Far-field Light Microscopy

Far-field Light Microscopy Christoph G Cremer, University of Heidelberg, Heidelberg, Germany Far-field light microscopy has reemerged as one of the major tools for studying the structure and dynamics of the human genome. This is due to the combination of multispectral molecular labeling and to the development of advanced instrumentation with increased optical, topological and size resolution. Historical Remarks Advanced article Article contents Historical Remarks State of the Art Improvements in Optical Resolution Improvements in Topological Resolution Improvements in Size Resolution Perspectives doi: 10.1038/npg.els.0005922 Light microscopy is regarded as one of the most important inventions in the history of humankind: in the hands of Antony van Leeuwenhoek and Robert Hooke, in the second half of the seventeenth century, it allowed the first glimpse into the microcosm of life. In the nineteenth century, progress in microscopical instrumentation and staining techniques made possible the discovery of the cell nucleus and of the chromosomes as basic constituents of eukaryotic genomes. The following further development of light microscopy and the steady increase in optical resolution constituted the basis of modern biology and medicine. State of the Art Conventional resolution limits imposed by the wave theory of light Since the work of Abbe and Rayleigh at the end of the nineteenth century, wave theory appeared to impose an absolute limit on the potential of light microscopy as a tool for studying the nanostructure of thick transparent specimens such as cells and cell nuclei. In an advanced conventional epifluorescence light microscope or confocal laser scanning fluorescence microscope (CLSM), the optical resolution is limited laterally to about 200 nm in the object plane and to about 1 mm (CLSM about 600 nm) in the direction of the optical axis of the microscope system, assuming biologically relevant conditions (Pawley, 1995). Achievements in human genome research using conventional resolution light microscopy The conventional optical resolution achieved, often in combination with advanced three-dimensional (3D) deconvolution techniques introduced by D. A. Agard and J. W. Sedat in the 1980s, is already sufficient to study many topics relevant to human genome structure: for example, to perform genome-wide cytogenetic analysis of mitotic chromosomes and to identify chromosomal band regions down to the level of several megabase pairs (Mbp) in deoxyribonucleic acid (DNA) content, using fluorescence in situ hybridization (FISH) techniques (Speicher et al., 1996); to visualize in human cell nuclei appropriately labeled chromosome territories, chromosome arm territories and still smaller chromatin domains down to the level of about 1 Mbp of DNA; to identify individual genes and estimate their spatial distribution by using in situ hybridization methods, especially FISH; to localize in live cells individual DNA sequences, using for example, lac operator/repressor recognition; to visualize protein and protein/ribonucleic acid (RNA) complexes related to genome function; to measure the mobility of individual protein and RNA molecules in the nucleus of living cells using fluorescence recovery after photobleaching (FRAP) or fluorescence correlation spectroscopy (FCS) techniques. Biological problems requiring further enhancement of resolution Compared with the typical size of nucleosomes (11 nm diameter), of the chromatin loops of individual genes (e.g. 100 kbp corresponding to a size estimate in the order of 100 nm) or of the biomolecular machines for replication, transcription, splicing and repair (typical size estimates several tens of nanometers), however, the present resolution of the light microscope is insufficient by far to answer many pressing questions in human genome structure research. Examples of such problems are the extent of an interchromatin domain space, the relative positioning of specific genes with respect to chromatin domains, the spatial structure and temporal dynamics of specific gene regions, the spatial requirements for accessibility of specific proteins to transcription factor binding sites, the assembly and dissassembly of genome-related biomolecular machines or modules and the analysis of small changes in the compactness of a specific gene ENCYCLOPEDIA OF LIFE SCIENCES & 2005, John Wiley & Sons, Ltd. www.els.net 1

region as a prerequisite for or as a consequence of transcription (Cremer and Cremer, 2001). Resolution enhancement by alternatives to far-field light microscopy Further increase in the resolution of light microscopes appears to be highly desirable. For many decades, however, such further improvement in resolution appeared to be impossible because of supposed limits imposed by the wave theory of light (see above). Consequently, alternative methods were developed. In particular, -ray crystallography and electron microscopy allowed further enormous progress in the elucidation of genome ultrastructure and related protein/dna complexes. In addition, surface-related techniques like atomic force microscopy (AFM) or near-field scanning optical microscopy (NSOM) allowed study of reconstituted genome structures on surfaces at a resolution considerably below 100 nm. These techniques contributed widely to genome structure research. None the less, only far-field light microscopy would allow the nondestructive study of nanostructures in the interior of cells, such as the 3D architecture of the human genome and its temporal dynamics in the nuclei of morphologically conserved and even living cells. Although in the past there has been occasional speculation on how to break the Abbe limit of resolution in far-field light microscopy, until recently there has not been any progress to an extent that has been really useful in biology. In the following, we shall discuss several major developments in far-field light microscopy (LM) with particular relevance to genome structure research. Improvements in Optical Resolution Since the beginning of the 1990s, Stefan Hell and his collaborators have designed, in a fully quantitative way, and then realized far-field light microscopical devices having a highly improved and practically usable optical resolution. This goal was achieved by using different methods to narrow the microscopic point spread function. The starting-point was the thoroughly calculated theoretical design of a 4Pi microscope where the spatial angle of the incident focusing wave was substantially enlarged compared with conventional microscopy, and with the experimental realization of such a microscope (Hell et al., 1994). A number of procedures are now available to modify ( engineer ) the point spread function (PSF) in such a way that, in combination with appropriate deconvolution techniques, an effective optical resolution of about 100 nm can be achieved. For example, in addition to 4Pi approaches, the PSF may be modified by standing-wave excitation (Bailey et al., 1993) or other means of structured illumination. In another approach, point spread function engineering was achieved by a sophisticated use of stimulated emission of radiation in the object (STED microscopy). Theoretical considerations put the limit of optical resolution of STED microscopy in the range of 25 nm, corresponding to about one-thirtieth of the exciting wavelength. Recently, by STED-4Pi microscopy, an optical resolution close to this limit has been realized also experimentally (Dyba and Hell, 2002). Such a far-field light optical resolution, previously thought to be physically impossible, would allow, for example, analysis of the architecture of chromatin domains with diameters of a few hundred nanometers. Figure 1 shows virtual microscopy examples of the degree of structural information that can be obtained by the various approaches. In cases where multispectral labeling is possible, for example, by using multicolor FISH, the effective structural resolution can be further increased (Figure 2). Improvements in Topological Resolution In a large percentage of cases, the scientific information one wants to obtain by microscopical analysis is the positioning and the mutual distance (topology) between specific identified structural elements. This is also the case in nuclear genome structure research where genome topology in three-dimensionally intact cell nuclei is under investigation (Cremer and Cremer, 2001). An improvement in topological resolution beyond the optical resolution limits is possible if the objects are labeled with fluorescent markers having different spectral signatures (van Oijen et al., 1999). This allows identification of the object elements due to their fluorescence lifetimes or excitation/emission spectra. The positions of the object elements and their mutual distances in the object space are then determined from their coordinates in the image data file, for example, with respect to a given coverslip position (spectral precision distance microscopy, SPDM) (Figure 3). Using confocal microscopy, a topological resolution of the order of about 35 nm laterally (in the object plane) and about 50 nm axially (perpendicular to the object plane, in the direction of the optical axis) has been achieved in topological studies of human nuclear genome regions (Esa et al., 2000). Recently, a confocal SPDM procedure has been applied to measure the measure with a precision of few tens of nanometers the position and distances of single fluorescent molecules. 2

(a) (b) (c) (d) Figure 1 Virtual microscopy of optical resolution improvement. (a) 3D visualization and projections to the lateral (x, y) plane and the vertical (x, z) and (y, z) planes of a computer simulation of a 1 Mbp chromatin domain of total diameter 500 nm (Cremer and Cremer, 2001) assumed to comprise 10 condensed 100 kbp domains formed by nucleosome chains, having an individual size in the 100 nm range. The small beads represent the individual nucleosomes. Here, a labeling of all 10 domains with the same spectral signature was assumed. (b) (d) Virtual microscopy representations (3D visualization, projections) of the monospectral computer simulated 1 Mbp domain shown in (a). For this, the simulated 3D data set of (a) was convoluted with the effective point spread function for the optical resolution assumed. (b) Virtual microscopy assuming a theoretical limit of STED microscopy (see text) of 25 nm. The positions and sizes of the ten 100 kbp regions are all resolved. (c) Virtual microscopy assuming an effective optical resolution of 100 nm (as achieved for example by 4Pi, STED, standing-wave excitation microscopy). Some structural resolution is still maintained. (d) Virtual microscopy assuming an effective optical resolution of 250 nm (achieved, for example, by merging confocal images obtained from different angles). All structural information has been lost, except the total size of the 1 Mbp domain. The bar indicates 100 nm. (Figure kindly provided by Dr G. Kreth, Heidelberg.) To achieve still better position and distance determinations with nanometer range precision under biologically relevant noise conditions, methods of point spread function engineering are required in addition to SPDM. The goal is to modify the of the microscope system in such a way that a maximum precision for determination of the position of the object can be achieved by the total fluorescence photon count given. For example, simultaneous labeling with 10 different spectral signatures would allow very detailed topological analysis of condensed (presumably genetically inactive) and decondensed (presumably active) small gene regions in 3D-conserved nuclei (Figure 4). To achieve the assumed multispectral labeling even under nondenaturing conditions will be a major challenge to molecular cell biology and biophysical chemistry. However, important new biological information would be obtainable already using two or three spectral signature labeling only; this should be feasible even at the present state of the art. In 2002, axial distance measurements approaching the range of a few nanometers and with a precision in the 1-nm range have become experimentally feasible using these methods (Albrecht et al., 2002). It is anticipated that with appropriate calibration procedures, topological resolutions of a similar distance and precision range will become possible. Such a performance might be even better than the biologically encountered variation in topology, for example, the distances between two or more labeled genomic nanostructures. In this case, precise topological measurements will allow measurement of the variation in the structural states of these sites, and thus provide important parameters for a better quantitative understanding of human genome dynamics using modeling and simulation approaches provided by scientific computing. 3

(a) (b) (c) (d) Figure 2 Virtual microscopy of improvement of effective structural resolution by multispectral labeling. (a) 3D visualization of a computer simulation of a 1 Mbp chromatin domain (see Figure 1a). In contrast to Figure 1a, however, here only eight of the ten 100 kbp domains are shown. Each of these eight domains has been assumed to be labeled with a different spectral signature. (b) (d) Virtual microscopy representations (3D visualization, projections) of the multispectral computer simulation of (a). Optical aberrations, for example chromatic deviations, detector and photon noise were neglected. (b) Virtual microscopy assuming a theoretical limit of STED microscopy of 25 nm. The positions and sizes of the eight labeled 100 kbp domains are all resolved; the domains are identified individually due to their distinct spectral signatures. (c) Virtual microscopy assuming an effective optical resolution of 100 nm. In contrast to the monospectral case (Figure 1c), the positions and sizes of the eight labeled 100 kbp domains with an individual size in the 100 nm range are all resolved; the eight labeled domains are also identified individually due to their distinct spectral signature. (d) Virtual microscopy assuming an effective optical resolution of 250 nm. In contrast to the monospectral case (Figure 1d), here the eight labeled domains are identified individually due to their spectral signatures; the positional information as given by the centers of the individual spherical effective diffraction images is also maintained. The size information, however, has been completely lost. The bar indicates 100 nm. (Figure kindly provided by Dr G. Kreth, Heidelberg.) Improvements in Size Resolution Besides positions and distances between specific elements of a structure, the size (diameter) of such elements also yields important scientific information. Examples of this in nuclear cell biology are the sizes of nuclear pores, of transcription factories, of replication factories, of inactive genes (thought to be generally more condensed) or of active genes (thought to be generally less condensed) and the size differences expected to be produced by differences in sequence repeats. Although electron microscopy offers the possibility of performing size measurements in the required range (down to the nanometer range), light microscopical approaches to such high-resolution size measurements would considerably facilitate measurements; they could be performed in three-dimensionally intact cells, eventually even in living ones. Furthermore, multispectral combinatorial labeling (Speicher et al., 1996) would allow simultaneous measurement of the sizes of multiple, specifically identified targets, such as the degree of condensation of multiple nuclear gene regions, on a cell-by-cell basis. In the case that presently used FISH procedures turn out not to be useful for such a purpose, in vivo labeling approaches may be used. Consequently, low-energy light optical nanoscopy would complement and extend the analysis achieved by particle and high energy photonic microscopy. A truly point-like fluorescent object (e.g. one fluorochrome molecule of 1 nm diameter) would result in a diffraction image with the smallest diameter (e.g. the full width at half-maximum (FWHM) of the PSF) achievable with the microscope system used. This ideal diffraction image (corresponding to the PSF) would be disturbed by larger objects; the larger the object diameter, the larger the deviation from the ideal diffraction image. This means that with precise 4

One spectral signature (a) (b) (c) Three spectral signatures Normalized intensity 0 100 200 300 (nm) 400 (d) (e) (f) Normalized intensity 0 100 200 300 (nm) 400 Figure 3 Principle of improvement of topological resolution by spectral precision distance microscopy (SPDM). The precise measurement of multiple positions, distances and angles between specifically labeled target sites (topology) by SPDM and related methods is based on the use of point-like objects in at least one spatial direction labeled with different spectral signatures. By spectrally discriminating image acquisition, the diffraction-limited intensity distribution of each object is then individually recorded. Thus, each spectral detection channel of the microscope contains only the diffraction-limited intensity distribution. After careful correction for chromatic and monochromatic aberrations, for each spectral channel, the image position corresponding to the geometrical image point may then be determined by appropriate algorithms, such as the maximum of the diffraction image, or the barycenter, the fluorescence intensity-weighted analog to the center of mass. Consequently, topological information about the object can be obtained considerably beyond the limit of optical resolution. (a) (c) One spectral signature (monospectral) labeling of three point-like fluorescent objects situated on a horizontal straight line, assumed to have a next-neighbor distance of 50 nm each. (a) Object positions in the object space. (b) Diffraction image obtained assuming an optical resolution of 250 nm. The spatial coordinates of the diffraction image have been normalized to the same scale as in the object space. (c) Normalized fluorescence intensity distribution along a horizontal line through the center of the diffraction image shown in (b). No positional contrast has been maintained. (d) (f) Three spectral signature labeling of the three point-like objects in (a). Here, each of the objects was assumed to have been labeled with a different spectral signature. (e) Diffraction image of (a) after spectrally discriminating image acquisition and correction of optical aberration. The diffraction images of the objects are strongly overlapping. (f) In the fluorescence intensity line scans through the diffraction images for the three spectral signatures, the maximum of fluorescence barycenter position of each of the diffraction images can be determined independently from the others, thus allowing determination of the geometrical image points of the three objects and hence their positions and distances from each other, with a precision much better than the optical resolution. The virtual microscopy diffraction images (b, e) were computed for a confocal microscope with NA ¼ 1.4/63 6 oil immersion objective. (Reproduced with permission from Cremer et al., 1999.) knowledge of the PSF of the system, object diameters even below the FWHM of the PSF of the system can be determined. Using special procedures like volume conserving algorithms, object diameters down to about half the wavelength of the exciting light (i.e. several hundreds of nanometers) have been determined by clsm. This minimum detectable diameter (size resolution) is not sufficient for many applications in cell biology, especially in human genome structure research. Using point spread function engineering with a considerably smaller FWHM of the PSF, it has become possible to measure fluorescent objects of considerably smaller size, down to about one-twentieth of the exciting wavelength. Another approach is to use otherwise modified point spread functions on the basis of spatially structured illumination, such as standing wave excitation/spatially modulated illumination (SMI) microscopy (Bailey et al., 1993; Albrecht et al., 2002). Using SMI far-field fluorescence microscopy and excitation frequencies in the visible light range, diameters (sizes) of small, isolated individual fluorescent objects have been reliably measured down to 40 nm, or about one-sixteenth of the exciting wavelength, with a standard deviation in the range of a few nanometers, or about a hundredth of the exciting wavelength. Theoretical considerations using virtual microscopy computer simulations suggested a size resolution (smallest object diameter that can be distinguished from a point like object) of about 20 nm for the visible light range. For example, this means 5

PSF engineered microscopy (a) (b) (c) Spectral precision distance microscopy (SPDM) Single signature (after convolution with 100 nm PSF) 100 nm 10 signatures (after convolution with 100 nm PSF) Reconstruction (with optical barycenters of each signature) Condensed 100 kb chain Decondensed 100 kb chain 10 kb 50 kb 100 kb Figure 4 Virtual microscopy of topological resolution improvement by multispectral spectral precision distance microscopy (SPDM). Above: (a) (c) Virtual microscopy 3D visualization of a computer simulation of an individual 100 kbp chromatin domain, assuming monospectral labeling (see Figure 1a). (a) Original simulation. (b) Assuming an optical resolution of 25 nm. (c) Assuming an optical resolution of 100 nm (compare Figure 1a). Although in (b) and (c) the overall size of the domain has been obtained correctly, in both cases all internal structural information has been lost. Below: Virtual SPDM of two 100 kbp chromatin domains. Here, it was assumed that the linear 100 kbp sequence was labeled in 10 sections of 10 kbp each, using a different spectral signature for each section. Representation of the computer simulation of a condensed (above) and a decondensed 100 kbp chromatin domain (nucleosome chain) labeled with the 10 spectral signatures as indicated. The small dots represent the individual nucleosomes. The bar is 100 nm. The monospectral (to the left) and the multispectral (to the right) virtual microscopy diffraction images were calculated assuming an optical resolution of 100 nm (half the width of the PSF). While in the monospectral diffraction image all structural information has been lost, following spectrally discriminating registration, in the multispectral image (shown after optical aberration correction) the 10 spectral diffraction images overlie each other. After the SPDM reconstruction (using the fluorescence intensity barycenters/ gravity centers of each of the independently registered diffraction images) the positions of the gravity centers of the ten 10 kbp sections of the two 100 kbp domains can be reconstructed. Thus, the topology of the two 100 kbp domains can be reconstructed to considerable detail. For example, the topology of the condensed and the decondensed 100 kbp domain can be clearly distinguished. Applying the same procedure to a specific 10-kb section (labeling the ten 1-kb sections of the 10-kb sequence with 10 different spectral signatures; performing SPDM), the topology of a gene region would be reconstructed down to the five-nucleosome level. Repeating the same procedure with multispectral oligonucleotide labeling on the 1 kbp level would allow a topological reconstruction on the molecular level. Such high topological resolution reconstructions, however, would require a positional and distance resolution in the range of 1 nm and below. Virtual microscopy simulations (Failla et al., unpublished data) indicate that such requirements can be realized under fluorescence photon yield conditions corresponding to single-fluorochrome labeling (Knemeyer et al., 2000). (Figure kindly provided by Dr G. Kreth, Heidelberg.) that the size of arrays of a few nucleosomes with labeled DNA might already be discriminated by such techniques. In the case of very small nuclear labels (e.g. an individual green fluorescent protein (GFP) connected to a specific DNA sequence) in combination with appropriate registration times, the apparent size obtained can be used to estimate the thermodynamic search volume of the labeled sequence. Thus, micromobility parameters (apparent effective diffusion coefficient) may be determined. These and other 6

(a) b linked DNA BCR2 BCR1 ABL BCR2 BCR1 β a α ABL b γ c Intensity barycenter (b) 52C3 B99E11 52C3 B99E11 (c) Figure 5 Analysis of genome nanostructure by three-color confocal microscopy and SPDM. (a) From left to right (schematically): Linear mapping of three probes labeled with different spectral signatures and covering the BCR ABL fusion region in bone marrow cells of patients with chronic myelogeneous leukemia (CLM); a complex folding of the 3D chromatin structure of this region in nuclei carrying the fusion region; the triangles resulting from the intensity barycenters of the three sequence sites labeled with three spectral signatures, after application of the SPDM procedure (from Esa et al., 2000). (b) From left to right (schematically): Linear mapping of two probes covering the BCR region on the intact chromosome 22; a complex folding of the 3D chromatin structure of this region. The topology of three sequences in the BCR ABL fusion gene region in bone marrow cell nuclei from leukemia patients determined by confocal SPDM. Shown are the relative mean positions of the fluorescence intensity barycenters, in relation to the observation volume of the CLSM used (transparent ellipsoid) (from Esa et al., 2000). (a) Left: The relative mean positions and distances (open dots) obtained by experimental three-color confocal SPDM of two BCR regions and one ABL region in t(9;22) Philadelphia chromosome territory regions of bone marrow cells of CML patients according to (a), measured in 75 nuclei (from Esa et al., 2000). Right: Confocal monospectral virtual microscopy diffraction image calculated for the experimental mean positions shown on the left (vertical axis denotes optical axis of the confocal microscope), using an experimental confocal point spread function (kindly provided by Dr H. Bornfleth, Heidelberg). In the case of monospectral labeling, all topological information about the BCR ABL fusion region has been lost (virtual microscopy image kindly provided by J. von Hase, Heidelberg.) quantitative geometrical parameters obtained by farfield light microscopy may be compared with the results of virtual microscopy simulations of nuclear genome structure. Perspectives The developments in far-field light microscopy described here will allow an optical resolution and a size resolution of the order of a few tens of nanometers and a topological resolution/colocalization resolution in the few nanometers range. For a number of problems in human genome research, this will allow us to reach the domain so far restricted to electron and -ray microscopy, as well as AFM and NSOMmethods, and thus synergistically complement and extend the nanostructural knowledge obtained by these latter methods. At the same time, however, all the established advantages of far-field light microscopy can be maintained, namely nondestructiveness, multispectral imaging, observations of the interior of 3D-conserved ( intact ) and eventually even living cells. Together with appropriate specific molecular labeling techniques down to the single molecule level (Knemeyer et al., 2000) and image analysis procedures, this will for the first time in the history of life sciences allow us to analyze the nanostructure of specific sites of the human genome and their dynamics in intact and living cells. Since now many major elements of life (such as the genome sequence, biochemistry, individual protein sequences and their structure) are known, quantitative analysis of the cellular, and in particular the nuclear, nanostructure and its dynamics will become one of the major issues in the understanding of the specific organization of cells and their differences, for example, in different stages of development, in different tissues or in different pathological conditions. Such an improved understanding of the four-dimensional geometry of life will be of utmost importance for a better understanding of the structural basis of genome reprograming, one of the fundamental problems of modern biology and medicine. 7

See also Digital Image Analysis Microscopy References Albrecht B, Failla AV, Schweitzer A and Cremer C (2002) Spatially modulated illumination microscopy allows axial distance resolution in the nanometer range. Applied Optics 41: 80 87. Bailey B, Farkas D, Taylor D and Lanni F (1993) Enhancement of axial resolution in fluorescence microscopy by standing-wave excitation. Nature 366: 44 48. Cremer T and Cremer C (2001) Chromosome territories, nuclear architecture and gene regulation in mammalian cells. Nature Reviews Genetics 2: 292 301. Cremer C, Edelmann P, Bornfleth H, et al. (1999) Principles of spectral precision distance confocal microscopy for the analysis of molecular nuclear structure. In: Ja hne B, Haussecker H and Geissler P (eds.) Handbook of Computer Vision and Applications, vol. 3, pp. 839 857. San Diego: Academic Press. Dyba M and Hell SW (2002) Focal spots of size l/23 open up farfield fluorescence microscopy at 33 nm axial resolution. Physical Review Letters 88: 163901-1 163901-4. Esa A, Edelmann P, Trakthenbrot L, et al. (2000) Three-dimensional spectral precision distance microscopy of chromatin nanostructures after triple-colour DNA labeling: a study of the BCR region on chromosome 22 and the Philadelphia chromosome. Journal of Microscopy 199: 96 105. Hell SW, Stelzer EHK, Lindek S and Cremer C (1994) Confocal microscopy with an increased detection aperture: type-b 4pi confocal microscopy. Optics Letters 19: 222 224. Knemeyer J-P, Marmé N and Sauer M (2000) Probes for detection of specific DNA sequences at the single-molecule level. Analytical Chemistry 72: 3717 3724. van Oijen AM, Köhler J, Schmidt J, et al. (1999) Far-field fluorescence microscopy beyond the diffraction limit. Journal of the Optical Society of America 16: 909 915. Pawley JB (ed.) (1995) Handbook of Biological Confocal Microscopy. New York: Plenum Press. Speicher MR, Ballard GS and Ward DC (1996) Karyotyping human chromosomes by combinatorial multi-fluor FISH. Nature Genetics 12: 368 375. Further Reading Agard DA and Sedat JW (1983) Three-dimensional architecture of a polytene nucleus. Nature 302: 676 681. Betzig E, Trautmann JK, Harris TD, Weiner JS and Kostelak RL (1991) Breaking the diffraction barrier: optical microscopy on a nanometric scale. Science 251: 1468. Cremer C and Cremer T (1978) Considerations on a laser scanning microscope with high resolution and depth of field. Microscopica Acta 81: 31 44. Dundr M, Hoffmann-Rohrer U, Hu Q, et al. (2002) A kinetic framework for a mammalian polymerase in vivo. Science 298: 1623 1626. Gustafsson MG, Agrad DA and Sedat JW (1999) I5M: 3D widefield light microscopy with better than 100 nm axial resolution. Journal of Microscopy 195: 10 16. Egner A, Jakobs S and Hell SW (2002) Fast 100-nm resolution threedimensional microscope reveals structural plasticity of mitochondria in live yeast. Proceedings of the National Academy of Sciences of the United States of America 99: 3370 3375. Lacoste TD, Michalet, Pinaud F, et al. (2000) Ultrahighresolution multicolor colocalization of single fluorescent probes. Proceedings of the National Academy of Sciences of the United States of America 97: 9461 9466. Politz JC, Browne ES, Wolf DE and Pederson T (1998) Intranulcear diffusion and hybridization state of oligonucleotides measured by fluorescence correlation spectrosocopy. Proceedings of the National Academy of Sciences of the United States of America 95: 6043 6048. Vogt S, Schneider G, Steuernagel A, et al. (2000) -ray microscopic studies of the Drosophila dosage compensation complex. Journal of Structural Biology 132: 123 132. Wachsmuth M, Waldeck W and Langowski J (2000) Anomalous diffusion of fluorescent proteins inside living cell nuclei investigated by spatially resolved fluorescence correlation spectroscopy. Journal of Molecular Biology 298: 677 690. 8