To access the private area of this site, please log in.

Username: 

Password:  

The Legacy of 20th Century Phage Research

ALLAN M. CAMPBELL

Posted 8 October 2010

Department of Biology, Stanford University, Stanford CA 94305

Phone: (650) 723–1170, Fax: (650) 725–1841, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

Bacteriophages were discovered during World War I (6, 27). In the 20 years after their discovery, they were studied with two principal aims: (i) as diagnostic tools for typing bacterial strains and (ii) as possible therapeutic agents against bacterial infection.

Starting about 1940, Max Delbrück and collaborators ushered in a "Golden Age" of phage research, during which the utility of phages in investigations of fundamental biological questions was exploited. That activity still exists, but since around 1970, many of the problems have become increasingly tractable in other systems.

In the late 20th century, research on phage was increasingly concentrated in certain areas, frequently with a view to diverse practical applications: the revival of phage therapy when antibiotics were no longer sufficient, the role of phage in bacterial pathogenicity, and phage in the biosphere. Progress and prospects in those areas are summarized by Campbell (3). I concentrate here on the period from 1940 to 1970, attempting to identify lines of research that have proven to be of lasting value.

The key paper that initiated the “Golden Age of Phage” (9) reported no new discoveries but introduced a new approach. It has sometimes been described as the beginning of “quantitative” phage research, but that is hardly a precise description. The phage literature of the time was filled with numerical results: times of lysis, phage titers under specific conditions, etc. What was new was the focus on those parameters most directly relevant to phage reproduction. It started with the fact that individual cells infected with phage soon thereafter lyse and liberate a crop of progeny phage. This was hardly a new concept, but the implications for deciding what was important to measure had been neglected.

The relevant numbers were the phage input per cell (multiplicity of infection), the time to lysis, and the number of progeny phage released at lysis. To synchronize the time of infection in all cells of a culture, rapid attachment was necessary, so Delbrück's early disciples dedicated much effort to determining the kinetics of attachment. Delbrück gathered converts to his New Virology by offering a summer course at Cold Spring Harbor, designed to attract the best minds among biologists and especially other physicists-turned-biologists like himself. It was decided to restrict their studies to seven coliphage isolates (types 1 through 7, or T1 through T7) and to measure the basic growth parameters in all of them to create a standardized set with which to try to understand what went on within the infected cells.

One important factor incorporated into the analysis was the variation in phage input from cell to cell. It is obvious that if a population of cells are infected with an average number of three phage particles per cell, some cells will receive exactly three, while some will receive two or five or one or none at all. Delbrück employed the precise predictions for random attachment to a uniform population of cells. The fact that different cells received different inputs was vaguely realized by previous workers, but many of them treated the system as though it were homogeneous and the variations were some kind of imperfection that should be glossed over. (And not all of them knew how to analyze the system with the aid of the Poisson distribution anyway.)

The methods used were largely microbiological and had been available for decades. The Delbrück school steered away from biochemistry of the type generally practiced at the time. It was not obvious which biochemical reactions were relevant. They preferred to quantitate phage by plaque counts. A plaque is undeniably the result of infection by a single viable phage particle, as verified by the linearity of plaque counts with input concentration (a point that had been emphasized by d’Herelle almost 20 years earlier).

The primary goal was to understand the mechanism of self-replication. Once the time interval in which replication occurs was delimited to a latent period of defined duration, it was hoped that intracellular growth kinetics could be determined by disrupting cells at various times and seeing how much phage was present. This approach, like much of early phage work, did not provide an answer but rather led to the discovery of various unexpected results of which perhaps the most important was Doermann's (7) finding that phage went through an “eclipse period," during which, from a cell destined to liberate phage if left undisturbed, not even the input phage could be recovered. Meanwhile, attempts to follow the intracellular growth of one phage by infection with and lysis by another led to the discovery of exclusion and of genetic recombination in phage.

The Ellis and Delbrück paper (and subsequent extensions, such as single-burst analysis) set the stage for many sophisticated investigations of phage development, of which relatively few (discussed below) left a legacy, recounted here. But the initial work remains of lasting value, and its importance should not be neglected or misunderstood, as it frequently is. For example, current research focuses on many phages with interesting properties (such as those infecting archaea), which have not been shown to form plaques or to generate linear dose-response curves. Much effort has been expended to demonstrate (frequently successfully) that these phages follow a growth pattern similar to that of T4 or λ. However, some authors have assumed (incorrectly, I believe) that such phages therefore constitute the same kind of sharp analytical tools for answering fundamental questions, as the classical coliphages did. The basic reason the classical phages were so useful in this respect is closely related to the ease with which the contents of a single plaque can be unambiguously interpreted as the descendants of a single-particle infection.

I pass now to more specific cases in which phage work made major contributions.

 

DNA AS GENETIC MATERIAL

The seminal experiment indicating that the genetic material of phage T2 is DNA (and DNA alone) was performed by Hershey and Chase in 1952 (11). As usual, a new finding can be best appreciated in the context of what was already known at the time.

By 1952, phages of the T series had been visualized by electron microscopy, although at a level that is crude by today's standards. It was known that their major components were DNA and protein, that most of the DNA was localized in the head, that phage particles attached to the cell by the tips of their tails, and that a few minutes after infection, the heads of those particles attached to cells were empty. It was the last observation that, according to Hershey, “practically forced” him to look chemically at whether the DNA had been injected into cells and (critically) whether removal of the empty head would curtail phage development. The actual experiment was extremely simple. Isotopically labeled phage particles were prepared by infecting bacteria in media containing radioactive phosphorus (to label DNA) or sulfur (to label protein). These labeled phage were then mixed with bacteria under conditions where attachment is rapid. A few minutes after attachment (when electron microscopy would have shown empty heads attached to tails) the infected cells were subjected to the shearing forces exerted by a Waring blender borrowed from the kitchen of a colleague. These stripped cells were then allowed to lyse, and the progeny phage were assayed for radioactivity; they proved to contain very little of the sulfur from the infecting particles, but most of the phosphorus.

The impact of the Hershey-Chase experiment of course went far beyond the community of phage workers. Most biologists had imagined that genes were composed of both protein and nucleic acid (sometimes called “nucleoprotein”) and that the protein was the informational part. The readiness of the larger world of biochemists and geneticists to accept the Hershey-Chase experiment as compelling evidence for the informational role of nucleic acid depended on historical factors.

The degree of chemical resolution was not very high. Most of the protein could be removed by blending, but some remained cell associated. This could of course be explained away as resulting from where the shear forces had cleaved the structure, but it might instead have meant that some proteins essential to successful infection needed to enter the cell along with the DNA (as is in fact the case for some viruses). The other major evidence that genes were made of DNA came from bacterial transformation, where the chemical purity of the DNA was much higher.

Volumes have been written from personal recollections about this particular turning point. One reason why some people found the phage experiment more convincing was that, by 1952, it was already clear (as I discuss shortly hereafter) that the phage had a genome rich in information and it seemed unlikely that all of that information could be carried by a small fraction of the protein.

Bacterial transformation, on the other hand, was just beginning to be credible as a general model for the nature of genes. For years its study had largely been confined to capsular polysaccharides in the pneumococcus, and although a few investigators (including Avery) were certainly aware that it might have much broader implications, it was only around 1952 that substantial evidence on transformability of other traits and in other bacteria accumulated. Second, the efficiency of transformation was sufficiently low that it was hard to guarantee that the specificity did not reside in a minor component, whereas the efficiency of phage infection was close to 100%.

At any rate, both kinds of results set the stage for an appreciation of the importance of DNA structure, which soon became available.

 

PHAGE GENETICS

 

Inception

Classical genetics was defined by mutations (or natural variation) and recombination (or, in Mendel's original paper, reassortment of chromosomes). Two types of phage mutants appeared early on: those extending host range (h) and those affecting plaque morphology. In the latter category, rapid-lysis mutants (r) of the T-even phages were conspicuous. Escherichia coli B cells infected with r mutants lyse more rapidly than those infected with wild-type (r+) phages. In mixed infection, r+ is dominant over r. The h mutants were isolated on mutant bacterial strains that were resistant to phage infection—a type of bacterial mutant on which Luria and Delbrück's demonstration of the spontaneous origin of bacterial variants was based (16).

One focus of phage work in the 1940s was the intracellular replication of phage. As mentioned earlier, artificial lysis of cells at various times after infection revealed an eclipse phase (an observation readily explained by the Hershey-Chase result; if only the DNA gets in, there is necessarily a stage before complete, infectious particles are present). The eclipse phase was demonstrated by lysing cells with chloroform. Interpretation of the results is complicated by the fact that chloroform lysis depends on synthesis of lysozyme encoded by the infecting phage. Chloroform damages the membrane and allows lysozyme to traverse the inner membrane layer and reach its peptidoglycan substrate (7). Chloroform damage collapses the membrane potential and prevents further macromolecular synthesis; consequently, any phage particles released after lysis must have been made before the time of chloroform action (29). Efforts to probe the course of infection were also made by superinfecting at various times with a phage other than the one under study. Those experiments turned out to be uninformative on the question they were designed to answer but instead disclosed unexpected phenomenology. (I remember Hershey once remarking that phage workers could rarely do experiments without being bothered by discoveries.)

When an unrelated phage was used to superinfect (e.g., T1 superinfection of T2-infected cells), mutual exclusion was discovered: individual cells liberated either T1 or T2, never both. When the superinfecting phage was a mutant of the original phage, it frequently contributed to the yield, and the proportion of the superinfecting type among the progeny could be used to infer the size of a “replication pool” of phage genomes at that time (28). The results were sometimes clouded by various exclusion mechanisms, which were absent or unimportant in certain phages (like λ). Such experiments led to a second discovery. If the superinfecting phage and the original infecting phage differed by two mutations (for example, if the superinfecting phage was an r mutant and the original phage was an h mutant), then the phage yield contained not only the parental types (r and h) but also recombinants between them (wild type and rh double mutants). Further study of recombination was usually done by simultaneous infection rather than by reinfection.

At the time recombination was discovered in phage, the process was already well known in eukaryotes. In crosses between parents carrying different mutations on the same chromosome, recombination results from exchanges between homologs at the first meiotic division. The further apart the mutations are on the chromosome, the greater the recombination frequency. Thus, genes could be placed in order along the chromosome. The near additivity of recombination frequencies at short distances was enhanced by the fact that recombination in adjacent intervals on the same individual was rarer than expected from random coincidence (a feature called interference). Even absent any relevant information on chromosomes, linkage maps could be constructed based on recombination frequencies alone.

In phage, meiosis was unknown, and chemistry that permitted equating the DNA molecule to the eukaryote chromosome was yet to be done. But linkage analysis proved feasible. In the construction of linkage maps it was assumed that the mutations used to detect recombinational events did not themselves affect the recombination process—an assumption that is approximately true in many cases but  might not have held for very small molecular distances. But the first thorough study of T2 recombination (12) did not disclose any glaring exceptions. For example, the recombination frequency observed in the cross of double mutant by wild type (rh × +) was about equal to that seen when single mutants were crossed (r+ × h+). (Classical geneticists had called these two configurations coupling and repulsion, respectively.) Thus, it was meaningful to speak of a recombination frequency between the r locus and the h locus. It was also found that independently isolated mutations of identical phenotype frequently mapped to different sites. So a recombination frequency between two mutant sites (r1 and r2) could be measured, and again the frequency of wild-type progeny from (r1+ × r2+) equaled the frequency of single-mutant progeny from (+ × r1 r2). The observed mutations fell into three linkage groups. Mutations from different linkage groups showed about 40% recombination (the maximum observed); mutations within a linkage group could be arranged in a linear order based on recombination frequencies; but the deviations from additivity at short distances were much more pronounced than in eukaryotes (a result formally known as subadditivity). Consistent with this subadditivity, comparison of three-factor crosses with two-factor crosses showed that double-exchange types were much more common than predicted from coincidence of single exchanges (a phenomenon dubbed “negative interference”).

Thus, before 1950, all the elements of a formal genetic analysis of the T-even phages were available. While most investigators were eager to translate linkage data into a picture of DNA structures, it was another 30 years before DNA sequencing became routinely available. In the meantime, the resolving power of fine-structure genetics (see below) greatly exceeded that of any operations performable on DNA itself. Much of the analysis of linkage relationships determined in the interim constitutes refinements, extensions, or rediscoveries of Hershey and Rotman's original findings, i.e., that T2 linkage maps resembled those of Drosophila or maize with two special features: the subadditivity of small recombination frequencies and the observed frequency of 40% recombination (rather than 50%) among unlinked mutations. It may also be noted that, although phage geneticists were fully aware of the similarities to eukaryotic genetics, nowhere did they introduce assumptions derived from eukaryotic genetics. Like most of the best phage experiments, phage genetics was self contained and not derived from any other field.

 

Fine-Structure Genetics

Extension of the Hershey/Rotman results went either upward (to the recombination pattern of whole genomes) or downward (to mapping of small regions of the genome). In the latter category, Seymour Benzer initiated a program to "run the map into the ground”; i.e., to create a genetic map of mutations very close to one another on the DNA, perhaps with sufficient resolution to include recombination frequencies between adjacent nucleotides.

The experimental system he chose was the rII mutants of phage T4. It had been observed incidentally that these mutants (i.e., the r mutants in one of Hershey's three linkage groups) not only made morphologically distinct plaques on the standard host, E. coli B but also failed to plate at all on another strain, E. coli K (λ). This meant that mutants could easily be isolated (even from an unmutagenized strain) based on plaque morphology on B and then tested for recombination and/or complementation on K. As Hershey and Rotman had shown, the recombination frequencies are high (several percentage points across the whole region), and given the ability to select wild-type recombinants on K, the only predicted lower limit to measuring recombination frequencies was set by the reversion rates.

Benzer began working on the T4 rII system in 1953 and stopped about 10 years later. He showed that small mutations could happen at hundreds of sites distinguishable by recombination. From crosses among deletion mutants, a linear linkage map could be inferred as expected if they were physical deletions from a linear DNA structure. Not all sites were equally mutable; some hot spots were represented many times, whereas others had only been hit once (implying that the screen was not yet saturated) (1). Mutagens acted preferentially at sites specific to the mutagen used. All these results were later understood in terms of actual chemical pathways of mutagenesis.

One of the most famous applications of mutagen specificity employed acridine dyes (which can cause deletions or additions of one to a few nucleotides from slippage during replication) and provided strong evidence for a three-letter code (5).

The physiological reason that rII mutants fail to plaque on K(λ) is not completely understood, even today. Some geneticists regard this as fortunate. Had the mechanism been clearly understood at the outset, more attention might have been paid to the physiology and less to the formal genetic results.

Considerable effort was expended on the use of fine-structure genetics in recombinational mapping (remembering always that this was during the many years before the DNA sequences were known). As indicated earlier, one basic assumption was that the number of exchanges in an interval depends only on the action of recombination on that stretch of DNA, independently of the mutations used to mark that DNA so that recombinations in defined intervals could be identified. Under that assumption an excess of double exchanges above random coincidence (coefficient of coincidence greater than one, or negative interference) implies a clustering of exchanges, whereas a coefficient of coincidence less than one (well known for eukaryotic chromosomes over much larger molecular distances) implies a distribution of exchanges more even than random. Both effects might occur on the same DNA, if clusters were evenly distributed. A coefficient of coincidence greater than one (and the associated subadditivity of recombination frequencies) had already been observed by Hershey and Rotman and seemed to become even more extreme at short distances (4).

Hopes of distilling such results into a mapping function that related recombination frequency to distance gradually evaporated as it became clear that the basic assumption was wrong. Today almost everyone agrees that most recombination at short distances is provoked by mismatches within the heteroduplex regions associated with genetic exchange. Phage experiments played a crucial role in the study of such heteroduplexes.

Many investigators were unwilling to attach strong credibility to ordering of genetic sites from three factor crosses when the double-crossover class and the rarer single-crossover class differed by less than a factor of 2 (which they frequently did, because of negative interference). With rII and other systems, the problem could be finessed if deletions were found that dissected the map between the markers of interest.

 

Linkage Analysis in Whole Genomes

T4 geneticists of the time also wanted to map the entire genome, not just the relatively small segment around rII. Hershey and Rotman had reported three linkage groups, but it was uncertain whether the genome comprised three separate molecules or a single molecule with substantial genetic distance between the three groups. Genetic markers at more locations were needed.

The desired markers came in the form of conditionally lethal mutations (10). These were of two main types: (i) suppressible by nonsense suppressor tRNAs, which were present only in certain laboratory strains of E. coli, and (ii) thermosensitive, which failed to plate at high temperature. The work was initiated by the chance discovery of suppressible mutants in T4 and λ, but the system was soon put to good use.

Before turning to the uses, it may be worth mentioning how much Benzer's rII work influenced the manner in which the conditional lethals were handled. Benzer's analysis was ingrained in the consciousness of younger phage workers. The rII mutants are a kind of conditional lethal, the lethality resulting from loss of function in two specific, closely linked genes. Suppressible and thermosensitive mutants instead could be isolated in any gene whose product was needed for plaque formation. They are located in dozens of genes distributed over the phage map. But the basic operational protocol for classifying the mutants by recombination and complementation was the same. The concepts are not all that complicated, of course; but Benzer's example made the time ripe for development of the system.

The conditional lethals were used for two purposes; first, simply as selectable markers with which the fate of the genome could conveniently be followed through recombination, deletion, and rearrangement; and second, as tools for exploring phage gene function. Their use as markers made it possible to map the whole T4 genome into one linkage group, connecting the three groups Hershey had found. The map turned out to be circular rather than linear.

As with fine-structure mapping, extension of this result to a unified theory ultimately proved to be largely unproductive. Initial theorizing postulated a mating pool of genomes engaged in successive rounds of random pairwise mating during infection (27). The 40% maximum of recombination frequencies was then assumed to mean that some genomes had, by chance, not mated with a heterologous partner. Whereas some of the parameters prescribed by the experimental results could be formalized independently of this assumption, it ultimately became necessary to rationalize them with the actual state of the intracellular DNA. This turned out to be quite different among phages; and generally, phage DNA did not replicate in monomer form, as the theory postulated. Instead, in both λ and T4, DNA is packaged from linear concatemers of multigenomic length. In λ packaging, DNA is cut at specific sites (cos), generating a linear map, whereas T4 and P22 DNA is cut into “headfuls,” where each progeny particle has slightly more than one genome's length of DNA, so that successive particles have different molecular endpoints, giving a circular map. Furthermore, intracellular T4 DNA is not just linear concatemers. Late replication is initiated by strand invasion, which generates a highly branched tangle of DNA that needs resolution of Holliday junctions to eliminate the branches and create a packageable substrate (20).

 

GENE FUNCTION AND REGULATORY PROGRAMS

The other use of conditional lethals (in determining gene function) proved to be an enormous success and underlies similar successes in other organisms. The first comprehensive exposition was given with T4 (9). Beyond simply matching proteins or phenotypes with genes came the determination of pathways. Phage assembly was the first posttranscriptional macromolecular pathway to be worked out; initially, it was achieved by the judicious use of single or multiple mutants blocked in specific steps (8). Some general properties emerged: (i) phage proteins assemble in defined order; (ii) assembly pathways are branched rather than linear, so that tails and heads, for instance, are assembled separately and then joined later; empty protein shells are finally assembled and later filled with DNA to give heads. Similar results were reported for λ and other phages.

Identification of phage functions was augmented by a body of work (initiated especially by Seymour Cohen), which, although undertaken because the Delbrück school had brought phages into such prominence, was derived from classical biochemistry and enzymology. Starting with the discovery that, in the DNA of phage T4, cytosine is completely replaced by hydroxymethylcytosine, a number of enzymes from T4-infected cells were found to carry out this transition, as well as other steps in the synthesis of phage DNA and its precursors. (See reference 19  for a summary.) At the time, very few viruses of any sort were known to work as enzymes. Within a few years, it was possible to identify most of these enzymes as products of phage genes known to be essential for plaque formation.

Regulatory pathways were likewise worked out, giving for each phage a genetic hierarchy wherein the transcription of certain genes is contingent on prior expression of other genes, so that genes could be classified not only by time of expression but also into causal chains. This was perhaps best exemplified by λ, because genes whose products interact are generally clustered. All the genes of λ are transcribed by the host RNA polymerase with its housekeeping sigma factor σ70; temporal control is exerted through λ-encoded repressors and antiterminators. In T4, on the other hand, σ70 is replaced at late times by a phage-coded σ; and in T7, a new core polymerase specific for T7 promoters replaces the host polymerase. The precise biochemistry of some of these effects is still being worked out, but their discovery and documentation were greatly facilitated by the technical advantages of the phage systems.

Some additional perspective was provided by the study of small, single-strand phages whose genomes were either DNA (23) or RNA (15). The DNA phages were the first natural substrates for host DNA polymerases where in vitro synthesis of a biologically active product was achieved. Except for initiation, these phages depend entirely on host genes for DNA synthesis; whereas some larger phages, like T4 and T7, encode a full complement of their own DNA genes. And the RNA phages were studied intensively in the 1960s by biochemists, for some of whom this was a kind of “last hurrah” in their use of prokaryotes before switching to molecular studies of eukaryotes.

The RNA phage work demonstrated that a phage could execute a temporal program of gene regulation solely by translational control (whose full extent had not been well appreciated until then). These phages have only three or four genes but manage to make their structural proteins (coat protein and assembly protein) in about the required 180:1 ratio and also to make polymerase early and turn it off late. All this is achieved through the action of secondary structure in sequestering the initiation sites for translation of assembly protein and polymerase; the ability of translation of the coat protein to uncover the initiation site of the polymerase gene; and the inhibition of polymerase initiation by accumulated coat protein.

 

LYSOGENY

It was noted already in the 1920s that many bacterial strains harbor phage, which is found in small amounts in the culture medium. Such strains were called lysogenic. A sharp definition at the individual cell level was first achieved by Lwoff and Gutmann (17), who grew a lysogenic culture of Bacillus megaterium, separated individual cells into liquid droplets with a micromanipulator, and plated them out to form colonies. Each such colony could generate a culture that also liberated phage. The liquid in droplets from which cells had been separated was assayed and found to contain no phage; however, an occasional cell lysed under the microscope, and the liquid in the droplet then contained many phage particles. So the phage genome is reproduced intracellularly, and phage liberation occurs sporadically when an occasional cell lyses spontaneously. Lwoff's group followed that work with the discovery that lysogenic cultures could be induced to undergo mass lysis, with phage liberation, by various chemical and physical agents such as UV light (18). Thus, every cell is capable of producing phage, and that potentiality is passed to both daughters at cell division.

In 1951, Esther Lederberg discovered that the common laboratory strain E. coli K-12 is lysogenic (14). The phage it carried was given the name “λ.” Because lysogeny is a heritable trait of bacterial strains, and K-12 was the only strain with a well-developed genetic system at that time, λ became the object of choice for studying how the prophage was maintained without destroying the cell and how it was partitioned at cell division. Some aspects of the problem, especially those related to gene regulation, were so appealing that by 1970 λ had displaced T4 as the prototypical phage (a status T4 had gained from the work of Hershey, Benzer, and the Caltech group).

It was soon shown that λ prophage could be mapped to a particular position on the bacterial chromosome. While the simplest interpretation was that the prophage was physically inserted into the chromosome, various observations persuaded some leading investigators that it was instead synapsed against it. Hard evidence for insertion based on genetic mapping became available only in the 1960s, and DNA sequencing of the lysogenic chromosome had to wait another decade. An important part of the genetic evidence (and the material later used in the first sequencing) came from the characterization of specialized transducing phages, which result from imprecise excision from the bacterial chromosome.

The aspect of lysogeny that attracted the most attention was how genes of the prophage were turned off. The seminal genetic work of Kaiser (13) showed that phage genes were repressed by a prophage-coded repressor, which turned off the two major promoters whose expression initiated lytic infection. Later, the repressor protein was isolated, and in vivo and in vitro studies complemented each other in demonstrating how a bistable molecular switch can work (22). The actual role of this switch in the decision to lysogenize remains under study even at present.

Most investigators have assumed that the direction a given cell takes depends stochastically on the random statistics of synthesis of small molecules early in infection. However, there is evidence that the size of the cell at the time of infection (and therefore its stage in the cell division cycle) can play a major role (25). Such studies have attracted attention because development of multicellular organisms requires some steps at which a cell becomes committed to a specific fate.

 

RESTRICTION/MODIFICATION

In 1953, Bertani and Weigle (2) reported that the host range of certain phages could be restricted by the host on which they had just been grown. For example, λ grown on E. coli B plated with very low efficiency on E. coli K and vice versa, although both plated well on E. coli C. This was not mutation and selection, because phage grown on K lost their ability to plate on K in a single passage on E. coli C. Furthermore, when phage from K infected C, those progeny phage from the first cycle that contained parental DNA in one strand still plated on K. So it seemed that the DNA of phage grown on K is somehow modified to allow plating on K. Restricting strains thus had not only a restriction system that destroyed foreign DNA but also a modifying mechanism that protected endogenous DNA from degradation. Subsequent studies, especially those of Werner Arber, led to a better understanding of the mechanism of restriction and modification and their biological significance. It soon became apparent that restriction enzymes cleaved DNA into large pieces, which could then be cleaved further by exonucleases such as ExoV.

 

Biochemical Mechanisms

Restriction endonucleases have been classified into several groups, exemplified by Type I and Type II. Type II endonucleases are familiar to most beginning students in molecular biology. They recognize specific DNA sequences (usually 4- to 8-bp palindromes) and cleave both strands within the recognition site (24). They have become standard reagents for recombination and cloning. Type I enzymes are more complicated. Their recognition sites are asymmetric and serve only as entry sites. Cutting occurs elsewhere on the DNA. Enzyme bound to the recognition site causes ATP-dependent movement of DNA toward the site, and cutting seems to take place where two bound molecules meet (26).

In both cases, modification is carried out by another polypeptide (with no homology to the restriction enzyme), which alters the DNA within the recognition site (usually by methylation) in both strands so that cutting no longer occurs. With Type II enzymes, the DNA is methylated in both strands at equivalent positions in the palindrome; with Type I enzymes, only one strand of the recognition site is methylated.

In vivo, modification protects endogenous DNA from destruction, but only very rarely are DNA molecules entering the cell from outside (as in phage infection) protected. The resistance of hemimethylated DNA to cutting by Type II enzymes provides a sufficient explanation for how endogenous DNA and exogenous DNA are distinguished: in semiconservative replication, each new strand is base paired to a methylated old strand, so the new strand has a whole division cycle to become methylated before it can be cut. This mechanism does not work for Type I enzymes, where only one strand of the recognition site is methylated; it seems that for Type I, endogenous DNA is protected by its cellular location.

Another important distinction is that in Type II, the restriction and modification enzymes independently recognize the same DNA site. However, Type I systems employ three peptides: a restriction subunit, a modification subunit, and a recognition subunit, which interacts with the catalytic subunits as well as with the DNA entry site. In both cases, the genes (two for Type II and three for Type I) are closely linked.

 

Biological Functions and Evolution

The natural function of restriction/modification seems to be to protect cells from invasion by foreign DNA, especially phage DNA. To be effective, restriction specificities must be highly polymorphic among the natural hosts of the phage—in particular, within a bacterial species. For if all E. coli strains had the same specificity, then any phage grown on E. coli would be immune to E. coli restriction. The evolution of Type I recognition has intriguing aspects: the recognition sites are bipartite, with 5’ and 3’ oligonucleotide sequences separated by an oligonucleotide spacer. Existing Type I systems seem to mix and match 5’ and 3’ oligonucleotides, as though natural sites were generated by some sort of homology-independent recombination. Restriction/modification constitutes a kind of addiction system best known for plasmids; if the plasmid is lost, the cell can be killed by the restriction enzyme once the modification enzyme is diluted or decays (21). Many plasmids elaborate toxins and antitoxins that behave in a similar manner. Superficially, such systems appear to stabilize the plasmid, because any cell that loses the plasmid is eventually killed by the toxin; however, this is largely an artifact of laboratory studies that start with pure cultures of plasmid-bearing cells. In the wild, where plasmid-bearing cells compete with plasmid-free cells as well as cells bearing other similar plasmids, a more plausible role is that a plasmid with such a system resists replacement by a natural plasmid that might invade the cell.

Type II systems are commonly encoded by phages or plasmids, whereas Type I systems are frequently encoded by chromosomes. For example, the E. coli K determinants (restriction, modification, and specificity) are adjacent on the chromosomes; furthermore, the B determinants (though not homologous to those of K) occupy the same location, so that the two specificities behave as alleles. The mechanisms that create and maintain such arrangements are barely understood.

So, restriction/modification systems continue to pose challenging problems, many of them closely related to phage biology.

 

REPRISE

I have tried here to identify important developments from the “Golden Age of Phage” that form an enduring legacy. I have addressed the initial work of the Delbrück school in special detail because, while it did not immediately result in important discoveries, it enabled the work that followed and remains essential to the design of current experiments. Current analysis of some of the classical phage systems, while more sophisticated and critical than ever, transcends the scope of this chapter.