Simulated Dynamics of Glycans on Ligand-Binding Domain of NMDA

Oct 11, 2017 - computational predictions have been experimentally confirmed ... (a) NMDA receptors consist of relatively autonomous parts, one of whic...
0 downloads 12 Views 3MB Size
Subscriber access provided by UNIVERSITY OF ADELAIDE LIBRARIES

Article

Simulated Dynamics of Glycans on LigandBinding Domain of NMDA Receptors Reveals Strong Dynamic Coupling between Glycans and Protein Core Anton V. Sinitskiy, and Vijay S. Pande J. Chem. Theory Comput., Just Accepted Manuscript • DOI: 10.1021/acs.jctc.7b00817 • Publication Date (Web): 11 Oct 2017 Downloaded from http://pubs.acs.org on October 13, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Journal of Chemical Theory and Computation is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

Fig 1 212x155mm (96 x 96 DPI)

ACS Paragon Plus Environment

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

146x192mm (96 x 96 DPI)

ACS Paragon Plus Environment

Page 2 of 40

Page 3 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

Fig 3 180x215mm (96 x 96 DPI)

ACS Paragon Plus Environment

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

308x549mm (96 x 96 DPI)

ACS Paragon Plus Environment

Page 4 of 40

Page 5 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

164x256mm (96 x 96 DPI)

ACS Paragon Plus Environment

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

graphics for TOC 480x167mm (72 x 72 DPI)

ACS Paragon Plus Environment

Page 6 of 40

Page 7 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

Simulated Dynamics of Glycans on Ligand-Binding Domain of NMDA Receptors Reveals Strong Dynamic Coupling between Glycans and Protein Core Anton V. Sinitskiy1*, Vijay S. Pande1,2* 1

Department of Chemistry, 2 Department of Computer Science and Department of Structural

Biology, Stanford University, Stanford, CA 94305, USA.

N-methyl-D-aspartate (NMDA) receptors, key neuronal receptors playing the central role in learning and memory, are heavily glycosylated in vivo. Astonishingly little is known about the structure, dynamics and physiological relevance of glycans attached to them. We recently demonstrated that certain glycans on the ligand binding domain (LBD) of NMDA receptors (NMDARs) can serve as intramolecular potentiators, changing EC50 of NMDAR co-agonists. In this work, we use molecular dynamics trajectories, in aggregate 86.5 μs long, of the glycosylated LBD of the GluN1 subunit of the NMDAR to investigate the behavior of glycans on NMDARs. Though all glycans in our simulations were structurally the same (Man5), the dynamics of glycans at different locations on NMDARs was surprisingly different. The slowest-timescale motions that we detected in various glycans in some cases corresponded to a flipping of parts of glycans relative to each other, while in other cases reduced to a head-to-tail bending of a glycan. We predict that timescales of conformational changes in glycans on the GluN1 LBD of

ACS Paragon Plus Environment

1

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 40

NMDARs range from nanoseconds to at least hundreds of microseconds. Some of the conformational changes in the glycans correlate with the physiologically important clamshelllike opening and closing of the GluN1 LBD domain. Thus, glycans are an integral part of NMDARs, and computational models of NMDARs should include glycans to faithfully represent the structure and the dynamics of these receptors.

Introduction N-methyl-D-aspartate (NMDA) receptors are neuronal transmembrane proteins playing the central role in learning and memory.1-3 These receptors are calcium-permeable ion channels that open only when agonist (glutamate) and co-agonist (D-serine or glycine) bind to it, and the voltage of the membrane changes. NMDA receptors (NMDARs) have been considered as important molecular targets for treating schizophrenia, depression and other neurological disorders.4-12 Like most proteins in human and other eukaryotic organisms, NMDARs are heavily glycosylated in vivo. Each receptor consists of two GluN1 subunits and two GluN2 or GluN3 subunits. At least 11 glycans are attached to the GluN1 subunit of NMDARs, at least 4 glycans to the GluN2A and at least 7 glycans to the GluN2B subunits.13-15 The role of glycans in NMDARs has previously been underestimated. Structural studies of NMDARs by X-ray or cryoEM are often based on the receptors or their fragments produced in non-glycosylating expression systems (e.g., E. coli) or even purposefully de-glycosylated.16-25 In the few structures of NMDARs that do involve glycans, the glycosylation pattern is significantly distorted by mutations introduced to crystallize the receptor, and the remaining glycan residues

ACS Paragon Plus Environment

2

Page 9 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

are only partially resolved.26-28 Molecular dynamics (MD) simulations of NMDARs or their parts historically omitted glycans.20,23,29-37 On the other hand, glycosylation has been shown to significantly affect physiological properties of NMDARs, as well as their trafficking to cell membranes.38-41 In particular, the removal of all glycans decreases EC50 for glutamate by a factor of 3.6±0.739 and reduces the ratio of the steady-state current amplitudes induced by 50 μM and 1 mM NMDA by 1.3±0.1.40 Computational modeling can provide unique information about the structure and dynamics of glycans, unavailable from experimental techniques alone due to the highly dynamic nature of glycans. Typical geometries of glycans, in the absence of the corresponding experimental data, can be reconstructed by equilibrating the system from a reasonable initial guess of its geometry, and metastable states can be reached by MD or Monte Carlo simulations and identified, for example, with the use of Markov state models (MSMs). The timescales of conformational transitions in the system and the corresponding structural changes can also be obtained with the use of MSMs or other computational methods. Computations in this case can guide further experimental investigation, for example, by suggesting an appropriate spectroscopic method with the temporal window matching the predicted timescales of conformational transitions. In this work, we clarify at atomic resolution the behavior of glycans on the GluN1 LBD of NMDARs. We simulate the GluN1 LBD as a separate protein, isolated from the rest of the receptor (Fig. 1a,b). This approach is justified by the experimentally established modular structure of NMDARs.42,43 Three Man5 glycans (Fig. 1c) were added to the GluN1 LBD to match the physiological pattern of glycosylation of NMDARs in vivo (Fig. 1b).44 Simulations of the full-length receptor (Fig. 1a), which is ~10 times larger than a single GluN1 LBD, would be too slow with the existing computational resources to reach the tens-of-microseconds sampling45

ACS Paragon Plus Environment

3

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 40

that we report here and, therefore, observe the reported conformational transitions in the glycans. In any case, we consider our results presented here as predictions that will guide further research, including experiments on full-length NMDARs. The choice of the GluN1 LBD as the part of the receptor to focus on is motivated by the role it plays in co-agonist binding, which provides a convenient location for modulating NMDAR activity and may be involved in neurological disorders such as schizophrenia.46-48 This work builds upon our previous study that demonstrated the influence of glycans attached to NMDARs on co-agonist potency.41 In that work, computational modeling indicated that a glycan attached to one lobe of the GluN1 LBD of NMDARs can transiently non-covalently bind to the other lobe, and thereby serve as a latch between the lobes, stabilizing the closed-clamshell conformation of the LBD domain. The closed-clamshell conformation of this part of the receptor is believed to be a necessary condition for the ion channel to open.16 Our computational predictions have been experimentally confirmed by electrophysiological recordings in full-length wild-type and mutated (to modify glycosylation) NMDARs.41 The strength of the interactions between this glycan and the opposite lobe of the GluN1 LBD domain was large enough to increase the coagonist EC50 in the full-length receptor by 50±3%, a change comparable by the order of magnitude to changes in the coagonist or agonist affinity in NMDARs with diseasecausing mutations.41 In this paper, we proceed to analyze the structure and dynamics of the glycans on the GluN1 LBD of NMDARs in more detail, revealing the main conformational transitions in the glycan shell, and estimating their timescales. In particular, we present new results on physiologically relevant interactions between the glycan shell and the protein core of this part of NMDARs.

ACS Paragon Plus Environment

4

Page 11 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

Results Simulations reveal a wealth of conformations of glycans on the GluN1 LBD of NMDA receptors To identify various conformational states of the glycans in the GluN1 LBD of NMDAR, we performed time-structure based independent components analysis (tICA)49,50 on all pairwise inverse distances between all O and N atoms in the glycans (Fig. 1d). In this way, we determine the collective variables (time-structure based independent components, tICs) capturing the slowest conformational changes in the glycan shell in the GluN1 LBD of NMDARs. The physical degrees of freedom (intermolecular distances) most closely correlated with each of the first 10 tICs are shown in Table 1 and Fig. 2. This approach allows for a visualization of the sampled conformational landscape of the studied system (Fig. 3a) [for further justification and details on the methodology, see Methods]. Note that an analysis of the GluN1 LBD opening and closure as a separate degree of freedom was presented earlier,41 while the interdependence between this degree of freedom and the tICs for the glycan shell is considered below in this section. Sampled conformations of the glycan shell cluster into several states, with rare transitions between these states that occur on up-to-microsecond timescales (Fig. 3a). The overall shape of the sampled conformational landscape in the subspace of the first ten tICs (Fig. 3a) resembles a ten-dimensional letter Ж. Interestingly, most paths between clusters are narrow and nearly straight, suggesting that most conformational transitions in the glycan shell can be described by one-dimensional reaction coordinates, each of which is defined as a linear combination of certain inverse interatomic distances. Below, we provide a detailed characterization of some of the

ACS Paragon Plus Environment

5

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 40

conformational states of the glycan shell defined as clusters in the tIC subspace (in order of decreasing characteristic timescales). In conformational state A of the glycan shell, sugar residue 3 in the glycan attached to amino acid residue N440 (this is the glycan serving as a latch between the two lobes of the protein) undergoes a chair-to-chair transition (Fig. 4a). As a result, the glycan at N440 assumes a shape resembling a three-legged stool, with sugar residue 3 as the seat of the stool, and sugar residues 2, 4 and 7 as the legs (the numeration of sugar residues used in this work is shown in Fig. 1c). State A is distinguished by large negative values of the first tIC (Fig. 3a). Rare transitions to this state could be explained by the fact that two oxygen atoms, one between residues 2 and 3, and the other between 3 and 7, need to move from equatorial to energetically less favorable axial positions in residue 3. Nevertheless, the resulting conformation is relatively stable, presumably due to more numerous van-der-Waals interactions between neighboring sugar residues. In conformational state B, two glycans attached to amino acid residues N471 and N771 noncovalently interact with each other in an approximately parallel orientation. In doing so, sugar residue 2 in the glycan at N471 and sugar residue 3 in the glycan at N771 locate close to each other (Fig. 4b). State B is identified by large positive values of tIC 5 (Fig. 3a). In conformational state C, the glycan at N440 shifts into the space between the lobes and assumes a compact shape (Fig. 4c). As we showed earlier, this glycan generally serves as a latch between the two lobes of the protein, stabilizing the closed-clamshell state of the protein. Typically, this glycan is relatively extended in the closed-clamshell closed-latch conformations, while is state C it is significantly more compact. State C is identified by large negative values of tIC 4 (Fig. 3a). Among all interatomic distances in the glycan shell, this tIC most closely correlates with the distance between atom O4 in residue 3 in the glycan at N440 and atom N2 in

ACS Paragon Plus Environment

6

Page 13 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

residue 1 in the glycan at N771. Conformations from state C are characterized by a much smaller value of this distance (~33 Å) than those from other states (~40-58 Å). Residue 1 at the head of the glycan at N771 serves here as a reference point for measuring the position of the glycan at N440 (in particular, its central residue 3), and the dynamics of the glycan at N771 does not seem to be involved in this degree of freedom. Thus, this degree of freedom actually describes the dynamics of the glycan at N440 relative to the protein core. In conformational state D, the glycan at N471 bends such that sugar residues 1 and 3 approach each other (Fig. 4d). State D is identified by large positive values of tIC 9 (Fig. 3a). In conformational state E, the glycans at residues N471 and N771 move away from each other and the distance between their tails increases (Fig. 4e). This state is characterized by large positive values of tIC 8 (Fig. 3a). In conformational state F, the glycan at N471 bends, and the tail of this glycan non-covalently binds to the head of the glycan at N771 (Fig. 4f). This state is characterized by large negative values of tIC 5 (Fig. 3a). In conformational state G, a chair-to-chair transition occurs in sugar residue 4 in the glycan at N471, and as a result, the glycan tends to form a compact structure, with the tail of the glycan (sugar residues 5 and 6) non-covalently binding to the head and the middle of the same glycan (Fig. 4g). This state is characterized by large negative values of tIC 3 (Fig. 3a). More conformational states can be identified, in particular, with the use of higher-order tICs. However, we limit ourselves to the explicit description of the above-listed seven clusters, since they represent the slowest transitions captured in our simulations, and provide an overall concept of the diversity of conformational transitions occurring in the glycans in the studied system.

ACS Paragon Plus Environment

7

Journal of Chemical Theory and Computation

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 40

Timescales of conformational changes in the glycans cover a wide range from nanoseconds to microseconds To study the conformational dynamics of this glycoprotein, we used Markov state models (MSMs) as an approach well suited for the case of rare transitions between multiple well-defined states (Fig. 3b).51 MSMs represent the dynamics of a biomolecule as it hops between multiple discrete states. MSMs serve as a powerful tool to process the results of multiple parallel MD simulations, and typically can capture processes on the conformational landscape with timescales comparable to the aggregate duration of all used MD trajectories, thereby, much longer than the duration of each single trajectory. The main limitation of MSMs as tools for molecular modeling is that the real-world systems are not precisely Markovian; that is, they retain some memory about past conformational transitions. However, the relatively isolated character of most clusters on the conformational landscape in the studied system (Fig. 3a) and relatively rare transitions between such clusters justify the assumption that the glycoprotein has enough time to equilibrate within each metastable state between the transitions, leading to the loss of memory and to Markovianity of transitions between clusters. Characteristic timescales of the processes in the glycan shell that we sampled in our simulations cover a wide range from nanoseconds to hundreds of microseconds. With the currently available sampling, we observe 4 different processes with characteristic timescales above or comparable to 100 μs (respectively, with the first-order reaction rate constants below 10 ms-1) , 3 timescales in the range between 10 and 100 μs (rate constants between 10 and 100 ms1

), 6 timescales between 1 and 10 μs (rate constants between 102 and 103 ms-1), and many more

timescales densely spread in the submicrosecond range (Table 2). For comparison, we previously estimated the timescale of the main physiologically important conformational transition in the

ACS Paragon Plus Environment

8

Page 15 of 40

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Chemical Theory and Computation

protein part of the same system (clamshell-like opening and closing of the GluN1 LBD) to be ~0.5 μs.41 Thus, the dynamics of glycans on the GluN1 LBD of NMDARs is complex, does not contain wide gaps in the hierarchy of characteristic timescales, and goes beyond the timescale of the physiologically relevant clamshell-like motion. The slowest conformational transitions in the glycan shell that we captured occur on the timescales of tens to hundreds of microseconds, which is of the same order of magnitude as the aggregate sampling (Table 2). A transition between state A and other states, accompanied by formation or vanishing of a ‘three-legged stool’ conformation of the glycan serving as a latch between two lobes of the GluN1 LBD, was observed only once in the simulations. Respectively, we can only give a lower bound estimate on this timescale, namely that it is significantly greater than the aggregate sampling in our simulations, that is >>100 μs (respectively, the rate constant