Chapter 6:  Basic Genetic Mechanisms

Introduction

Proteins constitute more than half the total dry mass of a cell, and their synthesis is central to cell maintenance, growth, and development. Protein synthesis occurs on ribosomes. It depends on the collaboration of several classes of RNA molecules and begins with a series of preparatory steps. First, a molecule of messenger RNA (mRNA) must be copied from the DNA that encodes the protein. Meanwhile, in the cytoplasm, each of the 20 amino acids from which the protein is to be built must be attached to its specific transfer RNA (tRNA) molecule, and the subunits of the ribosome on which the new protein is to be made must be preloaded with auxiliary protein factors. Protein synthesis begins when all of these components come together in the cytoplasm to form a functioning ribosome. As a single molecule of mRNA moves stepwise through a ribosome, the sequence of nucleotides in the mRNA molecule is translated into a corresponding sequence of amino acids to produce a distinctive protein chain, as specified by the DNA sequence of its gene. We begin by considering how the many different RNA molecules in a cell are made.

RNA Polymerase Copies DNA into RNA: The Process of DNA Transcription ¹

RNA is synthesized on a DNA template by a process known as DNA transcription. Transcription generates the mRNAs that carry the information for protein synthesis, as well as the transfer, ribosomal, and other RNA molecules that have structural or catalytic functions. All of these RNA molecules are synthesized by RNA polymerase enzymes, which make an RNA copy of a DNA sequence. In eucaryotes three kinds of RNA polymerase molecules synthesize different types of RNA, as described in Chapter 8. These RNA polymerases are thought to have derived during evolution from the single enzyme present in bacteria that mediates all bacterial RNA synthesis.

The bacterial RNA polymerase is a large multisubunit enzyme associated with several additional protein subunits that enter and leave the polymerase-DNA complex at different stages of transcription. Free RNA polymerase molecules collide randomly with the bacterial chromosome, sliding along it but sticking only weakly to most DNA. The polymerase binds very tightly, however, when it contacts a specific DNA sequence, called the promoter, that contains the start site for RNA synthesis and signals where RNA synthesis should begin. The reactions that ensue are outlined in Figure 6-2. After binding to the promoter, the RNA polymerase opens up a local region of the double helix to expose the nucleotides on a short stretch of DNA on each strand. One of the two exposed DNA strands acts as a template for complementary base-pairing with incoming ribonucleoside triphosphate monomers, two of which are joined together by the polymerase to begin an RNA chain. The RNA polymerase molecule then moves stepwise along the DNA, unwinding the DNA helix just ahead to expose a new region of the template strand for complementary base-pairing. In this way the growing RNA chain is extended by one nucleotide at a time in the 5'-to-3' direction (Figure 6-3). The chain elongation process continues until the enzyme encounters a second special sequence in the DNA, the stop (termination) signal, where the polymerase halts and releases both the DNA template and the newly made RNA chain.

By convention, when a DNA sequence associated with a gene is specified, it is the sequence of the nontemplate strand that is given, and it is written in the 5'-to-3' direction. This convention is adopted because the sequence of the nontemplate strand corresponds to the sequence of the RNA that is made.

Nucleotide sequences that act as start sites and stop signals for the bacterial RNA polymerase are illustrated in Figure 6-4. Nucleotide sequences that are found in many examples of a particular type of region in DNA (such as a promoter) are called consensus sequences. In bacteria strong promoters (those associated with genes that produce large amounts of mRNA) have sequences that match the promoter consensus sequences closely (as in Figure 6-4A), whereas weak promoters (those associated with genes that produce relatively small amounts of mRNA) match these sequences less well.

Only Selected Portions of a Chromosome Are Used to Produce RNA Molecules ²

As an RNA polymerase molecule moves along the DNA, an RNA/DNA double helix is formed at the enzyme's active site. This helix is very short because the RNA just made is displaced, allowing the DNA/DNA helix immediately at the rear of the polymerase to rewind (Figure 6-5). As a result, each completed RNA chain is released from the DNA template as a free, single-stranded RNA molecule, typically between 70 and 10,000 nucleotides long.

In principle, any region of the DNA double helix could be copied into two different RNA molecules - one from each of the two DNA strands. In reality, only one DNA strand is used as a template in each region. The RNA made is equivalent in nucleotide sequence to the opposite, nontemplate DNA strand. Which of the two strands is copied varies along the length of a single DNA molecule and is determined by the promoter of each gene. As illustrated in Figure 6-4, a promoter is an oriented DNA sequence that points the RNA polymerase in one direction or the other, and this orientation determines which DNA strand is copied (Figure 6-6). The DNA strand that is copied into RNA can be either different or the same for neighboring genes (Figure 6-7).

Both bacterial and eucaryotic RNA polymerases are large, complicated molecules, with multiple subunits and a total mass of more than 500,000 daltons. Some bacterial viruses, in contrast, encode single-chain RNA polymerases of one-fifth this mass that catalyze RNA synthesis at least as well as the host-cell enzyme. Presumably, the multiple subunit composition of the cellular RNA polymerases is important for various regulatory aspects of cellular RNA synthesis that have not yet been well defined.

This brief outline of DNA transcription omits many details. Other complex steps usually must occur before an mRNA molecule is produced. Gene regulatory proteins, for example, help to determine which regions of DNA are transcribed by the RNA polymerase and thereby play a major part in determining which proteins are made by a cell. Moreover, although mRNA molecules are produced directly by DNA transcription in procaryotes, in higher eucaryotic cells most RNA transcripts are altered extensively - by a process called RNA splicing - before they leave the cell nucleus and enter the cytoplasm as mRNA molecules. All of these aspects of mRNA production are discussed in Chapters 8 and 9, where we consider the cell nucleus and the control of gene expression, respectively. For now, let us assume that functional mRNA molecules have been produced and proceed to examine how they direct protein synthesis.

Transfer RNA Molecules Act as Adaptors That Translate Nucleotide Sequences into Protein Sequences ³

All cells contain a set of transfer RNAs (tRNAs), each of which is a small RNA molecule (most have a length between 70 and 90 nucleotides). The tRNAs, by binding at one end to a specific codon in the mRNA and at their other end to the amino acid specified by that codon, enable amino acids to line up according to the sequence of nucleotides in the mRNA. Each tRNA is designed to carry only one of the 20 amino acids used for protein synthesis: a tRNA that carries glycine is designated tRNA^Gly and so on. Each of the 20 amino acids has at least one type of tRNA assigned to it, and most have several tRNAs. Before an amino acid is incorporated into a protein chain, it is attached by its carboxyl end to the 3' end of an appropriate tRNA molecule. This attachment serves two purposes. First, and most important, it covalently links the amino acid to a tRNA containing the correct anticodon - the sequence of three nucleotides that is complementary to the three-nucleotide codon that specifies that amino acid on an mRNA molecule. Codon-anticodon pairings enable each amino acid to be inserted into a growing protein chain according to the dictates of the sequence of nucleotides in the mRNA, thereby allowing the genetic code to be used to translate nucleotide sequences into protein sequences. This is the essential "adaptor" function of the tRNA molecule: with one end attached to an amino acid and the other paired to a codon, the tRNA converts sequences of nucleotides into sequences of amino acids.

The second function of the amino acid attachment is to activate the amino acid by generating a high-energy linkage at its carboxyl end so that it can react with the amino group of the next amino acid in the protein sequence to form a peptide bond. The activation process is necessary for protein synthesis because nonactivated amino acids cannot be added directly to a growing polypeptide chain. (In contrast, the reverse process, in which a peptide bond is hydrolyzed by the addition of water, is energetically favorable and can occur spontaneously.)

The function of a tRNA molecule depends on its precisely folded three-dimensional structure. A few tRNAs have been crystallized and their complete structures determined by x-ray diffraction analyses. Both intramolecular complementary base-pairings and unusual base interactions are required to fold a tRNA molecule (see Figure 3-18). The nucleotide sequences of tRNA molecules from many types of organisms reveal that tRNAs can form the loops and base-paired stems of a "cloverleaf" structure (Figure 6-8), and all are thought to fold further to adopt the L-shaped conformation detected in crystallographic analyses. In the native structure the amino acid is attached to one end of the "L," while the anticodon is located at the other (Figure 6-9).

The nucleotides in a completed nucleic acid chain (like the amino acids in proteins) can be covalently modified to modulate the biological activity of the nucleic acid molecule. Such posttranscriptional modifications are especially common in tRNA molecules, which contain a variety of modified nucleotides (Figure 6-10). Some of the modified nucleotides affect the conformation and base-pairing of the anticodon and thereby facilitate the recognition of the appropriate mRNA codon by the tRNA molecule.

Specific Enzymes Couple Each Amino Acid to Its Appropriate tRNA Molecule ⁴

Only the tRNA molecule, and not its attached amino acid, determines where the amino acid is added during protein synthesis. This was established by an ingenious experiment in which an amino acid (cysteine) was chemically converted into a different amino acid (alanine) after it was already attached to its specific tRNA. When such "hybrid" tRNA molecules were used for protein synthesis in a cell-free system, the wrong amino acid was inserted at every point in the protein chain where that tRNA was used. Thus the accuracy of protein synthesis is crucially dependent on the accuracy of the mechanism that normally links each activated amino acid specifically to its corresponding tRNA molecules.

How does a tRNA molecule become covalently linked to the one amino acid in 20 that is its appropriate partner? The mechanism depends on enzymes called aminoacyl-tRNA synthetases, which couple each amino acid to its appropriate set of tRNA molecules. There is a different synthetase enzyme for every amino acid (20 synthetases in all): one attaches glycine to all tRNA^Gly molecules, another attaches alanine to all tRNA^Ala molecules, and so on. The coupling reaction that creates an aminoacyl-tRNA molecule is catalyzed in two steps, as illustrated in Figure 6-11. The structure of the amino acid-RNA linkage is shown in Figure 6-12.

Although the tRNA molecules serve as the final adaptors in converting nucleotide sequences into amino acid sequences, the aminoacyl-tRNA synthetase enzymes are adaptors of equal importance to the decoding process (Figure 6-13). Thus the genetic code is translated by two sets of adaptors that act sequentially, each matching one molecular surface to another with great specificity; it is their combined action that associates each sequence of three nucleotides in the mRNA molecule - that is, each codon - with its particular amino acid (Figure 6-14).

Amino Acids Are Added to the Carboxyl-Terminal End of a Growing Polypeptide Chain

The fundamental reaction of protein synthesis is the formation of a peptide bond between the carboxyl group at the end of a growing polypeptide chain and a free amino group on an amino acid. Consequently, a protein is synthesized stepwise from its amino-terminal end to its carboxyl-terminal end. Throughout the entire process the growing carboxyl end of the polypeptide chain remains activated by its covalent attachment to a tRNA molecule (a peptidyl-tRNA molecule). This high-energy covalent linkage is disrupted in each cycle but is immediately replaced by the identical linkage on the most recently added amino acid (Figure 6-15). In this way each amino acid added carries with it the activation energy for the addition of the next amino acid rather than the energy for its own addition - an example of the "head growth" type of polymerization described in Chapter 2 (see Figure 2-36).

The Genetic Code Is Degenerate ⁵

In the course of protein synthesis, the translation machinery moves in the 5'-to-3' direction along an mRNA molecule and the mRNA sequence is read three nucleotides at a time. As we have seen, each amino acid is specified by the triplet of nucleotides (codon) in the mRNA molecule that pairs with a sequence of three complementary nucleotides at the anticodon tip of a particular tRNA. Because only one of the many types of tRNA molecules in a cell can base-pair with each codon, the codon determines the specific amino acid residue to be added to the growing polypeptide chain end (Figure 6-16).

Since RNA is constructed from four types of nucleotides, there are 64 possible sequences composed of three nucleotides (4 × 4 × 4). Three of these 64 sequences do not code for amino acids but instead specify the termination of a polypeptide chain; they are known as stop codons. That leaves 61 codons to specify only 20 different amino acids. For this reason, most of the amino acids are represented by more than one codon (Figure 6-17) and the genetic code is said to be degenerate. Two amino acids, methionine and tryptophan, have only one codon each, and they are the least abundant amino acids in proteins.

The degeneracy of the genetic code implies either that there is more than one tRNA for each amino acid or that a single tRNA molecule can base-pair with more than one codon. In fact, both situations occur. For some amino acids there is more than one tRNA molecule, and some tRNA molecules are constructed so that they require accurate base-pairing only at the first two positions of the codon and can tolerate a mismatch (or wobble) at the third. This wobble base-pairing explains why so many of the alternative codons for an amino acid differ only in their third nucleotide (see Figure 6-17). The standard wobble pairings make it possible to fit the 20 amino acids to 61 codons with as few as 31 kinds of tRNA molecules; in animal mitochondria a more extreme wobble allows protein synthesis with only 22 tRNAs (discussed in Chapter 14).

The Events in Protein Synthesis Are Catalyzed on the Ribosome ⁶

The protein synthesis reactions just described require a complex catalytic machinery to guide them. The growing end of the polypeptide chain, for example, must be kept in register with the mRNA molecule to ensure that each successive codon in the mRNA engages precisely with the anticodon of a tRNA molecule and does not slip by one nucleotide, thereby changing the reading frame (see Figure 3-17). This precise movement and the other events in protein synthesis are catalyzed by ribosomes, which are large complexes of RNA and protein molecules. Eucaryotic and procaryotic ribosomes are very similar in design and function. Both are composed of one large and one small subunit that fit together to form a complex with a mass of several million daltons (Figure 6-18). The small subunit binds the mRNA and tRNAs, while the large subunit catalyzes peptide bond formation.

More than half of the weight of a ribosome is RNA, and there is increasing evidence that the ribosomal RNA (rRNA) molecules play a central part in its catalytic activities. Although the rRNA molecule in the small ribosomal subunit varies in size depending on the organism, its complicated folded structure is highly conserved (Figure 6-19); there are also close homologies between the rRNAs of the large ribosomal subunits in different organisms. Ribosomes contain a large number of proteins (Figure 6-20), but many of these have been relatively poorly conserved in sequence during evolution, and a surprising number seem not to be essential for ribosome function. Therefore, it has been suggested that the ribosomal proteins mainly enhance the function of the rRNAs and that the RNA molecules rather than the protein molecules catalyze many of the reactions on the ribosome.

A Ribosome Moves Stepwise Along the mRNA Chain ⁷

A ribosome contains three binding sites for RNA molecules: one for mRNA and two for tRNAs. One site, called the peptidyl-tRNA-binding site, or P-site, holds the tRNA molecule that is linked to the growing end of the polypeptide chain. Another site, called the aminoacyl-tRNA-binding site, or A-site, holds the incoming tRNA molecule charged with an amino acid. A tRNA molecule is held tightly at either site only if its anticodon forms base pairs with a complementary codon on the mRNA molecule that is bound to the ribosome. The A- and P-sites are so close together that the two tRNA molecules are forced to form base pairs with adjacent codons in the mRNA molecule (Figure 6-21).

The process of polypeptide chain elongation on a ribosome can be thought of as a cycle with three discrete steps (Figure 6-22):

1. In step 1, an aminoacyl-tRNA molecule becomes bound to a vacant ribosomal A-site (adjacent to an occupied P-site) by forming base pairs with the three mRNA nucleotides (codon) exposed at the A-site.

2. In step 2, the carboxyl end of the polypeptide chain is uncoupled from the tRNA molecule in the P-site and joined by a peptide bond to the amino acid linked to the tRNA molecule in the A-site. This central reaction of protein synthesis (see Figure 6-15) is catalyzed by a peptidyl transferase enzyme. Recent experiments with ribosomes that have been experimentally stripped of proteins show that this catalysis is mediated not by a protein but by a specific region of the major rRNA molecule in the large subunit (see Figure 3-23).

3. In step 3, the new peptidyl-tRNA in the A-site is translocated to the P-site as the ribosome moves exactly three nucleotides along the mRNA molecule. This step requires energy and is driven by a series of conformational changes induced in one of the ribosomal components by the hydrolysis of a GTP molecule.

As part of the translocation process of step 3, the free tRNA molecule that was generated in the P-site during step 2 is released from the ribosome to reenter the cytoplasmic tRNA pool. Upon completion of step 3, the unoccupied A-site is free to accept a new tRNA molecule linked to the next amino acid, which starts the cycle again. In a bacterium each cycle requires about 1/20th of a second under optimal conditions, so that the complete synthesis of an average-sized protein of 400 amino acids is accomplished in about 20 seconds. Ribosomes move along an mRNA molecule in the 5'-to-3' direction, which is also the direction of RNA synthesis (see Figure 6-3).

In most cells protein synthesis consumes more energy than any other biosynthetic process. At least four high-energy phosphate bonds are split to make each new peptide bond: two of these are required to charge each tRNA molecule with an amino acid (see Figure 6-11), and two more drive steps in the cycle of reactions occurring on the ribosome during synthesis itselfone for the aminoacyl-tRNA binding in step 1 (see Figure 6-31) and one for the ribosome translocation in step 3.

A Protein Chain Is Released from the Ribosome When Any One of Three Stop Codons Is Reached ⁸

Of the 64 possible codons in an mRNA molecule, 3 (UAA, UAG, and UGA) are stop codons, which terminate the translation process. Cytoplasmic proteins called release factors bind directly to any stop codon that reaches the A-site on the ribosome. This binding alters the activity of the peptidyl transferase, causing it to catalyze the addition of a water molecule instead of an amino acid to the peptidyl-tRNA. This reaction frees the carboxyl end of the growing polypeptide chain from its attachment to a tRNA molecule, and since only this attachment normally holds the growing polypeptide to the ribosome, the completed protein chain is immediately released into the cytoplasm. The ribosome releases the mRNA and dissociates into its two separate subunits (Figure 6-23), which can assemble on another mRNA molecule to begin a new round of protein synthesis by the process to be described next.

The Initiation Process Sets the Reading Frame for Protein Synthesis ⁹

In principle, an RNA sequence can be translated in any one of three reading frames, each of which will specify a completely different polypeptide chain (see Figure 3-17). Which of the three frames is actually read is determined by the RNA sequence, which determines how the ribosome assembles. During the initiation phase of protein synthesis, the two subunits of the ribosome are brought together at the exact spot on the mRNA where the polypeptide chain is to begin.

The initiation process is complicated, involving a number of steps catalyzed by proteins called initiation factors (IFs), many of which are themselves composed of several polypeptide chains. Because the process is so complex, many of the details of initiation are still uncertain. It is clear, however, that each ribosome is assembled onto an mRNA chain in two steps: only after the small ribosomal subunit loaded with initiation factors finds the start codon (AUG, see below) does the large subunit bind.

Before a ribosome can begin a new protein chain, it must bind an aminoacyl tRNA molecule in its P-site, where normally only peptidyl tRNA molecules are bound. (As explained previously, the peptidyl tRNA is translocated to the P-site during step 3 of the elongation reaction.) A special tRNA molecule is required for this purpose. This initiator tRNA provides the amino acid that starts a protein chain, and it always carries methionine (aminoformyl methionine in bacteria). In eucaryotes the initiator tRNA molecule must be loaded onto the small ribosomal subunit before this subunit can bind to an mRNA molecule. An initiation factor called eucaryotic initiation factor 2 (eIF-2) is required to position the initiator tRNA on the small subunit. One molecule of eIF-2 becomes tightly bound to each initiator tRNA molecule as soon as this tRNA acquires its methionine, and in some cells the overall rate of protein synthesis is controlled by this factor (see Figure 9-82).

As described in more detail in the next section, the small ribosomal subunit helps its bound initiator tRNA molecule find a special AUG codon (the start codon) on an mRNA molecule. Once this has occurred, the several initiation factors that were previously associated with the small ribosomal subunit are discharged to make way for the binding of a large ribosomal subunit to the small one. Because the initiator tRNA molecule is bound to the P-site of the ribosome, the synthesis of a protein chain can begin directly with the binding of a second aminoacyl-tRNA molecule to the A-site of the ribosome (Figure 6-24). Thus a complete functional ribosome is assembled, with the mRNA molecule threaded through it (Figure 6-25). Further steps in the elongation phase of protein synthesis then proceed as described previously (see Figure 6-22). Because an initiator tRNA molecule has begun each polypeptide chain, all newly made proteins have a methionine (or the aminoformyl derivative of methionine in bacteria) as their amino-terminal residue. The methionine is often removed shortly after its incorporation by a specific aminopeptidase; this trimming process is important because the amino acid left at the amino terminus can determine the protein's lifetime in the cell by its effects on a ubiquitin-dependent protein-degradation pathway (see Figure 5-39).

Evidently the correct initiation site on the mRNA molecule must be selected by the small subunit acting in concert with initiation factors but in the absence of the large subunit. This requirement helps to explain why all ribosomes are formed from two separate subunits. We shall now consider how the correct start codon is selected.

Only One Species of Polypeptide Chain Is Usually Synthesized from Each mRNA Molecule in Eucaryotes ¹⁰

A messenger RNA molecule will typically contain many AUG sequences, each of which can code for methionine. In eucaryotes, however, only one of these AUG sequences will normally be recognized by the initiator tRNA and thereby serve as a start codon. How does the ribosome distinguish this start codon?

Eucaryotic RNAs (except those that are synthesized in mitochondria and chloroplasts) are extensively modified in the nucleus immediately after their transcription (discussed in Chapter 8). Two general modifications are the addition of a unique "cap" structure, composed of a 7-methylguanosine residue linked to a triphosphate at the 5' end (Figure 6-26) and the addition of a run of about 200 adenylic residues ("poly A") at the 3' end. What part the poly A plays in the translation process is uncertain (see Figure 9-87), but the 5' cap structure is essential for efficient protein synthesis. Experiments carried out with extracts of eucaryotic cells have shown that the small ribosomal subunit first binds at the 5' end of an mRNA chain, aided by recognition of the 5' cap (see Figure 6-24). This subunit then propels itself along the mRNA chain in a scanning mode, carrying its bound initiator tRNA in a search of an AUG start codon. The requirements for a start codon apparently are not very stringent, since the small subunit usually selects the first AUG it encounters; however, a few nucleotides in addition to the AUG are also important for the selection process. For most eucaryotic RNAs, once a start codon near the 5' end has been selected, none of the many other AUG codons farther down the chain will serve as initiation sites. As a result, only a single species of polypeptide chain is usually synthesized from an mRNA molecule (for exceptions see p. 467).

The mechanism for selecting a start codon in bacteria is different. Bacterial mRNAs have no 5' cap structure. Instead, they contain a specific ribosome-binding site sequence, up to six nucleotides long, which can occur at several places in the same mRNA molecule. These sequences are located four to seven nucleo-tides upstream from an AUG, and they form base pairs with a specific region of the rRNA in a ribosome to signal the initiation of protein synthesis at this nearby start codon. Bacterial ribosomes, unlike eucaryotic ribosomes, bind directly to start codons in the interior of an mRNA molecule to initiate protein synthesis. As a result, bacterial messenger RNAs are commonly polycistronicthat is, they encode multiple proteins that are separately translated from the same mRNA molecule. Eucaryotic mRNAs, in contrast, are typically monocistronic, with only one species of polypeptide chain being translated per messenger molecule (Figure 6-27).

The Binding of Many Ribosomes to an Individual mRNA Molecule Generates Polyribosomes ¹¹

The complete synthesis of a protein takes 20 to 60 seconds on average. But even during this very short period, multiple initiations usually take place on each mRNA molecule being translated. A new ribosome hops onto the 5' end of the mRNA molecule almost as soon as the preceding ribosome has translated enough of the amino acid sequence to be out of the way. Such mRNA molecules are thus found in the cell as polyribosomes, or polysomes, formed by several ribosomes spaced as close as 80 nucleotides apart along a single messenger molecule (Figures 6-28 and 6-29). Polyribosomes are a common feature of cells. They can be isolated and separated from single ribosomes in the cytosol by ultracentrifugation after cell lysis (Figure 6-30). The mRNA purified from these polyribosomes can be used to determine if the protein encoded by a particular DNA sequence is being actively synthesized in the cells used to prepare the polyribosomes. These mRNA molecules can also serve as the starting material for the preparation of specialized cDNA libraries (discussed in Chapter 7).

In eucaryotes the nuclear envelope keeps transcription and protein synthesis separate. But in procaryotes, RNA is accessible to ribosomes as soon as it is made. Thus, ribosomes will begin synthesizing a polypeptide chain at the 5' end of a nascent mRNA molecule and then follow behind the RNA polymerase as it completes an mRNA chain.

The Overall Rate of Protein Synthesis in Eucaryotes Is Controlled by Initiation Factors ¹²

As we discuss in Chapter 17, the cells in a multicellular organism proliferate only when they are stimulated to do so by specific growth factors. Although the mechanisms by which growth factors act are incompletely understood, one of their major effects must be to increase the overall rate of protein synthesis, for cells must double their contents before they divide. What determines the rate of protein synthesis? When eucaryotic cells in culture are starved of nutrients, there is a marked reduction in the rate of polypeptide chain initiation. This is the result of inactivation of the protein synthesis initiation factor eIF-2 (see Figure 9-82). The initiation factors required for protein synthesis are much more numerous and complex in eucaryotes than in procaryotes, even though they perform the same basic functions. Many of the extra components may be regulatory proteins that respond to growth factors and help coordinate cell growth and proliferation in multicellular organisms. Less complex controls are needed in bacteria, which generally grow as fast as the nutrients in their environment allow.

The Fidelity of Protein Synthesis Is Improved by Two Proofreading Mechanisms ¹³

The error rate in protein synthesis can be estimated by monitoring the frequency of incorporation of an amino acid into a protein that normally lacks that amino acid. Error rates of about 1 amino acid misincorporated for every 10⁴ amino acids polymerized are observed, which means that only about 1 in every 25 protein molecules of average size (400 amino acids) should contain an error. The fidelity of protein synthesis depends on the accuracy of the two adaptor mechanisms previously discussed: the linking of each amino acid to its corresponding tRNA molecule and the base-pairing of the codons in mRNA to the anticodons in tRNA (see Figure 6-14). Not surprisingly, cells have evolved "proofreading" mechanisms to reduce the number of errors in both these crucial steps of protein synthesis.

Two fundamentally different proofreading mechanisms are used, each representative of strategies used in other processes in the cell. Both involve expenditure of free energy, since, as discussed in Chapter 2, a price must be paid for any increase in order in the cell. A relatively simple mechanism is used to improve the accuracy of amino acid attachment to tRNA. Many aminoacyl tRNA synthetases have two active sites, one that carries out the loading reaction shown earlier (Figure 6-11) and one that recognizes an incorrect amino acid attached to its tRNA molecule and removes it by hydrolysis. The correction process is energetically costly because to be effective it must remove an appreciable fraction of correctly attached amino acids as well. The same type of costly two-step proofreading process is used in DNA replication (see Figure 6-42).

A more subtle "kinetic proofreading" mechanism is used to improve the fidelity of codon-anticodon pairing. Thus far we have given a simplified account of this pairing. In fact, once tRNA molecules have acquired an amino acid, they form a complex with an abundant protein called an elongation factor (EF), which binds tightly to both the amino acid end of a tRNA and to a molecule of GTP. It is this complex, and not free tRNA, that pairs with the appropriate codon in an mRNA molecule. The bound elongation factor allows correct codon-anticodon pairing to occur but prevents the amino acid from being incorporated into the growing polypeptide chain. The initial codon recognition, however, triggers the elongation factor to hydrolyze its bound GTP (to GDP and inorganic phosphate), whereupon the factor dissociates from the ribosome without its tRNA, allowing protein synthesis to proceed. The elongation factor thereby introduces a short delay between codon-anticodon base-pairing and polypeptide chain elongation, which provides an opportunity for the bound tRNA molecule to exit from the ribosome. An incorrect tRNA molecule forms a smaller number of codon-anticodon hydrogen bonds than a correct one; it therefore binds more weakly to the ribosome and is more likely to dissociate during this period. Because the delay introduced by the elongation factor causes most incorrectly bound tRNA molecules to leave the ribosome without being used for protein synthesis, this factor increases the ratio of correct to incorrect amino acids incorporated into protein (Figure 6-31).

Many Inhibitors of Procaryotic Protein Synthesis Are Useful as Antibiotics ¹⁴

Many of the most effective antibiotics used in modern medicine are compounds made by fungi that act by inhibiting bacterial protein synthesis. A number of these drugs exploit the structural and functional differences between procaryotic and eucaryotic ribosomes so as to interfere with the function of procaryotic ribosomes preferentially. Thus some of these compounds can be taken in high doses without undue toxicity to humans. Because different antibiotics bind to different regions of bacterial ribosomes, they often inhibit different steps in the synthetic process. Some of the more common antibiotics of this kind are listed in Table 6-1 along with several other commonly used inhibitors of protein synthesis, some of which act on eucaryotic cells and therefore cannot be used as antibiotics.

Because they block specific steps in the processes that lead from DNA to protein, many of the compounds listed in Table 6-1 are useful for cell biological studies. Among the most commonly used drugs in such experimental studies are chloramphenicol, cycloheximide, and puromycin, all of which specifically inhibit protein synthesis. In a eucaryotic cell, for example, chloramphenicol inhibits protein synthesis on ribosomes only in mitochondria (and in chloroplasts in plants), presumably reflecting the procaryotic origins of these organelles (discussed in Chapter 14). Cycloheximide, on the other hand, affects only ribosomes in the cytosol. The difference in the sensitivity of protein synthesis to these two drugs provides a powerful way to determine in which cell compartment a particular protein is translated. Puromycin is especially interesting because it is a structural analogue of a tRNA molecule linked to an amino acid; the ribosome mistakes it for an authentic amino acid and covalently incorporates it at the carboxyl terminus of the growing polypeptide chain, thereby causing the premature termination and release of the polypeptide (see Figure 3-23). As might be expected, puromycin inhibits protein synthesis in both procaryotes and eucaryotes.

How Did Protein Synthesis Evolve? ¹⁵

The molecular processes underlying protein synthesis seem inexplicably complex. Although we can describe many of them, they do not make conceptual sense in the way that DNA transcription, DNA repair, and DNA replication do. As we have seen, protein synthesis in present-day organisms centers on the ribosome, which consists of proteins arranged around a core of rRNA molecules. Why should rRNA molecules exist at all, and how did they come to play such a dominant part in the structure and function of the ribosome?

Before the discovery of mRNA in the early 1960s, it was suspected that the large amount of RNA in ribosomes served a "messenger" function, carrying genetic information from DNA to proteins. Now we know, however, that all of the ribosomes in a cell contain an identical set of rRNA molecules that have no such informational role. In bacterial ribosomes, rRNA molecules have been shown to have catalytic functions in protein synthesis. As mentioned earlier, the major rRNA of the large ribosomal subunit appears to be the peptidyl transferase; in addition, the rRNA of the small ribosomal subunit forms a short base-paired helix with the initiation site sequence on bacterial mRNA molecules, positioning the neighboring AUG start codon at the P-site. A variety of specific base-pair interactions likewise form between tRNA molecules and bacterial rRNAs, although these interactions involve individual bases on the rRNA that are far apart in the nucleotide sequence, suggesting complex sets of interactions that depend on the tertiary structure of the rRNA.

Protein synthesis also relies heavily on a large number of proteins that are bound to the rRNAs in a ribosome (see Figure 6-20). The complexity of a process with so many interacting components has made many biologists despair of ever understanding the pathway by which protein synthesis evolved. The discovery that RNA molecules can act as enzymes, however, has provided a new way of viewing the pathway. As discussed in Chapter 1, early biological reactions probably used RNA molecules rather than protein molecules as catalysts. In the earliest cells tRNA molecules on their own may have formed catalytic surfaces that allowed them to bind and activate specific amino acids without requiring aminoacyl-tRNA synthetase enzymes. Likewise, rRNA molecules may have served by themselves as the entire "ribosome," folding up in complex ways to generate an intricate set of surfaces that both guided tRNA pairings with mRNA codons and catalyzed the polymerization of the tRNA-linked amino acids (see Figure 1-7). Over the course of evolution individual proteins have been added to this machinery, each one making the process a little more accurate and efficient, or adding regulatory controls. In this view the large amount of RNA in present-day ribosomes is a remnant of a very early stage in evolution, before proteins dominated biological catalysis.

Summary

Before the synthesis of a particular protein can begin, the corresponding mRNA molecule must be produced by DNA transcription. Then a small ribosomal subunit binds to the mRNA molecule at a start codon (AUG) that is recognized by a unique initiator tRNA molecule. A large ribosomal subunit binds to complete the ribosome and initiate the elongation phase of protein synthesis. During this phase aminoacyl tRNAs, each bearing a specific amino acid, sequentially bind to the appropriate codon in mRNA by forming complementary base pairs with the tRNA anticodon. Each amino acid is added to the carboxyl-terminal end of the growing polypeptide by means of a cycle of three sequential steps: aminoacyl-tRNA binding, followed by peptide bond formation, followed by ribosome translocation. The ribosome progresses from codon to codon in the 5'-to-3' direction along the mRNA molecule until one of three stop codons is reached. A release factor then binds to the stop codon, terminating translation and releasing the completed polypeptide from the ribosome.

Eucaryotic and procaryotic ribosomes are highly homologous, despite substantial differences in the number and size of their rRNA and protein components. The predominant role of rRNA in ribosome structure and function is likely to reflect the ancient origin of protein synthesis, which is thought to have evolved in an environment dominated by RNA-mediated catalysis.

Introduction

The long-term survival of a species may be enhanced by genetic changes, but the survival of the individual demands genetic stability. Maintaining genetic stability requires not only an extremely accurate mechanism for replicating the DNA before a cell divides, but also mechanisms for repairing the many accidental lesions that occur continually in DNA. Most such spontaneous changes in DNA are temporary because they are immediately corrected by processes collectively called DNA repair. Only rarely do the cell's DNA maintenance processes fail and allow a permanent change in the DNA. Such a change is called a mutation, and it can destroy an organism if the change occurs in a vital position in the DNA sequence.

Before examining the mechanisms of DNA repair, we briefly discuss the maintenance of DNA sequences from one generation to the next.

DNA Sequences Are Maintained with Very High Fidelity ¹⁷

The rate at which stable changes occur in DNA sequences (the mutation rate) can be estimated only indirectly. One way is to compare the amino acid sequence of the same protein in several species. The fraction of the amino acids that are different can then be compared with the estimated number of years since each pair of species diverged from a common ancestor, as determined from the fossil record. In this way one can calculate the number of years that elapse, on average, before an inherited change in the amino acid sequence of a protein becomes fixed in the species. Because each such change will commonly reflect the alteration of a single nucleotide in the DNA sequence of the gene encoding that protein, this value can be used to estimate the average number of years required to produce a single, stable mutation in the gene.

Such calculations always will substantially underestimate the actual mutation rate because most mutations will spoil the function of the protein and vanish from the population through natural selection. But there is one family of proteins whose sequence does not seem to matter, and so the genes that encode them can accumulate mutations without being selected against. These proteins are the fibrinopeptides - 20-residue-long fragments that are discarded from the protein fibrinogen when it is activated to form fibrin during blood clotting. Since the function of fibrinopeptides apparently does not depend on their amino acid sequence, they can tolerate almost any amino acid change. Sequence analysis of the fibrinopeptides indicates that an average-sized protein 400 amino acids long would be randomly altered by an amino acid change roughly once every 200,000 years. More recently, DNA sequencing technology has made it possible to compare corresponding nucleotide sequences in regions of the genome that do not code for protein. Comparisons of such sequences in several mammalian species produce estimates of the mutation rate during evolution that are in excellent agreement with those obtained from the fibrinopeptide studies.

The Observed Mutation Rates in Proliferating Cells Are Consistent with Evolutionary Estimates ¹⁸

The mutation rate can be estimated more directly by observing the rate at which spontaneous genetic changes arise in a large population of cells followed over a relatively short period of time. This can be done either by estimating the frequency with which new mutants arise in very large animal populations (in a colony of fruit flies or mice, for example) or by screening for changes in specific proteins in cells growing in culture. Although they are only approximate, the numbers obtained in both cases are consistent with an error frequency of 1 base-pair change in roughly 10⁹ base pairs for each cell generation. Consequently, a single gene that encodes an average-sized protein (containing about 10³ coding base pairs) would suffer a mutation once in about 10⁶ cell generations. This number is at least roughly consistent with the evolutionary estimate described above, in which one mutation appears in an average gene in the germ line every 200,000 years.

Most Mutations in Proteins Are Deleterious and Are Eliminated by Natural Selection ¹⁹

When the number of amino acid differences in a particular protein is plotted for several pairs of species against the time since the species diverged, the result is a reasonably straight line. That is, the longer the period since divergence, the larger the number of differences. For convenience, the slope of this line can be expressed in terms of the "unit evolutionary time" for that protein, which is the average time required for 1 amino acid change to appear in a sequence of 100 amino acid residues. When various proteins are compared, each shows a different but characteristic rate of evolution (Figure 6-32). Since all DNA base pairs are thought to be subject to roughly the same rate of random mutation, these different rates must reflect differences in the probability that an organism with a random mutation over the given protein will survive and propagate. Changes in amino acid sequence are evidently much more harmful for some proteins than for others. From Table 6-2 we can estimate that about 6 of every 7 random amino acid changes are harmful over the long term in hemoglobin, about 29 of every 30 amino acid changes are harmful in cytochrome c, and virtually all amino acid changes are harmful in histone H4. We assume that individuals who carried such harmful mutations have been eliminated from the population by natural selection.

Low Mutation Rates Are Necessary for Life as We Know It ¹⁹

Since most mutations are deleterious, no species can afford to allow them to accumulate at a high rate in its germ cells. We discuss later why the observed mutation frequency, low though it is, nevertheless, is thought to limit the number of essential proteins that any organism can encode in its germ line to about 60,000. By an extension of the same arguments, a mutation frequency tenfold higher would limit an organism to about 6000 essential proteins. In this case evolution would probably have stopped at an organism no more complex than a fruit fly.

While germ cells must be protected against high rates of mutation in order to maintain the species, the other cells of a multicellular organism (its somatic cells) must be protected from genetic change to safeguard the individual. Nucleotide changes in somatic cells can give rise to variant cells, some of which, through a process of natural selection, grow rapidly at the expense of the rest of the organism. In the extreme case the uncontrolled cell proliferation known as cancer results, which is responsible for about 30% of the deaths that occur in Europe and North America. These deaths are due largely to the accumulation of changes in the DNA sequences of somatic cells (discussed in Chapter 24). A tenfold increase in the mutation frequency would presumably cause a disastrous increase in the incidence of cancer by accelerating the rate at which somatic cell variants arise. Thus, both for the perpetuation of a species with 60,000 proteins (germ cell stability) and for the prevention of cancer resulting from mutations in somatic cells (somatic cell stability), eucaryotes depend on the remarkably high fidelity with which DNA sequences are maintained.

Low Mutation Rates Mean That Related Organisms Must Be Made from Essentially the Same Proteins ²⁰

Humans, as a genus distinct from the great apes, have existed for only a few million years. Each human gene has therefore had the chance to accumulate relatively few nucleotide changes since our inception, and most of these have been eliminated by natural selection. A comparison of humans and monkeys, for example, shows that their cytochrome c molecules differ in about 1% and their hemoglobins in about 4% of their amino acid positions. Clearly, a great deal of our genetic heritage must have been formed long before Homo sapiens appeared, during the evolution of mammals (which started about 300 million years ago) and even earlier. Because the proteins of mammals as different as whales and humans are very similar, the evolutionary changes that have produced such striking morphological differences must involve relatively few changes in the molecules from which we are made. Instead, it is thought that the morphological differences arise from differences in the temporal and spatial pattern of gene expression during embryonic development, which then determine the size, shape, and other characteristics of the adult. At the end of Chapter 8 we discuss the mechanisms that are thought to underlie such evolutionary changes in gene expression.

If Left Uncorrected, Spontaneous DNA Damage Would Rapidly Change DNA Sequences ²¹

The physicist Erwin Schroedinger pointed out in 1945 that, whatever its chemical nature (at that time unknown), a gene must be extremely small and composed of few atoms. Otherwise the very large number of genes thought to be necessary to generate an organism would not fit in the cell nucleus. On the other hand, because it was so small, a gene would be expected to undergo significant changes as a result of spontaneous reactions induced by random thermal collisions with solvent molecules. This poses a serious dilemma, since genetic data imply that genes are composed of a remarkably stable substance in which spontaneous changes (mutations) occur rarely.

This dilemma is real. DNA does undergo major changes as a result of thermal fluctuations. We now know, for example, that about 5000 purine bases (adenine and guanine) are lost per day from the DNA of each human cell because of the thermal disruption of their N-glycosyl linkages to deoxyribose (depurination). Similarly, spontaneous deamination of cytosine to uracil in DNA is estimated to occur at a rate of 100 bases per genome per day (Figure 6-33). DNA bases are also subject to change by reactive metabolites (including reactive forms of oxygen) that can alter their base-pairing abilities and by ultraviolet light from the sun, which promotes a covalent linkage of two adjacent pyrimidine bases in DNA (forming, for example, the thymine dimers shown in Figure 6-34B). These are only a few of many changes that can occur in our DNA (Figure 6-34A). Most of them would be expected to lead either to deletion of one or more base pairs in the daughter DNA chain after DNA replication or to a base-pair substitution (each C → U deamination, for example, would eventually change a C-G base pair to a T-A base pair, since U closely resembles T and forms a complementary base pair with A). As we have seen, a high rate of such random changes would have disastrous consequences for an organism.

The Stability of Genes Depends on DNA Repair ²²

Despite the thousands of random changes created every day in the DNA of a human cell by heat energy and metabolic accidents, only a few stable changes (mutations) accumulate in the DNA sequence of an average cell in a year. We now know that fewer than one in a thousand accidental base changes in DNA causes a mutation; the rest are eliminated with remarkable efficiency by DNA repair. There are a variety of repair mechanisms, each catalyzed by a different set of enzymes. Nearly all of these mechanisms depend on the existence of two copies of the genetic information, one in each strand of the DNA double helix: if the sequence in one strand is accidentally changed, information is not lost irretrievably because a complementary copy of the altered strand remains in the sequence of nucleotides in the other strand. The basic pathway for DNA repair is illustrated schematically in Figure 6-35. As indicated, it involves three steps:

1. The altered portion of a damaged DNA strand is recognized and removed by enzymes called DNA repair nucleases, which hydrolyze the phospho-diester bonds that join the damaged nucleotides to the rest of the DNA molecule, leaving a small gap in the DNA helix in this region.

2. Another enzyme, DNA polymerase, binds to the 3'-OH end of the cut DNA strand and fills in the gap by making a complementary copy of the information stored in the "good" (template) strand.

3. The break or "nick" in the damaged strand left when the DNA polymerase has filled in the gap is sealed by a third type of enzyme, DNA ligase, which completes the restoration process.

Both DNA polymerase and DNA ligase have important general roles in DNA metabolism; both function in DNA replication as well as in DNA repair, for example. The reactions that these two enzymes catalyze are illustrated in Figures 6-36 and 6-37, respectively.

DNA Damage Can Be Removed by More Than One Pathway ²³

The details of the excision step in DNA repair depend on the type of damage. Depurination, for example, which is by far the most frequent lesion that occurs in DNA, leaves a deoxyribose sugar with a missing base (see Figure 6-33). This exposed sugar is rapidly recognized by the enzyme AP endonuclease, which cuts the DNA phosphodiester backbone at the 5' side of the altered site. After excision of the sugar phosphate residue by a phosphodiesterase enzyme, an undamaged DNA sequence is restored by DNA polymerase and DNA ligase (see Figure 6-35).

A related repair pathway, called base excision repair, involves a battery of enzymes called DNA glycosylases. Each DNA glycosylase recognizes an altered base in DNA and catalyzes its hydrolytic removal. There are at least six types of these enzymes, including those that remove deaminated Cs, deaminated As, different types of alkylated or oxidized bases, bases with opened rings, and bases in which a carbon-carbon double bond has been accidentally converted to a carbon-carbon single bond. As an example of the general mechanism that operates in all cases, the removal of a deaminated C by uracil DNA glycosylase is shown in Figure 6-38A. The DNA glycosylase reaction produces a deoxyribose sugar with a missing base. Because this sugar phosphate is the same substrate recognized by the AP endonuclease, the subsequent steps in the repair process proceed in the same way as for depurinated sites. The importance of removing accidentally deaminated DNA bases has been directly demonstrated. In mutant bacteria that lack the enzyme uracil DNA glycosylase, the normally low spontaneous rate of change of a C-G to a T-A base pair is increased about twentyfold.

Cells have a separate nucleotide excision repair pathway capable of removing almost any type of DNA damage that creates a large change in the DNA double helix. Such "bulky lesions" include those created by the covalent reaction of DNA bases with large hydrocarbons (such as the carcinogen benzopyrene), as well as the various pyrimidine dimers (T-T, T-C, and C-C) caused by sunlight. In these cases a large multienzyme complex scans the DNA for a distortion in the double helix rather than for a specific base change. Once a bulky lesion is found, the phosphodiester backbone of the abnormal strand is cleaved on both sides of the distortion, and the portion of the strand containing the lesion (an oligonucleotide) is peeled away from the DNA double helix by a DNA helicase enzyme (discussed later). The gap produced in the DNA helix is then repaired in the usual manner by DNA polymerase and DNA ligase (Figure 6-38B).

The importance of these repair processes is indicated by the large investment that cells make in DNA repair enzymes. A comprehensive genetic analysis of a yeast suggests that these cells contain more than 50 different genes that code for DNA repair functions. DNA repair pathways are likely to be at least as complex in humans. Individuals with the genetic disease xeroderma pigmentosum, for example, are defective in a nucleotide excision repair process that can be shown by genetic analysis to require at least seven different gene products. Such individuals develop severe skin lesions, including skin cancer, because of the accumulation of pyrimidine dimers in cells that are exposed to sunlight.

Cells Can Produce DNA Repair Enzymes in Response to DNA Damage ²⁴

Cells have evolved a number of mechanisms to help them survive in a hazardous world. Often an extreme environmental insult activates a battery of genes whose products protect the cell from its effects. One such mechanism shared by all cells is the heat-shock response, which is evoked by the exposure of cells to unusually high temperatures. The induced "heat-shock proteins" include some that are thought to help stabilize and repair partially denatured cell proteins (see Figure 5-29).

Many cells also have mechanisms that enable them to synthesize DNA repair enzymes as an emergency response to severe DNA damage. The best-studied example is the SOS response in E. coli. In this bacterium any block to DNA replication caused by DNA damage produces a signal (thought to be an excess of single-stranded DNA) that induces an increase in the transcription of more than 15 genes, many of which code for proteins that function in DNA repair. The signal first activates the E. coli RecA protein (discussed later), which then destroys a negatively acting gene regulatory protein (a repressor) that normally suppresses the transcription of the entire set of SOS response genes. Studies of mutant bacteria deficient in different parts of the response indicate that the newly synthesized proteins have two effects. First, as would be expected, the induction of new DNA repair enzymes increases cell survival. When the mutants deficient in this part of the SOS response are treated with a DNA-damaging agent such as ultraviolet radiation, an unusually high proportion of them die. Second, several of the induced proteins transiently increase the mutation rate by greatly increasing the number of errors made in copying DNA sequences. While this has little effect on short-term survival, it is presumably advantageous in the long term because it produces a burst of genetic variability in the bacterial population and hence increases the chance that a mutant cell with increased fitness will arise.

The DNA repair system activated in the SOS response is not the only inducible DNA repair system known. Bacteria have another system that is activated specifically by the presence of methylated nucleotides in DNA, and there is at least one inducible DNA repair system in yeast cells. Some higher eucaryotic cells have been reported to adapt to DNA damage in similar ways.

The Structure and Chemistry of the DNA Double Helix Make It Easy to Repair ²⁵

The DNA double helix seems to be optimally constructed for repair. As discussed in Chapter 1, RNA is thought to have evolved before DNA, and it seems likely that the genetic code was initially carried in the four nucleotides A, C, G, and U. This raises the question of why the U in RNA has been replaced in DNA by T (which is 5-methyl U). We have seen that spontaneous C deamination converts C to U but that this event is rendered harmless by uracil DNA glycosylase (see Figure 6-38A). One can imagine how any repair enzyme designed to recognize and excise such accidents would be confused by the normal U nucleotides in a U-containing DNA molecule. Thus it is not surprising that U is not used in DNA.

This line of argument is strengthened by the observation that every possible deamination event in DNA yields an unnatural base, which can therefore be directly recognized and removed by a specific DNA glycosylase. Hypoxanthine, for example, is the simplest purine base capable of pairing specifically with C, but hypoxanthine is the direct deamination product of A. The addition of a second amino group to hypoxanthine produces G, which cannot be formed from A by spontaneous deamination and whose deamination product is likewise unique (Figure 6-39A).

A special situation occurs in vertebrate DNA, where selected C nucleotides are methylated at specific CG sequences associated with inactive genes (discussed in Chapter 9). As illustrated in Figure 6-39B, the accidental deamination of these methylated C nucleotides produces the natural nucleotide T, which forms a mismatched base pair with a G on the opposite DNA strand. To help protect methylated C nucleotides against such mutations, a special DNA glycosylase recognizes a mismatched base pair involving T in the sequence TG and removes the T. This DNA repair mechanism must be relatively ineffective, however, as methylated C nucleotides are common sites for mutations in vertebrate DNA. Even though only about 3% of the C nucleotides in human DNA are methylated, mutations in these methylated nucleotides account for about one-third of the single-base mutations that have been observed in inherited human diseases (see also Figure 9-71).

Whereas the chemistry of the bases ensures that deamination will be detected, accurate repair - and the fundamental answer to Schroedinger's dilemma - depends on the existence of separate copies of the genetic information in the two strands of the double helix. Only in the very unlikely event that both strands are damaged simultaneously at the same base pair is the cell left without one good copy to serve as a template for DNA repair. Even in this case mechanisms have evolved that are sometimes able to repair the damage. These repair mechanisms require that a second DNA helix of the same sequence be present in the cell, and they use genetic recombination mechanisms to transfer the missing information from one DNA helix to another - a process called gene conversion, which we discuss later.

Genetic information is stored in single-stranded DNA or RNA molecules only in some very small viruses with genomes of a few thousand nucleotides. The types of repair processes that we have described cannot operate on such nucleic acids, and the chance of a nucleotide change occurring in these viruses is very high. It seems that only organisms with tiny genomes can afford to encode their genetic information in a structure other than a DNA double helix.

Summary

The fidelity with which DNA sequences are maintained in higher eucaryotes can be estimated from the rates at which changes have occurred in nonessential protein and DNA sequences over evolutionary time. This fidelity is so high that a mammalian germ-line cell with a genome of 3 × 10⁹ base pairs is subjected on average to only about 10 to 20 base-pair changes per year. But unavoidable chemical processes damage thousands of DNA nucleotides in a typical mammalian cell every day. Genetic information can be stored stably in DNA sequences only because a large variety of DNA repair enzymes continuously scan the DNA and replace the damaged nucleotides.

The process of DNA repair depends on the presence of a separate copy of the genetic information in each strand of the DNA double helix. An accidental lesion on one strand can therefore be cut out by a repair enzyme and a good strand resynthesized from the information in the undamaged strand. Most of the damage to DNA bases is excised by one of two major pathways. In base excision repair an altered base is removed by a DNA glycosylase enzyme, followed by excision of the resulting sugar phosphate. In nucleotide excision repair a small region of the strand surrounding the damage is removed from the DNA helix as an oligonucleotide. In both cases the small gap left in the DNA helix is filled in by the sequential action of DNA polymerase and DNA ligase.

Introduction

Besides maintaining the integrity of DNA sequences by DNA repair, all organisms must duplicate their DNA accurately before every cell division. DNA replication occurs at polymerization rates of about 500 nucleotides per second in bacteria and about 50 nucleotides per second in mammals. Clearly, the proteins that catalyze this process must be both accurate and fast. Speed and accuracy are achieved by means of a multienzyme complex that guides the process and constitutes an elaborate "replication machine."

Base-pairing Underlies DNA Replication as well as DNA Repair ²⁷

DNA templating is the process in which the nucleotide sequence of a DNA strand (or selected portions of a DNA strand) is copied by complementary base-pairing (A with T or U, and G with C) into a complementary nucleic acid sequence (either DNA or RNA). The process entails the recognition of each nucleotide in the DNA strand by an unpolymerized complementary nucleotide and requires that the two strands of the DNA helix be separated, at least transiently, so that the hydrogen bond donor and acceptor groups on each base become exposed for base-pairing. The appropriate incoming single nucleotides are thereby aligned for their enzyme-catalyzed polymerization into a new nucleic acid chain. In 1957 the first such nucleotide polymerizing enzyme, DNA polymerase,was discovered. The substrates for this enzyme were found to be deoxyribonucleoside triphosphates, which are polymerized on a single-stranded DNA template. The stepwise mechanism of this reaction is the one previously illustrated in Figure 6-36 in connection with DNA repair. The discovery of DNA polymerase led to the isolation of RNA polymerase, which was correctly inferred to use ribonucleoside tri-phosphates as its substrates.

During DNA replication each of the two old DNA strands serves as a template for the formation of an entire new strand. Because each of the two daughters of a dividing cell inherits a new DNA double helix containing one old and one new strand (see Figure 3-13), DNA is said to be replicated "semiconservatively" by DNA polymerase.

The DNA Replication Fork Is Asymmetrical ²⁸

Autoradiographic analyses carried out in the early 1960s on whole replicating chromosomes labeled with a short pulse of the radioactive DNA precursor ³H-thymidine revealed a localized region of replication that moves along the parental DNA double helix. Because of its Y-shaped structure, this active region is called a DNA replication fork. At a replication fork the DNA of both new daughter strands is synthesized by a multienzyme complex that contains the DNA polymerase.

Initially, the simplest mechanism of DNA replication appeared to be continuous growth of both new strands, nucleotide by nucleotide, at the replication fork as it moves from one end of a DNA molecule to the other. But because of the antiparallel orientation of the two DNA strands in the DNA double helix (see Figure 3-10 and Panel 3-2, pp. 100-101), this mechanism would require one daughter strand to grow in the 5'-to-3' direction and the other in the 3'-to-5' direction. Such a replication fork would require two different DNA polymerase enzymes. One would polymerize in the 5'-to-3' direction (see Figure 6-36), where each incoming deoxyribonucleoside triphosphate carries the triphosphate activation needed for its own addition. The other would move in the 3'-to-5' direction and work by so-called "head growth," in which the end of the growing DNA chain carries the triphosphate activation required for the addition of each subsequent nucleotide. Although head-growth polymerization occurs elsewhere in biochemistry (see Figure 2-36), it does not occur in DNA synthesis; no 3'-to-5' DNA polymerase has ever been found (Figure 6-40).

How, then, is 3'-to-5' DNA synthesis achieved? The answer was first suggested in the late 1960s by experiments in which highly radioactive ³H-thymidine was added to dividing bacteria for a few seconds so that only the most recently replicated DNA, just behind the replication fork, became radiolabeled. This selective labeling method revealed the transient existence of pieces of DNA that were 1000 to 2000 nucleotides long, now commonly known as Okazaki fragments, at the bacterial growing fork. (Such replication intermediates were later found in eucaryotes, where they are only 100 to 200 nucleotides long.) The Okazaki fragments were shown to be synthesized only in the 5'-to-3' chain direction and to be joined together after their synthesis to create long DNA chains by the same DNA ligase enzyme that seals nicks during DNA repair (see Figure 6-37).

A replication fork has an asymmetric structure. The DNA daughter strand that is synthesized continuously is known as the leading strand, and its synthesis slightly precedes the synthesis of the daughter strand that is synthesized discontinuously, which is known as the lagging strand. The synthesis of the lagging strand is delayed because it must wait for the leading strand to expose the template strand on which each Okazaki fragment is synthesized (Figure 6-41). The synthesis of the lagging strand by a discontinuous, "backstitching" mechanism means that only the 5'-to-3' type of DNA polymerase is needed for DNA replication.

The High Fidelity of DNA Replication Requires a Proofreading Mechanism ²⁹

The fidelity of copying that is observed after DNA replication has occurred is such that only about 1 error is made in every 10⁹ base-pair replications, as required to maintain the mammalian genome of 3 × 10⁹ DNA base pairs. This fidelity is much higher than expected, given that the standard complementary base pairs are not the only ones possible. With small changes in helix geometry, for example, two hydrogen bonds will form between G and T in DNA. In addition, rare tautomeric forms of the four DNA bases occur transiently in ratios of 1 part to 10⁴ or 10⁵. These forms will mispair without a change in helix geometry: the rare tautomeric form of C pairs with A instead of G, for example. If the DNA polymerase accepts a mispairing that occurs between an incoming deoxyribonucleoside triphosphate and the DNA template, the wrong nucleotide can be incorporated into the new DNA chain, producing a mutation. The high fidelity of DNA replication depends on several "proofreading" mechanisms that act sequentially to remove errors brought about in these ways.

One important proofreading process depends on special properties of the DNA polymerase enzyme. Unlike RNA polymerases, DNA polymerases do not begin a new polynucleotide chain by linking two nucleoside triphosphates together. They absolutely require the 3'-OH end of a base-paired primer strand on which to add further nucleotides (see Figure 6-36). Moreover, DNA molecules with a mismatched (not base-paired) nucleotide at the 3'-OH end of the primer strand are not effective as templates. DNA polymerase molecules are able to deal with such mismatched DNAs by means of either a separate catalytic subunit or a covalently linked, separate catalytic site that clips off any unpaired residues at the primer terminus. Clipping by this 3'-to-5' proofreading exonuclease activity continues until enough nucleotides have been removed from the 3' end to regenerate a base-paired terminus that can prime DNA synthesis. In this way DNA polymerase functions as a "self-correcting" enzyme that removes its own polymerization errors as it moves along the DNA. Figure 6-42 illustrates how this proofreading process can correct a base-pairing error.

The requirement for a perfectly base-paired terminus is essential to the self-correcting properties of the DNA polymerase. For such an enzyme to start synthesis in the complete absence of a primer without losing any of its discrimination between base-paired and unpaired growing 3'-OH termini is apparently not possible. By contrast, the RNA polymerase enzymes involved in gene transcription need not be self-correcting: errors in making RNA are not passed on to the next generation, and an occasional defective molecule has no significance. RNA polymerases are able to start new polynucleotide chains without a primer, and an error frequency of about 1 in 10⁴ is found both in RNA synthesis and in the separate process of translating mRNA sequences into protein sequences.

Only DNA Replication in the 5'-to-3' Direction Allows Efficient Error Correction

The need for accuracy probably explains why DNA replication occurs only in the 5'-to-3' direction. If there were a DNA polymerase that added deoxyribonucleoside triphosphates in such a way as to cause chains to grow in the 3'-to-5' chain direction, the growing 5'-chain end rather than the incoming mononucleotide would carry the activating triphosphate. In this case the mistakes in polymerization could not be simply hydrolyzed away, since the bare 5'-chain end thus created would immediately terminate DNA synthesis. It is much easier, therefore, to correct a mismatched base that has just been added to the 3' end than one that has just been added to the 5' end of a DNA chain. Although the type of mechanism for DNA replication shown in Figure 6-41 seems at first sight much more complex than the incorrect mechanism depicted in Figure 6-40, it is much more accurate because it involves DNA synthesis only in the 5'-to-3' direction.

A Special Nucleotide Polymerizing Enzyme Synthesizes Short RNA Primer Molecules on the Lagging Strand ³⁰

For the leading strand a special primer is needed only at the start of replication; once a replication fork is established, the DNA polymerase is continuously presented with a base-paired chain end on which to add new nucleotides. But the DNA polymerase on the lagging side of the fork requires only about 4 seconds to complete each short DNA fragment, after which it must start synthesizing a completely new fragment at a site farther along the template strand (see Figure 6-41). A special mechanism is needed to produce the base-paired primer strand required by this DNA polymerase molecule. The mechanism involves an enzyme called DNA primase, which uses ribonucleoside triphosphates to synthesize short RNA primers (Figure 6-43). These primers are about 10 nucleotides long in eucaryotes, and they are made at intervals on the lagging strand, where they are elongated by the DNA polymerase to begin each Okazaki fragment. The synthesis of each Okazaki fragment ends when this DNA polymerase runs into the RNA primer attached to the 5' end of the previous fragment. To produce a continuous DNA chain from the many DNA fragments made on the lagging strand, a special DNA repair system acts quickly to erase the old RNA primer and replace it with DNA. DNA ligase then joins the 3' end of the new DNA fragment to the 5' end of the previous one to complete the process (Figure 6-44).

Why might an erasable RNA primer be preferred to a DNA primer that need not be erased? The argument that a self-correcting polymerase cannot start chains de novo also implies its converse: an enzyme that starts chains de novo cannot be efficient at self-correction. Thus any enzyme that primes the synthesis of Okazaki fragments will of necessity make a relatively inaccurate copy (at least 1 error in 10⁵). Even if the copies retained in the final product constituted as little as 5% of the total genome (for example, 10 nucleotides per 200-nucleotide DNA fragment), the resulting increase in overall mutation rate would be enormous. It therefore seems likely that the evolution of RNA rather than DNA for priming entailed a powerful advantage, since the ribonucleotides in the primer automatically mark these sequences as "bad copy" to be removed.

Special Proteins Help Open Up the DNA Double Helix in Front of the Replication Fork ³¹

The DNA double helix must be opened up ahead of the replication fork so that the incoming deoxyribonucleoside triphosphates can form base pairs with the template strand. The DNA double helix is very stable under normal conditions: the base pairs are locked in place so strongly that temperatures approaching that of boiling water are required to separate the two strands in a test tube. For this reason most DNA polymerases can copy DNA only when the template strand has already been separated from its complementary strand. Additional proteins are needed to help open the double helix and thus provide the appropriate exposed DNA template for the DNA polymerase to copy. Two types of replication proteins contribute to this processDNA helicases and single-strand DNA-binding proteins.

DNA helicases were first isolated as proteins that hydrolyze ATP when they are bound to single strands of DNA. As described in Chapter 5, the hydrolysis of ATP can change the shape of a protein molecule in a cyclical manner that allows the protein to perform mechanical work. DNA helicases utilize this principle to move rapidly along a DNA single strand; where they encounter a region of double helix, they continue to move along their strand, thereby prying apart the helix (Figure 6-45). We have previously described how a special DNA repair helicase functions in nucleotide excision repair (see Figure 6-38B).

The unwinding of the template DNA helix at a replication fork could in principle be catalyzed by two DNA helicases acting in concert - one running along the leading strand and one along the lagging strand. These two helicases would need to move in opposite directions along a DNA single strand and therefore would have to be different enzymes. Both types of DNA helicase, in fact, do exist, although in bacteria the DNA helicase on the lagging strand plays the predominant role, for reasons that will become clear shortly.

Single-strand DNA-binding (SSB) proteins - also called helix-destabilizing proteins - bind to exposed DNA strands without covering the bases, which therefore remain available for templating. These proteins are unable to open a long DNA helix directly, but they aid helicases by stabilizing the unwound, single-stranded conformation. In addition, their cooperative binding completely coats the regions of single-stranded DNA on the lagging strand, thereby preventing formation of the short hairpin helices that would otherwise impede synthesis by the DNA polymerase (Figure 6-46).

A Moving DNA Polymerase Molecule Is Kept Tethered to the DNA by a Sliding Ring ³²

On their own, most DNA polymerase molecules will synthesize only a short string of nucleotides before falling off a DNA template. This tendency to leave a DNA molecule quickly allows the DNA polymerase molecule that has just finished synthesizing one Okazaki fragment on the lagging strand to be recycled quickly to begin the synthesis of the next Okazaki fragment on the same strand. This rapid dissociation, however, would make it difficult for the polymerase to synthesize long DNA strands at a replication fork were it not for an accessory protein that functions as a regulated clamp. This clamp keeps the polymerase firmly on the DNA when it is moving, but releases it as soon as the polymerase stops.

How can a clamp prevent the polymerase from dissociating without at the same time impeding the polymerase's rapid movement along the DNA molecule? The three-dimensional structure of a clamp protein, determined by x-ray diffraction, indicates that it forms a large ring around the DNA helix. One side of the ring binds to the back of the DNA polymerase, and the whole ring slides freely as the polymerase moves along a DNA strand (Figure 6-47). The assembly of the clamp around DNA requires ATP hydrolysis by special accessory proteins that bind both to the clamp protein and to DNA; it is not known how the clamp is disassembled to remove it from the DNA.

The Proteins at a Replication Fork Cooperate to Form a Replication Machine ³³

Although we have discussed DNA replication as though it were carried out by a mixture of replication proteins that act independently, in reality most of the proteins are held together in a large multienzyme complex that moves rapidly along the DNA. This complex can be likened to a tiny sewing machine composed of protein parts and powered by nucleoside triphosphate hydrolyses. Although the replication complex has been best characterized in E. coli and several of its viruses, a very similar complex operates in eucaryotes (see p. 358).

The functions of the subunits of the replication machine are summarized in the two-dimensional diagram of the complete replication fork shown in Figure 6-48. Two identical DNA polymerase molecules work at the fork, one on the leading strand and one on the lagging strand. The DNA helix is opened by a DNA polymerase molecule clamped on the leading strand, acting in concert with a DNA helicase molecule running along the lagging strand; helix opening is aided by cooperatively bound molecules of single-strand DNA-binding protein. While the DNA polymerase molecule on the leading strand can operate in a continuous fashion, the DNA polymerase molecule on the lagging strand must restart at short intervals, using a short RNA primer made by a DNA primase molecule.

The efficiency of replication is greatly increased by the close association of all these protein components. The primase molecule is linked directly to the DNA helicase to form a unit on the lagging strand called a primosome. Powered by the DNA helicase, the primosome moves with the fork, synthesizing RNA primers as it goes. Similarly, the DNA polymerase molecule that synthesizes DNA on the lagging strand moves in concert with the rest of the proteins, synthesizing a succession of new Okazaki fragments. To accommodate this arrangement, its DNA template strand is thought to be folded back in the manner shown in Figure 6-49. The replication proteins are thus linked together into a single large unit (total mass > 10⁶ daltons) that moves rapidly along the DNA, enabling DNA to be synthesized on both sides of the fork in a coordinated and efficient manner.

This DNA replication machine leaves behind on the lagging strand a series of unsealed Okazaki fragments, which still contain the RNA that primed their synthesis at their 5' ends. This RNA must be removed and the fragments joined up by DNA repair enzymes that operate behind the replication fork (see Figure 6-44).

A Mismatch Proofreading System Removes Replication Errors That Escape from the Replication Machine ³⁴

Bacteria such as E. coli are capable of dividing once every 30 minutes, so it is relatively easy to screen large populations to find rare mutants that are altered in a specific process. One interesting class of mutants contains alterations in so-called mutator genes, which greatly increase the rate of spontaneous mutation. Not surprisingly, one such mutant encodes a defective form of the 3'-to-5' proofreading exonuclease (discussed earlier) that is a subunit of the DNA polymerase enzyme (see Figure 6-42). When this protein is defective, the DNA polymerase no longer proofreads effectively, and many replication errors that would otherwise have been removed accumulate in the DNA.

The study of other E. coli mutants that exhibit abnormally high mutation rates has uncovered another proofreading system that removes replication errors missed by the proofreading exonuclease. This mismatch proofreading system (also called a mismatch repair system) differs from most DNA repair systems in that it does not depend on the presence in the DNA of abnormal nucleotides that can be recognized and excised. Instead, it detects the distortion on the outside of the helix that results from the misfit between noncomplementary base pairs. But if the proofreading system simply recognized a mismatch in newly replicated DNA and randomly excised one of the two mismatched nucleotides, it would make the mistake of "correcting" the original template strand to match the error exactly half the time and would not therefore lower the overall error rate. To be effective, the proofreading system must be able to distinguish and remove the mismatched nucleotide only on the new strand, where the replication error occurred.

The recognition system used by the mismatch proofreading system in E. coli depends on the methylation of selected A residues in the DNA. Methyl groups are added to all A residues in the sequence GATC, but not until some time after the A has been incorporated into a newly synthesized DNA chain. Because only the new strands just behind a replication fork will contain GATC sequences that have not yet been methylated, these new DNA strands can be distinguished from old ones.

More recently, eucaryotic proteins have been discovered that are homologous in their amino acid sequence to several of the bacterial proteins that catalyze mismatch proofreading. As expected, when the genes that encode these proteins are deleted in a yeast cell, mutation rates can increase by 100-fold or more. There must, however, be some important differences between the bacterial and eucaryotic proofreading mechanisms, as the mechanism for distinguishing the newly synthesized strand from the parental template strand at the site of a mismatch cannot depend on DNA methylation as in bacteria, since some eucaryotes, such as yeasts and Drosophila, do not methylate any of their DNA. Newly synthesized DNA strands are known to be preferentially nicked, and it has been suggested that such nicks (single-strand breaks) provide the signal that directs mismatch proofreading to the appropriate strand in a eucaryotic cell (Figure 6-50).

Replication Forks Initiate at Replication Origins ³⁵

In both bacteria and mammals replication forks originate at a structure called a replication bubble, a local region where the two strands of the parental DNA helix have been separated from each other to serve as templates for DNA synthesis (Figure 6-51). For bacteria, yeasts, and several viruses that grow in mammalian cells, replication bubbles have been shown to form at special DNA sequences called replication origins, which can be as long as 300 nucleotides. For reasons that are not clear, the replication origins in mammalian chromosomes have thus far been very difficult to characterize at the molecular level.

For several well-defined replication origins, it has been possible to reproduce the fork initiation reaction in vitro. The in vitro studies reveal that fork initiation in bacteria and bacterial viruses starts in the manner indicated in Figure 6-52. Initiator proteins bind in multiple copies to specific sites at the replication origin, wrapping the DNA around them to form a large protein-DNA complex. This complex then binds the DNA helicase and loads it onto an exposed DNA single strand in an adjacent region of helix. The DNA primase also binds, forming the primosome, which moves away from the origin and makes an RNA primer that starts the first DNA chain. This quickly leads to assembly of the remaining proteins to create two replication protein complexes moving away from the origin in opposite directions (see Figure 6-51); these continue to synthesize DNA until all of the DNA template downstream of each fork has been replicated.

Replication fork initiation in eucaryotic chromosomes is discussed in detail in Chapter 8.

DNA Topoisomerases Prevent DNA Tangling During Replication ³⁶

When we draw the DNA helix (incorrectly) as a flat, ladderlike structure, we are ignoring the "winding problem" that arises during DNA replication. Every 10 base pairs replicated at the fork correspond to one complete turn about the axis of the parental double helix. Therefore, for a replication fork to move, the entire chromosome ahead of the fork would normally have to rotate rapidly (Figure 6-53), which would require large amounts of energy for long chromosomes. An alternative strategy is used during DNA replication: a swivel is formed in the DNA helix by proteins known as DNA topoisomerases.

A DNA topoisomerase can be viewed as a reversible nuclease that adds itself covalently to a DNA phosphate, thereby breaking a phosphodiester bond in a DNA strand. Because the covalent linkage that joins a topoisomerase to a DNA phosphate retains the energy of the cleaved phosphodiester bond, the cleavage reaction is reversible; resealing is rapid and does not require additional energy input. The rejoining mechanism is different in this respect from that of the enzyme DNA ligase, discussed previously (see Figure 6-37).

One type of topoisomerase (topoisomerase I) causes a single-strand break (or nick), which can allow the two sections of DNA helix on either side of the nick to rotate freely relative to each other, using the phosphodiester bond in the strand opposite the nick as a swivel point (Figure 6-54). Any tension in the DNA helix will drive this rotation in the direction that relieves the tension. As a result, DNA replication can occur with the rotation of only a short length of helix - the part just ahead of the fork. The analogous problem that arises during DNA transcription is solved in a similar way.

A second type of DNA topoisomerase (topoisomerase II) forms a covalent linkage to both strands of the DNA helix at the same time, making a transient double-strand break in the helix. These enzymes are activated by sites on chromosomes where two double helices cross over each other. When the topo-isomerase binds to such a crossing site, it (1) breaks one double helix reversibly to create a DNA "gate," (2) causes the second, nearby double helix to pass through this break, and (3) reseals the break and dissociates from the DNA. In this way type II DNA topoisomerases can efficiently separate two interlocked DNA circles (Figure 6-55). The same reaction prevents the severe DNA tangling problems that would otherwise arise during DNA replication. For example, mutant yeast cells have been isolated that produce, in place of the normal topoisomerase II, a version that is inactive at 37°C. When the mutant yeast cells are warmed to this temperature, their chromosomes remain intertwined at mitosis and are unable to separate. The usefulness of topoisomerase II for untangling chromosomes can readily be appreciated by anyone who has struggled to remove a tangle from a fishing line without the aid of scissors.

DNA Replication Is Basically Similar in Eucaryotes and Procaryotes ³⁷

Much of what we know about DNA replication comes from studies of purified bacterial and bacteriophage multienzyme systems capable of DNA replication in vitro. The development of these systems in the 1970s was greatly facilitated by the prior isolation of mutants in a variety of replication genes; these mutants were exploited to identify and purify the corresponding replication proteins.

Less is known about the detailed enzymology of DNA replication in eucaryotes, largely because it is difficult to obtain replication-deficient mutants. Nevertheless, the basic mechanisms of DNA replication, including both the geometry of the replication fork and the protein components of the multiprotein replication machine, are similar for procaryotes and eucaryotes (see Figure 8-35). The major difference is that eucaryotic DNA is replicated not as bare DNA but as chromatin, in which the DNA is complexed with tightly bound proteins called histones. As described in Chapter 8, histones form disclike structures around which the eucaryotic DNA is wound, creating a repeating structural unit called a nucleosome. Nucleosomes are spaced at intervals of about 200 base pairs along the DNA, which may be why new Okazaki fragments are synthesized on the lagging strand at intervals of 100 to 200 nucleotides in eucaryotes instead of at intervals of 1000 to 2000 nucleotides as in bacteria. Nucleosomes may also act as barriers that slow down the movement of DNA polymerase molecules, which could explain why eucaryotic replication forks move only one-tenth as fast as bacterial replication forks.

Summary

A self-correcting DNA polymerase catalyzes nucleotide polymerization in a 5'-to-3' direction, copying a DNA template with remarkable fidelity. Since the two strands of a DNA double helix are antiparallel, this 5'-to-3' DNA synthesis can take place continuously on only one of the strands at a replication fork (the leading strand). On the lagging strand short DNA fragments are made by a "backstitching" process. Because the self-correcting DNA polymerase cannot start a new chain, these lagging-strand DNA fragments are primed by short RNA primer molecules that are subsequently erased and replaced with DNA.

DNA replication requires the cooperation of many proteins, including (1) DNA polymerase and DNA primase to catalyze nucleoside triphosphate polymerization, (2) DNA helicases and single-strand binding proteins to help open up the DNA helix so that it can be copied, (3) DNA ligase and an enzyme that degrades RNA primers to seal together the discontinuously synthesized lagging-strand DNA fragments, (4) DNA topoisomerases to help relieve helical winding and tangling problems, and (5) initiator proteins that bind to specific DNA sequences at a replication origin and catalyze the formation of a replication fork at that site. At a replication origin a specialized protein-DNA structure is formed that subsequently loads a DNA helicase onto the DNA template; other proteins are then added to form the multienzyme "replication machine" that catalyzes DNA synthesis.

Introduction

In the two preceding sections we discussed the mechanisms by which DNA sequences in cells are maintained from generation to generation with very little change. Although such genetic stability is crucial for the survival of individuals, in the longer term the survival of organisms may depend on genetic variation, through which they can adapt to a changing environment. Thus an important property of the DNA in cells is its ability to undergo rearrangements that can vary the particular combination of genes present in any individual genome, as well as the timing and the level of expression of these genes. These DNA rearrangements are caused by genetic recombination. Two broad classes of genetic recombination are commonly recognized - general recombination and site-specific recombination.

In general recombination, genetic exchange takes place between any pair of homologous DNA sequences, usually located on two copies of the same chromosome. One of the most important examples is the exchange of sections of homologous chromosomes (homologues) in the course of meiosis. This "crossing-over" occurs between tightly apposed chromosomes early in the development of eggs and sperm (discussed in Chapter 20), and it allows different versions (alleles) of the same gene to be tested in new combinations with other genes, increasing the chance that at least some members of a mating population will survive in a changing environment. Although meiosis occurs only in eucaryotes, the advantage of this type of gene mixing is so great that mating and the reassortment of genes by general recombination is also widespread in bacteria.

DNA homology is not required in site-specific recombination. Instead, exchange occurs at short, specific nucleotide sequences (on either one or both of the two participating DNA molecules) that are recognized by a variety of site-specific recombination enzymes. Site-specific recombination therefore alters the relative positions of nucleotide sequences in genomes. In some cases these changes are scheduled and organized, as when an integrated bacterial virus is induced to leave a chromosome of a bacterium under stress (see Figure 6-80); in others they are haphazard, as when the DNA sequence of a transposable element is inserted at a randomly selected site in a chromosome.

As for DNA replication, most of what we know about the biochemistry of genetic recombination has come from studies of procaryotic organisms, especially of E. coli and its viruses.

General Recombination Is Guided by Base-pairing Interactions Between Complementary Strands of Two Homologous DNA Molecules ³⁹

General recombination involves DNA strand-exchange intermediates that require some effort to understand. Although the exact pathway followed is likely to be different in different organisms, detailed genetic analyses of viruses, bacteria, and fungi suggest that the major outcome of general recombination is always the same. (1) Two homologous DNA molecules "cross over"; that is, their double helices break and the two broken ends join to their opposite partners to re-form two intact double helices, each composed of parts of the two initial DNA molecules (Figure 6-56). (2) The site of exchange (that is, where a red double helix is joined to a green double helix in Figure 6-56) can occur anywhere in the homologous nucleotide sequences of the two participating DNA molecules. (3) At the site of exchange, a strand of one DNA molecule becomes base-paired to a strand of the second DNA molecule to create a staggered joint (usually called a heteroduplex joint) between the two double helices (Figure 6-57). The heteroduplex region can be thousands of base pairs long; we shall explain later how it forms. (4) No nucleotide sequences are altered at the site of exchange; the cleavage and rejoining events occur so precisely that not a single nucleotide is lost or gained. Despite this precision, general recombination creates DNA molecules of novel sequence: the heteroduplex joint can contain a small number of mismatched base pairs, and, more important, the two DNAs that cross over are usually not exactly the same on either side of the joint.

The mechanism of general recombination ensures that two regions of DNA double helix undergo an exchange reaction only if they have extensive sequence homology. The formation of a heteroduplex joint requires that such homology be present because it involves a long region of complementary base-pairing between a strand from one of the two original double helices and a complementary strand from the other. But how does this heteroduplex joint arise, and how do the two homologous regions of DNA at the site of crossing-over recognize each other? As we shall see, recognition takes place by means of a direct base-pairing interaction. The formation of base pairs between complementary strands from the two DNA molecules then guides the general recombination process, allowing it to occur only between long regions of matching DNA sequence.

General Recombination Can Be Initiated at a Nick in One Strand of a DNA Double Helix ⁴⁰

Each of the two strands in a DNA molecule is helically wound around the other. As a result, extensive base-pair interactions can occur between two homologous DNA double helices only if a nick is first made in a strand of one of them, freeing that strand for the unwinding and rewinding events required to form a heteroduplex with another DNA molecule. For the same reason, any exchange of strands between two DNA double helices requires at least two nicks, one in a strand of each interacting double helix. Finally, to produce the heteroduplex joint illustrated in Figure 6-57, each of the four strands present must be cut to allow each to be joined to a different partner. In general recombination, these nicking and resealing events are coordinated so that they occur only when two DNA helices share an extensive region of matching DNA sequence.

There is evidence from a number of sources that a single nick in only one strand of a DNA molecule is sufficient to initiate general recombination. Chemical agents or types of irradiation that introduce single strand nicks, for example, will trigger a genetic recombination event. Moreover, one of the special proteins required for general recombination in E. colithe RecBCD proteinhas been shown to make single strand nicks in DNA molecules. The RecBCD protein is also a DNA helicase, hydrolyzing ATP and traveling along a DNA helix transiently exposing its strands. By combining its nuclease and helicase activities, the RecBCD protein will create a single-stranded "whisker" on the DNA double helix (Figure 6-58). Figure 6-59 shows how such a whisker could initiate a base-pairing interaction between two complementary stretches of DNA double helix.

DNA Hybridization Reactions Provide a Simple Model for the Base-pairing Step in General Recombination ⁴¹

In its simplest form, the type of base-pairing interaction central to general recombination can be mimicked in a test tube by allowing a DNA double helix to re-form from its separated single strands. This process, called DNA renaturation or hybridization, occurs when a rare random collision juxtaposes complementary nucleotide sequences on two matching DNA single strands, allowing the formation of a short stretch of double helix between them. This relatively slow helix nucleation step is followed by a very rapid "zippering" step as the region of double helix is extended to maximize the number of base-pairing interactions (Figure 6-60).

Formation of a new double helix in this way requires that the annealing strands be in an open, unfolded conformation. For this reason in vitro hybridization reactions are carried out at high temperature or in the presence of an organic solvent such as formamide; these conditions "melt out" the short hairpin helices formed where base-pairing interactions occur within a single strand that folds back on itself. Bacterial cells could not survive such harsh conditions and instead use a single-strand binding protein, the SSB protein, to open their helices. This protein is essential for DNA replication as well as for general recombination in E. coli;it binds tightly and cooperatively to the sugar-phosphate backbone of all single-stranded regions of DNA, holding them in an extended conformation with their bases exposed (see Figure 6-46). In this extended conformation a DNA single strand can base-pair efficiently with either a nucleoside triphosphate molecule (in DNA replication) or a complementary section of another DNA single strand (in genetic recombination). When hybridization reactions are carried out in vitro under conditions that mimic those inside a cell, the SSB protein speeds up the rate of DNA helix nucleation and thereby the overall rate of strand annealing by a factor of more than 1000.

The RecA Protein Enables a DNA Single Strand to Pair with a Homologous Region of DNA Double Helix in E. coli⁴²

General recombination is more complex than the simple hybridization reactions just described. In the course of general recombination, a single DNA strand from one DNA double helix must invade another double helix (see Figure 6-59). In E. coli this requires the RecA protein, produced by the recA gene, which was identified in 1965 as being essential for recombination between chromosomes. Long sought by biochemists, this elusive gene product was finally purified to homogeneity in 1976, a feat that allowed its detailed characterization (Figure 6-61). Like a single-strand binding (SSB) protein, the RecA protein binds tightly and in large cooperative clusters to single-stranded DNA to form a nucleoprotein filament. This filament has several distinctive properties. The RecA protein has more than one DNA-binding site, for example, and it can therefore hold a single strand and a double helix together. These sites allow the RecA protein to catalyze a multistep reaction (called synapsis) between a DNA double helix and a homologous region of single-stranded DNA. The crucial step in synapsis occurs when a region of homology is identified by an initial base-pairing between complementary nucleotide sequences. The nucleation step in this case appears to involve a three-stranded structure, in which the DNA single strand forms nonconventional base pairs in the major groove of the DNA double helix (Figure 6-62). This begins the pairing shown previously in Figure 6-59 and so initiates the exchange of strands between two recombining DNA double helices. Studies in vitro suggest that the E. coli SSB protein cooperates with the RecA protein to facilitate these reactions.

Once synapsis has occurred, a short heteroduplex region where the strands from two different DNA molecules have begun to pair is enlarged through protein-directed branch migration, which can also be catalyzed by the RecA protein. Branch migration can take place at any point where two single DNA strands with the same sequence are attempting to pair with the same complementary strand; an unpaired region of one of the single strands will displace a paired region of the other single strand, moving the branch point without changing the total number of DNA base pairs. Spontaneous branch migration proceeds equally in both directions, and so it makes little progress and is unlikely to complete recombination efficiently (Figure 6-63A). Because the RecA protein catalyzes unidirectional branch migration, it readily produces a region of heteroduplex that is thousands of base pairs long (Figure 6-63B).

The catalysis of branch migration depends on a further property of the RecA protein. In addition to having two DNA-binding sites, the RecA protein is a DNA-dependent ATPase, with an additional site for binding and hydrolyzing ATP. The protein associates much more tightly with DNA when it has ATP bound than when it has ADP bound. Moreover, new RecA molecules with ATP bound are preferentially added at one end of the RecA protein filament, and the ATP is then hydrolyzed to ADP. The RecA protein filaments that form on DNA may therefore share many of the dynamic assembly properties displayed by the cytoskeletal filaments formed from actin or tubulin (discussed in Chapter 16); an ability of the protein to "treadmill" unidirectionally along a DNA strand, for example, could drive the branch migration reaction shown in Figure 6-63B.

General Genetic Recombination Usually Involves a Cross-Strand Exchange ⁴³

Exchanging a single strand between two double helices is presumed to be the slow and difficult step in a general recombination event (see Figure 6-59). After this initial exchange, extending the region of pairing and establishing further strand exchanges between the two closely apposed helices is thought to proceed rapidly. During these events a limited amount of nucleotide excision and local DNA resynthesis often occurs, resembling some of the events in DNA repair. Because of the large number of possibilities, different organisms are likely to follow different pathways at this stage. In most cases, however, an important intermediate structure, the cross-strand exchange, will be formed by the two participating DNA helices. One of the simplest ways in which this structure can form is shown in Figure 6-64.

In the cross-strand exchange (also called a Holliday junction) the two homologous DNA helices that initially paired are held together by mutual exchange of two of the four strands present, one originating from each of the helices. No disruption of base-pairing is necessary to maintain this structure, which has two important properties(1) the point of exchange between the two homologous DNA double helices (where the two strands cross in Figure 6-64) can migrate rapidly back and forth along the helices by a double branch migration; (2) the cross-strand exchange contains two pairs of strands: one pair of crossing strands and one pair of noncrossing strands. The structure can isomerize, however, by undergoing a series of rotational movements, so that the two original noncrossing strands become crossing strands and vice versa (Figure 6-65).

In order to regenerate two separate DNA helices and thus terminate the pairing process, the two crossing strands must be cut. If the crossing strands are cut before isomerization, the two original DNA helices separate from each other nearly unaltered, with only a very short piece of single-stranded DNA exchanged. If the crossing strands are cut after isomerization, however, one section of each original DNA helix is joined to a section of the other DNA helix; in other words, the two DNA helices have crossed over (see Figure 6-65).

The isomerization of the cross-strand exchange should occur spontaneously at some rate, but it may also be enzymatically driven or otherwise regulated by cells. Some kind of control probably operates during meiosis, when the two DNA double helices that pair are constrained in an elaborate structure called the synaptonemal complex (discussed in Chapter 20).

Gene Conversion Results from Combining General Recombination and Limited DNA Synthesis ⁴⁴

It is a fundamental law of genetics that each parent makes an equal genetic contribution to the offspring, one complete set of genes being inherited from the father and one from the mother. Thus, when a diploid cell undergoes meiosis to produce four haploid cells (discussed in Chapter 20), exactly half of the genes in these cells should be maternal (genes that the diploid cell inherited from its mother) and the other half paternal (genes that the diploid cell inherited from its father). In a complex animal, such as a human, it is not possible to check this prediction directly. But in other organisms, such as fungi, where it is possible to recover and analyze all four of the daughter cells produced from a single cell by meiosis, one finds cases in which the standard genetic rules have apparently been violated. Occasionally, for example, meiosis yields three copies of the maternal version of a gene (allele) and only one copy of the paternal allele, indicating that one of the two copies of the paternal allele has been changed to a copy of the maternal allele. This phenomenon is known as gene conversion. It often occurs in association with general genetic recombination events, and it is thought to be important in the evolution of certain genes (see Figure 8-74). Gene conversion is believed to be a straightforward consequence of the mechanisms of general recombination and DNA repair.

During meiosis heteroduplex joints are formed at the sites of crossing-over between homologous maternal and paternal chromosomes. If the maternal and paternal DNA sequences are slightly different, the heteroduplex joint may include some mismatched base pairs. The resulting mismatch in the double helix may then be corrected by the DNA repair machinery, which either can erase nucleotides on the paternal strand and replace them with nucleotides that match the maternal strand or vice versa. The consequence of this mismatch repair will be a gene conversion. Gene conversion can also take place by a number of other mechanisms, but they all require some type of general recombination event that brings two copies of a closely related DNA sequence together. Because an extra copy of one of the two DNA sequences is generated, a limited amount of DNA synthesis must also be involved. Genetic studies show that usually only small sections of DNA undergo gene conversion, and in many cases only part of a gene is changed.

Gene conversion can also occur in mitotic cells, but it does so more rarely. As in meiotic cells, some gene conversions in mitotic cells probably result from a mismatch repair process operating on heteroduplex DNA. Another likely mechanism in both meiotic and mitotic cells is illustrated in Figure 6-66.

Mismatch Proofreading Can Prevent Promiscuous Genetic Recombination Between Two Poorly Matched DNA Sequences ⁴⁵

As previously discussed, general recombination is triggered whenever two DNA strands of complementary sequence pair to form a heteroduplex joint between two double helices (see Figure 6-64). Experiments carried out in vitrowith purified RecA protein show that pairing can occur efficiently even when the sequences of the two DNA strands do not match well - when, for example, only four out of every five nucleotides on average can form base pairs. How, then, do vertebrate cells avoid promiscuous general recombination between the many thousands of copies of closely related DNA sequences that are repeated in their genomes (see p. 395)?

Although the answer is not known, studies with bacteria and yeasts demonstrate that the same mismatch proofreading system that removes replication errors (see Figure 6-50) has the additional role of interrupting genetic recombination events between imperfectly matched DNA sequences. It has long been known, for example, that homologous genes in two closely related bacteria, Escherichia coli and Salmonella typhimurium, generally will not recombine, even though their nucleotide sequences are 80% identical; when the mismatch proofreading system is inactivated by mutation, however, there is a 1000-fold increase in the frequency of such interspecies recombination events. It is thought, then, that the mismatch proofreading system normally recognizes the mispaired bases in an initial strand exchange and prevents the subsequent steps required to break and rejoin the two paired DNA helices. This mechanism protects the bacterial genome from the sequence changes that would otherwise be caused by recombination with foreign DNA molecules that occasionally enter the cell. In vertebrate cells, which contain many closely related DNA sequences, the same type of proofreading is thought to help prevent promiscuous recombination events that would otherwise scramble the genome (Figure 6-67).

Site-specific Recombination Enzymes Move Special DNA Sequences into and out of Genomes ⁴⁶

Site-specific genetic recombination, unlike general recombination, is guided by a recombination enzyme that recognizes specific nucleotide sequences present on one or both of the recombining DNA molecules. Base-pairing between the recombining DNA molecules need not be involved, and even when it is, the heteroduplex joint that is formed is only a few base pairs long. By separating and joining double-stranded DNA molecules at specific sites, this type of recombination enables various types of mobile DNA sequences to move about within and between chromosomes.

Site-specific recombination was first discovered as the means by which a bacterial virus, bacteriophage lambda, moves its genome into and out of the E. coli chromosome. In its integrated state the virus is hidden in the bacterial chromosome and replicated as part of the host's DNA. When the virus enters a cell, a virus-encoded enzyme called lambda integrase is synthesized. This enzyme catalyzes a recombination process that begins when several molecules of the integrase protein bind tightly to a specific DNA sequence on the circular bacteriophage chromosome. The resulting DNA-protein complex can now bind to a related but different specific DNA sequence on the bacterial chromosome, bringing the bacterial and bacteriophage chromosomes close together. The integrase then catalyzes the required DNA cutting and resealing reactions, using a short region of sequence homology to form a tiny heteroduplex joint at the point of union (Figure 6-68). The integrase resembles a DNA topoisomerase in that it forms a reversible covalent linkage to DNA wherever it breaks a DNA chain.

The same type of site-specific recombination mechanism can also be carried out in reverse by the lambda bacteriophage, enabling it to exit from its integration site in the E. coli chromosome in order to multiply rapidly within the bacterial cell. This excision reaction is catalyzed by a complex of the integrase enzyme with a second bacteriophage protein, which is produced by the virus only when its host cell is stressed. If the sites recognized by such a recombination enzyme are flipped, the DNA between them will be inverted rather than excised (see Figure 9-57).

Many other enzymes that catalyze site-specific recombination resemble lambda integrase in requiring a short region of identical DNA sequence on the two regions of DNA helix to be joined. Because of this requirement, each enzyme in this class is fastidious with respect to the DNA sequences that it recombines, and it can be expected to catalyze one particular DNA joining event that is useful to the virus, plasmid, transposable element, or cell that contains it. These enzymes can be exploited as tools in transgenic animals to study the influence of specific genes on cell behavior, as illustrated in Figure 6-69.

Site-specific recombination enzymes that break and rejoin two DNA double helices at specific sequences on each DNA molecule often do so in a reversible way: as for lambda bacteriophage, the same enzyme system that joins two DNA molecules can take them apart again, precisely restoring the sequences of the two original DNA molecules. This type of recombination is therefore called conservative site-specific recombination to distinguish it from the mechanistically distinct transpositional site-specific recombination that we discuss next.

Transpositional Recombination Can Insert a Mobile Genetic Element into Any DNA Sequence ⁴⁷

Many mobile DNA sequences, including many viruses and transposable elements, encode integrases that insert their DNA into a chromosome by a mechanism that is different from that used by bacteriophage lambda. Like the lambda integrase, each of these enzymes recognizes a specific DNA sequence in the particular mobile genetic element whose recombination it catalyzes. Unlike the lambda enzyme, however, these integrases do not require a specific DNA sequence in the "target" chromosome and they do not form a heteroduplex joint. Instead, they introduce cuts into both ends of the linear DNA sequence of the mobile genetic element and then catalyze a direct attack by these DNA ends on the target DNA molecule, breaking two closely spaced phosphodiester bonds in the target molecule. Because of the way that these breaks are made, two short single-stranded gaps are left in the recombinant DNA molecule, one at each end of the mobile element; these are filled in by DNA polymerase to complete the recombination process. As illustrated in Figure 6-70, this mechanism creates a short duplication of the adjacent target DNA sequence; such flanking duplications are the hallmark of a transpositional site-specific recombination event.

An integrase enzyme of this type was first purified in active form from bacteriophage Mu. Like the bacteriophage lambda integrase, it carries out all of its cutting and rejoining reactions without requiring an energy source (such as ATP). Very similar enzymes are present in organisms as diverse as bacteria, fruit flies, and humans - all of which contain mobile genetic elements, as we discuss next.

Summary

Genetic recombination mechanisms allow large sections of DNA double helix to move from one chromosome to another. There are two broad classes of recombination events. In general recombination the initial reactions rely on extensive base-pairing interactions between strands of the two DNA double helices that will recombine. As a result, general recombination occurs only between two homologous DNA molecules, and although it moves sections of DNA back and forth between chromosomes, it does not normally change the arrangement of the genes in a chromosome. Site-specific recombination, on the other hand, alters the relative positions of nucleotide sequences in chromosomes because the pairing reactions depend on a protein-mediated recognition of the two DNA sequences that will recombine, and extensive sequence homology is not required. Two site-specific recombination mechanisms are common: (1) conservative site-specific recombination, which produces a very short heteroduplex and therefore requires some DNA sequence that is the same on the two DNA molecules, and (2) transpositional site-specific recombination, which produces no heteroduplex and usually does not require a specific sequence on the target DNA.

Introduction

In our description of the basic genetic mechanisms, we have so far focused on their selective advantage for the cell. We saw that the short-term survival of the cell depends on the maintenance of genetic information by DNA repair, while the multiplication of the cell requires rapid and accurate DNA replication. On a longer time scale the appearance of genetic variants, on which evolution of the species depends, is greatly facilitated by the reassortment of genes and the occasional rearrangement of DNA sequences caused by genetic recombination. We shall now examine a group of genetic elements that seem to act as parasites, subverting the genetic mechanisms of the cell for their own benefit. These genetic elements are interesting in their own right. In addition, because they must heavily exploit the metabolism of the host cell in order to multiply, they serve as powerful tools for investigating the normal cell machinery.

Many DNA sequences can replicate independently of the rest of the genome. Such sequences have widely different degrees of independence from their host cells. Of these, virus chromosomes are the most independent because they have a protein coat that allows them to move freely from cell to cell. To varying degrees, the viruses are closely related to plasmids and transposable elements, which are DNA sequences that lack a coat and are therefore more host-cell-dependent and confined to replicate within a single cell and its progeny. More primitive still are some DNA sequences that are suspected of being mobile because they are repeated many times in a cell's chromosome. They move or multiply so rarely, however, that it is not clear if they should be considered as separate genetic elements at all.

We begin our discussion with viruses, which are the best understood of the mobile genetic elements. Then we describe the properties of plasmids and transposable elements, some of which bear a remarkable resemblance to viruses and may in fact have been their ancestors. The many repetitive DNA sequences in vertebrate chromosomes are discussed in Chapter 8.

Viruses Are Mobile Genetic Elements ⁴⁹

Viruses were first described as disease-causing agents that can multiply only in cells and that by virtue of their tiny size pass through ultrafine filters that hold back even the smallest bacteria. Before the advent of the electron microscope, their nature was obscure, although it was suspected that they might be naked genes that had somehow acquired the ability to move from one cell to another. The use of ultracentrifuges in the 1930s made it possible to separate viruses from host cell components, and by the early 1940s the generalization emerged that all viruses contain nucleic acids. The idea that viruses and genes carry out similar functions was confirmed by studies on bacteriophages, which are bacterial viruses. In 1952 it was shown for the bacteriophage T4 that only the phage DNA, and not the phage protein, enters the bacterial host cell and initiates the replication events that lead to the production of several hundred progeny viruses in every infected cell.

These observations led to the notion of viruses as genetic elements enclosed by a protective coat that enables them to move from one cell to another. Virus multiplication per se is often lethal to the cells in which it occurs; in many cases the infected cell breaks open (lyses) and thereby allows the progeny viruses access to nearby cells. Many of the clinical manifestations of viral infection reflect this cytolytic effect of the virus. Both the cold sores formed by herpes simplex virus and the lesions caused by smallpox, for example, reflect the killing of the epithelial cells in a local area of the skin.

As we shall see, the type of nucleic acid in a virus, the structure of its coat, its mode of entry into the host cell, and its mechanism of replication once inside all vary from one type of virus to another.

The Outer Coat of a Virus May Be a Protein Capsid or a Membrane Envelope ⁵⁰

Initially, it was thought that the outer coat of a virus might be constructed from a single type of protein molecule. Viral infections were believed to start with the dissociation of the viral chromosome (its nucleic acid) from its protein coat, followed by replication of the chromosome inside the host cell, to form many identical copies. After the synthesis of new copies of the virus-specific coat protein from virally encoded messenger RNA molecules, formation of the progeny virus particles would occur by the spontaneous assembly of these coat protein molecules around the progeny viral chromosomes (Figure 6-71).

It is now known that these ideas vastly oversimplify the diversity of virus life cycles. The protein shell that surrounds the nucleic acid of most viruses (the capsid), for example, contains more than one type of polypeptide chain, often arranged in several layers (Figure 6-72). In many viruses, moreover, the protein capsid is further enclosed by a lipid bilayer membrane that contains proteins. Many of these enveloped viruses acquire their envelope in the process of budding from the plasma membrane (Figure 6-73). This budding process allows the virus particles to leave the cell without disrupting the plasma membrane and, therefore, without killing the cell. Electron micrographs that emphasize the differences among viral coats are presented in Figure 6-74.

Viral Genomes Come in a Variety of Forms and Can Be Either RNA or DNA ⁵¹

As discussed earlier, the DNA double helix has the advantages of stability and easy repair. If one polynucleotide chain is accidentally damaged, its complementary chain permits the damage to be readily corrected. This concern with repair, however, need not bother small viral chromosomes that contain only several thousand nucleotides. The chance of accidental damage is very small compared with the risk to a cell genome containing millions of nucleotides.

The genetic information of a virus can, therefore, be carried in a variety of unusual forms, including RNA instead of DNA. A viral chromosome may be a single-stranded RNA chain, a double-stranded RNA helix, a circular single-stranded DNA chain, or a linear single-stranded DNA chain. Moreover, although some viral chromosomes are simple linear DNA double helices, circular DNA double helices and more complex linear DNA double helices are also common. Several viruses have protein molecules covalently attached to the 5' ends of their DNA strands, for example, and the DNA double helices from the very large poxviruses have their opposite strands at each end covalently joined through phosphodiester linkages (Figure 6-75).

A Viral Chromosome Codes for Enzymes Involved in the Replication of Its Nucleic Acid ⁵²

Each type of viral genome requires unique enzymatic tricks for its replication and thus must encode not only the viral coat protein but also one or more of the enzymes needed to replicate the viral nucleic acid. The amount of information that a virus brings into a cell to ensure its own selective replication varies greatly. The DNA of the relatively large bacteriophage T4, for example, contains about 300 genes, including at least 30 genes that ensure the rapid replication of the T4 chromosome in its E. coli host cell (Figure 6-76). T4 DNA replication has the unusual feature that 5-hydroxymethyl-C is incorporated in place of C in its DNA. The unusual base composition of the T4 DNA makes it readily distinguishable from host DNA and selectively protects it from nucleases encoded in the T4 genome that thus degrade only the E. coli chromosome. Still other proteins alter host cell RNA polymerase molecules so that they are unable to transcribe E. coli DNA and instead transcribe different sets of bacteriophage genes at different stages of infection, according to the needs of the phage.

Smaller DNA viruses, such as the monkey virus SV40 and the tiny bacterio-phage M13, carry much less genetic information. They rely heavily on host-cell enzymes to carry out their DNA synthesis, parasitizing most of the host-cell DNA replication proteins. Most DNA viruses, however, code for proteins that selectively initiate the synthesis of their own DNA, recognizing a particular nucleotide sequence in the virus that serves as a replication origin. This is important because a virus must override the cellular control signals that would otherwise cause the viral DNA to replicate in pace with the host cell DNA, doubling only once in each cell cycle. We do not yet understand very much about how eucaryotic cells regulate their own DNA synthesis, and the mechanisms used by viruses to escape from this regulation - which are much more accessible to study - provide insights into the host mechanisms.

RNA viruses have particularly specialized requirements for replication, since to reproduce their genomes they must copy RNA molecules, which means polymerizing nucleoside triphosphates on an RNA template. Cells normally do not have enzymes to carry out this reaction, so even the smallest RNA viruses must encode their own RNA-dependent polymerase enzymes in order to replicate. We now look in more detail at the replication mechanisms of the various types of viruses.

Both RNA Viruses and DNA Viruses Replicate Through the Formation of Complementary Strands ⁵³

Like DNA replication, the replication of the genomes of RNA viruses occurs through the formation of complementary strands. For most RNA viruses this process is catalyzed by specific RNA-dependent RNA polymerase enzymes (replicases). These enzymes are encoded by the viral RNA chromosome and are often incorporated into the progeny virus particles, so that upon entry of the virus into a cell, they can immediately begin replicating the viral RNA. Replicases are always packaged into the capsid of the so-called negative-strand RNA viruses, such as influenza or vesicular stomatitis virus. Negative-strand viruses are so called because the infecting single strand does not code for protein; instead its complementary strand carries the coding sequences. Thus the infecting strand remains impotent without a preformed replicase. In contrast, the viral RNA of positive-strand RNA viruses, such as poliovirus, can serve as mRNA and produce a replicase once it enters the cell; therefore the naked genome itself is infectious.

The synthesis of viral RNA always begins at the 3' end of the RNA template, starting with the synthesis of the 5' end of the new viral RNA molecule and progressing in the 5'-to-3' direction until the 5' end of the template is reached. There are no error-correcting mechanisms for viral RNA synthesis, and error rates are similar to those in DNA transcription (about 1 error in 10⁴ nucleotides synthesized). This is not a serious deficiency as long as the RNA chromosome is relatively short; for this reason the genomes of all RNA viruses are small relative to those of the large DNA viruses.

All DNA viruses begin their replication at a replication origin, where special initiator proteins bind and then attract the replication enzymes of the host cell (see Figure 8-34). There are many different replication pathways, however. The complexity of these diverse replication schemes reflects, in part, the problem of replicating the ends of a simple linear DNA molecule, given a DNA polymerase enzyme that cannot begin synthesis without a primer (see pp. 253-254). DNA viruses have solved this problem in a variety of ways: some have circular DNA genomes and thus no ends; others have linear DNA genomes that repeat their terminal sequences or end in loops; while still others have special terminal proteins that serve to prime the DNA polymerase directly (see Figure 6-75).

Viruses Exploit the Intracellular Traffic Machinery of their Host Cells ⁵⁴

All viruses have only a limited amount of nucleic acid in their genome, and so they must parasitize host-cell pathways for most of the steps in their reproduction. In fact, because viral products are usually synthesized in large amounts during infection, and because during its life cycle the virus follows a sequential route through the compartments of the host cell, virus-infected cells have served as important models for tracing the pathways of intracellular transport and for studying how essential biosynthetic reactions are compartmentalized in eucaryotic cells.

Enveloped animal viruses, in which the genome is enclosed in a lipid-bilayer membrane, have exploited the compartmentalization of the cell to an especially fine degree. To follow the life cycle of an enveloped virus is to take a tour through the cell. A well-studied example is Semliki forest virus, which consists of a single-stranded RNA genome surrounded by a capsid formed by a regularly arranged icosahedral (20-faced) shell composed of many copies of one protein (called C protein). The nucleocapsid (genome + capsid) is surrounded by a closely apposed lipid bilayer that contains only three types of polypeptide chains, each encoded by the viral RNA. These envelope proteins form heterotrimers that span the lipid bilayer and interact with the C protein of the nucleocapsid, linking the membrane and nucleocapsid together (Figure 6-77). The glycosylated portions of the envelope proteins are always on the outside of the lipid bilayer, and each trimer forms a "spike" that can be seen in electron micrographs projecting outward from the surface of the virus (Figure 6-77C).

Infection is initiated when an envelope protein on the virus binds to a normal cell protein that serves as its receptor on the host-cell plasma membrane. The virus then uses the cell's normal endocytic pathway to enter the cell by receptor-mediated endocytosis and is delivered to early endosomes (discussed in Chapter 13). But instead of being transferred from endosomes to lysosomes, the virus escapes from the endosome by virtue of the special properties of one of its envelope proteins. At the acidic pH of the endosome, this protein causes the viral envelope to fuse with the endosome membrane, releasing the bare nucleocapsid into the cytosol. The nucleocapsid is "uncoated" in the cytosol, releasing the viral RNA, which is then translated by host-cell ribosomes to produce a virus-encoded RNA polymerase. This in turn makes many copies of viral RNA, some of which serve as mRNA molecules to direct the synthesis of the structural proteins of the virus - the capsid C protein and the three envelope proteins.

The newly synthesized capsid and envelope proteins follow separate pathways through the cytoplasm. The envelope proteins, like the plasma membrane proteins of the host cell, are synthesized by ribosomes that are bound to the rough ER; in contrast, the capsid protein, like the cytosolic proteins of the cell, is synthesized by ribosomes that are not membrane bound. The newly synthesized capsid protein binds to the recently replicated viral RNA to form new nucleocapsids. The envelope proteins, in contrast, are inserted into the membrane of the ER, where they are glycosylated, transported to the Golgi apparatus, and then delivered to the plasma membrane (Figure 6-78).

The viral nucleocapsids and envelope proteins finally meet at the plasma membrane. As a result of a specific interaction with a cluster of envelope proteins, the nucleocapsid forms a bud whose envelope contains the envelope proteins embedded in host-cell lipids. Finally, the bud pinches off and a free virus is released on the outside of the cell. The clustering of envelope proteins as they assemble around the nucleocapsid during viral budding excludes the host plasma membrane proteins from the final virus particle.

Different Enveloped Viruses Bud from Different Cellular Membranes ⁵⁵

Viral envelope proteins are all transmembrane proteins that are synthesized in the ER. Like other ER proteins, they carry sorting signals that direct them to a particular cell membrane (discussed in Chapter 13). Their final location determines the site of viral budding. Epithelial cell lines, for example, can form polarized cell sheets when they are cultured on an appropriate surface, such as a collagen-coated porous filter. When viruses infect such polarized cells, which maintain distinct domains of apical and basolateral plasma membrane, some of them (such as influenza virus) bud exclusively from the apical plasma membrane, whereas others (such as Semliki forest virus and vesicular stomatitis virus) bud only from the basolateral plasma membrane (Figure 6-79). This polarity of budding reflects the presence on the envelope proteins of distinct apical or basolateral sorting signals, which direct the proteins to only one cell-surface domain; the proteins in turn cause the virus to assemble in that domain.

Other viruses have envelope proteins with different kinds of sorting signals. Herpes virus, for example, is a DNA virus that replicates in the nucleus, where its nucleocapsid assembles, and then acquires an envelope by budding through the inner nuclear membrane into the ER lumen; the envelope proteins therefore must be specifically transported from the ER membrane to the inner nuclear membrane, probably via the lipid bilayer that surrounds the nuclear pores. Flavivirus, in contrast, buds directly into the ER lumen, and bunyavirus buds into the Golgi apparatus, indicating that their envelope proteins carry signals for retention in the ER and Golgi membranes, respectively. After budding, the enveloped herpes virus, flavivirus, and bunyavirus particles become soluble in the ER and Golgi lumen, and they move outward toward the cell surface exactly as if they were secreted proteins; in the trans Golgi network they are incorporated into transport vesicles and secreted from the cell by the constitutive secretory pathway (discussed in Chapter 13).

Viral Chromosomes Can Integrate into Host Chromosomes ⁵⁶

The end result of the entry of a viral chromosome into a cell is not always its immediate multiplication to produce large numbers of progeny. Many viruses enter a latent state, in which their genomes are present but inactive in the cell and no progeny are produced. Viral latency was discovered when it was found that exposure to ultraviolet light induced many apparently uninfected bacteria to produce progeny bacteriophages. Subsequent experiments showed that these lysogenic bacteria carry in their chromosomes a dormant but complete viral chromosome. Such integrated viral chromosomes are called proviruses.

Bacteriophages that can integrate their DNA into bacterial chromosomes are known as temperate bacteriophages. The prototypic example is the bacteriophage lambda, discussed earlier. When lambda infects a suitable E. coli host cell, it normally multiplies to produce several hundred progeny particles, which are released when the bacterial cell lyses; this is called a lytic infection. More rarely, the free ends of the linear infecting DNA molecules join to form a DNA circle that becomes integrated into the circular host E. coli chromosome by a site-specific recombination event. The resulting lysogenic bacterium, carrying the proviral lambda chromosome, multiples normally until it is subjected to an environmental insult, such as exposure to ultraviolet light or ionizing radiation. The resulting cell debilitation induces the integrated provirus to leave the host chromosome and begin a normal cycle of viral replication. In this way the integrated provirus need not perish with its damaged host cell but has a chance to escape to other E. coli cells (Figure 6-80).

The Continuous Synthesis of Some Viral Proteins Can Make Cells Cancerous ⁵⁷

Animal cells, like bacteria, can offer viruses an alternative to lytic growth. Permissive cells permit DNA viruses to multiply lytically and kill the cell. Nonpermissive cells may allow the DNA virus to enter but not to replicate lytically; in a small percentage of such cells the viral chromosome either becomes integrated into the host cell genome, where it is replicated along with the host chromosomes, or forms a plasmida circular DNA moleculethat replicates in a controlled fashion without killing the cell. Such nonpermissive infections sometimes result in a genetic change in the host cell, causing it to proliferate in an ill-controlled way and thus transforming it into its cancerous equivalent. In this case the DNA virus is called a DNA tumor virus and the process is called virus-mediated neoplastic transformation. The most extensively studied DNA tumor viruses are two papovaviruses, SV40 and polyoma. Their transforming ability has been traced to several viral proteins that cooperate to stimulate quiescent cells to proliferate - that is, they drive the cells from G₀ into S phase. In permissive cells the shift to S phase (the phase of the cell cycle where DNA is synthesized) provides the virus with all of the host-cell replication enzymes required for viral DNA synthesis. When a provirus happens to make these viral proteins in a nonpermissive cell, they can override some of the normal growth control mechanisms in the cell and its progeny. By this means some DNA tumor viruses that infect humans are known to contribute to the development of some types of human cancers (although the great majority of human cancers are thought not to involve tumor viruses).

RNA Tumor Viruses Are Retroviruses ⁵⁸

For one group of RNA viruses, the so-called RNA tumor viruses, the infection of a permissive cell often leads simultaneously to a nonlethal release of progeny virus from the cell surface by budding and a permanent genetic change in the infected cell that makes it cancerous. How RNA virus infection could lead to a permanent genetic alteration was unclear until the discovery of the enzyme reverse transcriptase, which transcribes the infecting RNA chains of these viruses into complementary DNA molecules that integrate into the host cell genome. RNA tumor viruses - which include the first well-known tumor virus, the Rous sarcoma virus - are members of a large class of viruses known as retroviruses. These viruses are so named because as part of their normal life cycle they reverse the normal process in which DNA is transcribed into RNA.

The enzyme reverse transcriptase is an unusual DNA polymerase that uses either RNA or DNA as a template (Figure 6-81); it is encoded by the retrovirus RNA and is packaged inside each viral capsid during the production of new virus particles. When the single-stranded RNA of the retrovirus enters a cell, the reverse transcriptase brought in with the capsid first makes a DNA copy of the RNA strand to form a DNA-RNA hybrid helix, which is then used by the same enzyme to make a double helix with two DNA strands. The two ends of the linear viral DNA molecule are recognized by a virus-encoded integrase that catalyzes the insertion of the viral DNA into virtually any site on a host-cell chromosome (see Figure 6-70). The next step in the infectious process is transcription of the integrated viral DNA by host-cell RNA polymerase, producing large numbers of viral RNA molecules identical to the original infecting genome. Finally, these RNA molecules are translated to produce the capsid, envelope, and reverse transcriptase proteins that are assembled with the RNA into new enveloped virus particles, which bud from the plasma membrane (Figure 6-82).

Both RNA and DNA tumor viruses transform cells because the permanent presence of the viral DNA in the cell causes the synthesis of new proteins that alter the control of host-cell proliferation. The genes that code for such proteins are called oncogenes. Unlike DNA tumor viruses, whose oncogenes typically encode normal viral proteins essential for viral multiplication, the oncogenes carried by RNA tumor viruses are modified versions of normal host-cell genes that are not required for viral replication. Since only a limited amount of RNA can be packed into the capsid of a retrovirus, the acquired oncogene sequences often replace an essential part of the retroviral genome. In Chapters 15 and 24 we discuss how viral oncogenes have provided important clues to the causes and nature of cancer, as well as to the normal mechanisms that control cell growth and division in multicellular animals. We also discuss how the random integration of viral DNA into genomes can alter normal genes and thereby affect cell behavior (see Figure 24-24).

The Virus That Causes AIDS Is a Retrovirus ⁵⁹

In 1982 physicians first became aware of a new sexually transmitted disease that was associated with an unusual form of cancer (Kaposi's sarcoma) and a variety of unusual infections. Because both of these problems reflect a severe deficiency in the immune system - specifically in helper T lymphocytes - the disease was named acquired immune deficiency syndrome (AIDS). By culturing lymphocytes from patients with an early stage of the disease, a retrovirus was isolated that is now known to be the causative agent of AIDS, which has become a rapidly spreading epidemic that threatens to kill millions of people worldwide.

The retrovirus, called human immunodeficiency virus (HIV), enters helper T lymphocytes by first binding to a functionally important plasma membrane protein called CD4 (discussed in Chapter 23). There are two features of HIV that make it especially deadly. First, it eventually kills the helper T cells that it infects rather than living in symbiosis with them, as do most other retroviruses, and helper T cells are vitally important in defending us against infection. Second, the provirus tends to persist in a latent state in the chromosomes of an infected cell without producing virus until it is activated by an unknown rare event; this ability to hide greatly complicates any attempt to treat the infection with antiviral drugs.

Much current research on AIDS is aimed at understanding the life cycle of HIV. The complete nucleotide sequence of the viral RNA has been determined. This has made it possible to identify and study each of the proteins that it encodes. The three-dimensional structure of its reverse transcriptase (see Figure 6-81) is being used to help design new drugs that inhibit the enzyme. The nine genes of this retrovirus are displayed on the HIV genetic map in Figure 6-83.

Some Transposable Elements Are Close Relatives of Retroviruses ⁶⁰

Because many viruses can move into and out of their host chromosomes, any large genome is likely to contain a number of different proviruses. Most genomes are also likely to house a variety of mobile DNA sequences that do not form viral particles and cannot leave the cell. Such transposable elements range in length from a few hundred to tens of thousands of base pairs, and they are usually present in multiple copies per cell. One can consider these elements as tiny parasites hidden in chromosomes. Each transposable element is occasionally activated to move to another DNA site in the same cell by a process called transposition, catalyzed by its own site-specific recombination enzyme. These integrases, also referred to as transposases, are often encoded in the DNA of the element itself. Since most transposable elements move only very rarely (once in 10⁵ cell generations for many elements in bacteria), it is often difficult to distinguish them from nonmobile parts of the chromosome. It is not known what suddenly triggers their movement.

Transposition can occur by a variety of mechanisms. One large family of transposable elements uses a mechanism that is indistinguishable from part of a retrovirus life cycle. These elements, called retrotransposons, are present in organisms as diverse as yeasts, flies, and mammals. One of the best-understood retrotransposons is the so-called Ty1 element of yeasts. The first step in its transposition is the transcription of the entire transposable element, producing an RNA copy of the element that is more than 5000 nucleotides long. This transcript encodes a reverse transcriptase enzyme that makes a double-stranded DNA copy of the RNA molecule via a RNA/DNA hybrid intermediate, precisely mimicking the early stages of infection by a retrovirus (see Figure 6-82). The analogy continues as the linear DNA molecule uses an integrase to integrate into a randomly selected site on the chromosome. Although the resemblance to a retrovirus is striking, unlike a retrovirus, the Ty1 element does not have a functional protein coat and therefore can only move within a single cell and its progeny.

Other Transposable Elements Transfer Themselves Directly from One Site in the Genome to Another ⁶¹

Unlike retrotransposons, many transposable elements rarely exist free of the host chromosome; the transposases that catalyze their movement can act on the DNA of the element while it is still integrated in the host genome. The transposase binds to a short sequence that is repeated in reverse orientation at each end of the element, thereby holding these two ends close together while catalyzing the subsequent recombination event. The mechanism is closely related to that used by the retrovirus integrase (see Figure 6-70). For some transposable elements the transposition mechanism differs only in that the linear DNA molecule to be integrated must be cut out of a much longer DNA molecule, leaving a break in the vacated chromosome (Figure 6-84). This break is subsequently resealed, but in the process the DNA sequence is often altered, resulting in a mutation at the old chromosomal site.

Other transposable elements replicate when they move. In the best-studied example, a covalent connection is first made between the transposable element and a randomly selected target site; this connection then triggers a localized synthesis of DNA that results in one copy of the replicated transposable element being inserted at a new chromosomal site, while the other copy remains at the old one (Figure 6-85). The mechanism is closely related to the nonreplicative mechanism just described, and it starts in nearly the same way; indeed, some transposable elements can move by either pathway.

In addition to moving themselves, all types of transposable elements occasionally move or rearrange neighboring DNA sequences of the host genome. They frequently cause deletions of adjacent nucleotide sequences, for example, or carry them to another site. The presence of transposable elements makes the arrangement of the DNA sequences in chromosomes much less stable than previously thought, and it is likely that they have been responsible for many important evolutionary changes in genomes (discussed in Chapter 8).

Are the transposable elements also of evolutionary importance as the most ancient ancestors of viruses? Although the precursors of retroviruses were almost certainly retrotransposons, all present-day transposable elements rely heavily on DNA-based reaction mechanisms. But very early cells are thought to have had RNA rather than DNA genomes, so we must look to RNA-based mechanisms for the ultimate origin of viruses.

Most Viruses Probably Evolved from Plasmids ⁶²

Even the largest viruses depend heavily on their host cells for biosynthesis; no known virus makes its own ribosomes or generates the ATP it requires, for example. Clearly, therefore, cells must have evolved before viruses. The precursors of the first viruses were probably small nucleic acid fragments that developed the ability to multiply independently of the chromosomes of their host cells. Such independently replicating elements, called plasmids, can replicate indefinitely outside the host chromosome. Plasmids occur in both DNA and RNA forms, and, like viruses, they contain a special nucleotide sequence that serves as an origin of replication. Unlike viruses, however, they cannot make a protein coat and therefore cannot move from cell to cell in this way.

The first RNA plasmids may have resembled the viroids found in some plant cells. These small RNA circles, only 300 to 400 nucleotides long, are replicated despite the fact that they do not code for any protein. Having no protein coat, viroids exist as naked RNA molecules and pass from plant to plant only when the surfaces of both donor and recipient cells are damaged so that there is no membrane barrier for the viroid to pass. Under the pressure of natural selection, such independently replicating elements could be expected to acquire nucleotide sequences from the host cell that would facilitate their own multiplication, including sequences that code for proteins. Some present-day plasmids are indeed quite complex, encoding proteins and RNA molecules that regulate their replication, as well as proteins that control their partitioning into daughter cells. The largest known plasmids are double-stranded DNA circles more than 100,000 base pairs long.

The first virus probably appeared when an RNA plasmid acquired a gene coding for a capsid protein. But a capsid can enclose only a limited amount of nucleic acid; therefore a virus is limited in the number of genes it can contain. Forced to make optimal use of their limited genomes, some small viruses evolved overlapping genes, in which part of the nucleotide sequence encoding one protein is used (in the same or a different reading frame) to encode a second protein. Other viruses evolved larger capsids and consequently could accommodate more genes.

With their unique ability to transfer nucleic acid sequences across species barriers, viruses have almost certainly played an important part in the evolution of the organisms they infect. Many recombine frequently with their host-cell genome and with one another. In this way they can pick up small pieces of host chromosome at random and carry them to different cells or organisms. Moreover, integrated copies of viral DNA (proviruses) have become a normal part of the genome of most organisms. Examples of such proviruses include the lambda family of bacteriophages and the so-called endogenous retroviruses found in numerous copies in vertebrate genomes. The integrated viral DNA can become altered so that it cannot produce a complete virus but can still encode proteins, some of which may be useful to the host cell. Therefore, viruses, like sexual reproduction, can speed up evolution by promoting the mixing of gene pools.

The process in which DNA sequences are transferred between different host-cell genomes by means of a virus is called DNA transduction, and several viruses that transduce DNA with particularly high frequencies are commonly used by researchers to move genes from one cell to another. Viruses and their close relatives - plasmids and transposable elements - have also been important to cell biology in many other ways. Because of their relative simplicity, for example, studies of their reproduction have progressed unusually rapidly and have illuminated many of the basic genetic mechanisms in cells. In addition, both viruses and plasmids have been crucial elements in the development of the recombinant DNA technologies that will be described in Chapter 7.

Summary

Viruses are infectious particles that consist of a DNA or an RNA molecule (the viral genome) packaged in a protein capsid, which in the enveloped viruses is surrounded by a lipid-bilayer-based membrane. Both the structure of the viral genome and its mode of replication vary widely among viruses. A virus can multiply only inside a host cell, whose genetic mechanisms it subverts for its own reproduction. A common outcome of a viral infection is the lysis of the infected cell and release of infectious viral particles. In some cases, however, the viral chromosome instead integrates into a host-cell chromosome, where it is replicated as a provirus along with the host genome. Many viruses are thought to have evolved from plasmids, which are self-replicating DNA or RNA molecules that lack the ability to wrap themselves in a protein coat.

Transposable elements are DNA sequences that differ from viruses in being able to multiply only in their host cell and its progeny; like plasmids, they cannot exist stably outside of cells. Unlike plasmids, they normally replicate only as an integral part of a chromosome. Some transposable elements, however, are closely related to retroviruses and can move from place to place in the genome by the reverse transcription of an RNA intermediate. Although both viruses and transposable elements can be viewed as parasites, many of the DNA sequence rearrangements they cause are important for the evolution of cells and organisms.