Fluorescence microendoscopy for imaging deep within tissue. (a) Photograph of compound doublet microendoscope probes 350 μrn, 500 μm, and 1000 μrn in diameter. Each doublet probe is composed of two gradient refractive index (GRIN) microlenses, an endoscopic objective lens (shorter lens elements at the bottom of the photograph) and a relay lens (longer elements with dark coating). The endoscopic objective lens has a higher numerical aperture (NA) value and provides micron-scale resolution. The relay lens has a lower NA and provides the length needed to reach deep tissue (Jung et al. 2004). Each minor tick on the scale bar is 1 mm. (b) Optical schematic for two-photon microendoscopy. A GRIN microendoscope is inserted into tissue to image cells ~1 mm to > 1 cm beneath the external surface. A microscope objective lens focuses ultrashort-pulsed laser illumination (red beam) to just above the top face of the microendoscope probe. The laser focal spot is scanned laterally in this plane (scanning indicated by yellow arrows), typically in a raster pattern. The microendoscope probe projects and demagnifies the laser-scanning pattern to the focal plane within tissue (scanning indicated by yellow arrows), generally at a working distance ~150–800 μm from the tip of the endoscope probe. Within the GRIN microendoscope, the beam undergoes a gradual refocusing owing to the smoothly varying radial refractive index profile. A portion of the two-photon excited fluorescence generated at the focal volume in tissue returns back through the microendoscope probe and microscope objective and is captured by a photodetector. A computer controls the laser-scanning process and reconstructs the resulting image. Three-dimensional image stacks are acquired by combining 2D lateral scanning of the laser focal spot with axial adjustments to the position of the microscope objective, which changes the plane of laser-scanning both above the microendoscope and within tissue. (c) Average intensity projection of a 3D image stack acquired by two-photon microendoscopy, showing a CA1 hippocampal neuron expressing YFP under the control of the Thy1 promoter. The microendoscope probe extended through neocortical tissue but did not enter hippocampus and was positioned just dorsal to CA1 tissue layer stratum oriens. Images within the 3D stack (540 μm in total axial extent) were collected at working distances of 160–700 μm from the tip of the microendoscope, across a 185-μm-diameter field of view. The axial range of this stack covers the tissue layers stratum oriens and stratum pyramidale and part of stratum radiatum; thus, the basal dendrites, cell body, and proximal apical dendrites of the pyramidal neuron were sampled during image acquisition. The microendoscope used has lateral and axial optical resolution limits of ~0.9 ~m and ~10–12 μm, respectively (image courtesy of R. Barretto).

Laser-scanning techniques for two-photon fluorescence microscopy. (a) Multifocal two-photon microscopy. A microlens array creates a grid of laser beamlets. A 2D (x-y) scanning mirror sweeps the resulting grid of focal spots across an intermediate focal plane conjugate to the specimen plane. Unlike in conventional two-photon microscopy, the fluorescence signals from the array of focal spots in the specimen are imaged onto a camera. (b) Three-dimensional (3D) laser-scanning using a piezoelectric objective mount. Axial vibration of the mount together with the use of conventional scanning mirrors permits fast 3D scanning. (c) Data acquired by 3D scanning using a piezoelectric focusing mount. (Top) A reference image stack from mouse neocortical layer 2, in which cells are stained with the Ca²⁺ indicator Oregon Green BAPTA-1 am (green) and the astrocyte-specific marker sulforhodamine 101 (SR101) (red). Dual-stained astrocytes appear yellow. (bottom) The spiral 3D scan trajectory (red curve) cuts across three ellipsoids (green, blue, and yellow) that represent the volumes of interest (VOIs) of three example cell bodies selected from the reference stack. Scale bars: 25 μm. (d) Ca²⁺ activity traces. Line-scan segments within the VOIs selected from panel c were assigned to the three color corresponding cells. For each cell, the fluorescence signals from all sections were averaged, revealing the time courses of the Ca²⁺ signals (green, blue, and yellow traces at right). Stimulation via an extracellular micropipette consisted of several series of 5 bursts of 10 pulses at 100 Hz (1-s burst interval). Scale bars: 10 s (horizontal), 30% ΔF/F (vertical) (right). (Panels c and d are from Göbel et al. 2007.) (e) 3D laser-scanning using acousto-optic deflectors (AODs). Two successive pairs of orthogonally oriented AODs permit both lateral (x and y) and axial (z) steering of the focal spot by adjustment of the center acoustic frequencies and the net acoustic frequency chirp, respectively. Beam propagation is depicted for two different focal locations. (f) Two-photon fluorescence image of a rat CA1 pyramidal neuron obtained by 3D laser-scanning with AODs. The neuron was fluorescently labeled with 50 μM Alexa 594 and 200 μM of the Ca²⁺ indicator Oregon Green BAPTA-1 am. The image is a composite of maximum intensity projection images from eight image stacks that together span the entire cell. The overlaid yellow circles represent the magnitudes of Ca²⁺ transients (ΔF/F amplitudes) measured at a total of 146 recording sites during individual back-propagating action potentials. (Image courtesy of K. Kelleher, Y. Liang, and P. Saggau.) (g) Fast scanning of line illumination with auxiliary optical sectioning provided by temporal focusing. A cylindrical lens focuses a laser beam to a line incident on a diffraction grating, which spatially disperses the light’s wavelength components. A tube lens focuses each wavelength component to a line on the back aperture of the microscope objective, creating a beam with a smoothly varying wavelength gradient in cross section. The objective focuses the beam to a line of illumination on the specimen, such that all wavelength components recombine at the focal plane owing to temporal focusing effects (Tal et al. 2005). Thus, the laser pulses are of shortest duration at the specimen plane, which promotes optical sectioning. A galvanometric (galvo) mirror scans the line of illumination across the sample.

Three-dimensional imaging using linear contrast mechanisms. Fluorescence is depicted here as the contrast mechanism, but alternative contrast mechanisms are available for these techniques. (a) Objective-coupled planar illumination microscopy (OCPI). A cylindrical condenser lens focuses illumination to a sheet of light 4–10 μm thick, exciting fluorescence within an optically sectioned plane. The cylindrical lens and the objective lens are mechanically coupled and move together, enabling fast volume scanning. (b) Optical projection tomography (OPT). OPT reconstructs a 3D sample volume by acquiring a large set of 2D projection images from a rotating sample in a manner analogous to X-ray computed tomography (CT). Each pixel within an individual 2D projection contains a mixture of fluorescence signals originating from many sample voxels. By combining information from projections acquired over multiple angular orientations of the specimen, one can unmix or back compute these signals by using a Radon transform to determine the fluorescence intensities of the individual voxels. (c) Light field microscopy. An array of microlenses situated in an image plane creates a corresponding array of microimages. Light rays originating from different locations in the sample are focused by different microlenses to distinct microimages; rays from nearby locations are focused to the same microimage. Light rays coming from the same neighborhood in the specimen but emerging at different angles reach different pixels within the same microimage. (Light propagation is shown here only for a single neighborhood on the optical axis, with only four cones of light emerging at four distinct angles). The combination of both position and angular information in the set of raw 2D images provides a description of a light field. This in turn allows the sample to be viewed from different perspectives (Figure 4b) or to be synthetically focused at different depths (not shown). (d) Nonscanning holographic fluorescence microscopy. A spatial light modulator (SLM) projects a hologram-generating pattern that reflects two spherical waves of distinct curvature for each emitter in the sample. These two spherical waves interfere at the detector, creating a hologram of the sample (raw image) in which concentric circles encode the emitters’ lateral positions and depths in the circles’ locations and densities, respectively. Three holograms are sequentially recorded per image frame, each using a distinct phase offset for the reflection pattern on the SLM. The holograms are superposed and convolved with a holographic response function to reconstruct the spatial distribution of emitters.

Image volumes acquired using linear contrast mechanisms. (a) One slice from an image volume obtained by objective-coupled planar illumination microscopy, revealing neuronal somata labeled with the Ca²⁺-sensitive indicator Oregon Green BAPTA-1 am in the mouse vomeronasal sensory epithelium (Holekamp et al. 2008). (b) Light field microscopy images of Golgi-stained rat neurons. The four images shown were reconstructed from a single raw image, a portion of which is shown (yellow inset). Each image was constructed as the array of intensities from all pixels located at a specific position within the individual microimages, as schematized (orange inset: left pixels, blue; right pixels, red; center pixels, white; top pixels, green). The resulting images from the four distinct angular perspectives are shown within the color corresponding boxes. The degree of accessible parallax is most evident in the looping processes in the lower left of each image (Levoy 2009). (c) Three-dimensional (3D) gene expression maps in Drosophila melanogaster generated using optical projection tomography (OPT). Signals from autofluorescence (green) and bright field readouts of both β-galactosidase activity (blue) and exoskeleton (red) are shown (McGurk et al. 2007). (d) 3D reconstruction of an E12.5 mouse embryo using OPT imaging of fluorescence signals after immunostaining with a monoclonal anti-neurofilament 160 conjugated to Alexa Fluor 488 (green). The fluorescence image is superimposed on a bright field anatomical OPT image (gray) (image courtesy of J. Sharpe). (e) Time evolution of gene expression patterns (Scx-GFP) in mouse limb buds monitored by OPT. Two views of the 3D gene expression patterns are shown for 0, 13, and 19 h (Boot et al. 2008).

Microfluidic devices for automated manipulation of C. elegans under the microscope. (a) Microfluidic worm sorter consisting of control channels and valves that direct flow of worms. (b) A worm immobilized for imaging by suction against microposts. Fluorescent mec-4::GFP-expressing touch neurons are visible. (c) A microchamber chip for large-scale worm screening. The chip has been loaded with dye to facilitate visualization. Each chamber can be independently addressed and loaded. (d) A worm immobilized against circularly oriented microposts in preparation for imaging. (Inset) High-resolution fluorescence image acquired through a microfluidic chip’s glass substrate; a single touch neuron is indicated by a white arrow. (Panels a–d are from Rohde et al. 2007.) (e) Chip for determining how channel dimensions affect the dynamics and efficiency of worm larvae movement. The worms’ ages and velocities are (top to bottom) for each image: L1, υ < 0.1 mm/s; L2/3, υ < 1.4 mm/s; L4, υ = 1.0 mm/s; young adult, υ = 0.3 mm/s; and L4, υ < 0.1 mm/s (images courtesy of D. Albrecht and C. Bargmann). (f) Microfluidic worm sorter loaded with dye for visualization. This chip is equipped with additional temperature control and positioning mechanisms. (g) punc-25-YFP-RAB-5 expression in (top) wild-type and (bottom) unc-16^−/− mutants. An online machine vision algorithm automatically sorted ~400 worms/h on the basis of cell and puncta organization with 96.5% sorting accuracy after one round on the chip shown in f. White arrows indicate cell bodies, whereas the yellow arrow indicates a punctum (Chung et al. 2008). (h) Microfluidic chip for rapid fluorescence imaging of worm neural responses to olfactory stimuli (Chalasani et al. 2007). (i) Microfluidic worm corral loaded with dye for visualization. (j) Bright field image of immobilized worm in channel. (k) Fluorescence image of GFP-labeled synaptic puncta within a worm imaged in a microchannel (images i–k courtesy of D. Chiu). (l) Time-lapse sequence showing the development of an axon of the HSNL motor neuron, including the transport of GFP-labeled synapse-specific proteins down the axon (Allen et al. 2008).

Techniques for automated and multiscale microscopy. (a) Automated microscopy. Neurons were cotransfected with a morphological marker, mRFP (red), and a mutant huntingtin protein fused to GFP, Htt^ex1-Q₄₇-GFP (green). The expression levels of mutant huntingtin and cell fates were tracked automatically in thousands of neurons over multiple days (Arrasate et al. 2004). (b) Multiscale microscopy as illustrated in a series of increasingly magnified images of a medium spiny neuron from mouse nucleus accumbens. (Left) Mosaic of low-magnification, transmitted light images of mouse striatum. (Left inset) Contrast-enhanced, close-up of an individual neuron. (Middle) Two-photon fluorescence image of the same neuron labeled with Lucifer Yellow and Alexa Fluor 568. (Right) Tomographic reconstruction of a dendrite by electron microscopy and the corresponding segmentation of dendritic spines from the same neuron (images courtesy of E. Bushong, M. Martone, and M. Ellisman, available online at the Cell Centered Database, http://ccdb.ucsd.edu). (c) Array tomography. Volumetric image of mouse vibrissa somato-sensory cortex. The image reveals synapses labeled by rhodamine-conjugated antisynapsin I (small red objects), cellular nuclei (DAPI-DNA, blue), autofluorescence from blood cells (large red objects), and a subset of pyramidal neurons labeled by FITC-conjugated anti-GFP (green) and was rendered from 134 sections, each 200-nm thick (image courtesy of K. Micheva and S. Smith; Micheva & Smith 2007). (d) All-optical histology. Maximum intensity projections of fluorescently labeled vasculature in the mouse vibrissa somato-sensory cortex. A penetrating arteriole is highlighted in yellow (P.S. Tsai, P. Blinder, J.A. Kaufhold, B. Friedman, and D. Kleinfeld, unpublished data).

Microscopy in awake, head-restrained animals. (a) Apparatus for microscopy in an awake, head-restrained mouse. Cellular dynamics are imaged while simultaneously monitoring the movement of the mouse’s exercise ball. (b) Images of cerebellar cortex labeled by the Ca²⁺ indicator Oregon Green BAPTA-1 am (green) and the astrocyte marker sulforhodamine 101 (SR101) (red). SR101 co-staining allows online distinction between neurons (green) and astrocytic cells (yellow) in the molecular (left) and Purkinje cell (right) layers. (c) Example traces of displacements in x and z during mouse motion, showing the motion artifacts associated with running speed. Scale bars: speed, 20 cm/s; displacements, 5 μm; time, 10 s (panels a–c based on Nimmerjahn et al. 2007, 2009; Nimmerjahn 2009). (d) Contours of 48 Purkinje cell dendritic tree segments identified within two-photon Ca²⁺ imaging data acquired in an awake head-restrained mouse, superimposed on an image of the average fluorescence. (e) Ca²⁺ activity traces from four of the numbered dendritic tree segments shown in (d). Vertical ticks indicate Ca²⁺ spikes, as determined from threshold crossing events of the Ca²⁺ signals, after deconvolution to correct for the dye’s Ca²⁺ binding kinetics. Scale bars: 3 std. dev. (vertical) and 2 s (horizontal) (panels d and e based on Mukamel et al. 2007). (f, left) A false color time-projection image, acquired by two-photon microscopy over a 5-min period, 150 μm deep within the hind limb sensory cortex (S1HL) of an awake behaving mouse. The tissue was labeled by the Ca²⁺ indicator Calcium Green-1-am (green) and the astrocyte marker SR101 (red), such that neurons and astrocytes appear green and yellow, respectively. (Right) Regions of interest in which Ca²⁺ activity was monitored, example traces of which are shown in panel (h). (g) Simultaneous traces of the mouse’s locomotor speed and estimated brain displacements. Scale bars: speed, 20 cm/s; displacements, 5 μm; time, 10 s. (h) Ca²⁺ activity traces from four neurons outlined in (f), as obtained after correction of the image data for brain motion. Ticks indicate positive-going transients, with an estimated <5% motion-induced false positive error rate. Scale bars: 20% ΔF/F and 10 s (panels f–h based on Dombeck et al. 2007).

Portable microscopy using miniaturized optics and scanners. (a–c) Portable two-photon microscope based on a microelectromechanical systems (MEMS) scanning mirror. (a) Cutaway view of a computer-aided design (CAD) model of the microscope. NIR ultrashort laser pulses are delivered to the device via the photonic bandgap fiber. Illumination exiting this fiber passes through a collimating lens and reflects off the MEMS scanning mirror. A micro-optic assembly focuses the laser light onto the specimen and projects fluorescence signals into a multimode collection fiber. The collimating lens is mounted within a mechanical shuttle to allow focal adjustments in the plane of excitation within the sample. (b) Two-photon fluorescence image of neocortical capillaries in an anesthetized mouse, obtained using the microscope after labeling the blood plasma with the dye fluorescein-dextran. The frame acquisition rate was 4 Hz, and 8 frames were averaged. (c) Line-scanning data acquired over the region indicated in (b) and acquired at 560 Hz. As erythrocytes passed through the vessels, the fluorescence signal transiently diminished, creating the appearance of dark diagonal streaks. Velocities of individual erythrocytes were found by computing the slopes of individual streaks. Scale bars: 0.5 s (horizontal), 20 μm (vertical) (panels a–c based on Deisseroth et al. 2006; Piyawattanametha et al. 2007). (d) Portable fiber-optic spiral-scanning two-photon microscope. A photonic bandgap fiber guides ultrashort-pulsed laser illumination (red) to the microscope. The overhanging fiber cantilever is driven in a spiral pattern at mechanical resonance by the piezoelectric scanner. The micro-optic assembly images the light emitted from the fiber tip onto the specimen plane. Fluorescence emissions (green) are collected by a large-core multimode fiber and detected remotely by a photomultiplier tube (not shown). (e) Photograph of the microscope schematized in (d). (f) Regions marked in color exhibited Ca²⁺ activity within cerebellar Purkinje cell dendrites of anesthetized rats as recorded using the microscope of (d) and (e). Cells were stained with the fluorescent Ca²⁺ indicator Oregon Green BAPTA-1 am. (g) Normalized fluorescence traces (ΔF/F) reveal Ca²⁺ transients from the color corresponding regions in (f). Scale bars: 5 s (horizontal), 10% ΔF/F (vertical) (panels d–g based on Engelbrecht et al. 2008). (h–j) A high-speed 1.1-g miniaturized epi-fluorescence microscope for use in freely moving mice. (h) Cutaway view of a CAD model of the microscope. The fiber bundle delivers illumination from a mercury arc lamp and returns the fluorescence image for projection onto a high-speed camera. The microscope, which is mounted on the mouse’s head, contains three microlenses and a miniaturized gear assembly for focusing. (i, left) Colored lines mark the perimeters of 17 individual Purkinje cell dendritic tree segments that were identified in a freely behaving mouse using the microscope in (h). Purkinje cells were labeled with the Ca²⁺ indicator Oregon Green BAPTA-1 am. (Right) Average of four frames in which the blue-filled cell (left) exhibited a Ca²⁺ spike. (j) Mean waveforms for dendritic Ca²⁺ spikes are averages of 37 (red) and 23 (blue) Ca²⁺ spikes from the color corresponding filled cells in (i). Scale bars: 0.1 s (horizontal), 0.5% ΔF/F (vertical) (panels h–j based on Flusberg et al. 2008).

Stimulated emission depletion microscopy. (a) Jablonski energy diagram illustrating the competing processes of stimulated emission (red) and spontaneous fluorescence emission (green) following electronic excitation. Colored arrows indicate absorptive and radiative transitions. Black arrows indicate nonradiative vibrational relaxations. (b) In STED microscopy, two concentric laser beams are scanned together across the sample. An inner excitation beam excites fluorophores in the sample within a diffraction-limited focal spot (blue). An outer doughnut-shaped beam depletes the population of excited fluorophores at the periphery of the excited region (red) via stimulated emission. Thus, only the excited fluorophores within the central hole of the doughnut emit fluorescence. These remaining excited fluorophores can occupy an area of diameter well below the diffractive limit, even though both laser beams are subject to diffractive limitations. By tuning the power of the depletion laser beam, one can adjust the size of the region over which the population of excited fluorophores is fully depleted. This in turn modulates the diameter of the central fluorescence emission region and thereby the point-spread function. (c) Optical schematic of the propagation of the two laser beams used for STED microscopy. (d) STED image of vimentin (immunolabeled with the dye Atto 590) in a neurite of a retinoic acid differentiated human neuroblastoma (SH-SY5Y). (e) Confocal image of the same sample (images in d and e courtesy of S. Hell, acquired by supercontinuum STED microscopy as described in Wildanger et al. 2008). (f) Single section image of dendritic spines of a YFP-positive CA1 pyramidal neuron in an organotypic hippocampal slice, acquired by live-cell STED microscopy. (g) Volume-rendered reconstruction of a stack of STED images (Δz = 0.5 μm). Panels (f) and (g) are based on Nägerl et al. 2008.

Structured illumination microscopy (SIM). (a) A 3D SIM microscope employs periodic illumination patterns generated by the interference of multiple beams at the sample plane. To permit structured illumination imaging of the finest resolution, the spatial frequencies of the illumination patterns must closely approach the highest values the microscope can transmit. (b) The sequence of steps in 2D linear SIM. The sample is illuminated with a sequence of sine wave spatial patterns that are generated by the interference of two mutually coherent illumination beams. A set of images is acquired using at least three different orientations of the sine pattern, produced by using different in-plane rotations of the grating (three columns marked by angular orientation labels in color). Each pattern orientation is imaged using at least three different phase shifts, produced by using different in-plane translations of the grating (three rows marked by phase shift labels in white). These periodic illumination patterns generate moiré-like interference patterns when mixed (convolved) with the set of spatial frequencies present in the sample structure (i.e., in the spatial frequency transform of the distribution of fluorophores). Thus, although a microscope cannot directly transmit high spatial frequency information beyond the diffraction limit, deliberately inducing aliasing encodes information in the resulting beat pattern between the spatial frequencies present in the illumination and in the sample. Computationally extracting this encoded information and moving it back to its original position in frequency space, and then performing an inverse Fourier transform, yields a super-resolution image with about a twofold improvement over the resolution of conventional microscopy. In 3D SIM, three spatially coherent beams interfere at the sample, creating an illumination pattern with added structure in the axial direction. The axial component of the illumination structure encodes high-axial-resolution sample information in the image data, in a fashion similar to what happens in the lateral direction. The axial information does not require phase-shifting to extract but can be accessed simply by acquiring a focal series with the focal planes spaced twice as densely as normal. (c) Nonlinear SIM, for example using saturating illumination, provides even higher resolution than does linear SIM by applying sine wave illumination patterns that, at their peaks, produce saturating fluorescence excitation. This saturation adds a nonlinearity to the dependence of excitation on illumination intensity and thereby creates higher-order harmonics in the spatial modulation of the fluorescence excitation efficiency. To reconstruct images containing these higher-order spatial frequencies, a greater number of raw images must be acquired with different translations and rotations of the grating than that in linear SIM. Nonlinear SIM allows images to be constructed with about a fourfold improvement in resolution as compared with conventional microscopy but puts greater demands on the phototolerance of the sample. (d) 3D linear SIM reveals axonal and synaptic organization in wild-type Drosophila muscle 4 neuromuscular junction stained for Ank2-L (green) and colabeled with cell adhesion molecule Fas II (red). Insets show greater detail in densely labeled regions (Pielage et al. 2008).

Super-resolution microscopy based on localization of single molecules. (a) To create super-resolution images based on fluorescence signals from single molecules, a sequence of raw images is acquired in which only a sparse subset of the available fluorophores has been activated in each image. This procedure enables each fluorophore to be localized to within sub-diffraction-limited bounds. A super-resolution image results from the sum of the localized fluorophores. (b) Multicolor stochastic optical reconstruction microscopy (STORM) images of clathrin-coated pits labeled with Cy3-Alexa 647 (red) and microtubules labeled with Cy2-Alexa 647 (green) in a BS-C-1 cell shown at three different magnifications (Bates et al. 2007). (c) 3D STORM (left) and corresponding diffraction-limited (right) images of microtubule networks in kidney epithelial (BS-C-1) cells (Huang et al. 2008b). By introducing astigmatism into the optical pathway, the axial position of each fluorophore is encoded in the ellipticity of the spatial distribution of photons emitted from that fluorophore and captured on the camera. Axial position is indicated in color in the 3D-STORM image. (d) Multicolor photoactivated localization microscopy (PALM) image showing the distribution of adhesion complex proteins in a fibroblast (HFF-1) cell. Epi-fluorescence image of mCerulean-tagged actin (blue) overlaid with a PALM image of Dronpa-labeled paxillin ( green) and tdEos-labeled vinculin (red). Images are shown at three different magnifications (Shroff et al. 2007). (e) An image acquired by total internal reflection microscopy (TIRF) of a COS-7 cell expressing the lysosomal transmembrane protein CD63 fused to the photoactivatable fluorescent protein Kaede and the corresponding PALM image (left inset) (Betzig et al. 2006).

Emerging methods for coherent generation of intrinsic optical contrast: optical coherence tomography and coherent anti-Stokes Raman scattering microscopy (CARS). (a) Image acquired by adaptive optical, ultrahigh-resolution-optical coherence tomography (AO-UHR-OCT) of the retina of a live 35-year-old volunteer. The image was acquired at 4.5° nasal-retinal location by an AO-UHR-OCT system with 3.5-μm axial resolution. The focus of the adaptive optical system was set at the photoreceptor layers, and 500 axial depth scans were performed along a 0.5-mm line to generate the image. Fine details of the photoreceptor layer reveal individual human photoreceptors (bottom). Scale bars: 100 μm (top), 50 μm (bottom). Abbreviations for anatomical assignments: Ch, choriocapillaris and choroid; GCL, ganglion cell layer; I/OS, inner/outer segment junction; INL, inner nuclear layer; IPL, inner plexiform layer; OLM, outer limiting membrane; ONL, fibers of Henle with outer nuclear layer; OPL, outer plexiform layer; RNFL, retinal nerve fiber layer; RPE, retinal pigment epithelium; and VM, Verhoeff’s membrane created by end tips of cone photoreceptor outer segments (based on Zawadzki et al. 2008). (b, c) CARS microscopy images of lysophosphatidylcholine-induced swelling of myelin sheaths that envelope axons in excised guinea pig spinal tissue. The pump and Stokes laser beams for CARS microscopy were colinearly focused in the imaging plane of the neuronal axons. (b) CARS images of normal myelin sheathes obtained using excitation light (pump and Stokes beams) polarized parallel (left) or perpendicular (right) to the axons’ orientation. (c) CARS images of swollen myelin sheaths using excitation light that is polarized parallel (left) or perpendicular (right) to the axons’ orientation (panels b and c are based on Fu et al. 2007).

Second-harmonic generation (SHG) imaging reveals the dynamics of striated muscle sarcomeres and neuronal membrane voltage. (a–f) In vivo imaging of striated muscle sarcomeres by SHG microendoscopy. (a) Raw and (b) band-pass-filtered SHG images acquired using a 350-μm-diameter microendoscope inserted into the lateral gastrocnemius of an anesthetized mouse. (c) Map of sarcomere length variations within the images of (a) and (b). (d, e) Line-scan images acquired at 488 Hz in the extensor digitorum muscle of a human subject using a 350-μm-diameter microendoscope with digits of the hand held flexed (d) or extended (e). Scale bars: 100 ms (horizontal), 10 μm (vertical). (f, left) Time variations of mean sarcomere lengths from the same data sets as in (d) and (e). (Right) Average sarcomere lengths as determined from multiple images in each of three human subjects with the digits of the hand held either flexed (green) (n = 12, 17, and 11 trials, respectively) or extended (blue) (n = 10, 10, and 7 trials, respectively). Error bars are standard errors of the mean (a–f based on Llewellyn et al. 2008). (g–k) Imaging neuronal membrane potential by SHG microscopy. (g) SHG imaging of a layer 5 pyramidal neuron from a brain slice of mouse primary visual cortex stained with FM 4-64. Spines on a basal dendrite and an oblique dendrite are readily visible (insets). (h) Comparison between membrane contrast of the same neuron imaged with two-photon fluorescence (TPF) (top) and SHG (bottom). The strong fluorescence signal from the cytoplasm in the TPF image is not present in the SHG image. (i) SHG measurements of spines during action potential invasion. Representative layer 5 pyramidal neuron used for action potential measurements by FM 4-64 SHG imaging. A lower-resolution projection from an image stack reveals the neuron. A high-resolution image (inset), in which the dendritic spine is indicated by the yellow arrow. (j) A single action potential was initiated by current injection at the soma [scale bars: 50 ms (horizontal), 30 mV (vertical)], and (k) the resulting SHG signal changes were measured in the soma (average trace of four recordings, left) and individual dendritic spines (average trace of seven recordings, right). Each recording is a result of averaging ~40 action potentials. Scale bars: 50 ms (horizontal), 2.5% ΔSHG/SHG (vertical). (Images g–k are based on Nuriya et al. 2006.)

Fluorescent tool mice for microscopy studies. (a) Two strategies for creating fluorescently labeled Brainbow mice. (Top) In the Brainbow-1 construct, three sets of incompatible lox sites (shaded gray triangles) enable three mutually exclusive Cre-mediated recombinations. Excisions (dash-dot lines) of XFP-coding segments of DNA are possible. (Bottom) In the Brainbow-2 construct, two sequential invertible XFP-coding segments of DNA between pairs of loxP sites (blue triangles) undergo Cre-mediated recombinations. Both inversions (dashed lines) and excisions (dash-dot lines) are possible. In both strategies, only the first XFP following the promoter is expressed. (b) Fluorescence images of an axon tract in the brain stem and (c) the hippocampal dentate gyrus of Brainbow mice. The Brainbow mice from which these images were taken exhibited an estimated 90–160 distinct colors, as a result of the co-integration of several tandem copies of the transgene into the mouse genome and the independent recombination of each by Cre recombinase. The images were obtained by the superposition of separate red, green, and blue imaging channels (images a–c are based on Lichtman et al. 2008). (d) Fluorescent labeling using mosaic analysis with double markers (MADM). Chimeric genes containing loxP sites (blue triangles) between split exons of GFP and RFP, or vice versa, are positioned on homologous chromosomes. (XFP-coding regions are shown as halves of filled colored pentagons.) The chimeric genes on each chromosome are abbreviated as GR and RG, respectively. After DNA replication (G2 phase, left), Cre-mediated interchromosomal recombination of the reciprocally chimeric gene pair GR/RG (middle), followed by one of two types of chromosomal recombination during mitosis (G2-X or G2-Z segregation), generates a daughter-cell pair with different fluorescence qualities (right). Outcomes for conditional gene knock-out in daughter cells after mitosis are shown using symbols to signify the placement of wild-type (+) or mutant knocked-out (−) genes distal to the XFP-coding regions. Whereas unlabeled and double-labeled daughter cells are heterozygous for the gene mutation, singly labeled daughter cells are homozygous for either the knock-out mutation (GFP-labeled) or the wild-type gene (RFP-labeled). (e) Hippocampal dentate gyrus of a GR/RG; Nestin-Cre/+ mouse (P15) showing distinct labeling of neuronal cell types in the hillus, granule cell layer, and molecular layer. Blue structures indicate (4′-6-Diamidino-2-phenylindole) (DAPI) nuclear stain (images d and e are based on Zong et al. 2005). (f) Cerebral cortex of a GR/RG; Nestin-Cre/+ mouse (P2) showing green, red, and double-labeled neurons (image courtesy of J.S. Espinosa and L. Luo). (g) Single-neuron labeling with inducible Cre-mediated knockout (SLICK). (top) In the SLICK construct, YFP and CreER^T2 are coexpressed under the control of back-to-back copies of the Thy1 promoter. (bottom) In progeny of a cross between SLICK mice and transgenic mice created with the construct shown, administration of tamoxifen induces temporally controlled Cre-mediated conditional transgene expression. Use of a different construct enables conditional gene knock-out. (h) Conditional transgene expression using SLICK. Progeny of a cross between SLICK (line V) and R26R mice were treated for five days with tamoxifen. In the R26R Cre reporter strain, a transcriptional stop sequence is floxed upstream of a lacZ reporter gene coding for β-galactosidase at the ROSA26 locus. Two weeks following tamoxifen treatment, YFP-labeled pyramidal neurons in the CA1 hippocampal region exhibited robust lacZ expression revealed by immunofluorescent staining for β-galactosidase (images g and h are based on Young et al. 2008). Cyan, green, orange, red and yellow fluorescent proteins are abbreviated as CFP, GFP, OFP, RFP, and YFP, respectively.

Time-lapse fluorescence microscopy in live mammals. (a) In vivo two-photon microscopy images of the growing dendrite of a layer 2/3 pyramidal neuron that expresses GFP under the control of the Thy1 promoter in mouse somatosensory cortex (GFP-M line). The first three panels show the dendrite following cortical deletion of pten (phosphatase and tensin homolog on chromosome 10). CaMKII-Cre was used to excise floxed pten. Beginning in postnatal week 11, the mouse received daily doses (intraperitoneal injection) of rapamycin, which inhibits the same pathway as does pten. The last three panels show the same dendrite exhibiting a loss of filopodia-like spines near the tip (week 11–12) and stopped growth without retraction (week 12–15) (based on Chow et al. 2009). (b) Long-term in vivo two-photon imaging of layer 2/3 axon dynamics in adult mouse neocortex. Layer 2/3 neuron precursors were transfected with GFP by in utero electroporation at E16. Chronic glass-covered cranial windows were implanted in the hemisphere contralateral to the site of electroporation in adult mice (>3 months of age). At least four weeks after window implantation, callosal projections of layer 2/3 neurons in barrel cortex were imaged by in vivo two-photon microscopy at four-day intervals. Day 0 is the first day of imaging. Examples of axonal bouton dynamics are indicated by arrows (yellow arrows, stable boutons; green arrows, new boutons; red arrows, disappearing boutons). Note the extensive growth and retraction of the tip of the axon between days 12 and 40 (blue dashed lines). Images are projections of 8–16 slices, 1.5 μm apart, in which some distracting processes have been digitally removed for display purposes. Abbreviations: EPB, en passant boutons; TB, terminaux boutons (images courtesy of C. Portera-Cailliau). (c) In vivo time-lapse imaging of dendritic spine formation and elimination in the mouse cortex, as imaged at two time points through a thinned skull. Repeated two-photon imaging of two dendritic branches (top and bottom) at four (left panels) and six (right panels) weeks of age reveals spine elimination (arrowheads) and formation (large arrows) as well as filopodium turnover (small arrows) in control (top) and sensory-deprived (bottom) mouse barrel cortices (based on Zuo et al. 2005b). (d) Images of CA1 hippocampal pyramidal cell bodies expressing YFP under the control of the Thy1 promoter, as imaged by two-photon microendoscopy in a live mouse. On day 0, a guide tube was surgically implanted so that the microendoscope probe could be inserted repeatedly to the same tissue location just dorsal to CA1. The image acquired on day 239 after the initial surgery (green pseudocolor, left) closely resembles that acquired on day 241 (red pseudocolor, middle), as shown by the yellow portions in the merge of the two (right) (based on Deisseroth et al. 2006). (e) Spine imaging after stroke: long-term in vivo two-photon imaging of dendritic spines in layer 5 pyramidal neurons in the sparsely labeled primary somatosensory cortex of Thy1 -GFP-M mice. Imaging through a chronically implanted glass-covered cranial window began eight days before and ended three months after unilateral permanent middle cerebral artery (MCA) occlusion. Only dendrites in peri-infarct cortex, 1–3 mm away from the edge of infarct, were imaged. Panels show high-resolution in vivo imaging of dendritic spines before and after stroke at representative time points. Stable spines are indicated by yellow arrows. Thin blue arrows at day +4 after stroke indicate regions of local beading/swelling after stroke. Images are projections of 8–15 slices 1.5 μm apart (images courtesy of R. Mostany and C. Portera-Cailliau). (f) In vivo two-photon imaging of neocortical dendrites and axons (green) near an amyloid deposit (red) in cortical layer 1 of a mouse model of Alzheimer disease at six months of age. Although most spines (arrowheads) and varicosities (asterisks) were stable over two days, some structural changes [e.g., spine loss (arrows) and varicosity formation (double arrows)] occurred near the deposits (based on Tsai et al. 2004). (g) A dorsal root ganglion sensory axon of a Thy1 -GFP-S mouse in vivo before, immediately, and 48 h after transection with a small pin. The mouse was perfused after the 48-h time point, and the same axon (boxed area in +48 h panel) was re-identified in cut sections using confocal microscopy (fourth panel). Two bulbs formed on the transected axon (yellow insets; high power views below); one formed near the lesion site (left bulb), whereas the other died back from the lesion through acute axonal degeneration (Kerschensteiner et al. 2005). Degenerative fragments can be seen between the bulb and the lesion site along the original path of the axon. (For comparison to the in vitro images, the red insets show +48 h views of the bulb in vivo at higher magnification.) (Images are courtesy of T. Misgeld and M. Kerschensteiner.)