junk.pdfA New Way to Look at Oxidative StressShut Down, Don’t Stress OutA New Cell Model for Parkinson’s DiseaseWhere to Start? Alternate Protein Translation Mechanism Creates Unanticipated AntigensDevelopmental Milieu Infl uences a Gene’s Role in Tumor FormationA Relay-Signal Model of Nematode Vulval DevelopmentBMP Signaling Maintains Healthy Joint CartilageControlling the Timing of Gene Expression during Organ DevelopmentFor Gene Activation, Location MattersHigh Affi nity: Making Up for Being MaleMicroRNA Is a Major RegulatorSlime Mold Myosin Thick Filament Assembly DissectedSleeping, Waking, … and Glucose HomeostasisReconstructing Neural Circuits in 3D, Nanometer by NanometerOnly Connect: The Functional Architecture of Brain ConnectivityPaying Attention to MemoryOne Brain, One VisionA Clear View of Mycobacterial InfectionPredicting Risk of Mosquito- Borne Disease in Variable EnvironmentsEndangered Frogs Coexist with Fungus Once Thought FatalDid We or Didn’t We? Louse Genetic Analysis Says YesA Richer Map for Searching Scientifi c LiteratureSerial Block-FaceScanning Electron Microscopy to ReconstructThree-Dimensional Tissue NanostructureWinfried Denk*, Heinz HorstmannMax Planck Institute for Medical Research, Heidelberg, GermanyThree-dimensional (3D) structural information on many length scales is of central importance in biological research.Excellent methods exist to obtain structures of molecules at atomic, organelles at electron microscopic, and tissue atlight-microscopic resolution. A gap exists, however, when 3D tissue structure needs to be reconstructed over hundredsof micrometers with a resolution sufficient to follow the thinnest cellular processes and to identify small organellessuch as synaptic vesicles. Such 3D data are, however, essential to understand cellular networks that, particularly in thenervous system, need to be completely reconstructed throughout a substantial spatial volume. Here we demonstratethat datasets meeting these requirements can be obtained by automated block-face imaging combined with serialsectioning inside the chamber of a scanning electron microscope. Backscattering contrast is used to visualize theheavy-metal staining of tissue prepared using techniques that are routine for transmission electron microscopy. Low-vacuum (20–60 Pa H2O) conditions prevent charging of the uncoated block face. The resolution is sufficient to traceeven the thinnest axons and to identify synapses. Stacks of several hundred sections, 50–70 nm thick, have beenobtained at a lateral position jitter of typically under 10 nm. This opens the possibility of automatically obtaining theelectron-microscope-level 3D datasets needed to completely reconstruct the connectivity of neuronal circuits.Citation: Denk W, Horstmann H (2004) Serial block-face scanning electron microscopy to reconstruct three-dimensional tissue nanostructure. PLoS Biol 2(11): e329.IntroductionMany cellular structures are so small that they can only beresolved in the electron microscope. Furthermore, it is oftencrucial to visualize and reconstruct the three-dimensional(3D) structure of biological tissue. One prime example ofwhere 3D information is indispensable is in exploring theconnectivity of local networks of neurons. While axonal anddendritic processes have been traced using the light micro-scope from the very beginning of cellular neuroscience (Cajal1911), light microscopic tracing is only possible if staining isrestricted to a small subset of cells, as results, for example,from the Golgi method (Golgi 1873) or from the mosaicexpression of fluorescent proteins (Feng et al. 2000). How-ever, in many cases, in order to understand computationalalgorithms, the reconstruction of a complete neural circuitmay be necessary. F or this, the resolution of the lightmicrosc ope is insufficient because dendritic and axo nalprocesses can have diameters that are substantially belowthe wavelength of light. This lack of resolution (1) results inthe inability to resolve densely packed neighboring processes,which is absolutely necessary to reconstruct network top-ology, and (2) does not allow a sufficiently precise estimationof the neuronal geometry, which may be necessary forbiophysical modeling of cellular behavior. So far, only theelectron microscope (EM) can provide the spatial resolutionneeded to track neural processes or to identify synapsesunambiguously. Most commonly used to image biologicaltissue is the transmission electron microscope (TEM) (Ruskaand Knoll 1932), in which a broad beam of electrons isdirected at a sample that is thin enough to allow a substantialfraction of the electrons to pass through and then be focusedonto film or another electron-sensitive spatially resolvingdetector. Specimens are typically thin slices that are cut fromblocks of plastic-embedded tissue, with the resulting electronmicrographs providing a two-dimensional cross sectionthrough the tissue. Scanning electron microscopy (SEM)(Ardenne 1938a, 1938b), in which a tightly focused beam ofelectrons is raster-scanned over the specimen while secon-dary or backscattered electrons are detected, is used inbiological imaging mostly as a surface visualization tool,creating a 3D appearance but no actual 3D datasets.Truly 3D information in the TEM can be obtained usingeither tilt-series-based tomography (Hoppe 1981; Frank 1992;Baumeister 2002) or serial ultrathin sections (Sjostrand 1958;Ware and Lopresti 1975; Stevens et al. 1980). Tomography is avery promising tech nique for obtaining high-resolutionstructural data of macromolecules, organelles, and small cellsbut may not be applicable when larger volumes need to bereconstructed, because section thickness is limited to around1 lm. In the high-voltage EM, somewhat thicker sections canReceived May 23, 2004; Accepted July 29, 2004; Published October 19, 2004DOI: 10.1371/journal.pbio.0020329Abbreviations: 3D, three-dimensional; BSE, backscattered electron; EM, electronmicroscope; SBFSEM, serial block-face scanning electron microscope; SEM,scanning electron microscopy; TEM, transmission electron microscopeCopyright: Ó 2004 Denk and Horstmann. This is an open-access articledistributed under the terms of the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium,provided the original work is properly cited.Academic Editor: Kristen M. Harris, Medical College of Georgia*To whom correspondence should be addressed. E-mail: