*4.1. The Pre-Hydrolytic State Mimicked by AMPPCP, AMPPNP and ATP-*γ*-S*

In order to characterize the pre-hydrolytic state, we first investigated DnaB and BmrA in the presence of AMPPNP, AMPPCP, and ATP-γ-S. It however appeared that ATP-γ-S was completely hydrolysed during the rotor filling by BmrA (one hour of filling) and DnaB (overnight filling), as monitored by <sup>31</sup>P solid-state NMR experiments (see Figure S1), and was thus of no further use. We therefore focused on AMPPNP and AMPPCP. Since a major function of the DnaB helicase is to bind to DNA, protein samples were also prepared with the ATP analogue and single-stranded DNA (here a DNA-fragment of 20 thymidine nucleotides abbreviated as (dT)20). The presence of three signals in the 1D CP <sup>31</sup>P NMR spectrum (Figure 4A, left panel) indicates binding of the triphosphate AMPPCP to DnaB. However, the resonances of the phosphorus α and β are rather broad. This broadening might indicate inhomogeneities in the binding site in the environment of the ligand, or chemical-exchange broadening effects. In contrast, in the presence of DNA and AMPPCP, the <sup>31</sup>P resonances in the 1D CP

<sup>31</sup>P spectrum are very sharp (Figure 4B, left panel). This indicates that the DnaB:DNA complex fixes AMPPCP with high homogeneity.

**Figure 4.** Pre-hydrolytic states using the systems DnaB and BmrA. <sup>31</sup>P one dimensional spectrum cross-polarization (1D CP) spectrum (left panel) and the alanine region of <sup>13</sup>C-13C-DARR spectra overlay (right panel) of DnaB (blue) and DnaB:AMPPCP (red) (**A**), DnaB:DNA (blue) and DnaB:DNA:AMPPCP (red) (**B**), DnaB (blue) and DnaB:AMPPNP (red) (**C**), DnaB:DNA (blue) and DnaB:DNA:AMPPNP (red) (**D**). <sup>31</sup>P 1D CP spectrum (left panel), <sup>31</sup>P 1D spectrum (middle panel) spectrum and overlay of the alanine region of <sup>13</sup>C-13C-DARR spectra (right panel) of BmrA:AMPPCP (red) and BmrA (blue) (**E**). Red stars indicate hydrolysis products of AMPPCP. <sup>31</sup>P 1D CP spectrum of BmrA:AMPPNP (**F**). Spectra in (**A**) and (**B**) were adapted from Wiegand et al. 2019 [82] (http://creativecommons.org/licenses/by/4.0/.), spectra (**C**) and (**D**) were adapted with permission from Wiegand et al. 2016 [38]. Panels (**E**) and (**F**) on BmrA represent original data; Red stars (\*) indicate hydrolysis products of AMPPCP.

The 2D <sup>13</sup>C-13C DARR experiments recorded on DnaB:AMPPCP show not only chemical-shift perturbations when compared to the apo protein, but also dynamic changes, as can be seen in the extract of the alanine region (Figure 4A, right panel) by the disappearance of resonances, which could be assigned to the *N*-terminal domain [82], which is important for binding the DnaG primase within the primosome. As illustrated by the equivalent 2D <sup>13</sup>C-13C DARR experiment on the DNA-bound DnaB (Figure 4B, right panel), the binding of AMPPCP induces stronger CSPs due to larger conformational changes of the protein, but no dynamic effects of the *N*-terminal domain were observed.

In principle, AMPPNP and AMPPCP should have a similar effect on DnaB, as both should induce the pre-hydrolytic state. However, it is clear from the NMR spectra that the effects of these two analogues are very different. First, as highlighted by the 1D CP <sup>31</sup>P spectrum (Figure 4C, left panel), the presence of multiple resonances from the phosphate groups of AMPPNP indicates several structurally slightly different bound AMPPNP molecules. Interestingly, the 2D <sup>13</sup>C-13C DARR spectrum reveals that the

disappearance of the *N*-terminal domain resonances upon binding of AMPPCP is not observed in case of AMPPNP (Figure 4C, right panel). Also, we had observed that in presence of DNA, all AMPPNP is hydrolysed by the helicase [38]. Consequently, as shown by Figure 4D right panel, no AMPPNP is bound to the protein when DNA binds to the helicase, and the 2D <sup>13</sup>C-13C DARR spectrum of DnaB in the presence of AMPPNP looks highly similar to DnaB without the analogue, which is not detected in the <sup>31</sup>P spectra either (Figure 4D left panel).

BmrA also binds AMPPCP, as shown by the 1D CP <sup>31</sup>P spectrum (Figure 4E, left panel). However, the rate of AMPPCP hydrolysis is much higher, compared to DnaB, and degradation products of AMPPCP can be observed already four hours after the rotor filling in the supernatant of the NMR rotor, as shown in Figure 4E right panel (see red stars in the Figure). We recorded a 2D <sup>13</sup>C-13C DARR experiment of BmrA:AMPPCP (two days of acquisition), and the spectrum is virtually the same as the one of BmrA in the apo state. Possibly, AMPPCP has been rapidly hydrolysed, and an insufficient amount of AMPPCP only remained bound on BmrA. Similar to AMPPCP, AMPPNP binds to BmrA (Figure 4F), but was also rapidly hydrolysed (data not shown). The analysis of <sup>31</sup>P NMR spectra for protein samples containing lipids or DNA is more difficult due to the overlap between the <sup>31</sup>P γ- and β-phosphate signals from AMPPNP and those from lipid/DNA.

To overcome the hydrolysis problem with BmrA and to obtain a snapshot of the protein in its pre-hydrolytic state, we used an alternative approach, which is based on using mutant forms of the protein, which do bind ATP, but do not hydrolyse it. For this, catalytic residue/s can be mutated in order to make the protein inactive; still, one must take care that the protein retains its native fold. For BmrA, and also for other ABC transporters, it was shown that the mutation of the catalytic glutamate (E504 in BmrA) does not significantly affect the conformational change occurring upon nucleotide binding [23,99]. In contrast, the protein cannot achieve the pre-hydrolytic conformation when the nucleotide-binding Lys residue of the Walker A motif is mutated, here K380A [23,99]. We incubated the mutant E504A with ATP, and then sedimented it for analysis in the solid-state NMR rotor. While E504A is not completely inactive, it displays a very low ATPase activity (but still even crystals were obtained recently, PDB accession code 6R72, and a cryo-EM based structure was reported, PDB accession code 6R81) when compared to K380A, used as a fully inactive control (Figure 5A). After 40 h, only 50% of ATP is consumed, which allowed for the acquisition of 1D and 2D solid-state NMR experiments. The resulting 1D <sup>31</sup>P CP spectrum displays three narrow peaks corresponding to the three phosphate groups from the ATP bound to the protein (Figure 5B). The 2D <sup>13</sup>C-13C DARR spectrum displays CSPs and peaks appearing, both induced by the conformational and dynamic changes in the protein as a consequence of ATP binding (Figure 5C) [92]. γ β

Δ ○ **Figure 5.** Pre-hydrolytic states using the system BmrA-E504A (catalytically inhibited). Percentage of ADP (∆) and ATP (#) in the presence of the mutant BmrA-E504A (symbol filled in blue) or of the mutant, which does not bind the nucleotide, BmrA-K380A (symbol filled in green). BmrA-K380A was chosen as a negative control in order to exclude the possibility of an ATPase contaminant in the sample (**A**). <sup>31</sup>P 1D CP spectrum of BmrA-E504A:ATP (**B**). The overly of the alanine region of <sup>13</sup>C-13C-DARR spectra of BmrA-E504A (blue) and BmrA-E504A:ATP (red) (**C**). Results in panels (**A**) and (**B**) are original, and spectra in (**C**) were adapted from Lacabanne et al. 2019 [92] (http://creativecommons.org/licenses/by/4.0/.).

To summarize, our data show that analysis of the pre-hydrolytic state is difficult both for DnaB and BmrA, since first the corresponding ATP mimics do not behave in a homogenous manner, i.e., analogues which should yield similar states lead to different NMR spectra, and second, most popular analogues are actually hydrolysed by the helicase in presence of DNA, as well as by the ABC transporter.

With respect to the first point, the intriguing observation that the AMPPCP- and AMPPNP-induced pre-hydrolytic states show conformational differences might be linked to the proposition that one can further differentiate each pre-hydrolytic mimic, as discussed by Ogawa et al., and assign the different mimics to specific steps therein: ATP-γ-S for the initial pre-hydrolysis state, AMPPCP for the pre-isomerization state, ADP:BeF<sup>x</sup> for the middle pre-hydrolysis state and AMPPNP for the late pre-hydrolysis state [107]. It is difficult to establish a similar statement for DnaB, as one can also explain these differences by the fact that these analogues can behave differently from ATP in terms of their chemical properties: as examples for AMPPNP the oxygen, a hydrogen bond acceptor, is replaced by an NH<sup>2</sup> group, a possible hydrogen bond donor; AMPPCP has one oxygen atom less than ATP.

With respect to the second point, in the DnaB-DNA complex, only AMPPCP resisted to hydrolysis, and was the best choice to study DnaB and its DNA complex. It was however, rapidly hydrolysed in BmrA, which might be caused by the very high ATPase activity of BmrA, which is with an activity of 6.5 µmol·min−<sup>1</sup> ·mg−<sup>1</sup> one of the most active ABC transporters (one to three orders of magnitude higher than typical ABC transporters) [108]. Amongst AMPPNP and ATP-γ-S, which are the most used pre-hydrolytic state analogues for ABC transporters and helicases (see Figure 3B), neither proved useful here. Alternative strategies using mutant forms were successful to analyse a pre-hydrolytic mimic of the protein and presents a valuable alternative when ATP analogues fail to mimic the pre-hydrolysis states.
