to p27 mRNA, another negative regulator of the cell cycle. 3 These observations raise the possibility that abrogation of Elav-like protein activity in differentiated cells may coordinately destabilize mRNAs encoding negative regulators of the cell cycle. Indeed an important future goal will be to investigate whether this would lead to reentry into S phase and cell proliferation.