Contributions of the individual domains in human La protein to its RNA 3'-end binding activity.

The autoantigen La regulates the maturation of RNA polymerase III transcripts by binding to their poly(U) termination signal. The modular protein harbors a N-terminal RNA recognition motif (RRM), RRM1, and in the C-terminal domain, a second, atypical RRM2, in addition to a phosphorylation site, and a putative nucleotide binding site. This study presents a detailed investigation into the RNA 3'-end binding properties of La by using binding titration and competition assays with subsequent gel mobility shift analysis. Two truncation mutants containing one (La-RRM1) or both (La-RRM1-RRM2) RNA-binding domains were constructed, overexpressed, and purified. A K(d) value of 25 +/- 10 nm for La binding to a nonameric RNA ligand with the oligouridylate recognition sequence was obtained, discriminating with a specificity ratio of approximately 100 for this probe over a RNA ligand with a 3'-poly(A) stretch. The N-terminal La-RRM1 region was identified as the major contributor of these properties to La, manifested in a 5-fold lower K(d) of 5 +/- 3 nm and a slightly increased specificity ratio of 120 for the RNA ligand. The atypical RRM2 in the C-terminal domain of La has an unprecedented negative effect on 3'-end RNA recognition, as indicated by a higher K(d) value of 90 +/- 10 nm for the La-RRM1-RRM2 mutant but comparable specificity. Thus the C-terminal regions beyond RRM2 positively modulate the RNA binding affinity of La. Negative regulation, however, occurs through Ser(366) phosphorylation decreasing the binding affinity by 2-fold. ATP had no influence on RNA complex formation. The functional implications of these findings for the mechanism of action of La are discussed.

Human La protein (lupus antigen or La/SS-B) was originally identified as a major target of the autoimmune response in patients suffering from the autoimmune diseases Sjögren's syndrome and systemic lupus erythematosus (1). It is a conserved RNA-binding protein that recognizes specifically 3Ј-oligouridylate stretches (2). Associated with all RNA polymerase III transcripts where the 3Ј-UUU-OH sequence is added as the transcription termination signal, La is thought to play a central role in the metabolism of these RNAs by acting as a molecular chaperone that stabilizes and/or structures them for further processing (3). In HeLa cell extracts, the 5Ј-and 3Ј-end processing of tRNA precursors is influenced by La (4), and in vitro, La activity has been suggested in RNA polymerase termination, recycling, and initiation reactions (5)(6)(7)(8). Furthermore, a Walker A nucleotide-binding motif in La was postulated to confer ATP-dependent RNA/DNA and RNA/RNA helicase activity (9,10), but recently, the binding of La to double-stranded nucleic acids was shown to be independent of this site, and La does not appear to unwind double-stranded RNA (11). The Walker A motif has instead been implicated in binding to the 5Ј-triphosphate ends of nascent tRNAs, with the affinity negatively modulated by phosphorylation of Ser 366 (4,7). Cytoplasmic functions for La have also been described, including facilitation of translation or replication of viral RNA and stabilization of histone mRNA (12,13).
The 46.8-kDa human La antigen is a modular protein (14) and belongs to the RNA recognition motif (RRM) 1 protein or the ribonucleoprotein (RNP) superfamily. The RRM motif is an approximately 80-amino acid RNA-binding domain that contains several well conserved residues. Some of these residues cluster into two short submotifs, the octameric RNP-1 and the hexameric RNP-2 (15). Several members of the RNP family involved in single-and double-stranded RNA recognition carry multiple copies of the RRM in a tandem arrangement (15).
Protease digestion showed that the La protein can be divided in a N-terminal and a highly charged C-terminal domain connected by a 43-amino acid linker (16). The N terminus is highly conserved and can be subdivided further in a La protein-specific domain (La motif), amino acids 16 -75 (17), and a RNAbinding motif, RRM1, spanning amino acids 112-184 (18). Most of the autoantigenic epitopes are found in this domain of the protein with the immunodominant epitopes mapping to RRM1 (19). In contrast, the C-terminal region of the protein seems to have expanded differentially during evolution. A highly degenerate RNA recognition domain, RRM2, amino acids 230 -294, is present in metazoans but absent in yeast (20,21). This domain is fused to several auxiliary domains involved in RNA binding and functional specialization, such as a dimerization domain localized within residues 294 -348 (22), the putative ATP-binding motif, amino acids 333-340 (23), the phosphorylation site at Ser 366 (4,7), and a nuclear localization signal in the region of amino acids 382-408 (24,25). A bipartite binding model correlating the 5Ј-end binding activity with the C-terminal part and 3Ј-end recognition with the N-terminal region of La has been proposed (26), but to date, no comparative quantitative study dissecting the contributions of the individual domains to either of these RNA binding activities has been undertaken. K d values for general RNA binding of full-length La under various assay conditions with dif-ferent RNA substrates have been reported, and they range from 37 to 5 nM (27,28). Investigations with in vitro translated N-or C-terminal deletion mutants of La and human Y RNA localized the RNA binding activity within the N-terminal domain of the antigen (29). Similarly, the La motif in conjunction with RRM1 and several additional C-terminal residues were determined to form the minimal structure for high affinity interaction with tRNA precursors and viral RNAs (27,28). Moreover, regions within the C-terminal domain of La, distinct from the N terminus, may be critical for recognition of the 3Ј-end (28). In support of this notion, through immunoprecipitation and filter binding assays, La phosphorylation primarily occurring at Ser 366 was reported to impair poly(U) recognition (30,31). Furthermore, the presence of ATP was found to abrogate the 3Ј-UUU-OH affinity (30). A systematic examination of the 3Ј-end binding activity of RRM2, also localized within the C-terminal domain of La, is lacking. It could contribute to RNA binding or serve a different function.
To understand the role of the individual RRMs in La, as well as the C-terminal region outside these motifs, we constructed and purified two deletion mutants containing one or both RNAbinding units. Our assay system was designed to investigate the contributions of the individual protein regions solely to the 3Ј-end binding affinity of La. The two parameters, binding affinity and specificity, that characterize the RNA complex formation were determined by gel mobility shift analysis and compared with the truncated proteins under identical assay conditions. Furthermore, the influence of phosphorylation on the binding affinity of full-length La was studied. The effect of ATP on single-stranded RNA recognition was also probed in the presence of ATP or the nonhydrolyzable ATP mimic ATP-␥-S. We describe how our results relate to the binding model postulated for the activity of La in the pre-tRNA maturation pathway (26).

EXPERIMENTAL PROCEDURES
Materials-A cDNA encoding the human La antigen in the pET-8c expression vector was a generous gift of G. J. M. Pruijn (University of Nijmegen, Nijmegen, The Netherlands). The nonameric RNA sequences 9nt-U, 5Ј-UGCUGUUUU-3Ј, and 9nt-A, 5Ј-CCGAAAAAA-3Ј, were synthesized at the Genzentrum (Ludwig-Maximilian-Universitä t, Munich, Germany). The quantitation of radioactively labeled RNA probes was carried out on a Fuji phosphorus imaging system.
Construction of Plasmid DNA-DNA fragments encoding the truncated proteins La-RRM1 (amino acid 1-202) and La-RRM1-RRM2 (amino acids 10 -300) were derived by polymerase chain reaction. The primers were synthesized on the basis of the published human La sequence (14) with the addition of an NdeI and a EcoRI restriction site and cloned into the pET-22bϩ vector (Novagen). The primers used were 5Ј-AAA AAA ACA TAT GGC TGA AAA TGG TGA TAA TGA AAA GAT GGC T-3Ј and 5Ј-AGA ATT CCT ATT CCA CTT TAT TTT G-3Ј for the La-RRM1 mutant and 5Ј-AAA AAA ACA TAT GGC TGC CCT GG-3Ј and 5Ј-AGA ATT CTC ATT CTT TGT TCC TTA ATT G-3Ј for the La-RRM1-RRM2 mutant. The resulting constructs were confirmed by DNA sequencing and used for recombinant protein production.
Protein Expression and Purification-The full-length and truncated La proteins were overproduced in Escherichia coli BL21 (DE3) plysS (Stratagene). The cells were harvested 3 h after induction, resuspended in 150 mM NaCl, 0.02% NaN 3 , 50 mM Tris-HCl, pH 8.0, 5 mM EDTA, and 10 mg/ml Pefabloc (Roche Molecular Biochemicals), and sonicated. The lysate was cleared by centrifugation for 20 min at 30,000 ϫ g and 4°C to yield the supernatant with the overexpressed soluble protein. This supernatant was filtered through a 2-m filter and loaded onto a Resource Q anion exchange column (Amersham Pharmacia Biotech) on an Ä kta fast protein liquid chromatography system (Amersham Pharmacia Biotech) and eluted with 0.1 mM NaCl, 3 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM dithiothreitol, 25 mM Tris-HCl, pH 8.0 and 9% (v/v) glycerol (buffer A). The flow through fraction was immediately loaded onto a 5-ml Heparin Hitrap column (Amersham Pharmacia Biotech) and eluted over 10 column volumes with a linear gradient from 0.1 to 1 M NaCl in buffer A. The fractions containing the desired protein were pooled, diluted 5-fold, loaded onto a 10-ml poly(U)-Sepharose affinity column (Amersham Pharmacia Biotech), and eluted over 4 column volumes with the same gradient. The fractions containing pure recombinant protein were concentrated in Millipore Ultrafree 5-kDa molecular weight cut-off filter units. Purity was determined by Coomassie Blue-stained SDS-PAGE. The protein concentrations were determined from A 280 by using the calculated extinction coefficients of 40,500 M Ϫ1 cm Ϫ1 for La, 22,100 M Ϫ1 cm Ϫ1 for La-RRM1, and 28,700 M Ϫ1 cm Ϫ1 for La-RRM1-RRM2. The purified proteins were stored at 4°C. Circular Dichroism Measurements-All CD studies were performed on a Jasco J-715 spectrophotometer with a 0.1-cm-pathlength cuvette. The proteins were diluted to 200 g/ml in 0.4 mM Tris/HCl, pH 7.0, and 3 mM NaCl. The measurements were made at room temperature with a 1-nm bandwidth and 0.1-nm step size. Each spectrum is the average of eight recordings.
RNA Binding Assays-The synthetic RNA nonamer 9nt-U was 5Јend-labeled with T4 polynucleotide kinase (New England Biolabs) and [␥-32 P]dATP (Amersham Pharmacia Biotech) and purified on G25 spin columns (Amersham Pharmacia Biotech). For the gel mobility shift assays 300 ng (5000 cpm) of this RNA substrate were titrated with 90 pM to 7 M of the respective protein La, La-RRM1-RRM2, or La-RRM1 in 25 mM Tris-HCl, pH 8.0, 100 mM NaCl, 3 mM MgCl 2 , 0.1 mM EDTA, 0.5 mM dithiothreitol, and 0.5% Nonidet P-40 (buffer B). The influence of a ribonucleotide triphosphate was probed in the presence of 90 M ATP (Sigma) or ATP-␥-S (Sigma), respectively. The binding reactions were incubated on ice for 30 min in a total volume of 23 l. Upon addition of loading dye (40% sucrose, 0.01% (w/v) xylene cyanol), the solutions were immediately resolved on a prerun, precooled 9% native polyacrylamide gel at 4°C in 0.5ϫ TBE buffer (45 mM Tris-HCl, 45 mM boric acid, 0.1 mM EDTA). The gels were run for 1 h at 110 V and then vacuum dried onto Whatman No. 3MM chromatography paper. The dried gels were exposed to x-ray film and quantitated by phosphorus imaging.
Data Analysis-For each titration reaction the bands corresponding to the bound and the unbound probe were quantitated. The apparent equilibrium dissociation constant (K d ) was estimated from nonlinear least square fits of binding data to the Langmuir isotherm (32) using KaleidaGraph 3.5 software, where is the fraction of bound oligonucleotide probe and P is the total protein concentration. Each K d value is the average of at least two replicates.
Competition Assays-Relative dissociation constants for La, La-RRM1, and La-RRM1-RRM2 were determined in a competition assay. The full-length protein (27 nM), La-RRM1 (9 nM), or La-RRM1-RRM2 (270 nM) were incubated for 10 min on ice with varying concentrations of unlabeled 9nt-U or 9nt-A in a total volume of 22 l. After addition of 32 P-labeled 9nt-U (150 nM), the reaction mixtures were incubated for an additional 10 min to reach equilibrium at concentrations at which 10 -80% of the labeled RNA was bound. The reaction mixtures were resolved by electrophoresis as described for the gel mobility shift assays. Phosphorus imaging analysis was used to quantitate the bands corresponding to bound and unbound probe. The competitive binding of a protein to the labeled RNA probe and the unlabeled competitor RNA can be described by Equation 2 (33), where K rel is the ratio of K t , the apparent dissociation constant of the labeled probe, to K C , the apparent dissociation constant of the competitor, and P t , T t , and C t are the concentrations of the protein, the radiolabeled probe, and the competitor probe, respectively. Competition data were fit to Equation 2 by using KaleidaGraph 3.5 software. The errors were based on the statistical fit of the data and represent Ϯ S.D. Each data point was determined in two or three independent experiments.
In Vitro Phosphorylation/Dephosphorylation Assays-For the phosphorylation reactions 2.8 g of the full-length La protein was incubated overnight at 30°C with 500 units of casein kinase II (New England Biolabs) in the presence of 300 M ATP in the buffer supplied by the manufacturer in a total volume of 30 l. Similarly, 2.8 g of La protein were incubated with 5 units of protein phosphatase 1 (New England Biolabs). As a control for the loss of RNA binding activity, the same reactions were also carried out in the absence of those enzymes. Subsequent RNA binding assays were performed as described above.
Isoelectric Focussing-Isoelectric focussing was performed on a Phast system (Amersham Pharmacia Biotech) with Servalyt precotes gels (Serva) over a pH range of 3-10. The gels were electrophoresed at 2.5 mA for 30 min with a 300 VAh limit. The pI values were deduced by using the broad pI calibration kit (Amersham Pharmacia Biotech). Visualization of the proteins was performed by Coomassie Blue staining.

Construction and Purification of Deletion Mutants of Human
La Protein-Early studies suggested that La recognizes 3Јterminal uridylates in RNA (2,34). Because the full-length La protein contains two distinct RRM domains designated RRM1 (amino acids 112-184) and RRM2 (amino acids 230 -294) (Fig.  1A), two deletion mutants were designed, cloned, and overexpressed with the intent to delineate the RNA-binding determinants within the modular protein. The first construct, La-RRM1, was projected to be the minimal unit with specific RNA binding activity. Previous studies had found that neither the isolated La domain (27, 28) nor RRM1 alone (29) dispose of RNA recognition specificity. Furthermore, high affinity RNA binding by RRM1 in conjunction with the La-specific domain was achieved only in the presence of several subsequent Cterminal residues (28). Therefore, La-RRM1 was chosen to span the N-terminal region of La, comprising the La motif, amino acids 16 -111, the RNA-binding domain RRM1, and 17 subsequent amino acids (185-202) (Fig. 1A). These amino acids were added based on secondary structure predictions that revealed an ␣-helical element ending at residue 202. To investigate the additional contributions of RRM2 to RNA affinity, a second mutant comprising amino acids 10 -300, La-RRM1-RRM2, was designed. In this construct the putative RNAbinding region in the C-terminal domain was included.
The full-length La protein and the truncation mutants were all purified in very high yields in three steps by anion exchange, heparin, and poly-uracil affinity chromatography. UVvisible spectra confirmed that the recombinant protein was obtained RNA free. Coomassie Blue staining of SDS-PAGE gels established the purity to be Ͼ95% (Fig. 1B). The identity of the proteins was confirmed by N-terminal sequencing. Circular dichroism spectra verified that all proteins were folded (data not shown).
Characterization of the La Complex with a Single-stranded RNA Substrate-Equilibrium binding titration experiments were performed to determine the apparent dissociation constants K d of full-length La and the two deletion mutants La-RRM1 and La-RRM1-RRM2. Increasing amounts of wild-type or mutant protein were added to the nonameric RNA substrate 9nt-U, which carries a poly-uridylate sequence at the 3Ј terminus to mimic the binding site on tRNA precursors. Binding titrations for all three proteins were carried out under identical conditions.
With increasing protein concentrations, the RNA probe was recruited to slower migrating forms, because of specific complex formation (Fig. 2). Whereas for the La-RRM1 fragment only one slower migrating form was identified (Fig. 2C), for La-RRM1-RRM2 (Fig. 2B) and the full-length La ( Fig. 2A) two distinct bands of retarded mobility were detected, indicating that complexes of different oligomerization states were formed. For quantitation of the fractions of RNA complexed by protein, , all shifted bands were quantitated with a phosphorus imaging device. A plot of the binding data is presented in Fig. 2D. The versus protein concentration data were analyzed by a nonlinear least square fit to the Langmuir equation, affording a K d value of 25 Ϯ 10 nM for the full-length protein. The deletion mutant La-RRM1 binds with a 5-fold higher affinity as witnessed by a K d of 6 Ϯ 3 nM, suggesting that it is a sufficient functional unit for RNA recognition. La-RRM1-RRM2, which contains the second putative RRM domain, binds with an ϳ3-4-fold lower affinity relative to the full-length protein. The dissociation constant of this construct was determined to be 90 Ϯ 10 nM. All K d values reported are apparent dissociation constants, with the assumption that the protein binds as a monomer because the oligomerization state was not determined.
Binding Specificity-To assess the ability of La and its truncation mutants to discriminate for the specific RNA substrate 9nt-U, the stability of a protein complex with the 9nt-U substrate was challenged by titrating increasing amounts of unlabeled competitor into the binding reactions prior to the addition of the specific 9nt-U substrate. Either the 9nt-U substrate or a RNA nonamer containing a hexameric adenylate stretch at the 3Ј-end, 9nt-A, served as the specific or nonspecific competitor, respectively. The competition data were analyzed by a plot of the protein fraction bound to the specific RNA substrate after addition of the competitor RNA versus competitor concentration (Fig. 3). A least squares fit of the data for La and its truncation mutants to Equation 2 yielded K d values for the specific competitor probe 9nt-U in excellent agreement with those determined with the binding titration experiments (Table I). The binding affinity for the nonspecific 9nt-A competitor afforded apparent dissociation constants in the micromolar range. The relative binding affinities of the three proteins paralleled the affinities found with the specific probe, (La-RRM1-RRM2). Thus for all three proteins, the binding specificity is ϳ100-fold (Table I).
Influence of ATP-The basic region of human La contains the sequence 333 GRRFKGKG 340 representing a potential NTPbinding site of the consensus sequence GXXXXGKX Walker A motif found in a number of ATP-binding proteins (23). Inconsistent results have been reported for the RNA binding affinity of La in the presence of ATP. Although no influence of ATP was found in filter binding assays (31), abrogation of RNA binding in the presence of ATP was reported with immunoprecipitated complexes (30). Thus the complex stability in the presence of a 10-fold excess of ATP was investigated under the conditions established for the binding titration experiments described in the previous sections, and the results of these experiments are presented in Fig. 4. The RNA binding affinity of full-length La is maintained at 25 nM ATP. Similar results were obtained with the nonhydrolyzable ATP mimic ATP-␥-S (data not shown). An influence of ATP binding and ATP hydrolysis in mediating the sequence-specific 3Ј-end recognition of RNA can thus be excluded.
Influence of Phosphorylation-The residue Ser 366 of La resides within a conserved casein kinase II phosphorylation site with the consensus sequence (S/T)XX(D/E) (31,35). La phosphorylation was found to induce a large decrease in transcriptional activity (4, 7) that has been attributed to the decreased affinity of the phosphorylated protein for the 5Ј-end in tRNA precursors (4). In an attempt to determine the influence of phosphorylation on the 3Ј-end single-stranded RNA binding affinity, recombinant human La was treated with casein kinase II in the presence of ATP. The resulting phosphorylation products were fractionated by one-dimensional isoelectric focussing in a linear pH gradient (Fig. 5A). The casein kinase treatment  of the La antigen caused all of the bands to shift to more acidic forms with one predominant species identified at pI 6.6 and two minor bands at more acidic positions (Fig. 5A). La protein incubated under identical conditions but in the absence of enzyme served as a control. The recombinant untreated La protein was separated into one predominant product at a pI of ϳ6.7 consistent with the calculated pI of 6.68 and two minor isoforms with pI values ranging from 7.0 -7.4. To test whether the appearance of these isoelectric isoforms was a result of in vivo phosphorylation in E. coli, recombinant La was also treated with protein phosphatase 1 and subsequently resolved on an isoelectric focussing gel. No changes in the gel pattern were detectable, suggesting that the formation of additional bands was not induced by a modification in the phosphorylation structure (data not shown). Alternatively, the phosphatase may not be able to remove all the phosphates from the protein.
For quantitation of the RNA binding properties, the phosphorylated species as well as the untreated protein were subjected to binding titration assays with the radiolabeled RNA substrate and resolved on a native PAGE (Fig. 5B). Analysis of the fit to the binding isotherm reproducibly revealed a K d of ϳ90 Ϯ 7 nM for the phosphorylated species versus 47 Ϯ 2 nM for the untreated protein, accounting for a 2-fold decrease in binding affinity upon phosphorylation (Fig. 5C). The overall elevated K d values are likely to be a result of the extended incubation time at 30°C during the phosphorylation reaction. A severe decrease in binding activity was noted when the La protein was exposed to elevated temperatures for several weeks (data not shown). DISCUSSION One of the intriguing features of the La antigen is its modular nature. The full-length protein is composed of a N-terminal and a C-terminal domain connected by a 43-amino acid spacer that links two RNA-binding motifs. A bipartite binding model linking 3Ј-oligouridylate affinity to the N-terminal region and 5Ј-end RNA affinity to the C-terminal region has been proposed (4). Our assay system was designed to study the 3Ј-end poly(U) recognition specificity and to map the contributions of the individual protein domains to this activity. It was shown that La discriminates by 80-fold for a single-stranded nonameric RNA probe with an oligouridylate tail over RNA with a 3Ј-poly(A) stretch. The high stability of the La-RNA complex is manifested in an apparent dissociation constant (K d ) of 25 Ϯ 10 nM, comparable with the values of 17 nM determined for La binding to HIV-1 TAR RNA and 23 nM obtained for U1 RNA but higher than the K d values of 5 or 6 nM obtained for hY4 RNA and hY1 RNA, respectively (27,28). Because human Y RNAs contain a 5Ј-end triphosphate moiety, the 3-4-fold superior affinity for Y RNAs is likely a result of an additional interaction with the 5Ј-end. As further evidence, an ϳ3-fold diminished affinity of La for tRNA precursors lacking the 5Ј-pppG/A nucleotide was reported (4).
In the gel shift pattern for La, a second band of lower mobility appeared at higher protein concentrations, and protein dimerization could possibly account for the formation of this new species. In support of this interpretation, a dimerization domain had been mapped to the region of amino acids 294 -348 (22). This domain was found to be essential for the function of La in enhancing translation of poliovirus RNA and HIV TAR element containing mRNAs in vitro (22).
Contribution of the N-terminal La-RRM1 Domain to RNA Complex Stability-In our assay system, the La-RRM1 mutant binds with a 5-fold higher affinity to the oligouridylated RNA substrate than to full-length La. The specificity ratio of ϳ120 is comparable with the full-length protein. The results presented herein establish for the first time quantitatively that the Nterminal domain is not only sufficient for the binding of La to RNA, but indeed it confers sequence specificity to the protein.
In previous investigations, the recognition activities of several RRM domains were rarely found to possess any inherent sequence specificity and often required a second RNA-binding domain or other auxiliary domains (18,36). Likewise, RRM1 does not exhibit high affinity RNA binding on its own but only in conjunction with the La motif and additional C-terminal linker residues (27). It has been suggested that the La motif could adopt a RRM fold as well (26). Our analysis carried out with several secondary structure prediction algorithms, as well as the CD spectra, however, revealed predominantly ␣-helical consensus elements in this domain, in disagreement with a RRM topology. Conversely, the fold for the region of amino acids 112-184 was predicted to be consistent with a RRM domain. Thus even though the La motif and the linker residues are likely to participate in RNA contacts, we believe that they do this in a manner distinct from the RRM domain.
In contrast to full-length La, La-RRM1 apparently lacks the ability for higher order complex formation, because only one distinct RNA complex is observed in the gel shifts, consistent with a localization of the dimerization domain in the C-terminal region of the protein. This ability of isolated RRMs to exhibit superior binding to RNA relative to the full-length protein has been previously witnessed. For example, both the Sxl (sex-lethal) protein in Drosophila melanogaster and the poly(A) binding protein in Xenopus exhibit enhanced binding in the absence of the C-terminal residues (37,38). In both cases these regions were postulated to be involved in protein-protein interactions and to interfere with RNA binding of the fulllength protein in the absence of their physiological interaction partners. A possible candidate for an interaction partner of La could be the Ro antigen associated with La in hY RNA ribonucleoprotein particles (39).
Influence of La-RRM2 on RNA Complex Formation-The sequence of human La antigen revealed a second putative RRM in the C-terminal domain of La (21). Secondary structure prediction algorithms identify the elements necessary to maintain the ␤␣␤␤␣␤ fold of the RNP motif. The signature RNP-1 and RNP-2 motifs, which are usually involved in RNA contacts, are highly degenerate, however (Fig. 1C). To investigate the contribution of RRM2 to specific 3Ј-end RNA recognition, another mutant was designed by adding this motif. The results obtained with this construct were unexpected. Addition of RRM2 reduced the RNA binding affinity in the La-RRM1-RRM2 tan- dem unit by about 20-fold compared with the La-RRM1 mutant. Similarly, for a La mutant comprising amino acids 1-328, a diminished RNA binding affinity compared with the fulllength protein had been reported previously but was not quantified (4). La-RRM1-RRM2 maintained a specificity ratio of ϳ100, indicating that single-stranded RNA recognition is generally deteriorated rather than being restricted to 3Ј-end poly(U) recognition.
Several factors may induce the lower affinity of La-RRM1-RRM2. For instance, with a 43-amino acid spacer connecting the RRMs, the two domains could be situated as much as 150 Å apart. Thus lacking any further stabilization by the protein scaffold, the two domains may be unfavorably positioned for RNA binding. Restoration of a 5-fold lower K d value in the full-length La could be interpreted as evidence for this unfavorable orientation of the domains, although comparison of crystal structures of bound and free forms of tandem RRM domains indicate that RNA binding is often coupled with structural organization and rigidification of the interdomain linkers (40 -44). An alternative explanation would thus consider participation of other C-terminal auxiliary domains in the RNA binding affinity of La, such as the highly basic region between amino acid 328 and 344, absent in the mutant.
Features unique to RRM2 could also induce the inferior RNA binding affinity of La-RRM1-RRM2. The binding sites recognized by RRM domains in the complex structures previously characterized usually extend over 2-6 bases (40,45,46). Assuming that the 3Ј-UUU-OH binding site on RNA substrates is already occupied by the interaction with the La-RRM1 unit, RRM2 may have a RNA sequence-or structure-specific prefer-ence not satisfied with the nonamer employed in this study. A possible role for RRM2 could be the 5Ј-end stabilization in tRNA precursors (47), but it could also contribute to the affinity for RNAs lacking the UUU-OH recognition site, including a stem-loop structure in hepatitis B virus RNA and HIV-1 TAR RNA (27,48). Even though the latter binding activity was at least partially localized in RRM1, a contribution from RRM2 cannot be excluded without additional data.
Several RRM⅐RNA complexes have been structurally characterized, and they demonstrate the remarkable versatility of the RRM scaffold, presenting an enormous repertoire of RNA-binding surfaces to single-stranded and stem-loop RNA structures (40,46,49,50). Even though it is not possible to deduce a general RNA recognition code, comparison of the RNP signature regions in the RRM domains of La, Drosophila Sxl, and U1A spliceosomal protein, allows for identification of several features that might explain the poor RNA binding affinity of the La-RRM2 (Fig. 1A). Most strikingly, RRM2 contains Glu residues in the RNP-1 motif at the conserved positions 279 and 286, which are usually occupied by a positively charged Lys or Arg residue involved in RNA backbone interactions and an aromatic Phe/Tyr residue, respectively. In analogy, a negatively charged residue at the position equivalent to 279 also occurs in RRM1 of the apoptotic death signaling protein TIA-1, which does not exhibit any RNA binding activity (51). Furthermore, based on an inspection of the known RRM structures, amino acid types of Tyr 114 in RNP-2 and Phe 155 in RNP-1 are, as part of a solvent-exposed hydrophobic patch, conserved among RNP proteins. These residues engage in aromatic stacking interactions with the RNA bases and presumably attribute largely to the free energy of binding. Small aliphatic Cys or Leu residues replace the equivalent residues in the La-RRM-2. The diminished number of aromatic stacking contacts in the RNAbinding regions could perhaps lead to the low RNA binding affinity in La-RRM2. Conversely, the polar or charged Asn or Lys residue at position 116 in RNP-2, involved in a base specific contact in the other proteins, is replaced by the sequenceneutral Leu residue in La-RRM2. From a combination of all these interactions, it seems that La-RRM1 contains a reasonably well conserved RNP domain, whereas the diminished RNA binding properties of RRM2 are likely a result of its divergent amino acid sequence.
A difference in the nature of the complexes formed by La-RRM1-RRM2 versus the La-RRM1 mutant is observed in the gel retardation assays. In contrast to La-RRM1, the gel shift pattern reveals two distinct bands, indicating that the ability to form a second defined complex as observed in the full-length protein is restored through the addition of RRM2. Proteinprotein interactions could occur through RRM2, but the dimerization domain in the region of residues 294 -348 may be required for high affinity interaction. As further evidence, the RRM domain of the spliceosomal protein U2BЉ binds to its cognate RNA only in the presence of U2AЈ. This protein-protein interaction occurs through the RRM domain (49,52).
Influence of the Other C-terminal Domains in 3Ј-End Recognition-Several lines of evidence in this study suggest that apart from RRM2 other domains in the C-terminal region of La also influence the 3Ј-end affinity of the protein. The increased complex stability observed with full-length protein over the La-RRM1-RRM2 variant indirectly indicates a contribution of the regions further downstream to RRM2. Phosphorylation of La Ser 366 had been suggested to represent a control mechanism in the 5Ј-end processing events of tRNA precursors, because it abrogates the protection activity for these substrates. Binding experiments in this study showed that the 3Ј-end interaction is also affected through phosphorylation, indicated by a 2-fold decrease in binding activity compared with the unphosphorylated protein. A similar decrease was reported under nonequilibrium conditions in a filter binding assay (31). Whether regulation through modification of Ser 366 occurs strictly because of a change in charge distribution or whether conformational changes are involved remains to be determined. In contrast to another report (30), no influence of ATP or ATP-␥-S on 3Ј-end RNA recognition was detected. Indeed, recent results indicate that the Walker A motif is more likely to be involved in recognition of the 5Ј-ppp terminus of nascent tRNA transcripts (4). Such a role would be consistent with our data because it does not interfere with the 3Ј-UUU-OH binding activity of the protein and is not detected in our assay.
Functional Implications-By binding to the UUU-OH 3Ј-end motif of RNA polymerase III synthesized RNA, La stabilizes the precursor forms of these transcripts (26). Further processing of nascent RNA involves 3Ј-and 5Ј-end metabolism (53). Increasing evidence becomes available that La is integrating activities controlling both events, which had been roughly mapped to the N-and C-terminal regions of the protein, respectively (4,47). This study provides a systematic quantitative analysis of the affinity and specificity of La variants for the 3Ј-poly(U) tail and shows that both protein regions contribute either positively or negatively to this activity. The N-terminal domain comprising amino acids 1-200 was shown to be the functional unit for recognition of the 3Ј-poly(U) transcription termination signal. The superior affinity and specificity of this domain for the oligouridylated RNA, combined with the fact that it is the most highly conserved part of the protein, suggests that this part of the protein has evolved for 3Ј-end bind-ing. The lower K d value for RNA binding upon addition of the RRM2 points to a different target for this domain, consistent with the proposal that RRM2 may recognize a distinct determinant at the 5Ј-end of the nascent RNA polymerase III RNA (47). Furthermore, it can be concluded from our results that in contrast to the bipartite binding hypothesis, the C-terminal auxiliary domains are involved in both 3Ј-and 5Ј-end processing.