Supplementary data

CAVER Analyst 2.0: Analysis and Visualization of Channels and Tunnels in Protein Structures and Molecular Dynamics Trajectories

Adam Jurcik1, David Bednar2,3, Jan Byska4, Sergio M. Marques2,3, Katarina Furmanova1, Lukas Daniel2,3, Piia Kokkonen2,3, Jan Brezovsky2,6,7, Ondrej Strnad1, Jan Stourac1,2,3, Antonin Pavelka1,2, Martin Manak5, Jiri Damborsky2,3*, Barbora Kozlikova1,*

1 Human Computer Interaction Laboratory, Faculty of Informatics, Masaryk University, Botanicka 68a, 602 00 Brno, Czech Republic; 2 Loschmidt Laboratories, Department of Experimental Biology and Research Centre for Toxic Compounds in the Environment, Faculty of Science, Masaryk University, Kamenice 5/A13, 625 00 Brno, Czech Republic; 3 International Centre for Clinical Research, St. Anne’s University Hospital, Pekarska 53, 656 91 Brno, Czech Republic; 4 Visualization Group, Department of Informatics, University of Bergen, Thormøhlensgate 55, 5008 Bergen, Norway; 5 Department of Computer Science and Engineering, Faculty of Applied Sciences, University of West Bohemia, Univerzitni 8, 306 14 Plzen, Czech Republic, 6 Laboratory of Biomolecular Interactions and Transport, Department of Gene Expression, Institute of Molecular Biology and Biotechnology Faculty of Biology, Adam Mickiewicz University, Umultowska 89, 61-614 Poznan, Poland, 7 International Institute of Molecular and Cell Biology in Warsaw, Ks Trojdena 4, 02-109, Warsaw, Poland

*To whom correspondence should be addressed.

Contents

Implementation and configuration

Supplementary figures

· Supplementary Fig. X1 Representation of tunnel width evolution over time.

· Supplementary Fig. X2 Contour representation of the tunnel bottleneck.

· Supplementary Fig. X3 Mutagenesis window.

· Supplementary Fig. X4 Clip plane window.

Case study – Exploration of tunnel profile and bottleneck

References

Implementation and configuration

CAVER Analyst 2.0 is a JAVA-based software. Due to its highly modular architecture, it can be easily customized. CAVER Analyst 2.0 runs on a common hardware without the need for special hardware upgrades. This allows for seamless integration into existing IT environments. The application is supported by the following operating systems: Windows 8 and 10, Mac OS X 10.7.5 or later and major distributions of Linux including Fedora Core, Red Hat, and Ubuntu. The application can run on both 32-bit and 64-bit system architectures and requires JAVA version 1.8 or later.

The processing of small datasets requires at least 32-bit architecture with 2-4 GB RAM. Medium to large datasets require 64-bit architecture with 8 and more GB RAM, which is also the recommended configuration. AMD Radeon or NVIDIA GeForce dedicated graphics cards are recommended for the utilization of advanced visualization techniques. CAVER Analyst 2.0 application is distributed as a complete package with all the modules required for running the application, including the user guide.

The example data sets are available at http://caver.cz/fil/download/examples/caver_analyst2_example.zip.

Supplementary figures

Supplementary figure X1. Profile representation of the tunnel radius along its length, with the respective variation over time (top) and the amino acids forming the tunnel boundary (bottom). The tunnel comes from a representative MD simulation of a representative protein. The tunnel radius representation enables to reveal the potential bottlenecks and their stability. The bottom part is the ranking of the amino acids according to their influence on the tunnel boundary. Each set of colored lines corresponds to one amino acid and its influence over time. The length of these lines shows the position and size of the influenced tunnel area. The sets are colored according to a selected physico-chemical property. The amino acids can be ranked also according to these properties. The vertical slider enables to select a tunnel section which can be explored in detail using the contour visualization (see Figure X2).

Supplementary figure X2. Contour representation of the tunnel bottleneck from a representative MD simulation of a representative protein. The central part contains the set of overlapping contours which are aligned according to the surrounding amino acids. Each line corresponds to a single time step. The bars surrounding the contours correspond to the amino acids forming the bottleneck boundary. The height of the bars corresponds to the relative amount of time for which the given amino acid defined the bottleneck. The bars are positioned in co-centric circles, each circle corresponding to one physico-chemical property. The left vertical slider shows the range of areas of contours, influencing also the contour colors.

The contour representation can be mapped onto the 3D tunnel representation as well.

Supplementary figure X3. Mutagenesis window enabling the design of mutations targeting individual residues. The user can define a residue using its sequence number or picking it from a selection. Then the desired rotamer library and desired amino acid type can be selected. The window then lists the set of potential rotamers of a selected amino acid, along with their probability of occurrence, standard deviation, and collision penalty score. To better select the best rotamer candidate, the user can display the collisions of the newly introduced side-chain with the rest of the protein. After clicking on the Apply mutation button, the mutation is introduced and displayed in the Structure mutations panel of this window. This process can be repeated until all desired mutations of the protein are inserted. The final mutant can be exported to the PDB file format.

The 3D visualization demonstrates the influence of the mutation of one of the bottleneck residues, N52, on the main tunnel of the haloalkane dehalogenase DhaA. Left – the original N52 residue (dark blue) and the bottleneck sphere of the main tunnel (yellow). Right – mutated G52 residue, resulting in a thicker main tunnel.

Supplementary figure X4. Structure clip planes window enables the users to add several clip planes for each loaded structure. The clip planes can be attached to the structure – when manipulating with the structure by rotation or translation, the planes are moving as well. This functionality helps to create and explore an arbitrary cut through the structure. The user can set the distance of the clip plane from the protein, realign it according to the viewport, and invert the visibility of the structure. Each clip plane can be applied to the structure selections as well as to the computed tunnels. This functionality also enables the creation of a slice of a user-defined thickness. The 3D images demonstrate the usage of slices, also with the detected tunnel.

Case study

The following case study aims to demonstrate the usage of the newly developed and implemented visualization methods, providing the users with an interactive exploration of tunnel bottlenecks and tunnel width dynamics over the time in the MD simulations.

Case study – Exploration of tunnel profile and bottleneck

This study, based on the publications by Koudelakova et al., 2013 and Liskova et al., 2015, focuses on increasing the thermostability, kinetic stability, and resistance to organic cosolvents of the haloalkane dehalogenase DhaA. The study is based on designing mutations around the bottleneck of the main tunnel, which controls the flow of the solvent molecules into and out of the protein active site. The study included MD simulations of the wild type of DhaA and two of its mutants with modified access tunnels, DhaA80 and DhaA106. In the published papers the authors used traditional protein engineering approaches to design two subsequent mutations. This approach can be very time consuming, potentially does not lead directly to satisfying solutions and several mutations have to be designed. This was also the case of the designed DhaA80 mutant. Despite of the increased stability of this mutant, its reactivity did not change significantly. Therefore, the DhaA106 mutation designed next resulted in a satisfying balance between these two properties. Our case study aims to show how our newly proposed visualizations could help the design process by better understanding the influence of tunnel-lining residues. We demonstrate it on 500 time-step long MD trajectories of the DhaA wild type, and its mutants DhaA80 and DhaA106. The input trajectories, including the data describing the tunnels computed using the standalone CAVER 3.02 tool, can be downloaded from http://caver.cz/fil/download/examples/caver_analyst2_example.zip.

The visual analysis starts with the exploration of the main tunnel detected in the MD trajectory of the wild type DhaA. We load the trajectory to CAVER Analyst. Then we load the tunnels precomputed using the standalone CAVER tool. This computation might as well be performed by CAVER Analyst itself. On the top panel, we select Tunnel – Residue Graph, which opens a toolbox for analyzing tunnel profile over time. We select the main tunnel “tun_cl_001_1” in DhaA and compute the aggregated visualization (Figure 1).

The upper part of the visualization depicts the tunnel profile and its changes over time, while the bottom part shows the amino acids lining the tunnel. The colors of the bars correspond to the hydrophobic character of the respective amino acids. The user can choose to color them according to the hydrophobicity, partial charge of the side chain, electron-donating character, or even the significance to shaping the tunnel. The amino acids are vertically sorted according to their impact on the tunnel – the more significant impact of a given amino acid on the tunnel boundary results in the higher position of this amino acid. A description of the algorithms used for this visualization can be found in Byska et al., 2016.

Figure 1: The graph representation of the main tunnel in the wild type DhaA. The representation shows the tunnel width evolution over time (top) and the residues lining the tunnel along its centerline, sorted according to the extent of their influence on the tunnel boundary (bottom).

The visualization clearly shows that the tunnel bottleneck starts to form in the distance of approximately 9 Å from the tunnel active site and it continues almost to the end of the tunnel. Using the vertical slider (red) we select the right most part of the tunnel bottleneck for the subsequent analysis. On the top panel, we select Tunnel – Contours which opens, once computed, the visualization of the tunnel bottleneck shape over time (Figure 2). The bars surrounding the central contours represent the residues forming the tunnel bottleneck – or any other section chosen with the vertical slider from the previous view. The size of the bars corresponds to the relative amount of time that a given residue was lining the bottleneck region. The colors of the bars correspond to the physico-chemical properties of the respective bottleneck residues: hydrophobicity and partial charge. The colors of the contour represent the time frame in which the contour was computed – black represents the first time frame, orange the last time frame. A detailed description of the visualization algorithms can be found in Byska et al., 2015.

Figure 2: Contour representation of the main tunnel’s bottleneck in the wild type DhaA. The tunnel bottleneck is clearly defined and wide open, resulting in the limited stability in the presence of organic co-solvents (Koudelakova et al., 2013).

This visualization clearly shows that the shape of the bottleneck of the main tunnel in the wild type is quite consistent and the tunnel is wide open over time. This enables the water solvent and organic co-solvent DMSO to easily penetrate through the tunnel, which has significant impact on the protein stability and its inactivation by the co-solvent molecules. This observation is easy to derive from this visualization without the detailed study of the time steps through the entire simulation. From this visualization, it is also obvious that several amino acids (W141, T148, F149, V172, C176, K175, V245, F144, and A145) play a significant role in forming the tunnel bottleneck at the selected tunnel section over time.

The goal of the proposed mutagenesis was to limit the number of solvent molecules flowing through the enzyme tunnel, which is mostly limited by its bottleneck region. The first mutant, DhaA80, carried the substitutions of three residues (at positions 148, 171, and 176), from which two (148 and 176) are among the ones (at the selected tunnel section) suggested by the contour visualization. The mutations were obtained by focused directed evolution and provided less defined geometry and narrower bottleneck, which improved the protein stability, increased the resistance to co-solvents, but also lowered the catalytic efficiency. We can easily see this after we load the trajectory to CAVER Analyst and compute both Tunnel Graph (Figure 3) and Contour (Figure 4) visualizations for the tunnel “tun_cl_009_1” in DhaA80. The 9th tunnel was selected because it spatially corresponds to the main tunnel in the wild type of DhaA. This can be easily verified using the 3D view after both molecules are aligned using the align tool under the menu Structure – Alignment.

Figure 3: The tunnel graph representation of the tunnel in DhaA80 which spatially corresponds to the main tunnel in the wild type of DhaA. The tunnel is less prominent as there are fewer time steps when the tunnel was present in the structure during simulation than in the wild type DhaA (i.e., it was closed most of the time).

Figure 4: Contour representation of the bottleneck in the mutant DhaA80. The tunnel bottleneck is less structured and shows the narrower diameter than in the wild type, leading to the improved stability at the expense of its activity (Koudelakova et al., 2013).

Both tunnel graph and contour visualizations demonstrate how much the bottleneck region of a tunnel can fluctuate in space and shape during the MD simulation, proving that the tunnels are very dynamic features with properties that can highly vary in time. In this case, the change in the geometry of the bottleneck proved to be too dramatic, in spite of improving the stability and resistance to DMSO it also prevented the ligands (substrates/products) to pass through the tunnel to and from the active site.

The tunnel graph representation (Figure 3) clearly shows that the amino acid F176 influences the tunnel the most and it forms its bottleneck as well which is clearly visible from Figure 4.

The saturation mutagenesis at this site (described in Liskova et al., 2015), followed by screening and biochemical characterization, revealed that the substitution F176G indeed increases the catalytic activity of DhaA106, while retaining its high stability and resistance to organic co-solvents. This is consistent with the observed in our visualizations after we load the trajectory to CAVER Analyst. This time we evaluate the tunnel “tun_cl_005_1” in DhaA106, as it spatially corresponds to the same tunnels previously analyzed in DhaA and DhaA80.

It can be seen that the mutation only slightly increases the bottleneck space (Figure 5 and Figure 6) which preserves the improved stability from DhaA80. But as can be seen in these visualizations (and also from the Tunnel Statistics panel), the tunnel is now open more frequently. The tunnel graph (Figure 5) also shows that the bottleneck size now tends to change less erratically over time which is depicted by the smooth color gradient in the tunnel profiles.

Figure 5: The tunnel graph representation of the tunnel in DhaA106 which spatially corresponds to the main tunnel in the wild type of DhaA. The tunnel is more stable than in DhaA80 and it is clearly visible that at the beginning of the simulation the bottleneck is much wider than at the end of the simulation.

Figure 6: Contour representation of the bottleneck in the mutant DhaA106. The tunnel bottleneck is more stable in comparison with DhaA80, providing the optimal combination of stability and activity (Liskova et al., 2015).

References

Byska, J. et al. (2015) MoleCollar and Tunnel Heat Map Visualization for Conveying Spatio-Temporo-Chemical Properties Across and Along Protein Voids. Computer Graphics Forum, 34(3), 1-10.

Byska, J. et al. (2016) AnimoAminoMiner: Exploration of Protein Tunnels and their Properties in Molecular Dynamics. IEEE Transactions on Visualization and Computer Graphics, 22(1), 747-756.

Koudelakova, T. et al. (2013) Engineering Enzyme Stability and Resistance to an Organic Cosolvent by Modification of Residues in the Access Tunnel. Angewandte Chemie International Edition, 52(7), 1959-1963.

Liskova, V. et al. (2015) Balancing the Stability-Activity Trade-Off by Fine-Tuning Dehalogenase Access Tunnels. ChemCatChem, 7(4), 648-659.

1