<<  Chapter 8   —   Appendix A  >>

Chapter 9

Summary and Conclusions

Chapter doi:10.1101/restrictionenzymes_Summ-Conc

This book started with experiments in the early 1950s on a barrier to phage infection and “host-controlled variation,” which led to the discovery of DNA restriction and modification. Importantly, this modification was reversible and did not lead to mutations. This would herald the end of the distinction between genotype and phenotype and is reminiscent of the current nature versus nurture debate. The next breakthrough came in the shape of the Escherichia coli (EcoKI and EcoBI) restriction enzymes, which required S-adenosylmethionine (SAM) and ATP for restriction, followed by the discovery of the first enzymes (HindII and EcoRI) that did not require these cofactors. When EcoRI was found to produce staggered DNA ends, which could “restick” in vivo and/or in vitro, genetic engineering was born. Further studies resulted into the division in three types of enzymes: E. coli–like (Type I), EcoRI-like (Type II), and phage P1–like (Type III). The 1970s saw a revolution in recombinant DNA technology, while Alec Jeffreys started his analysis of eukaryotic DNA repeats, which would result in the development of the invaluable DNA finger printing technique allowing the solution of paternity cases, the identification of criminals and their victims, and the exoneration of the falsely accused. However, it was also the decade of cloning trouble due to modification-dependent (later named Type IV) restriction enzymes in E. coli that destroy cloned methylated DNA from other organisms.

The list of achievements of the 1980s includes DNA sequences of restriction genes (e.g., EcoRI and EcoKI) and many new restriction enzymes derived from different strains. For extensive biochemical and structural analysis, EcoRI and EcoRV became the enzymes of choice, because large amounts (“bathtubs” full) had to be produced for this work. Plasmids and phages such as lambda with mutant restriction enzyme sites helped to reveal the incredible specificity and fidelity of the enzymes for their DNA recognition site. Regular updates by Rich Roberts from 1976 onward of lists of enzymes and their properties eventually led to the REBASE website in the early 1990s. By 1993, nearly 1000 Type II enzymes with 200 specificities had been identified in many bacterial species, which would lead to a subdivision into 11 subtypes in 2003. In 2004 Alfred Pingoud edited a specialized book with 16 chapters dedicated to Type II restriction enzymes (including the first crystal structures), with a single chapter on the ATP-dependent Type I and III “molecular motors” of the SF2 superfamily. Mutant enzymes with longer recognition sequences were high on the wish list with the goal of preparing tools for gene therapy. This proved very difficult, although FokI looked promising with its separate recognition and catalytic cleavage domains, allowing the construction of chimeric fusion proteins. Different constructs of Type II catalytic domains with zinc fingers and improvements using TALE proteins met with varying success, and the off-target activity remained a worry. The new RNA-based method of genome editing using the CRISPR–Cas9 system may present a better alternative but also has its drawbacks, as reported in recent publications. The diverse studies described in this book provide clear evidence for tight control of potential genome alterations, and bacterial populations are not alone in this. This control allows a certain carefully contained level of changes in order to generate heterogeneity within populations, and large eukaryotic genomes have developed similar mechanisms to stabilize the genome. Taken together it is the proper balance between what is best for an individual cell versus what benefits the population and/or organism as a whole. Clearly, microorganisms in bacterial populations, and cells in tissues and individuals, do not allow their DNA to be so easily manipulated as genome engineers wish, and gene targeting will require further investigations and refinements before being of true benefit to human welfare.

The Type II enzymes prove to be incredibly versatile and diverse: Recognition sites can be palindromic, asymmetric, with ambiguities or indifferent internal bases, and/or differentially sensitive to methylated bases; the enzymes might have one or two catalytic sites, cleave DNA in one or two steps, with or without sliding and detaching from the DNA, and with or without looping. Crystal structures in combination with database searches have been useful to build evolutionary trees, which questioned the view held until the mid-1990s that the baffling lack of common features suggested independent convergence and not divergence from a common ancestor. Despite the lack of sequence similarity, the majority of the Type II enzymes have a common catalytic core, mainly with the PD…(D/E)XK motif, but also the HNH and GIY-YIG structural domains, and some the PLD domain. These catalytic domains seem to have been “mixed and matched” during the course of evolution. The structures of the current 50-odd Type II enzymes (and the first Type I and Type III enzymes) are revealing new and unexpected details about the mechanisms employed to prevent indiscriminate restriction: The catalytic cleavage domain may be hidden behind the recognition domain requiring a large conformational change dictated by the recognition domain or the methyltransferase to access the DNA; it may require dimerization (or multimerization) and/or two unmodified recognition sites; and it may require distortions, bending, or contortions of the DNA to properly position the two nucleolytic sites. A variety of other types of control of restriction emerged over the years and involve transcription regulation, control by “C” proteins, or the cognate methyltransferase (in the case of Type II systems). Various plasmids and phages employ DNA mimics and other proteins to inhibit Type I enzymes. Rather spectacular was the finding that the restriction subunit of EcoKI was degraded by the bacterial host's ClpXP protease during translocation of the EcoKI complex (but not before), when modification was impaired. Such extraordinary control of restriction was in sharp contrast to that of Type II enzymes, which destroy their own host DNA under similar circumstances. This led to the debate on recognition of “self” versus “non-self.” The “pro-self” camp stated that the restriction enzyme would not only destroy incoming DNA, but it would also enhance the frequency of horizontal transfer of restriction-modification systems by generating recombinogenic free DNA ends in the cell. These ideas are not mutually exclusive.

By 1993, only two dozen Type I and III enzymes were known, but this changed with the advent of whole-genome sequencing projects, which would lead to the identification of many more (putative) Type II restriction enzymes and to the realization that Type I and III enzymes are quite common in bacteria and archaea, like Type II enzymes. The evolution of the Type I DNA specificity genes became a hot topic in the 1980s, as new specificities could be generated via homologous recombination, unequal crossing-over, and transposition. Decades later this finding has become a finding of great importance to understand life-threatening bacterial infections in humans and other organisms. In 1988 Bill Studier proposed the collision model for Type I enzymes based on his work with phage T7, which proved to be correct in the following years and is apparently not limited to Type I enzymes. Extensive modeling of Type I enzymes led to the tentative conclusion for a common ancestor with one monomeric recognition domain and a separate catalytic domain for methylation, whereas the ATP-dependent molecular motor domains of Type I and III enzymes were assigned to the SF2 helicase superfamily, but proved to be translocases that do not open up the double helix. The breakthrough in 2012 concerned the structures of the Type I EcoKI and EcoR124I enzymes, which in the absence of crystals relied on single-molecule studies, and computer-assisted EM single-particle reconstructions. In 2015, the first structure of a Type III enzyme was published, that of EcoP15I, which indicated a division of labor of the two modification subunits: one for DNA recognition, the other for methylation. This threw light on the differential usage of ATP by Type III enzymes, which involves a large conformational change. The first long-awaited structures led to a comparison of the Type I enzymes with the Type IIB and IIG enzymes: A Type IIB is a motorless Type I system, whereas a Type IIG would be a half of a motorless Type I system. Of interest also are the Type IIG enzymes that are combined Type I–like restriction-modification enzymes that are either SAM-dependent or at least stimulated by SAM (like Type III enzymes). The major conformational change to DNA or protein, or both, to reposition the catalytic cleavage site of many Type II enzymes is apparently not limited to Type II enzymes and appears to be dictated in Type I and III enzymes by the MTase and also by cofactor SAM. The necessity for large conformational changes before cleavage indicates that perhaps the Type I, II, and III enzymes are not so different after all.

The new class of Type IV enzymes, defined in 2003 as modification-dependent restriction enzymes, recognizes a wide variety of DNA modifications at cytosine or adenine residues, and these enzymes have become important tools for research into DNA modifications in all kingdoms. The Type IV enzymes are highly diverse—a mix and match of various cleavage and DNA recognition domains. The Type IV enzymes present models to study eukaryotic modifications, and their role in the study of epigenetic phenomena will be obvious. Identifying new Type IV enzymes is of great importance for these studies, but unfortunately their genes are not easy to detect in whole-genome sequences.

The role of restriction and modification enzymes in bacterial pathogens and also in bacteria and archaea in our gut (the “microbiome”) has become a topic of great interest to the medical field. The presence of different Type I enzymes allows typing of different “non-typeable” pathogenic strains and should also be useful to analyze diet- or disease-induced changes in the microbiome. Restriction enzymes are linked to virulence via phase variation, which may be a common strategy to create phenotypic heterogeneity, in order to hide from the host immune system, or survive environmental changes as a population. Phase variation can occur via hypermutation at simple repeat sequences, or homopolymeric tracts located within the reading frame or promoter region in a subset of genes. This switches promoters “on” or “off” and causes frameshifts and/or alternative usage of translation initiation codons in different reading frames. And, obviously, if this switch affects a methyltransferase, this will have major effects on the methylome.

What will the future bring? More crystal structures may bring more surprises and remain a model for the much more complex eukaryotic DNA-recognition-cum-restriction-and/or-modification complexes such as that of, for example, the Xeroderma pigmentosum disease, or other complexes involved in DNA repair, recombination, or replication. The studies on phase variation will continue, with emphasis on the impact on microorganisms in the human world outside the laboratory, especially those of pathogens, and those in the gut. Of interest are some reports that need to be followed up: Type I enzymes may be linked to stress responses via associated anticodon nucleases or become phosphorylated, whereas both Type I and IV enzymes have been reported to be able to cut a replication fork. An important issue that still needs to be resolved is that of maintenance versus de novo methylation of EcoKI. Although EcoKI has a preference for methylation of hemimethylated DNA, in the presence of the small lambda Ral protein, the enzyme efficiently methylates recognition sites with either one or no methyl groups (i.e., it changes the enzyme from a maintenance to a de novo methyltransferase). Noreen Murray managed to generate EcoKI* mutants with Ral-independent de novo methylation. Interestingly, such mutants have a single high-affinity SAM-binding site in contrast to the wild-type enzyme, which has two high-affinity sites. How the ability to bind only one SAM molecule properly results in a de novo enzyme, without apparently affecting the methylation reaction as such, requires further investigation and remains a subject close to the author's heart.