Crystal cryocooling distorts conformational heterogeneity in a model Michaelis complex of DHFR

Abstract

Most macromolecular X-ray structures are determined from cryocooled crystals, but it is unclear whether cryocooling distorts functionally relevant flexibility. Here we compare independently acquired pairs of high-resolution data sets of a model Michaelis complex of dihydrofolate reductase (DHFR), collected by separate groups at both room and cryogenic temperatures. These data sets allow us to isolate the differences between experimental procedures and between temperatures. Our analyses of multiconformer models and time-averaged ensembles suggest that cryocooling suppresses and otherwise modifies side-chain and main-chain conformational heterogeneity, quenching dynamic contact networks. Despite some idiosyncratic differences, most changes from room temperature to cryogenic temperature are conserved and likely reflect temperature-dependent solvent remodeling. Both cryogenic data sets point to additional conformations not evident in the corresponding room temperature data sets, suggesting that cryocooling does not merely trap preexisting conformational heterogeneity. Our results demonstrate that crystal cryocooling consistently distorts the energy landscape of DHFR, a paragon for understanding functional protein dynamics.

Figure 1

Cryocooling induces changes in DHFR structural heterogeneity. The broad distribution of strong difference peaks in the isomorphous F_o-F_o 13cryo-05cryo map, in contrast to the relative flatness of the 13RT-05RT map, emphasizes that cryogenic freezing perturbed the two crystals differently, resulting in widespread and unpredictable changes to dynamics. Both the 13RT-13cryo and 05RT-05cryo difference density maps have significant peaks of both signs (green: positive, red: negative), suggesting that cryocooling idiosyncratically alters the same protein’s structure and/or dynamics in different crystals at the hands of different crystallographers. All four maps are shown contoured at 0.4 e^-/ų (van den Bedem et al., 2013), which in these cases corresponds to 2.61, 3.01, 2.43, and 4.55 σ, respectively. NADP⁺ in orange, folate in yellow. Model shown is qFit model of first structure in F_o-F_o subtraction.

Figure 2

Cryocooling alters sidechain rotamer heterogeneity preferentially in certain protein regions. Many residues have at least one rotamer in the 13RT or 05RT structure that is missing from the corresponding cryo structure (red+purple sidechains, top row). Smaller, but still substantial, sets of residues have at least one rotamer in the 13cryo or 05cryo structure that is missing from the corresponding RT structure (blue+purple sidechains, bottom row). Notably, many of these residues have altered heterogeneity consistently in both the 2013 and 2005 structure pairs (purple sidechains), as opposed to just one or the other. Glu101 (Figure S2) and Arg52 (Figure 6) are labeled in the 13RT panel.

Figure 3

Cryocooling affects sidechain and mainchain conformational heterogeneity idiosyncratically at the detailed level. Changes in the maximum sidechain (left) and mainchain (right) RMSD across all alternate-conformation combinations at each residue from the room-temperature to the cryogenic qFit model are poorly correlated between the 2013 and 2005 pairs. R² correlation coefficients are from linear least-squares fits. Diagonal lines are for visual comparison.

Figure 4

Room-temperature ensembles are consistently more flexible than cryogenic ensembles at several surface loops. C_α RMSF values averaged across the subset of the 75 total ensemble variants for each dataset with R_free within 0.02 of the lowest R_free (Table 2) are higher for each room-temperature ensemble (red lines) than the corresponding cryogenic ensemble (blue lines) at Arg52 and Tyr128 (labeled), although standard deviations are substantial (semi-transparent bars). These regions are explored further below (Figure 6, Figure 7). The Met20 loop (labeled) is flexible, but similarly so, in all four averaged traces.

Figure 5

Cryocooling rigidifies solvent molecules surrounding DHFR. The 13RT and 05RT structures have dramatically fewer ordered water molecules (red spheres) than the 13cryo and 05cryo structures (blue spheres) (see also Table 1). NADP⁺ and folate coloring as in Figure 1.

Figure 6

Shifted crystal contacts in cryogenic structures favor a new Arg52 sidechain conformation. In the 13RT and 05RT structures, Arg52 (gray, center) adopts rotamers that point toward folate (yellow), and that point toward the crystal lattice neighbor (pink) but are not close enough to be properly positioned for a hydrogen bond. By contrast, in both the 13cryo and 05cryo structures, the C_β-C_β distance to the nearest lattice residue (Asp70*) shrinks from 9.3 to 8.7 Å for the 2013 structures and from 9.0 to 8.7 Å for the 2005 structures. As a result, the lattice copy of Asp87 (Asp87*) shifts upward in this view such that Arg52 can adopt a new rotamer that forms hydrogen bonds (green dotted lines) to the lattice copy of Asp70 (Asp70*). A water molecule (red sphere) takes the place of the room-temperature rotamer near the folate. 2mF_o-DF_c electron density contoured at 1.9 σ (dark blue) and 0.6 σ (light blue) supports the modeled sidechain and water positions in all four qFit structures.

Figure 7

Cryogenically ordered water molecules induce ensemble loop ordering. The surface-exposed loop encompassing Asp127, Tyr128, and Glu129 is very flexible in the lowest-R_free 13RT ensemble (red, left panel) because it is relatively unconstrained by contacts to crystal lattice mates (left and top in this view, light pink). By contrast, consistently modeled ordered waters (blue spheres) unique to the cryogenic ensemble (blue, right panel) bridge the polar Asp127 and Glu129 sidechains via hydrogen bonds to stabilize a more unique loop conformation. This phenomenon is extremely similar for the 05RT and 05cryo models (not shown).

Figure 8

Temperature dependence of coupled conformational heterogeneity linking the two subdomains of DHFR. The largest dynamic CONTACT network (red) from both the 13RT and 05RT qFit models spans the two subdomains (top vs. bottom in this view) via one or both of the NADP⁺ cofactor and folate substrate (spheres, labeled for 13cryo) in the active site (van den Bedem et al., 2013). By contrast, using the same T_stress values (Methods) as input to CONTACT, none of the networks (different colors) from the corresponding 13cryo and 05cryo models connect the two subdomains via the active site.