Nobel Lecture: Single molecules, cells, and super-resolution optics^*

Figure 1

Apertures and near-field optics. (a) Electron beam-fabricated apertures in a silicon nitride membrane, with diameters as shown. Column at left shows light transmission through these apertures ([34]). (b) The concept of near-field scanning optical microscopy (NSOM) as a path to super-resolution ([10]). (c) A 50 nm aperture in a tapered glass pipette coated with opaque aluminum ([12]), made using techniques developed for patch clamp recording ([25]).

Figure 2

Breaking the diffraction barrier in the near field. (a) Microwave resonator with subwavelength aperture as used by [1]. (b) Resolved gratings having periods of $1/30$, $1/40$, and $1/60$ (top to bottom) of the 3 cm microwave wavelength. (c) Images of letters having linewidths of $1/15$ of the wavelength.

Figure 3

Breaking the diffraction barrier in the far field. (a) Diffraction-limited (top) and super-resolved test patterns in 1D (middle) and 2D (bottom) as seen by Lukosz in 1967, using linear grating or square grid masks inserted in the image path ([37]). Resolution is increased to 3 times the Abbé limit. (b) Exploiting nonlinearity in photoresist development and double patterning to create features beyond the Abbé limit during the production of integrated circuits. (c) Exploiting the a priori knowledge of a desired circuit pattern and comparing the distribution of scattered light from the actual pattern to the predicted distribution to measure sub-nm features to sub-Å precision in high volume semiconductor manufacturing. Images courtesy of Dr. Mehdi Vaez-Iravani, Applied Materials, Inc.

Figure 4

My first near-field microscope. (a) The microscope itself. (b) Test-pattern comparison of diffraction-limited imaging vs super-resolved NSOM. (c) Another resolved test pattern, i.e., an early lesson in learning how to sell my work. From [3].

Figure 5

Making NSOM a real tool. (a) Electron micrograph and (b) optical micrograph of an adiabatically tapered, aluminum coated single mode optical fiber used as a near-field probe ([13]). (c) Resolution comparisons with the probe. Left to right: electron, conventional optical, and near-field optical micrographs ([14]). (d) Schematic diagram of shear force feedback for regulating, at the nm scale, the distance from the aperture at the end of the probe to the specimen ([9]).

Figure 6

The golden age of NSOM. (a) Single bits of information (top) in a magneto-optic film recorded and readout by NSOM, compared with bits (white) recorded with conventional optics ([15]). (b) Near-field photolithography ([13]). (c) Fluorescence imaging of phase transitions in phospholipid monolayers ([29]). (d) Histological stained section from the monkey hippocampus ([13]).

Figure 7

Super-resolution fluorescence imaging of cells. Conventional wide field (left) and super-resolved NSOM (right) images of cytoskeletal actin in the same region of a fixed mouse fibroblast cell. From [8].

Figure 8

Single-molecule microscopy at room temperature. (a) Three views of the same field of carocyanine dyes molecules on PMMA as imaged by NSOM with three different polarizations as shown at top. (b) 1D schematic of the interaction of the electric field $\mathbf{e}$ emerging from the near-field aperture with the electric dipole moment $\mathbf{p}$ of a single molecule. (c) Resulting intensity $I(x)\propto {|\mathbf{e}(x)·\mathbf{p}|}^2$ recorded as the aperture is scanned across the molecule. From [7].

Figure 9

Mapping nanometric electric fields and measuring single dipole orientations. (a) Electric field components (rows) predicted near a subwavelength sized aperture at different distances (columns) from the aperture, compared to experimental components (center column) measured with single molecules. (b), (c) Orientations of single molecules determined by matching measured to predicted emission patterns. From [7].

Figure 10

Near-field cryogenic spectroscopy of quantum wells. (a) Experimental schematic, showing a near-field probe (left) exciting a multiple quantum well structure (bottom), with the resulting emission measured at a spectrometer (right). (b) Comparative near-field and far-field spectra at a single point (top), and spectral changes with position (bottom). (c) Images of emission from single exciton recombination sites isolated from one another in a 3D space of position $x$, $y$ and wavelength $\lambda$. From [27].

Figure 11

The concept of super-resolution localization microscopy. (a) A field of close-packed molecules unresolved, because their diffraction-limited images overlap. (b) The same molecules, after their mutual isolation in a higher dimensional space. (c) Super-resolution map of molecular positions after localization of each isolated molecule. From [4].

Figure 12

The Nyquist criterion and the labeling density problem. With many localized molecules per period, any spatial frequency is easily detectable (center left), but when the number drops to 2 per period or less, the frequency might be missed. As a result, resolution in localization microscopy is limited more by density of single-molecule labels than by the precision of their localization (top right). The minimum number of localizations required per diffraction-limited region increasing rapidly with dimensionality (lower right).

Figure 13

The development of photoactivatable GFP (PA-GFP). (a) UV-visible double absorption peaks of wild type GFP ([18]). (b) Pulse-chase experiment to trace the fate of wild type GFP locally photoactivated in one region of a cell ([62]). (c) Energy state diagram from single-molecule photoactivation of GFP ([20]). (d) Improved on/off contrast ratio of PA-GFP ([41]). (e) Pulse-chase experiment of the relative rates of nuclear and cytosolic diffusion of PA-GFP ([41]).

Figure 14

The development of PALM. (a) La Jolla Labs, also known as Harald’s living room. (b) The PALM team. Clockwise from upper left: Harald Hess, Mike Davidson, George Patterson, and Jennifer Lippincott-Schwartz.

Figure 15

Photoactivated localization microscopy (PALM). (a) Repeated rounds of weak photoactivation with violet light activates different subsets of molecules in a specimen. Summing their diffraction-limited spots produces a diffraction-limited image (left column), but summing the measured centers of all such spots produces the super-resolution PALM image (right column). (b) Diffraction-limited (upper left) and super-resolution PALM images in different regions, showing the distribution of transmembrane protein CD-63 in a 70 nm section cut through lysosomes in a COS-7 cell. From [16].

Figure 16

The importance of high molecular on/off contrast. (a) PA-GFP (left), with a poor contrast ratio, yields poor resolution in a PALM image of a focal adhesion, due to imprecise molecular localizations from surrounding background. The photoactivatable fluorescent protein Eos (right), with a high contrast ratio, achieves better resolution ([51]). (b) Diffraction-limited (left) and higher magnification PALM images (right) of islands of high contrast caged $Q$-rhodamine dye ([16]), demonstrating that PALM is not limited to just fluorescent proteins.

Figure 17

Applications of PALM. (a) Chemotaxis receptor clusters in E. coli ([23]). (b) Two-color diffraction-limited and PALM views of the spatial relationship of vinculin and paxillin in focal adhesions ([50]). (c) Spatial relationship of core promoters of transcription (green) relative to the nuclear membrane (red) at different stages of myogenesis ([61]). (d) Tracks of actin polymerization in dendritic spines of live cultured neurons ([21]).

Figure 18

Problems in super-resolution. (a) Overexpression of the target protein can alter the physiological state of the cell ([23]). (b) Exogenously introduced dyes often have limited affinity for the desired target (left, actin cytoskeleton) and high residual background (right, focal adhesion). (c) Chemical fixation can alter the ultrastructure, as seen here in the endoplasmic reticulum, before and after fixation.

Figure 19

Problems in live-cell super-resolution. Compared to diffraction-limited live imaging techniques, the various super-resolution methods ([45]) require the following: large increases in the amount of the fluorescence the specimen must produce (leading to rapid photobleaching), much higher illumination intensities (leading to rapid phototoxicity), and much longer acquisition times (leading to motion-induced artifacts and restricting investigations to slow dynamic processes).

Figure 20

Structured illumination microscopy excels for live imaging. (a) 2D live SIM image of the endoplasmic reticulum in an LLC-PK1 cell, taken from a movie of 1800 time points at 0.75 s intervals. Image courtesy of Dr. Dong Li, Janelia Research Campus, HHMI. (b) 3D live SIM images of mitochondria in a HeLa cell, showing internal structure and the dynamics of fission and fusion events. From [46].

Figure 21

Rapid, noninvasive 3D live imaging with lattice light sheet microscopy ([19]). (a) Tracks denoting the plus ends of growing microtubules, color coded by velocity, during different stages of mitosis, seen in relationship to chromosomes (orange). (b) Computationally extracted slices at different time points through a dividing LLC-PK1 cell, showing the 3D spatial relationship of chromosomes (green), endoplasmic reticulum (magneta), and mitochondria (yellow). (c) Rapid 3D shape changes in the protozoan Tetrahymena thermophila at 0.31 s intervals.

Figure 22

Adaptive optics enables deep imaging at high resolution. (a) Two-photon image of membrane-labeled neurons in the spinal cord of a live zebrafish embryo, 72 hours post fertilization. (b) Same region after adaptive optical correction using direct wavefront sensing ([60]), demonstrating recovery of signal and spatial resolution. (c) Adaptive optical correction of a sparse subset of neurons over a large portion of the zebrafish brain. (d) In vivo two-photon lateral and axial views of neural processes deep in the mouse cortex, before (left) and after (right) adaptive optical correction using indirect wave front sensing ([31]).