Extensible and Object-oriented System (Eos) Supplies a New Environment for Image Analysis of Electron Micrographs of Macromolecules

Takuo Yasunaga and Takeyuki Wakabayashi

Department of Physics, School of Science, University of Tokyo 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113, JAPAN

---- Short Running ----
Extensible system for macromolecular structure analysis

ABSTRACT

To study macromolecular structure by electron microscopy, a highly extensible and object-oriented system has been developed for image analysis. This system is named 'Eos' (Extensible and object-oriented system). The system described here supplies an environment with four types of supports: (i) a group of small tools for image analysis, (ii) tools for integration of small tools, such as 'Display2', (iii) tools for development, such as 'maketool', and (iv) object-oriented libraries for development of new tools. Using 'Eos', electron micrographs can be analysed by small tools and integration tools. Also, 'Eos' can be used to develop new tools based on new ideas because development tool and object-oriented libraries are provided. The examples of implemented small tools for image analysis include three-dimensional reconstruction of objects with helical symmetry, cluster analysis and contour expression.

INTRODUCTION

The function of macromolecules, such as protein and DNA, is attributable to their structure. X-ray crystallography and nuclear magnetic resonance (NMR) are useful to elucidate their structure at atomic level. In the field of macromolecular structure research, electron microscopy also has helped to elucidate structure at atomic level (Henderson et al., 1990) and single particle structure of macromolecules (Frank & Randermacher, 1992; van Heel & Frank, 1981). We have also studied the structure of actin filaments and thin filaments (Wakabayashi et al., 1975), complexes of actin filaments and myosin subfragment-1 (Toyoshima & Wakabayashi, 1985) and F1-ATPase structure (Fujiyama et al., 1990). To study the structure of individual molecules, electron microscopy is better than X-ray diffraction or NMR, because two-dimensional projections of specimens can be obtained directly. Electron microscopy is in the phase of technical development, and new ideas for image analysis should be tried to obtain new structural information. Therefore, we require an extensible environment where we can test new ideas for image analysis easily and quickly.

Some useful and powerful tools and packages such as FRODO (Jones, 1978; Jones, 1982), O (Jones, 1991) and X-plor (Bruger, 1992) have been supplied for the study of structures using X-ray diffraction and NMR. Many packages such as SPIDER (Frank et al., 1981), Tools for MRC (Moore et al., 1970; DeRosier and Moore, 1970; Wakabayashi et al., 1975; Amos and Klug, 1975), IMAGIC-5 (van Heel and Keegstra, 1981), SUPRIM (Stoops et al., 1991), and SEMPER (Saxton et al., 1979) have been also developed for electron microscopy. We have developed tools required to analyse images of electron micrographs and have recognised the importance of an environment suitable for developing new tools.

Some packages for general image analysis have been also supplied in the public domain. The 'Khoros' system (Konstatinides & Rasure, 1994) is one of the most powerful and extensible systems that can analyse multi-dimensional images. However, it is difficult to use it, because this system requires a considerable commitment of resources and expertise.

To overcome such difficulties, we have developed an extensible environment following the methods used by the 'Khoros' system to achieve extensibility. Because the essence of extensibility achieved by 'Khoros' is integration of small smart tools on 'cantata', a visual language, we constructed a group of small tools and additional tools with which to integrate these together. Furthermore, we constructed object-oriented libraries to improve extensibility and productivity.

Thus, the image analysis environment, which is named 'Eos' (Extensible object-oriented system), has been developed. At present, the system supplies four types of support to analyse images and to develop new tools: (i) a group of small tools for image analysis and three-dimensional reconstruction of objects with helical symmetry, (ii) integration tools such as 'Display2' and examples of 'makefile', (iii) tools for development such as 'maketool', and (iv) object-oriented libraries for development of new tools. We describe the 'Eos' development technique, how to use a group of image analysis tools, and how to develop new analysis tools.

MATERIALS AND METHODS

'Eos' system has been developed on UNIX workstations (HP9000/735, Hewlett Packard and Domain series, Apollo Inc.). 'Eos' was written in the ANSI-based C-language using an X-window interface. Originally, graphical tools such as 'Display2' have been developed on Apollo Domain series because useful two-dimensional and three-dimensional graphical libraries were available. X11R5, Xtoolkit, and Motif libraries must be installed on workstations to use graphical tools, because the tools have been ported to X-Window system (X11R5) with Motif widgets to achieve general portability.

Manuals are supplied as plain texts and HTML-formatted texts and are also available on a WWW server (html://tkyemghp.phys.s.u-tokyo.ac.jp/Eos). Plain and HTML manuals of each tool can be also obtained by using '-h' and '- html' options, respectively. Each of the tool manuals is almost completed, and general manuals are now being prepared.

RESULTS

'Eos' system concept

Figure 1 shows the concept of 'Eos', the environment for image analysis and tool development. 'Eos' supplies four types of support: (i) a group of small tools for image analysis such as three-dimensional reconstruction of objects with helical symmetry, (ii) tools for integration of small tools, such as 'Display2' and examples of 'makefile', (iii) tools for development such as 'maketool', and (iv) a group of objects, i.e., object-oriented libraries. The two former supports are used to analyse images of electron micrographs and two latter supports are used to develop new tools.

The group of small tools and integration tools

Using this group of small tools, 'Eos' users can analyse images of electron micrographs. 'Eos' supplies more than fifty small tools, such as histogram normalisation, median filter, low/high/band pass filtering, image correlation, Fourier and Fourier-Bessel transform, calculating angular distribution in Fourier space, printing images and contour maps, CTF correction. Cluster analysis using the 'distance' array and three-dimensional reconstruction of objects with helical symmetry can also be performed by combination of these small tools.

Figure 2 shows an example of a three-dimensional reconstruction of objects with helical symmetry using 'Eos'. Theories and fitting procedures for three-dimensional reconstruction were developed according to DeRosier and Moore (1970), Wakabayashi et al. (1975), and Amos and Klug (1975). A flow chart to obtain a three-dimensional image is shown in Fig. 2a. This indicates that the reconstruction is performed by a combination of small tools. The examples of 'Makefile' to integrate these small tools is also shown in Fig. 2b. Using this 'Makefile', the three-dimensional reconstruction of a filament in Fig. 3a is achieved by simply entering 'make' from a keyboard.

Figure 3 shows the result of image analysis. The three-dimensional image shown in Fig. 3c was reconstituted from the projected and low pass filtered image of an atomic model of actin filaments proposed by Lorenz et al. (1993). The tools of 'Eos' were used to read atomic coordinates from the Protein Data Bank, obtain the low pass filtered projected densities shown in Fig. 3a from atomic coordinates and reconstituted three-dimensional densities. In Fig. 3b the contour maps of two horizontal sections at different z-coordinates are shown. Because 'Eos' does not support volume rendering or solid modelling yet, the solid model shown in Fig. 3c was produced by 'xds' supplied by NCSA. To use such tools, many file-format converters such as mrc2hdf (mrcImage format to HDF by NCSA), mrc2viff (mrcImage format to Khoros format), mrc2tiff (mrcImage format to TIFF format), are provided.

Cluster analysis is an important technique in single particle analysis to determine orientation of particles (van Heel & Frank, 1981). Using 'Eos', we applied it to actin filaments. Phase residuals (Wakabayashi et al., 1975; Amos & Klug, 1975) were used as the 'distance' which provides the measurement of the difference between images. This approach was useful for structural analysis of actin filaments, because the change of subdomain arrangement could be detected. The tools for cluster analysis ("llFit", "clustShow", and "arrayClust") can be also integrated into a 'makefile'.

Both examples show that the integration of small tools by a UNIX 'make' tool is useful. Updated data can be obtained at the required time. Thus, we propose an idea for using UNIX 'make' to concatenate or pipeline small tools into one integrated tool, while 'make' is usually used to create executables from source codes.

Another integration tool supplied by 'Eos' is 'Display2'. This tool is suitable for interactive image analysis, e.g. to extract a region of interest. Image analysis techniques which need an interactive process can be integrated with 'Display2', because this tool has the ability to evoke further small tools.

Development tools and object-oriented libraries

'Eos' supplies development tools such as 'maketool' and object-oriented libraries to enable easy development and installation of new tools as shown in Fig. 1.

The 'maketool', which is a shell program, produces a special directory structure and prototypes of source codes. If a new tool named 'tool' is required to be developed, 'maketool' supplies prototypes of one header file (tool.h) and five source files (tool.c, argCheck.c, usage.c, init.c, util.c). Required values and information can be introduced into source codes interactively by using 'maketool'. The required information can be obtained from options of tools, a control file or both, whereas the way to input required data is described in 'Makefile' as shown in Fig. 2b. One can concentrate on the coding of "tool.c", because it is not necessary to modify most of the other source files such as "tool.h", "usage.c", "argCheck.c", or "init.c" if 'maketool' is used. This feature is also useful for documentation.

Furthermore, the development procedures of new tools are supported by object-oriented libraries. Tools can be modified or developed in a short time, because source-format is well-defined. The detailed method for developing a new tool is described in manuals.

Table 1 shows representative objects supported by 'Eos'. Objects are classified in three classes, i.e., 'General', 'DataManipulation', and 'DataExpression'. Tools for analysis have been created with combining these objects. An 'mrcImage' object is important, because it is tightly related to image data files of MRC format, i.e., CCP4 map format. Because both tools for electron microscopy by MRC and tools for X-ray crystallography by CCP4 can be applied to the files with MRC-format, an 'mrcImage' object is useful for comparing three- dimensional images reconstituted from electron micrographs with atomic structure obtained by X-ray crystallography.

DISCUSSION

The goal of this system, 'Eos', is to supply an extensible environment for the image analysis of macromolecular structure. The extensibility described here suggests both easy integration of small tools in analysing images and easy development of new tools. We realised or implemented this extensibility by supplying four types of supports as shown in Fig. 1.

'Eos' supplies examples of 'makefile' for integration of small tools as shown in Fig. 1 and 2b. We are modifying 'Display2' so as to integrate new small tools into 'Display2' using a control file to minimise necessary modification of source codes. When it is completed, 'Display2' will become more convenient.

'Eos' supplies prototypes of source codes using 'maketool' and object- oriented libraries as shown in Fig. 1. Easy and quick development of new tools enabled us to rapidly increase the number of supported tools for image analysis in our laboratory. In the future, modification of development tools such as 'maketool' will be improved by adopting a visual dialogue. Object-oriented libraries will be also improved by adding new functions and adopting the C++ language.

Acknowledgement

We thanks Drs K. Mitsuoka and S. Nagashima for useful discussions in developing "Eos" system. This work was supported by the grants: the Grant-in- Aid for Encouragement of Young Scientists and the grant from the JSPS Fellowships for Japanese Junior Scientists to T.Y. and the Grant-in-Aid for Specially Promoted Research from the Ministry of Education, Science and Culture of Japan to T.W. and the grant from the Biodesign Program, RIKEN to T.W.

REFERENCES

Amos, L.A. and A. Klug. (1975) Three-dimensional Image Reconstruction of the Contractile Tail of T4 Bacteriophage, J. Mol. Biol., 99, 51-73.2. Brunger, A.T. (1992) X-PLOR. Yale University Press, New Haven and London. DeRosier, D.J. and P.B. Moore. (1970) Reconstruction of Three-Dimensional Images from Electron Micrographs of Structures with Helical Symmetry, J. Mol. Biol., 52, 355-369. Frank, J., Shimkin, B., and Dowse, H. (1981) Spider - A Modular Software System for Electron Image Processing, Ultramicroscopy, 6, 343-358 Frank, J. and M. Randermacher. (1992) Three-dimensional Reconstruction of Single Particles Negatively Stained or in Vitreous Ice, Ultramicroscopy , 46, 241-262. Fujiyama, Y., K. Yokoyama, M. Yoshida, and T. Wakabayashi. (1990) Electron Microscopy of the Reconstituted Complexes of the F1-ATPase with Various Subunit Constitution Revealed the Location of the Gamma-subunit in the Central Cavity of the Molecule., FEBS Lett., 271, 111-115. Henderson, R., J.M. Baldwin, T.A. Ceska, C.F. Zemlin, E. Beckmann, and K.H. Downing. (1990) Model for the Structure of Bacteriorhodopsin Based on High- resolution Electron Cryo-microscopy, J. Mol. Biol., 213, 899-929. Jones, T.A. (1978) A Graphics Model Building and Refinement System for Macromolecules, J. Appl. Cryst., 11, 268-272.8. Jones, T.A. (1982) TOM: A Graphics Fitting Program for Macromolecules. In Computational Crystallography. D. Sayre, editor. Clarensdon Press, Oxford. 303- 317. Jones, T.A., Zou, J.Y., Cowan, S.W., and Kjeldgaard, M. (1991) Improved Methods for Building Protein Models in Electron Density Maps and the Location of Errors in These Models, Acta Crystalogr., A47, 110-119 Klug, A., F.H.C. Crick, and H.W. Wyckoff. (1958) Diffraction by Helical Structures, Acta Crystalogr., 11, 199-213. Konstantinides, K. and J. Rasure. (1994) The Khoros Software Development Environment For Image and Signal Processing, IEEE Trans. on Image Processing, 3, 243-252. Lorenz, M., D. Popp, and K.C. Holmes. (1993) Refinement of the F-actin Model against X-ray Fibre Diffraction Data by the Use of a directed mutation algorithm, J. Mol. Biol., 234, 826-836. Moore, P.B., Huxley, H.E., and DeRosier, D.J. (1970) Three-dimensional Reconstruction of F-Actin, Thin Filaments and Decorated Thin Filaments, J.Mol.Biol., 50, 279-295 Saxton, W.O., Pitt, T.J., and Horner, M. (1979) Digital Image Processing: The Semper System, Ultramicroscopy, 4, 343-354 Stoops, J.K., J. P. Schroeter, J. Bretaudiere, N.H. Olson, T.S. Baker, and D.K. Strickland. (1991) Structural Studies of Human a2-Macroglobulin: Concordance between Projected Vies Obtained by Negative-Stain and Cryo-electron Microscopy, J. Struct. Biol., 106, 172-178. Toyoshima, C. and T. Wakabayashi. (1985) Three-dimensional Image Analysis of the Complex of Thin Filaments and Myosin Molecules from Skeletal Muscle. V. Assignment of Actin in the Actin-Tropomyosin-Myosin Subfragment-1 Complex, J. Biochem., 97, 245-263. van Heel, M. and J. Frank. (1981) Use of Multivariate Statistics in Analysing the Images of Biological Macromolecules, Ultramicroscopy, 6, 187-194. van Heel, M. and Keegstra, W. (1981) IMAGIC: A Fast, Flexible and Friendly Image Analysis Software System, Ultramicroscopy, 7, 113-130 Wakabayashi, T., H.E. Huxley, L.A. Amos, and A. Klug. (1975) Three-dimensional Image Reconstruction of Actin-Tropomyosin Complex and Actin-Tropomyosin- Troponin T-Troponin I Complex, J. Mol. Biol., 93, 477-497.

Figure Legends

Table 1 : This table shows representative objects supplied by 'Eos' system. Three types of objects are shown. Figure 1: The concept of 'Eos' system and four types of supports by 'Eos'. Four shadow squares show four types of supports, i.e., "Group of Small Tools", "Integration Tools", "Development Tools", and "Group of Objects". When images are analysed, small tools are integrated using the integration tools. Image analysis techniques which need interactive processing can be integrated into 'Display2'. When a new tool is developed, a group of prototype source codes is supplied by a development tool, 'maketool'. Also, the new tool can be developed using a group of objects. The detailed explanation of "Group of Objects" is shown in Table 1. Figure 2: Three-dimensional reconstruction of filaments with helical symmetry as an example of 'Eos' usage. a. The schematic diagram of the procedure of three- dimensional reconstruction of specimens with helical symmetry. The left terms boxed with rectangles indicate 'Eos' objects and the right ones indicate image files with suffixes. Tools expressed in italic such as 'fft2d ' are supplied by 'Eos' system. b. An example of makefile for a 'make' tool. Using the UNIX 'make' tool, each of the small tools is integrated. Each of the data sets is assigned as an object of a particular type according to a suffix. By using this 'makefile', updated data can be obtained at the required time. Figure 3: The results of image analysis using three-dimensional reconstruction of specimens with helical symmetry. a. A projected image of an actin filament, which was produced from the atomic model of an actin filament (Lorenz et al., 1993). This image of a filament is treated as a model of micrographs recorded in an electron microscope so that the validity of the 'Eos' tools could be tested. b. Contour maps of horizontal sections at successive z-coordinates of a reconstructed image of a model actin filament. c. A reconstituted three-dimensional image of a model actin filament. A three-dimensional image (mrcImage) was transformed into NCSA-hdf format by 'mrc2hdf' and then the image was displayed with 'xds' supplied by NCSA.