Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`411
`
`RNA, the first macromolecular catalyst:
`the ribosome is a ribozyme
`Thomas A. Steitz1,2,3 and Peter B. Moore1,2
`
`1Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, CT 06520, USA
`2Department of Chemistry, Yale University, New Haven, CT 06520, USA
`3Howard Hughes Medical Institute, New Haven, CT 06520, USA
`
`Recently, the atomic structures of the large ribosomal
`subunit from Haloarcula marismortui and its complexes
`with substrates have been determined. These have pro-
`vided exciting new insights into the principles of RNA
`structure, the mechanism of the peptidyl-transferase
`reaction and early events in the evolution of this RNA–
`protein complex assembly that is essential in all cells.
`The structures of the large subunit bound to a variety of
`antibiotics explain the effects of antibiotic resistance
`mutations and provide promise for the development of
`new antibiotics.
`
`The RNA world hypothesis gained great support and
`prominence from the discovery that the RNA molecules of
`tetrahymena Group I intron and RNase P catalyze
`enzymatic reactions [1,2]. These results established that,
`in addition to its capacity to function as a carrier of genetic
`information, RNA can serve as an enzyme. The existence of
`these ribozymes made the RNA world hypothesis plaus-
`ible, namely that an early biological world once existed in
`which all enzymes were made of RNA rather than protein.
`However, given the functions of these ribozymes and those
`discovered subsequently, it could still be argued that they
`are not relics of an early stage of evolution but, rather,
`entities that evolved later. Although the existence of these
`ribozymes proves that RNA can be a catalyst and, thus,
`offers evidence for the possibility of a pre-protein RNA
`world, it does not prove it. Now that the ribosome has been
`shown to be a ribozyme there can be no doubt that there
`was a pre-protein RNA world.
`The atomic structures of the large ribosomal subunit
`and its complexes with substrates and products prove that
`RNA is the catalytic component of the macromolecular
`assembly that synthesizes proteins [3–6]. Because the
`catalytic element that synthesizes proteins must have
`preceded proteins, the ribosome and, thus, ribozymes must
`have preceded protein enzymes. The ribosomal structures
`enable us to explore how these large RNA –protein
`machines are put together, to ponder how they might
`have evolved and suggest how they catalyze peptide-bond
`formation.
`The ribosome is a macromolecular machine that carries
`out the mRNA-directed synthesis of proteins. It has two
`subunits, the larger of which sediments at 50S and has a
`
`Corresponding author: Thomas A. Steitz (eatherton@csb.yale.edu).
`
`mass of 1.5 megadaltons (MDa) in prokaryotes, and the
`smaller of which sediments at 30S and has a mass of
`0.8 MDa; together they form an assembly that sediments
`at 70S. In Escherichia coli the ribosome is approximately
`two-thirds RNA, and the large subunit contains 34
`proteins, and the small subunit 21 proteins. The small
`subunit contains the messenger decoding site, where
`interactions between codons in the mRNA and the antic-
`odons of tRNAs determine the order in which amino acids
`will be assembled into protein. The larger ribosomal
`subunit contains the site of peptide-bond formation –
`the peptidyl transferase centre. The ribosome has three
`tRNA-binding sites: the A-site binds aminoacyl-tRNA, the
`P-site binds peptidyl-tRNA and the E-site interacts with
`deacylated tRNAs that are being discharged from the
`ribosome. In addition to requiring a small subunit, which
`specifies the identity of the tRNA to be bound, and a large
`subunit, which stitches the polypeptide together, this
`assembly line machine makes use of two protein-factor
`components: (1) protein elongation factor Tu (EF-Tu),
`which delivers the aminoacylated-tRNAs to the ribosome
`and (2) protein elongation factor G (EF-G), which moves
`the assembly line device along its mRNA subsequent to
`peptide-bond formation. EF-Tu delivers aminoacyl-tRNA
`molecules to the ribosome and leaves upon hydrolysis of
`GTP only when the correct cognate tRNA has been
`delivered. EF-G, which is also GTP driven, facilitates the
`translocation of tRNA and mRNA after peptide-bond
`synthesis. Recent atomic structures of the large and
`small ribosomal subunits together with decades of
`biochemical and genetic studies of the ribosome have
`begun to elucidate, in atomic detail, how this 2.4 million
`molecular weight RNA–protein machine carries out a
`function that is central to all biology.
`
`Structural studies of the ribosome
`Electron microscopy played a key role in the early
`structural studies of the ribosome, beginning with the
`pioneering work of Palade [7] that contributed to the
`discovery of the ribosome, the first determination of the
`shapes of the large and the small subunits in the early
`1970s [8], and continuing with the cryo-electron micro-
`scopic investigations that have advanced to increasingly
`higher resolution [9,10]. At present, however, the only way
`to obtain an atomic structure of an assembly as large as the
`ribosome is by X-ray crystallography, and because this has
`
`http://tibs.trends.com 0968-0004/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0968-0004(03)00169-5
`
`1
`
`MTX1040
`
`

`

`412
`
`Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`long been evident, one might understandably ask why it
`took so long and what the events were that ultimately led
`to atomic structure determination. The essential first step
`– the growth of the first crystals of ribosomal subunits –
`was accomplished by Yonath and Wittmann in the early
`1980s [11]. The first crystals, although exciting in their
`prospect, diffracted poorly. Greater success was achieved
`with subunits obtained from other extremophiles, either
`thermophiles or halophiles, possibly because they are
`more robust to purification and perhaps because they are
`less flexible. Crystals of the 50S ribosomal subunit from
`Haloarcula marismortui (an archaeal halophile from the
`Dead Sea), first grown by Yonath and her collaborators in
`the mid-1980s, and their diffracting qualities were
`improved and by 1991 they diffracted to a resolution of
`3 A˚ [12]. Although these crystals possessed the marvellous
`property of diffracting to a resolution that makes an
`atomic-level structure determination possible in principle,
`they retained several pathologies. They were extremely
`thin, multiple, prone to non-isomorphism, radiation-
`sensitive and, often, as it turned out, twinned.
`Although crystals are essential for structure determi-
`nation, the challenge in solving the structure of such a
`large assembly is obtaining the phases associated with the
`hundreds of thousands of diffraction amplitudes their
`crystals yield. Just as the determination of the structure of
`myoglobin was a challenge because it was approximately
`one order of magnitude larger than the largest molecular
`structure that had been determined previously (in the
`1950s), the determination of the ribosome structure posed
`a similar challenge in the mid-1990s because it too was
`approximately one order of magnitude larger than the next
`largest asymmetric macromolecular structure that had
`been solved up to then. The only approach that appeared
`likely to succeed was the method of heavy-atom isomor-
`phous replacement pioneered by Max Perutz, in which
`heavy atoms are bound to specific sites in the crystal to
`make a derivative, and to follow the strategy of Kendrew,
`who began his myoglobin studies at low (5.5 A˚ ) resolution
`before proceeding to higher resolution. The Yale group
`(Nenad Ban, Poul Nissen, Jeffrey Hansen, Peter Moore
`and Thomas Steitz) began its crystallographic studies of
`the H. marismortui 50S subunit at very low resolution
`(16 A˚ ), which vastly decreased the effort required to assure
`that heavy-atom-derivative data were correctly analyzed
`and increased the diffraction signal obtained from heavy-
`atom-cluster compounds [13].
`Because the macromolecular assembly to be solved was
`supersized, a supersized heavy-atom compound was
`required. An 18-tungsten cluster compound – which, at
`the low resolution of 16 A˚ , has a diffraction signal that
`approaches 300 times that of a single tungsten atom – was
`used for the first derivative. To check that the positions of
`the tungsten-cluster compound bound to the 50S subunit
`had been correctly determined crystallographically, a
`cryo-electron microscopic reconstruction of the subunit
`determined at 20 A˚ resolution was used to phase the
`low-resolution diffraction from the 50S crystals by mol-
`ecular replacement. A difference electron-density map of
`the tungsten derivative calculated using these phases
`
`http://tibs.trends.com
`
`established that the heavy atoms were indeed correctly
`positioned [13].
`The first electron-density map of a ribosomal subunit
`that showed features expected of RNA was the 9 A˚
`resolution map of the H. marismortui 50S ribosomal
`subunit that was published in 1998 by the Yale group [13].
`This map proved that it was possible to solve the phasing
`problem for diffraction from crystals of objects as large as
`the ribosome. The strategies developed for the large
`ribosomal subunit were quickly followed by Venki Ramak-
`rishnan [14], who was studying the 30S subunit, and
`James Cate and Harry Noller [15], who were investigating
`the 70S ribosome. The structure of the large subunit was
`eventually refined at 2.4 A˚
`resolution (Fig. 1) and
`,100 000 atoms of RNA, protein, water and metal ions
`were placed with an accuracy approaching that of the best
`determined protein structures [3]. The coordinates depos-
`ited for the 3000 nucleotides of RNA and 27 proteins
`increased the known structural database for RNA by an
`approximate factor of four or five and revealed a surface
`area of protein–RNA interaction that is 30 times the
`surface area of tRNAGln that contacts Gln-tRNA synthe-
`tase. The structure determination has also made it
`possible to examine the structures of substrate and
`product-like ligands bound to the peptidyl transferase
`active site as well as a variety of antibiotics that target the
`peptidyl transferase centre and inhibit protein synthesis.
`These structures have provided insights into the prin-
`ciples of RNA architecture, the mechanism of the peptidyl
`transferase reaction and the evolution of this ancient
`macromolecular machine.
`
`Fig. 1. A space-filling model of the large ribosomal subunit from Haloarcula maris-
`mortui with a transition-state analogue bound viewed down the active site cleft.
`Bases are white, the sugar-phosphate backbone is orange and the substrate
`analogue is red. Proteins whose structures are established by the 2.4 A˚ resolution
`map are blue. Cyan ribbons represent proteins whose structures are indepen-
`dently known and have been approximately positioned using lower-resolution
`electron-density maps. Identification numbers are provided for all proteins, and
`‘CP’ designates the central protuberance. Reprinted, with permission, from Ref. [4].
`q (2003) American Association for the Advancement of Science (http://www.
`sciencemag.org).
`
`2
`
`

`

`Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`413
`
`Organization and stablization of the 50S subunit
`structure
`Early investigators of 23S rRNA could only wonder in
`amazement that this large polyanion, of nearly 3000
`negative charges, could fold to form a compact and stable
`structure [16]. We can now see how this is achieved. Three
`kinds of interactions stabilize the tertiary structure of 23S
`and 5S rRNA: (1) Mg2þ
`bridges, (2) RNA –RNA inter-
`actions that are largely of two types – long-range base
`pairs and a newly identified interaction called the A-minor
`motif, and (3) RNA–protein crosslinks. The 23S rRNA can
`be divided into six domains on the basis of a secondary
`structure [17], and the 5S rRNA can be thought of as the
`seventh domain. of the subunit. The shapes of all of these
`domains are highly irregular, although they fit together
`like the pieces of a jigsaw puzzle to form a compact object
`whose overall shape is essentially that of the entire
`subunit [3]. The interactions between rRNA domains
`and the large subunit are so extensive and intimate that it
`is not possible to tell by visual inspection where one
`domain ends and the next begins; the RNA structure of the
`subunit is monolithic, which is different from the domain
`organization of the small ribosomal subunit (Fig. 1).
`Perhaps not surprisingly, magnesium ions play an
`important role in stabilizing the compact structure of 23S
`rRNA by interacting with two or more phosphate groups
`from secondary-structure elements that are remote in the
`sequence, thereby allowing their close proximity [3]. These
`shared Mg2þ
`ions form a neutralizing link between
`phosphate groups whose non-bridging oxygens may be
`either inner- or outer-sphere ligands. Approximately 65 of
`the 108 Mg2þ
`ions identified thus far in the H. marismortui
`large ribosomal subunit stabilize the tertiary structure of
`23S rRNA this way. In addition to single Mg2þ
`-bridging
`sugar–phosphate backbones that are remote in the
`sequence, there are clusters of two or three Mg2þ
`ions
`that function the same way. There are also numerous
`
`monovalent cations bound to specific locations where they
`neutralize the negative charge of the phosphate and, thus,
`assist in specific rRNA positioning.
`A major portion of both the stability and specificity of
`the tertiary structure arises from specific RNA–RNA
`interactions, which largely involve the bases. There are
`base-pairs between nucleotides associated with different
`secondary-structure elements, many of which had pre-
`viously been predicted by phylogenetic sequence compari-
`sons [18]. There are ,100 such long-range base-pairs in
`H. marismortui 23S rRNA. An even more significant
`contribution to the RNA structure is made by a newly
`recognized interaction between adenine and the minor
`groove of RNA helices – an interaction that we term the
`‘A-minor motif ’ [19] (Fig. 2).
`Adenines are disproportionately abundant in the non-
`helical sequences of 23S rRNA, and many of these are
`completely conserved among the three kingdoms of life
`[20,21]. The adenine in an A-minor motif inserts its minor-
`groove face into the minor groove of a base pair in a helix,
`most often a GC pair, where it forms hydrogen bonds with
`one or both of the backbone 20 OH groups of the duplex
`(Fig. 2). Often, two or three consecutive adenines in a
`single-stranded region interact with successive base pairs
`of a helix in this way. There are 186 A-minor interactions
`in the H. marismortui large ribosomal subunit, and 68 of
`them involve adenines that are conserved across all three
`kingdoms [19]. A-minor motifs have both functional and
`structural significance. For example, the 30-terminal
`adenines of tRNAs bound in either the A-site or the
`P-site make A-minor interactions with 23S rRNA base
`pairs in the peptidyl transferase region of the large
`ribosomal subunit [19]. A-minor interactions also play an
`important role in assuring the fidelity of messenger
`decoding by the small ribosomal subunit, where A1492
`and A1493 in 16S rRNA interact via the minor groove with
`
`Fig. 2. The A-minor motif, an RNA– RNA interaction involving adenosines interacting in the minor groove of helices. (a) Examples of the three most prevalent kinds of
`A-minor motifs from the 23S rRNA of Haloarcula marismortui. Types I and II can only be formed by an adenine that makes specific interactions. Type III interactions can be
`made by other bases, but adenine is preferred. (b) The interaction between three consecutive adenines (an ‘A patch’) in helix (H-2 in yellow ball and stick with adenine
`bases shaded red) and the minor groove of helix 26 (H-26) shown in space-filling representation with underlying stick. A519, A520 and A521 of helix 2 make type III and type
`II and type I interactions, respectively. Reprinted, with permission, from Ref. [19]. q (2003) National Academy of Sciences, U.S.A.
`
`http://tibs.trends.com
`
`3
`
`

`

`414
`
`Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`Fig. 3. The structure of kink-turn 7 (KT-7) in the 23s rRNA of the Haloarcula marismortui ribosome. (a) The secondary structure of KT-7. The GC base-paired stem is red, the
`non-canonical base-paired stem is blue, and the bulged nucleotide is green. (b) Base pairing and stacking interactions in kink. The black triangle identifies an A-minor inter-
`action. (c) KT-7 in three dimensions. The backbone of the kinked strand is orange, and that of the unkinked strand is yellow. Broken lines indicate hydrogen bonds.
`Reprinted, with permission, from Ref. [23].
`
`a correctly paired codon–anticodon but not incorrectly
`paired codons and anticodons [22].
`
`RNA –protein interactions
`Interactions between 27 of the 31 proteins of the large
`subunit and rRNA are clearly crucial for the specific
`folding and stability of the large ribosomal subunit. Unlike
`proteins that bind to specific DNA sequences, ribosomal
`proteins bind to their specific RNA sites by recognizing
`unique RNA shapes through interactions that are largely
`with the sugar–phosphate backbone rather than through
`interactions with bases that would be sequence specific [3].
`Twenty-three of the 30 proteins that interact with RNA
`contact two rRNA domains or more. The ‘champion’ in this
`regard is L22, which is the only protein that interacts with
`sequences belonging to all six of the 23S rRNA domains.
`Among the unique secondary structures recognized by
`ribosomal proteins is a new RNA-secondary-structure
`motif that we call the ‘kink-turn’ or ‘K-turn’ [23] (Fig. 3).
`There are six K-turns in the 23S rRNA of H. marismortui,
`and they associate with about a third of the proteins in the
`large subunit. The RNA sequences that form K-turns have
`an asymmetric internal loop that is flanked by GC base
`pairs on one side and sheared GA base pairs on the other;
`an A-minor interaction occurs between these two helical
`stems. This structural motif has a kink in the phospho-
`diester backbone that causes a sharp turn in the RNA
`helix. The K-turns interact with proteins of unrelated
`structures in different ways, but interact with L7Ae and
`two homologous non-ribosomal proteins in the same way.
`The most unexpected features of ribosomal proteins are
`the tails that many of them possess, which penetrate into a
`forest of RNA helices within the interior of the ribosome [3]
`(Fig. 4). Although the globular regions of all large subunit
`proteins are partially exposed to solvent on one side and
`interact extensively with RNA on the other, 12 proteins
`include at least one sequence of significant length that has
`an extended, non-globular structure. Viewed in isolation
`these protein tails, which comprise ,26% arginine plus
`lysine and with abundant glycine and proline residues,
`look like random coils, and probably assume the confor-
`mations they display in the ribosome only upon interacting
`
`http://tibs.trends.com
`
`with rRNA. However, they are not random because the
`sequences of these protein tails are even more conserved
`than the sequences of the globular domains to which they
`are attached. They extend into the interior of the ribosome
`filling the gaps between RNA helices where they interact
`intimately and specifically with RNA groups all over their
`entire lengths.
`large
`Only 19 of the 31 proteins in this archaeal
`ribosomal subunit show clear sequence homology to
`proteins in the eubacterial
`large subunit, but all 31
`proteins are homologous to eukaryotic ribosomal proteins
`[3]. Almost all of the 19 proteins that are conserved across
`kingdoms have known important functions. In the recent
`structure of a eubacterial large ribosome subunit from
`Deinococcus radiodurans [24], six proteins that are not
`homologous to any H. marismortui proteins are bound in
`similar locations to the six H. marismortui proteins that
`are homologues of eukaryotic proteins. A few of the
`eubacterial and archaeal proteins are located in partial
`or non-overlapping positions. Thus, it appears that, at the
`time that the eubacterial kingdom diverged from the
`archaea and eukaryotes, the large ribosomal subunit had
`only 20 proteins.
`
`The ribosome is a ribozyme
`Of all the observations that have arisen from these
`structural studies of the large ribosomal subunit, the one
`that has the greatest functional and evolutionary signifi-
`cance is the finding that the site of peptide bond synthesis
`– the peptidyl transferase centre – is composed entirely of
`RNA [4]. Because the ribosome will catalyze peptide-bond
`formation using substrates that are small fragments of the
`aminoacyl- and peptidyl-tRNA substrates used by the full
`ribosome (Fig. 4), it has been possible to diffuse these
`substrates, as well as analogues of the reaction intermedi-
`ate, into crystals and to establish their structures bound to
`the peptidyl transferase centre. Indeed, when an aminoa-
`cyl-CCA, which binds to the A-site, and a peptidyl-CCA,
`which binds to the P-site, are diffused into the crystals, the
`product CCA is observed in the P-site and an elongated
`peptidyl-CCA is observed in the A-site (Fig. 5), showing
`
`4
`
`

`

`Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`415
`
`OCH3
`
`C
`
`O
`
`OH
`O−
`
`CH3
`
`H3C
`
`O
`
`A
`
`NH
`
`OH
`
`O P O
`
`O
`
`HO
`
`O
`
`+
`
`HN
`O
`
`HN
`
`O
`
`O
`
`C5
`
`NH
`
`C4
`
`S
`
`HN
`
`NH
`
`O
`
`C
`
`O
`
`OH
`O−
`
`HO
`
`O P O
`
`O
`
`C
`
`O
`
`OH
`O−
`
`A
`
`O
`
`OH
`
`OH
`
`O
`
`P O
`
`O
`
`OCH3
`
`50S ribosomes
`40 mM MgCI2
`No MeOH
`37°C
`
`HO
`
`C
`
`O
`
`OH
`O−
`
`H3C
`
`O
`
`CH3
`
`A
`
`NH
`
`OH
`
`O P O
`
`O
`
`O
`
`:NH2
`
`C
`
`O
`
`OH
`O−
`
`HO
`
`O P O
`
`O
`
`+
`
`C
`
`O
`
`OH
`O−
`
`A
`
`O
`
`O
`
`OH
`
`O
`
`P O
`
`O
`
`O
`
`HN
`O
`
`C5
`
`O
`
`NH
`
`C4
`
`S
`
`HN
`
`NH
`
`O
`
`CCA-pcb
`
`C-pmn
`
`CCA
`
`C-pmn-pcb
`
`Ti BS
`
`Fig. 4. A ribosome-catalyzed peptide-bond-forming reaction involving low molecular weight substrates. The reaction of CCA-phenylalanine-caproic acid-biotin (CCA-pcb),
`and C-puromycin (C-pmn) that yields CCA and C-puromycin-phenylalanine-caproic acid-biotin (C-pmn-pcb) is catalyzed by large ribosomal subunits. Reactions of this type
`are analogous to the peptidyl transferase reaction that occurs on intact ribosomes in vivo, and is referred to as the ‘fragment reaction’, because its substrates resemble the
`30 termini of aminoacyl- and peptidyl-tRNAs. Reprinted, with permission, from Ref. [5] (http://www.nature.com/nsb/).
`
`that the ribosome subunit is catalytically active in the
`crystals [5].
`The crystal structures of the large ribosomal subunit
`complexed with substrate and product analogues show
`that only rRNA is involved in the positioning of the A- site
`and P-site substrates, and only RNA is in a position to
`chemically facilitate peptide-bond formation [4–6]. A
`peptidyl-CCA bound in the P-site has its C74 and C75
`base-paired with two guanine residues of the P-loop and,
`correspondingly, the aminoacylated-CCA bound at the
`A-site has its C75 base-paired with a guanine residue of
`the A-loop. Furthermore, the A76 bases of the tRNA
`molecules bound in both the A- and P-sites make A-minor
`interactions. The orientations of the two single-stranded
`CCAs bound in these two sites are related by a twofold
`rotation axis in spite of the fact that the tRNA molecules to
`which they are attached are related to each other by a
`translation. The proposal [4] that this difference in the
`orientations of the 30 termini of the two tRNA molecules
`might help facilitate their translocation after peptide-bond
`formation is, as yet, untested.
`the subunit,
`Although the structure of most of
`including the peptidyl
`transferase centre, remains
`unchanged upon substrate binding, several nucleotides
`exhibit significant alterations in their positions. The
`most dramatic change is A2637 (A2602 in E. coli)
`whose base becomes positioned between the A-site and
`P-site CCAs and interacts with both. Likewise, the
`base of U2620 (U2585 in E. coli) moves and lies
`adjacent to the nascent peptide bond, and a possible
`role in peptide-bond formation is not ruled out.
`
`http://tibs.trends.com
`
`Although the structure of the large subunit containing
`both substrates bound simultaneously has not yet been
`established, an approximation of this complex can be
`achieved by superimposing the structures of the separ-
`ately determined A-site and P-site substrate complexes
`(Fig. 5c). In this hypothetical two-substrate complex, the
`a-amino group of the A-site amino acid is adjacent to the
`ester-linked carbonyl carbon of the peptidyl-tRNA it is to
`attack [6]. This positioning of the reactants by the
`ribosome alone might explain most of the catalytic-rate
`enhancement provided by the peptidyl transferase centre.
`We still do not know the extent to which or how the
`peptidyl-transferase centre might also chemically enhance
`the rate of peptide-bond formation. Importantly, no protein
`moiety is observed closer than 18 A˚ to the site of peptide-
`bond formation. Thus, protein cannot contribute to
`catalysis and, at present, there is no evidence for metal
`ion involvement. However, in the current model for A-site
`and P-site substrates bound to the centre (Fig. 5c) there
`are three RNA groups close enough to the reaction site to
`form a hydrogen bond with the attacking a-amino group:
`(1) the 20 OH of A76 of the tRNA in the P-site, (2) the N3 of
`A2486 (A2451 in E. coli) of 23S rRNA, and (3) the 20 OH of
`A2486. In part, these hydrogen-bond interactions help
`align the a-amino group for its nucleophilic attack, and
`there is reason to believe that hydrogen-bond formation by
`itself could enhance the reactivity of the a-amino group by
`two orders of magnitude [25]. A P-site substrate containing
`a 20 deoxy A76 is inactive in peptide-bond formation [26],
`consistent with a possible role for that 20 OH in catalysis.
`Furthermore, mutation of A2486 (A2451 in E. coli) to a
`
`5
`
`

`

`416
`
`Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`uracil reduces the rate of the chemical step of peptide-bond
`formation by two orders of magnitude and removes the pH
`dependence with a pK of 7.5 [27], again consistent with the
`possibility that A2486 plays a small but significant role in
`catalysis.
`Further insights into the structural basis for the
`catalysis of peptide-bond formation will require two
`experiments (at least). First, a substrate complex contain-
`ing both A- and P-site substrates bound simultaneously
`will be essential to obtain a more precise view of their
`relative orientation and to see if there are any confor-
`mational changes in the centre produced by the presence of
`both. Second, it is imperative to have a structure for the
`70S ribosome complexed with A- and P-site tRNA
`substrates at a resolution high enough (3 A˚ ) to accurately
`and independently position all of the atoms at the site of
`peptide-bond formation. Presumably, such a structure will
`explain why the 70S ribosome exhibits a 103 to 104-fold
`higher rate of peptide-bond formation than does the 50S
`subunit [28].
`
`Inhibition of the peptidyl transferase reaction by
`antibiotics
`Microorganisms conduct a form of bacterial warfare by
`making small molecule compounds that inhibit or kill
`other bacteria. Many of these bactericidal compounds
`work by blocking protein synthesis, targeting either the
`large or the small ribosomal subunit. Although many of
`these antibiotics will inhibit protein synthesis in all three
`kingdoms, a few are specific for eubacterial protein
`synthesis and are, therefore, useful in treating bacterial
`diseases in humans and animals. Medicinal chemists have
`improved several natural antibiotics by further chemical
`modifications.
`The crystal structures of 12 complexes between the
`large ribosomal subunit and antibiotics have been deter-
`mined (Fig. 6), and they demonstrate at least two modes by
`which they inhibit protein synthesis [29–31]. One class of
`antibiotics called macrolides (e.g. erythromycin, tylosin
`and azythromycin) bind to a site in the proximal part of the
`polypeptide exit tunnel adjacent to the peptidyl-transfer-
`ase centre, and all appear to inhibit protein synthesis
`largely by blocking the passage of nascent polypeptide
`down the exit tunnel. This location is consistent with the
`observation that some macrolide antibiotics, such as
`erythromycin, enable the synthesis of di- or tri- peptides
`[32]. The macrolides tylosin, carbomycin and azythromy-
`cin bind to the H. marismortui 50S subunit [30] in the
`same location that erythromycin and other macrolides
`bind to the eubacterial D. radiodurans large subunit [29]
`
`Fig. 5. Structural insights into peptide-bond formation. (a) A space-filling represen-
`tation of the complex between the Haloarcula marismortui large subunit and three
`intact tRNAs added in the positions that the tRNAs assume when bound to the A, P
`and E sites of the 70S ribosome [5]. rRNA is white, and ribosomal proteins are yel-
`low. The subunit, which is oriented in the crown view, has been cut in half along a
`plane that passes through the peptide exit tunnel, and the front of the structure
`has been removed to expose the polypeptide exit tunnel, which is ,100 A˚
`long
`and 12– 20 A˚ wide. The active-site area is boxed. (b) A close-up of the active site
`showing the peptidyl product CC-puromycin-phenylalanine-caproic acid-biotin
`(CC-pmn-pcb; green) bound to the A-loop (tan), and the deacylated product (CCA;
`violet) bound to the P-loop (blue). The N3 of A2486 (A2451 in Escherichia coli; light
`blue) is close to the 30 OH of the CCA, and the base of U2620 (U2585 in E. coli; red)
`
`has moved close to the new peptide bond and the 30 OH of A76 [5]. (c) A model of
`the peptidyl transferase centre of the large ribosomal subunit from H. marismortui
`with substrates bound to both the A-site and P-site. This model was obtained by
`superimposing the structure of an A-site substrate complex (PDB code: 1FGO) on
`the structure of a P-site substrate complex (PDB code: 1M90) [6]. The a-amino roup
`of the A-site substrate (purple) is positioned for a pro-S attack on the carbonyl car-
`bon of the ester linking the peptide moiety of the P-site substrate (green). Possible
`hydrogen-bonding interactions involving the a-amino group and the N3 of A2486
`(A2451 in E. coli) and the 20 OH of A76 are indicated. The 20 OH of A2486 (A2451 in
`E. coli) is also close enough so that is might interact. Panels (a) and (b) reprinted,
`with permission, from Ref. [5] (http://www.nature.com/nsb/). Panel (c) reprinted,
`with permission, from Ref [6]. q (2003) National Academy of Sciences, U.S.A.
`
`http://tibs.trends.com
`
`6
`
`

`

`Review
`
`TRENDS in Biochemical Sciences Vol.28 No.8 August 2003
`
`417
`
`exploit these many and varied small molecule-binding
`sites in the peptidyl transferase centre. Using this
`structural information, it might be possible to design
`novel antibiotics that will be effective against presently
`known antibiotic-resistance mutations that arise in the
`50S subunit.
`
`Evolution
`The existence of a peptidyl transferase centre that consists
`entirely of RNA as well as the very high level of sequence
`conservation around the peptidyl transferase centre and
`the 30S interface implies that the first ribosome was
`composed entirely of RNA. The fact that eubacteria share
`only 20 large subunit proteins with eukaryotes and
`archaea lends further support to the hypothesis that
`ribosomal proteins were late additions to the ribosome.
`The RNA within a 30–40 A˚
`radius of the peptidyl
`transferase centre is not only highly conserved among
`all three kingdoms, but contains almost no globular
`protein domains and is largely penetrated only by protein
`tail extensions. In vitro evolution of RNA oligonucleotides
`can produce small RNA molecules that catalyze reactions
`related to peptidyl
`transferase reaction and bind
`analogues of peptide-synthesis intermediates [33,34],
`suggesting that the appearance of a small RNA capable
`of catalysing peptidyl transferase was a plausible first step
`in the evolution of the ribosome. Presumably, the first
`peptides synthesized by such a primordial peptide synthe-
`sizing RNA might have been random copolymers. Possibly
`the production of even random sequence, basic peptides
`reminiscent of the ribosomal protein tails might have been
`useful in stabilizing the structures of ribozymes in the
`RNA world.
`The nature of the steps involved in the evolution of a
`simple peptide-synthesizing RNA domain into two sub-
`units of RNA that are capable of messenger-directed
`protein synthesis is less obvious at this time. We can look
`forward to future experiments that might illuminate the
`development of messenger-directed synthesis.
`
`Acknowledgements
`We thank our colleagues past and present on the ribosome project: N. Ban,
`P. Nissen, J. Hansen, T.M. Schmeing, D. Klein and B. Freeborn. The
`research from our laboratories summarized in this review was supported
`by grants from the NIH and The Agouron Institute.
`
`References
`1 Cech, T. et al. (1981) In vitro splicing of the ribosomal RNA precursor of
`Tetrahymena: involve