Light Microscopy

A half dozen optical tricks are used to produce contrast in light micrographs of biological specimens (Table 6-1 and Fig. 6-1). These are called wide-field methods, as a broad beam of illuminating light is focused on the specimen by a condenser lens.

Figure 6-1 light paths through various microscopes. A, Basic optical path in an upright light microscope. The condenser lens focuses light on the specimen. Light interacts with the specimen. The objective lens collects and recombines the altered beam. An ocular lens projects the enlarged image onto the eye or a camera. Processing optics produce contrast by phase contrast, differential interference, or polarization. B, Optical path in an inverted light microscope. C, Epi-illumination for fluorescence microscopy. The objective lens acts as the condenser to focus the exciting, short-wavelength light (green, in this example) on the specimen. Fluorescent molecules in the specimen absorb exciting light and emit longer-wavelength light (red, in this example). The same objective lens collects emitted long-wavelength light. A dichroic mirror in the light path reflects exciting light and transmits emitted light. An additional filter (not shown) blocks any short-wavelength light from reaching the viewer. D, Optical path in a transmission electron microscope. Electromagnetic lenses carry out the same functions as glass lenses in a light microscope. For visual observations, the electrons produce visible light from a fluorescent screen.

The classic light microscopic method is bright field, whereby the specimen is illuminated with pure white light. Most cells absorb very little visible light and thus show little contrast with bright-field illumination (Fig. 6-2A). For this reason, staining is used to increase light absorption and contrast. Because staining makes it difficult to see through thick tissues, specimens must also be relatively thin, about 1 mm for critical work. Slides for histologic and pathological study are produced by fixing cells with cross-linking chemicals, embedding them in paraffin or plastic, making sections with a microtome (a device that cuts a series of thin slices from the surface of a specimen), and staining with a variety of dyes (for examples, see Figs. 28-2, 28-5, 28-6, 28-7, 29-3, 29-8, 32-1, and 32-2). Alternatively, thin slices may be taken from frozen tissue and then stained. In either case, the cells are killed by fixation or sectioning prior to observation.

Figure 6-2 comparison of methods to produce contrast. a–d, Micrographs of a spread mouse 3T3 cell grown in tissue culture on a microscope slide, then fixed and stained with rhodamine-phalloidin, a fluorescent peptide that binds actin filaments. Contrast methods include bright field (A), phase contrast (B), differential interference contrast (C), and fluorescence (D). E–H, Micrographs of myofibrils isolated from skeletal muscle. Contrast methods include bright field (E), phase contrast (F), differential interference contrast (G), and polarization (H). The A-bands, consisting of parallel thick filaments of myosin (see Fig. 39-3), appear as dark bands with phase contrast and are birefringent (either bright or dark, depending on the orientation) with polarization.

(A–D, Courtesy of R. Mahaffy, Yale University, New Haven, Connecticut.)

Observations of live cells require other methods to produce contrast. In every case, these methods are also useful for fixed cells. Phase-contrast microscopy generates contrast by interference between light scattered by the specimen and a slightly delayed reference beam of light. Small variations in either thickness or refractive index (speed of light) can be detected, even within specimens that absorb little or no light (Fig. 6-2 B). Differential interference contrast (DIC) produces an image that looks as though it is illuminated by an oblique shaft of light (Fig. 6-2 C). What actually happens is that two nearby beams interfere with each other, producing contrast in proportion to local differences (gradient) in the refractive index across the specimen. Thus, a vesicle with a high refractive index (slow speed of light) in cytoplasm will appear light on one side (where the refractive index is increasing with respect to the cytoplasm) and dark on the other (where the refractive index is decreasing).

Fluorescence microscopy requires a fluorescent dye or protein in the specimen. Remarkable sensitivity makes fluorescence microscopy a powerful tool. Under favorable conditions, single fluorescent dyes or fluorescent protein molecules can be imaged. When a fluorescent molecule absorbs a photon of light, an electron is excited into a higher state. Nanoseconds later, a longer-wavelength (lower-energy) photon is emitted when the electron falls back to its ground state. For example, the fluorescent dye rhodamine absorbs green light (shorter wavelength) and emits red light (longer wavelength). Fluorescence microscopes use filters and special dichroic mirrors that reflect short wavelengths of light used to illuminate and excite fluorescent specimens but transmit the longer-wavelength emitted fluorescent light into the imaging system (camera). Strategically placed emission filters remove the exciting light reflected by the specimen so that only the fluorescent regions of the specimen appear bright. To provide fluorescence, a purified lipid, protein, or nucleic acid can be labeled with a fluorescent dye and injected into a live cell, where it will seek its natural location (see Figs. 37-6 and 38-9). Molecules labeled with a fluorescent dye can also be used to locate a target in a fixed and permeabilized cell. A powerful version of this strategy uses antibodies, proteins produced by the immune system (see Fig. 28-9), to react with specific molecular targets. Antibodies are tagged with fluorescent dyes and used to localize molecules in fixed cells by fluorescence microscopy (Fig. 6-3 E). This is called immunofluorescence. Another strategy is to label an oligonucleotide with a fluorescent dye to probe for nucleic acids with complementary sequences in fixed cells (see Fig. 13-15). Yet another approach is to localize individual structures, such as actin filaments, with a fluorescent dye attached to a small peptide that binds tightly to these filaments (Fig. 6-2 D).

Figure 6-3 fluorescence microscopy methods. A–C, Light micrographs of live fission yeast expressing GFP fused to myosin-I. A, Differential interference contrast (DIC). B, Standard wide-field fluorescence of the same cells. C, Stereo pair of a three-dimensional reconstruction of a stack of optical sections made by deconvolution of wide-field images. Removal of out-of-focus blur improves the resolution and contrast of small patches enriched in myosin-I. A stereo view is obtained by focusing your left eye on the left image and right eye on the right image. This can be achieved by holding the micrographs close to your eyes and then gradually withdrawing the page about 12 inches. D, Confocal fluorescence micrograph of fission yeast cells showing red microtubules and green Tea 1 protein (a protein involved in determining cell shape). This thin optical section eliminates the blur from fluorescence in other planes of focus. E–F, Fluorescence recovery after photobleaching. E, A fibroblast cell in tissue culture stained with fluorescent antibodies for the Golgi apparatus (yellow) and microtubules (green) and with the fluorescent dye DAPI for DNA (blue). F, A series of fluorescence micrographs of a fibroblast cell expressing GFP-galactosyltransferase, which concentrates in the Golgi apparatus. The GFP in a bar-shaped zone is bleached with a strong pulse of light, and the fluorescence is followed over time. After 2 minutes GFP-galactosyltransferase redistributes by lateral diffusion in the membranes to fill in the bleached zone.

(A–C, From Lee W-L, Bezanilla M, Pollard TD: Fission yeast myosin-I, Myo1p, stimulates actin assembly by Arp2/3 complex and shares functions with WASp. J Cell Biol 151:789–800, 2000. D, Courtesy of Hilary Snaith and Kenneth Sawin, University of Edinburgh, Scotland. E–F, Courtesy of J. Lippincott-Schwartz, N. Altan, and K. Hirschberg, National Institutes of Health, Bethesda, Maryland.)

The discovery of proteins whose amino acid sequence renders them naturally fluorescent, such as green fluorescent protein (GFP) from jellyfish, made fluorescence microscopy immensely valuable for observation of individual proteins in live cells. Typically, DNA-encoding GFP is joined to one end of the coding sequence for a cellular protein and introduced into cells, which then synthesize a fusion protein consisting of GFP linked to the protein of interest. GFP fluorescence marks the fusion protein wherever it goes in the cell and can be quantified to determine how many labeled molecules reside in a particular cellular location (Fig. 6-3). Ideally, the coding sequence for GFP fusion protein is inserted into the genome of the test cell in place of the wild type gene, and the fusion protein is shown to function normally by genetic or biochemical experiments. Where this is difficult or impossible (e.g., in most studies of metazoan cells), the GFP fusion protein can be produced from exogenous DNA or RNA introduced into the cell. Mutations in GFP can change its fluorescence properties, providing probes in a range of colors and with differing sensitivities to distinct biochemical parameters in the cell, such as pH, Ca²⁺ concentration, and kinase activity. When attached to different protein types, these probes allow two or more protein species to be visualized simultaneously in the same cell and can serve as “biosensors” to measure changes in the intracellular environment and in a protein’s behavior/interactions.

Dark-field microscopy and polarization microscopy have specialized uses in biology. In dark-field microscopy, the specimen is illuminated at an oblique angle so that only light scattered by the specimen is collected by the objective lens. Recall how easy it is to detect tiny dust particles in a beam of light in a dark room. The contrast is so great that single microtubules stand out brightly from the dark background. However, for the images to be interpretable, the specimen must be very simple, much simpler than a cell. A dark-field image of something as complicated as cytoplasm is very confusing, owing to multiple overlapping objects that scatter light.

Like dark-field microscopy, polarization microscopy produces a bright image on a dark background. When a specimen is viewed between two crossed polarizing filters, only light whose polarization state is modified by the specimen will pass through the second polarizer to the image. Polarization microscopy relies on a specimen’s crystalline order, or birefringence, to provide contrast. Birefringent specimens, such as filaments in striated muscle (Fig. 6-2 H) or microtubules in a mitotic spindle, are aligned enough that polarized light, oriented so that it vibrates along the length of the polymers, passes through more slowly than does light vibrating perpendicular to the polymers (much as a knife cuts through meat faster with the grain than across it). Most cells do not have sufficient birefringence to produce a useful image with a conventional polarization microscope. New methods are making this approach more applicable for future work.

Computer processing can greatly enhance contrast and remove optical artifacts from images. For example, computer-enhanced DIC can image single microtubules (see Fig. 34-7). New methods of image processing can even improve detection beyond the classic limit determined by the wavelength of light (about 0.2 mm with green light). A processing method called deconvolution produces clear fluorescence images of thick specimens by using an iterative computer process to restore light that is blurred out of focus to its proper focal plane. Starting with a stack of blurry images taken at different focal planes all the way through the specimen using a traditional wide-field microscope, this method produces a remarkably detailed three-dimensional image in sharp focus throughout (Fig. 6-3 C).

Confocal microscopy also produces thin optical sections of fluorescent specimens. Rather than illuminating with a wide beam of light, this method uses a point of laser light sharply focused in all three directions: x, y, and z. The point of light is scanned across the specimen in a raster pattern (checkerboard pattern, like the electron beam in a TV) to excite fluorescent molecules. Light emitted at each consecutive point in the specimen passes through a pinhole placed next to the detector to remove any light that does not come directly from each focal point. A computer reassembles the image from the fluorescence at each point in this checkerboard of fluorescence signals (Fig. 6-3 D; see also Figs. 14-17 and 14-18 Figs. 13-12, 14-2, and 44-23). A series of confocal images taken at different planes of focus can be used for three-dimensional reconstructions.

Electron Microscopy

A transmission electron microscope (Fig. 6-1 D) can resolve points below 0.3 nm, but the practical resolution is usually limited by damage to the specimens from the electron beam and the methods used to prepare specimens. Historically, the most common method used to prepare cells for electron microscopy was to fix the specimen with chemicals, embed it in plastic, cut the specimen into thin sections, and stain the sections with heavy metals (Fig. 6-4 F). With this technique, the resolution is limited to about 3 nm, but that is sufficient to bridge the gap between light microscopy and molecular structures. During the heyday of electron microscopy in cell biology, between 1950 and 1970, thin sections revealed most of what is known about the organization of organelles in cells.

Figure 6-4 electron micrographs. A, Scanning electron micrograph of developing flowers of the Western mountain aster. B–F, Transmission electron micrographs. B, Myosin-II minifilaments on a thin carbon film prepared by negative staining with uranyl acetate. C, Myosin-II minifilaments on a mica surface prepared by rotary shadowing with platinum. D, Freeze-fracturing. The cleavage plane passed through the cytoplasm and then split apart the two halves of the bilayer of the nuclear envelope. This fractured surface was then shadowed with platinum. The cytoplasm is in the upper left. Nuclear pores are prominent in the nuclear envelope. E, A cultured cell prepared by rapid freezing, fracturing, deep etching, and rotary shadowing with platinum. Membranes of the endoplasmic reticulum stand out against the porous cytoplasmic matrix. F, Thin section of a plasma cell, an immune cell specialized to synthesize and secrete antibodies.

(A, Courtesy of J. L. Bowman, University of California, Davis. C, Courtesy of J. Sinard, Yale University, New Haven, Connecticut. E, Courtesy of John Heuser, Washington University, St. Louis, Missouri. D–F, Courtesy of Don W. Fawcett, Harvard Medical School, Boston, Massachusetts.)

The highest resolution is attained with regular specimens, such as two-dimensional protein crystals rapidly frozen and viewed while embedded in a thin film of vitreous (i.e., amorphous, noncrystalline) ice (see Fig. 5-11A). This is called cryoelectron microscopy be-cause the stage holding the frozen specimen is cooled to liquid nitrogen temperature. Electron micrographs and electron diffraction of frozen crystals have produced structures of bacteriorhodopsin (see Fig. 7-8), aquaporin water channels (see Fig. 10-15), and tubulin (see Fig. 34-4) at resolutions of 3 to 4 nm. Computational image processing methods are used to calculate the three-dimensional structure of proteins in these regular specimens. These methods are similar to those used to calculate electron density maps from X-ray diffraction patterns (see Fig. 3-10). Although the resolution is limited and data collection is tedious in electron crystallography, electron microscopic images have the advantage of containing the phase information that is often difficult to ascertain with X-ray diffraction.

Electron microscopy is valuable for studying protein polymers and other large macromolecular specimens at less-than-atomic resolution. Diverse methods are used to prepare specimens and impart contrast. One way is to freeze filaments or macromolecular assemblies in vitreous ice, as described earlier (see Figs. 34-7 and 36-4A). A second is negative staining, whereby specimens are dried from aqueous solutions of heavy metal salts (Fig. 6-4 B). A shell of dense stain encases particles on the surface of a thin film of carbon and can preserve structural details at a resolution of about 1 nm. Alternatively, macromolecules dried on a smooth surface can be shadowed with a thin coat of metal evaporated from an electrode (Fig. 6-4 C). A variation of this approach that improves preservation is to freeze specimens rapidly, evaporate the ice surrounding the molecules, and then apply a coat of platinum (see Figs. 30-4 and 34-11).

Computer image processing of micrographs of certain types of structures can yield an average three-dimensional reconstruction of a molecular structure. Particles with helical symmetry, such as actin filaments (see Fig. 33-7) and microtubules (see Fig. 34-5), are analyzed by an image-processing method called deconvolution to reconstruct the three-dimensional structure. Single particles may also be reconstructed by first classifying images of thousands of randomly oriented particles into categories corresponding to different views. Then, an average three-dimensional structure is calculated computationally from this ensemble. One example is the Sec61p translocon associated with a ribosome (see Fig. 20-6). More recently, computing advances have led to the development of electron microscope tomography, in which many pictures are taken of a relatively thick specimen from different angles (by tilting the speci-men inside the microscope). Superimposition blurs each picture, but when they are merged together into a three-dimensional map, structures as complex as entire cells can be visualized at a resolution of a few nanometers.

Cells and tissues can also be frozen rapidly and prepared for electron microscopy without chemical fixation. In the freeze-fracture method, the frozen specimen is cleaved to expose the inside of the cells, and exposed surfaces are rotary-shadowed with a thin coat of platinum. This surface coat is then viewed by using a transmission electron microscope (Fig. 6-4 D). Frequently, the cleavage plane splits lipid bilayers in half to reveal proteins embedded in the plane of the membrane. If some of the frozen water in a fractured specimen is evaporated from the surface before shadowing, three-dimensional details of deeper parts of the cytoplasm can be revealed. A variation of this method involves extracting soluble molecules and membranes with mild detergents before freezing, fracturing, evaporating frozen water, and rotary-shadowing (Fig. 6-4 E; see also Fig. 1-13).

A scanning electron microscope (SEM) can be used on thicker specimens, such as whole cells or tissues that have been fixed, dried, and coated with a thin metal film. Here, an electron beam scans a raster pattern over the surface of specimens, and secondary electrons emitted from the surface at each point are collected and used to reconstruct an image (Fig. 6-4 A). The resolution of conventional SEM is limited, but nonetheless valuable, for studying surface features of cells and their three-dimensional relationships in tissues. SEMs that use special high-energy (field emission) guns to produce the electron beam have greatly improved resolution, and these have been very useful for studying cellular substructures, such as nuclear pores (see Fig. 14-6B).

Model Organisms

Ideal model organisms have completely sequenced genomes and facile methods to manipulate the genes, including replacement of a gene with a modified gene, by the process of homologous recombination. Haploid organisms with one copy of each chromosome after mitotic division are particularly favorable for detecting the effects of changes in genes, called mutations (Box 6-2). It is useful for a haploid organism to have a diploid stage with two copies of each chromosome and a sexual phase, during which meiotic recombination occurs between the chromosomes from the two parents. (See Fig. 45-7 for details on recombination.) This allows one to construct strains with a variety of mutations and facilitates mapping mutations to a particular gene. In addition, diploids carrying a lethal mutation of a gene that is essential for life can be propagated, provided that the mutation is recessive.

Budding yeast and fission yeast meet all of these criteria, so they are widely used to study basic cellular functions. These free-living haploid organisms have a tractable diploid stage in their life cycles. Moving between haploid and diploid stages greatly simplifies the process of creating and analyzing recessive mutations. This is important because most loss-of-function mutations are recessive. Even before their genomes were sequenced, the availability of yeast for genetic, biochemical, and microscopic analysis revolutionized research in cell biology. However, yeast are solitary cells with specialized lifestyles.

Multicellular organisms are required to study the development and function of tissues and organs. Flies, nematode worms, mice, and humans share many ancient, conserved genes that control their cellular and developmental systems, so flies and worms are popular for basic studies of animal development and tissue function. However, vertebrates have evolved a substantial number of new gene families (roughly 7% of total genes) and a large number of new proteins by rearranging ancient domains in new ways. Therefore, mice are used for experiments on specialized vertebrate functions, especially those of the nervous system, despite being more difficult to work with than flies and worms are. Although not an experimental organism, humans are included on this list because much can be learned by analysis of human genetic variation and its relationship to disease. Humans are, of course, much more eloquent than the model organisms when it comes to describing their medical problems, many of which have a genetic basis that can be documented by analysis of pedigrees and DNA samples. Arabidopsis is the most popular plant for genetics because its genome is small, reproduction is relatively rapid, and methods for genetic analysis are well developed. Its genome was the first of a plant to be completely sequenced. One drawback is the lack of methods to replace genes by homologous recombination (see later section).

By focusing on a limited number of easy-to-use model organisms, biological research raced forward in the last quarter of the 20th century. This focus does have liabilities. For one, these organisms represent a very limited range of lifestyles. Thousands of other solutions to survival exist in nature, and they tend to be ignored. At the cellular level, these liabilities are less severe, since most cellular adaptations are ancient and shared by most organisms.

Cell Culture

Regardless of the species to be studied, growing large populations of isolated cells for biochemical analysis and microscopic observation is helpful. This is straightforward for the unicellular organisms such as fungi or bacteria, which can be grown suspended in a nutrient medium. These organisms can also be grown on the surface of gelled agar in a petri dish. When single cells are dispersed widely on an agar surface, each multiplies to form a macroscopic colony, all descendents of a single cell. This family of cells is called a clone.

For multicellular organisms, it is often possible to isolate single live cells by dissociating a tissue with proteolytic enzymes and media that weaken adhesions between the cells. Many but not all isolated cells can be grown in sterile media, a method called tissue culture or cell culture. Terminally differentiated cells such as muscle or nerve cells do not reenter the cell cycle and grow. Cells that are predisposed to grow in the body including fibroblasts (see Fig. 28-4) and endothelial cells from blood vessels (see Fig. 30-13) will grow if the nutrient medium is supplemented with growth factors to drive the cell cycle (see Fig. 41-7). This is accomplished by adding fetal calf serum, which contains a particularly rich mixture of growth factors. Some cultured cells grow in suspension, but most prefer to grow on a surface of plastic or glass (Fig. 6-2), often coated with extracellular matrix molecules for adhesion (see Fig. 30-11). This is the origin of the term in vitro, meaning “in glass,” used to describe cell culture. Normal cells grow until they cover the artificial surface, when contacts with other cells arrest further growth. Dissociation and dilution of the cells onto a fresh surface allow growth to resume. Most “primary cells” isolated directly from tissues divide a limited number of times (see Fig. 12-15). Primary cells can become immortal, either through mutations or transformation by a tumor virus that overcomes cell cycle controls. Such immortal cells are called cell lines. Similar changes allow cancer cells to grow indefinitely. HeLa cells are a famous cell line derived from Henrietta Lax, an African-American patient with cervical cancer. HeLa cells have been growing in laboratories for more than half a century.

A variation on cell culture is to grow a whole organ or part of an organ in vitro. The requirements for organ culture are often more stringent than those for growing individual cells, but the method is used routinely for experiments on slices of brain tissue and for studying the development of embryonic organs.

Classical Genetics: Identification of Genes through Mutations

The approach in classical genetics is to identify mutations that compromise a particular cellular function and then to find the responsible gene(s). This approach is extremely powerful, especially when little or nothing is known about a process or when the gene product (usually a protein) is present at low concentrations. Yeast genetic studies have been spectacularly successful in mapping out complex pathways, including identification of the proteins that regulate the cell cycle (see Chapters 40 to 44) and the proteins that operate the secretory pathway (see Chapter 21).

Because one generally does not know the relevant genes in advance, it is important that mutations are introduced randomly into the genome and, ideally, limited to one mutation in each organism tested. A prerequisite for such a genetic screen is a good assay for the biological function of interest. Simplicity and specificity are essential, as interesting mutations may be rare, and much effort may be expended characterizing each mutation. The assay may test the ability to grow under certain conditions, drug resistance, morphologic changes, cell cycle arrest, or abnormal behavior. Mutations arise spontaneously at low rates, so often a chemical (e.g., ethyl methyl sulfonate or nitrosoguanidine) or radiation is used to increase the frequency of damage. Another approach is to insert an identifiable segment of DNA randomly into the genome. This simultaneously disrupts genes and marks them for subsequent analysis. Because the damage is random, the trick is to find the particular damage that changes the physiology of the organism in an informative way.

Haploid organisms are favorable for detecting mutations because damage to the single copy of a relevant gene will alter function, and either a loss of function or a gain of function can be detected with suitable test conditions (i.e., the ability to grow under certain conditions), biochemical assay, or morphologic assay. A disadvantage is that haploid organisms are not viable following the loss of function of an essential gene. Selecting for conditional mutant alleles allows the haploid organism to survive mutation of an essential gene under permissive conditions (e.g., low temperatures) but not under restrictive conditions (e.g., high temperatures). A further advantage of haploid organisms is that one can usually identify the mutated gene by a complementation experiment. Mutant cells are induced to take up a plasmid library containing fragments of the wild-type genome or cDNAs. Plasmids are circular DNA molecules that can be propagated readily in bacteria and, if suitably designed, in eukaryotes as well. Plasmids carrying the wild-type gene will correct loss-of-function mutations, allowing colonies of cells to grow normally. Plasmids complementing the mutation are isolated and sequenced. Additional tests are required to confirm that the wild-type gene in the plasmid corresponds to the mutant gene, as in some cases, raising the level of an unrelated gene can rescue a mutant phenotype. However, once this is done, the mutant gene can be isolated and sequenced to determine the nature of the damage. This complementation test can also be used to discover genes from other species that correct the mutation in the model organism. For example, genes for human cell cycle proteins can complement many cell cycle mutations in yeast (see Chapter 40). For gain-of-function mutations, a gene library from the mutant cell is inserted into plasmids, which are then tested for their ability to cause the altered phenotype in wild-type cells.

Genetics in obligate diploid organisms is more complicated. Many mutations will appear to have no effect, provided that the corresponding gene on the other chromosome functions normally. These recessive mutations produce a phenotype only after crossing two mutant organisms, yielding 25% of offspring with two copies of the mutant gene. (Consult a genetics textbook for details on Mendelian segregation.) Other mutations will yield an altered phenotype even when only one of the two genes is affected. These dominant mutations include simple loss of function when two wild-type genes are required to make sufficient product for normal function (called haplo-insufficiency); production of an altered protein that compromises the formation of a large assembly by normal protein subunits produced by the wild-type gene (called dominant negative); and production of an unregulated protein that cannot be controlled by partners in the cell (another type of dominant negative).

The classic method for identifying a mutated gene is genetic mapping. One observes the frequency of recombination between known markers and the mutation of interest in genetic crosses. This is usually sufficient to map a gene to a broad region of a parti-cular chromosome. If a complete genome sequence is available, the database of sequenced genes in the area highlighted by mapping is examined to look for sensible candidate genes. These candidates can then be studied to establish which one carries the mutation. Another approach is to make the mutation by inserting a piece of DNA (called a transposable element) randomly into the genome. If one of these insertions causes a mutant phenotype, the transposable element may be recovered together with some of the surrounding chromosome, which is sequenced to identify the disrupt-ed gene.

Once a gene required for the function of interest is sequenced (see Fig. 3-16), the primary structure of the protein (or RNA) is deduced from translating the coding sequence with a computer. Much can be learned by identifying RNAs or proteins with similar sequences or domains in the same or other species, particularly if something is known about the function of the corresponding gene product. Protein can often be expressed from a cDNA copy of the mRNA, tested for activity and binding partners, and (when fused to GFP or when used to make an antibody) localized in cells.

Further insights regarding function are often obtained by disruption of a gene. Genomic DNA can be used to construct a plasmid that contains two substantial regions of the chromosome (usually several thousand base pairs) flanking either the entire gene to be targeted or a significant portion thereof. In the plasmid, these “targeting” regions flank a selectable marker, for example, a gene encoding resistance to a particular drug that would normally kill the cells. If introduced into cells capable of homologous recombination, the targeting regions can recombine into the chromosome, thereby replacing the DNA between the targeting sequences with the selectable marker and disrupting the gene, ideally creating a null mutation. The selectable marker is used to enrich for cells with the disrupted gene. Gene disruption is readily accomplished in yeast and, with somewhat more difficulty, in vertebrate cells but is more complicated in flies, in which this gene-targeting technology is less well developed. Fortunately, an alternative method called RNAi (for RNA interference) can lower the levels of particular mRNAs from many cells, including those in worms and cultured cells of flies and humans (discussed later, and see Fig. 16-12 for details).

Genomics and Reverse Genetics

Thanks to large-scale DNA sequencing projects, nearly complete sequences of the coding regions of the most popular experimental organism are now available (see Figs. 2-4 and 2-9). When fully annotated (i.e., all sequences coding for genes have been identified and catalogued), these genome sequences will be the definitive inventory of genes. This is easier said than done, as accurate and complete identification of genes in raw sequence data is still challenging (see Chapter 12). The task has been aided by constructing databases containing millions of sequence fragments derived from cDNA copies of expressed genes (expressed sequence tags, or ESTs), which help to document the diversity of products created by transcription and RNA processing (see Chapter 15).

Nevertheless, even before genome annotation is complete, these sequences make possible a new approach for relating genes to biological function. Given the sequence of a gene of interest, the initial strategy is to search computer databases for proteins with similar sequences and known functions to try to predict what the protein might do. This is surprisingly fruitful, as many genes occur as extended families. First, one scans the protein sequence for conserved sequence motifs (regions of a few to several hundred amino acid residues). To accomplish particular tasks, for example, to be a protein kinase, proteins use motifs that arose early in evolution and are now widely scattered throughout the genome (see Fig. 25-4). Dozens of motifs are now known (and more are discovered daily), so finding such a motif in your protein can reveal that it binds to phosphorylated tyrosine, is an enzyme that methylates other proteins, or has one of the dozens of functions that are ascribed to particular motifs. Once predicted sequences have been analyzed, one can check when and where the gene is expressed in the organism, test the consequences of deleting the gene, or test for interactions of the protein with other proteins (see later section). These tests can be done one gene at a time or on a genomewide scale. For example, investigators created strains of budding yeast lacking each of the 6000 genes and tested for interaction of the products of each of these genes with the products of all other genes. These preliminary screening tests often yield some clues about function. Ultimately, however, function is understood only when representatives of each protein family are studied in detail by the biophysical, biochemical, and cellular methods described in the following sections.

Reverse genetics refers to the process of starting with a known gene and selectively disrupting its function. One common approach used in yeasts is gene disruption, described previously. For metazoans, gene disruption is also used, but the most widely used method of reverse genetics is RNAi (discussed later in the chapter in the section titled “Physiological Testing”).

Biochemical Fractionation

The biochemical approach (to the inventory) is to purify active molecules for analysis of structure and function. This requires a sensitive, quantitative assay to detect the component of interest in crude fractions, an assay to assess purity, and a battery of methods to separate the molecule from the rest of the cellular constituents. Assays are as diverse as the processes of life. Enzymes are often easy to measure. Many molecules are detected by binding a partner molecule. For example, nucleic acids bind complementary nucleotide sequences and sequence-specific regulatory proteins; receptors bind ligands; antibodies bind their antigens; and particular proteins bind partner proteins. More difficult assays reconstitute a cellular process, such as membrane vesicle fusion, nuclear transport, or molecular motility. Devising a sensitive and specific assay is one of the most creative parts of this approach. A second prerequisite for purification is a simple method for assessing purity. Various types of gel electrophoresis often work brilliantly (Box 6-3 and Fig. 6-5).

Figure 6-5 gel electrophoresis. A, Schematic diagram showing a (generic) gel with three sample wells and an electric field. B, Agarose gel electrophoresis of DNA samples stained with ethidium bromide. The lane on the left shows size standards. The middle lane has a bacterial plasmid, a supercoiled (see Fig. 3-18) circular DNA molecule carrying an insert (Fig. 6-8 provides details). The right lane has the same plasmid digested with a restriction enzyme that cleaves the DNA twice, releasing the insert. Although smaller than the circular plasmid, the empty vector runs more slowly on the gel because the linear DNA offers more resistance to movement than the supercoiled circular plasmid. C, Polyacrylamide gel electrophoresis of the Arp2/3 complex, an assembly of seven protein subunits involved with actin polymerization (see Fig. 33-13). All three samples are identical. In the left lane, the proteins are stained with the nonspecific protein dye Coomassie blue. The proteins in the other two lanes were transferred to nitrocellulose paper; each reacted with an antibody to one of the subunit proteins (ARPC2 and ARPC1). The position of the bound antibody is determined with a second antibody coupled to an enzyme that produces light and exposes a piece of film black. This method is called chemiluminescence.

(B, Courtesy of V. Sirotkin, Yale University, New Haven, Connecticut. C, Courtesy of H. Higgs, Dartmouth Medical School, Hanover, New Hampshire.)

With a functional assay and a method to assess purity, one sets about purifying the molecule of interest. Highly abundant constituents, such as actin or tubulin, may require purification of only 20- to 100-fold, but many important molecules, such as signaling proteins and transcription factors, constitute less than 0.1% of the cell protein, so extensive purification is required.

First, the cell is disrupted gently to avoid damage to the molecule of interest. This may be accomplished physically by mechanical shearing with various types of homogenizers or, where appropriate, chemically, with mild detergents that extract lipids from cellular membranes. Next, the homogenate is centrifuged to separate particulate and soluble constituents. If the molecule of interest is soluble, it can be purified by sophisticated chromatography methods (Box 6-4 and Fig. 6-6) given sufficient starting material.

Figure 6-6 chromatography. A, Affinity chromatography to purify poly A mRNAs with poly dT attached to beads. A mixture of RNAs is extracted from cells and applied to the column in a buffer containing a high concentration of salt. Only poly (A)⁺ mRNA binds and is then eluted with buffer containing a low concentration of salt. (rRNA, ribosomal RNA.) B, Gel filtration chromatography separates molecules on the basis of size. Large molecules (blue) are excluded from the beads and travel through the column in the void volume outside the beads. Smaller molecules (green) penetrate the beads depending on their size. Tiny molecules (red), such as salt, completely penetrate the beads and elute in a volume (the salt volume) equal to the size of the bed of beads. Material eluting from the column is monitored for absorbance of ultraviolet light (260 nm for nucleic acids, 280 nm for proteins) to measure concentration and then collected in tubes in a fraction collector. C, Anion exchange chromatography. The beads in the column have a positively charged group that binds negatively charged molecules. A gradient of salt elutes bound molecules depending on their affinity for the beads. For cation exchange chromatography, the beads carry a negative charge.

If a cDNA copy of the mRNA for a protein of interest is available, rare proteins or modified proteins can often be expressed in large quantities in bacteria, yeast, or insect cells. An advantage of this approach is that mutations can be made at will, including substitution of one or more amino acids or deletion of parts of the protein. Addition of domains can be useful for characterizing the protein such as the following:

If the molecule of interest is part of an organelle, centrifugation can be used to isolate the organelle. Typically, the crude cellular homogenate is centrifuged multiple times at a succession of higher speeds (and therefore forces). Particles move in a centrifugal field according to their mass and shape. Large particles such as nuclei pack into a pellet at the bottom of the centrifuge tube at low speeds, whereas high speeds are required to pellet small vesicles. These pellets may be enriched in particular organelles but are never pure. Next, the impure pellet is centrifuged for many hours in a tube containing a concentration gradient of sucrose. In sedimentation velocity gradients, particles are centrifuged in a gradient of sucrose (e.g., 5% sucrose in buffer at the top of the tube, increasing to 20% sucrose at the bottom). Because the motion of particles in a centrifugal field depends on the square of the distance from the center of the rotor (think of a spinning ice skater), the farther down the tube the particle travels, the faster it will go. However, the motion of particles in a centrifugal force field also depends on the difference between their density and that of the surrounding medium. Thus, the increasing density of sucrose gradient tends to slow the particle down. Ideally, the two factors counteract one another so that the particle moves at a constant rate, yielding the best separation. In sedimentation equilibrium gradients, particles move until their density equals that of the gradient, at which point they move no farther, regardless of how long or hard they are spun. Membrane-containing organelles can be isolated in this way in sucrose gradients. The small differences in size and buoyant density among many of the membrane-bound organelles limit the resolution of subcellular fractionation by sedimentation velocity and sedimentation equilibrium, so additional methods are useful in purifying preparations of organelles. For example, antibodies specific for a molecule on the surface of an organelle can be attached to a solid support and used to bind the organelle. Contaminating material can then be washed away. Certain particles, such as DNA or RNA molecules, are denser than sucrose. They can be centrifuged to equilibrium in gradients of dense salts, such as cesium chloride.

Once a protein of interest has been purified, the path to its gene(s) is relatively direct. Traditionally, each constituent polypeptide was cut into fragments by proteolytic enzymes, after which these fragments were isolated by chromatography and their amino acid sequence determined by Edman degradation (see Chapter 3). Given part of the amino acid sequence, the corresponding gene can then be identified in a genomic data base or isolated by using oligonucleotide probes as the assay (see next section).

Increasingly, proteins are identified by mass spectrometry. Proteins are fragmented by cleavage at specific sites with a proteolytic enzyme, such as trypsin, and the masses of the fragments produced are measured exactly with a mass spectrometer. If the protein comes from an organism with a sequenced genome, the gene encoding the protein can be identified by matching the experimental masses of the tryptic fragments with masses of all the peptides predicted from the genome sequence. The sensitivity of these methods has been improved to the point where a stained protein band on a gel suffices to identify the corresponding gene. Alternatively, fragments of known weight are bombarded inside the mass spectrometer under conditions that break the peptide backbone. Analysis of the masses obtained by fragmenting a particular peptide can be used to deduce the sequence of that fragment. Another method starts with isolation of cellular components composed of a complex mixture of proteins such as the nuclear envelope. The sample is digested with the proteolytic enzyme trypsin, fractionated by chromatography, and analyzed by mass spectrometry. Routinely, hundreds of proteins can now be identified in complex cellular structures.

Isolation of Genes and cDNAs

A variety of methods make isolation of specific nucleic acids relatively routine. Genomic DNA is isolated from whole cells by selective extraction. mRNAs are purified by affinity chromatography, taking advantage of their polyadenylate (poly A) tails (see Fig. 16-3), which bind by base pairing to poly dT attached to an insolu-ble matrix (Fig. 6-6A). Because DNA is easier to work with than RNA (e.g., it can be cleaved by restriction endonucleases and cloned), RNAs are usually converted to complementary DNA (cDNA) by reverse transcriptase, a viral DNA polymerase that uses RNA as a template.

Several options exist to purify a particular DNA from a complex mixture:

Figure 6-7 polymerase chain reaction. from the top, double-stranded DNA with a sequence of interest is denatured by heating to separate the two strands. An excess of oligonucleotide primers complementary to the ends of the sequence of interest are added and allowed to bind by base pairing. DNA polymerase synthesizes complementary strands, starting from the primers. This cycle is repeated many times to amplify the sequence of interest. Use of a DNA polymerase from a thermophilic bacterium allows many cycles at high temperature without losing activity.

Figure 6-8 DNA CLONING. A, Cloning of a segment of DNA into a plasmid vector. The vector is a circular DNA molecule with an origin of replication (Ori) that allows it to replicate in a host bacterium. Most vectors also include one or more genes conferring antibiotic resistance—in this example, resistance to ampicillin (Amp). This enables one to select only those bacteria carrying a plasmid by the ability to grow in the presence of ampicillin. Vectors also contain a sequence of DNA with multiple restriction enzyme digestion sites (see part B) for the insertion of foreign DNA molecules. In this example, a single restriction enzyme, EcoR1, is used to cut both the source DNA and the plasmid vector, leaving both with identical single-strand overhangs. The ends of the insert and the cut vector anneal together by base pairing and are then covalently linked together by a ligase enzyme, forming a complete circle of DNA. Plasmids are introduced into bacteria, which are then grown on ampicillin to select those with plasmids. Colonies of bacteria are screened for those containing the desired insert using, for example, DNA probes for sequences specific to the gene of interest. Figure 6-5B shows gel electrophoresis of a plasmid carrying an insert before and after digestion with a restriction enzyme to liberate the insert from the vector. B, Sequence-specific cutting of DNA with restriction enzymes. EcoR1 and BamH1 are two of the hundreds of different restriction enzymes that recognize and cleave specific DNA sequences. Both of these restriction enzymes recognize a palindrome of six symmetrical bases. Note that these enzymes leave overhangs with identical sequences on both cut ends that are useful for base pairing with DNA having the same cut. Other restriction enzymes recognize and cut from 4 to 10 bases.

Once a gene or cDNA has been cloned, it is sequenced and used to deduce the sequence of the encoded protein. Of course, analysis of a DNA sequence cannot reveal posttranslational modifications of a protein, such as phosphorylation, glycosylation, or proteolytic processing. Such modifications, which are often critical for function, can be identified only by analysis of proteins isolated from cells. This analysis entails mass spectrometry or amino acid sequencing.

Cloned cDNAs are used to express native or modified proteins in bacteria or other cells for biochemical analysis or antibody production. This approach has two advantages. First, the quantity of protein produced is often far greater than that from the natural source. Second, cloned DNA can readily be modified by site-directed mutagenesis to make predetermined amino acid substitutions and other alterations that are useful for studying protein function (Fig. 6-9). The behavior of mutant proteins in cells can provide evidence for the role of a given protein in particular cellular functions. Thus, biochemical, genetic, and molecular cloning approaches may be applied collectively to reveal the function of proteins.

Figure 6-9 in vitro mutagenesis of cloned dna. This is one of several types of PCR methods used to change one or more nucleotides (the symbol * in this example) in a cloned gene using a primer with altered bases. In this particular method, primer 1 has the altered base and is used to duplicate the entire plasmid. Primer 2 is used to synthesize the whole plasmid from the other end. After amplification with both primers, the two ends are ligated together, and the plasmid is produced in quantity by growth in bacteria.

Primary Structure

DNA sequences are now determined by automated dye-termination methods (see Fig. 3-16). The same automated dye-termination methods, when applied to cDNAs, are used to deduce the sequence of proteins and structural RNAs. Protein sequencing by Edman degradation is still occasionally used to detect modified amino acids (see Fig. 3-3); however, mass spectrometry is faster and more sensitive.

Subunit Composition

Gel electrophoresis of many isolated proteins has revealed that they consist of more than one polypeptide chain. Their stoichiometry can be determined from the size and intensity of the stained bands on the gel, but the only way to determine the total number of subunits is to measure the molecular weight of the native protein or protein assembly. The definitive method is a sedimentation equilibrium experiment carried out in an analytical ultracentrifuge. A sample of purified material is centrifuged in a physiological salt solution at relatively low speed in a rotor that allows the measurement of the mass concentration from the top to bottom of the sample cell. At equilibrium, the sedimentation of the material toward the bottom of the tube is balanced by diffusion from the region of high concentration at the bottom of the tube. This balance between sedimentation and diffusion uniquely defines the molecular weight of the particle. A less direct approach to measuring the molecular weight of the native protein or protein assembly is to measure the sedimentation coefficient (the parameter relating the rate of sedimentation to the centrifugal force) during centrifugation at high speed and to measure the diffusion coefficient separately, most often by analytical gel filtration (Fig. 6-6B). These two parameters are used to calculate the molecular weight. (Note that neither measurement separately is sufficient to measure molecular weights, despite numerous assertions in the literature that they are sufficient!) An advantage of the latter approach is that it can be used with impure material, provided that an assay is available that is applicable to the two types of measurements. Light scattering can also be used to estimate molecular weights.

Atomic Structure

X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy are used to determine the structure of proteins and nucleic acids at atomic resolution (see Fig. 3-8). Although X-ray crystallography has determined structures as large as the ribosome (see Fig. 17-7) and viruses (see Figs. 5-11 and 5-14), some large structures are currently outside the size range of this high-resolution method. Alternatively, large structures can be studied by electron microscopy of single particles or regular assemblies. If available from crystallography or NMR, atomic structures of subunits can be fit into lower-resolution reconstructions of large assemblies made by electron microscopy (see Figs. 36-4 and 36-10). NMR avoids the requirement to crystallize the protein to be studied, but the protein must be soluble at high concentrations, and NMR is difficult for proteins larger than 20 kD.

Biochemical Methods

Once a molecule of interest has been purified, finding partners with which it functions in the cell is often the next step. This requires a method to separate the macromolecular complex containing the molecule being studied away from other cellular proteins. One approach is affinity chromatography with the probe molecule attached by a chemical crosslink to an insoluble support, such as small beads. A popular variation is to express a probe protein fused to GST that can be bound with high affinity to a small molecule attached to beads. A crude cellular extract is run through the column with immobilized probe molecules and washed. Then molecules bound to the probe are eluted with high salt, extremes of pH, specific ligands, or, if necessary, with denaturing agents, such as urea. Eluted proteins are analyzed by gel electrophoresis and identified with antibodies, sequencing, or mass spectrometry. Eluted nucleic acids are cloned and sequenced.

An alternative to column chromatography is to mix beads with attached probe molecules with a crude cellular extract and then isolate the beads with bound molecules by centrifugation into a pellet. Bound molecules are eluted for analysis. Varying the concentration of such beads is a simple way to measure the affinity of the probe for its various partners. Antibodies are frequently used to separate a protein and its partners from crude extracts. An antibody specific for the probe molecule can be attached directly or indirectly to a bead and used to bind the protein of interest along with any associated molecules. This is called immunoprecipitation.

Proteins tagged with combinations of peptides can be purified by affinity methods along with tightly associated proteins. A popular method called TAP (tandem affinity purification) tagging adds to any protein of interest DNA sequences encoding two different peptide epitopes separated by a cleavage site for a highly specific viral protease. The cell makes the doubly tagged protein. The tagged protein, together with associated proteins, is purified from a cellular extract using immobilized antibodies to the outermost tag. The TEV protease, which has no natural targets in the cell, cleaves the tagged protein from the immobilized antibody. Then an entirely different set of reagents permits a second round of purification using the remaining tag. Two successive affinity steps remove most proteins that bind nonspecifically to the protein of interest or the affinity reagents. This is a quick method to purify stable protein complexes from crude whole-cell lysates.

Genetics

Given a mutation in a gene of interest, two genetic tests are used to search for partners: (1) identification of a second mutation that ameliorates the effects of the pri-mary mutation (a suppressor mutation, Fig. 6-10A–B) and (2) identification of a second mutation that makes the phenotype more severe, often lethal (an enhanc-er mutation [Fig. 6-10 C–E]). A specialized class of enhancer mutations, called synthetic lethal mutations, is particularly useful in the analysis of genetic pathways in yeast. In this case, mutations in two genes in the same pathway, if present in the same cell, even as heterozygotes (i.e., each cell having one good and one mutant copy of each gene), cannot be tolerated, so the cell dies. It is thought that each mutation lowers the level of production of some critical factor just a bit and that the combination of the two effectively means that the output of the pathway is insufficient for survival. These tests can be made with existing collections of mutations by genetically crossing mutant organisms. Alternatively, one can seek new mutations created by a second round of mutagenesis. The results depend on the architecture of the particular pathway. If the products of the genes in question operate in a sequence, analysis of single and double mutants can often reveal their order in the pathway. For essential genes in haploid organisms, a conditional allele of the primary mutation simplifies the experiment. Synthetic interactions (suppression or lethality) may also be discovered by overproduction of wild-type genes on a plasmid. Caution is required in interpreting suppressor and enhancer mutations, given the complexity of cellular systems and the possibility of unanticipated consequences of the mutations.

Figure 6-10 analysis of genetic interactions between two genes, m and n. The sizes of the arrows indicate the level of function of the gene product, usually a protein. The phenotype is indicated for each example. Mutant phenotype means an altered function dependent on gene products M and N. In the diagram, the symbol ⁺ indicates a wild-type allele, the symbol * indicates a suppressor allele, and the symbol D indicates a null mutation. A, Bypass suppression. Gene products M and N operate in parallel, with M making the larger contribution. Loss of M yields a mutant phenotype because N alone does not provide sufficient function. Mutation N* enhances the function of N, allowing it to provide function on its own. B, Suppression by epistasis. Products M and N act in series on the same pathway. Loss of M function blocks the pathway. Mutation N* allows N to function without stimulation by product M. C, Interactional suppression. Function requires interaction of gene products M and N. Mutation M⁻ interferes with the interaction. Suppressor mutation N* allows product N* to interact with M⁻. D, Synthetic lethal interaction when null mutations in either M or N are viable. The products of genes M and N operate in parallel to provide function. N provides sufficient function in the absence of M (DM) and vice versa. Loss of both M and N is lethal. E, Synthetic lethal interaction when null mutations in either M or N are lethal. Products M and N function in series. N can provide residual function even when M is compromised by mutation M⁻, and vice versa. When both M and N are compromised (M⁻, N⁻), the pathway provides insufficient function for viability.

(Redrawn from Guarente L: Synthetic enhancement in gene interaction: A genetic tool comes of age. Trends Genet 9:362–366, 1993.)

Another approach to find protein partners is called a two-hybrid assay (Fig. 6-11). This assay depends on the observation that some activators of transcription have two modular domains with discrete functions: One domain binds target sites on DNA, and the other recruits the transcriptional apparatus (see Fig. 15-19). The target gene is expressed if both activities are present at the transcription start site, even if the activities are on two different proteins. For the two-hybrid assay, the coding sequence of the protein whose partners are to be identified is fused to the coding sequence of a yeast protein that recognizes a target DNA sequence upstream of a gene that provides the readout of the assay. This so-called bait protein is expressed constitutively in yeast cells. A plasmid library is constructed consisting of cDNA sequences of all possible interaction partners (“prey”), each fused to the coding sequence of an “activator domain” and a nuclear localization sequence. This library of “prey” proteins is introduced into the “bait” yeast strain. The readout gene is expressed if a “prey” protein binds the “bait” protein and recruits the transcriptional apparatus. Many variations of this assay exist. One produces an enzyme that makes a colored product, so colonies of yeast with interacting proteins can be identified visually. In another version, the target gene encodes a gene essential for production of a particular amino acid, so only cells with a bait-prey interaction will grow on agar plates lacking that amino acid. Putative interactions must subsequently be tested carefully to define specificity, as false-positive results are common. Moreover, some valid interactions are missed owing to false-negative results.

Figure 6-11 one version of the yeast two-hybrid assay for interacting proteins. Interaction between “bait” protein and “prey” protein (bottom) brings together the two halves of a transcription factor required to turn on the expression of b-galactosidase. The DNA-binding domain of the GAL4 transcription factor binds a specific DNA sequence: GAL UAS. Generally, a library of random cDNAs or gene fragments is used to express test prey proteins as fusions with the activation domain.

Large-Scale Screening with Microarrays

Microarrays display thousands of tiny spots on a glass slide, each with a particular DNA sequence or protein (Fig. 6-12). This allows many reactions to be monitored in parallel. One type of microarray has cDNAs or oligonucleotides for thousands of genes. Probing such an array with complementary copies of mRNAs from a test sample reveals which genes are expressed. This can be used to find partners, because expression of genes contributing proteins to a particular pathway is often coordinated as conditions change. For example, unfolded proteins in the lumen of the endoplasmic reticulum trigger the expression of nearly 300 genes for proteins of the endoplasmic reticulum (see Fig. 20-11). Microarrays of thousands of different proteins can be used to test for interactions. For example, reaction of protein arrays with each yeast protein kinase, one kinase per slide, identified the substrates phosphorylated by each kinase (Fig. 6-12B).

Figure 6-12 large-scale analysis of gene expression and kinase activity with microarrays. A, Gene expression. PCR was used to make cDNA copies of mRNAs from two parts of the human brain. The cDNAs from cerebral cortex mRNAs were labeled with a red fluorescent dye, whereas those from the cerebellum were labeled with a green fluorescent dye. A mixture of equal proportions of the two fluorescent cDNA preparations was reacted with 384 different known cDNAs arrayed in tiny spots on a glass slide. The fluorescence-bound cDNAs were imaged with a microscopic fluorescent scanner similar to a confocal microscope. Yellow spots bound equal quantities of cDNAs from the two sources. Red spots bound more cDNA from the cortex, indicating a higher concentration of those mRNAs. Green spots bound more cDNA from the cerebellum, indicating a higher concentration of those mRNAs. B–C, Large-scale identification of substrates for a protein kinase. Thousands of different budding yeast proteins tagged with GST- and 6 histidines were overexpressed in yeast and purified by affinity chromatography. Each protein was spotted in duplicate on a glass slide, a small portion of which is shown here. B, The amount of bound protein in each spot was detected with a fluorescent antibody to GST (indicated by varying intensity of fluorescence from dark red to white). C, The slide was incubated with a yeast kinase in the presence of ³³P-ATP. Radioactive phosphorylated proteins were detected as pairs of dark spots by autoradiography. One pair is boxed.

(A, Courtesy of C. Barlow and M. Zapala, Salk Institute, La Jolla, California. B–C, Courtesy of Geeta Devgan and Michael Snyder, Yale University, New Haven, Connecticut. Reference: Zhu H, Bilgin M, Bangham R, et al: Global analysis of protein activities using proteome chips. Science 293:2101–2105, 2001.)

Reconstitution of Function from Isolated Components

The classic biochemical test of function is reconstitution of a biological process from purified components. This involves creating conditions in the test tube in which isolated molecules can perform a complex process normally carried out by a cell. The difficulty of the task depends on the complexity of the function. Successful reconstitution experiments reveal the molecular requirements and mechanisms involved in a process. Examples of successful tests include reconstitution of ion channel function in pure lipid membranes (see Chapter 10), protein synthesis and translocation of proteins into the endoplasmic reticulum (see Fig. 20-7), and motility of bacteria powered by assembly of actin filaments (see Fig. 37-12).

Anatomic Tests

No biological process can be understood without knowledge of where the components are located in the cell. Often, cellular localization of a newly discovered molecule provides the first clue about its function. This accounts for why cell biologists put so much effort into localizing molecules in cells. Cell fractionation, fluorescent antibody staining, and expression of GFP fusion proteins are all valuable approaches, illustrated by numerous examples in this book. For more detailed localization, antibodies can be adsorbed to small gold beads and used to label fixed specimens for electron microscopy (see Fig. 29-7).

GFP fusion proteins are particularly valuable because of the ease of their construction and expression and because they can be used to monitor both the behavior and dynamics of molecules within living cells. However, it should always be kept in mind that attaching GFP may affect either the localization or function of the protein being tested. Demonstration that a GFP fusion protein is fully functional, that is, that it can replicate the parent protein’s biochemical and biophysical properties, can be done only by genetic replacement of the native protein with the GFP fusion protein. This is routinely done in yeast but rarely for vertebrate proteins, as the required genetics are difficult or impossible. Instead, correct function is inferred from the fusion protein exhibiting morphologic, biochemical, and biophysical properties similar to those of the native protein. This is better than nothing but incorporates an element of wishful thinking.

The use of GFP fusions to study cellular dynamics has yielded many surprises, as structures that were thought to be inert have turned out to be remarkably dynamic. One powerful technique is to photobleach the GFP fusion protein in one part of the cell and to observe how the fluorescent proteins in other parts of the cell redistribute with time (fluorescence recovery after photobleaching, or FRAP; see Fig. 6-3E). The speed of fluorescence recovery into the photobleached area provides information on the mobility of the fusion protein (i.e., whether it diffuses freely, is immobilized on a scaffold, or is actively transported) and its interaction properties within the cell (see Figs. 7-11 and 14-4). These properties play important roles in how a protein functions within a cell, which cannot be determined by merely observing the protein’s steady-state distribution.

Proteins and other cellular components, including DNA, RNA, and lipids, can be labeled with fluorescent dyes to study their intracellular localization and dynamics. Fluorescent RNAs and proteins can be microinjected into cells. Fluorescent lipids can be inserted into the outer leaflet of the plasma membrane in living cells; from there, they move to appropriate membranes and then mimic rather faithfully the behavior of their natural lipid counterpart.

Physiological Tests

Although often obscured by technical jargon, just three methods are available to test for physiological function: (1) reducing the concentration of active protein (or other molecule), (2) increasing the concentration of active molecule, and (3) replacing a native protein with a protein that has altered biochemical properties. Biochemical, pharmacological, and genetic methods are available for each test, the genetic methods often yielding the cleanest results. These experiments are most revealing when robust assays are available to measure quantitatively how the cellular process under investigation functions when the concentration of native molecule is varied or an altered molecule replaces the native molecule. When done well, these experiments provide valuable constraints for quantitative models of biological systems, as described in the next section.

The definitive way to reduce the concentration of active protein or RNA is to prevent its expression. This option is available if the molecule is not required for viability. If a protein is essential, one can replace it with an altered version that is fully active under a certain set of conditions and completely inactive under other conditions (a conditional mutant). Proteins that are active at one temperature and inactive at another are widely used. Even then, it is difficult to control for the effects of temperature on all of the other processes in the cell. A second option is to put the expression of the protein or RNA under the control of regulatory proteins that are sensitive to the presence of a small molecule, such as a vitamin or hormone. Then, expression of the molecule can be turned on and off at will. This is commonly done for vertebrate cells by using promoters of gene expression engineered so that they can be turned on or off by the antibiotic tetracycline, which alters the ability of a bacterial protein (the tetracycline repressor) to bind particular regulatory sequences on DNA. A limitation of this technology is that some proteins are so stable that days are required to reduce their concentrations. During this time, cells may be able to compensate for the loss of the protein of interest.

RNA interference (RNAi) is a powerful method to reduce the concentration of a particular RNA, especially mRNAs (see Fig. 16-12). Introducing a double-stranded RNA copy of part of an RNA sequence into the cytoplasm generates a response that results in the degradation of the target RNA. Animals, fungi, and plants use this process to suppress expression of foreign RNAs, such as those introduced by viruses. If double-stranded RNA is introduced into cells, it is fragmented into pieces of about 21 nucleotides (see Fig. 16-12). Base pairing of these fragments with cellular RNAs having the complementary sequence (usually an exact match is required) targets the RNA for cleavage. To suppress a particular RNA in human cells experimentally, one synthesizes a double-stranded RNA including a sequence of 21 nucleotides matching the target cellular RNA. Introduction of this oligonucleotide into cells often (not always) results in destruction of the target RNA. If successful, the level of the targeted protein falls 5- to 10-fold as it is degraded naturally over the next several days. Loss of the protein may produce a cellular phenotype. RNAs and proteins can be depleted from Drosophila and Caenorhabditis elegans by using slightly different procedures. The simplicity of this approach makes RNAi very powerful and suitable for scaling up to study thousands of genes. However, false-negative results are common because some targeted protein usually remains. If the protein is an enzyme, a few protein molecules can turn over numerous substrate molecules and maintain function. One must also be cautious regarding other unanticipated consequences.

Another strategy is to inhibit a particular protein with a drug, inhibitory peptide, antibody, or inactive partner protein. Drugs as probes for function have a long and distinguished history in biology, but their use is hampered by the difficulty of ruling out side effects, including action on other unknown targets. One wag even asserted that “drugs are only specific for about a year,” roughly the time it takes someone to find an unexpected second target. Nevertheless, many drugs have the advantages that the onset of their action is rapid and their effects are reversible, so one can follow the process of recovery when they are removed. The use of libraries of small molecules to probe biological processes has been given the name chemical genetics.

If microinjected into cells, antibodies can be very specific, but the effects on their target must be fully characterized, and sufficient antibody must be introduced into the target cell to inactivate the target molecule. Some arginine-rich peptides, such as one from the HIV Tat protein, can also be used to carry inhibitory peptides across the plasma membrane into the cytoplasm. Other peptides can guide experimental peptides into various cellular compartments. It is also possible to inactivate pathways by the introduction of dominant negative mutants that can do part, but not all, of the job of a given protein. Dominant negative mutants of protein kinases are particularly effective. The active site is modified to eliminate enzymatic activity, but the modified protein can still bind to its regulatory proteins and substrates. This can interfere with signal transduction pathways very effectively by competing with functional endogenous kinases for regulatory factors and substrates. Dominant negative mutants offer the advantage that they can be expressed in many types of cells. However, all too often, little is known about the concentrations of these dominant negative agents or the full range of their targets.

The concentration of active protein can be increased by overexpression, for example, driving the expression of a cDNA from a very active viral promoter. Some expression systems are conditional, being turned on, for example, by an insect hormone that does not activate endogenous genes. Interpreting the consequences of overexpression tends to be more problematic than other approaches, as specificity of interactions with other cellular components can be lost at high concentrations.

Genetics is the best way to replace a native protein with a protein that has altered biochemical properties. Such gene replacement requires homologous recombination in the genome, which is not readily available in all experimental systems (Table 6-2). Examples of altered proteins include an enzyme with an altered catalytic function or a protein with altered affinity for a particular cellular partner. In the best cases, the altered protein is fully characterized before its coding sequence is used to replace that of the wild-type protein, and the cellular concentration of the altered protein is confirmed to be the same as the wild-type protein. On the relatively long time scale of such experiments (up to a year in vertebrates), interpreting the outcome may be compromised by the ability of cells to adapt to the change imposed by the gene substitution in un-known ways.