Portal de Periódicos da CAPES

Animating the model figure

2010; Elsevier BV; Volume: 20; Issue: 12 Linguagem: Inglês

10.1016/j.tcb.2010.08.005

1879-3088

Janet Iwasa,

Microtubule and mitosis dynamics

In all branches of scientific inquiry, researchers build models that enable them to visualize, formulate and communicate their hypotheses to others. In cell biology, our conceptual understanding of a process is typically embodied in a model figure. These visual models should ideally represent pre-existing knowledge of molecular interactions, movement, structure and localization but, in reality, they often fall short. Cell biologists have begun to look to the use of three-dimensional animation to visualize and describe complex molecular and cellular events. In addition to aiding teaching and communication, animation is emerging as a powerful tool for providing researchers with insight into the processes that they study. Two case studies focusing on the structure/function of the motor protein dynein and the structure of the centriole are discussed. In all branches of scientific inquiry, researchers build models that enable them to visualize, formulate and communicate their hypotheses to others. In cell biology, our conceptual understanding of a process is typically embodied in a model figure. These visual models should ideally represent pre-existing knowledge of molecular interactions, movement, structure and localization but, in reality, they often fall short. Cell biologists have begun to look to the use of three-dimensional animation to visualize and describe complex molecular and cellular events. In addition to aiding teaching and communication, animation is emerging as a powerful tool for providing researchers with insight into the processes that they study. Two case studies focusing on the structure/function of the motor protein dynein and the structure of the centriole are discussed. Over the past several years, there has been a steep increase in the use of animation to communicate dynamic molecular processes to a wide range of audiences. Biology students can view animations on numerous educational websites and in media packaged with their textbooks, and are increasingly presented with biological animations in classrooms and lecture halls. Studies in high school and graduate-level biology courses have shown that the use of animations in teaching has a positive impact; students who have viewed animations as part of their curriculum report a higher level of interest in the course material, and have shown greater memory retention and overall comprehension than students who did not view the animations [1McClean P. et al.Molecular and cellular biology animations: development and impact on student learning.Cell Biol. Educ. 2005; 4: 169-179Crossref Scopus (94) Google Scholar, 2Thatcher J.D. Computer animation and improved student comprehension of basic science concepts.J. Am. Osteopath. Assoc. 2006; 106: 9-14Google Scholar]. In the research arena, it is no longer unusual to be presented with an animation during a seminar. In many cases, animation is used to summarize and contextualize research on a specific molecular or cellular mechanism. The websites of an increasing number of journals feature animation that has been created by authors and animation is often used to present supplemental figures that can be downloaded or viewed online. These visualizations range in sophistication from basic line drawings to polished 3D-rendered animations, but all aim to provide the audience with information about a dynamic molecular process. At best, animations can be viewed as a visual molecular model that consolidates diverse data, from spatiotemporal information culled from light microscopy, structural data from sources such as X-ray crystallography and EM studies, and information about molecular interactions from biochemical and genetics assays. These visualizations can communicate a specific hypothesis for how a molecular process proceeds, and often can do so in a much more efficient and intuitive manner than a written description and with more accuracy and detail than a simplistic diagram or illustration. An example of this type of dynamic molecular model is shown in Figure 1. In collaboration with Tomas Kirchhausen (Harvard Medical School), I have created an animation that illustrates the process of clathrin-mediated endocytosis, focusing on the assembly and disassembly of the clathrin cage around a newly formed vesicle. A majority of the proteins shown in the animation are derived from crystal structures and the animation shows the progress of endocytosis in "real time" (based on light microscopy), such that the formation of the clathrin cage takes approximately one minute, and disassembly follows rapidly, spanning just a few seconds [3Kirchhausen T. Imaging endocytic clathrin structures in living cells.Trends Cell Biol. 2009; 19: 596-605Abstract Full Text Full Text PDF PubMed Scopus (150) Google Scholar]. Historically, physical 3D models of molecules have been used as thinking tools and have aided in scientific discovery (Box 1). In some cases, these models were created as an educational device, but were later brought into the laboratory and used to help researchers visualize and solve a problem. I believe that molecular animation will follow a similar trajectory, and that animations will increasingly become tools that enable thinking and discovery, in addition to aiding teaching and communication.Box 1The role of 3D models in past scientific discoveriesHistorically, physical models have provided researchers with insights that have led to important discoveries in chemistry and biology [12Chadarevian S.D. Hopwood N. Models: The Third Dimension of Science. Stanford University Press, 2004Google Scholar]. Some of the earliest molecular models were built by August Wilhelm von Hofmann in about 1860. These models, as illustrated in Figure I(a) , were essentially physical representations of chemical formulas, and showed the atoms arranged in a linear array. Hofmann's models were built for a public lecture on organic chemistry, and were used as an educational and communication tool. Chemists soon began to use models to show the 3D arrangement of atoms in space, including August Kekulé, who is credited for discovering the structure of benzene and benzene derivatives (b) and Jacobus van't Hoff, whose use of physical models is thought to have led to the earliest descriptions of stereochemistry [12Chadarevian S.D. Hopwood N. Models: The Third Dimension of Science. Stanford University Press, 2004Google Scholar].In biology, physical models have played an important role in major scientific findings. Linus Pauling was famously sick in bed with a cold when he realized the molecular structure of the alpha helix (Linus Pauling: A Centenary Exhibit, Oregon State University, http://pauling.library.oregonstate.edu/exhibit/). Bored, Pauling had drawn a polypeptide backbone on a sheet of paper and began to fold the sheet, eventually creating a helical structure (c). A similar "eureka" moment occurred for James Watson when pondering over the structure of DNA. Watson constructed a set of the four nucleic acid bases out of paper (d) and by manipulating the models he started to visualize how the bases might pair, giving him insight into the double-helical structure of DNA (DNA Interactive, produced by Dolan DNA Learning Center, Cold Spring Harbor Laboratory, 2003). Historically, physical models have provided researchers with insights that have led to important discoveries in chemistry and biology [12Chadarevian S.D. Hopwood N. Models: The Third Dimension of Science. Stanford University Press, 2004Google Scholar]. Some of the earliest molecular models were built by August Wilhelm von Hofmann in about 1860. These models, as illustrated in Figure I(a) , were essentially physical representations of chemical formulas, and showed the atoms arranged in a linear array. Hofmann's models were built for a public lecture on organic chemistry, and were used as an educational and communication tool. Chemists soon began to use models to show the 3D arrangement of atoms in space, including August Kekulé, who is credited for discovering the structure of benzene and benzene derivatives (b) and Jacobus van't Hoff, whose use of physical models is thought to have led to the earliest descriptions of stereochemistry [12Chadarevian S.D. Hopwood N. Models: The Third Dimension of Science. Stanford University Press, 2004Google Scholar]. In biology, physical models have played an important role in major scientific findings. Linus Pauling was famously sick in bed with a cold when he realized the molecular structure of the alpha helix (Linus Pauling: A Centenary Exhibit, Oregon State University, http://pauling.library.oregonstate.edu/exhibit/). Bored, Pauling had drawn a polypeptide backbone on a sheet of paper and began to fold the sheet, eventually creating a helical structure (c). A similar "eureka" moment occurred for James Watson when pondering over the structure of DNA. Watson constructed a set of the four nucleic acid bases out of paper (d) and by manipulating the models he started to visualize how the bases might pair, giving him insight into the double-helical structure of DNA (DNA Interactive, produced by Dolan DNA Learning Center, Cold Spring Harbor Laboratory, 2003). Cell biologists often employ a model figure when presenting a hypothesis about a cellular process. The model figure, which is generally presented in the discussion section of a research article, serves many potential purposes: to describe the current understanding of a molecular process; to give context to the study; to illustrate a specific hypothesis that the authors espouse; and to demonstrate how the current research has added to or altered a pre-existing model. While the main purpose of creating a model figure is typically to communicate a hypothesis to a wide audience, the process of constructing the model figure is often illuminative. When we start to draw a model figure of a process that has only been imagined, we must go through an active and creative mental exercise of concretely defining the properties of the model. What are the shapes, sizes and numbers of the main players? How do they interact with each other? As other scientific illustrators and photographers have noted (and I have observed), this process of illustrating a model, particularly in collaboration with others, can provide clarity as well as give rise to new ideas [4Frankel F. The power of the 'pretty picture'.Nat. Mater. 2004; 3: 417-419Crossref Scopus (10) Google Scholar, 5Frankel F. Reid R. Distilling meaning from data.Nature. 2008; 455: 30Crossref Scopus (124) Google Scholar, 6Goodsell D.S. Johnson G.T. Filling in the gaps: artistic license in education and outreach.PLoS Biol. 2007; 5: 2759-2762Crossref Scopus (31) Google Scholar]. While pencil and paper can be an ideal way to start visualizing the model figure, it is important to go through a reiterative process in which the model is further and further refined. Too often, model figures are oversimplified and give the false impression of knowing less about a process than we actually do. As we gain a greater understanding about the players in a cellular process (including their identity, dynamics, localization, density, shapes, sizes and stoichiometry), it is important that the model figures we create reflect this knowledge. Building more accurate molecular models is important in communication and in formulating better hypotheses. Researchers often rely on visual representation when considering new areas of investigation and planning experiments. However, using overly simplified schematics that are unable to convey complex information can cause us to miss key information and to design flawed experiments. Accuracy and detail are needed when considering molecular shape, size and number, and for conveying change over time. In many cases, a dynamic model or illustration is the best way to effectively explore and communicate a dynamic process. A model can be distilled into a series of stills that, in some cases, can more readily communicate a complex process to audiences than a moving visual. The use of animation often comes into the picture at the point where researchers are interested in communicating an essentially completed story. However, the process of creating and refining a dynamic visual model of a molecular mechanism can lead to important insights that might have an important role in earlier stages of the research process. Even in cases where little is known about a molecular process, creating a dynamic 3D model, especially one that can be modified in a reiterative manner as new information is gained, can lead to important insights that would be hard to achieve by other means. It is important to note that these animations are not necessarily meant to act as a consensus model to depict a mechanism that has received a stamp of approval by a majority of others in the field, or even by others in a research group. Rather, these visualizations are intended to show an individual's hypothesis of how a process might occur. This hypothesis can include aspects that are not yet supported by experimentation, and the visualization can act as a thinking and communication tool that might give insight into how this hypothesis would best be explored and tested experimentally. Researchers are frequently concerned that an animation showing a hypothetical model will bias viewers into believing that the model is "real" and well grounded in experimental evidence, which can be far from the case. Creating an animation usually requires including a level of detail that is not often included in a typical model figure, and visually communicating the fact that one or more aspects of the animation are purely conjecture, or only partially supported, is difficult to do; however, there are actions that can be taken that might help communicate ambiguity. When the research community supports a discrete number of competing models, a series of animations that show each of these models can be used to highlight the differences between the models. It is possible to visually communicate confidence in a model using different rendering styles (Box 2). Ultimately, however, it is up to the researcher presenting the model to communicate the level to which a model is supported by experimental evidence, and this remains true regardless of whether the model shown is a simple drawing or a sophisticated 3D animation.Box 2Communicating confidence through rendering styleWhen communicating a molecular or cellular process through animation, different rendering styles can help explain the level of confidence and the amount of experimental data that support a specific model. It is important to be selective about how and when animation is used and to be aware of how a visualization is perceived by the audience. A process that is poorly understood will be less misleading to the audience when shown as a simple 2D sketch rather than a 3D animation, while conversely, a well elucidated molecular mechanism will be best served by a visualization where detailed information in time and space are clearly conveyed. Animation software enables animators to transition easily between different rendering styles; however, animators are sometimes tempted to render animations in eye-catching styles that can unfortunately result in distraction from the key message or even miscommunication.Cartoon-style shading techniques, such as that shown in Figure I(a), can be used to illustrate a mechanism that is poorly understood, or to give a broad overview of a process. In (a), cellular components are shown in simple, colorful shapes and key proteins are exaggerated in scale to emphasize their role. There has been no attempt to show detailed protein or cellular structures, indicating that the viewer should focus on the overall process, rather than on individual proteins.Protein structures can be rendered in a number of styles, such as a surface rendering (b), and backbone (c). Rendering proteins as a surface can be done at different levels of smoothness and detail. In general, greater levels of surface detail indicate a higher emphasis on molecular structure and a higher confidence in how the model depicts protein conformations and protein–protein interactions. Regions of proteins that have not been crystallized are shown as a smooth cylinder in (b), indicating a lower level of confidence in how these domains are depicted than other domains whose structures are known.Illustrating the backbone of a protein structure can be highly informative in showing specific conformational changes (c). In general, this technique should be used only when the structures of the different conformations are available, and there is high confidence in how the protein structure moves between the conformations. In (c), the backbone is rendered as a thick cylinder, which gives a space-filling view of the proteins while still providing visual information about protein structure. When communicating a molecular or cellular process through animation, different rendering styles can help explain the level of confidence and the amount of experimental data that support a specific model. It is important to be selective about how and when animation is used and to be aware of how a visualization is perceived by the audience. A process that is poorly understood will be less misleading to the audience when shown as a simple 2D sketch rather than a 3D animation, while conversely, a well elucidated molecular mechanism will be best served by a visualization where detailed information in time and space are clearly conveyed. Animation software enables animators to transition easily between different rendering styles; however, animators are sometimes tempted to render animations in eye-catching styles that can unfortunately result in distraction from the key message or even miscommunication. Cartoon-style shading techniques, such as that shown in Figure I(a), can be used to illustrate a mechanism that is poorly understood, or to give a broad overview of a process. In (a), cellular components are shown in simple, colorful shapes and key proteins are exaggerated in scale to emphasize their role. There has been no attempt to show detailed protein or cellular structures, indicating that the viewer should focus on the overall process, rather than on individual proteins. Protein structures can be rendered in a number of styles, such as a surface rendering (b), and backbone (c). Rendering proteins as a surface can be done at different levels of smoothness and detail. In general, greater levels of surface detail indicate a higher emphasis on molecular structure and a higher confidence in how the model depicts protein conformations and protein–protein interactions. Regions of proteins that have not been crystallized are shown as a smooth cylinder in (b), indicating a lower level of confidence in how these domains are depicted than other domains whose structures are known. Illustrating the backbone of a protein structure can be highly informative in showing specific conformational changes (c). In general, this technique should be used only when the structures of the different conformations are available, and there is high confidence in how the protein structure moves between the conformations. In (c), the backbone is rendered as a thick cylinder, which gives a space-filling view of the proteins while still providing visual information about protein structure. The use of 3D molecular models and animations can provide important insights into the structure/function relationship of a protein or protein complex. This is especially true of classes of molecules, such as motor proteins, where understanding a molecular mechanism is dependent on the ability to visualize movements in space and time. Animations of motor proteins, such as kinesin, have been useful for describing a complex molecular process to broad audiences (http://valelab.ucsf.edu). Molecular models that can be animated might have a utility far beyond the visual demonstration of a well-supported mechanism. In instances where the mechanism and structure of a motor protein are not well understood, an animatable model can play an important role in the early stages of formulating and visualizing different hypotheses. Cytoplasmic dynein is a motor protein that transports vesicles and other cargo along microtubules. Cryo-EM, crystallography and sequence analysis have given researchers important clues about the structure of dynein and the dynamics of its movement, but the mechanism by which dynein is capable of walking processively along microtubules remains an area of active investigation. Working together with the Reck-Peterson group at Harvard Medical School, I have created an articulated model of dynein that can be viewed and manipulated in three dimensions (Figure 2C-D) . This 3D, animatable model was conceptualized after observing group members attempting to use a variety of 2D illustrations, physical models and hand gestures to describe and visualize their hypotheses regarding dynein structure and motion (Figure 2A-B). We hypothesized that laboratory members would benefit from having a structurally accurate 3D model that could be readily manipulated and updated as new information is gained. The model features a structure derived from several different sources; the microtubule-binding domain is from a published crystal structure [7Carter A.P. et al.Structure and functional role of dynein's microtubule-binding domain.Science. 2008; 322: 1691-1695Crossref PubMed Scopus (205) Google Scholar], the AAA domain and coiled-coil region are derived from crystal structures of homologous structures, and the linker domain is a simple flexible cylinder whose length is approximated from estimates taken from EM studies [8Roberts A.J. et al.AAA+ ring and linker swing mechanism in the dynein motor.Cell. 2009; 136: 485-495Abstract Full Text Full Text PDF PubMed Scopus (178) Google Scholar]. In addition, the model includes a microtubule (modeled to scale with dynein) that is highlighted at locations where the microtubule-binding domain of dynein has been shown to interact with the microtubule [7Carter A.P. et al.Structure and functional role of dynein's microtubule-binding domain.Science. 2008; 322: 1691-1695Crossref PubMed Scopus (205) Google Scholar]. Several manipulation handles are available, allowing the user to move and rotate each of the two dynein motor domains separately, or to bend and move the linker domains. The movement of the model is restricted to regions that have been observed to undergo conformational changes, and the degree of rotational freedom can be fixed to reflect experimental data. The model includes a built-in measuring tool that shows the distance between two points that can be moved in 3D space. The earliest articulated dynein model featured highly detailed molecular surfaces (Figure 2C) that were later deemed to be distracting and potentially misleading to audiences, who might believe that the model was based on a full crystal structure. A later version featured softer, smooth surfaces (Figure 2D) that better communicated the hypothetical nature of the dynein model. One major limitation of the articulated dynein model is that it can be viewed and manipulated only from within a commercial animation software application (Autodesk Maya). The model is being used by group members to visualize and explore the structure of dynein, and it has been instructive in revealing the conformations of dynein that are possible or impossible given structural and steric constraints. The ability of dynein to take steps of different lengths and to switch between protofilaments can be visualized by manipulating the model. Eventually, the dynein model will be used to create an animation that shows a proposed mechanism of how dynein walks on microtubules. A major hurdle in the widespread adoption of 3D molecular models is the requirement for highly specialized and often unintuitive software to create, view and manipulate the model. An increasing number of programs have started to support 3D visualization, including some programs that are commonly found on laboratory computers. Popular vector-drawing programs, such as Adobe Illustrator, have begun to support drawing and rendering of objects in 3D. It often comes as a surprise to researchers that even the ubiquitous Portable Document Format, or PDF, is currently 3D-capable. Since 2004, Adobe Acrobat Reader, a popular PDF viewer, has been able to display 3D models that are embedded into a PDF document. By clicking and dragging on a 3D image in the PDF, a user can rotate and manipulate the embedded model, as well as choose between different rendering and lighting styles. Recently, some research journals have started to publish articles that take advantage of 3D PDF technology [9Goodman A.A. et al.A role for self-gravity at multiple length scales in the process of star formation.Nature. 2009; 457: 63-66Crossref Scopus (109) Google Scholar], but its use is currently far from widespread. The 3D PDF format works well for viewing static structures of surface-rendered molecules, and might prove to be particularly useful in displaying cryo-EM and crystal structures in publications. The 3D PDFs can be used to view and manipulate 3D models created in animation software applications, where different models can be embedded side-by-side in a single document and easily compared. A good example of the use of a 3D PDF in a research setting comes from a collaboration with Tomer Avidor-Reiss at Harvard Medical School, whose group is interested in understanding the structure of the centriole. The centriole is a cylindrically shaped organelle exhibiting 9-fold symmetry. EM imaging has revealed the existence of a hollow central tubule with radiating spokes, a structure that has been termed the "cartwheel", at the center of the centriole. The Avidor-Reiss laboratory found that SAS-6, a conserved centriole protein, oligomerizes and appears to form the central tubule [10Gopalakrishnan J. et al.Self-assembling SAS-6 multimer is a core centriole building block.J. Biol. Chem. 2010; 285: 8759-8770Crossref Scopus (35) Google Scholar]. They proposed two structurally distinct models to describe how SAS-6 might form a tubule shape with 9-fold symmetry: by forming stacked rings, or by forming a helix. I have created two simple 3D models of the centriole showing these two possibilities, and a 3D PDF showing the two models next to one another (Figure 3a-b) . In addition to allowing users to intuitively zoom and rotate the models, the 3D PDF comes with a cross-section tool that gives users the ability to slice through a model with a semitransparent plane that can be adjusted for height and angle. The ability to manipulate and visualize a 3D model in this way was especially appropriate for researchers who, like the members of the Avidor-Reiss laboratory, study thin-section EM images. By creating thin-slice models in the 3D PDF, the researchers could clearly see how they might differentiate between the helical and stacked ring models. From a top view, an appropriately thin section reveals only a portion of the central tubule in the case of the helical model, while a similarly sliced section reveals the full central tubule in the case of the stacked ring model (Figure 3c-d). The use of 3D modeling and animation software can certainly benefit researchers by providing a means to visualize, explore and communicate ideas to others. There are, however, numerous challenges that currently make it difficult for researchers to adopt animation as a new research and communication tool for themselves. One major issue is the steep learning curve presented by commercial animation software, even for those who are technically inclined, and many hours of dedicated training are often required for even basic proficiency. Cost, both for training and for the software itself, can be prohibitive. While creating a 3D model of a protein or protein complex alone is unlikely to require a great deal of time for an experienced animator, using that model to create an animation can take weeks to months or more, depending on the complexity of the animation. Many of the difficulties described above stem from the fact that there is no 3D animation application that has been designed with molecular biologists in mind. Professional scientific animators, for the most part, rely on commercial software packages that have been created for large animation studios employed in the entertainment industry, and must adapt to using modeling and animation tools that are often less than ideal for creating molecular environments [11McGill G. Molecular movies… coming to a lecture near you.Cell. 2008; 133: 1127-1132Abstract Full Text Full Text PDF Scopus (39) Google Scholar]. However, online tutorials, custom scripts, toolkits and plug-ins created by a growing community of molecular animators have helped lower the threshold for those who are just beginning to explore animation (Box 3).Box 3Online animation resourcesThere is a growing number of online resources for those seeking to learn more about molecular and cellular animation. A few of these are listed below.•http://molecularmovies.orgThe Molecular Movies website, created by Gael McGill, is an excellent resource that features a large collection of animations organized by subject, tutorials on molecular animation and a plugin for Maya (the Molecular Maya Toolkit)•http://science.nichd.nih.gov/confluence/display/bvigThe homepage for the NIH Biological Visualization Interest Group, which hosts regular meetings to discuss molecular visualization technologies and to promote community building•http://www.grahamj.com/plugins-tutorials.phpThis site features molecular visualization plugins for Maxon Cinema4D that were built by Graham Johnson•http://iwasa.hms.harvard.eduThe website for the Cell Biology Visualization group at Harvard Medical School features animation projects and a tutorial on creating 3D PDFs There is a growing number of online resources for those seeking to learn more about molecular and cellular animation. A few of these are listed below.•http://molecularmovies.orgThe Molecular Movies website, created by Gael McGill, is an excellent resource that features a large collection of animations organized by subject, tutorials on molecular animation and a plugin for Maya (the Molecular Maya Toolkit)•http://science.nichd.nih.gov/confluence/display/bvigThe homepage for the NIH Biological Visualization Interest Group, which hosts regular meetings to discuss molecular visualization technologies and to promote community building•http://www.grahamj.com/plugins-tutorials.phpThis site features molecular visualization plugins for Maxon Cinema4D that were built by Graham Johnson•http://iwasa.hms.harvard.eduThe website for the Cell Biology Visualization group at Harvard Medical School features animation projects and a tutorial on creating 3D PDFs Due to the increase in visibility of high-quality molecular animations, more researchers are beginning to consider using animation to communicate and teach scientific concepts. But how do we make it easier for researchers to take advantage of 3D animation and modeling tools? A long-term solution might be to create animation software specially designed for the molecular and cellular research community, an effort that is likely to require years of effort from a large team of software developers. A more tractable solution is to build upon existing molecular visualization software (such as UCSF Chimera or PyMOL) or a 3D animation application. The latter approach is currently being undertaken for Autodesk Maya, a popular commercial animation platform (Box 3). Creating effective molecular models and animations is far from being simply an issue of having the right software, however. As cell biologists, we often rely heavily on some variant of a model figure to convey a complex and dynamic molecular/cellular process, and expect that this figure should effectively communicate a hypothesis to diverse audiences. In my experience, there is little to no emphasis in training programs on how to create an effective model figure. It is crucial that we start to invest in training programs and collaborations that support a new generation of scientists who are visual thinkers and communicators. This type of training should include an appreciation of the different types of modes of visualization, such as animation and 2D illustration, and an understanding of what types of visualization will best serve a particular purpose. Ultimately, creating better visualization will allow us to better articulate our hypotheses to others and ourselves, thereby enabling us to become more effective teachers, mentors and researchers. I thank Tomas Kirchhausen for collaborating on the clathrin animation project, Samara Reck-Peterson and members of her laboratory for collaborating on the dynein articulated model project, and Tomer Avidor-Reiss and Jay Gopalakrishnan for collaborating on the 3D PDF project. My work at Harvard Medical School is supported, in part, by the Giovanni Armenise Harvard Foundation.