Detailed contents of the original LIPS image dataset were previously reported [12]. Here, we focus on how to use the LIPS database website and the new additional content. LIPS database website visitors can find image and/or literature data by entering a query word (e.g., microtubule) in the search box on the home page. To obtain a LIPS image dataset, users should visit the “Images” page (http://hasezawa.ib.k.u-tokyo.ac.jp/lips/images.html; Figure 1A). There, users will see links to the three types of image datasets, as describe below.
Dataset I. Original serial optical sections
Dataset I is a collection of original optical sections of 18 kinds of fluorescent markers, which are listed in our previous report [12]. Users can select organelles using the tiled thumbnails (Figure 1B) and visit the user-choice page (Figure 1C). On this page, users can download the raw serial optical section 16-bit TIFF images by clicking on the thumbnails. To browse the downloaded images, we recommend using an image visualization tool such as ImageJ, which is freely available. For an overview of image visualization tools, please refer to the review by Walter et al. [1]. Representative examples of LIPS images are shown in Figure 2A. This dataset contains raw images that can be processed and analyzed with appropriate image analytic tools [1]. For example, three-dimensional models can be reconstructed from the raw serial sections with ImageJ plugins 3D Viewer (in ImageJ menu Plugins-3D-3D Viewer; Figure 2B, C).
Dataset II. Volume rendering data
Dataset I is a raw serial optical-section dataset that can be used to reconstruct three-dimensional models with appropriate image processing software. However, such three-dimensional reconstruction requires skill and labor. To overcome this problem, Dataset II allows easy viewing of the three-dimensional organizations of the target intracellular structures. The volume rendering data (Figure 2B) from all 930 pairs of guard cells are available. The original images are 180 frames that play back at 10 frames/sec. Users can easily examine the 360°-rotation GIF animations by simply clicking on the link in the web browser, without specific image visualization tools. We believe that this feature will be a valuable aid to education.
To more efficiently search and browse the images in LIPS datasets I and II, we prepared a database interface, named LIPService (Figure 2D). Users can visit the LIPService website from the “Images” page (http://hasezawa.ib.k.u-tokyo.ac.jp/lips/images.html). On this page, users can choose the “Filter” items of interest such as “Primary target” and “Stomatal aperture” from the dropdown list (Figure 2D, arrow). When users specify the filter condition, matching thumbnail images and links to download image datasets I and II appear (Figure 2D). We believe that the LIPService will be useful for image data mining, especially in studying the relationship between intracellular configuration and morphometrical parameters, such as stomatal pore width (aperture) or length.
Dataset III. Aligned images for localization analyses
Dataset III is a collection of aligned maximum intensity projections of the fluorescent serial optical sections. The step-by-step image processing with ImageJ used to make dataset III was described in our recent protocol paper [13]. All guard cell regions were aligned to a mean size of 304 × 119 pixels (19.5 × 7.6 μm). The fluorescent intensities were also normalized to an average of 0 with a standard deviation of 1. On the “Images” page, users can download dataset III as a ZIP file (52 MiB) containing 1,860 32-bit TIFF files (100–120 examples × 18 probes) with probe-name tags in the file names. Part of dataset III is presented as a tile in Figure 1D.
With dataset III, users can inspect their own fluorescence images of guard cells. To demonstrate this utility, we performed image clustering analysis of images obtained from another database, the Plant Organelle Database (http://podb.nibb.ac.jp/Organellome/) [6], using guard cell images of nuclei labeled by GFP fusions with a nuclear localization sequence (NLS-GFP; Additional file 1: Figure S1A), mitochondria labeled by DsRed fusions with the pre-sequence of the delta-prime subunit of mitochondrial F1-ATPase (F1-ATPase-δ-DsRed; Additional file 1: Figure S1B), and chloroplasts labeled by GFP fusions with CAS, a thylakoid membrane-localized protein (CAS-GFP) [14] (Additional file 1: Figure S1C). After preprocessing (Additional file 1: Figure S1D-H), image clustering analysis was performed using the freely-available image clustering software iCluster (http://icluster.imb.uq.edu.au/) [15] (Figure 3, Additional file 2: Movie S1) and the 55 selected LIPS images (5 examples × 11 probes) that users can download as a ZIP file (860 KiB). iCluster gathered images of the same organelles (Figure 3), including nuclei (NLS-GFP vs. HistoneH2B-RFP; Additional file 1: Figure S2A), mitochondria (F1-ATPase-δ-DsRed vs. Mt-GFP; Additional file 1: Figure S2B), and chloroplasts (CAS-GFP vs. autofluorescence; Additional file 1: Figure S2C). These data show the potential usefulness of this dataset for users’ on-demand localization analyses.