Structure of the Subunit Binding Domain and Dynamics of the Di-domain Region from the Core of Human Branched Chain α-Ketoacid Dehydrogenase Complex*

The homo-24-meric dihydrolipoyl transacylase (E2) scaffold of the human branched-chain α-ketoacid dehydrogenase complex (BCKDC) contains the lipoyl-bearing domain (hbLBD), the subunit-binding domain (hbSBD) and the inner core domain that are linked to carry out E2 functions in substrate channeling and recognition. In this study, we employed NMR techniques to determine the structure of hbSBD and dynamics of several truncated constructs from the E2 component of the human BCKDC, including hbLBD (residues 1-84), hbSBD (residues 111-149), and a di-domain (hbDD) (residues 1-166) comprising hbLBD, hbSBD and the interdomain linker. The solution structure of hbSBD consists of two nearly parallel helices separated by a long loop, similar to the structures of the SBD isolated from other species, but it lacks the short 310 helix. The NMR results show that the structures of hbLBD and hbSBD in isolated forms are not altered by the presence of the interdomain linker in hbDD. The linker region is not entirely exposed to solvent, where amide resonances associated with ∼50% of the residues are observable. However, the tethering of these two domains in hbDD significantly retards the overall rotational correlation times of hbLBD and hbSBD, changing from 5.54 ns and 5.73 ns in isolated forms to 8.37 ns and 8.85 ns in the linked hbDD, respectively. We conclude that the presence of the interdomain linker restricts the motional freedom of the hbSBD more significantly than hbLBD, and that the linker region likely exists as a soft rod rather than a flexible string in solution.

LBDs from other members of the ␣-ketoacid dehydrogenase family, the structure of hbLBD is a flattened ␤-barrel formed by two four-stranded anti-parallel ␤ sheets (7)(8)(9)(10). The lipoyl Lys 44 residue resides at the tip of a ␤-hairpin comprising a sharp ␤-turn that ties ␤-strands 4 and 5. The hbICD harbors the subunit binding site and forms a 24-meric cubic structure (11). The hbSBD is connected to the C terminus of hbLBD through a ϳ26-residue flexible linker and to the N terminus of hbICD by a ϳ25-residue tether. The hbSBD is the smallest globular protein domain known without disulfide bridges, metal ions, or prosthetic groups. The mammalian hbSBD binds E1 or E3 with different affinities of 10 Ϫ8 M and 10 Ϫ6 M, respectively. The chain folds of SBD from the Escherichia coli 2-oxoglutarate dehydrogenase complex (eoSBD) (12) and the B. stearothermophilus (also called Geobacillus stearothermophilus) PDC (bpSBD) have been determined previously by NMR spectroscopy (13). Each of these SBD is made up of two parallel helices, a fiveresidue helix-like turn and an irregular loop. The crystal structures of bpSBD in complex with E1 (14) or E3 (15) from the B. stearothermophilus PDC have also been solved. The structural information validates the view that E1 and E3 compete for essentially the same set of determinants in the E2 component of the B. stearothermophilus PDC (16) Unlike their prokaryotic counterparts, the mammalian and yeast PDCes employ SBD and a specific E3-binding protein (E3BP) to bind E1 and E3 components, respectively (17). The crystal structure of E3 bound to the E3 binding domain (hpE3BD) of E3BP from the human PDC has been elucidated (18,19). Whereas chain folds are conserved between hpE3BD and prokaryotic SBD, determinants in hpE3BD for binding human E3 differ from those in prokaryotic SBD for interactions with their cognate E3 (18). Despite these advances, to date the three-dimensional structure of hbSBD from the human BCKDC has not been reported. This information is essential for understanding the modes of E1 and E3 binding to the E2 core of the BCKDC catalytic machine.
In the present study, we employed multidimensional heteronuclear NMR techniques to determine the structures of hbSBD and dynamics of hbLBD, hbSBD, and a di-domain (hbDD, residues 1-168) containing both hbLBD and hbSBD domains and the interdomain linker region. We showed that the structures of the two domains are not perturbed by the presence of the linker. Although the linker region is disordered, it retains certain rigidity and thus imposes structural constraints that restrain the relative motion of the two domains. Analysis of backbone 15 N T 1 , 15 N T 2 , and 1 H-15 N NOE relaxation data further indicate that the dynamics of hbLBD is not significantly altered as a result of linking together the two domains. By comparison, hbSBD is less rigid and its dynamics is therefore affected more noticeably by the presence of the linker region.

MATERIALS AND METHODS
Sample Preparation-Three recombinant proteins containing various E2 domains were studied in this article, the di-domain protein (hbDD, residues 1-166), the lipoyl-bearing domain protein (hbLBD, residues 1-84), and the subunit binding domain protein (hbSBD, residues 104 -152). To facilitate protein expression and purification, the genes for hbDD and hbLBD were constructed such that each of the expressed recombinant proteins contains an extra methionine residue at the N terminus (at the Ϫ1 position), and a leucine, a glutamic acid along with six histidine residues at the C terminus. To produce the hbSBD protein, a tobacco-etch virus (TEV) protease cleavage site was cloned in the linker region of hbDD between positions 103 and 104. The cells harboring various plasmids were grown in LB medium with 0.025 mg/ml kanamycin at 30°C. The culture was grown to A 600 of 0.6 prior to induction with 1 mM isopropyl-1-thio-␤-D-thiogalactoside for 4 h. The cells were harvested by centrifugation, resuspended in a lysis buffer at pH 8, and lysed in a microfluidizer. The cell lysate was centrifuged at 18,000 ϫ g for 30 min at 4°C and the supernatant purified by HIS-select Ni-NTA resin (Qiagen). The column was washed with buffer at pH 7.5 containing 100 mM KCl and 50 mM KH 2 PO 4. The target protein was eluted with the same buffer containing 100 mM imidazole. The proteins were further purified to homogeneity by size exclusion chromatography using Sephacryl S-100 column in 50 mM phosphate buffer, pH 7.5, 100 mM NaCl. The Sephacryl S-100 column was connected to an AKTA fast performance liquid chromatography (FPLC) system (Amersham Biosciences). To produce isotopically labeled samples for NMR structural determination, the cell was cultured in M9 minimal medium, supplemented with 15 NH 4 Cl (1g/liter), 13 C-labeled glucose (2g/liter), and 13 C/ 5 N-labeled Celtone base powder (0.5 g/liter, Spectra Stable Isotopes). To produce hbSBD the Ni-NTA-purified di-domain protein was digested with the TEV protease and the His 6tagged protein isolated using the N i -NTA resin. The purified hbSBD contains an extra glycine in front of Glu 104 , which was left behind after the TEV cleavage. The purified protein was better than 95% pure, based on Coomassie Blue-stained gels. The NMR samples contained ϳ1-2 mM protein in 50 mM phosphate buffer at pH 7.5, 100 mM NaCl, 0.02% (w/v) NaN 3 , and 10% (v/v) D 2 O. To prepare D 2 O samples, concentrated protein solutions were lyophilized and redissolved in D 2 O.
NOE distance restraints were based on two-dimensional NOESY, three-dimensional 15 N-edited NOESY-HSQC, and 13 C-edited NOESY-HSQC experiments with a 120-ms mixing time. NOE peak regions were picked and quantified using the automatic picking routine of AURELIA. Automated NOE cross-peak assignments were performed using the CANDID module of CYANA (21), and backbone torsion angles were predicted from TALOS (22). Structure calculation was done using CYANA 2.0.19. One hundred random conformers were annealed in 10,000 steps for each cycle, and 20 structures with lowest target function in cycle 7 were selected for further analysis with programs MOLMOL (23) and PROCHECK-NMR (24). 15 N T 1 , 15 N T 2 , and steady state heteronuclear 1 H-15 N NOE data were obtained at 295 K using standard pulse sequences (25). Two sets of T 1 and T 2 experiments, and three pairs of the steady-state heteronuclear NOE experiments were collected at 600 MHz (14.1 T).
Relaxation Data Analysis-Relaxation rates and heteronuclear 1 H-15 N NOE enhancements were calculated from peak heights of the resonances in a series of 1 H-15 N HSQC spectra. The AURELIA program was employed for data analysis, including peak picking and peak intensity determination. Errors in 1 H-15 N NOE were expressed as the standard deviation of three pairs of repeated 1 H-15 N NOE experiments. The longitudinal and transverse relaxation rate constant, R 1 and R 2 , were obtained by the two parameter nonlinear optimization of a single exponential function using the Marquardt-based nonlinear least-square curve-fitting algorithm. Uncertainties were obtained from curve fitting with the Prism 3.00 program (GraphPad Software, Inc.).
Reduced Spectral Density Analysis-Dynamic parameters were contained in the spectral density function, J() at five frequencies: 0, N , H Ϫ N , H , and H ϩ N . An assumption was made that J() ϭ 1 / 2 ϩ 2 , where the first and second terms represent contributions to J() from the overall rotation and internal dynamics, respectively. By substituting a single value of J(0.87 H ) for linear combinations of J( H Ϫ N ), J( H ) and J( H ϩ N ) (26), three spectral density functions: J(0), J( N ), and J(0.870 H ), were determined from the experimental data for T 1 , T 2 , and NOE (27) (28), according to Equations 1-4, where R 1 and R 2 are the longitudinal and transverse relaxation rates, respectively; d ϭ ( o h␥ N ␥ H /8 2 )(r NH Ϫ3 ); c ϭ N ⌬/͌3; o is the permeability in vacuum; h is Planck's constant; ␥ H and ␥ N are the nuclear gyric ratios of 1 H and 15 N, respectively; N and H are the Larmor frequencies of 1 H and 15 N, respectively; ⌬ ϭ Ϫ170 ppm is the 15 N chemical shift anisotropy of polypeptide chains (29).
For rigid globule molecules, the overall rotational correlation time c was obtained from spectral density functions from Equation 5 (30). Fig. 1 shows the amino acid sequence and the secondary structure of the di-domain (hbDD, residues 1-166) of the E2 component from human BCKDC. The hbDD consists of three functional regions: the lipoyl-binding domain (hbLBD, residues 1-84), the linker region (residues 85-110) and the subunit-binding domain (hbSBD, residues 111-149). Fig. 2A shows the 15 N HSQC spectrum and resonance assignments of a clone containing the hbSBD domain (residues 104 -152). The assignments of the main chain H N , N H , C ␣ , C ␤ , and CЈ resonances of hbSBD and hbDD were based on CBCANH, CBCA(CO)NH, HNCO, and HN(CA)CO experiments. The assignments of side chain resonances were achieved by the analysis of 1 H-15 N-edited TOCSY-HSQC and 1 H-13 C-edited HCCH-TOCSY spectra. More than 80% of the spin systems and nearly all 1 H, 15 N, and 13 C backbone resonances of hbSBD were assigned. The NMR assignment data have been deposited in BMRB (accession number 7057).

Solution Structures of hbSBD-
The tertiary structure of hbSBD was determined from 634 distance restraints assigned by CYANA and 48 torsion angles (24,24). The superposition of the resultant family of 20 best structures of the hbSBD is shown in Fig. 2B. The coordinates have been deposited in PDB/RCSB (PDB accession number: 1ZWV). The structural statistics are given under supplemental materials. These structures have the total target function ranging from 0.02 to 0.15 (Å 2 ). The average root mean square deviations (RMSD) for backbone atoms are 0.28 Ϯ 0.08 Å for the well defined secondary structure regions (Pro 114 -Glu 122 , Lys 141 -Glu 149 ) and 0.41 Ϯ 0.10 Å for the segment over residues 114 -149. The average RMSD is 0.91 Ϯ 0.11 Å for heavy atoms and 1.00 Ϯ 0.09 Å relative to the mean structure. The quality of these structures is good as judged by the fact that 100% of the backbone torsion angles for non-glycine and non-proline residues fall in either the most favorable (76.4%) or the allowed regions (23.6%) of the Ramachandran plot (24). The ribbon representation of the tertiary fold of hbSBD (Fig. 2C) shows that the two parallel helices, i.e. helix H1 (Pro 114 -Glu 122 ) and helix H2 (Lys 141 -Glu 149 ), were stabilized by the hydrophobic interactions between Ala 115 , Val 116 , Leu 119 , Asn 123 in H1 and Lys 141 , Sequences for hbLBD (residues 1-84) and hbSBD (residues 111-149) are underlined. The outer linker segment between the two domains is shaded in gray.

Structure and Dynamics of the E2 Di-domain
Ile 144 , Leu 146 in H2. The two helices also interact with the long loop between them through the following hydrophobic interactions: Val 116 3 Ile 139 and Leu 140 ; Ala 120 3 Ile 125 and Leu 127 ; Ile 144 and Leu 148 3 Ile 125 , and Ile 144 3 Val 130 . However, the structure of the long loop is less defined than the helical structure regions. No long range NOE was observed for the six N-terminal residues of the linker region (residues 104 -109), suggesting that this segment is disordered.
Effect of the Linker Domain on the Structures of hbLBD and hbSBD-To probe possible structural effects caused by the tethering of hbLBD and hbSBD as in natural E2, we obtained the 15 N HSQC spectrum of hbDD (Fig. 3A) and compared it with those of hbLBD ( Fig. 1 in Ref. 8) and hbSBD ( Fig. 2A). The result shows that the bulk of resonances observed in hbLBD and hbSBD are present in the same positions as in the di-domain spectrum. It indicates that the overall tertiary folds of hbLBD and hbSBD were not affected by linking these two domains as in hbDD. However, significant chemical shift changes were observed in several resonances, suggesting the existence of structural perturbations. To identify the regions affected we have further assigned all resonances in Fig. 3A. Fig. 3B shows the chemical shift differences in amide resonances between those observed in hbDD and the corresponding resonances in hbLBD and hbSBD. The values are calculated with the equation: ⌬␦ ϭ [(⌬␦ NH ) 2 ϩ (0.154⌬␦ N ) 2 ] 1/2 , where ⌬␦ NH and ⌬␦ N are the chemical shift differences for NH and 15 N, respectively (31). The result showed that chemicals shift perturbations of all residues in hbLBD are less than 0.1 ppm suggesting the structure of hbLBD is not perturbed by the tethering of the linker to its C terminus (Fig. 3B). The three internal residues, Ile 11 , Ser 42 , and Leu 82 , exhibit slightly larger chemical shift perturbations (Ͼ0.05 ppm) than remaining residues in the hbLBD domain. The locations of Ile 11 , Ser 42 , and Leu 82 are shown in Fig. 3C. Interestingly, residues Ile 11 and Ser 42 are located in the vicinity of the lipoic acid attachment site of Lys 44 ␤-hairpin region that is away from the C terminus (8). Thus, the chemical shift perturbations of these two residues might indicate the presence of certain degree of interactions between the hairpin region and other part(s) of the di-domain.
In contrast, tethering hbSBD through the inter-domain linker causes considerable overall chemical perturbations in the hbSBD domain. As shown in Fig. 3B, Gly 107 , Leu 127 , and Leu 148 exhibit the largest changes in chemical shift (⌬␦ Ͼ 0.1 ppm) among the residues within hbSBD with Arg 108 , Leu 111 , Ala 112 , Thr 113 , Glu 142 , and Leu 145 also perturbed significantly (0.1 ppm Ͼ ⌬␦ Ͼ 0.05 ppm). Gly 107 is part of the disorder linker and its very large chemical shift change is expected. Leu 127 is located in the long loop region between the two helices and Leu 148 is located on helix 2 near the C terminus of hbSBD (Fig. 3D). The structure indicates that the long loop is located close to the N terminus and there are extensive interactions between these two regions. Thus, the conformational change in the N terminus of hbSBD caused by the presence of the linker region is likely to be transmitted to the loop region containing Leu 127 . The extended C-terminal linker residues (ϳ25 amino acids) of hbDD in turn could interact with helix 2, where Glu 142 , Leu 145 , and Leu 148 reside. However, we cannot discount the possibility that some of the chemical shift perturbations are caused by  Fig. 2. The spectrum was obtained at 600 MHz on a Bruker AVANCE600 spectrometer. The assignments for residues from hbLBD were shown in blue, for residues in the hbSBD region in red, and for residues in the outer linker residues between the two domains were depicted in green. B, sequence-dependent variation of the differences in chemical shift, ⌬␦, between amide resonances of hbDD, hbLBD and hbSBD. Definition of ⌬␦ is given in the text. C, mapping of residues in hbLBD with ⌬␦ Ͼ 0.05 ppm. D, mapping of residues in hbSBD with ⌬␦ Ͼ 0.05 ppm. SEPTEMBER 22, 2006 • VOLUME 281 • NUMBER 38 interaction with other part of the di-domain. It is also worth mentioning that mixing u-15 N hbLBD and u-15 N hbSBD samples causes no change in chemical shifts in the 15 N HSQC spectra of hbLBD and hbSBD, suggesting that the linker and the C-terminal residues of hbSBD are required for the observed changes to occur. Furthermore, the CSI plot of hbDD shows that secondary structures of the linked hbLBD and hbSBD are identical to those of the two corresponding isolated domains. Some resonances in the 8.3 Ϯ 0.5 ppm ( 1 H) region are attributed to residues Ser 85 , Glu 86 , Val 89 , Val 90 , Glu 91 , Thr 92 , Ala 94 , Ser 96 , Asp 98 , Glu 99 , His 100 , Thr 101 ,Gly 107 ,Arg 108 , and Thr 110 in the linker region (Fig. 3A). Interestingly, the consecutive sequence of five residues from His 102 to Ile 106 cannot be observed, suggesting that this segment is exposed.

Structure and Dynamics of the E2 Di-domain
Dynamics of hbLBD, hbSBD, and hbDD-Shown on the left panels of Fig. 4  Although the amide resonances of several residues in the linker region could be assigned, the considerable overlap of these residues (Fig. 3A) prevented unambiguous determinations of their relaxation parameters. The T 2 values of these resonances were also very short and reliable relaxation data could not be obtained. These results suggest that the linker region is disordered and partially exposed to the solvent. The relaxation data profile of hbLBD and the corresponding domain in hbDD are practically identical indicating the internal dynamics remains unchanged between the free and linked hbLBD. However, the R 2 values of hbLBD in the di-domain are higher than those of isolated hbLBD, and the situation is reversed for the R 1 profile, as a result of increases in the rotational correlation time in hbDD (see "Discussion" below). On the other hand, hbSBD shows more fluctuation than hbLBD both as a free domain and as a moiety of the di-domain. The ordering of the N-terminal region from the hbSBD domain in hbDD can be clearly discerned by the increase in NOE magnitude for the hbSBD residues.
To quantitatively assess the dynamic behavior of hbSBD, the reduced spectral density functions (J values) of the three domain constructs (free hbLBD, free hbSBD and hbDD) were calculated. The J values reflect the degree of motion in given frequency regions. Thus, the higher the value of J(), the more extensive motion is present at the MHz frequency region. The J values of hbSBD shown on the right panels of Fig. 4 (D-F). J(0), J(N), and J(H) represent those obtained at 0, 60 MHz, and 522 MHz, respectively. The hbSBD exhibits considerable sequence-dependent variation in dynamics, especially in the loop region. These results are consistent with the lower melting temperature and stability of hbSBD (T m ϭ 318 K and ⌬H G ϭ 20 kcal/mol) compared with hbLBD (T m ϭ 344 K and ⌬H G ϭ 80 kcal/mol) (32)(33)(34). The spatial variation of the dynamics associated with hbSBD can be seen more clearly from the sausage plot shown in Fig. 5, A and B. Upon tethering to the hbLBD in hbDD, the high frequency motions of hbSBD at 60 MHz and 522 MHz appear to be quenched, since there is little sequence variation at J(N) and J(522) corresponding to the latter domain in hbDD. In contrast, the low frequency motions at 0 MHz appear similar in both the free and linked states.
The rotational correlation times ( c ) of all residues can be calculated from the J values, as given in Equation 5. We exclude all residues with NOE less than 0.6 and remove residues with T1/T2 ratios outside one standard deviation of the mean in calculating c s (35). The two isolated domains, hbLBD and hbSBD have similar c values of 5.54 ns and 5.73 ns for hbLBD and hbSBD, respectively, which is consistent with their small sizes (Fig. 5C). Our measurements of the dimensions of the two molecules are 34 ϫ 25 ϫ 22 Å and 30 ϫ 26 ϫ 17 Å for hbLBD and hbSBD, respectively. Apparently hbSBD is a bit flatter and more asymmetric. We speculate that the shape difference and the more dynamic nature of hbSBD may render this domain diffuse slower than the corresponding globular protein of the same size. Relative to hbLBD, hbSBD shows larger residue-dependent variation in its c values, indicative of the presence of internal motions. On the other hand, the two linked hbLBD and hbSBD domains in hbDD exhibit distinctive rotational correlation times of 8.37 ns and 8.85 ns, respectively, with the smaller hbSBD domain surprisingly rotating more slowly than the larger hbLBD. The average c value for hbDD is 8.61 ns. We previously obtained a c value of ϳ9.5 ns for the 21 kDa globular protein E. coli thioesterase (36). In the present study, the observed overall c value of 8.61 ns for hbDD is significantly larger than that expected for the size of the two individual domains, but is in the range of that anticipated for a full-length 175-residue protein such as hbDD. The data suggest that the linker imposes significant rotational constraints on the two constituent domains such that hbLBD and hbSBD in hbDD are not rotating freely and independently of each other in the solution.

DISCUSSION
The structures of SBDs from B. stearothermophilus (bpSBD, 1w3d) (13) and 2-oxoglutarate dehydrogenase of E. coli (eoSBD, 1bbl) (12) have been solved, with both structures containing two ␣-helices and a short 3 10 -helix connected by loops. The crystal structures of the B. stearothermophilus bpSBD in complex with E1 (14) and E3 (15) have also been determined. The latter two structures reveal that essentially the same set of residues from bpSBD participates in forming the interfaces with E1 and E3, thus excluding the possibility of simultaneous binding of the prokaryotic SBD to both catalytic components. These interactions are dominated by a "charge zipper" consisting of salt bridges formed between positively charged Arg 136 , Lys 137 , Arg 140 , and Arg 157 of bpSBD and negatively charged counterparts from E1 or E3. Additional hydrophobic interactions involving Ile 130 , Met 132 , and Pro 133 from bpSBD also contribute to the stability of interfaces. The sequence of hbSBD from human BCKDC shows 47% similarity to bpSBD and 43% to eoSBD (Fig. 6A). The present structure of hbSBD also contains two helices; however, the 3 10 -helix is not present. An alignment of the above three SBD domains shows that SBD residues at the interface with cognate E3 are conserved. Specifically, residues Arg 117 , Arg 118 , Met 121 , and Arg 138 in hbSBD that correspond to the interfacial residues Arg 136 , Lys 137 , Arg 140 , and Arg 157 in hpSBD are present in roughly the same locations, with electrostatic surfaces of the two domains sharing a high degree of resemblance (Fig. 6, B and C). These results raise the possibility that similar interactions occur in the binding of hbSBD and bpSBD to their cognate E1 and E3. However, the replacement of Arg 140 in bpSBD by Met 121 in hbSBD could potentially alter the interaction with E3 when compared with  SEPTEMBER 22, 2006 • VOLUME 281 • NUMBER 38 bpSBD. For example, we have observed no change in the binding constant for E3 when the Arg 138 of hbSBD was mutated to Ala or Ile (data not shown), suggesting that Arg 138 in hbSBD may not be as important as Arg 157 in bpSBD for binding to the respective E3. Moreover, the side chains of Arg 117 and Arg 118 in hbSBD overlay very well with Arg 136 and Lys 137 in bpSBD, whereas the side chain of Arg 138 in hbSBD points to the direction opposite to that of Arg 157 in bpSBD (Fig. 6D). These differences in residue types and side chain orientations may account for the species specificity of SBD in interacting with the corresponding E3.

Structure and Dynamics of the E2 Di-domain
Recent cryo-electron microscopy reconstruction of the E1-E2 and E3-E2 subcomplex structures of the PDC from B. stearothermophilus, bovine kidney and Saccharomyces cerevisiae showed that at saturating concentrations in vitro, the E1 or E3 component is able to form an annular outer shell around the 60-meric dodecahedral E2 core (37)(38)(39)(40). The separation between the two shells is ϳ90 Å for the E1-E2 subcomplex, and 75 Å for the E1-E3 subcomplex of the B. stearothermophilus PDC. Similarly, the distance between the E1 shell and the E2 core is 50 Å for the E1-E2 subcomplex of the PDC from bovine kidney but may increase to 75 Å, depending on the occupancy of the E1 binding site. The architecture of the PDC is consistent with a model in which the LBD primarily reside and function inside the annual gap of E1 or E3 (37,38). To carry out its functional role in substrate channeling, the LBD must visit the three active sites in E1, E2, and E3 (6). Thus, the linker regions between LBD and SBD (the outer linker) and between SBD and the inner core domain (the inner linker) must be flexible (6,41,42). This notion is supported by the fact that in the above cryoelectron microscopic structures, the N-terminal LBD, the outer linker region, and the inner linker region cannot be observed. In the present study, the very high relaxation rates associated with the outer linker (residues 86 -110) of human E2b (Fig. 4, A  and B) provide the first direct evidence that this region is indeed flexible. The data lend strong support to the current concept on the mechanism of substrate channeling for mitochondrial ␣-ketoacid dehydrogenase catalytic machines.
Interestingly, while the flexibility is conserved, the outer linker in the hbDD of human BCKDC is abundant in negatively charged amino acids, which is in variance with the Pro/Ala-rich outer linkers in the E. coli PDC (43). The presence of negatively charged residues in the outer linker of bovine hbE2b may have resulted from the selective pressure for the hbLBD to bind the BCKD kinase for regulating BCKDC activity by reversible phosphorylation. We have shown previously that the truncation of the outer linker region immediately distal to the hbLBD of BCKDC prevents the docking of the BCKD kinase to this domain (44). The prokaryotic PDC and BCKDC are not regulated by reversible phosphorylation, therefore, the cognate LBD plays no role in kinase binding (45). At pH 7.5, about 50% of the amide resonances assignable to the linker region in hbDD were obtained (Fig. 3A). This result indicates that the amide groups from about 50% of the linker residues are partially protected from solvent exchange. These amide groups have T 2 values shorter than 10 ms, and the exchange rates of these amide protons are estimated to be in the range of 10 -100 s Ϫ1 . Significantly, 10 of 12 residues from Ser 85 to Ser 96 , with the exception of Glu 87 and Asp 88 , are observable. The results suggest that this Glu/Asp-rich segment linking hbLBD and hbSBD of human BCKDC is likely to be partially structured. This segment is immediately distal to the hbLBD domain; it is therefore tempting to speculate that this segment may have strong interaction with hbLBD. In fact, the segment may also be an integral part of the hbLBD domain, and may itself form a loop structure. However, we failed to observe any definitive NOEs between residues in this linker region in hbDD with certainty, probably because of the dynamic nature of the linker region. Chemical shift index analysis also indicated the absence of stable secondary structure elements for the linker region. A 32-residue synthetic peptide corresponding to the Ala/ Pro-rich outer linker between the innermost LBD and the SBD of the E. coli PDC was shown not to be random coils. Instead, it assumes a certain degree of rigidity, with the Ala-Pro peptide bonds in the all trans conformation (41,42). Based on this outer linker conformation and the results of limited proteolysis studies with the B. stearothermophilus PDC (46), it has been suggested that the movement of the LBD can be restricted by the linker between LBD and SBD. The limitation in translational freedom of the LBD may be necessary to affect active site coupling in the E2 core of the prokaryotic PDC (6). In the present study, the rigidity of the outer linker region can be inferred from the rotational correlation time, c , as determined from relaxation measurements. The c ϭ 5.54 ns and 5.73 ns at 295 K for hbLBD and hbSBD, respectively are consistent with the small molecular sizes of these two domains. These c values increase to 8.37 ns and 8.85 ns for the linked hbLBD and hbSBD domains, respectively, in the di-domain. These c values fall in the range expected for a globular protein of ϳ20 kDa. Thus, our results clearly indicate that the two tethered domains in the di-domain move in a correlated manner. These findings suggest that the linker between hbLBD and hbSBD exist as a soft rod and not a flexible string, since this segment imposes significant rotational restraints on the two domains. Surprisingly, the smaller hbSBD domain rotates slightly slower than the larger hbLBD domain in hbDD. This implies that not only the linker between hbLBD and hbSBD limits the free rotation of hbSBD, but also the C-terminal linker residues in hbDD (Lys 150 -Pro 166 ) may significantly restrict the motion of hbSBD.