Crystal structure of the translation recovery factor Trf from Sulfolobus solfataricus

During translation initiation, the heterotrimeric archaeal translation initiation factor 2 (aIF2) recruits the initiator tRNA i to the small ribosomal subunit. In the stationary growth phase and/or during nutrient stress, Sulfolobus solfataricus aIF2 has a second function: It protects leaderless mRNAs against degradation by binding to their 5′‐ends. The S. solfataricus protein Sso2509 is a translation recovery factor (Trf) that interacts with aIF2 and is responsible for the release of aIF2 from bound mRNAs, thereby enabling translation re‐initiation. It is a member of the domain of unknown function 35 (DUF35) protein family and is conserved in Sulfolobales as well as in other archaea. Here, we present the X‐ray structure of S. solfataricus Trf solved to a resolution of 1.65 Å. Trf is composed of an N‐terminal rubredoxin‐like domain containing a bound zinc ion and a C‐terminal oligosaccharide/oligonucleotide binding fold domain. The Trf structure reveals putative mRNA binding sites in both domains. Surprisingly, the Trf protein is structurally but not sequentially very similar to proteins linked to acyl‐CoA utilization—for example, the Sso2064 protein from S. solfataricus—as well as to scaffold proteins found in the acetoacetyl‐CoA thiolase/high‐mobility group‐CoA synthase complex of the archaeon Methanothermococcus thermolithotrophicus and in a steroid side‐chain‐cleaving aldolase complex from the bacterium Thermomonospora curvata. This suggests that members of the DUF35 protein family are able to act as scaffolding and binding proteins in a wide variety of biological processes.

During translation initiation, the heterotrimeric archaeal translation initiation factor 2 (aIF2) recruits the initiator tRNA i to the small ribosomal subunit. In the stationary growth phase and/or during nutrient stress, Sulfolobus solfataricus aIF2 has a second function: It protects leaderless mRNAs against degradation by binding to their 5 0 -ends. The S. solfataricus protein Sso2509 is a translation recovery factor (Trf) that interacts with aIF2 and is responsible for the release of aIF2 from bound mRNAs, thereby enabling translation re-initiation. It is a member of the domain of unknown function 35 (DUF35) protein family and is conserved in Sulfolobales as well as in other archaea. Here, we present the X-ray structure of S. solfataricus Trf solved to a resolution of 1.65 A. Trf is composed of an N-terminal rubredoxin-like domain containing a bound zinc ion and a C-terminal oligosaccharide/oligonucleotide binding fold domain. The Trf structure reveals putative mRNA binding sites in both domains. Surprisingly, the Trf protein is structurally but not sequentially very similar to proteins linked to acyl-CoA utilization-for example, the Sso2064 protein from S. solfataricus-as well as to scaffold proteins found in the acetoacetyl-CoA thiolase/high-mobility group-CoA synthase complex of the archaeon Methanothermococcus thermolithotrophicus and in a steroid sidechain-cleaving aldolase complex from the bacterium Thermomonospora curvata. This suggests that members of the DUF35 protein family are able to act as scaffolding and binding proteins in a wide variety of biological processes.
Sulfolobus solfataricus is a thermoacidophilic archaeon originally isolated from hot volcanic springs. It grows optimally at 350 K and in acidic environments with a pH between 2 and 4 [1]. Fundamental features of translation initiation are conserved between archaea such as S. solfataricus and eukaryotes [2]. The heterotrimeric translation initiation factor IF2 (a, b, and c) is required for binding of the methionine-loaded initiator tRNA (tRNA i ) to the small ribosomal subunit as well as the proper positioning of tRNA i in the ribosomal P-site in archaea and in eukaryotes. However, crenarchaeal archaeal initiation factor 2 (aIF2) lacks the tRNA i shuttle function of eukaryotic initiation factor 2 [3,4]. Interestingly, in S. solfataricus the aIF2c subunit has an additional function during the stationary growth phase and/or during nutrient starvation. Under these conditions, it protects mRNAs against exonuclease-mediated 5 0 -3 0 decay via binding to their triphosphorylated 5 0 -ends [5,6]. To reinitiate translation after nutrient replenishment, aIF2c must be displaced from mRNAs. The small protein translation recovery factor (Trf) was identified as the factor responsible for releasing aIF2c from bound mRNAs, thereby enabling re-initiation of translation after relief of nutrient stress [5].
A sequence alignment of Trf homologs from different Sulfolobus species revealed two conserved regions predicted to be similar to an N-terminal rubredoxinlike and a C-terminal oligosaccharide/oligonucleotide binding fold (OB-fold) domain [5]. Here, we report the crystal structure of the 124-amino acid (aa) S. solfataricus Trf protein (Sso2509) at a resolution of 1.65 A. Interestingly, the structure but not the sequence of Trf is similar to other proteins of the domain of unknown function 35 (DUF35) family [Protein Data Bank (pdb): pdb 3irb, 6et9 chain E, 6ok1 chain B, 5m3k chain B] that have been described to bind directly or indirectly to acyl-CoA [7,8] or act as scaffolding components of acyl-CoA-dependent enzyme complexes [8,9] as well as a CoA-independent bacterial acylase [10]. This suggests a general role of DUF35 family proteins as scaffolding or binding proteins in a variety of biological processes.

Materials and methods
A codon-optimized gene for production of Trf in Escherichia coli was obtained commercially from GenScript in a pUC57 plasmid vector. The gene encoded the fulllength native Trf protein sequence from S. solfataricus. The coding region was subcloned into a pET11a expression plasmid using the NdeI and BamHI restriction sites.
Escherichia coli BL21 (DE3) cells were transformed with this plasmid, and cells were grown at 37°C in Luria broth medium supplemented with 100 mgÁmL À1 ampicillin until an optical density (OD 600 ) of 0.8 was reached. Trf synthesis was induced by adding 1 mM isopropyl b-D-1-thio-galactopyranoside at 20°C. Sixteen hours after induction, the cells from a 1 L culture were harvested by centrifugation at 5000 g (10 min, 4°C) and disrupted by sonication in 20 mL lysis buffer (50 mM BisTris pH 5.8, 200 mM NaCl, 10 mM ß-mercaptoethanol, 200 mM imidazole) on ice. After cell lysis, a heat denaturation step was performed at 65°C for 15 min. Subsequently, the solution was centrifuged at 7000 g (20 min, 4°C). The supernatant containing Trf was diluted with dilution buffer (50 mM BisTris pH 5.8, 10 mM ß-mercaptoethanol, 200 mM imidazole) to 150 mL and loaded onto a SP-Sepharose column (GE Healthcare, Solingen, Germany, HiPrepTM SP FF 16/60). Trf was eluted with a buffer containing 1 M NaCl in a linear gradient. Fractions containing Trf were pooled, concentrated to 4 mL, and applied to a gel filtration column (Superdex 200 HR 26/60; GE Healthcare) pre-equilibrated with gel filtration buffer (50 mM BisTris pH 6.5, 50 mM NaCl, 5 mM ß-mercapthoethanol, 200 mM imidazole). Fractions containing Trf were pooled and concentrated to a final protein concentration of 175 lM. Protein purity was assessed by SDS/PAGE. MALDI mass spectrometry yielded a molecular weight of 14.78 kDa (calculated 14.78 kDa).
Crystallization of Trf was achieved by slowly removing imidazole from the buffer during a two-step dialysis of a gel filtration sample (4 mL/175 lM) of Trf against 1 L of an imidazole-free dialysis buffer (50 mM BisTris pH 6.5, 50 mM NaCl, 5 mM ß-mercapthoethanol) at 4°C in a dialysis volume of 500 mL. After 24 h, the buffer was exchanged, and dialysis was continued for additional 72 h. After 4 days of dialysis, crystals appeared in the dialysis chamber (Slide-A-Lyzer TM , Thermo Fisher, Waltham, MA, USA, 3.5 kDa molecular weight cutoff). A microscopic analysis revealed rod-like crystals being UV-active. These crystals were used for structure determination. The crystals were cryoprotected by adding 33% PEG-400 to the crystallization solution.
X-ray diffraction data were collected on the beamline station PXIII of the Swiss Light Source (Paul Scherrer Institut, Villigen, Switzerland). All diffraction data were obtained from a single crystal and processed with the XDS software package [11]. The positions of anomalous scatterers were determined using SHELXD [12]. The initial phases were obtained by SHELXE [13,14], and the initial model was built automatically by RESOLVE [15]. The initial model was extended by iterative rounds of model building with COOT [16] into the 2F o -F c electron density map and refined using the PHENIX software package [17]. The Ramachandran plots of the final structure showed no outlying residues, as assessed by model validation with the program MOLPROBITY [18]. The graphical representations were generated using CHI-MERA [19]. Proteins that are structurally similar to Trf were identified using the DALI server [20] and PDBeFOLD [21]. Ca RMSDs and Z-scores were calculated using the pairwise structure comparison option available at the DALI server.
Fluorescence anisotropy-based RNA binding assays were performed at 25°C using a Fluorolog 3 spectrometer (Horiba, Bensheim, Germany). Excitation and emission wavelengths were set to 492 and 521 nm. 5 0 -fluorescein-labeled RNAs were obtained commercially (Dharmacon, Lafayette, CO, USA) and deprotected according to the instructions of the manufacturer. Experiments were performed in a buffer containing 50 mM BisTris, pH 6.5, and 100 mM NaCl. A NaCl concentration of 100 mM in the buffer was necessary to prevent precipitation of the protein. The concentrations of the 5 0 -fluorescein-labeled RNAs were 100 nM. Trf was added stepwise until saturation was reached. Titration experiments were performed in triplicate. Each data point represents the average value with error bars indicating the standard deviation. Analytical gel filtration in a buffer containing 50 mM BisTris, pH 6.5, and 150 mM NaCl was performed as described previously [22]. A protein concentration of 50 lM was used. For molecular weight comparisons, we used the proteins PhS11 and PhFap7 as well as the highly stable PhFap7/PhS11 complex with molecular weights of 14.74, 20.18, and 34.92 kDa, respectively [18].

Results and Discussion
Structural information for the 124 aa protein Trf responsible for the release of mRNAs bound to aIF2c during the outgrowth phase in S. solfataricus was so far not available [5]. Here, we report the crystal structure of Trf determined from a cuboid needle-like crystal diffracting to a resolution of 1.65 A (Table 1). Previously, a sequence alignment suggested a N-terminal rubredoxin-like domain followed by a C-terminal OB-fold domain [5]. The X-ray structure (Fig. 1A) confirmed that Trf contains two domains, an N-terminal rubredoxin-like domain (aa: 19-51) and a C-terminal OB-fold-like domain (aa: 54-117). Additionally, a short a-helix (a1, aa 3-15) precedes the rubredoxin-like domain which is composed of two antiparallel bstrands (b1, aa 19-24; b2, aa 46-51) surrounding a loop region (aa 25-45), which complexes a zinc ion. The zinc ion is coordinated by four cysteine residues (aa 24, 27, 38, and 41) of the loop region (Fig. 1B). The b-barrel of the OB-fold-like domain is formed by five antiparallel b-strands (b3-b7, aa 54-65; 69-78; 83-88; 99-107; and 110-117) as is typical for this type of domain [23].
The asymmetric unit contains two molecules (chains A and B; Fig. 1A). For both molecules, the refined model begins with residue 1 and terminates with residue 119. The C-terminal residues His120, Leu121, His122, Asn123, and Phe124 could not be resolved due to insufficient electron density suggestive of a flexible C terminus. Overall, the structure contains the two protein chains A and B, two zinc ions, and 257 water molecules. Both monomers of the unit cell are virtually identical with an overall RMSD for Ca atoms 1-119 of 0.739 A (Fig. 1C). However, the dimer interface corresponds to a buried solvent-accessible surface area of only 789 A 2 as determined with PDBePISA. This argues against the formation of a stable Trf dimer in solution [24]. The retention volume of Trf on an analytical gel filtration column suggests that Trf in solution is indeed a monomeric protein (Fig. 1D).
An electrostatic surface representation of Trf shows a positively charged surface for b-strands b3, b4, and b5 which correspond to the first three b-strands of the OB-fold domain as well as for a part of the zinc ribbon motif (Fig. 2A). These two positively charged surface areas are located close to each other and might therefore represent a possible interaction platform for nucleic acids. This would correspond to the RNAbinding site in many other OB-fold proteins where often b-strands 2 and 3 sometimes ß-strand 1 represent the recognition and binding surface for their nucleic acid ligands [23,25]. Two other prominent determinants for RNA binding in OB-folds are the two loops connecting b-strands b1 and b2 (L 12 ) as well as b4 and b5 (L 45 ). In agreement with an RNA binding function for Trf, these loops contain a significant number of basic residues (R65 and K66 in L 12 and K105, K110, and K111 in L 45 ) augmented by aromatic residues (Y64 and Y70 in L 12 as well as Y112 in b4) that could contribute to RNA binding by providing base stacking interactions. Furthermore, many OB-fold proteins feature a solvent-exposed hydrophobic residue at the center of strand 3 which is often involved in stacking interactions with nucleic acid bases [23,25]. This residue is also present in Trf as Y85. However, in previous experiments using electrophoretic mobility shift assays (EMSA) no evidence for direct RNA binding by Trf was found [5]. We tested the binding of Trf to a single-stranded RNA and an RNA that contains a hairpin loop next to single-stranded segments which were 5 0 -fluorescein-labeled with fluorescence anisotropy measurements. Fluorescence anisotropy measurements are sensitive to binding events over a broader K D range compared to EMSA experiments. In titration experiments, we found that Trf bound these RNAs with K D values in the low micromolar range (Fig. 2B). Thus, Trf apparently has a genuine RNA-binding capacity in agreement with its proposed function. However, the additional presence of a structured hairpin loop element in one of the RNAs did not lead to an increased affinity of Trf for this RNA.
To find structurally similar proteins in the PDB, we used the DALI server [20] as well as PDBeFold [21].
Trf has structural homology to four other published structures (PDB 3irb, 6et9, 6ok1, and 5m3k, Fig. 3A). All four proteins also belong to the DUF35 family [7][8][9][10] but apparently have no RNA-related functions. The first example-a protein termed Sso2064-also occurs in S. solfataricus [7]. Functionally, Sso2064 is linked to acetyl-CoA binding related to functions in the biosynthesis of lipids and polyketide antibiotics based on a genome context analysis [7]. Sso2064 also crystallized as a homodimer similar to our observations for Trf. The superimposition of Trf with Sso2064 shows a similar overall fold with a Ca RMSD of 2.3 A for 108 aligned residues and a DALI Z-score of 10.4. The second DUF35 protein was recently identified as a scaffolding subunit in the acetoacetyl-CoA thiolase/high-mobility group-CoA synthase complex of Methanothermococcus thermolithotrophicus (pdb entry 6et9 chain E) which is part of the mevalonate pathway [8]. The DUF35 subunit (chain E) in pdb entry 6et9 aligns to Trf with a Ca RMSD of 2.2 A for 107 residues with a DALI Z-score of 12.7. Another example for a DUF35 protein acting as a scaffolding subunit in a CoA-dependent enzyme complex was found in the structure of a steroid side-chain-cleaving aldolase [9]  from the bacterium Thermomonospora curvata (pdb entry 6ok1, chain B). The structure of this protein aligns to Trf with a Ca RMSD of 2.4 A for 116 aligned residues (DALI Z-score 15.4). DUF35 protein family members can also occur as a scaffolding subunit in CoA-independent enzymes [10] as illustrated by the structure of a Friedel-Crafts acylase complex from the bacterium Pseudomonas protegens (pdb entry 5m3k, chain B). However, it should be noted that this enzyme complex is apparently structurally and evolutionarily related to CoA-dependent enzymes [10]. The DUF35 subunit of this complex aligns to Trf with a Ca RMSD of 2.5 A for 110 aligned residues (DALI Z-score 12.0). All DUF35 proteins structurally characterized so far bind a zinc ion in equivalent positions. In those cases where the DUF35 homolog was crystallized as a scaffolding protein in the context of larger hetero-oligomeric assemblies (6et9, 6ok1, and 5m3k) which are either dimers of heterodimers or dimers of heterotrimers, the DUF35 subunits are always involved in extensive interactions with the other subunits of the complex and never stabilize these complexes by homotopic interactions with each other.
In comparison with Sso2064, all other structurally characterized DUF35 family members including Trf lack an additional long N-terminal a-Helix (Fig. 3B). Despite the high degree of structural similarity between all proteins, the level of sequence identity between Trf and all other structurally characterized DUF35 proteins is rather low (22% for Sso2509, 24% for chain E of 6et9, 22% for chain B of 6ok1, and 19% for chain B of 5m3k). A sequence alignment of the proteins is shown in Fig. 3A. Importantly, the electrostatic surface potentials of the proteins differ considerably. The N-terminal helical parts of Trf as well as of chain E of 6et9 show a positive electrostatic surface potential, whereas the N-terminal helices of Sso2064 feature a negative surface potential due to the presence of an additional long N-terminal a-helix in this protein (Fig. 3C).
Furthermore, the parts of both domains forming the putative 'binding cleft' display a positive surface potential in Trf as well as in chain E of 6et9 and chain B of 5m3k, while Sso2064 and chain B of 6ok1 show a negative surface potential in this area (Fig. 3C). Thus, in regard to its overall structure and its surface properties Trf is most similar to chain E of 6et9. Interestingly, in chain E of 6et9 the positively charged 'binding cleft' between the two domains interacts with the thiolase domain of the enzyme complex. For Trf, this binding cleft could be either the interaction surface for aIF2c or for its target mRNAs. However, since S. solfataricus aIF2c features only a rather limited area of negative surface potential compared to the thiolase domain of M. thermolithotrophicus, a potential mRNA binding surface seems to be more likely. In analogy to the function of chain E of 6et9, Trf might function in order to establish a transient scaffolding state between Trf, aIF2c, and the bound mRNA to promote the release of bound mRNAs from aIF2c.
Overall, our results demonstrate that while sequentially dissimilar DUF35 proteins fold into highly similar structures, they differ considerably in their surface features and can therefore be recruited into a potentially wide variety of biological pathways with general scaffolding and binding functions.