Role of ATP for the in vitro assembly process of bacterial F-type ATP synthases. Since the assembly mechanism of the bacterial F-type ATP synthase, in particular for the soluble F1 module, is mostly still unclear, we aim to shed light on this process. We determine individual subunit interactions, relevant conditions and also investigate whether the assembly of all different subunits follows a specific order.
In a first step, we purified recombinant subunits (α, β, δ and γε) of the soluble part F1 individually. Attempts to express the γ-subunit in E. coli separately yielded insoluble aggregates (inclusion bodies). Therefore, we cloned the genes atpG and atpC bicistronically in the expression vector and purified the central stalk γε to generate a higher complex stability and protein solubility (Fig. S7). To investigate individual assembly steps we incubated different subunit combinations under different assembly conditions and determined with LILBID-MS and HPLC experiments which complexes did or did not form. The resulting (sub)complexes represent key steps in the assembly process. To ensure the relevance of the observed in vitro subcomplexes for the in vivo processes in the cell and for comparison of structure, complex stability and bioactivity, we then set out to produce homologous subcomplexes in vivo.
Incubation of recombinant α and β in vitro in ammonium acetate buffer without any additives shows no specific α/β-oligomerization into higher subcomplexes (Fig. 1a) suggesting that our experiment is missing something essential. The cytoplasmic concentrations of ATP in active cells is known to be approximately in the range of 2-5 mM[39] which prompted us to test the relevance of ATP for this assembly. As MS resolution can be affected by additives, we stayed in the lower range of the native concentration and incubated the same proteins with 2 mM ATP and MgCl2 (Fig 1). (Effect of ATP/Mg2+ on LILBID-MS spectra is shown in Supplemental Information Figure S1). Fig. 1b demonstrates that the addition of ATP/Mg2+ leads to specific αβ heterodimer formation only (no unspecific homodimers) with a charge state distribution (-1 to -3). Interestingly, no higher α/β subspecies can be identified, even though ATPases are known to form a hexameric head, containing three α- and β-subunits, respectively. As in vivo comparison, we purified and isolated an αβ complex from E. coli, which we then analyzed under the same in vitro conditions. The in vivo αβ complex showed the same heterodimer and no higher oligomeric states either (Fig. S2). It should be noted that the in vitro formed αβ heterodimers need the presence of ATP/Mg2+ not only for formation but in order to remain stable, as can be seen from measurements, for which the buffer is slowly diluted during the LILBID-MS run (Fig. S1). Reduction of the ATP/Mg2+ concentration during LILBID-MS goes along with an increase in the signal resolution but reduced complex stability. Interestingly the in vivo αβ complex seems to suffer less from reduced ATP/Mg2+ concentration (Fig. S2), which is the first hint that assembly of α and β in vitro and in vivo might lead to slightly different heterodimers.
For comparison with the LILBID-MS results we investigate the in vitro assembly of subunit α and β by an alternative analytical method: size exclusion chromatography (SEC) by HPLC, using the SEC-column with a protein separation range of 10k-700kDa (Fig. 1C). The chromatograms show a shift into a shorter elution peak time confirming αβ heterodimer formation only upon addition of mM ATP and MgCl2. LILBID and SEC experiments both show (Fig. 1a-c) that the additives ATP/Mg2+ are crucial for the in vitro assembly of subunits α and β of Na+-F1FO- ATP synthases of A. woodii.
Assembly studies of single isolated subunits with next neighbors
α, β, δ and γε Figure 2 gives an overview of all possible subcomplexes which we observed after incubation of different combinations of subunits. This provides information about next neighbor interactions and order of assembly steps within the F1- complex. As mentioned above we can observe αβ heterodimer assembly under the right assembly conditions – but no higher order oligomers, if just these two subunits are present (Fig. 1b). Similarly, we observe that upon incubating either subunit α or β with the γε complex, a single subunit α or β can bind to the γε complex, forming αγε or βγε complexes, respectively. These complexes are seen under soft laser conditions. Already medium laser desorption conditions, allow the ε-subunit to dissociate from the γε complex (Fig. S8c-d), giving rise to αγ or βγ complexes. This shows that α or β bind to the g-subunit, while the ε-subunit is only weakly bound to g and not directly involved in the interaction with α or β.
As subunit α and β alone can only form heterodimers and only a single α or β can bind to γ if incubated alone, the next step was to purify and incubate recombinant α, β and γε in presence of 2 mM ATP / MgCl2. Mass spectrometric analysis showed the formation of the α3β3γε-subcomplex which we could identify at a mass of 378.72 kDa (theoretical protein mass 378.24 kDa) and a charge state distribution from -1 to -5 (Fig. 3). In contrast, no higher subcomplex formation was observed if the same experiment was performed without ATP/Mg2+. This implies that these additives are vital for the entire in vitro reconstitution process providing complex stability.
Including the δ subunit to the incubation then led to its association with the preformed α3β3γε complex resulting in the fully assembled F1-complex α3β3δγε with a determined protein mass of 399.84 kDa (theoretical mass: 399.78 kDa) and a charge state distribution of -1 to -3 (Fig. 3d). The spectrum shows the fully assembled F1-complex, as well as subcomplexes α2β2γε, αβγε and αβγ, which could stem from incomplete assembly or dissociation. Surprisingly our spectra show no subcomplexes with an odd number of α or β subunits, indicating that the αβ heterodimer is the stable basic binding block from which the hexamer is build up. No δ is seen binding to any of the single subunits, which is in line with studies showing interaction of δ with α-subunits only if complexed with other F1-subunits[40]. No δ binding is seen for any of the mentioned subcomplexes either (Fig. 2a and Fig. S8a-b), suggesting the fully assembled αβ-hexamer to be a prerequisite for binding of δ. Combined the observed on-pathway subcomplexes en-route to the F1 complexes allow to propose the ATPase assembly to occur only via the binding of preformed αβ heterodimers onto the γ-subunit, to form the hexameric head, which then allows for binding of δ. As γ could not be purified separately we could not determine in the same manner, if ε is a prerequisite for binding of the αβ dimers onto γ.
For comparison we then attempted to purify and isolate all proposed on-pathway subcomplexes in vivo. All expected complexes (Fig. 2c) could be purified. As complexes lacking ε could be achieved, we conclude that the hexameric head forms around γ independently of ε. Mass spectrometric analysis (Fig. S6) reveals, as for the in vitro complexes, only complexes with the same amount of α and β subunits, supporting the αβ heterodimer as a basic building block. Incubation of δ with the in vivo complex as well confirms our in vitro results, which showed δ to only bind onto the complete αβ hexamer.
Interestingly, the in vivo α3β3γε complex (Fig. S6) shows a significantly higher complex stability even without addition of ATP/Mg2+ than the in vitro complex (Fig S2e-f). This difference could be very interesting, as it might show that there are deviations in the in vivo and in vitro assembly, possibly due to structural differences of the proteins. An alternative explanation would be steric hindrances due to the tags, which are part of our in vitro purified proteins. The in vitro studies were performed on N-terminal StrepI-α and His6-β, C-terminal His6-δ and γε (His6-tag located on subunit ε), while the in vivo complex was expressed with only one N-terminal His6-tag on the β-subunit. For comparison we purified a complex, which includes the same tags on every subunit as our isolated purified subunits, which we name α3β3γε* for distinction. The mass spectrum in figure S5a shows that the tags are not responsible for destabilization of the complex.
Remarkably the in vivo complexes are not only stable without ATP/Mg2+ addition which is crucial for the in vitro complexes, but even the last assembly step -association of δ to in vivo purified α3β3γε and α3β3γε* pre-complexes forming the full F1-complex (Fig. S5) is ATP/Mg2+ independent.
The role of ATP hydrolysis vs. nucleotide binding for the in vitro assembly process of bacterial F1-ATPase
The important role of ATP hydrolysis for the function of ATP synthases is unquestioned and we have as well observed the significance of ATP for complex assembly. Nevertheless, it is not yet clear, whether ATP hydrolysis or only nucleotide binding is required for the F1 assembly process. Therefore, we repeat the described assembly experiments with α, β and γε, but replace ATP by non-hydrolysable ATP analogues, such as ADP, adenylyl-imidodiphosphate (AMP-PNP) and adenosine 5’-[γ-thio]triphosphate (ATP-γ-S) with a concentration of ([ATP/ATP-analogues]/[Mg2+] = 2 mM).
The obtained MS-spectra with non-hydrolysable ATP analogues (Fig. 4b-d) show the successful formation of larger F1-subcomplexes. Interestingly all ATP analogs allow formation of α3β3γε apart from AMP-PNP, which generates only α2β2γε subcomplexes and not the expected α3β3γε complex; which could be due to differences in the chemical geometry of AMP-PNP. The free lone pair of nitrogen might sterically inhibit the assembly of the last αβ heterodimer to the complete α3β3 hexamer. Generally, this confirms that ATP hydrolysis is not required for assembly, but nucleotide binding is the essential factor triggering the in vitro complex formation.
Importance of charged residues in the catalytic sites of bacterial F-type ATP synthases.
After determining the crucial effect of ATP binding for ATPase assembly, we want to shed more light on this process. Recent studies have shown the importance of charged residues for the ability of Pi – binding at the catalytic sites of the αβ interface of the hexameric head. The replacement of charged amino acids with neutral residues induced reduced cell growth and loss of ATPase activity, suggesting the electrostatic interactions between amino acids to be crucial for initial phosphate binding in the catalytic sites[39]. Specifically four residues β(K155), β(R182), β(R246) and α(R376) were shown to be critical[41–43]. We want to deduce if these mutations affect ATPase activity via their influence on the ability of the complex to bind/hydrolyze ATP or if the effect might be even more drastic and influence the complexes assembly. Therefore, we introduced homologous point mutations in conserved regions in α (R363Q, R363K) and β (K159Q, R186Q, R251Q) in the A. woodii operon (Fig. 5).
To investigate the effect of these charged residues in the catalytic site we first analyze the binding efficiency of ATP to α, β and different mutants with nano electrospray-ionization (nESI) (Fig. 6), which allows to observe the number of bound ATP. The deconvoluted MS-spectrum in Fig. 6a shows three peaks corresponding to β[WT] with zero, one or two bound ATP’s. The dominating species is clearly the one with one bound ATP, indicating specific binding. In contrast peaks trailing of for the mutant β[K159Q] (Fig. 6b) suggest non-specific ATP binding only. Figure 6c shows the differences in ATP binding for the investigated α and β subunits and their respective mutants. The β[K159Q] mutant demonstrates a decrease of ATP binding around 60%, if compared to β[WT]. A less pronounced effect is seen for β[R251Q]. No significant differences in ATP binding were observed for α[WT] and its mutants which have a generally lower ATP-binding efficiency compared to β[WT].
Seeing that charged residues in the catalytic site can influence ATP-binding, we wanted to assess, if reduced ATP-binding affinity had an effect on the in vitro assembly process. Therefore, we performed the same HPLC and MS-experiments as for the wildtype α-β incubations. Figure 6d-h show with HPLC the formation of αβ heterodimers at an elution peak time of 10.8 min for α[WT] with each β construct. Similarly, the in vitro assembly of β[WT] with α[R363Q] (Fig. 6i) or α[R363K] (Fig. S3a) show heterodimer formation. Interestingly, the HPLC runs for α[WT] with β[K159Q] or β[R251Q] (Fig. 6e/g) the mutants, which reduced ATP binding, show additional peaks at around 11.7 min suggesting the presence of unassembled monomers and therewith less efficient binding.
In vitro assembly of the α and β mutant constructs was then probed by LILBID-MS. Interestingly, we observe that under the same experimental conditions, which show stable heterodimers of wildtype α and β (Fig. 1b), we do not detect stable αβ heterodimers for α[WT] incubated with β[K159Q] (Fig. 7a) or any of the b mutants (Fig. S4). This suggests that all b mutations affect stability to the extent that dimers can still be observed with HPLC (to a different degree) but are much reduced in the LILBID spectra. Similarly, we see clearly decreased dimerization for β[WT] with α[R363Q] (Fig 7c). A point mutation at the same position which retains the charge of the original amino acid (α[R363K]) shows no effect, as it does not hinder the formation of the αβ heterodimer (Fig. S3).
As our experiments so far indicate that the formation of the αβ heterodimer is decisive for further complex formation, the mutations, which impede the heterodimer, should affect the formation of higher order F1-complexes as well. Our mass spectra confirm that incubation of α[WT], β[K159Q] and the central stalk γε does not lead to the α3β3γε complex (Fig. 7b), which we observed with the wildtype β (Fig. 3c). The same result is obtained with all other β mutants, (β[R186Q], β[R251Q] and β[K159Q, R186Q, R251Q], highlighting the important role of the charged residues in subunit β not only for the catalytic activity but as well for the F1 assembly (Fig. S4). In contrast assembly of the mutant α[R363Q], β[WT] and the central stalk γε generates a larger complex – but interestingly not the expected hexameric head but subcomplex α2β2γε (Fig. 7d). This indicates that replacing the arginine with the neutrally charged glutamine in subunit α has an effect on the assembly process, which is less pronounced than the effect of the above-mentioned substitutions in subunit β, but still hinders the formation of the correct αβ hexamer.
Order of assembly steps into F1
The combined results obtained with LILBID and nESI-MS and HPLC allow us to establish all subcomplexes which form in vitro en route to the full complex as well as the necessary assembly conditions. As these assemblies take place in minimal environment, compared to a cell, we need to guarantee that the observed subcomplexes are not just the result of aggregation, but can occur in the same manner in the cell. Therefore we expressed the subunit combinations, which we found in our assembly experiments directly in E. coli cells for comparison. We were able to obtain all anticipated subcomplexes with the same stoichiometries. Based on these findings we can follow the individual assembly steps and establish their order for formation of bacterial solube F1 domain of F-type ATP synthases (Fig. 8).
Overall our results show that, α- and β-subunits assemble specifically to αβ heterodimers only in presence of nucleotides and Mg2+ (Fig. 1, Fig. 3 and Fig. 4) while surprisingly not forming higher subcomplexes (e.g. αβ2, α2β, α2β2 etc.) least of all the expected hexameric head (α3β3). The a and β subunits can both bind individually to the central stalk γ, but only little binding is observed and in a maximal copy number of one (Fig. S8c-d), unless a and β dimerize first. Three pairs of αβ heterodimers bind either to the central stalk γ or a preformed γε complex to generate stable α3β3γ or α3β3γε complexes, respectively. The relevance of these in vitro results is shown in our heterologously expressed subcomplexes α3β3γ and α3β3γε which we can isolate as stable and bioactive complexes in vivo (Fig. S6). The δ-subunit binds to the in vivo or in vitro pre-complexes containing the hexameric head (Fig. 3d and Fig. S5), which generates the full F1 complex.
As isolated subunit γ could not be purified in E. coli for stability reasons, we purified γε (Fig. S7), so our assembly studies did not include binding of ε and γ analogously to all the other binding experiments. The complete aggregation of the γ-subunit into inclusion bodies in the bacterial matrix during the overexpression, could suggest that the ε-subunit is essential for the stabilization of the central stalk (Fig. S7), which would place binding of ε to γ as a first step parallel to the formation of the αβ heterodimer, before further assembly steps occur. Nevertheless, expression of in vivo subcomplexes showed that α3β3γ can be expressed as a stable subcomplex, which indicates that binding of ε is not an essential prerequisite for any further assembly to the F1 complex. ε could therefore enter the assembly network at different times, as we indicate in Fig 8.
Interestingly, we observe some differences between our in vivo und in vitro complexes: the in vivo pre-formed complexes are stable without nucleotide- and Mg2+ and the binding of δ to any complexes including the hexameric head can then be performed in vitro without nucleotide- and Mg2+ addition. The same assembly experiment with the in vitro formed α3β3γε complex requires nucleotide- and Mg2+ addition, to hinder disaggregation of the complexes. This could be explained, if the F1-subcomplexes formed in vivo and in vitro shared the same assembly pathway and stoichiometry, but a structural change occurring in the cell to stabilize the formed complex were missing in vitro, which is a question we will come back to later. Fig. 8 depicts a full summary of the different assembly steps of the soluble part F1 as determined by our experimental data. The entire assembly process follows a choreographed pathway, which allows for an efficient construction of the F1 complex.
ATPase activity of in vitro assembled A. woodii αβ and α3β3γε complex
In general, many macromolecular complexes need molecular chaperones, which assist for example in the conformational folding or assembly processes. In mitochondria, it is known that the F1- assembly of the hexamer of alternating α- and β-subunits requires two specific chaperones, Atp11p and Atp12p, that bind transiently to β and α[31,44]. Our previous results demonstrate that the bacterial F1- assembly can take place without the presence of chaperones. This was a surprising finding, as a generally accepted hypothesis expects an Atp12p homolog, which has been identified only in proteobacteria[45] so far, to play an essential role in the bacterial F1- assembly. The question arises, if chaperones are not needed at all, or if their presence might be required, if not for assembly, then to guarantee correct function and activity.
For this reason, we investigated our in vitro assembled complexes for bioactivity. We performed ATP hydrolysis experiments with all of the separately expressed constructs (α,β, γε and ε) and the in vitro assembled αβ and α3β3γε complexes using an enzyme-coupled activity assay[37,38]. For in vitro complex formation the subunits were incubated with 4 mM MgCl2 and ATP with subunit stoichiometries of 1:1 for the αβ heterodimer, and 3:3:1 for α3β3γε. All subunits and subcomplexes showed moderate ATPase activity (Fig. 9a).
We could determine ATPase activities at 4 mM ATP for the β- and α-subunit of 0.15 ± 0.05·10-6 mol ATP / (mol protein · s) and 0.20 ± 0.14·10-6 mol ATP / (mol protein · s), respectively. These are quite similar and do not reflect the threefold higher ATPase activity of α previously seen in E.coli [30].
As isolated subunit ε was already known to bind ATP[46] and has to undergo large conformational changes to regulate the ATPase[47] we hypothesized that ε as well undergoes hydrolysis. Interestingly, the isolated ε-subunit showed a hydrolytic activity at 3.4 ± 0.1·10-6 mol ATP / (mol protein · s), which is one magnitude higher than that of the single catalytic β-subunit. Enzymatic hydrolytic activity of γε was measured at 3.1 ± 1.4·10-6 mol ATP / (mol protein · s) placing the overall activity of γε very similar to ε, suggesting no contribution from γ.
The ATPase activity calculated for the in vitro assembled αβ heterodimer is 0.69·± 0.22·10-6 mol ATP / (mol protein · s), assuming incubation leads to full complex formation. As this might not be the case, all molar activities of in vitro complexes have to be seen as a conservative estimate – if full complexation could be guaranteed the observed values would likely be higher. Nevertheless, the observed hydrolytic activity is about twice the sum of the individual α and β activities, supporting the relevance of the αβ heterodimer as a building block and working unit of the ATPase. Surprisingly this value stays clearly below the hydrolytic activity of ε.
The reconstitution based on the incubation of α, β and γε resulted in a hydrolytic enzymatic activity of 6.1±0.5·10-6 mol ATP / (mol protein · s). Interestingly, the activity of this in vitro assembly is not much higher than the sum of the ATPase activities of the utilized subunits. The increased activity (factor 1.5) of the of the now complexed subunits stems mainly from the increase of activity seen for the formation of the three required αβ heterodimers.
Similar measurements were performed with in vivo purified αβ and α3β3γε*. The ATPase activity for both is noticeably beyond anything we observed for in vitro complexes. (160 ± 14·10-6 and 900 ± 7·10-6 mol ATP / (mol protein · s) respectively) (Fig. 9a and 9b) These values cannot be compared directly with those obtained for in vitro complexes, for which the individual subunits don’t fully form into complexes. Nevertheless, the activity of the in vivo αβ heterodimer alone is already much higher than that of the in vitro assembled α3β3γε complex. Similarly, the in vivo α3β3γε* shows an activity way beyond the sum of its components. This indicates that the in vitro conditions allow for the correct assembly of the functional subunits, which then retain their individual ATPase activity, while the cellular environment allows for modifications, which increase the ATPase activity of the αβ heterodimer and the whole ATPase, way beyond that of the sum of its parts.
Stability studies with ion mobility MS: in vivo vs in vitro heterodimers
The observed differences in bioactivities as well as stability without ATP/Mg+ for our in vivo and in vitro complexes suggest that additional effects besides assembly are required to form an active complex. We assume conformational changes to play a role, possibly triggered by chaperones.
The relevance of the αβ heterodimer formed as a first step in the ATPase assembly and the differences observed in our in vivo and in vitro studies, makes the heterodimer a very interesting candidate to investigate, if the changes in activity can be correlated to structural changes. Noticeable structural rearrangements should be accompanied by a change in collision cross section (CCS), which can be monitored by ion mobility (IM) MS, while changes to the intrinsic stability can be revealed by differences in collision induced unfolding (CIU) experiments. To validate our assumption, we investigate the heterodimer for differences between in vivo and in vitro formed complex. We select in both cases the 20-times charged αβ heterodimer in our nESI mass spectra as precursor ion which is analyzed via ion mobility MS in dependence of the applied collision voltage. IM allows the separation of ions based on their mobility through an inert gas under the influence of an electric field. Increasing the collision voltage can lead to collision induced unfolding and then dissociation of the complex. The energy required to cause CIU can be correlated to the complex stability. In the resulting IM fingerprint of the in vivo assembled αβ heterodimer, the IM signal appears for low collision voltages at drift times about 17.5 ms (Fig. 10), which corresponds to the compact feature of the folded complex. Increase of collision voltage to around 175 V leads to protein unfolding, which is accompanied by an increase in CCS. The original IM signal decreases partially in favor of a signal at 19 ms, representing the corresponding unfolded structure of the αβ heterodimer. In contrast the in vitro assembled αβ heterodimer already unfolds at collision voltages about 155 V and appears at a slightly higher drift time (about 20 ms) compared to the in vivo assembled αβ. This indicates that the in vitro structure is less stabilized against unfolding than the in vivo assembled αβ. This can be seen more clearly in the difference plot (in vitro minus in vivo, Fig 10c), which helps to compare specific features of each fingerprint directly. Additionally it reveals a significantly broader IM signal for the compact, as well as the unfolded state for the in vitro assembled αβ heterodimer indicating a less homogenous structure for the in vitro than for the vivo αβ heterodimer. A more defined structure and increased stability of the in vivo heterodimer are structural properties, which can be correlated to the increased ATPase activity.