| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J. Biol. Chem., Vol. 278, Issue 41, 39801-39808, October 10, 2003
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
¶
From the
Departments of Biochemistry and Molecular Biology and Surgery, Keck School of Medicine, University of Southern California/Norris Comprehensive Cancer Center, Los Angeles, California 90089-9176
Received for publication, July 3, 2003 , and in revised form, August 4, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
AREs carry out their function through the binding of trans-acting factors. Numerous ARE-binding proteins have been characterized, including AUBF, AUF-1 (hnRNP D), hnRNPA1, hnRNPC, AUH, tristetraprolin (TTP), TIAR and TIA-1, KSRP, Hu proteins, and others (reviewed in Ref. 1). Although the function of some ARE-binding proteins remains unknown, a subset of them (AUF-1, TTP, KSRP, and BRF1) has been implicated in accelerating mRNA decay (15–18). Others, such as members of the Hu protein family, appear to stabilize ARE-containing mRNAs, presumably by competing with the destabilizing factors (1).
Hu proteins are a family of RNA-binding proteins that show homology to the Drosophila protein ELAV (embryonic lethal abnormal visual system) (19–24). There are four known Hu family members: the ubiquitous HuR(HuA) and neuronal-specific HuB (Hel-N1), HuC, and HuD. All four proteins contain three RNA recognition motif-type RNA-binding domains (RRMs), 
90-amino acid domains found in hundreds of RNA-binding proteins of diverse function (25). Evidence supporting a role of Hu proteins in antagonizing ARE-mediated RNA decay has been provided by a large number of studies. In vitro analyses have shown that these proteins can bind to AREs (19, 22, 26–31). Experiments in tissue culture, in vitro decay analyses, and in vivo studies have demonstrated direct effects of Hu proteins on the levels and/or stability of ARE-containing mRNAs (32–46). Hu proteins have also been implicated in cancer, as they are overexpressed in neuroblastoma and small cell lung cancer, two types of cancer that show stabilization of labile proto-oncogene mRNAs such as those of the MYC, MYCN, and FOS genes (35, 46, 47).
One of the most compelling studies supporting the idea that Hu proteins slow mRNA decay through competition with destabilizing ARE-binding proteins is an analysis of the dynamic regulation of interleukin-3 mRNA, which is stabilized by HuR and destabilized by TTP (48). The interleukin-3 study illustrates that the identity of the proteins bound to an ARE plays an important role in determining the fate of the mRNA. Which proteins are bound to a given ARE is determined by many factors: the intracellular location and concentration of the various proteins, their RNA-binding kinetics, their sequence preference, the minimal sequence required for their recruitment to the RNA, their potential to bind cooperatively, and their interactions with other proteins. Despite the importance of these characteristics, our knowledge about the mechanism by which ARE-binding proteins interact with their RNA targets remains very limited. Although studies of the interaction between these proteins and their natural RNA targets are important, a detailed mechanistic understanding of RNA binding by Hu proteins is difficult to obtain with the long and usually heterogeneous naturally occurring AU-rich elements. Use of these heterogeneous sequences complicates experimental interpretation, such as the determination of the identity of individual binding sites and the exact number of proteins bound (and therefore the affinity of the interaction). For this reason, we use a model system of short synthetic AU-rich sequences based on the prototype mRNA destabilizing element UUAUUUAUU (9, 10). Our aims are to gain detailed mechanistic insight into RNA binding by Hu proteins: to better define the size and the sequence of the RNA-binding target and the affinity and kinetic parameters for this interaction.
Of the four highly conserved family members, we have chosen HuD as a model system because its interaction with RNA has been best characterized biochemically, kinetically, and structurally (28, 49, 50). We use HuD RRM domain deletion mutants and the individual RRMs to explore the role of the different domains in binding and in determining target specificity, and in vitro genetics to further probe sequence preferences. Finally, we compare the sequence preferences of HuD with those of the related protein HuB and the destabilizing protein TTP.
|
EXPERIMENTAL PROCEDURES |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
Preparation of RNA Transcripts—The templates for RNA targets (see Figs. 1, 3, and 5) were generated by annealing the appropriate complementary DNA oligomers (generated by the University of Southern California Microchemical Core Facility) and ligating them into an SphI-cleaved pGEM-derived vector (pEP40) (52). Plasmids were linearized with AccI (which cuts just downstream of the RNA targets) and used for in vitro transcription as described (49). All in vitro transcribed RNAs were gel-purified before use. 5'-Biotinylated RNAs for Biacore experiments were ordered from Dharmacon Inc. (Lafayette, CO).
|
|
|
Gel Shift Analysis—Binding reactions were performed as described previously (49). Briefly, binding reactions were carried out in binding buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.5% Triton X-100) containing 1–2 fmol of [
-33P]-labeled in vitro transcribed RNA, to which were added 0.25 mg/ml bovine serum albumin, 1 mM dithiothreitol, and 0.5 mg/ml tRNA. 10 µl binding reactions were equilibrated at room temperature for 1 h before loading on 8% polyacrylamide Tris/glycine gels at 4 °C as described (49). Gel shift bands were analyzed using a Amersham Biosciences PhosphorImager, and ImageQuant software (Amersham Biosciences). KD values were calculated by plotting the logarithm of ratio of complex/free RNA against the logarithm of the protein concentration, which yields log(KD) as the x intercept. Lines were obtained by linear regression.
Surface Plasmon Resonance (Biacore) Analysis—The dynamics of RNA-protein interactions were studied using a Biacore 2000 (Biacore, Inc., Piscataway, NJ). 5'-Biotinylated RNAs were diluted to 1 µM in HBS buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% surfactant P20), heated at 80 °C for 10 min, cooled to room temperature, diluted to 20 fmol/µl in running buffer (10 mM Tris/HCl, pH 8.0, 150 mM NaCl, 5% glycerol, 62.5 µg/ml bovine serum albumin, 125 µg/ml tRNA, 1 mM dithiothreitol, 0.05% surfactant P20 (Biacore, Inc.)), and injected at 10 µl/min. 50–100 response units of RNA were coated per flow channel on each streptavidin-coated sensor chip (SA chip; Biacore Inc., Piscataway, NJ). Triplicate 2-min injections of four different protein concentrations (11, 5.5, 2.75, and 1.4 nM) were injected in random order over the sensor chip surface at 25 °C and a flow rate of 50 µl/min. The surface was regenerated by removing bound proteins with a 60-s 2 M NaCl injection followed by a 60-s binding buffer injection to rinse the needle. Sensorgrams were fit to a simple 1:1 Langmuir interaction model using the nonlinear data analysis program CLAMP with a correction for mass transport (53).
In Vitro RNA Selection—In vitro selection was performed according to Kenan et al. (54). An oligonucleotide library with a complexity of 413 (7 x 107), containing a randomized 13-nucleotide region (N)13 flanked by constant sequences (5'-TGGACGCGTCGACCTGCAGGCATGGG-(N)13GGCATGCAAGCTTCGTGCACCC-3'), a forward primer that contains a T7 promoter and anneals upstream of N13 (5'-AAAAATAATACGACTCACTATAGGGTGCACGAAGCTTGCATGCC-3'), and a reverse primer that can anneal downstream of (N)13 templates (5'-TGGACGCGTCGACCTGCAGGCATGGG-3'), were purchased from Integrated DNA Technologies, Inc. (Coralville, IA). Library oligonucleotides were converted into double-stranded DNA by annealing the forward primer, followed by primer extension using Klenow polymerase (New England Biolabs, Beverly, MA). Following PCR amplification using the forward and reverse constant primers, a fraction of the DNA pool representing the full library was in vitro transcribed using the MEGAshortscript T7 kit (Ambion, Austin, TX). 20 pmol of randomized RNAs and 20 pmol of biotinylated HuD were equilibrated for 30 min at room temperature in 100 µl of gel shift binding buffer. The binding reactions were then incubated with streptavidin-coated paramagnetic beads (Streptavidin Magnasphere; Promega, Madison, WI) for 30 min with gentle tumbling. 300 µl of streptavidin-coated paramagnetic beads were used per binding reaction. Beads were washed twice with 0.1 M NaOH, 0.1 M NaCl and twice with 0.1 M NaCl prior to the binding experiment. Bound RNA was purified by two phenol/chloroform extractions followed by ethanol precipitation and was amplified by reverse transcriptase-PCR using Superscript II, RNase H– reverse transcriptase (Invitrogen), and the constant forward and reverse primers, followed by in vitro transcription to generate RNAs for the next round of selection. A fraction of the cDNAs from each round was digested with HindIII and PstI and cloned into the corresponding sites in vector pEP40. Clones from round 0, 3, 4, and 5 were sequenced.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
FLU) resulted in a very modest loss of affinity. Further shortening of the AU-tract by removal of one or both of the subsequent flanking Us and the following 5'A (constructs 13AU3+5'U, 13AU3+3'U, 13AU3, and 13AU35'A; see Fig. 1) yielded RNA targets with moderately weakened affinities that were all in the same range (KD 200 nM). However, removal of the 3'A (13AU33'A) caused a further loss in affinity and removal of both As (11AU3) produced an RNA target that was bound much more weakly (Fig. 1). Thus, AU-rich tracts as short as 12–13 nucleotides bound relatively tightly, but further shortening to 11 nucleotides resulted in a substantial weakening of the interaction. We cautiously interpreted this to mean that the minimal binding sequence is 12–13 nucleotides long and chose the symmetrical sequence 13AU3 as a representative of the minimal target.
If such a short RNA target truly interacts with full-length HuD, one would expect all RRM domains to contribute to the binding interaction. The contribution of the different domains can be very clearly visualized by studying binding kinetics, because participation of individual domains will affect the association and dissociation rates of the binding reaction. This is exemplified by our previous studies of HuD interacting with a longer AU-rich tract (the 17-mer AU3), in which binding kinetics were analyzed using a surface plasmon resonance-based biosensor (Biacore) (49, 55). In interactions with AU3, RRM1 was the dominant RNA-binding domain and was essential for high affinity RNA binding; its removal dramatically reduced the ability of HuD to bind to AU3 by decreasing the association rate and increasing the dissociation rate for the HuD/RNA complex (28, 49). RRMs 2 and 3 provided an accessory binding function necessary for stable complex formation; removal of either of these RRMs increased the on and off rates of binding (49). On a tract that is too short to accommodate full-length HuD, one would expect the critical domain RRM1 and the adjacent RRM2 to be bound to the RNA (similar to two solved co-crystal structures (50)), whereas RRM3, which is linked to RRM2 by a flexible hinge, might not contact the AU-rich tract and thus might not stabilize the RNA/protein interaction. To determine whether RRM3 contributes to the binding of 13AU3, we compared the kinetics of binding of full-length HuD to 13AU3 with that of HuD lacking RRM3 (HuDRRM3) to the same RNA target. Removal of RRM3 destabilized the interaction of HuD with 13AU3 (Fig. 2), resulting in a 10-fold increase in the dissociation rate (kd). Thus, just as with the longer AU-rich tract, RRM3 is required for stable complex formation with 13AU3, supporting the notion that all HuD RRMs contact this 13-nucleotide minimal AU-tract.
|
The increased affinity of HuD for AU-rich tracts only slightly longer than 13 nucleotides is likely caused by the presence of additional overlapping binding sites, potentially facilitating protein recruitment. To examine by how much the AU-rich tract would need to be extended to fully bind two HuD molecules, we used gel shifts to monitor the recruitment of HuD molecules to AU-rich tracts extended by a single AUUU repeat at a time (Fig. 3). As we had observed previously (49), a small fraction of the 17-mer AU3 can recruit a second HuD molecule at high protein concentrations. However, with each AUUU extension, the recruitment of the second HuD molecule is facilitated, until all RNAs are occupied by two HuD proteins on an RNA target (AU7) containing 7 AUUU repeats (Fig. 3). The full bimolecular occupation of AU7 suggests that two individual binding sites are present on this target RNA. Indeed, AU7 is the smallest UU(AUUU)nAUU variant that carries two complete non-overlapping AUUUAUUUAUUUA (13AU3) sequences (Fig. 3B). Because the sequence consists of a repeated motif, it contains three overlapping binding sites, as well as the two non-overlapping ones. The fact that very little singly occupied AU7 RNA is seen, even at low HuD concentrations, suggests that the non-overlapping sites are preferentially occupied and hints at possible cooperativity during binding.
The Positions of As in a 13-nt AU-rich Tract Influence HuD Binding—When comparing the ability of HuD to bind to the different AU-rich sequences, we made an intriguing observation: the affinity of HuD for 13-mer 13AU3 was markedly stronger than that for the 13-mer AU2 (Fig. 1). The difference between these two tracts is the position and the number of As; 13AU3 contains four As separated by three Us each, two flanking As and two internal ones. In contrast, AU2 contains only three As, all of them internal. To further examine the role of the position of As in the AU-tract, we aligned the (AUUU)n-tracts with a binding consensus proposed on the basis of recently solved co-crystal structures of a HuD RRM1+2 fragment with two different AU-rich sequences (50)(Fig. 4). The co-crystal structures show the orientation of the polypeptide to be antiparallel to the RNA; the N-terminal RRM1 contacts 3' sequences whereas RRM2 contacts 5' sequences. Alignment of the AU-rich tracts with the proposed consensus binding sequence showed that only one of four possible registers of (AUUU)n sequences matches the consensus. All three of the other possible registers resulted in one or more As in unfavorable positions that allow only U or U/C. Thus, to be bound by RRM1+2, an eight-nucleotide (AUUU)n-type target sequence must conform to the sequence AUUUAUUU. If RRM3 is to bind as well, the tract must be extended at its 5' end. Alignment of 13AU3 shows a full AUUU-tract available for interaction with RRM3 (Fig. 4). In contrast, alignment of AU2 with the consensus shows that to prevent As at unfavorable positions, the sequence has to be shifted, causing a loss of nucleotides that appear to be required to contact either RRM3 or RRM1 (Fig. 4). The 11-mer 11AU3 encompasses the complete consensus sequence as determined from the co-crystal structures plus three additional 5'Us. However, it is not sufficient for tight binding of full-length HuD. In particular, the 3'A appears to be important, as evidenced by the weaker binding of 13AU33'A compared with 13AU3 (Fig. 1). This suggests that at least one additional nucleotide is required 3' of the proposed consensus sequence. The co-crystal structures did not show any such contacts, but the HuD fragment used for crystallization lacked the N-terminal 34 amino acids immediately upstream of RRM1. This region of the protein might make contacts to such a residue.
|
poly(U) RNA Is a Preferred Binding Target for HuD—The importance of the register of the AUUU-tract for binding suggested that the As present in the prototype mRNA instability element, UUAUUUAUU (9, 10), might play an important role in alignment of the protein and/or control of the binding affinity. To further study the role of the interspersed As, we replaced the flanking or middle As of 13AU3 by Us. This markedly increased the binding affinity of HuD (Fig. 5). When all the As in 13AU3 were replaced by Us (13U), HuD was able to bind to the target sequence even at a very low protein concentrations (Fig. 5). The presence of a second band at high HuD concentrations in gel shifts of 13U with HuD suggests that the structure of full-length HuD bound to poly(U) might be more compact than that of the protein bound to a mixed U-rich sequence.
To explore the role of the individual RRMs in poly(U) binding, we examined the binding of each RRM to 13U and 13AU3 (Fig. 6). As had been observed previously (49), binding by individual RRMs to AU-rich RNA was very weak. RRM1 and RRM2 showed binding at 5 µM, evidenced by a smear, and very weak binding by RRM3 was also observed (a very slight band is seen, though the probe band does not appear to be diminished). Binding of all three individual RRMs to 13U was stronger; RRM1 showed binding at 100 nM and shifted almost all the probe RNA at 5 µM. RRM2 also exhibited an increase in affinity, although it was much more modest. The most dramatic increase in binding was observed for RRM3, which barely bound 13AU3 but bound to 13U at a protein concentration of 100 nM, showing a very clear band. The difference in the types of bands produced by the various RRMs (diffuse smear or tight band) is reproducible and suggests that something about the nature of the complexes formed between the RNA and the various RRMs is different. The preference of the HuD RRMs for poly(U) is even more pronounced when a 32U-tract is used (Fig. 6, bottom). Because all of the RNA tracts are longer than the expected binding site for a single RRM, it is likely that more than one RRM is bound to each RNA. This might partially account for the smearing and for the increased height of the shifted bands at higher protein concentrations. From the increased affinity of all three RRMs for 13U compared with 13AU3, we conclude that the preference of HuD for poly(U) is reflected in the binding preferences of the individual RRMs.
|
In Vitro Selection Confirms a Preference for poly(U)—To confirm the surprising preference of HuD for U-rich tracts, we carried out an in vitro selection experiment using an RNA pool containing 13-nt randomized tracts. In vitro selection had previously been carried out for HuB and HuC (19, 26, 56). Three rounds of binding selection with HuB, using an RNA carrying a randomized region of 25 nucleotides, had yielded predominantly RNAs that contained U3–5-tracts (19) and binding selection using a 3'UTR library had selected RNAs that contained the sequence Pu(U/A)U3AU3(U/A)Pu (where Pu indicates a purine) (56). A similar iterative selection of HuC-binding RNAs indicated that HuC binds to RNAs carrying the sequence Pu(U)2–5 (Pu)1–2(U)2–5Pu (26).
To facilitate in vitro selection, we generated a HuD variant carrying a biotinylatable N-terminal tag and co-expressed the protein with biotin ligase in bacteria. This resulted in very efficient HuD biotinylation, allowing direct capture of the protein and bound RNA on streptavidin-coated paramagnetic beads. For the first round of selection, biotinylated HuD was incubated with sufficient randomized RNA to represent each individual RNA 1 x 106 times. After four such rounds of in vitro selection, RNAs were reverse transcribed, cloned, and subjected to sequence analysis (Fig. 7). The majority of selected RNAs were U-rich, consisting of stretches of U residues interrupted by other nucleotides. These U-rich tracts were characteristically two to four nt in length separated by one or two non-U residues and are similar to sequences found in the 3'UTRs of oncoprotein and cytokine mRNAs. Two such U-rich RNAs, a non-U-rich RNA that had also been selected and an RNA with scattered single Us (sequences 2, 5, 7, and 10; see Fig. 7, A and B), were tested for their ability to bind to HuD, using gelshift analysis. The two U-rich RNAs bound HuD with affinities comparable with that of 13AU3, whereas the scattered U and non-U-rich targets bound very poorly. Possibly, the latter targets are co-purified through RNA-RNA interactions of HuD-bound U-tracts. To examine the effects of further selection, a fifth round of selection was carried out. The increase in U content seen in the initial rounds (U content increased from 27% in the starting pool to 33 and 52% in RNAs from rounds 3 and 4, respectively) appeared to plateau (52% Us in round 5). However, the distribution of Us within the sequence showed continued progression to longer tracts, exhibiting an increase in U-tract length from one-two Us in the starting pool to maximally three, four, and seven Us in rounds 3, 4 and 5, respectively (Fig. 7, C and D). Alignment of round 5 sequences did not show a clear sequence preference, merely a preference for U-tracts interspersed with G, A, or C. The in vitro selection experiments thus support the preference of HuD for poly(U)-tracts but do not indicate an obvious preference for the interspersed non-U residues. This suggests that U-richness is the prime source of HuD affinity.
|
Comparison of HuD Sequence Specificity to That of Other AU-binding Proteins—To explore whether the strong affinity of HuD for poly(U) is unique to this protein, we compared the ability of two other AU-rich-binding proteins, HuB and TTP, to bind to AU-rich versus poly(U)-containing RNAs. HuB is highly homologous to HuD, showing 90% amino acid identity. As mentioned above, its RNA-binding preference had been suggested to be short U-rich tracts separated by purines (19). To our knowledge, the affinity of HuB for poly(U) had never been measured. Like HuD, HuB is thought to prevent mRNA decay by occluding AU-rich sequences. In contrast, TTP (also known as Nup475/TIS11) is an mRNA destabilizing AU-rich-binding protein and appears to promote mRNA degradation by recruiting labile AU-rich mRNAs to the exosome (17). In vitro selection of randomized RNA with recombinant TTP had yielded RNAs very similar to the prototype mRNA instability element, UUAUUUA(U/A)(U/A) (57). Comparison of the full-length TTP protein to its isolated RNA-binding domain (a 77-amino acid region containing the tandem TTP zinc fingers) has indicated that the RNA-binding ability and indeed the ability to destabilize tumor necrosis factor mRNA are maintained in the isolated RNA-binding domain (58). poly(U) can be used to deplete TTP from cellular extracts, thereby removing its mRNA destabilizing activity, suggesting this protein can bind well to poly(U) (17); however, competition experiments have suggested that poly(U) is not a strong target for TTP (59). We therefore compared the ability of HuB and the RNA-binding domain of TTP to bind to 13AU3 versus 13U using gel shift analysis (Fig. 8). HuB showed binding properties similar to those of HuD, exhibiting a marked preference for poly(U). In contrast, the TTP fragment bound well to 13AU3 but not at all to 13U at the concentrations tested. Thus, we conclude that a high affinity for poly(U) is not a general characteristic of AU-binding proteins but may be a general characteristic of proteins of the Hu family.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
13 nts, a variety of RNAs carrying non-overlapping targets separated by different lengths of non-AU-rich sequences can now be tested to assess cooperativity. These studies will be important to carry out, as cooperativity of ARE-binding proteins can importantly affect RNA target occupancy; on naturally occurring heterogeneous AREs, strong cooperativity could result in homopolymeric coating of a long tract of mixed sequence by a given protein. Differences in HuD affinity for the 13-mer AU2 and 13AU3, which both consist of the same motif (triple U-tracts separated by As) but differ in the register of the sequence presented to the protein (Fig. 1), indicate that HuD binding depends not only on AU-tract length but also upon the exact position of the A and U nucleotides. This suggested that the positions of As within the binding tract might have important effects on binding. To further analyze this, we aligned these RNAs with a binding consensus that had been proposed on the basis of two HuD/RNA co-crystal structures (Fig. 4). Comparison of this alignment with RNAs that bound well (Fig. 1) suggested that RRM1 requires a triple U region to make optimal contacts. Because RRM1 is most critical for RNA binding (removal of this RRM from HuD strongly reduces binding affinity (49)), one would expect a triple U-tract to be a prominent requirement of a HuD RNA target. Because the protein binds in antiparallel orientation to the RNA (based on the co-crystals (50)), positioning of the triple U-tract near the 3' end of an AU-rich sequence leaves the most sequence for interacting with RRMs 2 and 3. Thus, short tracts in which the UUU sequence is away from the 3' end (e.g. AU2) cannot bind HuD well, because insufficient contacts remain for the C-terminal end of the protein.
The importance of the positions of As within the AU-rich tract prompted us to examine the effects of the removal of As. In vitro selection experiments with the highly conserved Hu proteins HuB and HuC had indicated a preference for U-tracts separated by purines (19, 26, 56). Therefore, we were somewhat surprised to observe an increased affinity of HuD for the RNA upon mutation of As to Us (Fig. 5). A 13-nucleotide tract consisting uniquely of Us (13U) was a much better target than 13AU3. This suggests that the presence of As in destabilizing elements might favor the recruitment of other proteins and/or prevent HuD from binding too tightly to AREs. We also observed a second shifted band of 13U at HuD concentration of 50 nM and higher, suggesting that HuD bound to poly(U) makes a more compact structure. This could occur if the sequence composition of the RNA affects its position on the RRMs. In both co-crystal structures of HuD RRM1+2 with AU-rich RNA, the RNA shows a two-nucleotide kink consisting of a flipped out and stacked A and U (50). The six-amine group of the A makes a hydrogen bond to Gln-48. It is conceivable that replacement of A by U, resulting in loss of the contact to Gln-48, causes this stack and kink to disappear. The bases might then shift, allowing the RNA to take a much more direct path across the RRMs. The hydrogen bond with Gln-48 is unlikely to be critical, as evidenced by the affinity of HuD for poly(U). Indeed, preliminary molecular modeling suggests that the A or both of the flipped out residues can be removed without distortion to the rest of the RNA/protein interface.2 This would result in poly(U) contacting RRM1 and shortens the RRM1-interacting RNA tract by one or two nucleotides. A shorter contact region of HuD on poly(U) would mean that a tract of 13 Us would contain one or more additional overlapping binding sites, which would facilitate protein recruitment and thus increase the affinity.
To further examine why the affinity of HuD for 13U is higher than for 13AU3, we analyzed the sequence preference of the individual RRMs (Fig. 6). We conclude that 13U likely presents improved protein/RNA contacts to all RRMs, because each individual RRM shows a preference for 13U. RRM3 shows the most pronounced increase in affinity for 13U compared with 13AU3 and thus may play a more prominent role in RNA binding when a U-tract lies at the 5' end of the target sequence. RRM1 and RRM3 both show binding at a protein concentration of 100 nM, whereas binding by RRM2 is much weaker. Thus, RRM1 and RRM3 might play more dominant roles in interacting with poly(U). The U-richness of the binding tract might therefore be most important in the outmost regions of the bound site (which contact RRM1 and RRM3), resulting in better tolerance for non-U residues in the middle of the target sequence. In accordance with this, mutation of the flanking As to U in 13AU3 increased the affinity more than mutation of internal As (Fig. 5). An affinity of RRM3 for poly(U) appears to contradict previous reports of poly(A) binding by RRM3 (26, 27). However, this poly(A) binding activity is limited to tracts 
150 nucleotides in length (26, 27). In agreement with these observations, we find that full-length HuD and all individual RRMs bind poorly to poly(A)-tracts of 13–32 nucleotides.3 A role for long poly(A) tails in recruiting HuD to RNA is supported by in vitro decay analyses of GAP-43 mRNA, in which stabilization by HuD was dependent on poly(A) tails of at least 150 As (60). All observations made to date suggest that the contacts made between Hu proteins and long poly(A) tails are of a different nature than Hu/ARE contacts and are likely related to the ultrastructure of long poly(A)-tracts (27), which have been proposed to form double helices in vivo (61). Such higher order structures may present interaction sites for RRM3 and/or other parts of Hu proteins distinct from the conventional RNA-binding surface. Elucidating the mechanism by which RRM3 binds to long poly(A) tails would be of great interest.
To confirm the preference of HuD for poly(U), we performed an in vitro selection experiment using RNA containing a randomized tract of 13 nucleotides. Based on the studies described above, and the limited size of the library (a complexity of (4)13), we expected a rapid enrichment for poly(U). Such an enrichment did occur, but relatively slowly, and even after five rounds of binding selection, the longest poly(U)-tract was (U)7. During this fifth round, we began seeing occasional loss of nucleotides from the (N)13-tract. These observations point to a counterselective pressure preventing the accumulation of poly(U). The most likely factor that could prevent poly(U) accumulation is the interpretation of U-tracts as termination signals by T7 polymerase (62, 63). Based on our observed preference of HuD for poly(U), we speculate that a selection against poly(U) may have influenced the outcome of in vitro selection analyses of HuB and HuC, as well. Thus, we surmise that though our selection experiment supports a preference of HuD for poly(U), it is unable to yield the most tightly bound RNA, 13U. Our observations indicate that the difficulty in selecting consecutive Us should be taken into account when designing in vitro selection experiments and that U-rich RNAs should perhaps be designed and tested separately.
Although mammalian RNA polymerase III terminates at U-tracts, polymerase II does not (64), so that poly(U) in the 3'UTR of mRNAs is not expected to cause transcription elongation problems. Indeed, a search of the 3'UTR data base (65) indicates that many mRNAs with long U-tracts exist. One remarkable example is SEC22L1, a vesicle-trafficking protein with a short 3'UTR that contains a (U)22-tract (66). The significance of many of these U-tracts remains unknown and may be influenced by sequence context. In a variety of proto-oncogene and other growth-regulatory mRNAs, poly(U)-tracts in the 3'UTR lie in or near AREs (8, 11). For example, human FOS mRNA has a (U)5A(U)4A(U)7-tract, MYCN mRNA contains a (U)10-tract and several (U)5-tracts, and the VEGF transcript has several (U)6–8-tracts in its 3'UTR. Hu proteins have been demonstrated to be able to bind to these proto-oncogene UTRs in vitro and in cells (22, 28, 34–37, 40, 45, 46, 59, 68). Based on the sequence composition, under conditions of limiting protein concentration, one might expect Hu proteins to occupy the U-tracts and TTP or other proteins to occupy tracts containing more interspersed As. This idea is supported by competitive binding experiments with HuR and TTP (59). In intact cells, actual occupancy will depend on the intracellular location and level of the various proteins and other factors such as protein/protein interactions and cooperative binding.
Our analysis of HuB and TTP showed that the high affinity for poly(U), also seen with HuB, is not a general feature of ARE-binding proteins (Fig. 8). The lack of affinity of TTP for poly(U) meshes with the observed inability of a (U)32-tract to confer instability on a heterologous 
-globin RNA (9), in contrast to UU(AUUU)nAUU sequences, which did accelerate mRNA decay. It will be important to examine the sequence preferences of all AU-rich-binding proteins suspected of involvement in mRNA (de)stabilization, to shed further light on the role of U-tracts in regulating mRNA stability. In addition, it would be useful to study a collection of AREs containing defined mixes of AU-rich and poly(U) sequences. Use of designed AREs with simplified sequence tracts (such as combinations of canonical UU(AUUU)nAUU and poly(U)-tracts) would help circumvent problems of overlapping binding sites and sequence heterogeneity. In vitro studies would help elucidate cooperativity and sequential occupation of sequences and would yield detailed mechanistic insight that could be applied, among others, to the design of small molecule inhibitors. Such inhibitors might be of clinical importance in situations in which improper mRNA stabilization leads to overexpression of labile mRNAs, such as tumor necrosis factor overexpression in inflammatory disease (1, 5) and proto-oncogene overexpression in cancer cells inappropriately expressing Hu proteins (34, 35, 47, 68). Recent experiments in which TTP was overexpressed demonstrate that the modulation of mRNA decay can be used to suppress oncogenesis (67). This observation emphasizes the importance of understanding the interactions of proteins with the AREs of labile mRNAs, not only at a functional level but at the detailed mechanistic level, as well.
|
FOOTNOTES |
|---|
¶ To whom correspondence should be addressed. Tel.: 323-865-0655; Fax: 323-865-0158; E-mail: ilaird{at}usc.edu.
1 The abbreviations used are: ARE, AU-rich element; RRM, RNA recognition motif; TTP, tristetraprolin; UTR, untranslated region; nt, nucleotide.
2 Dr. Ian Haworth, personal communication.
3 S. Park-Lee and I. A. Laird-Offringa, unpublished results.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONEXPERIMENTAL PROCEDURESRESULTSDISCUSSIONREFERENCES |
|---|
