Portal de Periódicos da CAPES

Novel Relational Database for Tissue Microarray Analysis

2003; American Medical Association; Volume: 127; Issue: 4 Linguagem: Inglês

10.5858/2003-127-0492-nrdftm

1543-2165

Rita Shaknovich, Ashlyn Celestine, Lin Yang, Giorgio Cattoretti,

Molecular Biology Techniques and Applications

Tissue microarray (TMA) analysis is becoming broadly accepted as an efficient and expeditious method in the field of proteomics.1–3 It allows simultaneous analysis of hundreds of samples or clinical cases on a single slide by placing small (<1 mm in diameter) cores of fixed, embedded tissue in a paraffin block in arrays. The limits of the technique are dictated by the size of the slide, the diameter of the cores, and the distance between the cores. The technology is available and user-friendly, allowing an easy and quick creation of TMAs and even potentially making this technology applicable for diagnostic studies. In our experience, the rate-limiting step for TMA is data acquisition and manipulation.45 Image recording, scoring, and analysis have to be conveniently integrated in order to successfully use this technology. At this time, Beecher Instruments (Silver Spring, Md), maker of the most popular TMA instrument, and image analysis companies providing TMA services are not offering integrated data management packages that would suit the average TMA user.By using Microsoft Excel (Microsoft Corporation, Seattle, Wash) and Adobe Photoshop 5.5 (Adobe Systems Incorporated, San Jose, Calif), which are PC and Macintosh compatible, we created a relational database4 able to provide integrated image viewing, analysis, and scoring. This arrangement allows easy navigation between multiple folders, viewing images, scoring, and entering data in the same spreadsheet. Microsoft Excel is a widely used and user-friendly program that does not require any programming skills. It allows creation of a large database with up to 65.536 rows and 256 columns. Results of chip analysis of 15 000 genes from tumor cases and corresponding clinical data are presented in Microsoft Excel format and are available online, for example, on the National Institutes of Health Web site (http://llmpp.nih.gov/lymphoma) for lymphomas.The first step in construction of a TMA is selection of cases from a database and creation of the template or spreadsheet, which identifies the position of each case and controls in the TMA block. Figure 1 demonstrates the stepwise progression from the linear database (Figure 1, number 1) to the layout of cases identical to the TMA (Figure 1, number 2), and finally to the simplified color-coded version with the substitution of the case number by the position identifier in the TMA (Figure 1, number 3). During viewing and photographing of the slide in Figure 1 (see number 4), we refer to cores (cases) by position. Please note that the example shown in Figure 1, featuring 9 specimen and 27 cores, is a simplified version of TMA for the purpose and ease of demonstration. The TMA in our study contains 150 cores.Each case in the database is identified by unique position(s) of each core on the template, for example, A1 or B5 or K15; the letter designates the column in the template and the number designates the position in the row. Triplicate cores have been suggested as the gold standard.6 Ideally, adjacent cases in the database will lie together in each row in the block, thus facilitating identification.The next step is image acquisition. We used Adobe Photoshop 5.5 to acquire images from 3-μm-thick hematoxylin-eosin–stained TMAs, as well as various immunohistochemical and immunofluorescence stains. For the image acquisition, each core in the TMA is identified based on the row and column position on the slide. We used a Spot-2 CCD camera (Diagnostic Instruments Incorporated, Sterling Heights, Mich) mounted on a Nikon Eclipse E800 microscope (Nikon Inc, Melville, NY) and linked to a G4 Apple computer (Apple Computers, Cupertino, Calif). One image per core is taken and saved as a JPEG compressed file, and then named as the position identifier (A1.jpg, A2.jpg, A3.jpg, etc). Each image acquisition takes less than 1 minute, including field selection, manual focusing, and identification.A set of images for every stain is saved in a unique folder with a unique folder name that is explanatory for the stain (eg, Folder Name: Vimentin/Folder Content: A1.jpg through J15.jpg). Each image measures 1315 × 1033 pixels (1.6 Mpixels) at 72 dpi for 3.9 MB, JPEG compressed with high resolution to a 1.8-MB file.We used a ×20 objective for hematoxylin-eosin and immunostains, a ×40 objective for close-up and immunofluorescence, and ×4 for low-power images of the whole core. We found that under optimal focus and light conditions, Adobe Photoshop allows easy virtual zooming of the image up to 200% or more, while retaining excellent resolution, mimicking changing microscope objective during real microscopy. In addition, 3-color immunofluorescence microphotographs can be split in separate wavelengths, and 1- or 2-color combinations examined at 1 time (eg, DAPI nuclear counterstain plus Bcl-6–fluorescein isothiocyanate).The third step is linking the images with the database. We used a hyperlink function to link each cell to an image [=HYPERLINK(“StainX/A2.psd”,)], where StainX is the name of the folder and A2.psd is the name of the image in that particular cell. Selecting a hyperlinked cell in Excel automatically opens up the corresponding image in Adobe Photoshop. Creation of these links can be automated by Excel programming (see detailed instruction at http://ICG.cpmc.columbia.edu/cattoretti/Protocol/Immunohistochemistry/ScaleUp.html).Once the accuracy of the correspondence of the links with each case is checked, an unlimited number of hyperlinked stain columns can be added by simply copying and pasting an existing column and changing the Folder address with the Find-Replace function. Macintosh users need to link to images defined with the .psd suffix in order to open them in Adobe Photoshop for Mac. Instructions can be found at http://ICG.cpmc.columbia.edu/cattoretti/Protocol/Immunohistochemistry/TissueArray.html.A more detailed description of the Excel program can be found in the program instructions and is not reported here. The folders with the images and the relational database can be allocated on the same hard disk (in our case an 80-GB FireWire Maxtor device, Maxtor Corporation, Milpitas, Calif) or on CD-ROMs. An example of a relational database is illustrated in Figure 2. As we continue to acquire images (Figure 2, numbers 4 through 6), we expand the database by adding the columns under the name of the stain next to the case number (Figure 2, number 1) and the core identifier (Figure 2, number 2). Each position in the column is hyperlinked to the corresponding image with the specific address. Clicking on the position (Figure 2, number 3) results in opening of the corresponding image.The last step is data entry. Several images from different stains can be simultaneously viewed on the screen, scored, and the data can be entered manually into additional cells, adjacent to the image cell in the spreadsheet. Additional standard program features can be used to verify such functions as spelling and duplications.The cases and scores within the Excel spreadsheet can be manipulated and sorted at will as a normal database and additional data (biological, clinical, etc) are added. Data can be exported for statistical analysis.This method allows examination of multiple cores and multiple stains (including fluorescence) of the very same case with ease. This flexibility would be impossible at the microscope, requiring changing slides, stains, light sources, and identifying the correct core.Scoring can be checked by independent investigators, and images can be printed or shared over a network. In addition, the collection of images becomes an educational resource, easily accessible to multiple users.We are currently using this method to evaluate antigen expression for more than 30 markers on TMAs containing 80 diffuse large cell lymphomas and 20 control cases.