Structural study of MCPIP1 N-terminal conserved domain reveals a PIN-like RNase

Protein expression, purification and crystallization

Sequences encoding the MCPIP1 ND (residues 1–297) and ND-ZF (residues 1–334) were respectively inserted into pGEX-6 p-1 vectors (GE Healthcare), and Escherichia coli BL21 (DE3) cells were transformed with these vectors for expression. For Se-derived ND protein expression, E. coli B834 cells were transformed with the cloned vector. All E. coli cells were cultured at 16°C after adding Isopropyl β-d-1-thiogalactopyranoside (IPTG) to induce gene expression. Cells were harvested and then lysed by ultrasonication. After centrifugation at 30 700g, supernatants were applied to a Glutathione Sepharose 4B (GE Healthcare) column in order to remove other proteins. PreScission protease (GE Healthcare) was used to cleave the fusion proteins to obtain the ND and ND-ZF protein constructs after elution from the column by Glutathione. Amino acids 108–113 of MCPIP1 (L-V-P-R-G-G) are similar to those of the thrombin recognition cleavage site (L-V-P-R-G-S). Thus, thrombin protease (Novagen) was used to obtain the NCD (resides 112–297) and NCD-ZF (residues 112–334) protein, which were then further purified separately by Hitrap Q and Hitrap S before crystallization.

Crystallization was performed by the hanging drop vapor diffusion method. Proteins NCD, Se-NCD and NCD-ZF were concentrated to 7 mg/ml (in 50 mM NaCl, 20 mM Tris–HCl pH 8.0 buffer), 5 mg/ml (in 50 mM NaCl, 20 mM Tris–HCl pH 8.0 buffer) and 6 mg/ml (in 300 mM NaCl, 20 mM Tris–HCl pH 8.3 buffer), respectively, for crystallization. After optimization, an NCD crystal of size 200 × 10 × 10 µm³ was obtained in 1 μl of the above protein solution mixed with 1 μl of reservoir solution (0.1 M Bis–Tris pH 5.5, 7% PEG 3350). The Se-NCD crystal used for SAD data collection was obtained using the same condition. The single NCD-ZF crystal of size 200 × 200 × 50 µm³ used for data collection was grown in 4% Tacsimate pH 5, 6% PEG 3350. The NCD-ZF crystal was dehydrated in a solution of 30% ethanol, 4% Tacsimate pH 5, 15% PEG 3350 for a few seconds before mounting. All crystals were flash-frozen in liquid nitrogen before data collection.

Data collection and structure determination

Native NCD and Se-derived crystal diffraction data sets were collected at BL-17 A, Photon factory, KEK, Japan. The wavelength used for anomalous diffraction data collection was 0.9790 Å. Native NCD-ZF crystal diffraction data was collected at BL-17I, SSRF, China. All data sets were processed and scaled using the HKL2000 package (15). Due to anisotropic diffraction of NCD-ZF crystal, the ellipsoidal truncation and anisotropic scaling was applied to NCD-ZF data using the online Diffraction Anisotropy Server (16). The phasing of the NCD structure was solved using the Single-wavelength Anomalous Diffraction (SAD) method with the hkl2map program (17–20). The NCD-ZF structure was solved by the Molecular Replacement (MR) method using the NCD structure as a model (21,22). Mg²⁺-derived NCD crystals were obtained by soaking the NCD crystal in a reservoir containing 15 mM MgCl₂. Mg²⁺-derived crystal data was collected at BL-17A, Photon factory, KEK, Japan. The Mg²⁺ ion-containing structure was solved by the MR method using the NCD structure as a model. All refinements were carried out using the Phenix.refine and Coot programs (23,24). Table 1 shows the crystallographic data and refinement statistics for the two structures. Figures were prepared using PyMol (25).

Synthesis of [³²P]-labeled RNAs and in vitro RNase activity determination

In vitro transcription was carried out using a MEGAshortscript^TM Kit (Ambion, Inc). UTP was replaced by 90% α-[³²P]-UTP (PerkinElmer) to produce isotope-labeled mRNAs. A T7 RNA polymerase promoter was inserted upstream of the mouse IL-6 3′-UTR by PCR using cDNA from mouse endothelial cells. Amplified IL-6 3′-UTR and a control template in the kit were used as templates to synthesize different mRNAs. TURBO DNase was added to the reaction system after transcription to remove the DNA template. Enzymes and free nucleotides were removed by phenol:chloroform extraction and alcohol precipitation.

To measure RNase activity, purified [³²P]-labeled RNA was incubated with purified MCPIP1 constructs at 37°C. Denaturing acrylamide gels were run and autoradiography was used to measure the amount of RNA left after digestion by different proteins.

In vivo RNase activity determination

Luciferase reporter assays were used according to a previously reported method (26,27). The MCPIP1 cDNA was inserted into pcDNA4^TM/TO/myc-His A (Invitrogen) to express myc-tagged full-length MCPIP1 protein. Mutations were introduced into the vector directly using a quick-change single mutation method. The PGL3-IL-6 3′-UTR vector was generated by inserting the IL-6 3′-UTR sequence into a PGL-3 Luciferase Reporter Vector (Promega). 293A cells were transfected with 1 μg plasmid of each construct and 0.1 μg IL-6 3′-UTR PGL3 plasmid using Lipofectamine^TM 2000 (Invitrogen) according to the manufacturer’s instructions. A Renilla luciferase reporter pRL-TK (Promega), which is not sensitive to MCPIP1, was included for normalization of transfection efficiencies. Cells were harvested 48 h later and lysed in NP40 lysis buffer. After centrifuging for 20 min at 12 000g, 4°C, the supernatant was tested for luciferase activity using a luciferase assay system (Promega).

The N-terminal conserved domain degrades IL-6 mRNA 3′-UTR in vitro

To determine the minimal part of MCPIP1 that is required for its RNase activity, we constructed four truncations, spanning residues 1–297, 112–297, 1–334 and 112–334 (referred to hereafter as ND, NCD, ND-ZF and NCD-ZF, Figure 1A). These four recombinant protein truncations were incubated respectively (50 ng each) with [³²P]-labeled IL-6 mRNA 3′-UTR (50 ng) for 20 min in the presence of 5 mM Mg²⁺ (the reaction was stopped before the RNA was completely digested to reveal differences between the different truncations). All truncations were able to digest [³²P]-labeled IL-6 mRNA 3′-UTR, but their RNase activities were different (Figure 2A). ND-ZF had the highest RNase activity while that of NCD was the lowest, suggesting that both the first 111 N terminal residues and the zinc-finger motif are required for efficient digestion of substrate mRNA. In addition, ND had comparable activity to ND-ZF, and both constructs had much higher activity than constructs which lacked the N-terminal region (NCD and NCD-ZF), implying that the N-terminal region is more important for RNase activity than the zinc-finger domain. From the above results, we conclude that while the conserved domain of MCPIP1 (112–297) is sufficient for RNase activity in vitro, the zinc-finger domain and the N-terminal region may also play important roles in RNA digestion in vitro and in vivo.

We also tested the in vitro metal ion selectivity and the RNA specificity of the RNase activity. Ca²⁺, Fe²⁺, Mg²⁺, Mn²⁺ and Zn²⁺ (at a final concentration of 5 mM) were added separately to a system containing 50 ng [³²P]-labeled IL-6 mRNA 3′-UTR and 50 ng ND-ZF to test their influence on RNase activity. High RNase activity was observed both in the presence of Mg²⁺ and Mn²⁺, while in the presence of Ca²⁺, Fe²⁺ and Zn²⁺, ND-ZF lost almost all its RNase activity (Figure 2B). Since the zinc finger might participate in RNA recognition, ND-ZF was also incubated respectively with IL-6 mRNA 3′-UTR and control mRNA in the presence of 5 mM Mg²⁺ to test for RNA specificity. No RNA was left after 1 h incubation in the presence of ND-ZF (Figure 2C). These results indicate that MCPIP1 is a Mn²⁺ or Mg²⁺ dependent RNase and that substrate specificity is lost in vitro in the absence of the C-terminal region of MCPIP1.

Overall structure of the conserved domain of MCPIP1

To obtain an atomic resolution structure for further analysis, we crystallized NCD and NCD-ZF and solved their structures. The NCD crystal belongs to space group P4₁ with unit cell parameters a = b = 56.4 Å, c = 113.2 Å. Matthews coefficient analysis (30) suggested that there are two molecules in one asymmetric unit at a solvent content of 43.08% (V_M = 2.16 ų/Da). The initial phases were obtained using the multiwavelength anomalous diffraction method. The N-terminal 22 residues (i.e. 112–133) and residues 216–222 are missing in the final model due to the absence of interpretable electron density. Residues 134–215 and 223–296 can be well traced in the density map, while residues 216–222 extended into the solvent region without interacting with neighboring molecules in the crystal and thus tended to be flexible. The NCD-ZF crystal belongs to space group P3₁21 and has a unit cell parameter a = b = 113.5 Å, c = 78.0 Å. The initial phases were obtained by MR using the NCD structure as a search model. As was the case for NCD, there are two molecules in one asymmetric unit, and loop 216–222 was well defined in the structure of NCD-ZF. This loop forms a positively charged arm whose significance will be discussed below. The CCCH zinc-finger motif (297–334) was totally disordered in our structure of NCD-ZF, although SDS–PAGE analysis confirmed that the CCCH zinc finger was not degraded in the crystal. Residues 134–296 were well build in chain A, while residues 188–190 were not able to be traced in chain B of the NCD-ZF structure due to the weak density. A stereo view of chain A of NCD-ZF structure was shown in Figure 3A. The structures of NCD and NCD-ZF are quite similar, and the RMSD between the two structures is 0.479 Å (Figure 3B). Secondary structures were labeled according to NCD-ZF structure in Figures 1B and 3B. The reliably modeled range of NCD-ZF structure was also indicated in Figure 1B as untraceable residues are marked with dot above them.

NCD and NCD-ZF were packed in an identical dimer arrangement in their respective crystals. However, analytical ultracentrifugation indicated that the monomer was the predominant form of MCPIP1 NCD in solution (data not shown). The dimer we observed may be a crystallographic artifact which has no physiological significance.

The NCD structure adopts a Rossmann-like α/β/α sandwich fold (Figure 3A and B). β-strands 1, 2, 3 and 6 form a β-sheet surrounded by sets of α-helices at both sides. Dali results show that MCPIP1 NCD shares highest structural homology with T5 D15 5′-exonuclease (PDB: 1exn) (31,32) (RMSD 2.4 Å for 92 Cα of aligned residues), suggesting a previous assumption that the conserved domain of MCPIP1 is a nuclease. Alignment of the NCD-ZF structure with that of T5 D15 5′-exonuclease showed that the central four-stranded β-sheet of the NCD-ZF structure is very similar to that of T5 D15 5′-exonuclease (Figure 3C). There was a good fit for the three α-helices (α 1, α 2 and α 5) surrounding the β-sheet in these two structures. While the arm in the T5 D15 5′-exonuclease structure that is responsible for nucleotide binding was not observed in our structure, a flexible positively charged region (residues 214–222) was found near the location of the nucleotide-binding arm. Since MCPIP1 N-terminal conserved domain shared remote homology with PIN domain as suggested in an early study (8), we also compared the NCD-ZF structure with PIN domain of SMG6 (PDB: 2hww) (33). A good fit in the central part of these structures was shown in Figure 3D. Although the sequence identity was <10%, the three structures share highly conserved catalytic center (shown in Figure 3E).

The catalytic center has several residues important for RNase function

Analysis of the electrostatic surface potential of the NCD-ZF structure revealed a negatively-charged pocket which is presumably the catalytic center of this RNase domain (Figure 4A). Consistent with a previous prediction (8), this pocket is composed of several conserved acidic residues including Asp141, Asn144, Asp225, Asp226, Asp244 and Asp248.

To obtain an accurate structure of the catalytic center and identify which residues are important for processing RNA, we solved the structure after soaking magnesium into the NCD crystal. Although many nucleases require two metal ions for complete enzymatic activity, only one magnesium ion was found in the catalytic center of MCPIP1. The magnesium ion forms six coordination bonds with surrounding atoms including an oxygen from residue Asp 226 and five water molecules, forming an octahedron (Figure 4B). The length of these bonds ranged from 2.0 to 2.3 Å, consistent with the empirical value of 2.07 Å (34). Asp141, Asp225 and Asp244 form hydrogen bonds with these water molecules which help to maintain the magnesium ion in the correct position, suggesting that they may be important for the RNase activity of MCPIP1. While Asn144 and Asp248, which were previously predicted to be involved in the catalytic center (8), were 6.5 and 8.4 Å away from the magnesium ion, they had no interaction with either the magnesium ion or the water molecules around the magnesium ion, suggesting that they probably do not contribute to the RNase activity. Taken together, the structure of the MCPIP1 conserved domain containing a magnesium ion shows that Asp 141, Asp 225, Asp 226 and Asp 244 should be important for its RNase activity.

In vivo studies confirm which residues of the catalytic center are essential

Luciferase assays were used to determine whether specific residues in the catalytic center could affect RNase activity in vivo. An IL-6 3′-UTR sequence was inserted upstream of the luciferase gene in a PGL3 vector and a full-length MCPIP1 cDNA was inserted into a pcDNA4 vector to be overexpressed. The PGL3-IL-6 3′-UTR and pcDNA4–MCPIP1 plasmids were cotransfected into 293A cells. Luciferase activity was measured 48 h after transfection. MCPIP1 protein expression was confirmed by western blotting (Figure 5B). In general, expression levels of MCPIP1 mutants were higher than those of the native protein, but some showed lower activity or even no activity at all. Results showed complete loss of RNase activity in single mutants D141A, D225A, D226A, D244A (Figure 5A), even though these mutations enhanced the level of protein expression. This result is easy to interpret since residue Asp 226 interacts directly with the magnesium ion while residues Asp 141, Asp 225 and Asp 244 hold the magnesium ion in position by coordination bond bridges mediated by water molecules. Mutation of any of these residues would disrupt the elaborate coordination bonds and hydrogen bond network which holds the magnesium ion in the correct place allowing native RNase activity. Also these four residues lie in the conserved catalytic center as suggested by superposition of NCD-ZF with 1exn and 2hww (Figure 3E). In contrast to these mutants, mutant N144A had the same RNase activity as native MCPIP1; this result is in contrast to previous studies but is consistent with the assumptions based on the crystal structure that Asn 144 may not participate in digesting RNA. Asn 144, which previous computational studies have suggested to be important, appears to play no significant role in RNase activity as measured by luciferase assays. These findings are consistent with findings from the structure that the magnesium ion is important in RNA digestion and that residues interacting with the magnesium ion directly or indirectly are all crucial for RNA decay.

A putative RNA-binding site is revealed by in vivo luciferase assays

The catalytic center of MCPIP1 can be clearly seen in our MCPIP1 structure. We noticed that four residues (Arg 214, Arg 215, Lys 219 and Arg 220) formed a positively charged arm located close to the catalytic center (Figure 4A). When we superposed our MCPIP1 structure with that of T5 5′-exonuclease [PDB: 1exn (31)], we found that this positively charged arm was in the same direction with a positively charged arch which is thought to be the nucleotide-binding region of T5 5′-exonuclease (oval region in Figure 3C). Since the positively charged arm was flexible, we hypothesized that this positively charged arm in MCPIP1 may also be involved in mRNA binding. In order to verify this speculation, a luciferase activity assay was used to test the RNase activity of different mutants in vivo (Figure 5A). The single mutation R214A yielded a high level of luciferase activity which was similar to the negative control in which there was no MCPIP1 protein expression, suggesting a complete loss of RNase activity. Mutation of Lys 219, Arg 215 and Arg 220 to alanine also yielded higher luciferase activity compared to wild-type MCPIP1, indicating that these residues may affect MCPIP1 RNase activity. Considering the location of these positively charged residues and their influence on RNase activity, we believe these residues may play a role in RNA binding. Our results indicate that these two positively charged antiparallel β-strands are likely to be a potential RNA-binding site.