Crystal structure and possible dimerization of the single RRM of human PABPN1

In most eukaryotes, poly(A) tails are attached to the 3′-ends of mRNAs. They influence almost all aspects of mRNA metabolism: transport from nuclear to cytoplasm, translation, and turnover. To mediate its biological functions, poly(A) tails are covered in sequence-specific manner by two types of conserved poly(A) binding proteins, PABPC in the cytoplasm and PABPN1 in the nucleus. PABPC (Pab1p in Saccharomyces cerevisiae1) is present in all eukaryotes. It plays important roles in the initiation of translation2 and in mRNA decay.3 PABPC contains four copies of RRMs [the RNA recognition motif, also known as RBD (RNA-binding domain)] in its N-terminal portion. RRM1 and RRM2 are most similar to RRM3 and RRM4, respectively. The first two RRMs are mainly responsible for specific binding to poly(A). In the crystal structure of an human PABPC (RRM1/2) bound to poly(A), the β-sheet surfaces form a platform that binds an extended conformation of the oligonucleotide. Each RRM interacts with about four nucleotides.4 PABPN1 (also named PABP2, PABII) is the only nuclear poly(A) binding protein identified in mammalian so far. It can improve synthesis of the poly(A) tails by stimulation of PAP [poly(A) polymerase] and control poly(A) tail length. Under native conditions, there was equilibrium between monomers, dimmers, and higher oligomers of the protein.5 Electron microscopy of PABPN1-poly(A) complexes suggests that oligomerization of the protein is functionally significant. The complexes can form spherical particles which are in equilibrium with filamentous complexes. This dynamic structure may function as a molecular ruler to determine the length of poly(A) tail.6 Although PABPN1 binds poly(A) with high affinity and specificity as PABPC, it just contains a high-conserved single RRM, which separates the argine-rich C-terminal domain from the acidic N-terminal domain which is essential for the stimulation of PAP. Both argine-rich domain and the RRM are required for poly(A) binding, but the latter contributes the high specificity.7 Intriguingly, despite those sequences which are identified as RRM family, the RRM of PABPN1 has little sequence similarity with any of the RRMs of known structure, including the RRMs of PABPC. Here, crystals of the single RRM of human PABPN1 in two space groups have grown from the same conditions, unusually, in the same droplet. We determined the three-dimensional structures of the two crystals independently. The first structure was determined by single wavelength anomalous diffraction (SAD) method at 2.0 Å resolution with R factor = 17.4% (Rfree 20.7%) in the cubic crystal space group I23, with unit cell dimensions of a = b = c = 92.35 Å. The second structure was determined by molecular replacement at a resolution of 2.82 Å with R factor = 22.6% (Rfree 29.0%) in the trigonal space group P31, with unit cell dimensions of a = b = 59.34 Å, c = 80.60 Å. The overall structure is very similar in both crystal forms, adopting a fold similar to the canonical RRM topology. However, two apparent features of PABPN1 RRM are distinguished from other RRMs: the loop3 deviation and the observed dimerization in the crystal structure. Cloning, expression, and purification. The DNA coding sequence for the RRM of human PABPN1 (NCBI entry code NP_004634), corresponding to amino acids 167–254, was amplified from human brain cDNA library (BD biosciences) by PCR using the following oligonucleotide primers containing artificial NdeI and XhoI sites(in bold): 5′-CTGTGCATATGGAGGCTGATGCCCGTTC-3′ and 5′-GACCTCGAGGCCTGGTCTGTTGGTTCGTTTTG-3′. The PCR product was digested with NdeI and XhoI and was ligated into the NdeI and XhoI sites of the bacterial expression vector pET22b (incorporating a C-terminal polyhistidine tag). The cloning junctions were confirmed by DNA sequencing. This recombinant plasmid was transformed into Escherichia coli strain BL21 (DE3). Cells were grown at 310 K in Luria–Bertani medium containing 50 μg/mL ampicillin and were harvested and sonicated in 20 mM Tris-HCl buffer pH 8.0 containing 300 mM NaCl. The lysate was clarified by centrifugation at 15,000g for 30 min. The soluble fraction was applied to nickel-chelating resin (Amersham Biosciences) preequilibrated with equilibration buffer [20 mM Tris pH 8.0, 300 mM NaCl]. The resin was washed with wash buffer [20 mM Tris-HCl pH 8.0, 50 mM imidazole, 300 mM NaCl] and the target protein was eluted with elution buffer [20 mM Tris-HCl pH 8.0, 200 mM imidazole, 300 mM NaCl]. The eluate was buffer exchanged with 20 mM Tris-HCl pH 7.9, 150 mM NaCl, and loaded onto a Superdex 75 column (Amersham Biosciences) preequilibrated with the same buffer. The flow-through fraction, which contained the target protein, was buffer exchanged with crystallization buffer [50 mM Na2HPO4-KH2PO4 pH 5.6, 100 mM NaCl] and concentrated for crystallization assays to 25 mg/mL by centrifugal ultrafiltration (Millipore). The purity of protein was then analyzed on SDS–PAGE (better than 95% purity) and was judged to be suitable for crystallization. To prepare the Se-Met-derivatized protein, the Leu230M mutant was constructed by replacing the TTG codon for Leu230 with ATG. The construct was transformed into Escherichia coli strain B834 (Novagen) and expressed using M9 medium supplemented with SeMet and the amino acids (leucine, isoleucine, valine, phenylalanine, lysine, and threonine). The purification protocols are the same as for the wild-type protein described earlier. Crystallization of the wild-type protein was performed at 283 K using the hanging-drop vapor diffusion method and initial trials were carried out by using screening Crystal Screen I and Crystal Screen II reagent kits (Hampton Research). The best crystals were grown by mixing 2.0 μL of protein solution and an equal volume of reservoir solution containing 0.1M HEPES (pH 7.5), 4.2M sodium chloride. Quite unexpectedly, crystals with two different geometric shapes were observed in the same droplet. The spindly and cubic crystals grew to their full dimensions after 1 month and 2 months, respectively. Crystals of the selenomethionine substituted Leu300M mutant were obtained as above but using a reservoir solution containing 0.1M HEPES (pH 7.5), 3.5M sodium chloride. Before data collection, all crystals were transferred into a cryoprotectant solution consisting of 0.1M HEPES (pH 7.5), 3.5M sodium chloride, 10% glycerol. One native data set to 2.0 Å resolution was obtained from a wild-type protein crystal with cubic shape at 100 K. The data were collected on an in-house MAR Research dtb345 imaging-plate detector, using the copper Kα radiation from a Rigaku Micro007 rotating-anode generator. A selenomethionyl single-wavelength anomalous diffraction (SAD) data set to 2.1 Å resolution from the selenomethionine substituted Leu230M crystal was collected on beamline 3W1A of the BSRF (Beijing Synchrotron Radiation Facility) at a wavelength of 0.9792 Å at 100 K and processed using HKL2000.8 The second native data set to 2.82 Å resolution from the spindly crystal were also collected on beamline 3W1A of the BSRF. Both native data were integrated and scaled using Automar 1.6 [marresearch G.m.b.H.]. Because an initial attempt to solve the structure by molecular replacement using the known structures of RRMs as the search models did not give an appropriate solution, the cubic crystal structure was determined by SAD using SeMet-labelled data. The cubic crystal and the selenomethionine substituted Leu230M crystal are in space group I23 with similar cell parameters. They contain one molecule in the asymmetric unit corresponding to a solvent content of ∼59%. SAD data from selenomethionine substituted Leu230M crystals were used to find the Se atom position; phases were calculated and refined by SOLVE.9 The resulting phases were improved by solvent flattening using the program RESOLVE.10 An initial model was constructed into the 20–2.3 Å cut-off electron density map by using the program ARP/wARP,11 which allowed automatic building of 75 residues. The structure was refined against the first native data to 2.0 Å resolution using REFMAC5.12 The model was completed in COOT13 and refinement was carried out with REFMAC5 and CNS.14 The final refined model, including 80 of the expected 88 amino acid residues, 97 water molecules, has an R factor (Rfree) of 17.4% (20.7%). The quality of model was checked by using the program PROCHECK.15 The Ramachandran plot reveals that 94.2% of the residues are in the most favored region, with an additional 5.8% in the additionally allowed region. The spindly crystal belongs to space group P31. The asymmetric unit contains four molecules with a solvent content of ∼35%. The structure was determined by Molrep16 using the crystallographic RRM dimer as the search model, which was produced from the previous refined coordinate by a twofold axis of symmetry. The complete model was built manually using the COOT and refinement was carried out with REFMAC5. The final model, with the R factor of 22.6% and the Rfree of 29.0%, contains 352 residues (33 water molecules). The model was also validated using PROCHECK. The Ramachandran plot shows that no residues were found in disallowed regions. Data-collection and refinement statistics are summarized in Table I. Illustrations were prepared with PyMOL.17 The dimer interaction was analyzed using LIGPLOT.18 Solvent-accessible and buried surface were calculated using Protein–Protein Interactions server.19 We determined a crystal structure of human PABPN1 RRM at 2.0 Å resolution by using single wavelength anomalous diffraction methods. The refined model contains 80 amino acids in one RRM molecule in the asymmetric unit. Eight residues (two N-terminal and six C-terminal residues) lacking electron density were not included in the model. Similar to the canonical RRM fold, PABPN1 RRM is composed of a four-stranded antiparallel β-sheet spacially arranged in the order β4β1β3β2 from left to right when facing the sheet and two roughly perpendicular helices (α1, α2) packed against the β-sheet [Fig. 1(a)]. In addition, loop5 (between α2 and β4) has a single β-strand (β′) that extends the main β-sheet. The two conserved motifs, RNP1 (residues 213–220) and RNP2 (residues 174–179), are juxtaposed on the central strands, β3 and β1, respectively. Obviously, the loop3 (between β2 and β3), also known as the variable loop,20 differs markedly in orientation from the previously reported RRMs, except the corresponding RRM domain of ePABP2 from Xenopus laevis [Fig. 1(b)], which solution NMR structure was solved recently (PDB entry 2JWN) by CESG (Center for Eukaryotic Structural Genomics). Xenopus ePABP2 is a cytoplasmic protein and it is an RNA binding protein with preference for poly(A).21 Like PABPN1, ePABP2 contains only a single RRM domain and an arginine-rich C-terminal domain. PABPN1 RRM has 70% sequence identity with ePABP2 RRM, and both of them fold into a very similar structure, including the loop3 orientation, with RMSD (root-mean-square deviation) for main chain atoms of 1.38 Å. These information suggest PABPN1 RRM and ePABP2 RRM may have a similar mode of poly(A) recognition. (a) Ribbon figure of human PABPN1 RRM structure (stereo view). The RNP1 and RNP2 are highlighted in red and yellow, respectively. (b) Backbone trace representation of PABPN1 RRM (green) compared with ePABP2 RRM (blue, pdb code:2jwn) and other RRMs of known structure(grey), including RRM1 and RRM2 of PABPC (pdb code:1cvj), RRM1 of HUD (pdb code:1fxl), RRM1 and RRM2 of hnRNP A1 (pdb code:1ha1; 2up1), RRM1 of SXL (pdb code:3sxl; 1b7f), RRM of P14 (pdb code:2f9d), RRM of Daz-associated protein 1 (pdb code: 2dgs; 2dh8). The unique loop3 (β2/β3 loop) of PABPN1 RRM is marked by black arrow. Although adopting similar topology, human PABPN1 RRM shares low sequence similarity with PABPC RRM1 and RRM2, with identity of 20% and 23%, respectively. Their superposition results in RMSD values of 2.524 Å (PABPN1 RRM − PABPC RRM1) and 2.152 Å (PABPN1 RRM − PABPC RRM2) for main chain atoms. Deviations are largely restricted to loop1 and loop3. Structural differences within loop3 and loop1 appear to be a function of poly(A) binding and intermolecular interactions, as will be discussed below. In the cocrystal structure of human PABPC (RRM1/RRM2),4 Ser12, Tyr14, Asp45, Met46, Leu52, Tyr54, Tyr56, Asn58, Met85, Trp86, and Gln88 from RRM1 are involved in specific hydrogen bonds plus stacking interactions and van der Waals contacts to adenine bases. According to the structure-based sequence alignment of PABPC RRM1 and PABPN1 RRM [Fig. 2(a)], the analogous positions for these residues in PABPN1 RRM are occupied by Ser173, Tyr175, Asp206, Lys207, Lys213, Phe215, Tyr217, Glu219, Ile245, Pro246, and Arg248, respectively. Only four residues (Ser173, Tyr175, Asp206, Tyr217) are conserved and two residues (Phe215, Ile245) have similar physicochemical properties in PABPN1. Others have notable differences in biochemical characteristics or/and spacial orientation. Especially, due to the orientation difference of loop3, Asp206, Lys207, and Lys213 show a main chain shift of about 8 Å, 10.8 Å, and 3.6 Å away from the analogues of PABPC RRM1, respectively. Asn100, Phe102, Lys104, Asn105, Ser127, Tyr140, Phe142, and His144 from RRM2 contribute to the adenine bases discrimination. The relevant residues in PABPN1 RRM are substituted by Ser173, Phe175, Glu177, Asn178, Ser127, Phe215, Tyr217, and Glu219, respectively. Only one of those residues is conserved in PABPN1 (Asn178) [Fig. 2(b)]. Furthermore, surface topology and electrostatics also are widely divergent between PABPN1 RRM and PABPC RRM1 [Fig. 2(c)]. For example, the loop3 of PABPN1 RRM, possessing more strongly electropositive potential than that of PABPC RRM1, may play more important role in specific poly(A) binding. Taken together, this comparison and the structural information imply that the mode of poly(A) binding may indeed be different between PABPN1 and PABPC. (a) Structure-based sequence alignment of the RRM of human PABPN1, RRM1 and RRM2 of human PABPC. The RNP2 and RNP1 motifs, which correspond to the first and third β-strand, respectively, are boxed. Amino-acid residues involved in PABPC-poly(A) interactions are in the shadow. α-helices, β-strands, and 310-helices are denoted by Greek characters, α, β, and η, respectively. Strictly conserved residues are highlighted with red boxes. These amino acid sequences were aligned by Multalin,22 and the figure was generated using ESPript.23 (b) Stereo diagram showing the superposition of PABPC RRM1, RRM2 and PABPN1 RRM. The amino acid residues, which form specific hydrogen bonds to adenine bases in PABPC-oligo(A) complex, are shown in red (RRM1 of PABPC) and blue (RRM2 of PABPC). Amino acid side chains of PABPN1 RRM at positions equivalent to those of PABPC RRM1 and RRM2 are shown in green. (c) Comparison of the electrostatic surface plots between PABPC RRM1 and PABPN1 RRM. The positive surface was drawn in blue and negative surface in red. The crystallographic asymmetric unit of the RRM crystal in cubic form contains only one monomer. However, this monomer appears to form a dimer with its crystallographic symmetry (CS) mate. Thus, we used the CS dimer as a search model to successfully solve the structure of the crystal in trigonal space group P31, which includes two noncrystallographic asymmetry (NCS) dimers in the asymmetric unit. The two NCS dimers (chain A/B, chain C/D) and CS dimers are very similar with RMSD for main chain atoms of 0.275 Å and 0.289 Å, respectively. We use Protein–Protein Interactions server to analyze the monomer interface in the CS dimer and to examine which intermolecular interactions stabilize the observed dimer-like arrangement of the RRM. According to this analysis, 775.85 Å2 of accessible surface are buried in the monomer interface, corresponding to 16.09% of the total accessible surface. The subunit interface area has a length of 29.02 Å and a breadth of 17.41 Å, and 62.2% of the interface is contributed by nonpolar atoms. Thus we conclude that at least the 10 kDa RRM fragment is apt to be dimeric and may have characteristics of true functional dimmers, although the peak position makes it difficult to indisputably distinguish between dimer and monomer in the last step of the purification procedure. This is also supported by the NMR structure of the corresponding RRM domain from Xenopus laevis (also dimeric). Two main interface residue segments are involved in intermolecular interactions according to the analysis of LIGPLOT [Fig. 3(a)]. The β2 strand (residues 198–206, 208) in one monomer forms main chain hydrogen bonds and hydrophobic contacts with the β2 strand in the neighboring monomer in an antiparallel fashion, resulting in an eight-stranded antiparallel sheet platform [Fig. 3(b)]. In addition, there are hydrophobic interactions and hydrogen bonds contributed by some residues (181–186, 189) on loop1 and α1, which provide additional forces to make this dimer stable [Fig. 3(c)]. Dimer interactions of PABPN1 RRM. (a) Two-dimensional plots give an overview of the mode of dimerization in PABPN1 RRM crystal structure. Sticks diagrams show (b) the residues on β2 strand in one monomer forms main chain hydrogen bonds with the β2 strand in the neighboring monomer in an antiparallel fashion, and (c) some residues on loop1 and α1 provide additional dimeric forces. The relevant residues are labeled in (b) and (c). PABPN1 has many forms under native condition. There was equilibrium between monomers, dimers and higher oligomers. In OPMD patients, it accumulates and forms intranuclear inclusions.24 The self-interaction is promoted by the C-terminal domain and compatible with poly(A) binding of PABPN1.7 The crystal structure of PABPN1 RRM suggests that, besides the C-terminal domain, the RRM may also contribute the interaction of PABPN1 and stabilize the soluble or insoluble inclusions under some conditions. In the PABPC-poly(A) cocrystal structure, two tandemly arranged RRMs create a continuous RNA recognition surface that performs specific binding to poly(A).4 As the key domain of the PABPN1-poly(A) complex, the single RRM specially binds to poly(A), whereas, the argine-rich C-terminal domain contributes nonspecific contacts.7 In sight of the known structures of RRM-RNA complexes, one RRM interacts with no more than six nucleotides,25, 26 except in structured RNAs, such as U1 snRNA hairpin II and U2 snRNA hairpin IV, in which seven and 12 nucleotides can be recognized specially.27, 28 The selected experiment suggested that base recognition is mediated exclusively by the PABPN1 RRM,7 and the minimal specific binding site size is 10–11 nucleotides according to quantitative evaluation of the change in PABPN1 affinity with the lengths of a series of ribooligoadenylates.5, 29 Consequently, PABPN1 specifically recognizes poly(A) by two possible ways: either two RRMs cooperate towards the base discrimination, or poly(A) possesses regular secondary structure. In NCS and CS dimer of PABPN1 RRM, the eight antiparallel β-sheet platform may provide a larger poly(A) binding interface to reach the high specificity. However, this would imply that the same surface of the monomer is able to recognize both 5′ to 3′ and 3′ to 5′ oriented oligonucleotides. The second one is very attractive. Conformational flexibility of RNA may also play a role in governing interactions with RRM-containing proteins. In the crystal structure of the U2B″-U2A′ protein complex bound to RNA, 12 nucleotides of U2 snRNA hairpin IV are contacted by the RRM domain of U2B″ protein. The superposition of PABPN1 RRM and the U2B″ RRM bound to a fragment of U2 snRNA provides a possibility for PABPN1-poly(A) binding mode [Fig. 4]: oligonucleotides (10–11 adenine bases) possessing a regular secondary structure are recognized by the RRM monomer of PABPN1. Whether or not, the real mechanism of specific PABPN1-poly(A) recognition may need to be elucidated with the determination of a PABPN1-oligo(A) complex in the future. Stereo ribbon figure showing the superposition of PABPN1 RRM (green) and the U2B″ RRM (orange) bound to a fragment (5′-6AUUGCAGUACC16-3′) of U2 snRNA. We are grateful to Dr. Congzhao Zhou and Dr. Zhi Liang for reading the manuscript. We thank Dr. Zhiqiang Zhu, Dr. Jiang Yu, Dr. Hui Rong, Dr. Weili Yang, and Dr. Huihao Zhou for many useful discussions.