Journal of microscop

Journal of Microscopy, Vol. 231, Pt 1 2008, pp. 134–143

Received 5 June 2007; accepted 31 January 2008

Knife-edge scanning microscopy for imaging and reconstructionof three-dimensional anatomical structures of the mouse brain

D. M AY E R I C H ∗, L . A B B O T T† & B . M c C O R M I C K ∗∗Department of Computer Science, Texas A&M University, TX 77843, U.S.A.

†Department of Veterinary Integrative Biosciences, Texas A&M University, TX 77843, U.S.A.

Key words. High-throughput, microvessel, optical microscopy, serialsectioning, three dimensional.

Summary

Anatomical information at the cellular level is importantin many fields, including organ systems development,computational biology and informatics. Creating data sets atresolutions that provide enough detail to reconstruct cellularstructures across tissue volumes from 1 to 100 mm3 has provento be difficult and time-consuming. In this paper, we describea new method for staining and imaging large volumes oftissue at sub-micron resolutions. Serial sections are cut usingan automated ultra-microtome, whereas concurrently eachsection is imaged through a light microscope with a high-speed line-scan camera. This technique, knife-edge scanningmicroscopy, allows us to view and record large volumes of tissuein a relatively small amount of time (approximately 7 mm2 s−1).

The resolution and scanning speed of knife-edge scanningmicroscopy provides a new method for imaging tissueat sufficient resolution to reconstruct maps of cellulardistribution and morphology. We show that these techniquespreserve the alignment of serial sections accurately enoughto allow for reconstruction of neuronal processes andmicrovasculature. Expanding these techniques to other tissuesopens up the possibility of creating fully reconstructed cellularmaps of entire organs.

Introduction

The study of spatial distribution, morphology and inter-connection of cells in the nervous system has been a primaryfocus since the beginning of neuroscience. The study ofneural connections started with the development of stainingtechniques by Camillo Golgi, who used them initially to studyinter-relationships of neurons and glia in the olfactory bulb(Golgi, 1875) and cerebellar cortex (Golgi, 1873, 1874).The staining methods developed by Golgi were also used

Correspondence to: D. Mayerich. Tel: (979) 255-1571; fax: (979) 847-8578;

e-mail: [email protected]

by Santiago Ramon y Cajal, who surmized that the visible‘mesh-like’ arrangement of fibres were actually complexnetworks of independent inter-connected neurons (Cajal,1911).

Reconstructing and modelling the three-dimensional (3D)anatomical structure of individual cells in situ is vital tounderstanding their function in larger units. To create detailed,3D cellular maps of organisms, we require methods forcreating volumetric data sets from blocks of tissue or, ideallywhole organs. These data sets should also be of sufficientresolution to allow cells and their processes to be identified andtraced. Reconstructing thin cellular processes also requiresthat individual sections be accurately aligned and in properregistration.

There are several well-understood methods for achievingvolumetric data sets at the cellular level. Ultra-microtomyand immunohistochemical labelling have given rise to thereconstruction of gene expression in the mouse brain(Reymond et al., 2002). Although methods have beendeveloped to account for issues that arise in section alignmentand warping (Ju et al., 2003), relatively thick sections (10–20 μm) are required to sample the data set in order to completethe process within reasonable time constraints. Confocal lightmicroscopy is a well-known method for extracting high-resolution details from tissues stained with fluorescent dyes(Lacey, 1989; Herman & Jacobson, 1990; Handbook of ConfocalMicroscopy, 1995; Spector & Goldman, 2006). These methodscan be used to construct a volumetric data set (Agard, 1984);however, the presence of backscattered light limits these datasets to sampling that takes place on or very near the specimensurface. In addition, the point spread function along the opticalaxis limits the resolution of optical sectioning. Multi-photonmicroscopy (Denk et al., 1990) allows deeper sectioning butthe problem of still being limited to sampling relatively near thetissue surface remains. Extension of multi-photon microscopyto depths exceeding 1 mm requires photo-ablation (Tsai et al.,2003). The problem of limited optical resolution along theimaging axis remains.

C© 2008 The AuthorsJournal compilation C© 2008 The Royal Microscopical Society

K N I F E - E D G E S C A N N I N G M I C RO S C O P Y 1 3 5

The use of serial block face scanning electron microscopy(SBF-SEM) (Denk & Horstmann, 2004) for gathering 3Dvolumetric data at sub-cellular resolution provides a methodfor reconstructing small blocks of tissue. Using an in-chamberultra-microtome, serial sections (less than 50-nm-thick) areremoved, allowing the block face to be scanned repeatedly bythe SEM. Although this sequential sectioning and scanningproduces an aligned data set at sufficient resolution to followeven the smallest diameter cellular processes and resolvecellular organelles, the time required to resolve a single imageprohibit it from being used for large volumes of tissue, such asa whole mouse brain.

In this paper, we describe knife-edge scanning microscopy(KESM), a novel technique using light microscopy for dataset acquisition that allows us to section and image blocks oftissue at the cellular level of detail quickly and automatically.Our imaging techniques allow us to cut and image embeddedspecimens at a rate of approximately one section every 3 s.These sections are sampled at sub-micron intervals in all threespatial dimensions, maximizing the spatial resolution to allow3D reconstruction of the entire tissue sampled. The use of serialsectioning allows us to overcome limits in optical resolutionalong the imaging axis as well as removes the need for compleximage processing, such as deconvolution (Wiener, 1964).

This technology provides a means of extracting full-scalecellular microstructure from large tissue samples. We havetested KESM using several en bloc staining techniques that canbe used to answer many open questions in neuroscience. Theseinclude creating full-scale soma maps of the brain, exploringthe connectivity of brain microvasculature, and gatheringlarge statistical samples of neuron morphology.

Methods

Our primary goal was to develop methods for the acquisitionof high-resolution volumetric data sets from biological tissuesand organs. To this end, we developed and are using the knife-edge scanning microscope (KESM). The KESM was designed toact simultaneously as both a microtome and a microscope,to allow sections as thin as 0.5 μm to be cut from tissueembedded in plastic. We constructed the instrument froma stacked series of mechanical stages, a knife–collimatorassembly, a microscope and a high-sensitivity line-scan camera(Fig. 1). Embedded tissue was sectioned using a diamondknife. Concurrent with cutting the tissue, each section wasscanned through the microscope and the subsequent imageswere compressed and stored for additional processing to yield3D reconstructions.

KESM design

The most distinctive feature of our design was the knife–collimator assembly. This assembly provides illumination aswell as the means to cut individual sections of tissue (Fig. 2).

Fig. 1. Knife-edge scanning microscope. The KESM is constructed withthree Aerotech high-precision stages (A), a diamond knife and illuminator(B), modified microscope (C) and high-sensitivity line-scan camera (D).

Fig. 2. Objective, knife and specimen photograph (top) and cartoon(bottom) illustrating light transport through the diamond knife and intothe objective for imaging.

C© 2008 The AuthorsJournal compilation C© 2008 The Royal Microscopical Society, Journal of Microscopy, 231, 134–143

1 3 6 D. M AY E R I C H , L . A B B O T T A N D B . M c C O R M I C K

The knife–collimator assembly consists of a removablediamond knife module rigidly mounted to a granite bridge suchthat the top knife facet is 45◦ to the vertical and perpendicularto the optical axis of the microscope. When perfectly aligned,the top facet of the knife and the specimen plane of themicroscope coincide. The collimator assembly was slightlyadjustable, allowing rotation around two axes for precisionorientation of the knife.

The motion of the tissue block under the stationary diamondknife was controlled by a series of mechanical and air-bearing stages, allowing movement in all three dimensions.The objective was positioned such that the edge of the knifebisected the objective field-of-view. To improve the resolutionof our images, all cutting was performed under water and usedwater-immersion objectives.

To maintain the highest possible bandwidth for imaging,specific hardware was devoted to image capture. The high-sensitivity camera sent data directly to a camera server. Thedigital camera worked in conjunction with two frame-grabberboards. The twin boards worked concurrently to send datafrom the camera into a 3-GB frame buffer stored in the server’smain memory.

KESM construction

The specimen stage is constructed by stacking three differentstages to provide movement along all three axes. The stageswere constructed and assembled by Aerotech Inc. accordingto our specifications. Movement along the cutting axis (x-axis)is performed by an ABL20030 air-bearing stage. This stageprovides smooth motion but will oscillate slightly around afixed position. Because of this, all imaging is performed whenthe stage is moving at a constant velocity. Movement along theedge of the knife (y-axis) is performed by an ALS130 lineardrive stage and movement up and down (z-axis or depth) isprovided by an AVL112 vertical lift stage. Both of these latterstages are mechanical, not air-bearing, stages and remain veryrigid at a fixed position. All imaging is performed while theseaxes are stationary so the rigidity is required. Both the x- andy-axes have a feedback resolution of 20 nm and the z-axis has afeedback resolution of 25 nm. All feedback error is well belowthe maximum resolution of the objective, and therefore anymisalignment will be too small to be visualized in the images.

Imaging was performed using a DALSA CT-F3 high-sensitivity line-scan camera (Waterloo, Ontario, Canada). Themaximum speed of the camera (and thus, the limit of our dataacquisition rate) is 44 kHz. Since the camera provides a line of4096 eight-bt pixels, our maximum data rate is 180 MB s−1.Camera triggering is controlled by either an internal clock oran external Transistor–Transistor-Logic (TTL) trigger signal.In our case, the TTL signal was sent from the x-axis stagecontroller. This allowed us to key camera imaging to the stageposition and gave us two advantages over using an internaltrigger. First, the sampling rate could easily be set based onstage position. Second, basing camera firing on stage position

made image capture independent of stage velocity. Therefore,any small variance in stage velocity did not adversely affectimage quality.

Image capture from the camera was supported by twoPIXCI D3X frame grabber cards (EPIX Inc.). Both cards arerequired in order to support the maximum data rate for thecamera. To capture and store images, we used the applicationprogramming interface (API) provided with the PIXCI cards.This software also allows some preliminary image processing(e.g. formatting) and storage. All stage motion and cameracontrol applications were programmed in C++, allowingautomatic control and image capture. The PIXCI API was usedto format the images as TIFF files and store them on disk afterevery section.

Magnification of the tissue was done with a modifiedNikon E600FN brightfield microscope (Nikon InstrumentsInc., Melville, NY, USA). Two objectives were used to acquirethe data sets described in this paper. Nissl-stained tissue wasimagedusingaNikonFluor10×objectivewitha0.3numericalaperture (NA). Golgi tissue was imaged using a Nikon Fluor40× objective with a 0.8 NA. Both objectives are designedfor water immersion and have a working distance of 2.0 mm.The NA of each of these objectives allowed us to sample at600- and 300-nm intervals, respectively (Born & Wolf, 1999).A 2.5× coupler was used to match the intermediate field ofview of the microscope to the CCD sensor of the line-scancamera.

En bloc staining

For Nissl staining, mice were deeply anaesthetized usingketamine and xylazine, and then perfused transcardially using50 mL of room temperature phosphate-buffered saline (pH7.4), followed by 250 mL of cold 4% phosphate-bufferedparaformaldehyde (pH 7.4). The mice were then perfused with100 mL of a solution of 0.1% thionin dye in deionized water,and the bodies were placed in the refrigerator (4◦C) for 24 h.This is a novel way to introduce stain into the whole mousebrain, since most staining occurs on sectioned tissue. Perfusionof the thionin stain speeds up the staining process, which waspreviously accomplished in our laboratory entirely throughdiffusion, taking two or more months to complete. In this case,after 24 h the brains were then removed from the calvaria andplaced in a fresh solution of 0.1% thionin and left at 4◦C for7 days. The brains were destained and dehydrated through agraded series of ethanols starting with 50% ethanol and waterand increasing to 100% ethanol over a time period of 6 weeks.After three changes of acetone (2–4 days in each solution),the brains were then embedded in araldite plastic following astandard protocol (Abbott & Sotelo, 2000), with the exceptionthat each step needed to infiltrate the brains with aralditetook 24 h. KESM sectioning requires that the whole brainsbe completely dehydrated and infiltrated with araldite plastic.Normal plastic embedding is typically carried out on muchsmaller pieces of tissue, so we have modified the processing



steps to allow us to embed whole mouse brains that we can cutusing the KESM.

For Golgi–Cox staining of individual neurons, mice weredeeply anaesthetized using isoflurane inhalant anaesthesiaand then decapitated. The brains were quickly removedand placed into a Golgi–Cox fixation solution containing1% potassium chromate, 1% potassium dichromate and 1%mercuric chloride in deionized water. The brains were leftin this solution, in the dark, at room temperature for 10–16 weeks. The brains were then rinsed in deionized waterovernight in the dark, and placed in a 5% ammoniumhydroxide solution in deionized water for 7–10 days, in thedark and at room temperature. The brains were then rinsedagain in deionized water at room temperature for 4 h andthen dehydrated through a graded series of alcohols andembedded in araldite using the same protocol and timesdescribed previously for Nissl-stained mouse brains.

Results

In this paper, we demonstrate the use of KESM to sectionand image three data sets. Since each data set can consist

of thousands of sections, we provide examples of individualsections as well as images showing cross sections of each 3Ddata set. The first data set is mouse brainstem and spinalcord that was stained en bloc with the Nissl stain, thionin.This data set was cut into 1-μm-thick sections and the tissuesections were imaged at a 0.6-μm resolution (Figs 3 and 4).The sampling resolution used by the KESM is large enough inall three dimensions to allow full reconstruction of cell soma inthe brain. In addition, Nissl-stained tissue can be convenientlycounterstained to explore the relationship of cell soma to other,more specific structures.

We provide an example of this in the second data set, wherewe followed a similar Nissl-staining process and introducedIndia ink in order to enhance the contrast of the vasculature.We then sectioned the entire mouse olfactory bulb (Figs 5and 6). A video displaying a series of several hundreddownsampled sections as well as a video showing severalorthogonal sections through the data set at a higher resolutionis available as supplementary information (Supplementaryvideo 1 and 2). Again, the sampling resolution is high enoughto image even the smallest capillaries that make up the brainmicrovasculature. Analysis of these images allows us to study

Fig. 3. Coronal section of mouse brain stem. The complete section is shown (left) with two close-up inserts (right). The overall resolution of the image is0.6 μm pixel−1 and the section thickness is 1 μm.



Fig. 4. Re-sectioning of the data set constructed using sections like thosein Fig. 3. A large cross section of the data set (mouse brain stem and spinalcord) is shown (top) as well as a close-up (bottom). The section thickness(resolution along the vertical axis) is 1 μm. The labels indicate the pialsurface (PS), central canal (∗) and a blood vessel (V).

the relationship between brain microvasculature and the cellsin the olfactory bulb.

Finally, we stained a mouse brain en bloc using a slightmodification of the Golgi–Cox method (see Methods) andscanned several thousand sections of olfactory bulb and cortexto demonstrate fine alignment of sections and the abilityto segment fine fibres (Figs 7 and 8). For the Golgi-stainedbrain, we used 0.5-μm-thick sections and imaged with a 40×objective, which provided a sample resolution of 0.3 μm. Evenwhen using a high-resolution objective, Golgi-stained tissuecan be difficult to resolve using standard techniques becauseof the low resolution along the imaging axis. When thinsections are taken, each section provides less information onits own; however, we are able to extract more detailed spatialinformation for the fibre in three dimensions.

Given a 3D data set, standard techniques exist for surfaceextraction and visualization. The alignment of consecutivesections is such that contours can be created using softwarepackages such as Amira for visualization of vasculature andneurons (Fig. 9). Generally, some processing is required in orderto remove or reduce lighting artefacts; however, these result inlittletonodatalossandcanbeappliedastheimagesarestored.

Discussion

Imaging of biological tissues and organs at a cellular level is anemergent technology, which promises to elucidate mammalianbrain anatomy in unprecedented detail and also has thepotential to elucidate cellular organization in other organsboth normal and those with pathology related to a wide rangeof disorders. We have shown how our techniques can helpto reveal the density, morphology and inter-connectedness ofneurons in the rodent brain. Our KESM instrument makespossible the 3D microscopy of large biological specimens thesize of an entire adult mouse brain. The instrument hasbeen designed to volume digitize a specimen (e.g. a plastic-embedded tissue block) at a maximum sampling resolution of160 nm in the image plane at rates up to 200 M pixels s−1.Depth resolution in this case is typically 500 nm using a40× objective, and 1000 nm using a 10× objective. Ourinitial results on brain tissue show the feasibility of the KESMmethod to gather large, high-quality light-microscope data setsof whole tissue and organ samples. The speed of the KESM issuch that an entire embedded mouse brain (1 cubic cm) canbe scanned in one month of one-shift operation, yielding anuncompressed volume data set of 15 terabytes (15000 GB).Although our primary research interests are focussed on thecentral nervous system, colleagues have expressed interestin other organ samples including liver, kidney, heart andlung.

Data mining

In addition to cell positions, the structure of the cell somafound in Nissl-stained tissue can provide some informationconcerning the type and orientation of the cell. With thedevelopment of segmentation algorithms designed to parsethese complex data sets, this opens the door to the constructionof complete maps of cells containing cell type, position andorientation. This would provide a high-resolution atlas towhich other data sets can be mapped and related. The speed atwhich KESM data sets can be imaged also allows biologists tocounterstain structures of interest with Nissl. These counter-stained structures can then be put in context by mapping thevolume (using the Nissl-stained cells as fiducials) to the whole-brain map.

The use of Golgi–Cox staining allows a more comprehensivestudy of the morphology of cells in the nervous system.Although the percentage of cells stained is relatively small, thesample size is quite high when compared with the small regionsthat can be imaged using optical sectioning. This providesa much larger statistical sample from which to characterizecell morphology. In addition, the cells are imaged ‘in context’,making their relationships to other cells apparent. This isimportant in neural simulations and provides a broader basefor the stochastic generation of cells and connectivity incomputational models.



Fig. 5. Coronal section of the mouse olfactory bulb. The complete section is shown (left) along with two close-up inserts (right). The tissue is stained withNissl along with perfusion of India ink through the vascular system. The optical resolution of the image is 0.6 μm pixel−1 and the section is 1-μm-thick.

Fig. 6. Sagittal re-sectioning of the India ink and Nissl-stained olfactory bulb shown in Fig. 5. The sections are 1-μm-thick (giving 1-μm resolution alongthe vertical axis).

Future work

One current limitation of KESM is the lack of significantresearch on en bloc staining techniques. We have developednovel methods for staining brain tissue with both Golgi and

Nissl stains; however, these stains can supply only limitedinformation. Clearly other absorption light microscopy stainsare of interest. Transgenic animals should provide a goodsolution to the problem of en bloc staining but would requiresome technological modifications to our current techniques.



Fig. 7. Coronal sections of Golgi-stained mouse cortex. The optical resolution along the x and y axes is 0.3 μm pixel−1 and the section thickness is 0.5μm. The complete section is shown (left) along with two close-up inserts (A and B). Because Golgi-stained tissue is sparse, small cross sections of the fibresare generally only visible in a single section. We have shown a close-up (C) created by combining 20 consecutive sections from the associated region intoa single image to show the fibre structure of the cell processes.



Fig. 8. Sagittal re-sectioning of the data set shown in Fig. 7. A single sliceis shown (top). Because Golgi-stained tissue contains sparse data, fibrescan be more easily seen by combining several slices (middle). A singleneuron is shown by combining 300 slices (bottom). The section thickness is0.5 μm (vertical axis).

In particular, since many transgenic animals are engineeredto express fluorescent proteins, we need to modify our currentsystem to allow for fluorescence imaging, and this modificationis currently underway.

The KESM can be classified as a small-scale planing machinethat uses a single-point cutting tool. An important obstacle toobtaining high-quality data is self-excited oscillations (chatter)generated during the cutting process, when sequential sectionsare being removed from the sample (McCormick et al., 2005).As is visible in some of our sample images, chatter continuesto be a problem both in the time it takes to filter it outand the overall quality of the final data set. Altering thecutting velocity for each pass has significantly suppressed thechatter, as described earlier. Modifications to the instrumentto produce a chatter-free operation by increasing the rigidityof the cutting tool also have helped significantly to reducechatter. In addition, development of en bloc staining andembedding methods yielding specimens that are embedded inharder plastic resins also have helped to reduce the problemof chatter. The exact explanation of the chatter phenomenon

Fig. 9. Surface reconstructions of blood vessels (top) from the mouseolfactory bulb (Figs 5 and 6) and cells stained with Golgi–Cox (bottom)from mouse cortex (Figs 7 and 8). The neuron labelled in purple (withneighbouring axon labelled in yellow) are also shown in Fig. 9 (bottom).Nearby astrocytes (green) are shown supporting a blood vessel (red).

encountered with the KESM and its suppression has yet tobe completely elucidated. Extensive modelling of the cuttingprocess, including the study of chatter and chatter suppressionhas been conducted in support of the (metal) machiningindustry, but the precision sectioning of plastic at sub-microthickness has hitherto attracted virtually no attention from themechanical engineering community. Traditionally this grouphas focussed on the work piece geometry and surface quality,and has ignored the tissue ribbon (‘chip’) of interest in physicalsectioning and imaging of tissue.

The image acquisition speed (approximately 7 mm2 s−1

using a 10× objective) provided by the KESM as well as theimage resolution (600 nm pixel−1 using a 10× objective)can allow us to create full-scale maps of entire organs in areasonable time frame. Although we are currently limited



to tissue volumes the width of the field of view of theobjective (2.5 mm at 10×), we are developing cutting andalignment techniques that can be used to combine adjacentvolumes and thereby eliminate constraints on specimensize.

Application in clinical studies

Alzheimer’s disease (AD), the most common cause of dementiain the elderly, is a complex central nervous system disorderprimarily characterized by progressive loss of cognitiveabilities. Pathological hallmarks of AD are extra-cellularamyloid plaques, intra-cellular neurofibrillary tangles andneuronal cell death (Yang et al., 2005). Current researchin the pathogenesis of AD also has implicated variouscomponents of the neurovascular unit, which contributeto the dynamic regulation of microvascular permeability.Specifically, atherosclerosis, degeneration of endothelial cellsand decreased microvascular density leading to cerebralhypoperfusion are vascular factors that play a role in thepathogenesis of AD (Hawkins & Davis, 2005; Mateo et al.,2006); cerebral hypoperfusion in particular is a major clinicalfinding in AD (Yang et al., 2004). Ultimately, the criticalpathological feature in AD is significant loss of corticalcholinergic neurons, which leads to memory loss and dementiain affected individuals. Numerous reports link AD pathology tothe presence of extra-cellular beta-amyloid deposition and tothe presence of pathologic or altogether missing microvessels.It has been suggested that these microvascular changes mayresult in perivascular cholinergic neuron loss that is causal inthe cognitive decline of AD (Kalaria, 2005). However, to date,no one has documented specific co-localization of pathologicor missing microvasculature, amyloid and missing cholinergicneurons in the AD neocortex at the cellular level. We proposethattheKESMistheideal instrumenttocarryoutthenecessary3D studies needed to elucidate the complex localization ofmultiple components involved in AD pathology. Combinationsof staining protocols such as India ink to fill blood vessels,including capillaries (Lauwers et al., 2007), Nissl to localizeneuronal cell bodies and immunohistochemistry for beta-amyloid deposition (Bard et al., 2000) could be used to answerthe question of 3D co-localization of microvascular deficits,areas of neuronal cell loss and beta-amlyloid plaque deposition,using the KESM.

Acknowledgements

The authors would like to thank Bernard Mesa for his designefforts and advice on the KESM knife–collimator assembly.

This work was funded in part by National Instituteof Neurological Disorders and Stroke grant no. R01 NS54252, National Science Foundation grant no. 0079874,

Texas Higher Education Coordinating Board grant no. ATP-000512–0146-2001, the Office of the Vice President forResearch at Texas A&M University and the Department ofComputer Science at Texas A&M University.

References

Abbott, L.C. & Sotelo, C. (2000) Ultrastructural analysis ofcatecholaminergic innervation in weaver and normal mouse cerebellarcortices. J. Comp. Neurol. 426, 316–329.

Agard, D.A. (1984) Optical sectioning microscopy: cellular architecturein three dimensions. Annu. Rev. Biophys. Bioeng. 13, 191–219.

Bard, F., Cannon, C., Barbour, R., et al. (2000) Peripherally administeredantibodies against amyloid bold beta-peptide enter the central nervoussystem and reduce pathology in a mouse model of Alzheimer disease.Nat. Med. 6, 916–919.

Born, M. & Wolf, E. (1999) Principles of Optics: Electromagnetic Theoryof Propagation, Interference and Diffraction of Light, 7th edn. CambridgeUniversity Press, London.

Cajal, S.R.N.Y. (1911) Histologie du Systeme Nerveux de l’Homme etdes Vertabras, Vol. 2. Consejo Superior de Investigaciones Cientificas,Madrid.

Denk, W. & Horstmann, H. (2004) Serial block-face scanning electronmicroscopy to reconstruct three-dimensional tissue nanostructure.PLoS Biol. 2, e329.

Denk, W., Strickler, J.H. & Webb, W.W. (1990) Two-photon laser scanningfluorescence microscopy. Science 248, 73–76.

Golgi, C. (1873) Sulla struttura della sostanza grigia del cervello. GazzettaMedica Italiana 33, 244–246.

Golgi, C. (1874) Sulla fina anatomia del cervelletto umano. Rendiconti delR. Istituto Lombardo di Scienze e Lettere 7, 69–72.

Golgi, C. (1875) Sulla fina struttura dei bulbi olfattori. Rivista Sperimentaledi Freniatria e Medicina Legale 1, 403–425.

Hawkins, B.T. & Davis, T.P. (2005) The blood-brain barrier/neurovascularunit in health and disease. Pharmacol. Rev. 57, 173.

Herman, B. & Jacobson, K. (1990) Optical Microscopy for Biology. Wiley-Liss, New York.

Ju, T., Warren, J., Eichele, G., Thaller, C., Chiu, W. & Carson, J. (2003)A geometric database for gene expression data. In: Proc. 2003Eurographics/ACM SIGGRAPH Symp. Geom. Process. 43, 166–176.

Kalaria, R.N. (2005) Vascular factors in Alzheimer’s disease. Int.Psychogeriatr. 15, 47–52.

Lacey, A.J. (1989) Light Microscopy in Biology: A Practical Approach, 2ndedn. IRL Press 2nd, Oxford.

Lauwers, F., Cassot, F., Lauwers-Cances, V., Puwanarajah, P. & Duvernoy,H. (2007) Morphometry of the human cerebral cortex microcirculation:general characteristics and space-related profiles. Neuroimage 39, 936–948.

Mateo, I., Llorca, J., Infante, J., et al. (2006) Case-control study of vascularendothelial growth factor (VEGF) genetic variability in Alzheimer’sdisease. Neuro. Lett. 401, 171–173.

McCormick, B.H., Mayerich, D.M. & Wiercigroch, M. (2005)Nanomachining for high-resolution scanning of mammalian brainmicrostructure. In: Proc. 11th Int. Conf. Frac. Turin, Italy.

Pawley, J. (1995) Handbook of Biological Confocal Microscopy. Plenum Press.New York.



Reymond, A., Marigo, V., Yaylaoglu, M.B., et al. (2002) Humanchromosome 21 gene expression atlas in the mouse. Nature 420, 582–586.

Spector, D.L. & Goldman, R.D. (2006) Basic Methods in Microscopy. ColdSpring Harbor Laboratory Press, Cold Spring Harbor, New York.

Tsai, P.S., Friedman, B., Ifarraguerri, A., et al. (2003) All-optical histologyusing ultrashort laser pulses. Neuron 39, 27–41.

Wiener, N. (1964) Extrapolation, Interpolation, and Smoothing of StationaryTime Series: With Engineering Applications. MIT Press, Cambridge, MA.

Yang, S.P., Bae, D.G., Kang, H.J., Gwag, B.J., Gho, Y.S. & Chae, C.B. (2004)Co-accumulation of vascular endothelial growth factor with β-amyloidin the brain of patients with Alzheimer’s disease. Neurobiol. Aging 25,283–290.

Yang, S.P., Kwon, B.O., Gho, Y.S. & Chae, C.B. (2005) Specificinteraction of VEGF 165 with beta-amyloid, and its protectiveeffect on beta-amyloid-induced neurotoxicity. J. Neurochem. 93, 118–127.

Supplementary Material

The following supplementary material is available for thisarticle.

Video S1. This video shows several sections of tissue takenfrom the mouse olfactory bulb. Cells are skained with Nissl andvasculature is labeled using india ink. The images are shownat low resolution in order to see the large-scale structure.

Video S2. This video shows a higher resolution region ofthe olfactory bulb (supplementary video 1). The data set isconstructed using a stack of images cropped from the originaldata set and re-sectioned and visualized using Amira to showthe alignment of consecutive sections as well as highlight thecellular and vascular structure in a small region.

This material is available as part of the online article from:http://www.blackwell-synergy.com/doi/10.1111/j.1365-2818.2008.02024.x(This link will take you to the article abstract).

Please note: Blackwell Publishing are not responsible forthe content or functionality of any supplementary materialssupplied by the authors. Any queries (other than missingmaterial) should be directed to the corresponding author forthe article.


Documents

Journal of microscop