**2. The Di**ff**erent Hydrolysis States and the ATP-Mimics Used to Induce Them**

The pre-hydrolytic state, where ATP is bound to the protein, is often already associated with protein conformational changes [21–24]. In this state, the γ-phosphate adopts a tetrahedral geometry (note that a similar discussion also holds for GTP). This geometry can change to a trigonal-bipyramidal geometry generating a pentavalent terminal phosphate group [25,26]. Most of the analogues are mimicking the pre-hydrolytic state because their γ-phosphate (in case of nonhydrolyzable analogues) or the γ-phosphate-mimicking group adopts a tetrahedral geometry [27]. The most commonly used nonhydrolyzable analogues are: AMPPNP (adenylyl imidodiphosphate) [28], AMPPCP (adenylyl methylenediphosphate) [29], AMPCPP (alpha,beta-methylene-triphosphonate) and ATP-γ-S (adenosine 5'-(gamma-thiotriphosphate)) [30]. In addition, the pre-hydrolytic state appears to be mimicked by ADP-BeF<sup>x</sup> [27] (Figures 2 and 3A). BeF<sup>x</sup> forms a strictly tetrahedral complex (specific to the pre-hydrolytic state); a penta-coordinated bipyramidal geometry (describing the transition state, see below) is excluded in this case [26,27]. The nonhydrolyzable ATP analogues are not completely resistant to hydrolysis. While the rate of hydrolysis of these analogues is indeed significantly lower, several of them can still be hydrolysed by many ATPases [31–38]. This behaviour can differ from protein to protein, and an analogue can fail to mimic a pre-hydrolytic state or may mimic a different/uncomplete pre-hydrolytic state in certain cases [26,39–41]. This difference can be observed between distinct protein families, or even within the same family [42], as shown in this work for the two model systems discussed.

The transition state (the "ATP-is-ready-to-be-split" state) can be accessed by an associative and dissociative mechanism, which represent the two extreme cases discussed in the literature [43–48]. In the case of an associative mechanism, the phosphorus possesses a pentavalent geometry. The nucleophilic attack of a water molecule at the γ-phosphate, forming a H2O-P bond, in this scenario occurs before the leaving group departs and before the P-O bond breaks (similar to a SN2 nucleophilic substitution, see Scheme 1B,C). In contrast, in the case of a dissociative mechanism, the nucleophilic attack of a water molecule at the γ-phosphate occurs after the leaving group was released, generating a metaphosphate intermediate before it collapses onto the acceptor nucleophile (similar to a SN1 reaction). The transition state can be simulated by employing three prominent mimic groups in combination with ATP or ADP: aluminium fluoride (ADP:AlFx) [27,49,50], magnesium fluoride (ADP:MgFx) [51], and vanadate (ADP:Vi) [52]. In some enzymes, ATP hydrolysis is required prior to the binding of the transition-state mimic [53,54]. In structural studies, aluminium fluoride is most frequently used as a mimic of the γ-phosphate in the transition state, as evidenced by analysing the number of deposited structures in the PDB database (Figure 2A). When the analogue in the presence of ADP is complexed with the protein, ADP:AlF<sup>x</sup> is believed to mimic the transition state of an ATP molecule. Two configurations of this analogue have been observed: ADP:AlF<sup>3</sup> and ADP:AlF<sup>4</sup> <sup>−</sup>. In the ADP:AlF<sup>4</sup> <sup>−</sup> mimic (two-thirds of in the

PDB deposited AlFx-containing structures) the AlF<sup>4</sup> <sup>−</sup> group is in a squared-planar geometry and forms an octahedral complex with two oxygen ligands in the apical positions. While one ligand is provided by the β-phosphate, the other ligand comes from the hydrolytic water molecule in the attack position next to the phosphorus atom. It is believed that such a structure mimics the interaction of the catalytic water molecule with the γ-phosphate in the anionic transition state for phosphoryl transfer [49,55]. AlF<sup>3</sup> (one-third of in the PDB deposited AlFx-containing structures) is in a trigonal-planar geometry forming a bipyramidal complex resembling the geometry of the transition state [49,55]. γ γ

−

γ

γ

γ

**Figure 2.** Distribution of the different NTP analogues based on the number of structures in the Protein Data Bank in December 2019. Number of Protein Data Bank structures for each analogue (**A**). Number of ABC transporters (blue, search query "ABC transporter + protein data bank accession codes of the ATP-analogue") and helicases (red, search query "DNA helicase + PDB ID of the ATP-analogue") Protein Data Bank structures for each analogue (**B**).

γ − γ **Figure 3.** ATP analogues mainly used for structure determination. Left panel is the chemical structure and the right panel shows the protein-bound-structure of AMPPNP (**A**), AMPPCP (**B**), ATP-γ-S (**C**), AMPCPP (**D**), ADP:BeF<sup>x</sup> (**E**), ADP:AlF<sup>x</sup> (**F**), ADP:MgF<sup>3</sup> <sup>−</sup> (**G**), ADP:Vi (**H**) from the Protein Data Bank (https://www.rcsb.org). The protein-bound-structures of ATP analogues were generated after alignment of their adenosine moieties. In order to compare the molecules among each other, only the first 100 structures with the smallest RMSD were selected for AMPPNP, AMPPCP, ATP-γ-S and AMPCPP.

γ

−

β

MgF<sup>3</sup> <sup>−</sup> shows nearly the same geometry as AlF<sup>3</sup> but carries a negative charge similarly to the anionic γ-phosphate in the transition state. AlF<sup>3</sup> and MgF<sup>3</sup> <sup>−</sup> are structurally similar and have similar scattering factors for X-rays; therefore, it has been suggested that MgF<sup>3</sup> <sup>−</sup> is present in some crystal structures, which are indicated as containing NDP:AlF<sup>3</sup> [56]. Indeed, Mg2<sup>+</sup> ions are usually present in the samples as cofactors of NTP hydrolysis. In contrast to X-ray, NMR can differentiate the two metal fluorides so that in few cases, the presence of MgF<sup>3</sup> <sup>−</sup> in the active site was directly shown [56–58]. For more information about such cases and the use of metal fluorides as ATP or phosphate analogues, we refer the reader to the two comprehensive recent reviews [55,58]. Finally, vanadate-containing ATP:Vi or ADP:Vi are used as a transition-state mimic for a variety of proteins. Vanadate is an oxoanion of vanadium which shares structural and chemical similarities with phosphate molecules mimicking the hydrolysis transition state [52,59]. It is known that the simple form of the oxoanion (VO<sup>4</sup> <sup>3</sup>−) can adopt a penta-coordinated, trigonal bipyramidal geometry around the central vanadium in presence of ADP [60]. These properties make the vanadate a phosphate mimic of the transition state for phosphoryl transfer so that vanadate acts as an inhibitor for some ATPases. As previously described by Davies et al. [52], vanadate can be used to mimic phosphoryl transfer, and structures of different protein families including myosin [61,62], dynein [63], kinesin [40], ABC transporters [60,64,65], heat shock protein (Hsp70s) [66], NS3 helicase (dengue virus) [67], nucleoside-diphosphate kinase [68] or F1-ATPase [69] are reported. The main advantage of vanadate is that it can form covalent bonds with the oxygens of phosphate groups from ADP or other ligands [52]. Interestingly, this is not always the case [68], as there are structures where a vanadate is not bound to ADP, but still stabilizes the transition state. It is also noteworthy that vanadate does not work as an inhibitor or as transition-state mimic for all proteins with ATPase activity [70].

Finally, the post-hydrolytic state corresponds to a situation where the nucleotide diphosphate and the previously associated γ-phosphate are separated, but both are still bound to the protein, or, alternatively, where the γ-phosphate is already released, and only ADP is bound to the protein. The post-hydrolytic state where the γ-phosphate is not released can be mimicked not only by an orthophosphate [67,71,72] but also by a sulphate ion, SO<sup>4</sup> <sup>2</sup><sup>−</sup> [71,73–75]. Note that sulphate ions have only two ionisable oxygens (with pK<sup>a</sup> below 2) [76].

The overall conformational variability of NTP analogues can be seen by overlaying the structures extracted from the PDB and by aligning them on their nucleoside parts (Figure 3). AMPPNP and ATP-γ-S adopt a wider range of conformations (Figure 3A,C) than AMPPCP and ADP:BeF<sup>x</sup> (Figure 3B,E), although this allows for a qualitative statement only, since the total numbers of deposited structures in the PDB are different (see Figure 2). AMPPNP and ATP-γ-S thus seem to adapt their conformation to the protein-binding pocket, while for AMPPCP and ADP:BeF<sup>x</sup> it may be the protein that adapts. For the transition-state analogues, it is difficult to make the same comparison due to the small number of structures available. However, ADP:AlF<sup>x</sup> shows a significant distribution of structures as well (Figure 3F).

In sum, from the eight mainly used analogues for structural studies, five are used to mimic the pre-hydrolytic state: AMPPNP, AMPPCP, ATP-γ-S, AMPCPP and ADP:BeFx, three to mimic the transition state: ADP:AlFx, ADP:MgF<sup>x</sup> and ADP:V<sup>i</sup> , and ADP and ADP:SO<sup>4</sup> <sup>2</sup><sup>−</sup> to mimic the post-hydrolytic state. Note that also other NTP analogues exist that differ structurally through the introduction of atoms or groups (e.g., fluorescent probes, biotin groups, etc.) on the base, sugar, or triphosphate regions of the molecule [77–79]. A complete overview is given in reference [77].
