Role of AAA3 Domain in Allosteric Communication of Dynein Motor Proteins.

Cytoplasmic dynein, an AAA+ motif containing motor, generates force and movement along the microtubule to execute important biological functions including intracellular material transport and cell division by hydrolyzing ATP. Among the six AAA+ domains, AAA1 is the primary ATPase site where a single ATP hydrolysis generates a single step. Nucleotide states in AAA3 gate dynein's activity, suggesting that AAA3 acts as a regulatory switch. However, the comprehensive structural perspective of AAA3 in dynein's mechanochemical cycle remains unclear. Here, we explored the allosteric transition path of dynein involving AAA3 using a coarse-grained structure-based model. ATP binding to the AAA1 domain creates a cascade of conformational changes through the other domains of the ring, which leads to the pre-power stroke formation. The linker domain, which is the mechanical element of dynein, shifts from a straight to a bent conformation during this process. In our present study, we observe that AAA3 gates the allosteric communication from AAA1 to the microtubule binding domain (MTBD) through AAA4 and AAA5. The MTBD is linked to the AAA+ ring via a coiled-coil stalk and a buttress domain, which are extended from AAA4 and AAA5, respectively. Further analysis also uncovers the role of AAA3 in the linker movement. The free energy calculation shows that the linker prefers the straight conformation when AAA3 remains in the ATP-bound condition. As AAA3 restricts the motion of AAA4 and AAA5, the linker/AAA5 interactions get stabilized, and the linker cannot move to the pre-power stroke state that halts the complete structural transition required for the mechanochemical cycle. Therefore, we suggest that AAA3 governs dynein's mechanochemical cycle and motility by controlling the AAA4 and AAA5 domains that further regulate the linker movement and the power stroke formation.

Introduction

Molecular motors are important biological machines that drive many key biological functions such as cell divisions, intracellular material transports, and, more predominantly, the communications between cells by walking on actins or microtubules.[1−3] Among the different families of the cytoskeletal motor, dynein is an ATPase-associated microtubule (MT)-dependent motor protein that has no common ancestor with myosin and kinesin.[4−7] Cytoplasmic dyneins are essential for the transport of different cargoes such as mRNA, organelles, vesicles, and viruses inside living cells.[8−12] It also acts as an anchor for the nuclei, Golgi, and centrosome during mitosis to position them properly.[13−16] However, because of its highly complex structure, a comprehensive characterization of the mechanochemical cycle and motility of cytoplasmic dynein is a challenging task. Recently, researchers from both the theoretical and experimental expertise are trying to explore this fascinating biological machine extensively.[4,5,17−24]

Different crystal structures and electron microscopy (EM) studies reveal that cytoplasmic dynein consists of two heavy chains that are the crucial catalytic and mechanical site for motility.[4,25−29] The core of each chain comprises a hexameric ring with six non-identical AAA+ (AAA1 to AAA6) motile (Figure ). Each AAA+ unit is subdivided into large (AAAL) and small (AAAS) subdomains. The large subdomain contains five standard parallel β sheets with H0-H4 helices, and the small subdomain contains a five-helix bundle.[26] A 10 nm long, extended linker domain, which generates a mechanical force for the displacement, spans the diameter of the hexameric ring and swings between AAA2 and AAA5 depending on the nucleotide state of dynein.[30,31] A 15 nm long coiled-coil stalk domain with a small globular MT binding domain (MTBD) protrudes from AAA4, and a strut or buttress extended from AAA5 makes contact with it.[32] Movement of dynein is associated with ATP hydrolysis in the AAA+ ring, which drives a cascade of conformational changes in the ring.[19,33,34] Due to the unique sequence and structure of each AAA+ domain, it is a daunting task to assign the role of each domain. Among the six AAA domains, AAA1 to AAA4 contains nucleotide binding sites, and AAA5 and AAA6 domains do not contain the walker A and walker B ATP binding and hydrolysis motifs.[4,35] ATP hydrolysis at AAA1 is integrated with the kinetics of dynein stepping. Dutta et al. investigated the mechanochemical pathway for ATP hydrolysis at AAA1 with the linker movement.[36] ATP binding to the AAA1 domain causes the formation of the pre-power stroke state, which leads to the AAA1L/2L interface closing, the MTBD detachment from the MT, and the linker movement to the bent conformation. ATP hydrolysis and phosphate release engender partial opening of the AAA1L/2L interface along with the attachment of the MTBD with the MT and the linker shifting to its straight conformations. The role of other ATP binding sites in the mechanochemical cycle is not clearly known yet. Mutations that revoke ATP binding or hydrolysis at AAA2 and AAA4 domains have a negligible effect on the velocity of dynein; however, they affect processivity to some extent.[22] On the other hand, ATPase mutations at AAA3 diminish dynein’s velocity and MT-dependent activity by an order of magnitude.[4,22,33] Although single ATP binding at AAA1 is sufficient for dynein’s stepping along MTs, the conformational changes within the AAA-ring domain that underlay the drastic impediment of dynein’s motility due to blocking of ATP hydrolysis at AAA3 are still unknown. Bhabha et al. reported X-ray and EM structures of yeast dynein with nonhydrolyzable ATP analogue AMP-PNP.[37] They suggest that AAA3 acts as a switch by regulating the transmission of different conformational changes between AAA1 and the linker. Both DeWitt et al. and Nicholas et al. suggest that ATP hydrolysis at AAA3 elevates the “MT-gate” to facilitate MT release and fast movement of dynein.[38,39] Thirumalai and his group investigated the pathway of allosteric transition in dynein computationally.[40] They observe that the interactions between the linker and AAA2-insert loop is persistent when AAA3 is bound to ATP and thus dynein is locked in a nonfunctional repressed state.[40] Though all these studies shed some light on the role of AAA3 in dynein’s mechanochemical cycle, a detailed investigation is needed for the clear understanding of the allosteric transition pathway of AAA3 in the mechanochemical cycle and power stroke formation of dynein.

Figure 1

Overview of the monomeric structure of dynein showing important structural components. (A) AAA+ ring comprising six AAA+ domains, AAA1 to AAA6 (shown in different colors), a linker domain spanning over the ring, and a long stalk domain attached with a globular microtubule binding domain (MTBD). (B) Post-power stroke state of dynein where four nucleotide binding domains have AMP-PNP, a nonhydrolyzable ATP analogue. The linker is in a straight conformation. (C) Pre-power stroke state of dynein where AAA1 has an ATP hydrolysis transition state analogue AOV, AAA2 has ATP, and AAA3 and AAA4 have ADP in their binding sites. The linker has a bent conformation in the pre-power stroke state.

Here, we developed a coarse-grained structure-based model (SBM)[41−43] to probe the gating mechanism of AAA3 using two end-state crystal structures. The SBM has been used extensively to study complex systems like motor proteins earlier.[36,44−49] A physical model can be built by using the structural information present in the Protein Data Bank. One can perturb the model by incorporating several interactions like ATP binding, ADP/Pi release, MT/actin binding, etc., to study the structural changes due to these perturbations. Other than native states, proteins acquire different conformations to execute their functions. Therefore, stabilizing interactions that are coming from other conformations should be incorporated to obtain a better characterization of the functional landscape. The SBM allows the mixing of the stabilizing interactions coming from other states. In addition, the SBM permits the possibility of “cracking”, which stabilizes the functional intermediates. Thence, the SBM can capture different scales of motions from native to non-native states relevant for different functions.

Results and Discussion

As our present study aims to explore the control of allosteric communication of dynein by its AAA3 domain during switching function, we constructed our model to follow the responses of other domains when AAA3 remains active and inactive during the transition. A multi-basin SBM was developed by mixing the topologies (see the Computational Methods section for details) of two end-state crystal structures: (i) the post-power stroke state where AAA3 along with other nucleotide binding sites have a nonhydrolyzable ATP analogue, AMP-PNP (PDB ID: 4W8F), and the linker is in the straight conformation[37] (Figure B) and (ii) the pre-power stroke state where AAA3 has ADP, AAA2 has ATP, and AAA1 has a transition state analogue ADP–vanadium complex (AOV) in its binding pockets and the linker is in the bent conformation (Figure C) (PDB ID: 4RH7).[27] From several experimental studies, it has been found that during the post- to pre-power stroke transition, the AAA1/2 cleft closes, the linker shifts from a straight to a bent conformation, and the MTBD gets detach from the MT.[30,31] In our present model, we have not considered the MTBD with MT explicitly; however, we have built our simulation strategy in such a way that we can draw a conclusion about the MTBD movement in an implicit way. It is also important to mention that, in our study, ATP is not involved explicitly; however, we can track the conformational changes implicitly.

Figure shows us the important conformational changes between the two end states. The contact map was calculated at the smog@ctbp online server[50−52] using a 0.6 nm default cutoff for Cα coarse-graining. We observed that the contacts of the linker with AAA5 and AAA2-insert loop (AAA2-IL), which are present in the post-power stroke state, disappear in the pre-power stroke state and a new set of contacts between AAA3/linker appear. We also find that new contacts are formed at AAA1L/2L and AAA5L/6L interfaces in the pre-power stroke state (Figure ).

Figure 2

Differential contact map of the post-power stroke and the pre-power stroke states. Blue and red contact pairs indicate the unique contacts of the pre-power stroke and the post-power stroke states, respectively.

To execute the post- to pre-power stroke transition when AAA3 is in the active and inactive states, we performed two types of multi-basin SBM simulations: (i) type-I and (ii) type-II. In type-I, we prepare a mixed topology with the contact pairs of both the pre- and post-power stroke states, which allows a smooth transition from one state to the other. We termed it as a nonrepressed simulation. In type-II, the topology is generated in a similar way; however, AAA3 lacks the information of the unique contact pairs of the pre-power stroke state, which creates repression of AAA3 in the post-power stroke state or ATP-bound state. However, other domains are free to move from one state to the other. Due to the repression of AAA3 motion, we termed it as a repressed-state simulation.

Allosteric Control of MTBD Movement

Previous studies have established that ATP binding to the AAA1 domain leads to the closing of the gap between AAA1 and AAA2 domains, which generate a large number of conformational transitions that propagate through the other domains and create AAA5 and AAA6 closing, the linker bending along with the MTBD detachment.[30,31] In our study, we investigated the changes in different domains using distance, RMSD, and the fraction of native contact (Q) as the order parameters. For the type-I scenario, we have calculated the distances of AAA1/2, AAA5/6, linker/AAA2, and linker/AAA5 over 50 trajectories to examine the cleft closing and linker shifting during simulations (Figure ). We have performed single-basin SBM simulations that uniquely stabilize the post- or the pre-power stroke state. Figures S1 and S2 represent the distribution of the distances between different domains in the pre- and post-power stroke states, respectively. In each plot, we put a dotted line corresponding to the peak of the distribution as obtained from two end-state single-basin simulations. All the distances are calculated by taking one residue from each domain (Figures S3 and S4). Please see the Supporting Information for details. Figure A,B shows that the AAA1/2 and AAA5/6 clefts close completely as the distances reach the pre-power stroke state values. The linker also shifts from the straight to the bent conformation as the distance between the N-terminal of the linker and AAA2L decreases and that between the linker and AAA5L increases to the values similar to the pre-power stroke state (Figure C,D). To confirm the complete transition, we have calculated the distribution of Qpre (fraction of unique native contact pairs of the pre-power stroke state formed during simulations) of each domain where the values from 0 to 1 indicate the post- to pre-power stroke transition (Figure ). We first calculated the Qpre values of each domain from single-basin simulations of the pre- and post-power stroke states (Figure S5). For the linker domain, we considered the unique contact pairs of the N-terminal linker with AAA3 because the linker only makes contact with AAA3 in the pre-power stroke state. From the distribution plot, we observed that, for the post-power stroke state, the peak values of the distribution range from 0.3 to 0.45 for AAA1 to AAA6 domains (Figure S5B), whereas the values for the pre-power stroke state are between 0.9 and 1 (Figure S5A). For the linker, the values are zero at the post-power stroke state and 1 at the pre-power stroke state. (Figure S5C) For type-I simulations, Figure A shows that the peak values of all the distributions are between 0.9 and 1.0, which signify a complete transition of each domain to the pre-power stroke state. Figure C shows a peak value at 1 for the linker/AAA3 contact pairs that indicate the bent conformation in type-I. We have also superimposed the structures obtained from type-I simulations with the pre-power stroke state crystal structures. Figure S6 shows that AAA1/2 and AAA5/6 and the linker domains superimposed completely. In the type-II scenario, as we know, AAA3 is in the switch off/inactive state, we examined all the major changes over 50 trajectories again so that we can compare them with dynein’s active condition. We noticed that the AAA1/2 cleft closes completely in all of the trajectories to attain the pre-power stroke state conformations (Figure A), whereas the AAA5/6 cleft closes partially (Figure B) as the distance cannot reach the pre-power stroke value and the linker remains in straight conformations (Figure C,D). There is a small increase in linker/AAA2 distance from the post state value because of the AAA1/2 cleft closing. We have calculated Qpre for the type-II scenario (Figure B) and found that AAA1 and AAA2 give peaks close to 1.0 and the AAA6 peak value is close to 0.9, which indicates a complete conformational change for these domains. For AAA3, the peak value is 0.6, which is far from the pre-power stroke state due to repression. The Qpre value of AAA3 in the post-power stroke state is 0.45. This small change (from 0.45 to 0.6) may be originated from the conformational changes of the other domains that force AAA3 to make few contacts. The peak values for AAA4 and AAA5 are 0.75 and 0.8, respectively, which suggest that AAA4 and AAA5 are unable to execute the full conformational changes to the pre-power stroke state. Also, for the linker (Figure C), we found a distribution with the peak value of 0.0 that implies that the linker remains straight and cannot move to the bent conformation. Figure S7 represents superimposed structures between conformations obtained from type-II simulations and the pre-power stroke state crystal structure. We observed that AAA5/6 domains are not completely superimposed and the linker remains in the straight conformation.

Figure 3

Distance plots of AAA1/2, AAA5/6, linker/AAA2, and linker/AAA5, which are calculated over multiple trajectories from the type-I scenario. The dotted line in each plot indicates the values of the distances in the pre- and post-power stroke states. (A, B) Plots showing that both the AAA1/2 and AAA5/6 clefts close completely as the distances reach the pre-power stroke state values. (C, D) Plots indicating the linker bending as the linker/AAA2 distance decreases and the linker/AAA5 distance increases to the pre-power stroke state values.

Figure 4

Fraction of unique native contact pairs of the pre-power stroke state (Qpre) of each domain in type-I and type-II simulations. The peak values of the distributions close to 1.0 indicate the pre-power stroke state. (A) In type-I simulations, the peak values for AAA1 to AAA6 are between 0.9 and 1.0, which imply a complete conformational change from the post- to the pre-power stroke state. (B) For type-II, AAA1, AAA2, and AAA6 show peak values between 0.9 and 1.0. AAA4 and AAA5 show peak values much less than 1, which indicate that AAA4 and AAA5 are unable to execute full conformational changes to the pre-power stroke state. Due to the repression in the AAA3 domain, it shows a peak value close to 0.6. (C) For the linker, the unique contact pairs of linker/AAA3 were considered because the linker only makes contact with AAA3 at the pre-power stroke state. In type-I simulations, the linker shows a peak value at 1.0; however, in type-II, the value is 0.0.

Figure 5

Plots of different domain distances in type-II simulations. (A) AAA1/2 cleft closes completely as the distance reaches the pre-power stroke state value. (B) AAA5/6 cleft closes partially as the distance cannot reach the pre-power stroke state value. (C) Linker/AAA2 distance does not change significantly from the post-power stroke state values. (D) Linker/AAA5 distance also remains at the post-power stroke state values. The dotted line in each plot indicates the values of the distances in the pre- and post-power stroke states.

We calculated the root-mean-square deviation (RMSD) of each domain (AAA1 to AAA6) from its initial position in both type-I and type-II simulations (Figure ). The distribution plots of RMSD (Figure ) show that the AAA4 and AAA5 have lesser RMSD values in type-II simulation as compared to type-I (Figure D,E), which indicates minimal structural changes of AAA4 and AAA5 from its post-power stroke conformation in type-II. AAA3 is repressed in type-II simulation, so AAA3 has very little change in RMSD (Figure C). For the other three domains, AAA1, AAA2, and AAA6, complete structural changes have occurred in both type-I and type-II as evident from Figure A,B,F, respectively.

Figure 6

Distribution plots of RMSD of each domain in type-I and type-II simulations. RMSD was calculated w.r.t. the crystal structure of the post-power stroke state of each domain. (A–F) AAA1 to AAA6 domains. Pink and blue colors indicate type-I and type-II, respectively. Note that AAA4 and AAA5 have lesser RMSD values in type-II as compared to type-I simulations, which imply minimal structural changes from their initial conformations. For AAA3, a little change in the RMSD value in type-II is obvious as AAA3 motion is repressed.

If we refer to the structure of dynein in Figure A, we can observe that a small globular MTBD domain is connected with the ATPase ring via a long coiled-coil stalk and a small buttress region that are extended forms of AAA4 and AAA5, respectively. The AAA3 domain is directly connected to AAA4 and allosterically to AAA5 via AAA4. So, there is a high possibility that any repression in AAA3 motion can affect AAA4 and AAA5 either directly or indirectly. When ATP binds to the AAA1 domain, the gap between AAA1 and AAA2 closes, which simultaneously creates a cascade of domain motions that propagates through the ring when AAA3 is in the ADP-bound state or active state and ultimately leads to the linker shifting and the MTBD movement. However, when AAA3 is in the ATP-bound state or inactive state, the propagation of domain motion through the ring is inhibited, and AAA4 and AAA5 domains cannot execute its full conformational change to create a proper mechanical force for the MT detachment. From our observation, it could be inferred that the inhibition of the complete motion of AAA4 and AAA5 domains indirectly favors the MT attached state of the MTBD as both the coiled-coil regions are unable to pull the MTBD for the detachment. Thus, AAA3 repression inhibits MTBD detachment and slows down dynein movement.

Control on the Linker Movement

The linker is an important element that generates force for the movement. We investigated the effect on the linker movement due to the AAA3 domain repression. From Figure C, we observed that the linker remains in the straight conformation in type-II simulations. However, to explore the thermodynamic pictures of different linker conformations, we calculated the free energy profile using an umbrella sampling technique (see Computational Methods for details) implemented in GROMACS, and the distance between the N-terminal linker and AAA2 was taken as an order parameter (Figure S4A,B). In type-I, we observed that the bent conformation of the linker is stabilized more than the straight one. However, in type-II, a larger stabilization of the straight conformation along with a high energy barrier going from the straight to the bent conformations of the linker makes the straight conformations of the linker much more favorable than the bent conformations on AAA3 repression (Figure ).

Figure 7

Free energy profiles of the linker in different conformations in both type-I and type-II scenarios. Here, the N-terminal linker/AAA2 distance is taken as an order parameter. A high linker/AAA2 distance indicates that the linker is in the straight conformation, and the low value represents the bent conformation of the linker. In type-I, the bent conformation is more favorable than the straight one. Note that, in the type-II scenario, the straight conformation is much more stabilized compared to the bent state.

We searched for the reason why the linker favors the straight conformation in the repressed state. Recent studies of Goldtzvik et al. have shown that the AAA2 insert loop (AAA2-IL) has an important role in stabilizing the linker in the straight conformation as the interaction between AAA2-IL and the linker prevents the linker from bending.[40] However, from contact map analysis (Figure ), we understand that the linker makes contact with AAA5 along with AAA2-IL in its straight conformations. Thirumalai and his group have elucidated the underlying mechanism using a coarse-grained self-organized polymer model.[40] They found that the ATP-bound state of AAA3 stabilizes the linker/AAA2-IL interactions that prevent the linker bending. However, what happens to the linker/AAA5 interactions or how important those interactions are for the regulation of switching function is not very clear from this study. In our present study, we have focused on both the linker/AAA5 and linker/AAA2-IL interactions to get a clear idea about the allosteric transition path. In the post-power stroke state, the linker/AAA2-IL distance is around 0.7 nm (Figure S2E), and the linker/AAA5 distance is around 2.2 nm (Figure S2D). In our type-I simulations, we found that the distance between linker/AAA2-IL increases from 0.7 to 2.0 nm initially due to the closing of the AAA1/2 cleft. The distance further increases to 3.5 nm, which is the pre-power stroke state value with the linker bending (Figure A). In type-II simulation, the AAA2-IL/linker distance increases from the post-power stroke state value to 2.0 nm due to the AAA1/2 cleft closing (Figure B). We have already seen that the linker/AAA5 distance does not change significantly and stays in the post-power stroke state distance in type-II (Figure D). This observation suggests that the AAA3 repressed state is unable to break the linker/AAA5 interactions. We can connect this fact with our previous observation that the linker/AAA5 contacts become stabilized because AAA5 is unable to complete its full conformational change. Some experimental groups also proposed that the rearrangement of AAA4 and AAA5 are needed for the dislodging of the linker.[37,53] Our present observations are in good agreement with the experimental work by Nicholas et al.[39] They investigated the effect of dynein-MT attachment in the presence of tension depending on different nucleotide states of AAA1 and AAA3 using optical tweezers. They pointed out a crucial role of the linker in the gating mechanism of dynein via AAA3. They observed that, if tension is absent or applied via the C-terminus of dynein, ATP at AAA1 facilitates MT release only if AAA3 is in the post-hydrolysis state. Instead, when tension is applied through the linker, ATP binding to AAA3 is sufficient to lift the regulatory gate. To further substantiate our observations, we have mutated the interactions between the linker and AAA2-IL by providing those particular contact pairs a repulsive potential in the repressed state. We observed that the linker/AAA5 interactions do not break due to the mutations of linker/AAA2-IL interactions and the linker prefers the straight conformation (Figure S8A,B), which agrees with our previous suggestion that a AAA5 conformational change is crucial.

Figure 8

Values of linker/AAA2-IL distances in type-I and type-II scenarios. (A) In type-I simulations, the linker/AAA2-IL distance increases initially from 0.7 nm (the post-power stroke value) to 2 nm, and it increases further to 3.5 nm (the pre-power stroke state) with the linker bending. (B) Linker/AAA2-IL distance increases from 0.7 to 2 nm; however, it cannot reach the pre-power stroke state value in type-II.

Next, we tried to find out what happens if we mutate the interactions between AAA5 and the linker. In a similar way, we mutated the contact pairs between the linker and AAA5 in type-II simulations. From Figure B, we observed that the linker/AAA5 distance increases, which indicates the destabilization of the straight conformations of the linker in type-II. Next, we observed that the linker gradually bends to the pre-power stroke state as the distance between linker/AAA2 decreases (Figure A). To make sure about the conclusion, we have calculated Qpre of the linker and found that the peak arises at 0.85 (Figure C), which signifies the linker bending. However, 15% linker/AAA3 contacts cannot form properly due to AAA3 repression. Figure D shows a representative structure where the linker bends after linker/AAA5 interactions have been mutated. From this observation, we suggest that AAA3 repression stabilizes linker/AAA5 interactions, which halt the linker movement; however, after removing the interactions, the linker moves to the bent state. The experimental work by Nicholas et al.[39] also showed that when tension was applied through the linker, MT-gate lifted in the presence of ATP at AAA3. To investigate the conformational change of AAA4 and AAA5 due to linker/AAA5 mutations, we have calculated Qpre for AAA4 and AAA5. From Figure S9, we noticed that AAA4 and AAA5 showed peak values much less than type-I simulations. From these observations, we understand that the ATP-bound AAA3 repressed the motion of AAA4 and AAA5 domains so that they cannot execute their full conformational change to the pre-power stroke state, and as a result of that, the linker favors the straight conformation because the linker/AAA5 interactions get stabilized. As we understand that the conformational change of AAA4 and AAA5 is very crucial for the linker movement, we forcefully change the AAA4 and AAA5 domains to the pre-power stroke state (see Computational Methods for details) in the type-II scenario to examine what happens to the linker. From Figure S10A,B, we infer that the linker completely bends to the pre-power stroke state as the linker/AAA2 distance decreases and the linker/AAA5 distance increases. The Q value for the linker is also close to 1.0 (Figure S10C). We performed another type of simulation where the AAA2 domain was repressed (type-III) to show that the effects we were getting were not potential artefacts of our model and were unique to the AAA3 domain only. From Figure S11, we notice that the AAA1/2 and AAA5/6 clefts close completely and the linker shifts from the straight to the bent conformation after AAA2 repression. Figure S11D shows that peak values of Qpre of each domain are close to 1.0 except that of AAA2, which implies a complete conformational change of other domains to the pre-power stroke state.

Figure 9

Conformational changes of the linker where the linker/AAA5 interactions have been mutated in the type-II scenario. (A, B) Plots showing that both the linker/AAA2 and linker/AAA5 distances increase initially; however, after some time, the linker/AAA2 distances decrease to the pre-power stroke values, which indicate linker bending. (C) Qpre value for the linker shows a peak at 0.85, which signifies the linker bending. Because of the AAA3 repression, the linker cannot make full contact with AAA3. (D) Structural representation showing that the linker completely bends after mutations.

The entire observations suggest that the ATP-bound AAA3 gates dynein’s mechanochemical cycle by directly governing the AAA4 and AAA5 motions, which are important for the power stroke formation and the linker movement. The AAA3 repressed state resists the allosteric communication from AAA1 to AAA4 and AAA5 that hampers the full conformational change of these two domains. As a consequence, MTBD detachment from the MT through conformational changes within the coiled-coil stalk and the buttress is prevented, and the linker/AAA5 interactions get stabilized to preserve the straight conformation of the linker. In support of this conclusion, Rao et al.[54] recently found from their experimental work that MT detachment requires dissociation of the linker from AAA5 and that the docking of the linker to AAA5 in the straight conformation is required for the strong MT binding registration of the stalk coiled coil. They demonstrated that the buttress is very important for dynein motility. When they truncated the buttress to prevent stalk–buttress interactions, they observed that the MT binding affinity was reduced significantly and that the tension-induced transition from weak to strong MT binding was prevented. The linker controls conformational changes of the buttress through the docking and undocking to and from AAA5 that control the strong and weak MT binding states of the motor. They proposed that the linker/AAA5 interactions induce conformational changes within the buttress that result in the sliding of the stalk helices into the α or γ registries to induce increased MT binding and that preventing the detachment of the linker from AAA5 inhibits the transition into the weak MT binding β registry. This experimental work supports the importance of linker/AAA5 interactions in the AAA3-mediated gating functions.

Conclusions

In our present study, we used a structure-based model to gain insight into how AAA3 motion governs other different domain motions, which are important for allosteric communication during power stroke formations. We have generated two models: in one case, we allowed all the domains to move freely from the post-power stroke (i.e., AMP-PNP/ATP-bound state) to the pre-power stroke state (i.e., ADP at AAA3), and in another model, we repressed AAA3 motion in the ATP-bound conformation intentionally where the topology lacks the information of the pre-power stroke state of AAA3; however, other domains are free to move. Previously, Thirumalai and his group investigated the molecular mechanism behind this AAA3 regulatory switching function.[40] Their computational study using a self-organized polymer model proposed that the ATP binding to the AAA3 stabilizes linker/AAA2-IL interactions that prevent linker bending. We already know that the linker makes contact with AAA5 along with AAA2-IL in the straight conformation. However, the importance of linker/AAA5 interactions was not very clear from their study. Many experimental works have already shown that the linker/AAA5 interactions are crucial for the maintenance of the straight conformation of the linker.[39,54] Rao et al.[54] demonstrated that the impairment of functional linker/AAA5 interactions results in the weak MT binding β registry, suggesting that linker/AAA5 interactions are required for the tension-induced and strong MT binding α registry. In our present work, we focused on both the linker/AAA5 and linker/AAA2-IL interactions to find out the underlying mechanism. We noticed that the conformational change in the AAA3 domain, going from the ATP- to ADP-bound conformation, triggers AAA4 and AAA5 domains’ motions, which eventually generate a proper force to pull the MTBD from the MT through the stalk and buttress. However, when ATP hydrolysis at AAA3 is hampered indirectly by inhibiting its conformational change, AAA4 and AAA5 cannot move to the pre-power stroke state completely, and it indirectly restricts the stalk domain movement, which in turn leads to the high affinity of the MTBD with the MT and slows down dynein motility. Our present work is directly supported by the recent works of Rao et al.[54] and Nicholas et al.[39] Both the experimental studies suggest the importance of linker/AAA5 interactions in stalk-helix sliding and MT detachment. When the linker is docked to the AAA5, the buttress cannot undergo the necessary conformational change to induce the weak MT binding β registry to cause MT release. Therefore, the linker must undock from AAA5 to facilitate rear head detachment to support rapid forward movement. In support of this conclusion, Rao et al.[54] have also demonstrated that a dynein motor with the stalk helices cross-linked into the strong MT binding α registry moves at a significantly reduced speed. Again, when they applied tension through the linker, they found that ATP binding to AAA3 is sufficient to promote MT release.

Bhabha et al. suggested that the linker, which is a mechanical element, is also important for the ATPase activity of dynein.[37] From their cryo-EM data, they propose that the unbinding of the linker from AAA5 and bending promote ATP hydrolysis at AAA1. The N-terminal linker at the AAA3 position influences the conformation of the R finger at AAA2 in a proper catalytic component, which is important for ATP hydrolysis at AAA1.[4,37] Here, we observed that the repressed state of dynein prevents unbinding of the linker from AAA5 (Figure C) and favors the straight conformation, which is also suggested by the free energy calculations (Figure ). As AAA1 is the primary ATPase site, each ATP hydrolysis at AAA1 triggers dynein’s single step. The linker in the straight conformation basically inhibits the ATP hydrolysis at AAA1, which causes the retardation of ATPase activity of dynein. Our study suggests how AAA3 acts as a switching motif in the mechanochemical cycle of dynein (Figure ). We conclude that AAA3 gates the mechanochemical cycle of dynein with the assistance of AAA4 and AAA5 that directly controls the power stroke formation and the linker movement.

Figure 10

Model showing that ATP hydrolysis at AAA3 regulates the activity of dynein. From no nucleotide state, AAA1 and AAA3 bind ATP. The ATP-bound AAA3 resists the propagation of conformational transition from AAA1 to AAA5. Thus, the mechanochemical cycle cannot proceed further. However, after ATP hydrolysis at AAA3 dynein gains its activity that leads to the allosteric communication of ATP-induced AAA1 through the other domains like AAA4 and AAA5, thereby facilitating the linker movement and MT release.

Computational Methods

We used the Cα coarse-grained model to reduce the structural complexity and long time scale problem for this large molecular machine. We took two crystal structures from the Protein Data Bank (PDB): (i) yeast dynein in the post-power stroke state where AAA1 to AAA4 are bound to ATP analogues, AMPPNP (PDB ID: 4W8F)[37] and (ii) the pre-power stroke state of human dynein where AAA1, AAA2, and AAA3 are bound to AOV, ATP, and ADP, respectively (PDB ID: 4RH7).[27] As we have two structures from two different species, their sequences are different. In order to develop both the structure from the same sequences, we modeled our pre-power stroke structure from the SWISS-MODEL server[55] by taking the FASTA sequence of a yeast dynein and human dynein as a structural template. Total residues in both the structures are 2483, including linker, AAA1, AAA2, AAA3, AAA4 (with a small extended part of coiled coil), AAA5 (with a small extended part of buttress), and AAA6 domains.

Building of a Multi-Basin Structure-Based Model

Cα coarse-grained models were constructed, and topology files were generated for the post- and pre-power stroke structures using the SMOG@ctbp[50−52] online server. We developed a multi-basin structure-based model by mixing the topologies from both the states. Let us consider that the native contact pairs present in the post- and pre-power stroke states are Mo and Mc, respectively. The native contact pairs are calculated directly from two crystal structures with the same sequences using SMOG@ctbp server taking 0.6 nm as a cutoff distance. Some contact-pairs are identical in both the post- and pre-power stroke states, and they are called common or shared contact pairs ((Mshared = Mo ∩ Mc). Some contacts are unique to a particular state; they are MoUnique for the post-power state and McUnique for the pre-power state. For the post- to pre-power stroke transition, we have generated a mixed topology file incorporating the common contact pairs and the unique contact pairs from both the states. The Hamiltonian for multi-basin SBM iswhere HB and HNB represent the local bonded and nonbonded components, respectively, and superscript O and C represent the open/post-power state and the closed/pre-power state, respectively. The general form of the local bonded Hamiltonian is

The first term r is the distance between two consecutive residues i and i+1, and it is harmonically constrained with respect to its native distance r0 by a spring constant Kr, where Kr = 200 kJ mol–1 Å–2. The second term θ represents the angle between residues of i, i+1, and i+2 and is constrained by a harmonic spring constant Kθ, where Kθ = 40 kJ mol–1 rad–2, and it is constrained with respect to its native value θ0. The dihedral angle potential is constituted by the third term, which delineates the rotation of the backbone involving successive residues from i to i+3, where Kφ(1) = 2Kφ(3) and Kφ(1) = 1 kJ mol–1.

The general form of the nonbonded Hamiltonian HNBO(unique), HNBC(unique), and HNB(shared) is represented by

HNBO(unique) and HNBC(unique) refer to the unique contact pairs of the post- and pre-power stroke states, respectively. HNB(shared) represents the common pairs present in both the states. If i and j residues are in contact, then Δ = 1; otherwise, Δ = 0. A repulsive potential is implied to the non-native pairs (Δ = 0). ε and εr was set to 1.0 kJ mol–1.

All simulations were performed using the GROMACS MD engine. For type-I simulations, the mixed topology includes the common contact pairs along with the unique contact pairs of both the post-power stroke and pre-power stroke states. We start our simulation from the post-power state, and the system gradually visits the pre-power state through different intermediates. For type-II, the mixed topology was constructed including the common contact pairs of both the states, the unique contact pairs of the post-power stroke, and the unique contact pairs of the pre-power stroke state for each domain excluding the intradomain contacts of AAA3; however, the contacts of AAA3 with other domains remain intact. For mutation purposes, we considered a repulsive potential for a particular domain–domain contact pair where Δ = 0. When AAA4 and AAA5 domains are forced to change to the pre-power stroke state, we took ε = 1.5 kJ mol–1 for the unique contact pairs of the pre-power stroke state of AAA4 and AAA5.

SBM Simulations

Both the initial structures were relaxed in the structure-based Hamiltonian, and to collect different equilibrium ensembles, we performed Langevin dynamics at 90 K (reduced temperature where T* = 0.75) and at low friction limit to improve sampling. The equation of motion for Langevin dynamics iswhere ζ represents the friction coefficient, –∂({r⃗}) is the conformational force, and Γ⃗(t) is the random force that satisfies where the integration time step is 0.0005τL, where . Here, m is the mass of the Cα atom, σ is the van der Waals radius of Cα, and ϵ is the solvent-mediated interactions. We performed 50 simulations (a particular subset of each time vs distance plot is shown in Figures S12 and S13) in each case where each simulation lasts for 109 steps.

Free Energy Calculations

For the free energy calculations (Figure ), we used an umbrella sampling technique implemented in GROMACS. The distance between the N-terminal linker and AAA2 was considered as an order parameter to explore the stability of the linker conformations in the nonrepressed and AAA3-repressed conditions. In each case, we considered 46 windows where the distance varies from 7.8 to 3.1 nm. In each window, a 1000 kJ/mol/nm2 biased force was applied to maintain the particular distance, and equilibration and production runs were performed for 107 and 108 steps, respectively. From the histogram of distances, obtained from each window, we performed the weighted histogram analysis method (WHAM)[56] to extract the free energy profile.