Crystal structure of the small form of glucose-inhibited division protein A from Thermus thermophilus HB8



Molecules of tRNA contain many variously modified nucleosides that are important for tRNA function. These modified nucleosides are most often found in the anticodon region, and they contribute to the accurate recognition of the mRNA codon, which is necessary for high-fidelity translation.

The gene of glucose-inhibited division protein A (gidA) was first isolated in association with a glucose-inhibited division phenotype of Escherichia coli.1 The gidA genes are well conserved among a wide range of prokaryotes, implying that they play a fundamental role within the cell. The GidA protein was recently shown to be involved in the biosynthesis of the 5-methylaminomethyl-2-thiouridine (mnm5s2U), which is found in the wobble position of bacterial tRNAs.2 This modification is necessary for stabilizing codon-anticodon interactions. Inactivation of gidA in E. coli caused a two-base translational frameshift.2 The synthesis of mnm5s2U requires several enzymatic reactions, and the details of the reactions remain unknown. The recently presumed reaction pathway is shown in Figure 1(A).3

Figure 1.

A: Proposed pathway for the biosynthesis of mnm5s2U.3B: Sequence alignment of GidAsmall, GidA from T. thermophilus HB8, T. maritima, and E. coli. The sequence numbering corresponds to GidAsmall. Strictly and highly conserved residues are shaded in black and gray, respectively. The secondary structural elements of GidAsmall are shown as arrows and cylinders. β-Strands and α-helices involved in the Rossmann fold domain are shown in blue and cyan. Those in domain 2 are shown in red. C: Effect of crystal annealing. A comparison of diffraction from the same Hg derivative crystal before (left) and after (right) crystal annealing. The annealing cycle was repeated three times.

Humans and yeast possess a GidA homologue, Mto1.4, 5 Mto1 is a mitochondrial protein that plays a role in the biosynthesis of the 5-cmnm group of cmnm5s2U of mitochondrial tRNA.6 A subset of organisms has a second gidA gene, which encodes a smaller GidA protein. Thermus thermophilus HB8 has a small gidA encoding 232 amino acids (GidAsmall), in addition to the gidA of conventional size encoding 597 amino acids [Fig. 1 (B)]. The function of GidAsmall is unknown. No structural information about GidA and its homologues has yet been reported so far. To provide structural insight into the function of GidA, we determined the crystal structure of GidAsmall.

Materials and Methods.

Crystals of the T. thermophilus GidAsmall were obtained as previously described.7 The X-ray data were collected at the BL45XU8 and BL44B29 of SPring-8. Crystals were soaked for about 10 s in a mother liquor containing 23% glycerol as cryoprotectant, and then they were mounted in a cryo-gas stream at 90 K. Before the data collection, a few cycles of crystal annealing were carried out, in which the crystal was removed from the cryostream, incubated in the cryoprotectant solution for about 1 min, and then replaced into the nitrogen cryostream. The crystal structure of GidAsmall was determined at 1.65 Å resolution by the MAD method of the Hg derivative. The data were indexed, integrated, and scaled with HKL200010 and were further processed by using CCP4 suite programs.11

One mercury atom was located in one monomer of GidAsmall in a crystallographic asymmetric unit using the program SOLVE.12 RESOLVE13 was used for the density modification, assuming 46.3% solvent content. RESOLVE-refined phases yielded a good-quality experimental electron density map suitable for model building. Structure refinement was carried out with CNS.14 At this stage, strong clear density near well-conserved Tyr213 was recognized as a molecule of FAD. A summary of the refinement statistics is shown in Table I. The atomic coordinates and structure factors have been deposited in the PDB with accession code 2CUL.

Table I. Crystallographic Statistics
Crystal dataNativeHg remoteHg edgeHg peak
  • Values in parentheses are for the highest resolution shell

  • a

    Rsym = Σhkl{(Σi|Ihkl,i − 〈Ihkl〉|)/ΣiIhkl,i}.

  • b

    Figures of merit from SOLVE12 and after solvent flattering with RESOLVE.13

  • c

    Rwork = Σhkl||Fo| − |Fc||/Σhkl|Fo|. Rfree is calculated with 10% of the total reflections held aside throughout the refinement.

 Space groupP3221P3221  
 Unit-cell parameters (Å)    
  a = b78.5178.62  
Data collection    
 Wavelength (Å)1.00001.02001.00861.0080
 Resolution range (Å)40.0–1.6540.0–2.5040.0–2.5040.0–2.70
 No. of reflections1120529151287105970113425
 No. of unique reflections284631588882266741
 Rsym (%)a4.1 (11.3)9.9 (22.9)9.8 (17.2)9.5 (18.0)
 Completeness (%)99.1 (91.4)100 (100)96.8 (98.0)99.0 (99.2)
 MAD phasing    
  Mean figure of meritb 0.32, 0.49  
 Resolution (Å)30–1.65   
 Rworkc/Rfree (%)18.0/19.9   
 Average B-factor (Å2)    
  Main-chain atoms11.5   
  Side-chain atoms16.2   
  All protein atoms13.8   
  Bond length (Å)0.006   
  Bond angles (°)1.4   

Results and Discussion.

Crystal annealing.

Most of the native crystals and all of the derivative crystals gave poorly shaped diffraction spots when the crystals were frozen directly. Although one cycle of crystal annealing was not enough, a few cycles of it, described in Materials and Methods, significantly improved the shape of the diffraction spots, as well as the resolution limit [Fig. 1 (C)].

Overall structure of GidAsmall.

The crystals contained one GidAsmall molecule per asymmetric unit. Dynamic light-scattering experiments indicated that GidAsmall is monomeric in solution. Figure 2(A) shows the structure of GidAsmall. Although FAD was not added to the crystallization solvent, it was noncovalently bound to GidAsmall at full occupancy [Fig. 2(B)]. FAD was bound in an elongated conformation, with the isoalloxazine portion pointing toward the molecular surface of the protein. The phenyl ring of Tyr213, which is completely conserved among GidA proteins, stacks against the si face of the isoalloxazine ring of the FAD. The phenolic oxygen atom of Tyr213 is 3.6, 3.6, and 4.7 Å from the FAD atoms O4, N3, and N5.

Figure 2.

A: Ribbon diagram of the overall structure of GidAsmall. β-Strands and α-helices involved in the Rossmann fold domain are shown in blue and cyan, respectively. Domain 2 inserted in the Rossmann fold domain is shown in red. B: Fo − Fc omit map for FAD, contoured at 3σ generated after a cycle of simulated annealing without including the FAD. Residues and waters approximately positioned to form either hydrogen bonds or van der Waals contacts with FAD are shown. C: UV-vis absorbance spectra of GidAsmall collected before (orange) and after (black) adding dithionite.

The overall fold of GidAsmall consisted of two segments. The main segment [shown in blue and cyan in Fig. 2 (A)] belongs to a Rossmann fold, which is characteristic of a dinucleotide-binding fold. The remainder is shown in red in Figure 2(A). We refer to these two parts as the Rossmann fold domain and the domain 2, respectively. A structure similarity search using DALI15 identified about 30 structures, the z scores of which were >10.0. They were all oxidoreductases and included Rossmann folds. No similar fold to that of domain 2 was detected by DALI.

GidAsmall is an FAD-binding protein.

GidA proteins contain a well-conserved dinucleotide-binding consensus sequence at the N-terminus [Fig. 1 (B)]. This sequence is located on the loop connecting β1 and α1 in the Rossmann fold domain of GidAsmall, which is a feature commonly observed among FAD- and NAD(P)-binding Rossmann fold proteins. Furthermore, the topology of GidAsmall satisfies typical features of flavoproteins proposed by Vallon.16 It has been reported that GidA of Myxococcus xanthus binds FAD.17 The structure of GidAsmall clearly showed that this is a flavoprotein and that FAD is a genuine cofactor of GidA. The absorption spectra of GidAsmall confirms it.

GidAsmall shows two broad absorbance peaks that are characteristic of flavoprotein, and they disappear by addition of reductant dithionite [Fig. 2 (C)].

It remains unknown whether GidA catalyzes oxidation-reduction reactions. However, considering the above-mentioned points, GidA is likely to posses FAD-dependent redox activity, which would be required for the enzymatic reaction pathway of tRNA modification.

Prediction of the structure of GidA.

Figure 1(B) compares the amino acid sequences of GidAsmall and GidA proteins of conventional size. The 597-amino-acid GidA of T. thermophilus shares 48% and 47% identity with the GidA proteins of Thermotoga maritima and E. coli. GidA has two long inserted sequences, compared to GidAsmall [insert 1 and insert 2 shown in Fig. 1 (B)]. Insert 1 was found to be included in domain 2. Thus, GidA possesses a much larger domain 2 than GidAsmall because of the addition of insert 1. Insert 2 is located in the C-terminus. To predict the structure of GidA, the amino acid sequences around insert 1 and insert 2 were each analyzed by the fold recognition method such as 3D-PSSM,18 FUGUE,19 and mGenTHREADER.20 Although no possible fold was given a score reflective of a high level of confidence, the results of mGenTHREADER might be nonetheless noteworthy. mGenTHREADER gave Rossmann fold proteins as the top three scoring structures predicted from both sequences around insert 1 and insert 2. The highest-scoring structure predicted from the sequence around insert 1 was adenylylsulfate reductase (PDB code 1JNR).21 On the other hand, the highest-scoring fold calculated from the sequence around insert 2 was respiratory complex II-like fumarate reductase (PDB code 1QLA).22 The E-values given for these folds were 0.053 and 0.026, respectively, which indicate a “medium” confidence level. It is of interest that these proteins both belong to the succinate dehydrogenase/fumarate reductase family. They consist of three domains: a Rossmann fold domain, which binds FAD, a capping domain, and a helical domain. mGenTHREADER suggests that the structure of domain 2 including insert 1 of GidA could be similar to that of the capping domain of adenylylsulfate reductase [colored red in the left panel of Fig. 3 (A)]. On the other hand, mGenTHREADER suggested that the structure of the C-terminal region of GidA, including insert 2, could be similar to that of the helical domain of fumarate reductase [colored green in the right panel of Fig. 3 (A)]. The sequence identity between the corresponding areas of GidA and adenylylsulfate reductase, and GidA and fumarate reductase are 11.3% and 12.3%, respectively. The structural composition of GidA would be similar to that of the succinate dehydrogenase/fumarate reductase family, although the sequence identity is low. The catalytic site of this family is formed by the residues of the Rossmann fold domain and the capping domain.

Figure 3.

A: Structures of adenylylsulfate reductase (left)21 and fumarate reductase (right).22 The Rossmann fold domain is shown in blue. The areas that were predicted as possible folds for inserts 1 and 2 are colored red and green, respectively. B: The model for GidA, created by the Rossmann fold domain of GidAsmall (blue), the predicted structures of inserts 1 (red) and 2 (green). The models for inserts 1 and 2 were created by the alignment of the sequence of GidA onto the structures of the capping domain of adenylylsulfate reductase and the helical domain of fumarate reductase using the SCWRL3.0 program for the prediction of protein side-chain conformations.23C: Electrostatic potential mapped to the molecular surface of the GidA model in the same orientation as that shown in (B). Basic (blue) and acidic (red) regions colored according to the scale. Visualization, molecular surface building, and electrostatic potential calculations were carried out with GRASP.24 FAD is represented by a yellow space-filling model.

Figure 3(B) shows a model of the structure of GidA created by using the structure of the Rossmann fold domain of GidAsmall and the above-mentioned predicted structures. In the GidA model, conserved positively charged residues are localized on the surface where the isoalloxazine ring of FAD is exposed [Fig. 3 (C)]. It is reasonable to assume that direct interaction with tRNA takes place in this positively charged area, and this explanation may thus support the validity of the model. The GidAsmall lacking inserts 1 and 2 appears as a shortcut structure of GidA, and GidA would be the complete functional form required for tRNA modification.

The structure of GidAsmall presented here is the first structure of GidA proteins. It revealed that GidA is a flavoprotein and that the invariant Tyr213 plays the important role in FAD binding. The structure of GidAsmall also provided insights into the structure of GidA. The crystal structure of GidA and further biochemical studies are still required to achieve a full understanding of its reaction mechanism in the tRNA modification pathway.


The authors thank Drs. Y. Kawano and T. Hikima, and Mr. H. Nakajima for their kind help with the data collection. This study was performed under the “Structurome” Project of RIKEN at the Harima Institute. The protein sample preparation carried out by Prof. S. Kuramitsu, Dr. A. Ebihara, and their coworkers is greatly appreciated.