Active Site Breathing of Human Alkbh5 Revealed by Solution NMR and Accelerated Molecular Dynamics

Abstract

AlkB homolog 5 (Alkbh5) is one of nine members of the AlkB family, which are nonheme Fe²⁺/α-ketoglutarate-dependent dioxygenases that catalyze the oxidative demethylation of modified nucleotides and amino acids. Alkbh5 is highly selective for the N⁶-methyladenosine modification, an epigenetic mark that has spawned significant biological and pharmacological interest because of its involvement in important physiological processes, such as carcinogenesis and stem cell differentiation. Herein, we investigate the structure and dynamics of human Alkbh5 in solution. By using ¹⁵N and ¹³C_methyl relaxation dispersion and ¹⁵N-R₁ and R_1ρ NMR experiments, we show that the active site of apo Alkbh5 experiences conformational dynamics on multiple timescales. Consistent with this observation, backbone amide residual dipolar couplings measured for Alkbh5 in phage pf1 are inconsistent with the static crystal structure of the enzyme. We developed a simple approach that combines residual dipolar coupling data and accelerated molecular dynamics simulations to calculate a conformational ensemble of Alkbh5 that is fully consistent with the experimental NMR data. Our structural model reveals that Alkbh5 is more disordered in solution than what is observed in the crystal state and undergoes breathing motions that expand the active site and allow access to α-ketoglutarate. Disordered-to-ordered conformational changes induced by sequential substrate/cofactor binding events have been often invoked to interpret biochemical data on the activity and specificity of AlkB proteins. The structural ensemble reported in this work provides the first atomic-resolution model of an AlkB protein in its disordered conformational state to our knowledge.

Figure 1

Comparison of the x-ray structure of apo Alkbh5 with solution NMR data. (a) Agreement between the observed and calculated RDCs obtained by SVD to the coordinates of the x-ray structure of apo Alkbh5 (PDB: 4NJ4 (21)). SVD fitted using the full set of experimental RDCs (open black circle) or RDCs from backbone amides in secondary structures (blue circles) fails to converge (because of the lack of a single orientation of the alignment tensor that satisfies all the experimental data), resulting in predicted RDC values close to zero and high R-factors (see Table 1). Exclusion of RDCs from the protein active site from the SVD analysis results in good agreement between experimental and back-calculated RDCs (red circles). (b) Amide groups whose experimental RDCs are in good agreement with back-calculated values are shown as red spheres on the x-ray structure of apo Alkbh5. Smaller blue spheres indicate the location of RDCs that were not included in the final SVD analysis (i.e., RDCs from loop or active-site residues). Residues that are missing from the x-ray structure (Leu¹⁴⁵–Gly¹⁴⁹) are indicated by a black curve. The Cys²³⁰–Cys²⁶⁷ disulfide bridge is shown as yellow sticks. αKG is modeled in the protein active site and shown as green spheres. (c) Agreement between measured and back-calculated C_α/C_β secondary chemical shifts. Back-calculation of the NMR chemical shifts was done using Sparta+ (67) and the crystal structure of apo Alkbh5 (21). R² values for linear regression of the C_α and C_β data are shown. To see this figure in color, go online.

Figure 2

Crystal structure of apo Alkbh5 (PDB: 4NJ4 (21)) showing the localization of the five cysteine residues (yellow sticks). α-helices are colored pink. β-strands are colored light blue. Missing residues in the x-ray structure (Leu¹⁴⁵–Gly¹⁴⁹) are indicated by a black curve. αKG is modeled in the protein active site and shown as green spheres. To see this figure in color, go online.

Figure 3

ps-ns conformational dynamics in apo Alkbh5. (a) ¹⁵N R₂, R₁, and R₂/R₁ data measured at 800 MHz and 25°C are graphed versus residue index. The secondary structure calculated from the x-ray coordinates using the software MolMol (68) and the secondary structure propensity calculated from the assignment of the NMR backbone resonances using the software Talos+ (69) are shown in the uppermost panel. (b) R₂/R₁ ratios are plotted on the structure of apo Alkbh5 according to the color bar. Residues that are missing from the x-ray structure are indicated by a black curve. The Cys²³⁰–Cys²⁶⁷ disulfide bridge is shown as yellow sticks. αKG is modeled in the protein active site and shown as green spheres. To see this figure in color, go online.

Figure 4

μs-ms conformational dynamics in apo Alkbh5. (a) 800 MHz ¹⁵N and ¹³C_methyl exchange contributions to the transverse relaxation rate (R_ex) measured at 15 (blue) and 25 (red) °C are graphed versus residue index. 30% transparency was applied to the curve describing the R_ex data at 15°C. (b) 800 MHz ¹⁵N and ¹³C_methylR_ex values measured at 25°C are plotted on the structure of apo Alkbh5 according to the color bar. Residues that are missing from the x-ray structure are indicated by a black curve. The Cys²³⁰–Cys²⁶⁷ disulfide bridge is shown as yellow sticks. αKG is modeled in the protein active site and shown as green spheres. (c) Examples of typical 800 MHz ¹⁵N (left) and ¹³C_methyl (right) relaxation dispersion data at 15 (blue), 20 (green), and 25 (red) °C are shown. Data are shown for the Ala¹²⁷-N_H (left) and Leu¹⁷⁶-C_methyl resonances, with the experimental data represented by open diamonds and the best-fit curves for a two-site exchange model as solid lines. Similar plots for all the analyzed resonances are shown in Fig. S4 and S5. To see this figure in color, go online.

Figure 5

aMD/RDC refinement of protein conformational ensemble. (a) Schematic of the aMD/RDC ensemble calculation protocol developed in this work. A preliminary pool of structures is generated by aMD and subsequently filtered to maximize the agreement between experimental and back-calculated RDCs. The ensemble size (i.e., the number of conformers in the ensemble) is increased until the minimal R-factor is reached. (b) R-factor versus ensemble size for the aMD/RDC ensemble refinement of apo Alkbh5. A 20-member ensemble is required to fit the experimental RDCs.

Figure 6

Comparison of the Alkbh5 structural ensemble with solution NMR data. (a) Agreement between the observed and back-calculated RDCs for the aMD/RDC refined ensemble of Alkbh5. Note that all the 188 experimental RDCs’ data were included in the calculation (see Table 1). (b) Agreement between measured and back-calculated C_α/C_β secondary chemical shifts. Back-calculation of the NMR chemical shifts was done in Sparta+ (67). Predicted C_α/C_β secondary chemical shifts are averaged over the aMD/RDC ensemble. R² values for linear regression of the C_α and C_β data are shown. (c) Sausage representation of the aMD/RDC ensemble generated using the software Pymol (70). Cartoons are colored according to the B-factor, as indicated by the color bar. B-factors were calculated using the formula $B_i=8{\pi }^2{U}_i^2$, where B_i and U_i are the B-factor and mean-square displacement of atom i, respectively. The overlay of the 20 conformers in the ensemble is shown in Fig. S3. To see this figure in color, go online.

Figure 7

Comparison between solution and crystal structures of human Alkbh5. (a) Sausage representations of the aMD/RDC ensemble (left), x-ray structure of apo Alkbh5 (center), and x-ray structure of Alkbh5 complexed with Mn²⁺ and αKG (right). Cartoons are colored according to the B-factor, as indicated by the color bar. B-factors for the aMD/RDC ensemble were calculated using the formula $B_i=8{\pi }^2{U}_i^2$, where B_i and U_i are the B-factor and mean-square displacement of atom i, respectively. The Cys²³⁰–Cys²⁶⁷ disulfide bridge is shown as yellow sticks. αKG is shown as green spheres. (b) Close-up views of the αKG binding pocket in the aMD/RDC ensemble (left), x-ray structure of apo Alkbh5 (center), and x-ray structure of Alkbh5 complexed with Mn²⁺ and αKG (right). For the aMD/RDC ensemble, the conformer with the widest αKG binding pocket is displayed. One αKG molecule is modeled in the binding site of the apo protein (left and center panels). Alkbh5 is shown as a transparent light blue surface. αKG is shown as red spheres. To see this figure in color, go online.