Crystal structure of the bovine mitochondrial elongation factor Tu.Ts complex.

The three-dimensional structure of the bovine mitochondrial elongation factor (EF)-Tu.Ts complex (EF-Tumt.Tsmt) has been determined to 2.2-A resolution using the multi-wavelength anomalous dispersion experimental method. This complex provides the first insight into the structure of EF-Tsmt. EF-Tsmt is similar to Escherichia coli and Thermus thermophilus EF-Ts in the amino-terminal domain. However, the structure of EF-Tsmt deviates considerably in the core domain with a five-stranded beta-sheet forming a portion of subdomain N of the core. In E. coli EF-Ts, this region is composed of a three-stranded sheet. The coiled-coil domain of the E. coli EF-Ts is largely eroded in EF-Tsmt, in which it consists of a large loop packed against subdomain C of the core. The conformation of bovine EF-Tumt in complex with EF-Tsmt is distinct from its conformation in the EF-Tumt.GDP complex. When domain III of bovine EF-Tumt.GDP is superimposed on domain III of EF-Tumt in the EF-Tumt.Tsmt complex, helix B from domain I is also almost superimposed. However, the rest of domain I is rotated relative to this helix toward domain II, which itself is rotated toward domain I relative to domain III. Extensive contacts are observed between the amino-terminal domain of EF-Tsmt and domain I of EF-Tumt. Furthermore, the conserved TDFV sequence of EF-Tsmt also contacts domain I with the side chain of Asp139 contacting helix B of EF-Tumt and inserting the side chain of Phe140 between helices B and C. The structure of the EF-Tumt.Tsmt complex provides new insights into the nucleotide exchange mechanism and provides a framework for explaining much of the mutational data obtained for this complex.

Protein biosynthesis is the process by which the ribosome translates the sequence of nucleotides in a mRNA into the sequence of amino acids in a protein. During the cyclic elongation phase, the ribosome is assisted by elongation factors (EFs) 1 (1,2). In prokaryotes, elongation factor EF-Tu promotes the binding of aminoacyl-tRNA (aa-tRNA) to the A-site of the mRNA-programmed ribosome in the form of the ternary complex aa-tRNA⅐EF-Tu⅐GTP (3). Upon cognate interaction between the codon of mRNA and the anticodon of aa-tRNA, GTP is hydrolyzed, and EF-Tu⅐GDP is released from the ribosome. The nucleotide exchange factor, EF-Ts, binds the EF-Tu⅐GDP complex mediating the release of the GDP and forms a stable EF-Tu⅐Ts complex (1). The high concentrations of GTP in the cell help dissociate EF-Ts leaving the active EF-Tu⅐GTP complex. EF-Tu⅐GTP binds another aa-tRNA, forming a new ternary complex, and the cycle is repeated.
Mitochondria contain a highly specialized protein biosynthetic machinery that is responsible for the synthesis of 13 polypeptides of the electron transport chain and the ATP synthase in the inner membrane (4). The mitochondrial translational system has a number of unique features including an altered genetic code, unusual protein-rich ribosomes, and tRNAs that lack many of the conserved residues found in canonical tRNAs (5)(6)(7). Despite these differences, mammalian mitochondria possess a translational elongation machinery with significant similarities to that of bacteria (8).
Mitochondrial EF-Tu (EF-Tu mt ) is highly conserved and is 55-60% identical to bacterial EF-Tu (9). The three-dimensional structure of the EF-Tu mt ⅐GDP complex has been determined at 1.94-Å resolution (10). The overall structure is similar to that observed in the Escherichia coli and Thermus aquaticus factors, but the nucleotide-binding domain (domain I) of EF-Tu mt is in a different orientation relative to the rest of the structure compared with that observed in prokaryotic EF-Tu (11,12). Furthermore, domain III is followed by a short 11-amino acid extension that forms one helical turn. This extension seems to be specific to the mitochondrial factors and has not been observed in any of the prokaryotic factors.
Bovine liver EF-Ts mt is 338 amino acids in length. The amino terminus of the mature protein has been determined and indicates that removal of the 55-residue import signal leaves a mature protein of 283 amino acids (13). Bovine EF-Ts mt is only 25-35% identical to its bacterial homologs, and primary sequence alignments are a challenge due to the limited conservation observed between the mitochondrial and prokaryotic factors.
Previous crystallographic studies have determined the structure of the EF-Tu⅐Ts complexes from E. coli and Thermus thermophilus (14,15). The E. coli complex is a heterodimeric complex, whereas the T. thermophilus complex is heterotetrameric. In these complexes, EF-Ts makes extensive contacts with both domain I and domain III of EF-Tu. E. coli EF-Ts is organized into four structural modules (14): the amino-terminal domain (residues 1-54), the core domain (residues 55-179 and 229 -263), the dimerization or coiled-coil domain (residues 180 -228), and the carboxyl-terminal module (residues 264 -282). The core domain can be further divided into subdomain N (residues 55-140) and subdomain C (residues 141-179 and 229 -263). The amino-terminal domain, subdomain N, and the carboxyl-terminal module interact with domain I of EF-Tu, whereas subdomain C interacts with domain III of EF-Tu. In the T. thermophilus (EF-Tu⅐Ts) 2 complex, each EF-Tu moiety interacts with two EF-Ts subunits through a bipartite interface, which explains the need for an EF-Ts dimer in the nucleotide exchange reaction (15).
The mechanism of guanine nucleotide exchange in these factors is suggested to occur through a three-part mechanism. The interaction between EF-Tu and EF-Ts results in the disruption of the Mg 2ϩ ion binding site, which leads to a reduced affinity of EF-Tu for guanine nucleotides. This process is induced by the intrusion of the side chains of Asp 80 and Phe 81 (E. coli numbering) from EF-Ts between helices B and C in domain I of EF-Tu. A peptide flip in the phosphate binding loop (P-loop) of EF-Tu destabilizes the binding of one ␤-phosphate oxygen of GDP. Finally, the movement of EF-Tu helix D affects the binding of the sugar and base of the bound nucleotide. The coordinated effects of this three-part mechanism lead to the release of GDP.
In the current study, we present the crystal structure of the bovine EF-Tu mt ⅐Ts mt complex in the absence of any bound nucleotide at 2.2-Å resolution. The structure provides the first picture of the structure of the mitochondrial nucleotide exchange factor and provides insight into the mechanism of nucleotide exchange for this protein complex.

MATERIALS AND METHODS
Subcloning and Expression-The ligation independent cloning technique (Novagen) was used to remove the carboxyl-terminal His 6 tag on mitochondrial Bos taurus EF-Tu mt (9). The plasmid was purified by the miniprep procedure (Qiagen) and used as a template for PCR amplification using forward primer 5Ј-GACGACGACAAGATGTGATAAGA-AGGAGATATACATATGGCTGTGGAGGCCAAGAAGACC-3Ј and reverse primer 5Ј-GAGGAGAAGCCCGGTTAACTCCACTTGATGTTC-TTGTCCTCC-3Ј. The sequence required for ligation independent cloning is shown in bold in both primers, the ribosomal entry site is underlined, and the start codon is shown in italics in the forward primer. A stop codon in the reverse primer is also shown in italics.
The PCR product was inserted in the pET30 Ek/LIC vector according to the manufacturer's instructions (Novagen). The annealed vector and insert were transformed into NovaBlue heat competent E. coli cells (Novagen). A clone containing the insert, determined by restriction enzyme digestion, was sequenced using the BigDye Terminator v.3.1 kit (Applied Biosystems), verifying the removal of the His 6 tag. Isolated plasmids were then transformed into electrocompetent E. coli BL21(DE3) cells for expression.
Instead of transforming the SeMet autotroph E. coli strain Bl21(DE3)834, the normal BL21(DE3) strain was used for the expression of SeMet protein. SeMet media (50 ml) containing kanamycin at 25 g/ml were inoculated with E. coli BL21(DE3) cells expressing untagged B. taurus EF-Tu mt and grown overnight at 37°C. The cells were spun down and resuspended in 100 ml of fresh SeMet media, which were used to inoculate 1850 ml of SeMet media containing 25 g/ml kanamycin. The cells grew at 37°C to an A 600 of 1.2-1.5, at which protein expression was induced with 0.25 mM isopropyl ␤-D-thiogalactopyranoside for 8 h (also at 37°C). The cells were collected by centrifugation and resuspended in 2 ml of lysis buffer (40 mM KCl, 50 mM Hepes-NaOH, pH 7.5, 10% (v/v) glycerol, 15 M GDP, 10 mM MgCl 2 , 3 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), and 0.5 mM NaN 3 ) per gram of wet cells and frozen at Ϫ20°C. The cells were ruptured in an EmulsiFlex-05 high pressure cell homogenizer (Avestin) at ϳ15,000 p.s.i. and ultracentrifuged at 50,000 rpm for 75 min at 4°C. The supernatant was filtered through a 0.45-m filter (Sartorius) and loaded on a 120-ml Source 30Q column (Amersham Biosciences) equilibrated in lysis buffer. The protein was eluted with a 400-ml linear gradient of 0 -25% S30Q-B buffer (lysis buffer containing 1 M KCl). The flow rate was 2 ml/min, and 5-ml fractions were collected. Fractions containing the protein of interest as determined by SDS-PAGE were pooled. The pooled protein was precipitated overnight with ϳ70% (NH 4 ) 2 SO 4 and collected by low-speed centrifugation. The pelleted protein was resuspended in buffer HIC A (1.5 M (NH 4 ) 2 SO 4 , 50 mM Hepes-NaOH, pH 7.5, 10% glycerol, 5 mM MgCl 2 , 15 M GDP, and 1 mM DTT). The pH of the buffer was adjusted to 7.5. A portion of the protein was loaded on a 1-ml ISO Resource column (Amersham Biosciences) equilibrated in HIC A. The purity of EF-Tu in the flow-through fractions was close to 95% as assessed by SDS-PAGE, and no further purification was necessary. Fractions containing the protein were pooled and concentrated in a YM-30 Centricon tube (Amicon), glycerol was added to a final concentration of 20%, and the sample was frozen in liquid nitrogen and stored at Ϫ80°C.
Purification of EF-Ts mt -A single colony of cells expressing carboxylterminally His 6 -tagged EF-Ts mt (13) was grown in 4 liters of LB media containing 25 g/ml kanamycin at 37°C while shaking until an A 600 of 0.7-0.9 was reached. Protein expression was induced with 0.25 mM isopropyl ␤-D-thiogalactopyranoside for 3.5 h (also at 37°C), and the cells were harvested by centrifugation at 4°C. The wet cells were resuspended in EF-Ts mt lysis buffer (150 mM KCl, 50 mM Tris-HCl, pH 7.6, 10% glycerol, 2 mM MgCl 2 , 15 M GDP, 2 mM ␤-mercaptoethanol, and 0.5 mM PMSF) to a total of 50 ml and stored at Ϫ20°C. The thawed cells were broken using the EmulsiFlex-05 high pressure cell homogenizer (Avestin). The lysed cells were subjected to ultracentrifugation at 50,000 rpm for 75 min at 4°C. The pH of the supernatant was adjusted to ϳ8 with Tris-HCl and applied to a nickel-nitrilotriacetic acid column (Amersham Biosciences) equilibrated in EF-Ts mt lysis buffer. The column was washed with NH 4 Cl wash buffer (1 M NH 4 Cl, 50 mM Tris-HCl, pH 7.6, 5 mM MgCl 2 , 10% glycerol, 15 M GDP, 2 mM ␤-mercaptoethanol, and 0.5 mM PMSF) in order to minimize the fraction of EF-Ts mt that would otherwise be associated with the E. coli EF-Tu. The E. coli EF-Tu⅐EF-Ts mt complex was eluted with EF-Ts mt elution buffer (40 mM KCl, 50 mM Tris-HCl, pH 7.6, 10% glycerol, 2 mM MgCl 2 , 15 M GDP, 2 mM ␤-mercaptoethanol, and 0.5 mM PMSF) containing 40 mM imidazole-HCl, pH 8. The remainder of the protein bound to the column was eluted with EF-Ts mt elution buffer containing 250 mM imidazole-HCl, pH 8. This latter portion contains mainly EF-Ts mt .
The eluted portion containing EF-Ts mt was clarified in a tabletop centrifuge at 15,000 rpm for 10 min at 4°C before being loaded on a 10-ml Source Q column (Amersham Biosciences) equilibrated in EF-Ts mt Source Q Buffer A (40 mM KCl, 50 mM Tris-HCl, pH 7.6, 2 mM MgCl 2 , 10% glycerol, 1 mM DTT, and 0.5 mM PMSF). The protein was eluted with a gradient from 0% to 50% Source Q Buffer B (Source Q Buffer A containing 0.5 M KCl). Fractions containing EF-Ts mt were pooled and concentrated in a YM-10 Centricon tube (Amicon). Glycerol was added to the concentrated protein to a final concentration of 20%, and the sample was stored at Ϫ80°C.
Crystallization-Samples containing untagged SeMet substituted EF-Tu mt and carboxyl-terminally His 6 -tagged EF-Ts mt were thawed on ice. The concentrations of both proteins were determined by the Bradford method (17) using bovine serum albumin as the standard. The two proteins (ϳ5 mg of each) were mixed in a molar ratio of 1:1.3 EF-Tu mt ⅐Ts mt and left on ice for approximately 1 h. The mixture was then dialyzed against 1.5 liters of 20 mM Hepes-NaOH, pH 7.5, 75 mM KCl, 5 mM EDTA-NaOH, pH 8, 1 mM DTT, and 0.5 mM NaN 3 overnight and dialyzed the next morning for 2 h against a new 0.5 liter of the same buffer. The 1.5-ml sample was then clarified in a tabletop centrifuge at 15,000 rpm for 10 min at 4°C and applied to a 120-ml (2 ϫ 60-cm) Sepharose S100 HR gel filtration column (Amersham Biosciences) equilibrated in buffer GF (20 mM Hepes-NaOH, pH 7.5, 75 mM KCl, 1 mM DTT, and 0.5 mM NaN 3 ). Fractions containing the complex were concentrated in a YM-50 Centricon tube (Amicon) to ϳ6.5 mg/ml, and crystallization trials were set up.
The SeMet substituted protein complex was crystallized by the sitting drop vapor diffusion technique by mixing 1 l of protein solution with 1 l of reservoir solution (14 -18% PEG 8k, 100 mM Tris-HCl, pH 7.6, 200 mM trisodium citrate⅐2H 2 O, 2 mM DTT, and 1 mM NaN 3 ) followed by equilibration against the reservoir at 6°C. One day after setup, the drop was streak-seeded using native crystals. The crystals appeared within 1 day after this and grew to a maximum of 500 ϫ 200 ϫ 100 m in 2-3 weeks.
Data Collection and Processing-A crystal of SeMet substituted protein was transferred to stabilization buffer (20% (w/v) PEG 8k, 200 mM trisodium citrate⅐2H 2 O, 100 mM Hepes-NaOH, pH 7.4, and 10 mM DTT) and transferred 2 h later to cryobuffer (22% (w/v) PEG 8k, 200 mM trisodium citrate⅐2H 2 O, 100 mM Hepes-NaOH, pH 7.4, 10 mM DTT, and 10% glycerol). After 18 h in this buffer, the crystal was frozen in a stream of liquid nitrogen gas at 100 K at the EMBL BW7a beamline at the DORIS storage ring, DESY (Hamburg, Germany). An x-ray fluorescence spectrum at the selenium K absorption edge was obtained by tuning the wavelength of the incident x-ray beam and used to determine the wavelengths necessary for subsequent data collection. A three-wavelength dataset was collected at the selenium peak ( ϭ 0.9818 Å), inflection point ( ϭ 0.9824 Å), and high energy remote ( ϭ 0.9716 Å), in that order. Data (360°) were collected for each wavelength, with 1°oscillation in each frame, but only 220°of data for each wavelength were used for structure determination. The data were processed with Denzo and Scalepack (18). The crystal belongs to space group C2 and has cell dimensions a ϭ 94.0 Å, b ϭ 149.3 Å, c ϭ 70.0 Å, and ␤ ϭ 125.2°. The asymmetric unit contains one EF-Tu mt ⅐Ts mt complex and has a solvent content of 52%.
Structure Determination-MAD phase calculation using all three wavelengths and subsequent structure refinement were carried out in CNS (19). Eleven of 12 possible SeMet sites in EF-Tu mt were successfully located. The MAD phases were used as MLHL restraints in the initial cycles of refinement to 2.5-Å resolution. The density-modified experimental three-wavelength electron density map calculated at 2.5Å was easily interpretable. Upon refining to 2.2 Å, the target was switched to MLF, and refinement was carried out against the remote data without anomalous contribution. Models were visualized in O (20). Residues 44 -55 and 84 -111 of EF-Tu mt had no density initially and could not be modeled after refinement. Residues 112, 117-119, 225-228, 251-256, 295-298, 310 -314, and 452 of EF-Tu mt initially had badly defined density but could be built after refinement. The region of EF-Tu mt containing residues 182-187 had bad density even after refinement indicated by their high B-factors. SeMet 375 of EF-Tu mt was observed to have a double conformation and was refined accordingly in the last step of refinement. EF-Ts mt had poor density for residues 187-193 and no density for 330 -338. During refinement, residues 187-193 and 330 -331 were built. Residues 332-338 were the only ones that could not be built. A total of 322 water molecules were assigned to the asymmetric unit. The quality of the final model was inspected with PROCHECK (21), and the secondary structure was assigned with DSSP (22). The final model, having been refined from 30.0 to 2.2 Å, has a R work ϭ 21.8% and a R free ϭ 24.7%. The root mean square deviation for bond lengths is 0.007 Å and 1.3°for bond angles. There are no outliers in the Ramachandran plot, and there are only 4 residues in the generously allowed region from a total of 645 residues. See Table I for data collection and refinement statistics. Fig. 1, A and B, and Figs. 4 -6 were made with PyMOL (23).

RESULTS
Overall Structure-The crystallographic asymmetric unit contains one heterodimeric complex of EF-Tu mt ⅐Ts mt , and 322 water molecules could be modeled. The overall structure of the complex (Fig. 1A) is similar to that of the E. coli EF-Tu⅐Ts complex (14). As in the E. coli complex, the amino-terminal domain and subdomain N of EF-Ts mt contact domain I of EF-Tu mt , whereas subdomain C interacts with domain III. The buried surface area, when considering these contacts, constitutes 3140 Å 2 in the mitochondrial complex compared with 2623 Å 2 in the E. coli complex. When also considering the contacts made between domain I and the carboxyl-terminal module in the E. coli complex, the buried surface area in this complex is 3624 Å 2 . This kind of interaction is not observed in the mitochondrial complex because EF-Ts mt is shorter than E. coli EF-Ts and the last 7 residues could not be modeled in the mitochondrial protein.
EF-Ts mt -The structure of EF-Ts is known for both the E. coli and T. thermophilus factors (14,15). The structure of E. coli EF-Ts has been divided into five domains, the aminoterminal domain, subdomains N and C of the core region, the coiled-coil region (which is part of the dimerization domain), and the carboxyl-terminal module. There is a duplication of the secondary structure motif in subdomain N and C that gives rise to a local internal pseudo-symmetry ( Fig. 1, B and C). The corresponding structure of the dimeric T. thermophilus EF-Ts is created by the three-stranded anti-parallel ␤-sheet on each subunit interacting to form a truly symmetric ␤-sandwich corresponding to the ␤-sandwich structure observed between subdomain N and C of the core of E. coli EF-Ts (24).
The basic structure of EF-Ts mt has interesting similarities and also rather striking differences compared with that of the E. coli factor (Fig. 1, B and C). In contrast to T. thermophilus EF-Ts, both E. coli and mitochondrial EF-Ts function as monomers in solution, and comparisons will therefore be made between the mitochondrial and E. coli factors. Overall, the aminoterminal domain is similar between the two factors. The two where F obs and F calc are the observed and calculated structure factor, respectively, and k is a scaling factor. e R free is identical to R on a subset of test reflections not used in refinement. f r.m.s.d., root mean square deviation.
structures then begin to show some deviations. The cores of E. coli EF-Ts and EF-Ts mt are both formed by a ␤-sandwich. However, the number of ␤-strands and their arrangement differ between them. Most dramatically, the coiled-coil domain found in E. coli EF-Ts (helices ␣9, ␣10, and ␣11 in Fig. 1B) is almost completely eroded in the mitochondrial factor (Fig. 1A, loop region between ␣8 and ␣12).
The amino-terminal domain of bovine EF-Ts mt is structurally similar to the amino terminus of the EF-Ts from both E. coli and T. thermophilus (14,15). It consists of the helical segments ␣1-␣3 (Fig. 1, A and C). The insertion of 4 extra residues in the bovine protein located in helix ␣3 adds one helical turn to this helix and moves the amino-terminal domain slightly closer to helix D of EF-Tu mt compared with EF-Ts in the two other complexes. Helix ␣3 has a considerable kink in its middle at sAla 99 (in the following sections, residues will be referred to by the prefixes u and s if they belong to EF-Tu or EF-Ts, respectively). The kink in the mitochondrial factor is more pronounced than that seen in E. coli at sGly 44 , whereas ␣3 is only slightly bent in T. thermophilus EF-Ts.  1. A, cartoon representation of the bovine EF-Tu mt ⅐Ts mt complex. EF-Tu mt is colored orange, and EF-Ts mt is colored violet (amino-terminal domain), green (subdomain N), and blue (subdomain C). The amino terminus and carboxyl terminus of both proteins are labeled along with domain I, II, and III and helices B, C, and D of EF-Tu mt and the secondary structure elements of EF-Ts mt . Not shown in cartoon are four 3 10 -helices of EF-Ts mt (see the Fig. 2 legend). B, cartoon of E. coli EF-Ts (helix ␣13 and connecting loop have been omitted for clarity) (14). The overall structure of the E. coli EF-Ts molecule is very similar to that of bovine EF-Ts mt . It has been colored as described for the domains of EF-Ts mt , and additionally, the E. coli EF-Ts coiled-coil motif consisting of helices ␣9, ␣10, and ␣11 is shown in red. C, secondary structure projection of the bovine EF-Ts mt and E. coli EF-Ts structures. Secondary structure elements common with E. coli EF-Ts are named as in E. coli EF-Ts. In bovine EF-Ts mt , the amino-terminal domain consists of ␣1, ␣2, and ␣3; subdomain N consists of ␤1, ␤2, ␣4, ␣5, ␤2Ј, ␣6, ␣7, ␤3, and ␤4Ј; and subdomain C consists of ␤4, ␤5, ␣8, ␣12, and ␤6.
Subdomain N of EF-Ts mt consists of four helical segments (␣4-␣7) and a five-stranded ␤-sheet (␤1, ␤2, ␤2Ј, ␤3, and ␤4Ј) (Fig. 1, A and C). In E. coli EF-Ts, subdomain N of the core consists of four helical segments (␣4-␣7) and a three-stranded ␤-sheet (␤1-␤3) (Fig. 1B). Helix ␣3 from the amino-terminal domain leads into ␤-strand ␤1 in subdomain N of the core. This strand makes a turn and runs anti-parallel to strand ␤2. Strand ␤2 is followed by a short helix, ␣4, which contains the important aspartic acid and phenylalanine residues that play a role in the nucleotide exchange reaction. This section then leads into helix ␣5. In E. coli EF-Ts, ␣5 is followed by another helical segment (␣6). However, in EF-Ts mt , there is a large insertion of residues between helices ␣5 and ␣6, creating an extra ␤-strand (␤2Ј) consisting of five residues that runs antiparallel to ␤-strand ␤3 (Fig. 1C). The polypeptide chain then turns ϳ90°perpendicular to the ␤-sheet and forms a short helix (␣6) consisting of 5 residues. This helix is followed by a loop remarkably similar in size and shape to the one found in T. thermophilus EF-Ts between helices ␣7 and ␣8 (15) and distinct from the corresponding region in E. coli EF-Ts, where ␣6 leads directly into ␣7 (Figs. 1 and 2). The loop in EF-Tu mt is followed by a helix that corresponds to ␣7 in E. coli EF-Ts. This helix leads into the last ␤-strand, ␤3, of the ␤-sheet in subdomain N.
Subdomain C of the core of EF-Ts mt begins where strand ␤3 leads into strand ␤4 (Fig. 1). An insertion of 3 residues between ␤3 and ␤4 compared with E. coli EF-Ts extends these strands in bovine EF-Ts mt and stabilizes the amino terminus of strand ␤2Ј in the mitochondrial factor. In E. coli EF-Ts, strand ␤4 leads directly into strand ␤5. In EF-Ts mt , there is another insertion between ␤4 and ␤5 that folds back across the pseudo 2-fold symmetry axis of EF-Ts to subdomain N of the core domain. It forms a loop and a small ␤-strand, ␤4Ј, consisting of 3 residues. It is fixed in place by hydrogen bonds to ␤1, and the chain then returns back to form strand ␤5 in subdomain C of the core domain. This organization means that the ␤-sheet in subdomain N of the core consists of five ␤-strands, instead of three ␤-strands as in the other two EF-Tu⅐Ts complexes. Residues 255-262 form a loop between ␤5 and ␣8 that contains a 3 10 -helix (residues 259 -261) (data not shown). This loop is similar to but 2 residues shorter than the one seen in T. thermophilus EF-Ts between ␤2 and ␣5 of the second EF-Ts moiety of the (EF-Tu⅐Ts) 2 complex (15). In T. thermophilus, this region contains the conserved TDFV sequence motif that interacts with EF-Tu.
The region between helices ␣8 and ␣12 corresponding to the coiled-coil segment in E. coli EF-Ts is reduced to a large loop in bovine EF-Ts mt that packs against subdomain C of the core domain (Fig. 1A). Due to the less well defined density at the carboxyl terminus of mitochondrial EF-Ts mt following sGly 331 , the last 7 residues of EF-Ts mt could not be modeled. In the E. coli EF-Ts structure, this region is followed by a loop and an additional helix, ␣13, which interacts with domain I of EF-Tu. The carboxyl terminus of T. thermophilus EF-Ts is found at the position to which EF-Ts mt could be modeled.
The EF-Tu mt ⅐Ts mt Interface-There are basically three major areas of contact between EF-Tu mt and EF-Ts mt , two regions involving contacts with domain I and one major region contacting domain III of EF-Tu mt . The amino-terminal domain of EF-Ts mt interacts with residues in helix D and the loop leading into helix C of domain I of EF-Tu mt . Subdomain N of the core contacts EF-Tu mt through residues in helix C and the region between helices B and C. Finally, subdomain C of the core makes contacts with domain III of EF-Tu mt .
Contacts between the amino-terminal domain of EF-Ts mt and domain I of EF-Tu mt result in a significant displacement of helix D, which moves away from domain III. Helix ␣1 from the amino terminus of bovine EF-Ts mt interacts primarily with helix D of EF-Tu mt (Fig. 3A). The methylene groups from the side chain of sLys 60 , sLeu 63 , sMet 64 , and sPhe 74 form a hydrophobic pocket for uLeu 194 . Numerous electrostatic contacts are also present (Fig. 3A). The side chain of uGlu 190 is within hydrogen bonding distance with the amino-terminal amine of sSer 56 (the amino-terminal residue of the mature protein) and its carbonyl oxygen (Fig. 3A). Residue sLys 60 forms a salt bridge to uGlu 193 O⑀ 1  and sIle 75 from the amino terminus of helix ␣2 of EF-Ts mt makes hydrogen bonds with the backbone carbonyl of residue uAsn 154 from the loop leading to helix C of EF-Tu mt .
Subdomain N of the core of EF-Ts mt plays a critical role in the rearrangements in EF-Tu mt that lead to the release of the bound nucleotide. Residue sPhe 140 from the conserved TDFV sequence motif found in all EF-Ts sequences is inserted directly between helices B and C of EF-Tu mt as observed in the E. coli and T. thermophilus complexes (Figs. 3B and 6). The conserved Asp residue (sAsp 139 in EF-Ts mt ) makes tight hydrogen bond contacts with the backbone amides of uHis 130 (2.8 Å) and uAla 131 (3.1 Å) of helix B (Fig. 3B). Residue sAsn 144 in the loop between ␣4 and ␣5 of EF-Ts mt makes a hydrogen bond contact with uGlu 163 in helix C of EF-Tu mt . In addition, the backbone carbonyl of sLeu 206 at the very end of ␣7 in EF-Ts mt makes hydrogen bonds with the side chain of uGln 160 at the end of helix C in EF-Tu mt (Fig. 3B). A slight difference in the coordination of the P-loop in domain I of EF-Tu by EF-Ts can be observed among the mitochondrial, E. coli, and T. thermophilus complexes. The mitochondrial residues sArg 109 and sArg 113 from ␣3 coordinate the carboxyl side chain of mitochondrial uAsp 67 . In E. coli, sLys 51 coordinates uAsp 21 , whereas T. thermophilus EF-Ts has both sLys 50 and sArg 53 contacting uAsp 21 .
Subdomain C of the core of EF-Ts mt contacts domain III of EF-Tu mt . The relative position of domain III of EF-Tu mt in relation to subdomain C of the EF-Ts mt core is different from that in the E. coli complex. When helices B and C from EF-Tu mt and the regions from the core of EF-Ts mt that interact with EF-Tu mt are superimposed, there is a rotation of ϳ18°of domain III around helix B in a clockwise manner when looking from the amino terminus toward the carboxyl terminus of this helix. The difference is minimal at the interface to EF-Ts mt and progresses further away from the domain III/EF-Ts mt interface. In comparison, the E. coli and T. thermophilus complexes superimpose almost perfectly for domains II and III. A hydrophobic patch on EF-Tu mt consisting of the methylene groups of uLys 368 , uPro 369 , uVal 371 , uLeu 397 , and uMet 399 faces residues sGly 275 , sMet 276 , sLeu 302 , and sLeu 303 on EF-Ts mt . The carboxylic oxygens of uGlu 396 are held in place by interactions from sGln 271 N⑀ 2 and sArg 267 N 2 and more weakly by sArg 268 N⑀ and sArg 268 N 2 (Fig. 3C). There is also contact between sHis 231 at the end of ␤4 and uGlu 402 in the loop between strands c3 and d3 in EF-Tu mt . The side chain hydroxyl group of sTyr 229 stabilizes sHis 231 through a hydrogen bond to N␦ 1 .
EF-Tu mt -The conformation of bovine EF-Tu mt in complex with EF-Ts mt is distinct from the EF-Tu mt ⅐GDP complex (10). When superimposing domain III of bovine EF-Tu mt in the GDP and EF-Ts mt complexed forms, helix B in domain I in the two structures is almost superimposed (Fig. 4A). However, the remainder of domain I is rotated ϳ25°as a rigid body relative to this helix away from domain III toward domain II. The solventaccessible surface of EF-Tu mt from the EF-Ts mt complex decreases by 140 Å 2 compared with the GDP-bound form as a result of the movement of the carboxyl-terminal part of helix C away from domain III. Besides the changes in the P-loop and near the guanine nucleotide binding site, which will be described later, there are no changes in this domain. This is similar to the changes observed in domain I of E. coli EF-Tu between the GDP and EF-Ts complexed forms (12,14). Surprisingly, domain II of EF-Tu mt in the EF-Tu mt ⅐Ts mt complex is also rotated ϳ18°as a rigid body toward domain I relative to domain III when compared with the EF-Tu mt ⅐GDP complex (Fig. 4A). This rotation is exemplified by the movement of the loop between ␤-strands b2 and c2 in domain II by ϳ7.8 Å. The shape of the "hole" between the three domains of EF-Tu mt is slightly altered, mainly as a result of the relative movements of domain I and the linker between domains I and II. The rotation of domain II is probably a result of the movement of domain I, which mediates its effect through the linker between these two domains, because there are no major changes in domain III.
As a result of the domain rearrangements, some of the interface interactions between domain II and III have changed upon EF-Tu mt ⅐Ts mt complex formation from the EF-Tu mt ⅐GDP complex (Fig. 4B). In the EF-Tu mt ⅐GDP structure, residues uHis 295 and uSer 296 , from the loop between ␤-strands e2 and f2, of domain II interact with residues uMet 385 , uGln 412 , uPro 413 , and uMet 414 from domain III (10). Residue uHis 295 is held in place by uMet 385 and uMet 414 from domain III and uArg 335 from domain II. In the EF-Tu mt ⅐Ts mt complex, the peptide bond between uGly 294 and uHis 295 has flipped ϳ170°, causing the following residues in the loop to assume a different conformation than that observed in the EF-Tu mt ⅐GDP complex. The relative movement of the C ␣ atoms of uHis 295 and uSer 296 is 6.1 and 5.8 Å, respectively, when domain III is superimposed in the two structures. In the equivalent E. coli structures, there is essentially no change between the relative positions of domain II and III upon complex formation with EF-Ts (12,14). The interface between domains II and III is also basically the same in the E. coli EF-Tu⅐GDPNP⅐Phe-tRNA⅐kirromycin complex, 2 indicating that in the bacterial EF-Tu, these domains act together as a rigid body. The loop between ␤-strands e2 and f2 in domain II of EF-Tu mt is similar but not identical in shape or position to the corresponding region in domain II of the T. thermophilus EF-Tu⅐Ts complex (15).
Rearrangements Leading to Guanine Nucleotide Exchange-The dramatic conformational changes observed in domain I of EF-Tu mt are critical for the nucleotide exchange process. The changes lead to the disruption of the binding site for the guanine nucleotide. The changes observed encompass portions of EF-Tu mt that interact with the guanine base and regions binding the ␤-phosphate and Mg 2ϩ ion. The GTP-binding site is defined by three consensus sequence elements (for review, see Ref. 26). The first sequence motif, GXXXXGK(S/T), also called the P-loop, coordinates the ␣and ␤-phosphate groups of the bound nucleotide. The second motif, DXXG, which is part of the switch II region, is involved in the coordination of the ␥-phosphate. The third motif, NKXD, determines the specificity for the guanine base. Furthermore, a conserved threonine residue from the switch I region coordinates the Mg 2ϩ ion.
The mechanism of guanine nucleotide exchange in EF-Tu is suggested to occur through a three-part mechanism. The interaction between EF-Tu and EF-Ts results in the disruption of the Mg 2ϩ ion binding site that is induced by the intrusion of the side chains of Asp 80 and Phe 81 (E. coli numbering) from EF-Ts between helices B and C in domain I of EF-Tu. A peptide flip in the P-loop of EF-Tu destabilizes the binding of one ␤-phosphate oxygen of GDP, and the movement of EF-Tu helix D affects the binding of the sugar and base of the bound nucleotide.
The region in EF-Tu mt (K 182 ADAVQ 187 ) containing part of the conserved NKXD motif involved in nucleotide binding was very difficult to model and had less well defined density even after the final refinement. The orientation of this loop is different from that of the EF-Tu mt ⅐GDP complex (Fig. 5). In E. coli, sLys 23 in the amino-terminal domain of EF-Ts makes an electrostatic contact with uAsp 141 from the loop following the NKXD motif. In EF-Tu mt , the aspartic acid residue found in E. coli EF-Tu (uAsp 141 ) is replaced by uGln 187 that points in a different direction than in the E. coli protein. The side chain of uGln 187 forms two hydrogen bonds with the side chain of uAsn 154 in the mitochondrial EF-Tu mt ⅐Ts mt complex. The mitochondrial residue equivalent to E. coli sLys 23 (sLys 78 ) is conserved in EF-Ts mt but makes a salt bridge with uAsp 188 (3.6 Å). A similar orientation of the equivalent of EF-Tu mt uLys 182 is seen in two other G-proteins complexed with their nucleotide exchange factor, namely, that of Ras⅐Sos1 and Ran⅐RCC1 (27,28). The conformation observed in the EF-Tu mt ⅐Ts mt complex could represent a state in the exchange mechanism distinct from that observed in the E. coli and T. thermophilus complexes.
As mentioned, two of the dominant residues in the interaction between EF-Tu and EF-Ts are from the conserved TDFV motif in subdomain N of the core of bovine EF-Ts mt , namely, sAsp 139 and sPhe 140 (Figs. 3B and 6). The phenyl ring of sPhe 140 sits in a hydrophobic region with uTyr 133 and uLeu 167 and between the side chains of uHis 130 and uHis 164 . This position inserts sPhe 140 between helices B and C and clearly disrupts the binding site for the Mg 2ϩ ⅐GDP. Residues 84 -111 containing the switch I region of EF-Tu mt could not be modeled.
This region is involved in large conformational changes between the GDP-and GTP-bound state of EF-Tu (29,30). A peptide flip between residues uVal 66 and uAsp 67 causes the P-loop to change conformation as seen in other complexes of G-proteins and their exchange factors (14,15,31).
Mutations versus Structure-A number of mutations have been made in residues of EF-Ts mt that were predicted to interact with EF-Tu mt . These predictions were based on the alignment of the sequences of the bacterial and mitochondrial factors and on the structure of the E. coli EF-Tu⅐Ts complex (14,(32)(33)(34). The variants made have been divided into three groups, based on their location in the structure of EF-Ts (Table  II and Fig. 6). The first group is located in the amino-terminal domain. In the E. coli EF-Tu⅐Ts complex, sArg 12 makes an electrostatic contact with uGlu 152 and a hydrogen bond with the backbone carbonyl of uPro 111 (3.4 Å). It plays an essential role in the interaction of these two factors (32). The corresponding position in EF-Ts mt is sArg 67 . This residue makes electrostatic interactions with uGlu 198 and a hydrogen bond contact with the backbone carbonyl of uPro 157 (2.9 Å). The adjacent residue sArg 68 makes salt bridge contacts with uGlu 201 in helix D of EF-Tu mt (Fig. 3A). However, a triple mutant in this region of mitochondrial EF-Ts mt (R67A, R68A, and K69A) that eliminates these interactions is as active as wild-type EF-Ts mt . Hence, unlike the E. coli complex, these interactions between EF-Tu mt and EF-Ts mt do not contribute significantly to the guanine nucleotide exchange in the mitochondrial system.
Contacts between helix ␣2 in the amino-terminal domain of EF-Ts mt and helix D in domain I of EF-Tu mt also were predicted to involve residues sPhe 74 and sIle 75 . In the crystal FIG. 4. A, superposition of EF-Tu mt domain III from EF-Tu mt ⅐GDP and the EF-Tu mt ⅐Ts mt complex viewed from the EF-Tu mt ⅐Ts mt interface. EF-Ts mt has been removed for clarity. EF-Ts mt complexed EF-Tu mt is colored orange, and EF-Tu mt in the GDP conformation is shown in wheat. Residues 84 -111 of the EF-Tu mt ⅐GDP complex have been omitted for clarity. Helices B, C, and D of domain I of EF-Tu mt are labeled. Helix B from domain I superimposes well in the two conformations. The rest of domain I is rotated/twisted away from domain III by ϳ25°. Note the movement of domain II in the background. B, stereo-view of the interface between domain II and III of EF-Tu mt around the loop of domain II marked by an asterisk in A (not all secondary structure is shown). Domain II is seen at the left, and domain III is seen at the right, with the same coloring as described in A. Selected residues have been labeled. The C ␣ atom of uHis 295 moves by ϳ6.1 Å. See "Results" for further explanation. structure of the bovine EF-Tu mt ⅐Ts mt complex, sPhe 74 makes hydrophobic contacts with the side chains projecting from helix D of EF-Tu mt , although it is not tightly nestled in this pocket. The backbone amide group of sPhe 74 forms a weak hydrogen bond to the backbone carbonyl oxygen of uAsn 154 on EF-Tu mt (3.3 Å). The backbone amide group of sIle 75 also hydrogen bonds to the backbone carbonyl oxygen of uAsn 154 (3.2 Å), and its side chain may make a hydrophobic contact with the C ␤methylene of this residue. Three different mutants in these residues have been made (Table II) (34). Conversion of both sPhe 74 and sIle 75 to alanine results in a 2-fold reduction in the activity of EF-Ts mt , suggesting that the hydrophobic interactions play some role in stabilizing the interaction of EF-Ts mt with EF-Tu mt . Replacement of both residues with methionine converts this sequence into one identical to that observed in E. coli EF-Ts. This variant has an activity indistinguishable from that of the wild-type factor. Clearly, a methionine at both positions is sufficient to maintain strong contacts between these two factors, allowing GDP exchange to occur readily. It is essential to maintain hydrophobic residues in this region because replacement of sPhe 74 and sIle 75 with glutamic acid residues completely inactivates EF-Ts mt .
Mutational analysis shows that mutation of sLys 78 and sLys 79 to alanine in EF-Ts mt has no effect on the activity of EF-Ts mt (Table II). Based on the structural information, the mutation of sLys 78 could be expected to destabilize the interaction of EF-Ts mt with the amino-terminal of helix D following the NKXD motif in EF-Tu mt . Apparently, however, this contact is not important. The adjacent sLys 79 points away from EF-Tu mt , and its mutation is not expected to have an effect on the activity of EF-Ts mt .
Mutation of either sAsp 139 or sPhe 140 from the conserved TDFV motif of EF-Ts mt to alanine reduces the activity of EF-Ts mt about 5-fold (Table II). This observation indicates that disruption of this region of EF-Tu mt plays a role (although not an essential one) in the ability of EF-Ts mt to promote guanine nucleotide exchange on EF-Tu mt . Replacement of both sAsp 139 and sPhe 140 with alanine completely inactivates EF-Ts mt , demonstrating that EF-Ts mt must be able to disrupt the region between helices B and C in EF-Tu mt to promote nucleotide exchange. The mutation of both these residues to alanine does not completely inactivate E. coli EF-Ts, indicating that they are not essential for the function of the bacterial factor (33).
EF-Ts mt interacts with domain III of EF-Tu mt over a significant surface (924 Å 2 ). One residue predicted to be important in this region of contact is sHis 231 located in subdomain C of the core of EF-Ts mt (Figs. 3C and 6). The side chain of this residue interacts with the loop between ␤-strands c3 and d3 of domain III. The N⑀ 2 of the imidazole ring of sHis 231 hydrogen bonds to an oxygen of uGlu 402 , and N␦ 1 is hydrogen bonded to sTyr 229 , whereas hydrophobic contacts are made with the side chains of uLeu 397 and uMet 399 . Mutation of sHis 231 to alanine results in the loss of the side chain hydrogen bond and leads to a 6-fold reduction in the ability of EF-Ts mt to stimulate the activity of EF-Tu mt (Table II). Subdomain C also makes significant hydrophobic contacts with domain III of EF-Tu mt . Two of these contacts involve sLeu 302 and sLeu 303 . Conversion of these residues to valine and methionine, respectively (the residues found in E. coli EF-Ts), had no effect on the activity of EF-Ts mt . This observation is reasonable because these changes maintain the hydrophobic interactions between the two factors. However, conversion of the two residues to glutamic acids completely inactivated EF-Ts mt , indicating that the hydrophobic contacts between EF-Ts mt and domain III of EF-Tu mt are essential for their interaction, as seen in the E. coli complex (34).
There are hydrogen bond contacts between sArg 267 and sArg 268 of EF-Ts mt and uGlu 396 of EF-Tu mt . However, mutation of both sArg 267 and sArg 268 to alanines did not affect the ability of EF-Ts mt to promote guanine nucleotide exchange with EF-Tu mt , indicating that these contacts do not play a significant role in the interaction between these two factors. DISCUSSION One of the striking differences between EF-Ts mt and E. coli EF-Ts is the loss of the majority of the coiled-coil domain in the mitochondrial factor. The deletion of the coiled-coil motif in E. coli EF-Ts affects its ability to compete with guanine nucleotides for binding to EF-Tu (35). The concentration of either GDP or GTP needed to dissociate the mutant EF-Tu⅐Ts complex was 2 orders of magnitude lower than that for the wildtype complex. Furthermore, an E. coli strain harboring the EF-Ts mutant is resistant to phage Q␤ (36). The Q␤ complex is responsible for the replication of the single-stranded RNA genome of coliphage Q␤. Both EF-Tu and EF-Ts are components of the Q␤-polymerase complex along with the Q␤-replicase subunit and ribosomal protein S1 (37)(38)(39). The region that corresponds to the T. thermophilus coiled-coil region of one monomer is reduced in size in bovine EF-Ts mt and forms a ␤-strand, ␤2Ј, in subdomain N. The coiled-coil region of the second EF-Ts moiety in the T. thermophilus EF-Tu⅐Ts complex, which is equivalent to the coiled-coil motif of E. coli EF-Ts, is reduced to a large loop packing against subdomain C of bovine EF-Ts mt . The coiled-coil motif is conserved in bacteria and chloroplasts. It has been suggested that the coiled-coil motif of E. coli EF-Ts is involved in an isomerization step within EF-Ts during the formation of the EF-Tu⅐GDP⅐EF-Ts complex that acts as an intermediate in the nucleotide exchange reaction (35). This type of isomerization step has been studied with the nucleotide exchange reaction taking place with the ternary EF-Tu⅐thioGDP⅐EF-Ts complex (40). Furthermore, it has been suggested that the binary EF-Tu⅐Ts complex and the ternary complex with either GDP or GTP are structurally distinct (41). Clearly, the absence of the coiled-coil motif in EF-Ts mt suggests that any isomerization reaction occurring with the mitochon- FIG. 5. Conformational changes at the guanine base binding site in domain I of EF-Tu mt . EF-Tu mt ⅐GDP and the EF-Tu mt ⅐Ts mt complex are colored wheat and orange, respectively. The GDP moiety is shown in gray stick representation, with nitrogens colored blue, oxygens colored red, and phosphorus colored yellow. Residues uAsn 181 , uLys 182 , and uAsp 184 from the NKXD motif (also shown in stick representation), along with helices A, C, and D and the P-loop, have been labeled. The loop following the NKXD motif in EF-Tu mt is in a distinct conformation after complex formation with EF-Ts mt . The P-loop, as in other G-protein⅐nucleotide exchange factor complexes, is seen to have a different conformation due to a peptide flip in this loop. drial factor must involve other parts of the molecule.
The observation that the EF-Ts proteins have the highest sequence identity in the amino terminus obviously signifies the importance of this domain in the nucleotide exchange reaction. The fact that mutation of the aspartic acid and phenylalanine residues from the TDFV motif in E. coli and bovine mitochondrial EF-Ts yielded different results reflects the idea that the exchange reaction is a result of several effects mediated by EF-Ts, perhaps mainly by the amino terminus. The relative importance of each effect appears to vary with the factors from different translational systems.
The movement of part of the NKXD motif and the following loop could be a significant contribution to the exclusion of GDP from EF-Tu mt , and other G-proteins in general. The Ran⅐RCC1 complex is suggested to release the base of the guanine first, followed by the phosphate moiety (28). This complex and that of Ras⅐Sos1 (27) both posses a NKXD motif that is structurally different from that seen in E. coli and T. thermophilus EF-Tu⅐Ts but somewhat similar to the one in the EF-Tu mt ⅐Ts mt complex. However, the weak density in this region of our structure does not convincingly suggest the specific mechanism at work. A higher mobility of the guanine binding region was suggested as the main reason for the low affinity of EF-Tu mt for nucleotides (10).
The interaction between EF-Ts mt and EF-Tu mt ⅐GDP leads to quite large conformational changes in EF-Tu mt . These conformational changes involve a considerable shift in the interactions of the three domains of EF-Tu mt . Of particular interest is the difference in the interface between domain II and III of EF-Tu mt , as well as their relative position between the GDP and EF-Ts mt complexed forms. These differences have not been observed in the other EF-Tu⅐Ts complexes from E. coli and T. thermophilus (14,15). Thus, in the bovine EF-Tu mt ⅐Ts mt complex, domains II and III do not act as a rigid body as observed in E. coli and T. thermophilus EF-Tu.
When superimposing the parts of EF-Ts that interact with EF-Tu along with EF-Tu helices B and C of domain I from E. coli, T. thermophilus, and B. taurus, there is a relatively large shift in the position of domain III of EF-Tu mt relative to domain III in the other complexes that are almost superimposed. The three complexes differ mainly in the distance between helix D of EF-Tu and helix ␣1 of EF-Ts. Helix ␣1 in EF-Ts mt is extended compared with the corresponding helices in the prokaryotic factors, and it lies closer to helix D than in either of the other two bacterial complexes. The implications of these differences must await further biochemical and structural studies. a Numbering refers to the precursor protein and is different from the numbering in the indicated reference, which refers to the mature protein.
FIG. 6. Position of previously mutated residues in the bovine EF-Tu mt ⅐Ts mt interface. See Table II for an overview of mutations made in mitochondrial EF-Ts mt . Some of the mutated residues of EF-Ts mt are shown in stick representation and labeled along with helices B and C from EF-Tu mt . Coloring of the EF-Tu mt ⅐Ts mt complex is as described in Fig. 1A. See "Results" for discussion.