This essay was written by Peter Rosenthal and was first published in the 2009 Mill Hill Essays. An updated version was published in the Mill Hill Essays anthology.  

In 1665 Robert Hooke published Micrographia, a book which revealed the world seen through a microscope. His engravings presented observations of microscopic animal, vegetable, and mineral worlds including the eye of a fly, the “cellular” structure of cork, and snow flakes. Micrographia stimulated the Dutch scientist Anton van Leewenhoek to take up microscopy, leading to his discovery of the world of microbes. Hooke was also concerned with the science of optics and the construction of better microscopes in order to obtain better observations. He would be astonished at the capabilities of modern microscopes, and the biological knowledge that they have uncovered. This essay examines how we can use direct imaging of biological structure at the smallest scale to link cell behaviour with molecular events.

The light microscope is still central to all biological research, with increasing importance given to live cell imaging both in vitro and in vivo. Light microscopy can only resolve objects (i.e. recognize them as distinct) when they are separated by a distance greater than half the wavelength of light, or about half of one micron (a micron is a millionth of a meter). Cells range in size from one to a hundred microns, and therefore many cellular processes can be studied with light, and there are a wide range of light microscopy techniques. The proteins within cells are too small to be distinguished by light microscopy, but in many techniques their movements can be inferred by tracking light from fluorescent tags that are attached to the proteins. These tags are available in a full palette of colours. “Super-resolution” light microscopy techniques are now being developed that employ various tricks to resolve fluorescently labelled proteins separated by distances of only twenty nanometers (a nanometer is a billionth of a meter). The precise location of labelled molecules can reveal, for example, the shapes of sub-cellular organelles and protein filaments such as microtubules with vesicles moving along them.

The components

Many of the basic building blocks of cells including proteins, DNA and RNA have been isolated, purified, and their structures determined at high resolution using X-ray crystallography. These macromolecules may be several nanometers in diameter and each consists of thousands of atoms. The distances between the atoms are a few tenths of a nanometer, and X-rays with wavelength of one-tenth of a nanometer (also called one Ångstrom) are perfect to resolve atomic spacings. X-ray lenses for such wavelengths have not been developed and therefore there are no high resolution X-ray microscopes. Instead, the proteins are crystallized and an analysis of the diffraction patterns observed when X-rays are fired at the crystal reveals a detailed map of the protein. Computer analysis of scattering from a crystal, which provides a regular packing of protein molecules, makes up for the absence of a focussing lens. The energetic X-rays may damage some of the proteins in the crystal, but because the crystal contains many identical copies of the protein, this will not affect the overall scattering if it is kept to a minimum. The protein maps can be interpreted by fitting a model of the chain of amino acid subunits corresponding to the known chemical composition of the protein. A correct model is one that agrees with the experimental diffraction data and satisfies known rules for chemical bonding in the placement of each component atom. The resulting three-dimensional model shows the unique fold of the protein chain and reveals the way proteins use chemistry to function as enzymes, change shape in response to molecular signals, and serve as scaffolds. There is now a database of more than 50000 protein structures from many organisms determined by X-ray crystallography or by NMR (nuclear magnetic resonance) spectroscopy. The structures of many proteins, from the very first to be solved, haemoglobin, have been extremely successful in explaining aspects of cell biological behavior and the molecular basis of human disease.

Armed with knowledge about the identity, function, and structure of many individual proteins, biologists are now looking to explore the structural mechanisms used by protein complexes to regulate events in cells and signalling between cells. Examples include arrays of receptors in bacteria that sense chemical signals in the environment and tell the bacterium which way to swim, or the complex protein architecture at nerve synapses that regulates signaling between neurons. X-ray crystallography has tackled some large multi-component assemblies including viruses and the ribosome, a large molecular machine that translates the genes into proteins. But many multi-protein complexes involve transient interactions between components, exchange of partners, and sometimes display large motions related to their functions. Biology therefore needs methods to acquire high resolution structural snapshots of a world less regular than the inside of a crystal. Understanding the next level in a hierarchy of structural organization requires three-dimensional imaging on a nanometer scale.

Seeing with electrons

To image at a smaller scale than light microscopy, a smaller wavelength must be used. Electrons, which have properties of both particles and waves, have a wavelength roughly 1/1000 of a nanometer, and this is more than adequate to resolve the distance between atoms in molecules. An electron microscope consists of an electron beam illuminating a specimen and electro-magnetic lenses to focus and magnify the image. The transmission electron microscope was invented early in the 20th century and became an essential tool for biological microscopy. Because electrons are scattered strongly on interacting with matter, the electron beam must travel in a vacuum down the microscope column, and the specimens themselves must be thin. The need to use a vacuum posed problems for biological specimens which typically inhabit water-based environments. In the early days of electron microscopy observations were restricted to “dead” specimens obtained, for example, by cutting thin sections of cells that were chemically fixed and stained with metal salts to provide strong contrast. This chemical processing destroyed the highest resolution features, but nevertheless preserved enough of the real cell to be extremely interesting. The electron microscope revealed a sub-cellular world of organelles previously invisible to light microscopy including the endoplasmic reticulum, mitochondrial membranes, ribosomes, and viruses.

In 1959, exactly 50 years ago and nearly 300 years after Hooke’s Micrographia, the physicist Richard Feynman gave a famous lecture entitled “There’s Plenty of Room at the Bottom” which invited young physicists to think about building machines on the atomic and molecular scale, including miniature computers and motors. Feynman’s manifesto for a new field of nanotechnology claimed that construction and control at these atomic length scales would not violate any laws of physics. He drew inspiration from the mechanisms used by the processes of life to manipulate information on the same molecular scale. A significant part of his discussion focussed on the need for better electron microscopes to image the products of nanotechnology. He challenged physicists to build microscopes that could resolve a hundred times more finely than the one nanometer resolution limit of electron microscopes of his era. Feynman did not see this as a route to fundamental physics of particles, but rather as a way to study “the strange phenomena that occur in complex situations” such as in solid state physics (e.g. transistors). Biology clearly represents another such complex situation where new and interesting phenomena arise, though ultimately they must be reducible to the laws of physics and chemistry. Feynman’s view of the importance of the electron microscope to biology was expressed as follows: “It is very easy to answer some of these fundamental questions in biology: you just look at the thing!”

“I call architecture frozen music” – Goethe

Today electron microscopes can indeed resolve features such as the individual atoms in a germanium crystal, or carbon atoms in two dimensional sheets of graphene. Some of the instrument improvements include more coherent electron beams, improved electron optics, better vacuum technology, more stable sample stages, and computer control, all of which have contributed to removing the blur from images of atoms.

What about biological specimens? In addition to better instruments, a major advance has been the development of cryomicroscopy in which unstained, frozen-hydrated specimens are imaged at very low temperatures, down to near liquid nitrogen temperature (-190° Celsius) or liquid helium temperature (-269° Celsius). An aqueous solution containing the specimen is prepared as a thin film and rapidly frozen. If the specimen is thin enough, a glass or vitreous solid is formed without forming ice crystals that damage cell structure. The specimen is thus preserved in a frozen state that can be imaged in the microscope vacuum. This avoids the need for chemical fixation and staining which would distort the living architecture and destroy high resolution features. Figure 1 shows a state-of-the-art electron microscope for biological research.

300 keV field emission gun cryomicroscope
Figure 1. 300 keV field emission gun cryomicroscope at the National Institute for Medical Research.

Biological specimens are more easily damaged by the imaging electron beam than many specimens studied by physical scientists. This is because biological structure is “soft” matter, held together by weaker and more transient forces than the strong glue that holds inorganic crystals together. A solution is to use low dose electron microscopy, which seeks to minimize electron exposure and maximize the information recorded before the specimen is destroyed. This is facilitated by computer control of the electron beam to prevent over-exposure. Radiation damage is also reduced in the frozen-hydrated state, because chemical products of damage cannot diffuse to new sites where they could cause more damage.

Cryomicroscopy can be applied to a wide range of biological problems and specimen types. An important area of application has been the structure determination of membrane proteins, which are difficult to crystallize outside the membrane. Many of these are important potential targets of drugs so understanding the structures will be useful in drug design. Thin, two-dimensional crystals of membrane proteins can be studied by a combination of imaging and diffraction methods to yield high resolution structures as with X-ray crystallography. Cryomicroscopy is especially useful for determining the structure of large proteins and assemblies that are difficult to crystallize, often because of their flexibility or heterogeneity. Low dose images of molecules embedded in a thin layer of vitreous ice can show many different views because they are randomly oriented (See Figure 2). If the molecules are of identical or similar structure the different views may be combined by computer analysis (called single particle reconstruction) to create a three dimensional map of the structure. Such single particle maps of symmetric viruses can approach near-atomic resolution. Large molecular machines that undergo structural changes, such as the ribosome, can be imaged in different functional states.

Molecules of a protein complex (pyruvate dehydrogenase core enzyme)
Figure 2. Molecules of a protein complex (pyruvate dehydrogenase core enzyme) imaged in a thin, frozen film with a three-dimensional map of the complex computed from images (inset).

Excitement now surrounds the possibility of imaging frozen-hydrated cells, from small bacteria to mammalian cells. These cells can be grown on support films, then plunge-frozen and imaged in the microscope, provided the specimens are much thinner than one micron. Thick specimens must be frozen at high pressure and sliced into sections. Images show a complex, crowded environment containing protein assemblies, protein filaments, vesicles, and tubules (See Figure 3). The image of a lysosome, a subcellular organelle containing degradative enzymes, shows the beautiful preservation of concentric rings of membranes (See Figure 4).

Slice of a three-dimensional tomogram showing the edge of a frozen-hydrated cell
Figure 3. Slice of a three-dimensional tomogram showing the edge of a frozen-hydrated cell and a computational model for membrane organelles (Image courtesy of Sebastian Wasilewski).

Two cells may be highly similar, but each is a truly unique structure that changes with time, and therefore images of different cells cannot be combined into one map. This is also true of some viruses, organelles, and many protein molecular machines captured in various states of motion. Rather, we can obtain all the unique views of such specimens by recording images of each example from many directions, usually achieved by tilting the specimen in the electron beam. This strategy for obtaining three-dimensional maps or volumes is called tomography, perhaps most familiar from its medical applications. Because all the images are recorded from a single frozen specimen, much of the high resolution structural information will have faded due to radiation damage, and the resolution will not reach the atomic scale, but may be restricted to about three nanometers. Though information about proteins at the chemical level will not be accessible, it should be possible through this technique to reveal structural networks of protein and membranes important in describing, for example, how the cytoskeleton determines the shape of a cell.

Image of a lysosome, a sub-cellular organelle.
Figure 4. Image of a lysosome, a sub-cellular organelle.

New experimental methods could make the study of cell structure by cryomicroscopy easier and more powerful. The identification of individual proteins in the crowded environment of the frozen-hydrated cell will be facilitated by the use of protein-specific electron-dense labels that provide strong contrast. Imaging the same cell by “correlated” light microscopy and electron cryomicroscopy connects information on different length scales. Perhaps the greatest limitation of frozen-hydrated images is that they are static. However, dynamic events may be studied by comparing cells (or sub-cellular structures) in different states and inferring structural transformations. When viewed on the nanometer scale, the cell appears complex and irregular, providing a diversity of molecular arrangements. If cell architecture is frozen music, it certainly is not the music of the spheres. It is dissonant, modern.

Computing cell architecture

Images of unstained, frozen-hydrated specimens have weak contrast, and whether recorded on film or digital camera, the specimen may be invisible to the eye. Computational analysis of images is therefore required to extract information from the images. The more successfully this is done, the more confidently one can build a model of cell architecture and draw conclusions that will be important to biology. The computer thus becomes an essential addition to the microscope optics. To actually see the cell or a protein complex hidden in the images requires calculation of a threedimensional map that explains the appearance of the specimen in each of the images, which is dependent upon the particular view of the specimen and also how imperfect microscope optics influence the experimental image. This typically requires an iterative strategy that involves cycles of map calculation, comparison of the map with experimental image data, and map improvement until a map that optimally matches the experimental data is achieved.

The structures observed through the microscope, whether drawn by hand or computed from images, have to be interpreted. The threedimensional maps of the cell are crowded and complex, and their interpretation is difficult. Analysis involves segmentation, or “drawing” computational surfaces to delineate the different compartments and organelles within the cell, building protein filaments, and false-colouring these sub-cellular structures according to type. Computer analysis of maps can help enforce objective analysis. While such models may look spectacular in their complexity, how can we be sure that they are in fact accurate or useful representations of the cell? When the experimental work is at sufficient resolution (say better than 5 nanometres), the shapes of individual protein molecules become discernible, and then they can be compared with structural models determined at higher resolution by X-ray crystallography (or NMR spectroscopy). This represents an independent validation of the imaging experiment.

One approach to building a computational model of the cell is to correlate known structural templates for macromolecules with matching locations in the experimental map, and in this sense build a cell containing proteins, nucleic acids, membranes, and protein filaments (See Plate 3). This directly links cell architecture with the known structure and function of the building blocks. It would not be surprising if these models share some agreement with an artist’s conceptions of the cell based on the known list of components. However, the experimentally determined model of the cell will tell us about organizational principles that could not have been deduced or imagined from the parts.

A great emphasis is now placed on understanding complex biological behavior at the systems level (See Richard Goldstein’s Mill Hill Essay on Systems Biology in 2005). The systems could refer to anything from an individual cell’s response to a signal, pattern formation in a developing embryo, or memory storage in a neural network, and their study requires both experimental and theoretical approaches. A structural model of the cell at the nanometer scale includes all the interactions between proteins, membranes, and nucleic acids. A description of a model so complex may focus more on large-scale, structural self-organization rather than the individual components. Models derived from experimental images provide still frames that are starting points for computational simulations of the cell. Such simulations may describe mechanical or chemical interaction between components, with appropriate simplification of their properties to reflect only their functional or structural contributions on the next hierarchical level. The hope is that simulations will demonstrate how collective molecular activity accounts for larger scale behavior of living cells.

Electron microscopes can be applied to the construction and analysis of nanotechnology. Similarly, they will be used to image and understand the creations of synthetic biology from molecular machines to organisms. In the multi-disciplinary spirit of Hooke’s Micrographia, high resolution imaging with electrons will continue to explore previously unseen landscapes in physical science and biology.

Leave a comment


email* (not published)