Project 1: The Mouse Brain Library Project 2: Internet Microscopy (iScope) Project 3: Neurocartographer and Segmentation of the MBL Project 4: The Neurogenetics Tool Box
















Principal Investigator/Program Director Williams, Robert W.



During the course of this proposal, we will extend the breadth and depth of the MBL to reflect the increase in subjects, types and numbers of images and through-focus series (Project 2), and information concerning regional segmentation of the brain (Project 3). We will provide the tools with which individuals can measure and analyze images, either working over the Internet or by downloading customized datasets onto their own computers. In addition, the MBL will be the key source for CNS phenotypes that are such a critical part of the gene mapping analysis system described in Project 4.

1. Extend the MBL

We will extend the depth and breadth of the MBL on three fronts. First, we will increase the numbers of mice in the MBL by supplementing the current collections and by adding inbred strains, RI strains, and advanced intercrosses. Second, we will introduce brains that have been processed with stains other than what is currently in the MBL (i.e., Nissl stains). Finally, we will increase the range of images that will be stored in the MBL.

Additional Brains

Every brain that is added to the MBL originates in the Genotyping and Mouse Colony Core (Core B). The animals are perfused and the brains are shipped to the Neurohistology Core (Core A), where the brains are processed. For detailed methodologies, please see the respective Core descriptions.

Completing the RI strains. The current roster of RI strains in the MBL (BXD, AXB, BXA, and BXH) can be seen in Table 1. In the first year of the project, we will increase the number of animals representing each strain to at least 12 in order to improve the statistical power of subsequent analysis.

Advanced Intercrosses. As detailed in the description of the Genotyping and Mouse Colony Core (Core B), Dr. Williams and colleagues have been undertaking a large project to generate an advanced intercross between C57BL/6 and DBA/2 (the parental strains for the BXD RI strain). They currently have over 1400 tenth-generation (G10) mice from this B6D2 intercross. During years 1 and 2 of the project, we will add at least 600 of these mice to the MBL. These mice range in age from P46 to P76. In year 3 we will add a minimum of 200 additional aged animals (2 years old and up). All G10 mice will be fully genotyped at 350 markers.

An additional 600 G10 animals will be held for celloidin or immunohistochemical processing (see below) in years 3 and 4 or will be used for whole brain dissection in years 4 and 5.

Complete Standard Inbred Strains. As shown in Table 1, the MBL already includes 20 inbred strains. As with the RI strains above, we will increase the number of animals representing each strain to at least 12. In addition to our standard horizontal and coronal planes of section, we will include some sagittally processed sections as well.

Additional RI sets. We will add more RI sets to the MBL. We are currently planning on adding the AKXD series (AKR/J x DBA/2J) and CXB. The Genotyping and Animal Facilities Core (Core B) has already produced the complete CXB set.

Potential Problems and Pitfalls. We are aiming to process a large number of animals over the course of the proposal, and a question can be fairly raised as to whether this is overly ambitious. In the 2 years from the inception of the MBL, we have processed over 600 brains with no funds directly earmarked for this purpose. The histological processing was performed by the equivalent of a half-time technician. Moreover, for each brain, 2 one-in-five series were generated (one for Rosen and one for Williams) whereas only one series will be used by all members of the program project. Given the level of staffing in the Neurohistology Core and in this project, we have every reason to expect that we can comfortably process the numbers of mice proposed.

Questions can also be raised as to our choice of subjects in the MBL. For example, we are not planning to include mutant or knockout mice. Instead, we are heavily weighted toward RI strains and the advanced intercross G10 progeny. This choice was made because our focus is on the source of normal variation in the CNS. This important question has never been systematically addressed using quantitative methods. That being said, the infrastructure would be in place for the addition of knockout and mutant mice should funding become available.

Additional Stains

The purpose of adding more stains to the MBL is to improve our ability to determine borders of neuroanatomical regions in the CNS. This is especially important for Project 3 (NeuroCartographer), in which we will use additional stains to enhance our ability to properly segment the tissue.

Celloidin-embedded tissue. Every brain currently in the MBL is cut in celloidin and stained for Nissl substance—a stain particularly well suited for determining neuroanatomical architecture and for performing cell counts. As detailed in the description of the Neurohistology Core, a number of other histological stains can be applied to tissue processed in this matter. During the course of the present proposal, we will begin using alternative stains on sections adjacent to those stained for Nissl substance in order to investigate their utility as additions to the MBL. Of the choices available, myelin staining (using the Loyez method) is likely to improve the ability to distinguish a variety of subcortical structures, including the striatum and thalamus. We therefore view the staining of adjacent series of sections to be an aid for the neuroanatomical analysis conducted on the Nissl-stained tissue.

Immunohistochemistry. While staining for Nissl substance provides an excellent overall picture of brain anatomy, other, more specialized stains can improve the delineation of certain neuroanatomical borders. Choline acetyltransferase (ChAT), for example, stains the ascending cholinergic system of the forebrain, and the distribution of ChAT-positive fibers and cells is regionally distinct. We will use this stain on 4 brains from each of the RI strains currently in the MBL (BXD, AXB, BXA) as well as 12 mice from the tenth-generation advanced intercross. We will also investigate the possibility of using other immunohistochemical stains (parvalbumin) and histochemical stains (AChE) as aids.

Potential Problems and Pitfalls. The potential technical problems associated with processing celloidin-embedded tissue and Immunohistochemistry are discussed in the description of the Neurohistology Core.


In addition to the 25 µm/pixel and .5 µm/pixel resolution images, we will also acquire images a 1 µm/pixel resolution for the purpose of better visualizing individual neuroanatomical ROIs. Of greater importance to the design of the MBL in the future is the integration of these static images with the through-focus QuickTime movies being generated by the iScope (Project 2) as well as with the segmentation vectors provided by NeuroCartographer (Project 3). Because this integration requires precise alignment of images, global fiducial coordinates must be established which will require unique methodologies. In the sections below, we detail the procedure by which images will be acquired for the MBL.

A microscope stage fitted with LED sensors as described in Project 3 will be used to manipulate the slides for imaging. The sensors will determine an absolute point of origin for the slide by computing the intersection of the left and bottom sides of the slide. All other points on the slide will be described on an X,Y coordinate system with this point as the origin. As the slide is moved into position for photography, its travel along this X,Y coordinate space is recorded by electronic digital length gauges. As each image is captured, the coordinates of the bottom left and top right of the image are recorded. Each image therefore has two fiducial marks embedded with it, which, along with the point of origin, will allow proper alignment of images by triangulation. By establishing these fiducial marks, we can safely rotate and align individual images to allow for ease of measurement without losing the ability to link that image to through-focus series (Project 2) or information concerning segmentation (Project 3).

Several types of images will be generated. Most will be contrast-optimized 8-bit gray-scale images that have been compressed and saved as high-quality JPEGs (quality level 8 or 9). At the highest resolution, each compressed image requires about 0.6 to 1 megabytes of storage space. In each case, we will image the entire slide (25 µm/pixel), each section individually on the slide (.5 µm/pixel), and then 40-50 ROIs a 1 µm/pixel. These regions will include, but will not be limited to, olfactory bulb, caudate/putamen complex, nucleus accumbens, basal forebrain, hypothalamus, septum, globus pallidus, amygdala, lateral geniculate nucleus, medial geniculate nucleus, ventrobasal and ventrolateral nucleus, anterior thalamic nuclei, lateral dorsal nucleus, posterior nuclei, hippocampal subregions, red nucleus, deep mesencephalic nuclei, inferior and superior colliculi, pontine nuclei, periacqueductal gray, and various cerebellar regions.

All images in the MBL have been and will continue to be imaged with a Micro Nikkor 60 mm camera lens with the exception of the 1 µm/pixel images, which will be captured using a Nikon microscope with a 0.5´ objective. Images are digitized using a Kodak DCS560 digital camera in 16-bit gray scale mode. The CCD has a pixel count of 3060 x 2036. These digital images are subsequently transferred to Adobe PhotoShop for sharpening and rotation. During the first year of the proposal, we will be considering the implementation other imaging technologies, such as FlashPix <>, which may allow greater image depth and quality (see below).

The current collection of MBL gray-scale images has not been acquired with optical density standards. We have strived to optimize the range of contrast in every section of the collection. Thus faintly stained sections have been contrast enhanced to more closely match well-stained sections. In general, optical density values for Nissl-stained specimens are not of much analytic use (but see work of Zilles et al. 1980, 1985). However, calibrating the collection is relatively simple and will be of use to some image analysts, ourselves included. During slide photography sessions we will photograph a linear gradient neutral density filter (1 OD/cm from Edmund Scientific) in both axes. During batch processing of images, but prior to any image manipulation, we will superimpose 5-pixel-wide bands taken from the photograph of the linear gradient neutral density filter along the lower and right edges of the digital image. Color image calibration can be carried out using the simple LED device described by Beach and Duling . Video camera calibration can also be carried out using a straightforward method that employs a single optical density filter described by Baldock and Poole .

We will also adjust for shading corrections for the light source using a relatively simple technique. We will remove the slide or section and photograph the "uniform" light source, then divide each pixel value in the image by that in the empty field and multiply the product by ~250. This can be done automatically in batch mode by any of a number of programs, including PhotoShop 5.5, NIH Image, Image-Pro Plus (v 3.0 from Media Cybernetics), IP Spectrum, and IDL.

Potential Problems and Pitfalls. There are several potential difficulties in acquiring these images for the database. It is essential that static images located in the MBL be of exceptional quality, and we believe that the equipment being used will provide sharp images with good tonal range. Further, the ability to sharpen, change contrast, and otherwise adjust the image digitally in Adobe PhotoShop will ensure high quality.

In addition to image quality, the inclusion of fiducial points with each image is essential, as discussed above. It is important that these fiducial marks be transferable between the projects where slides will be manipulated. All projects will have identical stages made, and we will each have identical calibration slides. Before any imaging is to occur on a given day, the stage will be calibrated using this slide.

The use and construction of this specialized stage is an innovative aspect of the current proposal. Dr. Nissanov has a working prototype of this stage already operational. While we are confident that we will be able to generate fiducial coordinates for all slides and sections, we are considering several rapid methods to create global series of coordinates. One possibility is to create a grid overlay to fix to the bottom of each slide before imaging. Thus, each image would be taken with grid marks visible within the plane of focus.

Finally, our ability to process the amount of images can be called into question. On the average, it takes approximately 1 minute to acquire each digital image (although this may be improved with the faster Macintosh G4 machine). It will therefore take approximately 20–30 minutes to produce the 25 µm/pixel and .5 µm/pixel image for a brain cut in the horizontal plane. A coronally sliced brain will take approximately 45 minutes. To acquire the 40–50 1 µm images will take another 75 minutes. We therefore conservatively estimate that a half-time imaging technician could process 3 slides/day for a total of approximately 600 slides/year. If that throughput is not enough, we will increase that technician’s time on the imaging workstation





2. Changes to the Underlying Data Structure