An on-line archive of reconstructed hippocampal neurons

https://doi.org/10.1016/S0165-0270(98)00091-0Get rights and content

Abstract

We have developed an on-line archive of neuronal geometry to encourage the use of realistic dendritic structures in morphometry and for neuronal modeling, located at web address www.neuro.soton.ac.uk. Initially we have included full three-dimensional representations of 87 neurons from the hippocampus, obtained following intracellular staining with biocytin and reconstruction using Neurolucida. The archive system includes a structure editor for correcting any departures from valid branching geometry and which allows simple errors in the digitisation to be corrected. The editor employs a platform-independent file format which enforces the constraints that there should be no isolated branches and no closed loops. It also incorporates software for interconversion between the archive format and those used by various neuronal reconstruction and modelling packages. The raw data from digitisation software can be included in the archive as well as edited reconstructions and any further information available. Cross-referenced tables and indexes are updated automatically and are sorted according to a number of fields including the cell type, contributor, submission date and published reference. Both the archive and the structure editor should facilitate the quantitative use of full three-dimensional reconstructions of neurons from the hippocampus and other brain regions.

Introduction

The three dimensional structure of neurons is of interest both as a means of cell classification and for understanding the complex influence of dendrites on neuronal properties (Uylings et al., 1986, Major et al., 1994, Spruston et al., 1994, Ishizuka et al., 1995, Mainen et al., 1996). Various studies (Li et al., 1994, Pyapali and Turner, 1994, Spruston et al., 1993, Migliore et al., 1995, Turner et al., 1995, Zador et al., 1995, Pyapali and Turner, 1996, Pyapali et al., 1998) have examined the variation of structure within neuronal populations under different conditions in the hopes of determining differences which may correlate with observed single cell and network behaviour. Realistic neuronal structures have also been used in modelling studies of single-cell behaviour (Turner, 1984a, Turner, 1984b, Turner et al., 1991, Claiborne et al., 1992, Spruston et al., 1993, Migliore et al., 1995, Henze et al., 1996) and it has been shown that understanding some features of neuronal behaviour demands a detailed quantitative description of a neuron's dendritic arborization. The use of such detailed information is particularly important for understanding integrative properties of dendrites (Yuste and Tank, 1996, Lipowsky et al., 1997) and the complex interactions between cells subserved by axonal arborizations.

Quantitative data on neuronal structure are, however, relatively costly to obtain, requiring many hours of either manual tracing or computer-based reconstructions to digitise a single complex cell. The most detailed visualization technique for individual neurons requires some form of intracellular dye staining, usually performed during physiological recording. This technique offers the advantages of allowing access to structure/function relationships on a single cell level, of achieving rapid transport of the dye throughout the living cell for dense staining and of staining only a single cell in the tissue, thereby enhancing reconstructions.

It is therefore particularly important to maximise the use made of these detailed data, in the form of an open database which is easily accessible and with cells available in common formats. The availability is particularly important since no such databases currently exist for open dissemination of quantitative neuronal information and obtaining neuronal structure information in particular can be very difficult, beyond an occasional single neuron. In addition, most neuronal modeling requires assessing average results over several neurons within a class, to decrease the effect of random variations in dendritic architecture. We have developed the present archive system to assist in making such data readily available in a tested and convenient format for use in morphometric and statistical analysis and in neuronal modelling. The archive is both easily extendible and readily accessible to many different computer platforms. It presently contains a first series of 87 cells from published datasets which were all obtained using a Neurolucida three-dimensional reconstruction system (Pyapali and Turner, 1994, Turner et al., 1995, Pyapali and Turner, 1996, Mott et al., 1997, Pyapali et al., 1998; Microbrightfield, Colchester, VT). In this report we describe the content and structure of the morphology archive in order to facilitate the use of these data.

Section snippets

Data structures

The archive is located at www.neuro.soton.ac.uk and includes data both in the native format as supplied from the digitisation software, and in a simpler, three-dimensional standardised format (given the extension `SWC' in the archive) which enforces certain elementary geometrical constraints. The SWC files contain an optional header which may be of any length, but which should contain certain key fields as listed in Table 1 for the automated cross referencing to work correctly. These data are

Structure editor

The structure editor is written in java for the convenience of platform-independence and use across the internet. At present it can read files in Neurolucida (ascii), Genesis (Bower et al., 1994) or `SWC' format (the format used in the archive for indexing as described above). Trivial errors of connectivity are corrected on loading files, so, for example, where a new branch starts at the same, or almost the same, coordinates as a point already defined in the structure, but without an explicit

Archive indexing and submissions

Rather than using a database system and associated search methods, the most important aspects of the archive are visual and include the cell geometry. All the archiving is done with hypertext indexes and thumbnail pictures of the cells. The processing is fully automated, requiring one file, in the SWC format, per cell. Each time the archive is reindexed, various images are made of each cell, parameters calculated and html pages created. The final entry for one cell is shown in Fig. 1. This

Examples

Fig. 2 shows the main window of the morphology editor. Data may be read locally or from a web site. In the latter case, the site should contain a text file `list.html' detailing all the cells available. The name of the site can be entered in the top bar. Available cells will be shown in the menu on the right. The viewing style is selected with the menu on the second row. The first menu selects between three-dimensional rotation of the whole cell or echoing with a cube only. The second gives the

Conclusion

Although neurons are routinely studied physiologically, there remain many uncertainties in the relationship between cellular function and structure of complex neurons. These include at the most basic level the passive electrotonic structure of the cell which is largely determined by its geometrical structure, to which this archive is devoted. The single cell anatomical data may be modified to include dendritic spines, a common structural feature of pyramidal cells, as well as voltage-dependent

Acknowledgements

This work was supported by the Wellcome Trust (H.V. Wheal and R.C. Cannon) and by grants from the NIH (AG13792 to D.A. Turner) and Veterans Affairs (Merit Review to D.A. Turner).

References (22)

  • N. Ishizuka et al.

    A quantitative analysis of the dendritic organization of the pyramidal cells in the rat hippocampus

    J Comp Neurol

    (1995)
  • Cited by (0)

    View full text