The goal of the Allen Brain Atlas (ABA) is to create a detailed, cellular-resolution, genome-wide map of gene expression in the mouse brain. The complete sequencing of the mouse genome, and availability of techniques to probe gene expression amenable to scale-up and automation, made this an ambitious but achievable goal. To this end, the Allen Brain Atlas has created a platform for high-throughput in situ hybridization that allows a highly systematic approach to analyzing gene expression in the brain.
In situ hybridization (ISH) is a technique that allows the cellular localization of mRNA transcripts for specific genes. Labeled antisense probes, specific to a particular gene, are hybridized to cellular (sense) transcripts and subsequent detection of the bound probe produces specific labeling in those cells expressing the particular gene. Historically the predominant method of labeling has been to incorporate radioactively labeled nucleotides into the nucleic acid probe, allowing the detection of bound probe by film and/or emulsion autoradiography. More recently, non-isotopic methods have been developed that involve tagged nucleotides detected using immunocytochemical methods.
The platform used for the ABA utilizes this latter non-isotopic approach, with digoxigenin-labeled nucleotides(DIG) incorporated into a riboprobe produced by in vitro transcription. The project selected this non-isotopic methodology because it allows for visualization of cellular morphology to a greater extent than radioactive measures. This method produces a color precipitate that fills the cell body, in contrast to autoradiography that produces scattered silver grains surrounding each labeled cell. In addition, non-isotopic procedures are more environmentally friendly, a significant consideration for a project on the scale of the ABA. Traditionally, non-isotopic methods have been considered less sensitive than radioactive methods. To enhance the ability to detect low level RNA expression, the ABA has incorporated a tyramide signal amplification (TSA) step into the protocol that greatly increases sensitivity. The specific methodology is described in detail within the ABA Data Production Processes document.
Colorimetric detection of bound probe is generated by the alkaline phosphatase substrates nitroblue tetrazolium (NBT) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP) that produce a vivid blue/purple particulate reaction product. Unlike dark-field illumination of the silver grains produced by emulsion autoradiography following radioactive ISH, this product is somewhat difficult to see at low magnification. However, at higher magnification the label is easily visible and allows very clear cellular localization. It should be noted that there are certain reproducible types of background associated with this technique that are important to recognize when interpreting this data. Cell-dense areas such as the hippocampus tend to have a light blue label that is not particulate in nature and appears in negative controls (no probe added). A second type of background is a pink color that tends to label very cell-dense regions such as the granule cell layer of the cerebellum and to a lesser extent non-neuronal cells throughout the brain. The color difference between this pinkish color and the true purple particulate label make for an easy determination of signal from noise for this type of background. These issues of signal and noise are explored in more detail in the Cross-Platform Validation Experiments document, in which ABA data is compared to radioactive ISH data. Additionally, extensive quality control steps are taken throughout multiple stages of data preparation and analysis, to ensure consistent reagent quality and facilitate the generation of reliable ISH data. Examples of these control procedures are described in detail in the ISH Platform Controls document.
To produce a comprehensive analysis of gene expression, methodologies have been developed to allow even coverage for a given gene throughout the mouse brain. Essentially this involves cutting serial brain sections onto serial slides, and applying the same probe to evenly spaced slides from this series. This allows for the analysis of a particular gene on 25 µm sections every 100-200 µm t hroughout the entire brain. Early in the project each gene was processed in the coronal and sagittal plane of section. Expression data for genes demonstrating particularly compelling specificity of expression will be replicated at higher sampling density in the coronal plane of section that most neurobiologists and anatomists are familiar.
To facilitate the annotation of ABA ISH data, and to create a standardized 3D framework upon which to superimpose gene expression data, the project is generating detailed reference atlases of the mouse brain to accompany the ABA data. Three-dimensional alignment and registration of ISH data to a standardized atlas requires the atlas to be prepared in an identical fashion to the ISH data, namely 8-week old male C57BL/6J mice brains with unfixed, fresh-frozen tissue preparation. The reference atlases are available for viewing side-by-side and at the same plane of section as the ISH data. See the Coronal and Sagittal Allen Reference Atlases document for more details.
Following digital image acquisition the ISH data enters the informatics data pipeline (IDP), in which the images of raw ISH data are post-processed to facilitate visualization, signal quantification and binning by neuroanatomical expression patterns. ISH data for a given gene is aligned to the reference atlas through a three dimensional registration process that alleviates issues of different planes of section between brains. See Informatics Data Processing documentation for a detailed description of this process. This combination of anatomic localization and relative RNA expression level data is presented through the versatile web based ABA application, http://www.brain-map.org, that will serve as a valuable resource for the neuroscience community.