Discovery of Deoxyribonucleic Acid (DNA)

Friedrich Miescher discovered an unknown compound, later identified as chromatin, in 1869. He extracted a gelatinous material from various cells (initially human pus), and discovered it contained much inorganic phosphorus. This newly identified biochemical material was named ‘nuclein’ because it was always associated with what the histologists designated nuclei. During the period 1885-1900, it was discovered that beside phosphorus, ‘nuclein’ was also rich in a carbohydrate (later identified as a deoxypentose) and in the organic bases adenine, thymine, guanine, and cytosine. The linear structure of the purified organo-phosphate polymer was finally solved by Phoebus Levene (period 1909-1929). At that time, the DNA polymer was thought to be the scaffold of some important elements within the chromatin. No connection was made between this ‘boring long polymer with only four types of nucleotides’ and the molecular basis of transmission of hereditary characteristics that geneticists were eagerly seeking.

Detailed study of polymeric DNA began in 1928 when Fred Griffith suspected that a “genetic transforming principle” was associated with the ‘nuclein’. However, it was only in 1944 that Oswald Avery and his research group,¹ using almost-pure DNA from Streptococcus cell extracts, inferred that DNA contains genetic information. It took another year before Avery demonstrated that the transforming activity disappeared after DNAse treatment.² Then the race to identify the detailed chemical structure of the ‘genetic’ DNA really started. First, Rollin Hotchkiss³ confirmed the genetic nature of DNA, while Erwin Chargaff⁴ discovered that adenine with thymine and guanine with cytosine always exist in a 1:1 ratio, although the ratio of G+C/A+T varies from species to species. Based on these crucial observations, together with the very first crystallographic data of DNA fibers obtained by Rosalind Franklin working in the laboratory of Maurice Wilkins, and based on competitor Linus Pauling’s suggestion that DNA could have an helical shape, Francis Crick and James Watson proposed in 1953 the double helix structure of DNA,^5,6 which revolutionized our concept of the transmission of genetic characters. Next came the identification and purification of the first DNA restriction enzymes^7,8 that recognize a defined sequence in DNA and cut it specifically. Together with the invention of techniques for DNA sequencing,^9,10 these advances allowed the development of recombinant DNA technology¹¹ and opened the field of modern molecular biology.¹²

Discovery of Ribonucleic Acids (RNAs)

It was not until later in the 20th century that scientists realized there are two types of nucleic acids, DNA and RNA, the latter involving ribose instead of deoxyribose and uridine (or pseudouridine, the ‘fifth ribonucleoside’—see below) instead of thymidine. The reason was that little attention was given to the presence of RNases, and any ‘RNA’ identified in cell extracts was just a mix of degradation products a few nucleotides long.^13,14 Degradation of DNA by metal-dependent DNases was easier to avoid. Thus while DNA research was progressing well, the chemistry of the second type of nucleic acid (RNA) remained obscure until the 1950s. Only after introducing detergent (as for DNA preparation¹⁵), associated with phenol for purification, were the first long RNA polymers finally identified in 1956-58 (ribosomal RNA¹⁶ and ‘soluble’ RNA—now called transfer RNA¹⁷).

Wide interest in these new types of nucleic acids emerged only after Crick hypoth esized in 1955 (but published only in 1958) that an RNA molecule should be the inter mediate between DNA and proteins (known as the ‘RNA adaptor hypothesis¹⁸), and later on advanced the ‘Wobble hypothesis’ for decoding mRNA.¹⁹ Initially, Crick thought the adaptor molecules might be the small RNA molecules that were known to be present in cell extracts, until the ‘soluble’ RNAs (tRNAs), able to be specifically aminoacylated,²⁰ were identified and characterized in 1958. The concept of messenger RNA and regulatory mechanisms in the synthesis of proteins was formulated in 1961 (refs. 21, 22). The genetic code was finally solved and officially presented during a Cold Spring Harbor Symposium²³ in 1966. In the meantime (1965-67) the first fully sequenced tRNAs specific for alanine,²⁴ tyrosine,²⁵ serine²⁶ and phenylalanine,²⁷ all from yeast were fully sequenced. These sequences included the identification and location of no less than 17 different noncanonical nucleosides, among them two hypermodified nucleosides N6-isopentenyladenosine (i⁶A) and wyosine (yW). The first crystals of tRNAs were produced and the first three dimensional structure of one of them²⁸ was finally solved in 1974. This was the birth of structural biology of the nucleic acids.

Modified Nucleosides in Genomic DNAs

During the period 1920-45, naturally occurring nucleic acid polymers (DNA and RNA) were thought to contain only four canonical nucleosides (ribo-or deoxy-derivatives): adenosine, cytosine, guanosine, and uridine or thymidine. However, after analyzing a picrate precipitate from a hydrolysate of DNA of avian tubercule bacilli, Johnson and Coghill²⁹ detected a minor amount of a methylated cytosine derivative (m⁵dC, Fig. 1). This report was later disputed by Vischer et al³⁰ because they could not reproduce the result, but Johnson and Coghill were in fact correct. Only in 1948 was the presence of m⁵dC in DNA from calf thymus^31,32 firmly established using the new technique of paper chromatography of DNA hydrolysates.^31,32 This was followed in 1958 by the detection of N6-methyl adenine (m⁶dA) in microbial DNA.³³ It was not until much later—1964—that the methylation of cytosines and adenosines within DNA molecules was shown to occur by enzymatic post-replicative modification (see below the section concerning enzymes). A surprise discovery during the period 1953-63 was that the DNA of some bacterial viruses lacks deoxycytosine (dC) or deoxythymidine (dT) and instead contains 5-hydroxymethyldeoxycytosine³⁴ (hm⁵dC), 5-hydroxymethyldeoxyuridine³⁵ (hm⁵dU) or simply deoxyuridine³⁶ (dU). These modified cytosines or thymidines (100% replacing of the standard base dC or dT completely), unlike the m⁵dC and m⁶dA in bacterial and mammalian DNA, are generated at the precursor level (prereplicative modification) and subsequently incorporated into phage DNA by the bacteriophage polymerase.³⁷ However, the hm⁵dC in phage DNA can be further glucosylated at the polymer level^38,39 by direct transfer of glucose from UDP-glucose to form hexosylated derivatives glc-hm⁵dC / glc-hm⁵dU and even di-glucosylhydroxy methyl deoxy uridine glc-glc-hm⁵dC. Note that a minor amount of hm⁵dU and glc-hm⁵dU (also designated Base J) have been found recently in genomic DNA of flagellated protozoa of the order Kinetoplastida (Trypanosoma brucei for example) and in the closely related unicellular alga Euglena gracilis (see chapter by Sabatini et al). In this case hydroxylation of deoxyribothymine and the subsequent glycosylation step occur at the polymer level. Later (1972-81) came the discovery of new uridine derivatives containing putrescinyl-, glutamyl-or dihydroxpentyl groups linked to C5 of the uracil ring (symbolized by Put-m⁵dU, Glu-m⁵dU and Dhp⁵dU respectively^40,41). In the case of Dhp⁵dU, glucose or gluconolactone-1-phosphate can be further attached on one of the two free hydroxyl groups leading to hypermodified Glc-Dhp⁵dU and GlcP-Dhp⁵dU, respectively. In these latter cases, depending on the type of chemical alteration, the extent of replacement of the canonical dT by modified dU derivatives was estimated to be 15-60%. In E. coli phage Mu, a substantial number of adenines were found modified to N6-carbamoylmethyl adenine (ncm⁶A, 15% of dA), while in S. elongatus phage S-2L 100% of adenines are methylated to 2-aminoadenine⁴² (m²A) or N2-N6-dimethyladenine⁴³ (m^2,6A). Also, in the phage DDV1 infecting Shigella sonnei, a trace amount of 7-methylguanine (m⁷dG, about 1% of dG) was found. In contrast, other types of noncanonical modified deoxynucleosides would most probably be identified were more bacterial and phage DNAs to be explored—an endeavor that unfortunately has been much neglected in the past decade (discussed in the chapter by Forterre and Grosjean). Quite recently (1983-87), N4-methylcytosine (m⁴dC) and also deoxyinosine (dl) were identified in some bacterial DNA,^44-46 especially from thermophiles. The selective advantage of m⁴dC over m⁵dC at high temperatures is thought to be to avoid production of mutagenic m⁵dU resulting from heat-induced deamination of m⁵dC, and m⁴dC is indeed more resistant to deamination at high temperature than m⁵dC (discussed in ref. 47). A surprising recent discovery (2005) is that the phosphoryl group in bacterial DNA can be thiolated to form a phosphoro thionate linkage of the S_p chiral configuration;^48,49 the mechanism remains to be elucidated (commented by Eckstein⁵⁰).

Figure 1

Modified bases (and phosphate) in DNAs. In boxes are the chemical structures or the description of the chemical composition of adducts to selected atoms of pyrimidine ring (upper part) and purine ring (bottom part) found in cellular (nuclear) genomic (more...)

In conclusion, so far relatively few naturally occurring modified deoxynucleosides have been identified in genomic DNAs (summarized in Fig. 1). The most common modifications are simple methylation of either the C5 atom of the cytidine ring (m⁵dC in almost all kinds of organisms) or the exocyclic amine groups of adenine (m⁶dA mainly in bacteria and archaea) or cytidine (m⁴dC mainly in thermophilic bacteria and Archaea). Unusual, deoxynucleosides (sometimes hypermodified) are confined to bacteriophages and viruses (reviewed in refs. 40-41).

Modified Nucleosides in Coding and Noncoding RNAs

In the case of RNA, the story is very different and far more complex. In contrast to DNA, we now know that every position of a pyrimidine or a pyrimidine ring (Figs. 2 and 3 respectively) can be posttranscript ionally modified, not only by methylation or hydroxymethylation, but also by deamination, transglycosyl ation, acetyl ation, reduction, thiolation, oxidation, ribosylation, formylation, isomerization, selenat ion, or multiple group additions or transfer …. singly or sequentially (Fig. 2). Moreover, the 2’-hydroxyl group of the ribose moiety can be methylated (alone or in combin ation with base modific ations) or ribosylated with a bulky adenosine-5’-phosphate group. To date 110 - 119 (depending on how certain ‘hypermodified’ modified nucleosides are considered) naturally occurring modified nucleosides have been identified in different types of RNAs, not only tRNAs and rRNAs, but also mRNAs and snRNAs like sn/snoRNAs, miRNAs, and chromosomal RNAs. The most widespread RNA modifications are base or ribose methylations (symbolized by mX or Xm respectively) and isomerization of uridine into pseudouridine (Psi). The majority of hypermodified ribonucleosides occur in transfer RNAs; these modifications include long lateral chains or multiple substituents on two or more atoms of the same purine or pyrimidine ring (see below).

Figure 2

Modified bases and ribose in RNAs. In the boxes are the various types of chemical groups that can be enzymatically attached to selected atoms of a pyryrimidine ring (in red in the color version available on the Web) during maturation of RNA precursor (more...)

Figure 3

Modified bases and ribose in RNAs. In the boxes are the various types of chemical groups that can be enzymatically attached to selected atoms of a purine ring (in red in the color version available at www.landesbioscience.com/curie) during maturation (more...)

How was this vast body of information on the identity and location of the many modified nucleosides in RNA acquired? The story starts only in 1951, after the discovery of m⁵dC in DNA, when W. Cohn and his colleagues⁵¹ used paper chromatography of an acid hydrolyzate of enriched ‘soluble RNA’ of yeast to identify a new compound in addition to the four expected ribonucleosides. This compound, initially designated by a question mark ‘?’, was shown later in 1957-58 to be 5-ribosyluridine, also called the ‘fifth nucleoside in RNA’ (now designated pseudouridine^52,53). Pseudo uridine accounts for about 4% of the molecular weight of the total constituent nucleosides in yeast tRNAs and is the most abundant modified nucleoside identified so far in all kinds of tRNA as well as rRNA. The next modifications to be identified (about one year later) were 2’-O-methylribose derivatives (Xm = Cm, Um, Gm, Am; these are the second most abundant class.^54,55 Also identified in the same period were 5-methylribouridine (m⁵U, ribothymine or ribo-T, one of which occurs in almost every tRNA), 5-methylribocytosine (m⁵C, also abundant in mammal ian tRNAs, up to 2-3 per tRNA) and a few other simple methylated adenine and guanine derivatives (m¹A, m²G, m¹G, m⁷G…; refs. 56-59). Mainly from sequence information of tRNAs from yeast, E. coli and mammals, many other modified nucleosides were identified, including N6-isopentenyladenosine (i⁶A and its variant ms²i⁶A), wybutosine (yW—see chapter by Urbonavicius et al), and N6-carbamoylthreonineadenosine (t⁶A). By 1970, thus just 20 years after the discovery of Psi in RNAs, 35 well-characterized modified nucleosides had been identified compared to only five in DNA (reviewed in the book by RH Hall,⁶⁰ the only one available to date dealing with RNA and DNA modifications. For details concerning chemical structures, occurrence and classification of modified ribonucleosides, as well as on metabolic pathways, and enzymes catalyzing RNA modification reactions, consult the MODOMICS database at http://modomics.genesilico.pl; see also Appendix 1 by Rother et al and additional Web links in Appendix 6 in this volume. Other useful sources of information are in references 61-64.

Degree, Extent, and Pattern of Nucleoside Modifications

A modified nucleoside at a given position within a population of RNA molecules may not be present in all of them so that the molar ratio (or % proportion) of a given nucleoside in a population of RNA molecules (referred to as the degree of modification) can be less than 1/1 (less than 100% modified) at a given site. The degree of modification may vary according to the physiological conditions (oxygen concentration, temperature, availability of metabolic intermediates or cofactors, metabolic stress, malignancy…) of the cell from which the RNA came, thus creating a micro-heterogeneity in the RNA population (‘modivariants’; see chapter by Giegé and Lapointe). In some cases, modivariants can be separated by simple chromatographic procedures. For example, the molar ratio of ribothymine (m⁵U, ribo-T) at position 54 in the T-Psi loop of all types of tRNAs is usually 1/1 (100% U-54 methylated), while the molar ratio of thiolation on C2 of the ring of the same uridine-54 in the tRNA of thermophiles (harboring m⁵s²U instead of m⁵U, as in Thermus thermophilus—see chapter by Noma et al) can be less than 1/1 or even zero, especially when the organism is grown at temperatures below that optimal for growth. It is important to remember that in the RNA modification data banks (tRNA, rRNA, snRNA) the presence of a given modified nucleotide at a given position of an RNA molecule is indicated (m⁵s²U as in the example above), but never the degree of modification of that particular base or ribose. This caveat is particularly relevant for ribosomal RNA, where for instance the degree of modification of a particular Psi or 2’-O-methylribose (Xm) can be very low. Since the DNA genome in principle exists in only one copy per cell, the notion of degree of modification does not apply to DNA. However in some microorganisms this is not the case: Synechococcus for example has about 10 copies of the chromosome, while in certain hyperthermophilic and halophilic archaea, this number can be as high as 20 copies of the chromosome. Then the notion of degree of methylation should apply.

The extent of nucleic acid modification concerns the relative amount of a given modified nucleoside that exists at several positions within a given RNA or a DNA molecule, usually expressed as % replacement of total nucleosides (or total of a particular canonical one) in the whole nucleic acid molecule. For example, the extent of post-replicative modification of dC into m⁵dC for bacterial, archaeal, and eukaryal genomic DNA is generally 1-8 % of the total dC, except for mammalian and plant DNAs where m⁵dC can reach 30% of total dC. In phage DNAs, where modifications arise by a prereplicative event, the extent of modification can reach 100%. The extent of total modifications in tRNA molecules from plant and mammals is also high (up to 25%), whereas that in homologous tRNAs from bacteria is lower (2-15%—reviewed in refs. 40, 60).

The pattern of modifications in RNA/DNA is a more complex, qualitative concept. Here, comparison of different nucleic acids is made by taking into the account type, location and diversity of modifications which of course differ greatly from one type of nucleic acid to another (for example DNA versus RNA, or rRNA versus tRNA or mRNA). More interesting is that distinct and characteristic patterns of modification exist between homologous nucleic acids from phylogenetically distant organisms (see Fig. 1), as well as between tRNAs of the same organism (see below). The pattern of modification is the ‘fingerprint’ or ‘identity card’ of a RNA molecule, in the same way that a restriction pattern is the ‘fingerprint’ or ‘identity card’ of a DNA molecule. As more sequences of RNAs themselves (not the sequence of their genes or RT-PCR products) become available, this important feature of nucleic acids will become more evident.

Nucleosides Found in Coding and Noncoding RNAs

Figure 4, shows the symbols of 107 structurally distinct modified ribo nucleo sides identified so far in different RNAs from various Eukarya, Bacteria, or Archaea.⁶⁵ The information comes primarily from RNA sequence data and from analysis of RNA nucleoside composition by thin-layer chromatography, high performance liquid chromatography and/or mass spectrometry (for examples see refs. 66-70). Symbols indicated in normal characters (in red in the version on the Web Site of this chapter), initalics, bold (in blue) or in normal charaters, underlined (in black) correspond to modified nucleosides found in tRNAs, rRNAs, or in both t+rRNAs, respectively. Organelle (mitochondrial and chloroplastic) tRNAs and rRNAs contain their own set of modified nucleosides, some of which (like cmnm⁵U, k²C, τm⁵U, τm⁵s²U, f⁵C, f⁵Cm) are not present in cytoplasmic RNAs of the eukaryotic host cell. The corresponding modification enzymes, now encoded in the host genome, are believed to have originated from ancient bacterial endosymbionts. Therefore, while present in the Eukaryal domain, mitochondrial-modified nucleosides should be considered as ‘bacterial by origin’ or at least belonging to both Eukarya and Bacteria (they are boxed in the intersector E-B in Fig. 4). Symbols of modified nucleosides outside the circles correspond to those found in eukaryal mRNAs (normal characters, in red) and snRNAs (italics, bold, in green) or in both mRNAs and snRNAs (italics, underlined, in black). Five members of this eukaryal group are unique to mRNAs and/or snRNAs, while others are also present in eukaryal tRNAs and/or rRNAs (indicate by an arrow).

Figure 4

Phylogenetic distribution of modified nucleosides present in RNAs from the three domains of life. Symbols are written differently according they were found in tRNAs, rRNAs, mRNAs and/or in sn(o)RNAs. For details see text. A color version of this image (more...)

Figure 4 shows that half of the modified ribonucleosides are domain specific. These presumably arose later during more than evolution, after the separation of organisms into the three domains. About one fifth of the other modified nucleosides are located within overlapping sectors of the circles and thus found in two or more domains: either between Eukarya and Bacteria (E+B), or Bacteria and Archaea (B+A) or Archaea and Eukarya (E+A). The remaining fifth of modified nucleosides are present in all kinds of organisms (E+B+A). They are the simplest types of modification, several are found in all types of RNAs. From this observation, it has been inferred⁷¹ that they correspond to relics of modified nucleosides that were present in primordial organisms existing before the three biological domains separated. However, the reality may not be so simple. Symbols like m¹G or m⁵U within the central common sector E+B+A correspond to modifications that are located in different positions and in different types of RNA molecules, each of them being produced by site-specific as well as RNA-specific enzymes that do not necessarily belong to the same protein family. Some cases most probably represent convergent rather than divergent evolution, so that the evolutionary history of the emergence of RNA modification machinery is complex (see for examples refs. 72,73, also chapters by Czerwoniec et al, by Myllykallio et al and by Forterre and Grosjean).

Concerning doubly modified nucleosides of the type xNm (like m²Gm or ac⁴Cm), a majority of them were found so far in archaeal RNAs. They correspond in fact to combinations of simple methylation of the ribose (Gm or Cm) and of enzymatic alteration of the base (m²G or ac⁴C), each of the ‘independent’ modific at ions being found within the three overlapping E+B+A sectors, or in the E+A sectors (Fig. 5). Thus, while modified nucleosides like xNm’s are indeed found mainly in archaeal RNAs, the corresponding modification enzymes may not necessarily be unique to archaea.

Figure 5

Localization of ‘doubly modified’ nucleosides at the base and the ribose (in red in the color version available at www.landesbioscience.com/curie) of RNAs of Archaea. Lines point out which ones among these hypermodified nucleosides correspond (more...)

Lastly, modifications like imG, imG2, mimG, yW, OHyW, o2yW and OHyW*, or preQo, preQ1, Q, oQ, gluQ, manQ, GalQ and G+, or nm⁵U, or cmnm⁵U, mnm⁵U are merely intermediates of the same phylogenetically related stepwise metabolic reaction chain (see chapters by Urbonavicius et al for wyosine derivatives, by Iwata-Reuyl and de Crécy-Lagard for queuosine derivatives, and by Bessho and Yokoyama for the modified uridines series). Consequently, the real diversity of naturally occurring modified nucleosides as it appears in Figure 4 could probably be reduced from 107 to about half truly distinct, biosynthetically unrelated types of chemical structures in RNAs. However, exploring more RNAs, especially from extremophiles, may uncover new types of modified nucleoside.

Nucleosides Found in Genomic DNAs

Figure 6 summarizes the types of modified deoxyribonucleosides found in DNA from different origins. Symbols for modified deoxynucleosides found in genomic DNA of a cell, or in DNA of bacteriophages and eukaryal viruses, are indicated in different types of boxes. In the cases of cellular genomic DNA, almost all (if not all) modified deoxynucleosides are formed by post-replicative enzymatic modification processes; their extent ranges from 1% to 30% (refs. 40,60). In viruses, on the other hand, modified deoxynucleosides are derived either by post-replicative modification processes or via incorporation of modified deoxynucleotide precursors directly into DNA by the virus DNA-dependent DNA polymerase (prereplicative process). In this later case the extent of DNA modification can reach 100%. In the case of viruses and bacteriophages, only few modified deoxynucleosides are found to be common (m⁵dC, m⁶dA and hm⁵dU, there are also those found in genomic DNA of bacteria or Eukarya (reviewed in refs. 40,41).

Figure 6

Phylogenetic distribution of modified deoxynucleosides present in genomic DNA from the three domains of life. Distinction is made according their origins: from cellular/nuclear DNAs (in circles) or from viruses/bacteriophages DNA (in squares). This figure (more...)

Discovery of RNA Modification Enzymes

The first evidence for existence of enzymes able to modify nucleic acids at the polymer level came in 1962-63. After incubating transfer RNA with E. coli cell extract and S-AdoMet labeled in the methyl group, three groups^81-83 demonstrated independently that radioact ivity appeared in methylated bases in RNA. The first identified modification enzyme⁸² was tRNA:m⁵U54 methyltransferase, now designated TrmA in Bacteria and Trm2 in Eukarya. Soon after followed the discovery of similar activities for other methyl transfers specific for the formation of m¹G, m⁷G, m²A, m⁶A, m^2,2G and m⁵C in E. coli transfer RNAs⁸⁴ and four additional distinct activities for the formation of m⁶A, m^6,6A, m⁷G and m⁵C in ribosomal RNA.⁸⁵ These simple but important experiments illuminated a new feature of RNA metabolism, namely that methyl group incorporation can take place after polymerization and not, as was shown earlier (1958) for bacteriophage DNA, by incorp oration of deoxyribonucleotide triphosphate analogs (such as m⁵dCTP) during replication.³⁷ Since the pioneering work on RNA methylation, all subsequent modifications identified in RNAs of many different types of cells have been found to occur the same way, i.e., by enzymatic posttranscriptional alteration of a base and/or of the ribose at the RNA precursor level. Many other RNA processing enzymes catalyzing reactions as diverse as 5’-and 3’-trimming, 5’-capping, RNA-splicing, CCA and polyA addition likewise act posttranscriptionally (RNA maturation process, see chapter by Hall and Li). The precise interplay of these various types of RNA alterations allows in fine to produce fully mature RNAs with many new chemical ‘decorations’ as described in preceding paragraphs.

Later in 1975, a completely different type of RNA methyltransferase using 5,10-methylene tetrahydrofolate (CH2-THF) instead of S-AdoMet as methyl donor was discovered in Streptococcus faecalis (ref. 86, and chapter by Myllykallio et al). Thus while S-AdoMet is by far the major cellular source of methyl groups (and is often called the ‘universal methyl donor’), an alternative solution exists for methylating RNAs.

An enzyme catalyzing the insertion of a guanine in tRNA (via a transglycosylation reaction)⁸⁷ was identified in rabbit erythrocytes 1973-75. It was only few years later that the physiological function of this ‘G’-inserting enzyme was discovered:^88-89 the insertion of a deazaguanine derivative in the anticodon of few selected tRNAs. This enzyme, now designated tRNA-guanine-34 insertase (abbreviated tgt), removes the encoded guanine located at the first position of the anticodon of precursor tRNA by cleaving the canonical C1-N1 glycosidic bond and inserting in its place a premodified 7-deazaguanosine derivative precursor (or a guanine as in the original observation of Farkas and coworkers;⁸⁷ see chapter by Iwata-Reuyl and de Crécy-Lagard). A similar type of enzyme was recently found in Archaea (in 1997). In this case,⁹⁰ formation of archaeosine (G⁺, another type of deazaguanine derivative) at position 15 in the D-loop of archaeal tRNAs depends on a similar, phylogenetically related tRNA-guanine-15 insertase designated a-tgt. It should be mentioned that formation of pseudouridine in RNA proceeds by a similar mechanism, except that it is the genetically encoded uracil base that is replaced in RNA after a 180° rotation and reformation of a noncanonical C1-C5 glycosidic bond (cis-transglycosylation or isomerization reaction—see chapter by Mueller and Ferre d’Amare).

Another remarkable recent discovery (1996-97) is that some RNA modification enzymes are ‘guided by RNA’. This was first demonstrated in the case of enzymatic form ation of 2’-O-methyl ribose in yeast and mammalian rRNAs,^91,92 immediately followed by the same discovery in the case of Psi formation also in rRNAs.^93,94 This observation has since been extended to the formation of 2’-O-methylribose and Psi in many other RNAs (tRNAs, snrRNAs, snoRNAs) of Eukarya and/or Archaea; however, neither bacteria, nor organelles examined so far use this ‘RNA-assisted’ type of enzyme, in fact ‘RNA-assisted’ multiprotein enzymatic complex (see chapters by Gagnon et al, by Grozdanov and Meier and by Karijolich et al). Note that a given ribose methylation or uridine isomerization in RNA can be mediated by a ‘classical’ all–protein enzyme in one organism, while in another organism the same modification is catalyzed by the RNA-assisted multiprotein machinery (see for examples refs. 95-99). This observation raises interesting questions about the evolutionary pressures that favour one type of RNA modification system over the other. Perhaps the main advantage for a cell using an ‘RNA-assisted’ enzyme machinery instead of an ‘only protein’ enzyme is that, with only few proteins (besides the enzyme) required for elaborating the RNA-guided RNA machineries, and with a huge array of guide RNAs (of which the sequence is more versatile than that of proteins) many more nucleosides in RNAs can be targeted. However, this might not be the sole advantage (discussed in chapters by Gagnon et al, by Grozdanov and Meier and by Karijolich et al).

Discovery of DNA Modification Enzymes

At almost the same time as tRNA:m⁵U54 methyltransferase was discovered (1963), however before the first sequence-specific restriction enzyme was identified^8,12 (and the importance of restriction/modification self-defence mechanism in bacteria was recognized), enzymatic activities for ‘post-replicative’ methylation in polymeric DNA were beginning to be identified.¹⁰⁰ Partially purified S-AdoMet-dependent methyltransferases of E. coli were shown to catalyze the formation of m⁵dC and m⁶dA in double-stranded DNA.¹⁰¹ Similar enzymes were subsequently identified in many other types of bacterial and eukaryotic cells, as well as in certain bacteriophages (reviewed in refs. 40,41; and chapters by Coffin et al; by Cheng and Blumenthal and by Jeltsch and Jurkowski). Enzymatic post-replicative DNA glucosylation (in fact formation of hyper-modified glucopyranosyloxymethyluracil, base J) was discovered with DNA of bacteriophages^38,39,102 before it was found in eukaryotic DNA^103,104 (see chapter by Sabatini et al). In the 1980s, enzymes catalyzing formation of m⁴dC in bacteria were discovered,⁴⁴ and also a new family of demethyl/dealkyl-methylases acting on both RNA and DNA (AlkB family of enzymes—see chapter by Falnes et al). Another family of dual-enzymes exists, catalyzing the conversion of C-to-U in single-stranded DNA or RNAs and cellular mRNAs (Apobec deaminases—see chapter by Smith). These deaminases play an essential role in cellular defense against viruses and allow new opportunities for variability in gene expression