Spectroscopic glimpses of the transition state of ATP hydrolysis trapped in a bacterial DnaB helicase

The ATP hydrolysis transition state of motor proteins is a weakly populated protein state that can be stabilized and investigated by replacing ATP with chemical mimics. We present atomic-level structural and dynamic insights on a state created by ADP aluminum fluoride binding to the bacterial DnaB helicase from Helicobacter pylori. We determined the positioning of the metal ion cofactor within the active site using electron paramagnetic resonance, and identified the protein protons coordinating to the phosphate groups of ADP and DNA using proton-detected 31P,1H solid-state nuclear magnetic resonance spectroscopy at fast magic-angle spinning > 100 kHz, as well as temperature-dependent proton chemical-shift values to prove their engagements in hydrogen bonds. 19F and 27Al MAS NMR spectra reveal a highly mobile, fast-rotating aluminum fluoride unit pointing to the capture of a late ATP hydrolysis transition state in which the phosphoryl unit is already detached from the arginine and lysine fingers.

A denosine triphosphate (ATP)-driven motor proteins play a key role in various cellular processes 1 . For example, motor proteins belong to the class of ATPases, which hydrolyze ATP into ADP (adenosine diphosphate) and inorganic phosphate to gain chemical energy allowing such enzymes to drive further chemical or mechanical events 2 . Structural insights into the functioning of these molecular machines are not straightforward to obtain, neither by X-ray crystallography, nor by cryo-electron microscopy or NMR spectroscopy due to the difficulty in trapping the intermediate catalytic states occurring during ATP hydrolysis. Diverse ATP analogues can be employed to mimic different stages of ATP hydrolysis as closely as possible 3,4 which, in combination with molecular dynamics simulations 5 , can give mechanistic insights into complex biomolecular reaction coordinates. Of particular interest in unravelling the ATP hydrolysis reaction mechanism is the transition state of the phosphoryl (PO 3 − ) transfer reaction (see Fig. 1 for a sketch of the limiting case of an associative ATP hydrolysis mechanism 6 ). Metal fluorides have been found to mimic such states for structural studies, mostly using X-ray crystallography 7,8 . The number of deposited protein structures containing analogues such as AlF 4 − , AlF 3 and MgF 3 − has strongly increased in the past years 7 . AlF 4 − forms together with the phosphate oxygen atom of ADP as well as an apical water molecule an octahedral complex mimicking the "in-line" anionic transition state of phosphoryl transfer, whereas AlF 3 and MgF 3 − form trigonal-bipyramidal complexes 7 . The formation of AlF 4 − or AlF 3 is controlled by pH, the latter being favored at lower pH values 9 . However, some concern regarding their discrimination, e.g. distinction between AlF 3 and MgF 3 − , has been raised and it could be shown by 19 F NMR spectroscopy that some complexes, which were believed to contain AlF 3 , contain instead MgF 3 − 9 . Similarly, in lower resolution X-ray structures (>2.8 Å) AlF 3 cannot be distinguished unambiguously from AlF 4 − 9, 10 .
We present magnetic resonance approaches using EPR and solid-state NMR to obtain spectroscopic insights into the transition state of ATP hydrolysis which we trap for the oligomeric bacterial DnaB helicase from Helicobacter pylori (Hp, monomeric molecular weight 59 kDa) by using the transition-state analogue ADP:AlF 4 − . The motor domain of the helicase belongs to P-loop fold nucleoside-triphosphatases (P-loop NTPases), one of the largest protein families, which includes motor proteins like myosins, kinesins, and rotary ATPases. About 10-20% of genes in any genome encode for diverse P-loop NTPases 11 . In these enzymes, ATP or guanosine triphosphate (GTP) molecules are bound to the so-called Walker A motif GxxxxGK[S/T] of the signature P-loop of the motor nucleotide-binding domain (NBD) 12,13 . Bacterial DnaB helicases, which use the energy of ATP hydrolysis to unwind the DNA double helix, belong to the ASCE division of P-loop NTPases. The members of this division are characterized by an additional β-strand in the P-loop and a catalytic glutamate (E) residue next to the attacking water molecule [14][15][16] . Within the ASCE division, DnaB helicases are attributed to the RecA/F1 class 16 . Generally, P-loop NTPases need to be activated before each turnover because otherwise, they would promptly consume the entire cellular stock of ATP and GTP. As inferred from the comparative structure analysis of NTPases with transition-state analogues, such as NDP:AlF 4 − or NDP:MgF 3 − , the activation is mostly achieved by the insertion of a positively charged hydrolysis-stimulating moiety (usually, a positively charged arginine or lysine "finger" or a potassium ion) between the αand γ-phosphates 7, [17][18][19] . As shown by MD simulations, linking of αand γ-phosphates by the stimulating moiety leads to rotation of the γ-phosphate group yielding a hydrolysis-prone conformation of the triphosphate chain 20 . In DnaB helicases, activation is achieved through the interaction with the neighboring domain that provides a pair of hydrolysisstimulating Arg and Lys residues. RecA-type ATPases generally differ from most other P-loop NTPases in that their stimulating residues, which operate in a tandem, interact upon activation only with the γ-phosphate, but not with the α-phosphate group. This interaction triggers the ATP hydrolysis; the triggering mechanism, however, has yet to be determined 21 .
HpDnaB belongs to the class of SF4-type, ring-shaped DnaB helicases, and only two crystal structures in complex with ATP analogues have been reported so far, namely the one from Aquifex aeolicus 22 (AaDnaB, PDB accession code 4NMN) Bacillus stearothermophilus 21 (currently Geobacillus stearothermophilus, BstDnaB, PDB 4ESV). AaDnaB is complexed with ADP only and BstDnaB with GDP:AlF 4as well as singlestranded DNA, which is similar to our ADP:AlF 4 − complex that we study in presence and absence of DNA herein.
W-band electron-electron double resonance (ELDOR)-detected NMR (EDNMR) [23][24][25][26] and electron-nuclear double resonance (ENDOR) 27,28 allow for the positioning of the divalent metal ion within the active site by identifying nuclei in its vicinity. The native Mg 2+ co-factor is replaced for such studies by the EPRobservable paramagnetic Mn 2+ analogue 29 (the biological functionality is maintained to about 80% under such conditions compared to the one observed in presence of Mg 2+ ) 30 . 19 F, 27 Al and 31 P nuclear resonances were observed among others in the EDNMR spectra, proving the binding-mode of ADP:AlF 4 − . The extracted 31 P hyperfine coupling constants and the detection of 19 F and 27 Al nuclei in the proximity of the co-factor point to a coordination of the Mn 2+ ion to the β-phosphate of ADP and AlF 4 − .
Solid-state NMR can identify the amino-acid residues involved in the coordination of the ATP analogue. Protons are of particular interest as their resonance frequencies can contain information regarding their engagement in hydrogen bonds. Fast magic-angle spinning (MAS) nowadays provides sufficient spectral resolution for proton-detected sidechain studies 31 . Indeed, proton detection at fast MAS has become an important tool in structural biology in the past years for unraveling protein structures [32][33][34][35][36][37][38][39][40][41] , to characterize RNA molecules 42 and protein-nucleic acid interactions [43][44][45] , and to address protein dynamics [46][47][48][49][50][51][52] . A key advantage of solid-state NMR is the straightforward sample preparation, which simply consists of sedimentation from solution into the solid-state NMR rotor without requiring crystallization steps 53,54 yielding long-term Fig. 1 Sketch of the associative ATP hydrolysis mechanism with a trigonal-bipyramidal transition state. ‡ indicates the transition state. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-25599-z stable protein samples 55 . A further advantage is the sensitivity of NMR in identifying hydrogen bonds, which is often not achievable by standard structure determination techniques, such as X-ray crystallography or cryo-electron microscopy (EM), in which resolution in the order of 1 Å (for cryo-EM a slightly lower resolution might be sufficient 56 ) has to be achieved. We have already previously reported that the transition state of ATP hydrolysis is accessible for DnaB by employing the ATP-analogue ADP:AlF 4 − 57 , but the direct identification of hydrogen bonds required for characterizing the noncovalent interactions driving molecular recognition of both, ATP and DNA, was hardly possible and only spatial proximities derived from 31 P-13 C/ 15 N correlation experiments or the proton chemical-shift values were explored 44 .
We herein identify protein residues engaged in hydrogen bonding to the phosphate groups of nucleotides (ADP:AlF 4 − and DNA) by (i) measuring high-frequency shifted proton resonances characteristic for hydrogen-bond formation 58 , (ii) probing spatial proximities in dipolar-coupling based proton-detected 31 P, 1 H correlation experiments at fast MAS (105 kHz) and (iii) using the temperature dependence of 1 H chemical-shift values as a probe for hydrogen bonding, an approach well known in solution-state NMR [59][60][61] , and recently extended to the solid state 62 . Note, that we herein report a proton-detected 31 P, 1 H correlation spectrum at fast MAS frequencies using a sub-milligram sample amount, which was so far, to the best of our knowledge, not possible with any of the previous equipment. This is an important step for proving hydrogen bonding in protein-nucleic acid complexes ranging from proteins involved in DNA replication or virus assemblies by solid-state NMR and to derive nucleotide-binding modes, even in quite large systems as the one we looked at. From a combination of (i)-(iii), key contacts between the ADP phosphate groups and residues located in the Walker A motif were identified, as well as two hydrogen bonds to the phosphate groups of the two DNA nucleotides. To complement our spectroscopic characterization of the ATP hydrolysis transition state, we performed 19 F and 27 Al MAS experiments to access information about bound AlF 4 − . The spectra indicate a fast rotation of the AlF 4 − unit implying that AlF 4 − is not rigidified by coordinating protein residues indicating that the ADP:AlF 4 − trapped state of DnaB possibly describes a late transition state, just after the bond fission, but before the release of the phosphate group from the catalytic pocket.

Results
EPR enables the positioning of the metal ion co-factor within the active site. Binding of ADP:AlF 4 − to the protein is revealed in EDNMR experiments, which employ the hyperfine couplings between a paramagnetic center and nearby nuclei to detect the latter. EDNMR has been used to characterize transition states of ATP hydrolysis, often in the context of ABC transporters for which such a state is successfully mimicked by ADPvanadate 63,64 . Figure 2a shows the Mn 2+ EDNMR spectrum of DnaB complexed with ADP:AlF 4 − (red) compared to the reference spectrum of DnaB complexed only with ADP (cyan), here using a non-13 C/ 15 N labeled protein (see below for 13 C/ 15 N labeling). While in both spectra couplings to 31 P nuclei are observed, additional peaks for 19 F and 27 Al are detected only for the ADP:AlF 4 − bound state consistent with the presence of AlF 4 − in the NBD of DnaB. The weakly coupled 19 F is also observed by ENDOR (see Supplementary Fig. 2 for the spectrum and discussion of extracted parameters) and indicates its proximity to Mn 2+ . To corroborate that these resonances are due to DnaBbound Mn 2+ :ADP:AlF 4 − and to rule out that these correlations originate from the formation of the Mn 2+ :ADP:AlF 4 − complex in solution, we recorded EDNMR spectra on a frozen control solution in the absence of protein and indeed we do not observe any 19 F and 27 Al resonances (purple spectrum in Fig. 2a). Interestingly, in the presence of protein, two groups of 31 P resonances are detected: a hyperfine-split doublet (denoted 31 P d in Fig. 2a) and an unresolved doublet (denoted 31 P u ). Davies 31 P Electron-Nuclear DOuble Resonance (Davies ENDOR) 27 experiments were performed on the Mn 2+ -containing protein complex (Fig. 2b) to extract the hyperfine tensor A of the doublet. Line shape simulations yield a large A iso value of 4.7 MHz (for all 31 P hyperfine tensor parameters extracted from the spectrum see Supplementary Table 1). This value is similar to published values for an ADP:Mn 2+ complex in which the Mn 2+ ion binds symmetrically to the two ADP phosphate groups 65,66 or an ATP:Mn 2+ complex 67 . Mims ENDOR 28 experiments were performed to detect the small A iso value of the in EDNMR unresolved doublet which is determined to be 0.3 MHz (see Fig. 2c). Mims ENDOR measurements on the control solution did not show this doublet. We assign the large A iso value (4.7 MHz) to 55 Mn- 31 Pβ and the small A iso value (0.3 MHz) to 55 Mn- 31 Pα hyperfine couplings indicating that the Mn 2+ ion is located much closer in space to the Pβ atom of ADP than to the Pα atom. This assignment is supported by Density Functional Theory (DFT) calculations of the hyperfine coupling tensors performed on small clusters mimicking the Mn 2+ coordination sphere extracted from the available crystal structures of SF4 helicases (BstDnaB:GD-P:AlF 4 -:DNA 21 and AaDnaB:ADP 22 ) although it has to be noted that the uncertainty in the exact metal ion position due to insufficient resolution of the electron density and the initial presence of Ca 2+ instead of Mg 2+ in the 4ESV structure might be significant and influence the results of the calculations (Supplementary Table 1 and Supplementary Fig. 2).
We additionally performed EDNMR experiments using uniformly 13 C/ 15 N labeled DnaB complexed with Mn 2+ :AD-P:AlF 4 − . While the same 19 F, 27 Al and 31 P features discussed above are present in the spectrum, additional intense 13 C and 15 N resonances are observed ( Supplementary Fig. 3). In combination with the absence of such resonances in the corresponding reference spectrum measured in the absence of protein ( Supplementary Fig. 3), this provides further evidence for binding of Mn 2+ :ADP:AlF 4 − to DnaB. Note that the EPR experiments were performed on the protein complex in absence of DNA in contrast to most solid-state NMR experiments described below. As described in earlier work, the ADP:AlF 4 − states in presence and absence of DNA are highly similar 57 and we thus recorded EPR experiments only on one of these complexes.  Fig. 3b for the atomic numbering) recorded at 17 kHz MAS 57 . Two narrow resonances are detected for both, the Pα and Pβ of ADP (at −6.0 and −7.1 ppm, respectively) as well as for the DNA phosphate groups (at 0.5 and −1.1 ppm). The latter observation reflects that two DNA nucleotides bind to one DnaB monomer, which is characteristic for SF4-type helicases 44,57 . Proton-detected NMR experiments at fast MAS frequencies (>100 kHz) allow the identification of protons engaged in hydrogen bonds requiring only small amounts of protein in the order of 0.5 mg. The 1 H NMR chemical-shift value serves as a sensitive indicator for the formation of hydrogen bonds: a de-shielding effect is observed if protons are engaged in such interactions 44,45,58,69 . However, the chemical shift alone is not a sufficient criterion to prove hydrogen bonding. We therefore extend the experimental approaches to directly detecting such interactions by the presence of throughspace 31 P, 1 H dipolar couplings in hPH correlation experiments at 105 kHz MAS. The hPH spectra were recorded with two different 1 H-31 P CP contact times (1.5 and 3.5 ms) on a 13 C, 15 N uniformly labeled, deuterated and 100% back-exchanged sample of DnaB in which the ADP and the DNA remained at natural abundance. Note that this deuterated version of the protein has been chosen over a fully protonated sample to increase the intrinsically rather low signal-to-noise ratio in such a large protein due to the narrowing of the proton resonances by roughly a factor of three attributed to the dilution of the proton dipolar network (see Supplementary Fig. 4 for the proton line-widths determined for a deuterated and fully protonated sample) 70 . Figure 3c shows the rather sparse 2D hPH correlation spectrum (with 3 ms CP contact time) of the DnaB complex and indeed protein-phosphate correlations to all four 31 P resonances observed in Fig. 3a are visible. The CP-based hPH experiment proves spatial proximities between proton nuclei in the vicinity of the phosphate groups. An INEPT-based experiment transferring polarization directly over the hydrogen bond via the J-couplings (typical 2 J( 31 P-1 H) values are in the order of 3 Hz 71,72 ) was not successful due to a too short proton transverse relaxation time compared to the required INEPT transfer delay period (see Supplementary Fig. 5). The resonance assignments shown in Fig. 3c and Supplementary Fig. 6 (CP contact time of 1.5 ms) were obtained using the deposited proton chemical-shift values (BMRB accession code 27879). The hPH spectra reveal intense signals and thus spatial correlations between Pβ of ADP and S206, G208, K209 and T210, all located in the conserved Walker A motif of the P-loop in the motor domain of the helicase 73 . Note that for all mentioned amino acids correlations to the backbone amide protons are observed, except for K209 for which additional sidechain Hζ protons are detected. For the Pα resonance of ADP only weak correlations are observed, the strongest one to S211 and an unassigned resonance, 40   . The assignments of the peaks to the nuclear resonance frequencies are shown. The cyan spectrum is reproduced from ref. 64 . Asterisks mark a broad, currently unassigned feature that is possibly due to Mn double quantum or combination lines 23  possibly an arginine residue, which has been detected in previous NHHP experiments 44 (R242 or the "arginine finger" R446 from a neighboring DnaB subunit). The main difference in the spectrum recorded at shorter CP contact times ( Supplementary Fig. 7) is that correlations to the ADP and DNA protons (sugar and base) present in the spectrum recorded at 3.5 ms contact time (highlighted in light red and green in Fig. 3c) are absent. It is important to note that the herein described hPH experiments appear to be much more selective for detecting direct coordination partners than the previously described 1 H-1 H spin-diffusion based NHHP and CHHP experiments 74 and possibly also TEDOR experiments 75 , thereby providing a more detailed picture of the local geometry around the phosphate groups of ADP and DNA 4,76 than reported previously 44 . The hPH spectrum in Fig. 3c also contains important information regarding the DNA coordination. Actually, only two intense backbone amide correlations to the two DNA phosphate groups, D374 in case of P1 and G376 in case of P2 are observed. Together with our previous observation of K373 forming a salt-bridge to P2 via the lysine sidechain, only three contacts seem to coordinate the DNA in this molecular recognition process.
The high-frequency shifts of their amide protons and their spatial proximity to the phosphate ADP group already point to the engagement of K209 and T210 in hydrogen bonding as discussed above. To further verify this, we determined the temperature dependence of their chemical shifts between 294 and 302 K (sample temperatures, see Methods Section). Due to their characteristic chemical shifts (and thus their isolated position in the 2D fingerprint spectrum) the temperature dependences could be directly extracted from 2D CP hNH experiments. It is well known from solution-state NMR that the chemical shifts of protons in strong intramolecular hydrogen bonds experience only a weak temperature dependence 60,61 as recently also shown by solid-state NMR 62 . However, for protons in rather weak hydrogen bonds, the resonances become significantly more shielded upon increasing the temperature, due to an increase in the average hydrogen-bond length. Figure 4 shows the temperature dependence for residues identified in the hPH spectra (left column). Indeed, K209 and T210, previously identified as forming hydrogen bonds to the Pβ of ADP, show an almost vanishing temperature coefficient (slope of the corresponding linear regression). Similar values are found for D374 and G376 The protein resonance assignment is taken from ref. 44 (BMRB accession code 27879). Regular-printed residue labels: Chemical-shift deviation to reported proton shifts <0.05 ppm. Italic-printed residue labels: Chemical-shift deviation to assigned proton shifts ≥0.05 ppm. All proton shifts are assigned to amide backbones, except the ones indicated by an asterisk, which are associated to sidechain atoms. Correlations between the phosphate groups and ADP or DNA are highlighted in green and light red/purple, respectively. The assignments of the DNA proton resonances are based on average chemical-shift values reported in the BMRB database (www.bmrb.wisc.edu). The pink dashed lines highlight signals from insufficiently suppressed DNA in solution.
( Supplementary Fig. 7) assumed to be involved in DNA coordination. In contrast, Fig. 4 (right column) shows resonances associated with a larger temperature coefficient thus not being involved in hydrogen bonds (for all extracted temperature coefficients see Supplementary Fig. 7).
Solid-state NMR shows that the AlF 4 − unit is highly mobile. The AlF 4 − unit can be detected in 19 F-and 27 Al-detected MAS experiments. Figure 5a displays the 19 F MAS spectrum of DnaB:ADP:AlF 4 − in the presence and absence of DNA. Interestingly, only one 19 F resonance line at around −146 ppm is detected for the protein-bound AlF 4 − group pointing to a fast chemical-exchange process, most probably a rotation of the unit (vide infra, for the 19 F spectrum in the absence of protein see Supplementary Fig. 8). The additional sharp 19 F resonances visible in the spectra are attributed to the excess of AlF 4 − and related species present in the supernatant of the NMR rotor (roughly 50 weight percent after sedimentation 77 ). Around 4% of the AlF 4 − remains in the supernatant after the rotor-filling step. The resonance assignments displayed in Fig. 5a are based on reported solution-state NMR assignments 78,79 . A similar chemical-exchange process has also been observed for the RhoA/ GAP:GDP:AlF 4 − complex 80 , for the GTPase hGBP1 81 and for the motor protein myosin 78 in solution-state 19 F NMR experiments.
The 27 Al satellite transition NMR spectrum (SATRAS, Fig. 5b) of the sideband family is observed at δ iso = −0.2 ppm pointing to an octahedral coordination geometry of the 27 Al nucleus 82 . The spectrum allows to extract the quadrupolar coupling constant (C Q ) which amounts to only~570 kHz. The central m = 1/2↔m = −1/2 transition is observed at a similar resonance shift indicating a small contribution of the second-order quadrupolar shift. The C Q -value is significantly lower than expected for a six-fold oxygen/fluorine coordinated aluminum species. C Q -values for crystalline aluminum hydroxyfluorides are typically in the order of 5 MHz 83 . We attribute this effect to a rotation on the NMR time scale of the AlF 4 − unit around an axis inclined by an angle θ with respect to the direction of the principal component of the electric field gradient tensor (V zz , see Fig. 5c). The angle θ must be close, about 5-10°, to the magic angle (54.7°) leading to the significant reduction of the anisotropy of the quadrupolar interaction (see Fig. 5d). Similar observations were made for the DnaB complex in the absence of DNA (see Supplementary Fig. 9). Alternatively, the reduction of the quadrupolar interaction could be achieved by a rotational diffusion process, in which the angle θ varies randomly and is on average close to the magic angle. Note that the coordination of AlF 4 − to the β-phosphate of ADP is also reflected in a lowfrequency shift of the corresponding 31 P ADP resonance (−4.5 ppm compared to the ADP-bound state 57 ), which is a similar trend as observed for aluminophosphate gels and glasses 84 .
The rotational motion or even diffusion of this unit (with a correlation time shorter than the inverse quadrupolar coupling constant) reflects the absence of tight binding either to the protein (e.g. via hydrogen bonds to the fluorine atoms) or to the metal ion co-factor. The 27 Al isotropic chemical-shift value of close to 0 ppm is characteristic for an octahedrally coordinated Al-species, in our case most likely formed by four fluoride ligands, one oxygen ligand from the ADP phosphate backbone and one water molecule originating from an "in-line" geometry of phosphoryl transfer 7 or the catalytic glutamate as observed in the BstDnaB structure 21 .
Homology modelling points to a free rotating AlF 4 − detached from lysine and arginine fingers in SF4 helicases. We performed homology modelling based on the available bacterial helicase structures to investigate whether the dynamic behaviour of the AlF 4 − moiety could be related to the activation mechanism of RecA NTPases. Although the crystal structure of the HpDnaB dodecamer is available (PDB accession code 4ZC0 85 ), its low resolution of 6.7 Å and the absence of either DNA or of bound nucleotides prevents its use for modelling the ADP:AlF 4 − interactions in the catalytic site of a DNA-bound protein. Therefore, we reconstructed the mechanism of the activation from analysis of the DNA-and Ca 2+ :GDP:AlF 4 − -containing BstDnaB structure (PDB accession code 4ESV, resolution 3.2 Å) 21 . In the BstDnaB structure, the Ca 2+ :GDP:AlF 4 − moieties are bound to five out of six catalytic centres ( Supplementary Fig. 10). Furthermore, the positions and orientations of the AlF 4 − moieties differ among the five catalytic sites (Supplementary Fig. 10). By considering these different configurations as mimics of different reaction intermediates, the reaction steps could be reconstructed in the following way: Generally, the interaction with a stimulating moiety enables the nucleophilic attack on the γ-phosphate group by an apically positioned water molecule 17 Fig. 6b, c and Supplementary  Fig. 8, suggests that Arg and Lys residues are able, together, to twist/tilt the γ-phosphate group, which is mimicked by AlF 4 − in Fig. 6b, c. While in Fig. 6a, b the stimulating Lys and Arg residues are H-bonded to AlF 4 -, its further movement away from the nucleotide, as seen in Fig. 6c, leads to the weakening of H-bonds or even their entire dissociation (note the longer distances indicated in Fig. 6c), possibly yielding an almost unbound AlF 4 − unit tilted relative to its catalytic position (compare Fig. 6c with Fig. 6a) in agreement with our solid-state NMR observations of a nearly freely rotating AlF 4 − moiety.

Discussion
The transition state of ATP hydrolysis in the bacterial DnaB helicase from Helicobacter pylori has been trapped by using the mimic ADP:AlF 4 − . Such metal fluorides have been successfully In our work, EPR experiments allow the localization of the metal ion co-factor with respect to the ADP:AlF 4 − unit. In the transition state of ATP hydrolysis for HpDnaB, the Mn 2+ ion is in spatial proximity to the β-phosphate group of ADP as well as the AlF 4 − unit (Fig. 7a) as concluded from the large 31 P hyperfine coupling constant to the Pβ (a significantly smaller one is found for the Pα atom) and 19 F and 27 Al resonances observed in EDNMR, respectively. The structures of the only SF4-type helicases solved crystallographically, namely the BstDnaB:GD-P:AlF 4 − :DNA and AaDnaB:ADP complexes (Acquifex aeolicus, PDB 4NMN 22 ), support the finding of a Mn 2+ coordination to the β-phosphate group as also supported by the DFT calculations of the 31 P hyperfine tensors revealing the same trends as observed experimentally (Supplementary Table 1, Supplementary Fig. 2). A similar experimental observation by EPR has been made for DbpA RNA helicase in complex with ADP 86 .
Hydrogen bonds were identified spectroscopically by combining the information of high-frequency shifted proton resonances, spatial proximities probed in hPH correlation experiments and proton chemical-shift temperature coefficients (see Table 1 for a summary). Similar to other P-loop NTPases, in the HpDnaB transition state trapped by solid-state NMR, residues S206, G208, K209, T210 and S211 of the Walker A motif were identified in coordinating the ADP phosphate groups by their backbone amino groups and by the sidechain of K209 yielding a dense hydrogen-bond network (see Fig. 7a for a schematic representation). DnaB helicases are characterized by a unique ARP[G/S] xGK[T/S] sequence of the Walker A motif with an Ala residue instead of Gly in the first position 16 . Homologous residues were found to coordinate the phosphate chain in the crystal structure of DnaB from Bacillus stearothermophilus (currently Geobacillus stearothermophilus) crystallized with Ca 2+ :GDP:AlF 4 − and DNA (PDB accession code 4ESV 21 , see Fig. 7b and Supplementary Table 2 for the averaged distances to the oxygen atoms of the phosphate groups). An important difference is the only partial occupation of NBDs in BstDnaB with the transition-state analogue, whereas for HpDnaB all binding sites are occupied 57 and highly symmetric as revealed by the absence of evident peak splitting in the hPH spectra.
The hPH spectrum reveals two key contacts in DNA recognition by DnaB, namely the coordination of D374 and G376 to the two structurally distinct DNA phosphate groups P1 and P2. The de-shielded proton resonances in combination with the almost vanishing temperature coefficient found for D374 point to an engagement of these two protons in hydrogen bonding (Figs. 3  and 4). In previous studies, we have also identified the sidechain of K373 in forming a salt-bridge to P2 44 , which is also supported by the hPH spectrum showing a correlation of the resonance of the DNA phosphate group P2 to the K373 sidechain (Fig. 3c). These contacts are identical to those found in the crystal structure of the BstDnaB:DNA complex (backbone amide proton of E382 and G384 and the sidechain of R381) 21 thus revealing similarities in DNA recognition for these two SF4-type helicases. The protein proton resonances contacting the DNA are not broadened or even split into several peaks indicating that all six DnaB subunits engage the DNA in a highly similar way, pointing to a closed hexamer rather than an extended open structure as observed for BstDnaB (see Supplementary Fig. 11) 21 . This again agrees with our observation of a full saturation of all six NBDs with Mg 2+ :ADP:AlF 4 − therefore still indicating structural differences in the position of DnaB monomers in the HpDnaB and BstDnaB helicase complexes with transition-state analogues and DNA. An important feature revealed in our NMR analysis is the free rotational diffusion of the AlF 4 − unit mimicking the departing phosphate group during ATP hydrolysis. The averaging of the 27 Al quadrupolar coupling constant in combination with the single 19 F resonance observed indicate that the AlF 4 − unit (Fig. 5) is not coordinated tightly by the protein anymore, in contrast to the ADP for which we have observed a dense network of hydrogen bonds (Figs. 3 and 4). In contrast to the BstDnaB structure, in the HpDnaB complex studied herein ADP:AlF 4 − moieties are present in all six catalytic pockets 57 . The observed uniformity, however, comes in contradiction with the sequential operation of catalytic subunits, as observed in several studied oligomeric P-loop ATPases 87,88 . Their subunits operate one after another so that the catalysis in one subunit is thermodynamically promoted by the substrate binding to the other subunit 89,90 . Hence, only one site stays at any moment in the conformation catalytically active for ATP hydrolysis.
The here reported free rotational diffusion of all six AlF 4 − moieties within the tight hexamer of HpDnaB (Fig. 5) could be explained in the following way (see also the extended discussion in Supplementary Note 1): The exergonic binding of the first ADP:AlF 4 − moiety to a HpDnaB subunit (subunit 1) brings it into its catalytically active, DNA-bound configuration with the Arg (R446) and Lys (K444) fingers of the adjoining subunit 2 interacting with the ADP:AlF 4 − :H 2 O cat complex in the "catalytic" position ( Fig. 6a). This suggestion is supported by our earlier observation that ADP:AlF 4 − binding alone induces protein conformational changes and preconfigures the protein for DNA binding 57    Our results demonstrate that magnetic resonance is highly suitable to obtain structural and dynamic insights into the transition state of ATP hydrolysis of a bacterial DnaB helicase trapped by aluminum fluoride allowing a more profound understanding of the functioning of such complex motor proteins. EPR reveals the coordination of the metal ion co-factor to the β-phosphate group of ADP as well as to the AlF 4 − unit, whereas protondetected hPH solid-state NMR experiments combined with temperature dependences of proton chemical-shift values allow for identifying hydrogen bonds, which are crucial for the molecular recognition process of ADP and DNA binding to the DnaB helicase. NMR is one of the most sensitive techniques in proving hydrogen bonding with the additional advantage of shedding light onto dynamic processes, herein the free rotational diffusion of the AlF 4 − unit mimicking the phosphate group transferred during ATP hydrolysis.

Methods
Sample preparation. Protein expression and purification. The protein was cloned into the vector pACYC-duet1 (using the forward primer 5'-agtcatatggatcatttaaagcatttgcag-3' containing a NdeI restriction site and reverse primer 5'atactcgagttcaagttgtaactatatcataatcc-3' containing a XhoI site), and expressed in the E. coli strain BL21 Star (DE3) (One Shot® BL21 Star™ (DE3) Chemically Competent E. coli, Invitrogen™) 53 . Natural abundance and 13 C-15 N labeled HpDnaB was prepared in buffer A (2.5 mM sodium phosphate, pH 7.5, 130 mM NaCl) as described in ref. 53 . In short, DnaB was recombinantly expressed in presence of 13 C-glucose (2 g/L) and 15 N-ammonium chloride (2 g/L) as sole sources of carbon-13 and nitrogen-15. In case of the deuterated protein, the protein was expressed in D 2 O in presence of deuterated 13 C-glucose. The back-exchange was achieved by purifying the protein in a protonated buffer (2.5 mM sodium phosphate, pH 7.5, 130 mM NaCl).
NMR sample preparation. 0.3 mM HpDnaB in buffer A was mixed with 5 mM MgCl 2 • 6H 2 O and consecutively 6 mM of an NH 4 AlF 4 solution (prepared by incubating 1 M AlCl 3 solution with a five-fold excess of 1 M NH 4 F solution (compared to AlCl 3 ) for 5 min) and 5 mM ADP and incubated for 2 h at 4°C. Under these conditions a full occupation of binding sites has been observed 57 .1 mM of (dT) 20 (purchased from Microsynth) was added to the complexes and reacted for 30 min at room temperature. The protein solution was sedimented 53,54,92 into the MAS-NMR rotor (16 h at 4°C at 210,000 × g) using home-built tools 93 . In case of the DnaB:ADP:AlF 4 − complex the DNA addition step was omitted.
EPR sample preparation. For EPR experiments, natural abundance DnaB was concentrated to 48 mg/ml (850 μM) using a Vivaspin 500 centrifugal filter with a cut-off of 30 kDa. The concentrated protein was incubated in presence of 6 mM ADP, 170 μM Mn 2+ and 7 mM NH 4 AlF 4 for 2 h at 4°C. After 2 h, glycerol was added to a concentration of 20%. The final concentrations were: DnaB 690 μM, ADP 5 mM, Mn 2+ 138 μM and NH 4 AlF 4 6 mM. An identical protocol was used for experiments performed on 13 C/ 15 N labeled DnaB.
Solid-state NMR experiments. Solid-state NMR spectra were acquired at 11.7, 14.1 and 20.0 T static magnetic-field strengths using an in-house modified Bruker 3.2 mm ( 19 F and 27 Al NMR) probe and a 0.7 mm ( 1 H NMR) triple-resonance ( 1 H/ 31 P/ 13 C) probe. The MAS frequencies were set to 17 and 100/105 kHz, respectively. The 2D spectra were processed with the software TOPSPIN (version 3.5, Bruker Biospin) with a shifted (2.0 or 3.0) squared cosine apodization function and automated baseline correction in the indirect and direct dimensions. For 1 H-detected experiments, the sample temperature was set to 293 K 93 and varied in the range of 294-302 K for the temperature-dependence studies. A fast adjustment of the temperature in the bore of the magnet (typically causing B 0 instabilities) was achieved by a bore heating system implemented by the instrument manufacturer. This is crucial for detecting the rather small temperature dependences of proton chemical-shift values (on the order of several ppb/K) 62 . For 19 F (recorded at 14.1 T) and 27 Al (recorded at 11.7 T) MAS-NMR experiments, the sample temperature was adjusted to 278 K. 1 H and 31 P-detected spectra were analysed with the software CcpNmr (version 2.4.2) 94-96 and referenced to 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS). 19 F and 27 Al spectra were referenced to internal standards. For more detail see the Source Data file. All samples were measured repeatedly. The samples were at least prepared twice and yield identical NMR spectra. The temperature chemical-shift gradients were analysed with MATLAB, version 9.6.0 (R2019a).
EPR experiments. All experiments were conducted on a Bruker Elexsys E680 EPR spectrometer (Bruker Biospin) operating at W-band frequencies (~94.2 GHz). ENDOR measurements used a 250 W radiofrequency (rf) amplifier. The temperature was generally set to 10 K.
Electron-electron double resonance (ELDOR)-detected NMR spectra were acquired with a shot repetition time of 1 ms and the echo-detected hole-burning sequence t HTA -T-t p -τ -2t p -τ-echo, with t HTA = 50 µs, T = 10 µs, t p = 100 ns, τ = 1400 ns and an integration window of 1400 ns. The frequency of the high-turning angle (HTA) pulse was incremented in steps of 0.1 MHz over the measured range. A+/− phase cycle on the first π/2 pulse of the echo was used to eliminate unwanted coherence transfer pathways. The power of the HTA pulse, generated by the ELDOR channel of the spectrometer, was optimized such that the observed lines were as intense as possible without being broadened by saturation effects. The nutation frequency ν 1 at the centre of the resonator was ca. 6 MHz or ca. 12 MHz, denoted as low and high HTA pulse power, respectively. The settings were held constant between protein samples and the corresponding control samples. Yet it is important to note that exact reproducibility of peak intensities between runs may be difficult with the resonator used because the resonator profile strongly affects line intensities in EDNMR, and hence a careful experimental setup is required.
Davies ENDOR spectra were acquired with a shot repetition time of 5 ms and with the sequence t inv -T-t p -τ -2t p -τ-echo, where during the time T, an rf pulse was applied. The inversion pulse length was set to 200 ns, and the rf pulse length to 50 µs. The echo was integrated symmetrically around the echo maximum over a time of 400 ns. Due to enormous time overhead on this particular spectrometer, we did not use stochastic acquisition mode and used 10 shots per point.
Mims ENDOR spectra were acquired with a shot repetition time of 2.5 ms and with the sequence t p -τ-t p -T-t p -τ-echo, where during the time T, an rf pulse was applied. The interpulse delay τ was set to 1200 ns, corresponding to the phase memory time T m , where detection of small hyperfine couplings is most sensitive 97 , and the rf pulse length to 25 µs.
Raw EDNMR data were background corrected with a Lorentzian line that was fitted to the central hole, and normalized to the signal intensity far off-resonance, i.e. the peak intensity corresponds to the relative hole depth. Fitting of the EPR spectra was performed with EasySpin (version 6.0.0).
DFT calculations of hyperfine tensors. DFT calculations were performed on small clusters mimicking the coordination sphere of the metal ion co-factor extracted from the PDB structures (BstDnaB: accession code 4ESV and AaDnaB: accession code 4NMN, see Supplementary Fig. 2). Hydrogen atoms were added to saturate terminating groups and their positions were optimized on a TPSS 98 /def2-SVP 99 level using TURBOMOLE (version 6.0) 100,101 . In all TURBOMOLE SCF calculations, an energy convergence criterion of 10 −7 E h and in all geometry optimizations an energy convergence criterion of 5 × 10 −7 E h was chosen. The integration grid was set to m4 and the RI approximation was used. Hyperfine coupling tensors were calculated in the ADF suite (version 2013) 102 on a B3LYP 103,104 /TZ2P 105 level of theory. The INTEGRATION keyword was set to 6.0 and in the SCF calculation an energy convergence criterion of 10 −6 E h was used.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The NMR and EPR spectra can be accessed at https://doi.org/10.3929/ethz-b-000501034. The following PDB structures were used in this study: 4ZC0, 4NMN, 4ESV, 3CMW and 3PUW. All experimental NMR parameters are provided as a Source Data file. Protein resonance assignments are available from the BMRB database (www.bmrb.wisc.edu, accession code 27879). Source data are provided with this paper.