The starting conformation of the SK_Mtu•ATP•Mg²⁺ complex was generated using the coordinates of SK_Mtu from the structure of the SK_Mtu•ADP•Mg²⁺•shikimate complex (PDB: 1WE2) (Pereira et al., 2004) after removal of the bound ADP and shikimate. The Mg²⁺ ion and all solvent molecules were retained. The coordinates of ATP were derived from the structure of the SK_Mtu•AMPPCP•shikimate complex (PDB: 1ZYU) (Gan et al., 2006) by aligning the 1WE2 and 1ZYU structures using the Cα atoms of their respective protein components and converting the bound AMPPCP to ATP through the replacement of the relevant carbon atom by oxygen. Subsequently, the protein and shikimate coordinates of the 1ZYU structure were removed to generate the SK_Mtu•ATP•Mg²⁺ complex. For consistency, the corresponding apo simulations utilized the same starting structure after removal of ATP•Mg²⁺. For unliganded SK_Cbu, the starting structure was obtained directly from the coordinates found in PDB: 3TRF (Franklin et al., 2015) after removal of all ligands except solvent.

The starting structures, generated as discussed above, were equilibrated using previously described protocols (Hajredini et al., 2020). Briefly, the systems were parameterized within the CHARMM36m force-field (Huang et al., 2017), solvated with TIP3P water molecules, energy minimized and then equilibrated, first in an NVT ensemble, and then in an NPT ensemble to generate starting structures for the enhanced sampling, REST2, (Wang et al., 2011) simulations. Subsequently, REST2 production runs were carried out for 200 ns using 14 replicas spanning a 300–400 K temperature range. The final exchange probabilities for the SK_Mtu•ATP•Mg²⁺, SK_Mtu, and SK_Cbu simulations were 23.3 ± 8.2, 24.2 ± 8.4, and 29.0 ± 0.6%, respectively. The first 40 ns in each of the simulations were used as “burn-in” periods and discarded, and all subsequent analyses were carried out utilizing the remaining 160 ns. As shown in Supplementary Figure S2, 40 ns was more than sufficient to ensure equilibration of the key observables (θ and |h|, see below) probed in the simulations.

To describe the global conformational landscape of SK_Mtu, a cylindrical coordinate frame, comprising of an angle θ and a rise |h| (Figure 1A), was defined as in the case of Wzc_CDΔC, described previously (Supplementary Figure S3A) (Hajredini et al., 2020). The centers of mass of the Cα atoms of two segments on helix α1 (that contains the Walker-A motif), comprising residues 15–19 (P1, brown) and 21–25 (P2, yellow), were used to define a vector that was aligned to the z-axis of the molecular frame. Each structure within the REST2-generated ensembles was rotated and translated such that point P3 (cyan), defined by the center of mass of the Cα atoms of the first turn of helix α2 (33–37), was placed on the x-axis along the positive direction. θ is then defined as the angle between the xy-projection of a vector extending from point P1 to P4 (blue; defined by the center of mass of the Cα atoms of residues 155–159 that form the first turn of helix α8). The rise |h| is defined as the absolute value of the distance between points P1 and P3 along the z-axis. Thus, the angle θ and the rise |h| provide measures of the outward rotation and the up/down movement, respectively, of the Walker-A motif relative to the Walker-B and DxD motifs. The crystal structures of SK_Mtu (listed in Supplementary Table S1) were projected onto the cylindrical coordinate frame using the same protocol. For the structures of the SK enzymes from other organisms (listed in Supplementary Table S1), a similar procedure was utilized to define the points P1-P4 (described in Supplementary Table S2) and subsequently, θ and |h|.

We performed two sets of enhanced sampling simulations utilizing the replica exchange with solute scaling (REST2) approach. REST2 represents a modification of the Hamiltonian scaling protocol of the replica exchange with solute tempering (REST) scheme (Liu et al., 2005) that greatly enhances its efficiency by decoupling the acceptance rate from the solvent. For the REST2 simulations, protocols similar to those described before (Hajredini et al., 2020) were utilized with either unliganded (apo) or ATP•Mg²⁺-bound SK_Mtu. To analyze the global conformational states populated by these two ensembles we utilized the cylindrical coordinate frame defined in Figure 1A. Note that this frame is similar, but not identical, to that previously defined for Wzc_CDΔC (Hajredini et al., 2020) (compare Supplementary Figure S3A) given the significant structural differences between the two proteins. The orientations of the α8 helix (carrying P4) and α1 (carrying P1 and P2) in SK_Mtu are similar to the α2-α3 pair in Wzc_CDΔC (the latter carries the Walker-A sequence) (Hajredini et al., 2020). In lieu of a missing Wzc_CDΔC α4 equivalent in SK_Mtu, the α2 helix (that hosts the second Asp, D34, of the DxD motif) was utilized to define P3. Thus, while this definition allows for a qualitative comparison of the global conformational states sampled by SK_Mtu and Wzc_CDΔC with respect to their key catalytic elements, a quantitative equivalence should not be expected.

Projection of the conformational ensembles of apo-SK_Mtu (Figure 1B, top panel) and the SK_Mtu•ATP•Mg²⁺ complex (Figure 1C, top panel) onto the 2-dimensional θ-|h| space suggests differences in global conformations between the two cases. The presence of ATP•Mg²⁺ induces a more closed conformation, characterized by smaller θ values, compared to a more open conformation (larger θ values) seen for apo-SK_Mtu. We designate these global states as a Walker Closed State (WCS, seen in the SK_Mtu•ATP•Mg²⁺ complex), and a Walker Open State (WOS, seen in apo-SK_Mtu) to distinguish them from the closed and open classifications of the relative conformation of the SB and the LID domains traditionally used in the SK literature (Hartmann et al., 2006). Contrasts between the θ-|h| projections of SK_Mtu and Wzc_CDΔC are evident upon comparing Figures 1B,C with Supplementary Figures S3B,C. In the case of Wzc_CDΔC, the OS is characterized by high values of both θ and |h|, while the CS exhibits low values for both variables (Supplementary Figures S3B,C), suggesting that these variables are positively correlated (Hajredini et al., 2020). In contrast, the regions of maximal density for the WOS (Figure 1B, top panel) and the WCS (Figure 1C, top panel) suggest a weak anti-correlation between the θ (WOS: 196 ± 4°, WCS: 188 ± 3°) and |h| dimensions (WOS: 4.6 ± 0.7 Å, WCS: 4.9 ± 0.4 Å), with lower θ and higher |h| for the WCS, and the opposite for the WOS. Further, the apparent separation between the WOS and the WCS seems less prominent than that between the OS and CS of Wzc_CDΔC (compare Figures 1B,C with Supplementary Figures S3B,C; also see Supplementary Figure S4A). As an independent validation of our simulations, we note that the regions in θ-|h| space spanned by the WOS and the WCS in our REST2 ensembles are well represented (Supplementary Figure S4A) in the several available crystal structures of SKs (and their complexes) from a variety of organisms (listed in Supplementary Table S1).

Representative structures of apo- (Figure 1B, bottom panel) and ATP•Mg²⁺-bound (Figure 1C, bottom panel) SK_Mtu extracted from the regions of maximal density in θ−|h| space of the corresponding ensembles show altered conformations of key catalytic site elements. In the presence of ATP•Mg²⁺ i.e., in the WCS, the Walker-A Lys (K15) assumes a downwards orientation and contacts the β− and γ−phosphates of ATP; the first Asp (D32) of the DxD motif forms a hydrogen bond with the Walker-A S16, thus functionally mimicking the absent Walker-B Asp (Leipe et al., 2003). The D32-S16 interaction is necessary for the optimal coordination of Mg²⁺ (Figure 1C, bottom panel). Indeed, as shown in Supplementary Figure S4B, the D32,Cγ−S16,Oγ distance is well defined and shows a narrow distribution (3.7 ± 0.2 Å) within the ensemble. In the absence of ligands, i.e., in the WOS, K15 rotates outwards to establish a hydrogen-bond with S77 of the Walker-B motif (Figure 1B, bottom panel) in a significant fraction of the structures. In this configuration, the D32-S16 distance shows a broad distribution (6.0 ± 1.2 Å), suggestive of increased disorder, with a maximum reflective of significantly increased separation (Supplementary Figure S4B) suggesting a conformation that is not conducive for Mg²⁺ coordination. Conformations displaying the K15-S77 hydrogen bond have been observed in several crystal structures of SK_Mtu, that are, not surprisingly, free of Mg²⁺. The structures (representative examples shown in Supplementary Figure S4C) (Hartmann et al., 2006) that display the K15-S77 interaction have the highest values of θ, and correspondingly, the lowest values of |h|, and exist at the outer edges of the WOS. Taken together, these results suggest that the global conformations of SK_Mtu are coupled to the conformational state of the Walker-A Lys as was observed for Wzc_CDΔC (Hajredini et al., 2020; Hajredini and Ghose, 2021).

As mentioned above, in a significant fraction of the structures that comprise the WOS of SK_Mtu, the Walker-A K15 forms a hydrogen bond with the Walker-B S77 (Figure 1B), generating a conformation that mirrors that seen in the Wzc_CDΔC OS. In the latter case, the equivalent Walker-A Lys (K540) forms a salt bridge with the Walker-B D642. The β-strand housing the Walker-B motif in SK_Mtu is displaced relative to that in Wzc_CDΔC, this enables S77 from the degraded Walker-B in the former to assume a spatial position that is approximately equivalent to the Walker-B D642 in the latter to facilitate an interaction with K15 (Figure 2A). It would therefore appear from our current results, that small local structural rearrangements allow the polar Ser to substitute for the now missing Walker-B Asp in preserving the interaction with the Walker-A K15 to maintain the open conformation (WOS) in SK_Mtu. However, despite these similarities, there are differences in stability between the open conformations involving the Walker-A Lys of SKs and BY-kinases. In Wzc_CDΔC, this interaction mode, as assessed through the K540,Nξ−D642,Cγ distance distribution, is well defined over the entire OS ensemble, showing a single peak with a maximum at ∼3.3 Å (green trace in Figure 2B). The corresponding distance, K15,Nξ–S77,Oγ in SK_Mtu, shows a broad distribution with the dominant peak centered at ∼2.8 Å but containing only approximately 50% of the population in the WOS (pink trace in Figure 2B; this disorder is also evident from the bottom panel of Figure 1B). This suggests that while the orientation of the Walker-A Lys is well-defined in Wzc_CDΔC, it is somewhat more dynamic in SK_Mtu. The reasons for this difference are analyzed in detail below.

As mentioned earlier, the conformational dynamics of the LID and SB domains have been suggested to be critical for function in the SKs in serving to stabilize the reaction-compatible state, and in facilitating product release (Blanco et al., 2013). The dynamics of these entities have been suggested as suitable targets for drug discovery (Prado et al., 2016). For the sake of completeness, we also analyzed the local conformational variability in the REST2-generated ensembles of SK_Mtu. There is, not unexpectedly, a slight decrease in overall flexibility upon binding ATP•Mg²⁺ (Supplementary Figure S5A), and transition from the WOS to the WCS. The LID domain that is the most dynamic region of the protein in the unliganded state remains so in the presence of ATP•Mg²⁺, though there is some decrease in its overall flexibility. The SB domain is also dynamic in both states of SK_Mtu, albeit less so than the LID. To probe the presence of opening/closing motions of the SB domain with respect to the protein core, we defined a distance (d_47-109, Supplementary Figure S5B, top left panel) between the Cα atoms of D47 (one of the more dynamic parts of the SB domain and therefore likely to sense these motions, if any) and R109 (near start of the LID domain; one of the more rigid parts of the protein in its ATP•Mg²⁺-bound state). In the absence of ATP•Mg²⁺ a somewhat bimodal distribution of distances is seen (Supplementary Figure S5B top right panel). Analyses of structures corresponding to the two maxima indicate an open conformation with greater separation between the LID and SB domains (Supplementary Figure S5B bottom right panel), and a more closed conformation where this distance is reduced (Supplementary Figure S5B bottom middle panel). Only the latter conformation is seen in the ATP•Mg²⁺-bound state (Supplementary Figure S4B bottom left panel).

While S77 of SK_Mtu appears to substitute for the Asp/Glu residue of a canonical Walker-B motif in maintaining the WOS through its interaction with the Walker-A Lys, this position is poorly conserved in SKs. In contrast to the ϕϕSLGGG sequence in SK_Mtu, the consensus sequence for Walker-B motifs of SKs is ϕϕϕTGGG (Leipe et al., 2003); the corresponding consensus in BY-kinases is ϕϕϕϕDT (Leipe et al., 2002). Notably, in Wzc_CDΔC, the conserved Walker-B Thr (T643) participates in multiple interactions that also serve to stabilize the OS (Figure 3A, top panel). T643 forms a hydrogen bond through its backbone carbonyl with the K540 sidechain (supplementing the interaction of the latter with D642), and an additional hydrogen bond through its sidechain hydroxyl with the backbone carbonyl of M531 on the adjacent β-sheet. Projection of the unliganded Wzc_CDΔC ensemble (Hajredini et al., 2020) onto the 2-dimensional space spanned by the K540-T643 (D1: K540,Nξ−T643,O) and T643-M531 (D2: T643,Oγ1−M531,O) distances, indicates a single major state (white arrow, Figure 3A, bottom panel). This robust network of interactions (K540-D642, K540-T643 and T643-M531) is the likely reason for the well-defined orientation of K540 in the OS (as is evident from the green trace in Figure 2B). This set of interactions is drastically altered in the CS (T643 now interacts T532, D642 interacts with T541, the latter interaction is necessary to optimally engage Mg²⁺) and this is the likely cause of the well-defined separation between in the OS and the CS upon comparing the apo- and ATP•Mg²⁺ complexes of Wzc_CDΔC (Supplementary Figures S3B,C).

To test whether a similar interaction mode involving the Walker-B Thr also stabilizes the WOS in SKs that carry the more conventional ϕϕϕTGGG sequence, we performed an additional set of REST2 simulations on the unliganded state of Coxiella burnetii SK (SK_Cbu). Inspection of the structural ensemble of apo-SK_Cbu reveals the presence of hydrogen bonds between the backbone carbonyl and sidechain hydroxyl of the Walker-B Thr (T81) with the sidechain of the Walker-A K18 and the backbone carbonyl of L10 (that lies on the adjacent β-sheet), respectively (Figure 3B, top panel), mirroring the interaction mode seen in Wzc_CDΔC. However, unlike in Wzc_CDΔC, projection of the resultant ensemble onto the 2-dimensional space spanned by the K18-T81 (D1: K18,Nξ−T81,O) and T81-L10 (D2: T81,Oγ1−L10,Ο) distances reveals a more diffuse set of states (Figure 3B, bottom panel). Nevertheless, a disruption in the T81-K18 interaction leads to a corresponding disruption of the T81-L10 interaction (red arrow in Figure 3B, bottom panel) suggesting that these two sets of interactions are coupled as in Wzc_CDΔC. It is of note that the canonical Walker-B Thr in SK_Cbu forms two sets of the hydrogen bonds involving both its backbone and its sidechain to stabilize the WOS. This contrasts a single hydrogen bond involving the sidechain of the unique Walker-B Ser (S77) in SK_Mtu. One can therefore expect the WOS in SK_Cbu to be somewhat more closed than in SK_Mtu. Inspection of Figure 3C shows that this is indeed the case with reduced θ values (SK_Cbu: 190 ± 5°, SK_Mtu: 196 ± 4°) and a corresponding increase in the |h| values (SK_Cbu: 4.8 ± 0.6 Å, SK_Mtu: 4.6 ± 0.7 Å).

Based on the discussion above, an SK Walker-B motif may be generalized by the following sequence: ϕϕαβGGG; the α- and β-positions represent structural equivalents of the BY-kinase Walker-B Asp and Thr, respectively (Figure 4A). In unliganded Wzc_CDΔC, both the α- and β-positions are optimal in that they contain polar residues. Thus, the Walker-A Lys (K540) can form hydrogen bonds with both the Walker-B Asp (D642 at the α−position) and with the Walker-B Thr (T643); the latter being a polar residue at the β−position is then able to utilize its sidechain to hydrogen bond with M531. This generates a mesh of interactions (shown schematically in Figure 4B) leading to a well-defined orientation of the Walker-A Lys (K540) in a stabilized OS. In contrast, the fact that only one of these two positions contains a polar residue for the two SKs analyzed here, with T81 (β) in SK_Cbu (Figure 4C) or S77 (α) in SK_Mtu (Figure 4D). Thus, only a subset of these interactions is possible in the SKs resulting in less well-defined orientation for the Walker-A Lys and a more diffuse WOS.