Viewing the cell’s inner life via light microscopy

There are four research areas in the Department of Pharmaceutical Chemistry. Viewing the cell’s inner life via light microscopy is a research challenge within physical biology.

The challenge

Structural biology methods such as x-ray crystallography, NMR spectroscopy, and cryo-electron microscopy allow scientists to determine the three-dimensional structures of individual protein molecules. These serve as vital guides to designing drugs that will interact with them and therapeutically alter their activity.

But proteins routinely combine into macromolecular machinery and subcellular organelles to perform key biological tasks—and the higher-order architectures of such complexes are too large for x-ray and NMR to capture, yet too small to be seen in detail via conventional light microscopy.

More importantly, such structural biology methods have typically required isolating purified molecules—which does not allow the observation of the protein dynamics and interactions with other cellular components that are needed to fully integrate individual structures into biological systems in their native environment, the cell.

Examples of our methods and research include

Super-resolution microscopy: photo-switchable fluorescence and sub-diffraction resolution

Since its invention more than 400 years ago, light microscopy has been an indispensable tool across almost all fields of biological and biomedical research. It has the power to directly visualize fine structures and dynamic processes at the tissue, cellular, and even subcellular level.

In recent decades, numerous new light microscopy techniques have been invented with ever more analytical power. For example, fluorescence microscopy, by specifically tagging intracellular structures or components with fluorophores—molecules that absorb and then emit specific wavelengths of light—allows the tracking of the location, movement and activities of specific biomolecules in cells. In addition, confocal microscopy and two-photon microscopy (a fluorescence imaging technique) provide three-dimensional sectioning capabilities that are essential to looking deeper into tissue samples.

Breaking the diffraction limit in spatial resolution

However, the capability of light microscopy to resolve sample features is limited by the diffraction of light to roughly one half of its wavelength. For visible light, this diffraction limit is several hundred nanometers. Thus many cellular organelles and subcellular structures appear only as glowing blurs.

sizes of various biological entities and the diffraction limit
Cell 143(7), December 2010

(Left to right) A mammalian cell, a bacterial cell, a mitochondrion (an organelle inside cells), an influenza virus, a ribosome (molecular machine that synthesizes proteins in cells), a green fluorescent protein (used for molecular labeling), and a small molecule (thymine).

Department scientists have pioneered and are extending a technique that overcomes these limits called Stochastic Optical Reconstruction Microscopy (STORM). This technique can routinely provide ten times greater spatial resolution (an order of magnitude), thus resolving subcellular features as small as 20 nanometers laterally and 50 nm axially—approaching the size of an individual protein molecule. (Even finer resolution can be achieved here in specialized experiments). The method can also capture such images rapidly enough to view dynamic biological processes inside cells.

comparison of cell structure resolution
Curr Opin Chem Biol, Oct. 2008, 12(5): 505-514, Bates et al.

A comparison of sub-cellular structure resolution of microtubules (green) and clathrin-coated pits (red) in monkey kidney epithelial cells. (A) Two-color conventional immunofluorescence (in which antibodies linked to fluorescent dyes stain molecular targets) versus (B) STORM images and (C) the further magnified STORM image of the boxed region.

STORM starts by reducing the intensity of the light that activates the conversion of photo-switchable fluorophores from a dark state to another chemical state that can glow. As a result, only some of the millions of protein labels in a cell will light up (switch on) at any one time. This is the random (stochastic) part. If all the fluorophores lit up once, there would be just a glowing blur.

In the STORM method, a series of photographs are taken and software tracks the 3D coordinates of individual fluorophores (the center positions of the fluorescent spots, aka single-molecule localization), producing 3D constellation images that precisely map the labeled macromolecules and/or complexes. With further compiled images, their dynamics are captured via video.

The STORM principle: (At left) In conventional fluorescence microscopy all fluorophores glow at once. (Middle) STORM uses photo-switchable fluorescent probes, and randomly (stochastically) activates only a fraction of them to glow in each snapshot, allowing for localization of individual molecules. (Right) A super-resolution image is reconstructed from those localizations.

Curr Opin Chem Biol, Oct. 2008, 12(5): 505-514, Bates et al.

Top row: Example of such super-resolution imaging by high-precision localization of photo-switchable fluorophores. (A) Schematic of cell with structures of interest (gray filaments) labeled with photo-switchable fluorophores, initially none are activated by light source. Red box indicates area shown in B-C-D. (B) A sparse set of fluorophores are activated such that their images (large red circles) do not overlap. The position of each activated fluorophore is fitted by finding the geometric centers (centroids), the black crosses. (C) In subsequent activation cycle, a different set of fluorophores are activated and have their positions determined. (D) With sufficient number of fluorophores localized, a high-resolution image is constructed by plotting the fluorophore positions (tiny red dots).

Bottom row: Comparison of same sub-cellular structures (microtubules) from kidney epithelial cells seen via conventional immunofluorescence (E, G) and STORM images (F,H), with (G) and (H) showing boxed region in (E).

Improving on super resolution

In addition to making discoveries about the structures and activities of macromolecular complexes inside living cells, department scientists have continually worked at improving the methodology. These efforts include:

  • New experimental approaches and image analysis algorithms to visually record faster processes in living biological systems.
  • Quantitative super-resolution microscopy to count the absolute number of molecules in a structure.
  • Better super-resolution microscopy in thicker tissue samples.

Live imaging of cell microtubules via compressed sensing

Department scientists have overcome a key limitation of super-resolution microscopy methods based on single-molecule switching: the trade-off between spatial and temporal resolution. In other words, accumulating enough individual fluorophore activation events to assemble into images takes time. For example, achieving an image with 50 to 70 nm spatial resolution via single-molecule on-off switching events typically requires several thousand camera frames, taking tens of seconds to acquire. This limits time resolution—how brief-lived a biological activity can be viewed.

STORM assembling fluorophore positions

STORM cumulatively assembles highly accurate fluorophore positions over time to generate sub-diffraction imaging.

One way to capture more information faster is to increase the density of activated fluorophores so that each camera frame samples more molecules, but then fluorescent spots or signals overlap, so it becomes difficult to precisely locate individual fluorophores.

In collaboration with others, researchers here developed a technique using compressed sensing (a signal processing technique that can precisely recover signals from highly noisy measurements) to analyze images with highly overlapping fluorescence. By accounting for the fluorophores’ point spread function (the tendency of a light image to spread out concentrically from a single point), the method reduces the overlap of their glow to a collection of sparsely spread out dots representing the positions of the fluorophores. This method allows an activated fluorophore density up to ten times (an order of magnitude) greater than single-molecule fitting methods and, by requiring the acquisition of fewer camera frames, increases imaging speed to 15 times that of prior methods.

STORM imaging of immunostained microtubules
Nature Methods 9, 721–723 (2012), Zhu et al.

STORM imaging of immunostained microtubules (dynamic components of the cytoskeleton and cell structure) in fruit fly (Drosophila) cells using compressed sensing. All scale bars are 300 nanometers.

(Left column) Conventional fluorescence image and a raw image frame captured during STORM data acquisition.

(Middle column) The results of single-molecule fitting—locating each fluorophore by accounting for its concentrically spreading glow to determine its position—reconstructed from 100 and 500 frames of camera images.

(Right column) Result for compressed sensing using the same set of images.

STORM movie of microtubule assembly dynamics in a living Drosophila cell expressing tubulin (proteins that make up microtubules) fused to photoswitchable probes. The movie has a time resolution of (capture events as brief as) three seconds. Reconstructed from 4,349 camera frames (77 seconds), it plays 11 times faster than real time. The scale bar is one micrometer (a millionth of a meter). This video does not include audio.

This improved image analysis algorithm was demonstrated via the live-cell imaging of the dynamic activity of microtubules—protein aggregates vitally involved in cell structure, molecule transport, movement, and replication. Using compressed sensing, immunostained microtubules not discernable via older methods were resolved using as few as 100 camera frames. Ultimately, a super-resolution movie of Drosophila cells assembling tubulin proteins into microtubules demonstrated a much faster time resolution of three seconds with a spatial resolution of 60 nanometers.

Applying sub-diffraction microscopy to view organelle domains: the centrosome

Super-resolution microscopy pioneered and advanced by department scientists is being applied to study the structural organization of macromolecular complexes inside living cells. For example:

Centrosomes are organelles that play integral roles in animal cells of diverse tissues. They help generate (nucleate) and are the main organizing centers for microtubules (cylindrical elements formed from tubulin proteins) that are, in turn, vital elements of the cytoskeleton that forms key structures in all cells and plays dynamic roles, including the intracellular transport of molecular containers (vesicles), protein complexes, and organelles.

Centrosomes are involved in critical steps in stem cell duplication, embryonic development, and the equal distribution of chromosomes in cell replication. When differentiated into basal bodies in ciliated cells, they are essential to the formation of structures for cellular signal transduction and movement. Thus centrosomes are central to normal function—and to diseases ranging from cancers to genetic disorders called ciliopathies.

But a major swathe of the molecular architecture of these important organelles was not known prior to analyses by department researchers and their collaborators applying two methods of sub-diffraction microscopy.

Oxford University Press

Traditional representation of basic centrosome structure (note the amorphous depiction of PCM).

Centrosomes are comprised of two centrioles (barrel-shaped, perpendicular cylinders encircled by microtubule blades) surrounded by the pericentriolar material (PCM). The PCM is known to promote microtubule nucleation: It increases in size, a process known as centrosome maturation, during cell division, recruiting gamma-tubulin ring complexes (microtubule-nucleating centers) from the cytosol. Yet for more than a century the PCM was commonly described only as an amorphous cloud.

While mass spectrometry revealed the numerous protein components of the PCM, and electron and immunofluorescence microscopy identified some PCM structures, its electron density left its molecular details and higher-order organization unknown or hypothetical.

Department scientists and their collaborators applied super-resolution microscopy to view and quantitatively map the distribution and orientation of proteins critical for centrosome maturation (as determined by gene mutation and RNA interference studies) as a way to determine the overall architecture of the PCM. Their experiments focused on the proteins in centrosomes from Drosophila, model organism fruit flies.

Initially, they used 3D structured illumination microscopy (SIM), which illuminates samples with shifting grid patterns of light and then applies software to process the resulting images to generate double the resolution of conventional light microscopy.

SIM imaging revealed distinct differences in the quantitative distribution and organization of the proteins within the PCM, with one domain, nearest the centriole wall (proximal layer) made up of two components, including pericentrin-like protein (PLP), in a donut-like ring (toroid) while the other proteins were farther away with a broad, matrix-like distribution. A few proteins, including PLP, spanned both domains.

Closer examination of PLPs using STORM’s greater resolution discovered they were in distinct molecular clusters around the centriole and formed fibrils extending into the matrix domain to make a spoke-like scaffolding, symmetrically organized by the mother centriole’s nine-fold symmetry. However, that architecture shifts to create a gate-like molecular opening during the formation of the daughter centriole, with one fibril cluster missing in STORM images.

Far from amorphous, such sub-diffraction microscopy revealed that the PCM is comprised of two distinct structural domains with separate functions. For example, the research also revealed, among other details, that the PLP fibrils extending from the mother centriole form a radial scaffold that recruits other proteins and crucially organizes the PCM outer matrix during centrosome maturation.


Subdiffraction microscopy found PLP forms fibrils extend radially from the centriole wall to support the 3D organization of the pericentriolar material (PCM). During centrosome maturation, the PCM is organized into distinct domains. This architecture shifts to allow during daughter centriole formation amid cell replication.

Source: Nature Cell Biology, online 21 October 2012

STORM images

STORM images of PCM proximal layer in Drosophila cells immunostained for PLP. (Left) End-on view. (Middle) side view. (Right) Mother centrioles during centrosome maturation phase of cell cycle, note missing cluster.

Source: Trends in Cell Biology, March 2014, Vol. 24, No. 3; Nature Cell Biology, online 21 October 2012