A Mechanism for Stop Codon Recognition by the Ribosome: a Bioinformatic Approach

Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

HYPOTHESIS A mechanism for stop codon recognition by the ribosome: A bioinformatic approach

VALERY IVANOV,1,2 ARTEMY BENIAMINOV,1 ANDREI MIKHEYEV,2 and ELVIRA MINYAT1 1 V+A+ Engelhardt Institute of Molecular Biology, Russian Academy of Sciences, 119991 Moscow, Russia 2 Moscow Institute of Physics and Technology, 141700 Moscow Region, Dolgoprudny, Russia

ABSTRACT Protein synthesis in ribosomes requires two kinds of tRNAs: initiation and elongation. The former initiates the process (formylmethionine tRNA in prokaryotes and special methionine tRNA in eukaryotes). The latter participates in the synthesis proper, recognizing the sense codons. Synthesis is also assisted by special proteins: initiation, elongation, and termination factors. The termination factors are necessary to recognize stop codons (UAG, UGA, and UAA) and to release the complete protein chain from the elongation tRNA preceding a stop codon. No termination tRNA capable of recognizing stop codons by their anticodons is known. The termination factors are thought to do this. In the large ribosomal RNA, we found two sites that, like tRNAs, contain the anticodon hairpin but with triplets complementary to stop codons. One site is hairpin 69 from domain IV; the other site is hairpin 89, domain V. By analogy, we call them termination tRNAs: Ter-tRNA1 and Ter-tRNA2, respectively, even though they transport no amino acids, and suggest that they directly pair to stop codons. The termination factors only aid in this recognition, making it specific and reliable. A strong argument in favor of our hypothesis comes from vertebrate mitochondria. They are known to acquire two new stop codons, AGA and AGG. In the standard code, these are two out of six arginine codons. We revealed that the corresponding anticodons, UCU and CCU, have evolved in Ter-tRNA1 of these mitochondria. Keywords: genetic code; large rRNA; phylogeny; RNA world; termination; Ter-tRNA; translation; vertebrate mitochondria

INTRODUCTION: FOUNDATION OF THE There are two alternative opinions: (1) the RFs only HYPOTHESIS recognize stop codons, and (2) the triplets (anticodons) in the small rRNA directly interact with stop codons Termination is the final process of a ribosomal cycle (though with the help of RFs; Tate & Brown, 1992)+ The and begins with the recognition of a stop codon in the first opinion prevails and is strongly supported by the mRNA by the ribosome+ This is followed by hydrolysis finding that RF2 forms crosslinks with a piece of mRNA of an ester bond that connects the complete polypep- near the decoding site (Brown & Tate, 1994)+ Moreover, tide chain to the peptidyl tRNA, removal of the deacyl- the tripeptides in RF1 and RF2 were localized, and ated tRNA from the ribosome, and dissociation of the when mutated, affected specificity of stop codon rec- latter into small and large subunits+ Obligatory partici- ognition (Ito et al+, 2000; we shall discuss this important pants of the termination process are proteins called the article further in a special section)+ release factors (RFs)+ The eubacterial proteins RF1 The site of protein synthesis in the ribosome is formed and RF2 participate in recognition of stop codons: RF1 by several structural domains of the large rRNA+ Two of works with UAG and UAA, whereas RF2 traces UGA them, IV and V (see Fig+ 1) are discussed in this work+ and UAA (Scolnick et al+, 1968)+ A single eukaryotic We paid attention to hairpin 69 (domain IV), whose protein, eRF1, serves all the three codons: UAG, UGA, terminal loop consists of 7 nt+ It, like the anticodon loop and UAA (reviewed by Kisselev & Buckingham, 2000)+ of tRNAs, has the triplet (CUA) in the middle of the loop exactly as tRNA does+ In this case, it is for a stop codon Reprint requests to present address: Valery Ivanov, Center for , + + , , UAG rather than for an amino acid codon (Fig 1) Biophysical Sciences and Engineering University of Alabama Bir- , , mingham, CBSE 221, 1530 3rd Avenue South, Birmingham, Ala- Moreover hairpin 69 together with its close neigh- bama 35294-4400, USA; e-mail: irina@uab+edu+ bors, can be depicted as a clover leaf, typical for tRNAs 1683 Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

1684 V. Ivanov et al.

FIGURE 1. Domains IV and V of E. coli 23S rRNA+ Small numbers are the nucleotide positions+ Small bars mark every 10th nucleotide+ Large numbers designate relevant hairpins+ Stop anticodons in 7-nt loops of hairpins 69 and 89 are shown in a magnified view near the corresponding stop codons and release factors+ Three relevant crosslinks are shown by the slanted lines (Mueller et al+, 2000)+ Curved two-headed arrows connect the segments with possible stacking+ A zigzag line indicates the crosslink between the loop of hairpin 69 and 16S rRNA (decoding site)+ Wower’s crosslink (J+ Wower, pers+ comm+) is seen above hairpin 89+ A set of rectangular lines below the hairpin 89 is WC-pairing in a suggested pseudoknot (Ivanov et al+, 1999)+ For all the details see text+

(see Fig+ 2A)+ Here, it is a mimic of a tyrosine su(ϩ3) or Ter-tRNA1+ Interestingly, m2G, which is frequently Tyr-tRNA (Fig+ 2B)+ It is clear that the number of iden- met in the 10th position in tRNAs at the bottom of a tical nucleotides, having identical positions (except for DHU-stem (Saenger, 1984), is similarly disposed in the the CCA stem) is too great to be just a random coinci- Ter-tRNA1 (Fig+ 2A)+ dence+ Especially impressive is the similarity between There is a Ter-tRNA2 as well, which is found in do- the T⌿C-stem and its ribosomal mimic+ We termed main V (Fig+ 1)+ In this case, it is confined to hairpin 89, these combined four rRNA hairpins the termination tRNA whose 7-nt terminal loop imitates the tRNA anticodon Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

Stop codon recognition with large rRNA 1685

FIGURE 2. A clover leaf depiction for hairpins 68–71 of E. coli 23S rRNA (A) and su(ϩ3)Tyr-tRNA (B)+ Identical elements are shaded+ Below, see pairing of the stop codon with its anticodon+

loop containing the anticodon UCA for the stop codon 3+ Are stop anticodons present in hairpins 69 and 89 in UGA+ organisms of different taxons or not? We formulate the following hypothesis:

ARCHIVE INFORMATION Stop codons in mRNA are recognized by the stop anticodons of the tRNA mimics, Ter-tRNA1 and The data we used on the large rRNA sequences are Ter-tRNA2, covalently built into the large rRNA, available at the sites www+rna+icmb+utexas+edu and and termination proteins RF1 and RF2 promote http://rrna+uia+ac+be/lsu+ We have analyzed more than this recognition by increasing its accuracy and 500 sequences belonging to bacteria, eukaryotes, chlo- fidelity. roplasts, and mitochondria+ A search for the particular regions including hairpins 69 and 89 and their neigh- In this article, we restrict ourselves to the consider- borhood was facilitated by the fact that they are the ation of termination in bacteria and organelles, mito- most conserved regions in the large rRNA+ chondria, and chloroplasts+ These organelles are far Atom coordinates of the Haloarcula marismortui 50S descendants of bacteria that formerly inhabited eukary- subunit (Ban et al+, 2000) were obtained from the Pro- otic cells+ Presently we do not exclude that this concept tein Data Bank, accession number 1FFK+ The program is also true for eukaryota and archaebacteria+ How- Insight 2 for the workstation Silicon Graphics displayed ever, the protein mimics of RNA in eukaryota might go the three-dimensional structures, which are necessary so far to develop entirely “proteinic” recognition of stop for answering Crucial Question 1+ codons, so that this mechanism no longer exists in higher organisms+ ANSWERING THE CRUCIAL QUESTIONS 1+ Are hairpins 69 and 89 present near mRNA in the CRUCIAL QUESTIONS A-site, at the interface between the small and large Prior to specifying the announced hypothesis and in- ribosomal subunits, or not? specting its predictions, three crucial questions must be answered: The data obtained by crosslinking summarized in Muel- ler et al+ (2000; see also Fig+ 1) help to answer this 1+ Are hairpins 69 and 89 present near mRNA in the question regarding hairpin 69 with optimism: 23S rRNA A-site, at the interface between the small and large hairpin 69 (nt 1911–1920) crosslinks to 16S rRNA hair- ribosomal subunits, or not? pin 44 (nt 1408–1411)+ In addition, according to the 2+ Can the noncanonical pair A-C be formed, when data of X-ray studies (Cate et al+, 1999; Yusupov et al+, necessary, during pairing of stop codons and stop 2001), the loop of the hairpin 69 makes direct contact anticodons or not? (Such a pair is necessary for the with the decoding site in the small subunit (hairpin 44, anticodons of hairpins 69 and 89 to pair with the nt 1494 and 1408–1410)+ Unfortunately, the X-ray data UAA stop codon+) from the large subunit (Ban et al+, 2000) does not show Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

1686 V. Ivanov et al. the loop of the hairpin 69+ In this respect, the recent CmCA inherent to Trp-tRNA+ In a test of binding to a paper by Yusupov et al+ (2001) is especially important+ ribosome, this tRNA binds to a stop codon UGA, but It confirms cryo-electron microscopy data on the close not to serine or tryptophan codons+ In vitro experi- proximity between the loop of helix 69 and the anti- ments confirmed that this tRNA exhibits suppressor ac- codon hairpin of A-site tRNA, but also testifies to the tivity toward the UGA codon+ In this connection, one agile nature of hairpin 69+ Because the secondary struc- can be reminded that a suppressor Trp-tRNA from ture of rRNA is universal for different species of organ- E. coli (Hirsh & Gold, 1971) also reads UGA codon and isms in the region of hairpins 68–71 and 89–92 (Gutell, has similar CCA anticodon, as well as the modified (by 1996), the answer to the above question should be a point mutation) D-stem+ Thus, some change in the universal as well+ Therefore, concerning hairpin 69, the position or mobility of the anticodon hairpin, as well as answer is affirmative+ modification of a nucleotide of the anticodon, can lead Regarding hairpin 89, the situation is more complex+ to reading of the UGA codon by the CCA anticodon of As it follows from the coordinates of atoms of the large tryptophan+ One more relevant illustration is the fact subunit (Ban et al+, 2000), a turn of about 408 is nec- that m5C pairs with A in the wobble position when de- essary for hairpin 89 to reach the required position+ coding UGA as selenocysteine by a special Sec-tRNA This is also true for the 70S ribosome (Gabashvili et al+, (Hatfield & Diamond, 1993)+ 2000; Yusupov et al+, 2001)+ A phylogenetically con- It follows from all the above observations that the served hinge near the bottom of hairpin 89 (see Fig+ 1) recognition CCA/UGA is realized in some codon– may be of interest in this respect+ There is another anticodon pairs, with various mechanisms being in- theoretical argument for the ability of hairpin 89 to make volved in the A@C interaction+ a significant turn, namely, the possibility of the forma- A comment is required on the modification in E. coli’s tion of a transient pseudoknot between the terminal middle base of the CUA stop anticodon (hairpin 69)+ loop of hairpin 89 and an adjacent part of the peptidyl- This nucleotide is 3-methyl-⌿U (Merouch et al+, 2000) transferase ring (Ivanov et al+, 1999), indicating the po- in most prokaryotes+ Its grouping 3 r N-CH3 does not tential mobility of the hairpin 89 (see Fig+ 1)+ Lastly, prevent Watson–Crick (WC) pairing, as a 1808 rotation گ Ϫ there are experimental data that testify to a change of of the base around the glycosidic bond brings 5 / N-H , + گ Ϫ , the position of hairpin 89 depending upon the func- group in place of 3 / N-CH3 However the modified U tional state of a ribosome+ “The crosslink between U2474 is incapable of pairing with G (according to the wobble and G2502/A2503 (Escherichia coli numbers) forms scheme)+ The tempting suggestion is that such a mod- upon direct UV irradiation of tRNA-70S ribosome com- ification serves exactly to prevent the Ter-tRNA1 anti- plexes with tRNA in 20 mM Mg (Fig+ 1)+ Under these codon from recognizing UGA which is recognized by conditions, you can see some of U2474 crosslinking to Ter-tRNA2+ protein L6+ When the same crosslinking was carried out on 70S ribosomes without tRNA or on 50S ribo- 3+ Are stop anticodons present in hairpins 69 and 89 in somal subunits (this time in 10 mM Mg) only crosslink- organisms of different taxons or not? ing to protein L6 was observed” (J+ Wower, pers+ comm+)+ A summary of the conservation of sequences in the 2+ Can the non-canonical A-C pair be formed, when loops of hairpins 69 and 89, including the middle triplet, necessary, during pairing of stop codons and stop is presented in Table 1 for different life lineages+ The anticodons, or not? Such a pair is necessary for the main peculiarity of the table is seen at once: The nu- anticodons of hairpins 69 and 89 to pair with the cleotide occurrence in the stop anticodon is far from UAA stop codon: random+ It is either close to 100% and 0%, or roughly divided equally between 2 or 3 nt+ (The last case is for 59-CUA and 59-UCA the first position of the anticodon in loop 89 of archaea, @|| |@| where only G is forbidden+) AAU-59 AAU-59 This pattern can hardly be explained by the model of direct stop codon recognition by protein+ It is no busi- We use the @ sign to show that a third participant, ness of an RF in such a model to check the sequences Release Factor, might be involved in helping the A@C of “stop anticodon+” Nevertheless, the pattern of abun- interaction, for example, by A or C protonation resulting dance calls for explanation+ Indeed, one can see from in a wobble A ϩ C pair+ By X-ray analysis, the proton- Table 1 that in loop 89, the stop anticodon 59-UCA may ated A ϩ C pairs were shown to exist in ribo-duplexes+ be substituted with a triplet 59-CCA, which is an anti- They were isogeometric with the wobble G-U pair (Jang codon for the tryptophan codon 59-UGG+ It is especially et al+, 1998)+ evident in eubacteria, where the “wrong” anticodon oc- In support of the A@C interaction, a minor tRNA from curs in half of the cases+ Again, the A@C interaction, beef liver can be mentioned (Diamond et al+, 1981)+ whose existence was demonstrated above, can ex- This tRNA accepts serine, while having an anticodon plain this observation+ Another question then arises: Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

Stop codon recognition with large rRNA 1687

TABLE 1+ Abundances of each nucleotide types (percentage) in every position (1 to 7) from 59 end of hairpins 69 and 89’s loops in different life domains and chloroplasts+

Eubacteria Chloroplast Eukarya Archaea

ACGU A C G U ACGUACG U

Loop 69 1 100 0+00+00+0 100 0+00+00+00 100 0+00+00+0 100 0+00+00+0 298+30+01+60+099+10+00+00+9 100 0+00+00+0 100 0+00+00+0 30+0 100 0+00+00+0 100 0+00+00 0+099+30+00+70+0 100 0+00+0 40+40+00+099+60+00+00 0+0 100 0+00+00+799+30+09+80+090+2 599+20+80+00+00 97+03+00+00+00 99+30+70+00+048+851+20+00+00 60+00+00+0 100 0+00+00+0 100 0+00+00+0 100 0+00+00+0 100 793+10+06+90+086+10+013+90+00 0+01+398+70+022+00+078+00+0 Loop 89 10+00+0 100 0+00+00+0 100 0+00+00+0 100 0+00+00+0 100 0+0 20+87+00+092+20+95+90+093+24+45+80+789+10+085+40+014+6 30+045+90+054+10+023+60+076+40+05+80+094+226+836+60+036+6 40+0 100 0+00+00+0 100 0+00+00+098+60+01+40+0 100 0+00+0 5 100 0+00+00+098+00+02+00+094+91+41+42+263+436+60+00+0 60+049+20+050+80+083+30+016+71+51+50+097+00+048+80+051+2 7 100 0+00+00+0 100 0+00+00+0 100 0+00+00+0 100 0+00+00+0

Stop anticodons are present in the position (3,4,5)+ The data for Archae and Eukarya are given for completeness+ They are not discussed in text for the absence of experimental data comparable to those in prokaryota+

why does the Ter-tRNA2 with a CCA anticodon not data for such an assignment? Among others, mutation recognize the tryptophan codon in mRNA leading to data cannot be avoided+ abortive termination? The answer is that the process of codon recognition in the elongation stage is much faster than in the termination stage (reviewed in Parker, 1989), Mutations in hairpins 69 and 89 so that Trp-tRNA will be a winner in competition with O’Connor and Dahlberg (1995) obtained a number of Ter-tRNA, just like the suppressor tRNAs+ natural and artificial mutations exactly in hairpins 69 As to eukarya and archae, which use only one RF, it and 89 and checked how these mutations promote read- is not clear which hairpin it uses, 69 or 89, if any+ It may through of stop codons in the lac operon of E. coli+ Note well be that these life domains had evolved the RF with that increase in readthrough over a control indicates a the ability to directly recognize stop codons+ defect in termination+ The authors have demonstrated Here, a general statement is appropriate+ As to the that the mutations in each of those hairpins, if effective, elongation stage, the requirement of isostericity at will usually suppress termination at both UGA and UAG each step (triplet by triplet) is very strict: Devia- in one stroke+ This is understandable, because any tions from WC pairing are only permitted within the defect in either structure could slow the entire process, frames of a wobble scheme (or the extended wob- thereby increasing the probability of readthrough+ Yet, ble scheme for mitochondria). There is no reason one of the mutations in hairpin 69, C1921, although to think that the termination stage, which com- strongly suppressive in termination at UAG, is abso- pletes polypeptide synthesis, must strictly follow lutely silent in suppressing termination at UGA+ Of sim- the same pairing scheme that elongation does. For ilar, though smaller, selectivity is the mutation U1914 example, a quadruplet code for termination of transla- (Table 2)+ We can only speculate as to why the mutation will be discussed further+ tion A1914 is of the wild-type phenotype (Table 2)+ It is not improbable that RF1 or modifications in the loop can differently stabilize GU, AA, and GA pairs in the ASSIGNMENT OF RF1 AND RF2 TO THE wobble position+ Table 2 shows that, generally, the mu- CORRESPONDING TER-tRNAs tations in hairpin 69 suppress stronger termination at Phylogenetic analysis has shown that 7-nt loops in hair- UAG (for one exception), whereas mutations in hairpin pins 69 and 89 contain anticodons for stop codons 89 suppress stronger termination at UGA (one excep- UAG and UGA, respectively+ Each of the stop anti- tion)+ Because RF1 works with UAG, and RF2 works codons can also potentially recognize UAA with the with UGA, the above experimental data are not con- allowance of the A@C pairing+ Are there experimental trary to our assignment, based upon phylogenetic analy- Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

1688 V. Ivanov et al.

TABLE 2+ Effects of mutations in hairpins 69 and 89 of rRNA on stop REGARDING THE PAPER BY ITO ET AL. , + codon readthrough in E. coli (O’Connor & Dahlberg 1995) (2000): “A TRIPEPTIDE ‘ANTICODON’ b-galactosidase activity at DECIPHERS STOP CODONS IN MESSENGER RNA” Hairpin Mutation UGA UAG This important recent paper concerns the same prob- Wild type 24 27 lem considered here: how does a ribosome read stop A1914 21 21 U1914 33 55 codons? The authors have proved that certain tripep- 69 ϪA1916 86 79 tides of RF1 and RF2 respond specifically to stop co- ϩAA1916 49 76 dons+ In particular, the tripeptides discriminate G against C1921 25 45 A in the stop codons, supposedly by 2-amino group of + , , ϩG2457 78 54 G However the title of the above paper as well as the A2458/C2493 93 71 main conclusion that “the discriminator tripeptide of bac- C2458/C2493 118 78 terial release factors is functionally equivalent to the A2493 127 96 anticodon of transfer RNA” can make one think that C2493 182 140 only protein recognizes stop codons+ While discussing ϪU2493 100 103 the paper by Ito et al+ (2000), we emphasize that it is 89 C2492 64 50 quite possible for RF1 and RF2 to recognize the RNA– A2460 76 55 RNA duplex of the stop codon with the anticodon of G2460/A2490 81 63 Ter-tRNA, rather than the stop codon alone+ Experi- / G2460 U2490 67 56 mental data from the paper under discussion agrees / G2460 C2490 80 54 with both models+ However, the model with Ter-tRNA C2477 25 25 has an important advantage; it does not require the A2477 25 25 search for a special amino acid residue responsible for recognition of the universal U in the first position for all the stop codons+ U finds its natural counterpart A as the nearly universal base in stop anticodons of the loops in hairpins 69 and 89 (Table 1)+ Moreover, the discrimi-

nator role of the NH2-group of G becomes more understandable in the model with recognition by RF of an RNA sis: Ter-tRNA1 (hairpin 69) works with RF1, whereas duplex+ Indeed, the situation is similar to recognition of Ter-tRNA2 (hairpin 89) works with RF2+ a DNA duplex by the peptide antibiotics of distamycin or

netropsin, in which the above NH2-group of G in the mi- A further argument for participation of nor groove also plays a discriminator role (Zimmer & , + Ter-tRNA2 (hairpin 89) in the Wahnert 1986) The rigid structure of the RNA duplex recognition UGA can be an essential factor responsible for the remark- able selectivity of RFs+ This selectivity is quantitatively Arkov et al+ (1998) spotted a special fragment of 23s discussed in the study by Nakamura et al+ (2000)+ rRNA of E. coli within domain I (nt 74 to 136), which, It must be stressed that the idea of stop codon rec- when added, increased readthrough of UGA in vivo+ ognition by a complementary triplet (in small rRNA) has Below, we demonstrate that a decanucleotide from the repeatedly been suggested and discussed (reviewed in above fragment can form a perfect 9-nt duplex plus Tate & Brown, 1992)+ There are also data on the muta- one GU pair with the terminal part of hairpin 89, includ- tions in the large rRNA that affect accuracy of termina- ing the full 7-nt loop: tion (O’Connor & Dahlberg, 1995; Murgola et al+, 2000)+ Unlike others, (1) we suggest a model for stop codon 59-UGAUAUGAAC (decanucleotide 87–96 from a deciphering by the large rRNA rather than the small fragment 74–136) rRNA, and (2) we not only indicate the locations of stop GCUAUACUUG-59 (terminal part of hairpin 89, nt anticodons for each release factor but also call atten- 2463–2472) tion to the finding that the stop anticodons are part of the real anticodon loops in hairpins 69 and 89, which belong to tRNA mimics and are incorporated into the Underlined is the loop with the UCA stop anticodon in large rRNA+ the middle (cf+ Fig+ 1)+ Formation of this duplex, if it occurs, must block recognition of the UGA stop codon by the stop anticodon in the loop of hairpin 89 and MITOCHONDRIA RUN EXPERIMENTUM CRUCIS therefore to affect termination+ This result additionally supports the role of hairpin 89 in recognition of UGA in Is it possible, starting from known facts, to choose be- termination of translation in eubacteria+ tween a model with direct recognition of stop codon by Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

Stop codon recognition with large rRNA 1689 protein and the model of recognition by an anticodon the ribosome of vertebrate mitochondria, the question present in a Ter-tRNA (though with assistance of RF)? naturally arises about the location of the anticodon within Yes, it is+ Assume that the genetic code for some spe- the loop+ The answer comes from mitochondrial tRNAs+ cies of living organisms has acquired a new stop co- At least two cases are known where a tRNA has the 6-nt don, that is, the one that is absent as a stop codon in anticodon loop: mt Tyr-tRNA (Cyprinus carpio) and mt the universal code+ If our hypothesis is correct, then a Asp-tRNA (Balaenoptera musculus): corresponding anticodon must appear in the loops of hairpins 69 or 89+ On the other hand, if an alternative 59- UGGG[UUGUAG]CCCG -39 hypothesis of direct deciphering of the stop codon by 59-UAACU[UUGUCA]AGUUA-39 RF solely is correct, then no corresponding anticodon will appear in these loops+ Although such organisms Because the specificity of these tRNAs is known, it are unknown, there are mitochondria that have a ge- addresses the position of an anticodon+ Here, anti- netic code different from the standard one+ Thus, an codon loops are shown in brackets, anticodons are overwhelming majority of species have mitochondria bold+ One can see that these tRNAs contain anti- that use UGA as a codon for tryptophan rather than for codons in a position from the third to fifth nucleotide a stop codon (reviewed in Osawa et al+, 1992)+ More from 59 end of the 6-nt loop+ interesting for our concept is a case of human mito- Now, going back to the Ter-tRNA1 with the suppos- chondria (and the majority of other vertebrates) that edly terminator hairpin 69, several dozen mitochondrial acquire two new stop codons: AGA and AGG (Osawa sequences of vertebrates have been analyzed, some et al+, 1992)+ The stop codons correspond to a pair from of which are listed in Table 3+ the six arginine codons of the standard genetic code+ We arrived at the conclusion that vertebrate mitochondria, except for those of fish, have acquired the CCU or UCU stop anticodons (for the new AGG A peculiarity of the large rRNA structure in and AGA stop codons), which are disposed just as mitochondria of vertebrates anticodons of the tRNAs with 6-nt loops are. Note, Which hairpin, 69 or 89, might be expected to acquire allowing for A@C interaction, the CCU can accommo- anticodons UCU and CCU corresponding to the new date AGA stop codon in addition to AGG+ stop codons AGA and AGG? Hairpin 89 can be ex- And where are anticodons for the old stop codons, cluded, as its stop codon UGA had been donated to UAG and UAA, in the hairpin 69 of vertebrate mito- tryptophan (except of plant mitochondria, which keep chondria? They are known to stay in use+ Table 3 the stop UGA)+ The disappearance of UGA as a stop shows that triplets CUA or CUG are universally present codon is accompanied by the loss of RF2 itself (Lee at the 39 end of the loop and complementary to the et al+, 1987)+ Thus, only hairpin 69 matters in this case+ old UAG stop codon (CUG must form a wobble GU Unexpectedly, a complication arises: mitochondria of pair)+ Pairing to UAA by triplets CUA and CUG re- vertebrates have hairpin 69 with a 6-nt loop instead of quires the A@C interaction again+ Remarkably, the the 7 nt present in the standard anticodon loop+ The 7-nt new and old anticodons are overlapped by 2 nt, which anticodon loop in tRNA contains its anticodon just in the makes it possible for a quadruplet termination code middle+ With the 6-nt anticodon loop of the Ter-tRNA1 in in most cases (Tables 3 and 4)+

TABLE 3+ Examples of mitochondrial rRNA sequences of hairpin 69 of vertebrates and stop codon usage in the corresponding genomes+

Number Organism Hairpin 69 UAA UAG AGA AGG of genes

Mammals Homo sapiens cgcggu[a-cccua]accgug 9 3 1 0 13 Mus musculus cgcggu[a-uccug]accgug 10 2 0 1 13 Balaenoptera musculus cgcggu[a-uccug]accgug 11 1 1 0 13 Birds Corvus frugilegus cgcggu[a-uccua]accgug 11 1 0 1 13 Struthio camelus cgcggu[a-uccua]accgug 7 2 2 2 13 Reptile Chrysemys picta cgcggu[a-uccua]accgug 7 4 0 2 13 Amphibium Xenopus laevis cgcggu[a--ucug]accgug 8 3 2 0 13 Fish Latimeria chalumnae cgcggu[a-uccug]accgug 11 1 1 0 13 Cyprinus carpio cgcggu[uauuuug]accgug 11 2 0 0 13 Oncorhynchus mykiss cgcggu[a-uuuug]accgug 11 2 0 0 13

Hairpins’ loops are in brackets+ New stop anticodons are underlined+ Overlapping new and old stop anticodons are shown in bold+ Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

1690 V. Ivanov et al.

TABLE 4+ Occurrence of different stop codons in mt DNA of selected vertebrates+

Mammals Aves Fish Amphibia Reptiles

Homo Balaenoptera Mus Corvus Strutio Cyprinos Oncorinchus Latimeria Xenopus Chrisemys Alligator Gene sapiens musculus misculus frugilegus camelus carpio mykiss chalumnae laevis picta mississip.

ND1 UAA UAA UAGaa AGG AGG UAA UAGcc UAA UAGga UAGga UAGga ND2 UAGaa UAGga UAGaa UAA UAGaa UAGgg UAA UAA UAGag UAGaa UAA COX1 AGA UAA UAA AGG AGG UAA UAA UAA UϩpolyA AGG UAUggcc* COX2 UAGca UAA UAA UAA UϩpolyA UϩpolyA UϩpolyA UϩpolyA UϩpolyA UAA UϩpolyA ATPase8 UAGgc UAA UAA UAA UAA UAGua UAA UAA UAA UAA UAA ATPase6 UAA UAA UAA UAA UAA UAA UAA UAA UAA UAA UAA COX3 UϩpolyA UϩpolyA UAA UϩpolyA UϩpolyA UAA UAA UAA UϩpolyA UϩpolyA UϩpolyA ND3 UϩpolyA UAA UAA UAA AGA UAGgg UϩpolyA UϩpolyA UϩpolyA UAGgu UAA ND4L UAA UAA UAA UAA UAA UAA UAA UAA AGA UAA UAA ND4 UϩpolyA UϩpolyA AGG UϩpolyA UAA UϩpolyA UϩpolyA UϩpolyA UϩpolyA UϩpolyA UAA ND5 UϩpolyA UAA UAA AGA AGA UAA UAA UAA UAA UAGac UAA ND6 UAA UAA UAA UAA UAGgg UAA UAGgu AGA AGA AGG AGG Cyt b UϩpolyA AGA UAA UAA UAA UϩpolyA UϩpolyA UAGcc UAGuc UAA UϩpolyA

*This amazing termination signal UAU is discussed in text+

Stereochemistry of the 6-nt loop determines favorable for stop codons’ binding+ It avoids a kink within disposition of stop anticodons the codon–anticodon duplex in the 6-nt loop, even if it would have four stacked base pairs with two overlap- The standard 7-nt anticodon loop in tRNAs has two ping stop triplets (Fig+ 3)+ Note, it is the FH conformer favorable conformations: HF and FH (Fuller & Hodg- that has been supposed for the tRNA in the A-site of a son, 1967)+ The FH structure has an anticodon with ribosome, that is, exactly in the site, where a stop co- stacking to the 39 end of the loop, whereas the HF don is to be recognized+ contains the anticodon stacked to the 59 end of the loop+ In the 6-nt loop of hairpin 69 from mitochondria of , vertebrates the 4 nt are used by two overlapping stop SPECIAL CASES anticodons: the old (4,5,6) and new (3,4,5) (see Fig+ 3)+ Positions (1, 2) from the 59 end of the loop, are en- “As usual, interpreting ribosomal function is never gaged by the nonparticipating nucleotides in anticodon straightforward” (Porse & Garrett, 1999)+ Some organ- pairing+ Therefore, it is the FH-conformation that is most isms invent diverse versions of a general theme+

The special case of fish The mitochondrial genetic code of fish generally differs from that of other vertebrates+ It does not include the new AGG and AGA stop codons, but preserves UAG and UAA (UGA codes for tryptophan)+ Therefore, it is interesting to see how the absence of stop codons AGG and AGA manifest itself in hairpin 69 sequences+ Table 3 illustrates that the 6-nt loops of fish may differ from the loops of other vertebrates+ Instead of the CCU or UCU anticodons recognizing stop codons AGG and AGA, the corresponding fragment in mitochondria of fish looks like UUU (rarely AUU)+ As to the old UAA and UAG codons, a triplet, which is positioned at (4,5,6), as is in the case in other vertebrates, can recognize them+ A single exception that we found is an ancient fish, Lat- imeria chalumnae, the close ancestor of amphibia+ It, like mitochondria of the genuine amphibium Xenopus lae- , + FIGURE 3. Three-dimensional presentation of a 6-nt loop that shows vis uses the new stop codon AGA (Tables 3 and 4) Re- disposition of the overlapping stop anticodons in the loop of hairpin markably, loop 69 in the mitochondria of Latimeria looks 69 in the large rRNA from vertebrate mitochondria+ Clearly, a non- like loops of hairpins 69 in the nonfish vertebrates with kinked 4-bp duplex with the stop codons is only possible for “new” anticodon positioned at (3,4,5) and the “old” anticodon (4,5,6,) stack- the possibility of a quadruplet anticodon–codon pairing ing at the 39 side (FH conformation)+ scheme: 59-CCUG/59-UAGA (with an A@C pair)+ Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

Stop codon recognition with large rRNA 1691

The special case of alligator pairs+ Importantly, it is shown in NMR experiments that the 7-nt anticodon loop is capable of pairing the penta- Like a majority of vertebrates, reptiles have the 6-nt anti- nucleotides (Geerdes et al+, 1980)+ Stereochemically, codon loop in the mitochondrial Ter-tRNA1 with the same this is possible in a 7-nt loop with 5-bp stacking at one typical sequence as exemplified in Figure 4 by Sphe- end of the loop with 2 nt looped out at the other end nodon punctatus. However, in alligators, this loop has 7, (Fuller & Hodgson, 1967)+ but not 6 nt+ Moreover, the loop contains no triplets for This is one more example that deciphering the ge- stop codons, either old or new, whereas the stem se- netic code in termination can be much more flexible quences of hairpin 69 are identical in both reptiles (Fig+ 4, than anticipated+ In addition to the large-scale usage of Alligator mississippiensis)+ Therefore, we turned to hair- A-C pairs, the stop codon–anticodon duplex may ex- pin 89 (Ter-tRNA2)+ Figure 4 shows that its loop con- pand beyond the limits of a triplet scheme and even be tains a quadruplet CCUA for the overlapping stop codons a pentaplet+ UAGG at the 39 end, just like other reptiles do in hairpin No doubt, the number of the diverse cases is not 69, rather than 89+ We do not know why alligator limited to those that we found+ switched the decoding hairpin+ Another strange thing about termination in the alligator’s mitochondria is that the translation of mRNA from one of alligator’s mito- CONCLUSION chondrial genes, COX1, stops at a triplet UAU (Janke & The evidence for existence of Ter-tRNA mimics, which Arnason, 1997), which is a sense codon for tyrosine (see are covalently incorporated into the large rRNA, sup- Table 4)+ Why does UAU play a role of stop codon? ports the concept of the “RNA world” by demonstration Above, the quadruplet stop codons were discussed+ of how early preribosomes would have recognized stop Extending UAU to quartet, even to quintet UAUGG in codons before release factors evolved+ In this connec- the mitochondrial COX1 mRNA, helps one to answer tion, we want to mention the studies (reviewed by Spi- this question+ Figure 4 shows that the quintet codon is rin, 1985) in which it was shown by experiment that the possible in this case and that the presence of a C-U ribosome under special conditions is capable of work- mismatch is compensated by two adjacent, strong G-C ing without translational factors+ Indeed, the protein factors only catalyze the ribosome work making it fast, faultless, and controllable+ Our observation that acquir- ing new stop codons by mitochondria of vertebrates works in parallel with the emergence of corresponding anticodons in the Ter-tRNAs shows that the RNA world is “alive” even now+

ACKNOWLEDGMENTS The authors thank Lev Kisselev, Alexey Bogdanov, Ed Tri- fonov, Maxim Frank-Kamenetskii, Victor Zhurkin, Nick Uly- anov, Alex Mankin, Bob Jernigan, Albert Dahlberg, Robert Zimmerman, Mans Ehrenberg, and Steve Harvey for discus- sions, and Jacek Wower, who acquainted us with his unpub- lished data on crosslinking to hairpin 89+ We are also thankful to Bob Jernigan and Jake Maizel for support+ V+I+I+ is especially grateful to Steve Harvey, who provided the opportunity to work in the stimulating atmosphere of his group at the University of Alabama, Birmingham, and to Jason Mears, who corrected our English usage+ This work was supported in part by the Russian Fund for Basic Research, Grant No+ 01-04- 48670, State Support for Leading Scientific Schools, Grant No+ 00-15-97837 (Russia), Grant EIMB RAS-200, and Grant Fogarty 1R03 TW01465-01+

Received January 18, 2001; returned for revision March 9, 2001; revised manuscript received October 1, 2001

REFERENCES

Arkov AL, Mankin A, Murgola EJ+ 1998+ An rRNA fragment and its FIGURE 4. The special case of mt rRNA of alligator+ (Explanation in antisense can alter decoding genetic information+ J Bacteriol text+) 180:2744–2748+ Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

1692 V. Ivanov et al.

Ban N, Nissen P, Hansen J, Moore PB, Steitz TA+ 2000+ The com- Ch+ 2000+ Unique structure and stabilizing roles of the individual plete atomic structure of the large ribosomal subunit at 2+4Å pseudouridine residues in the 1920 region of E. coli 23S rRNA+ resolution+ Science 289:905–920+ Nucleic Acids Res 28:2075–2083+ Brown CM, Tate WP+ 1994+ Direct recognition of mRNA stop signals Mueller F, Sommer I, Baranov P, Matadeen R, Stoldt M, Wohnert J, by Escherichia coli polypeptide chain release factor+ J Biol Chem Gorlach M, van Heel M, Brimacomb R+ 2000+ The 3D arrange- 269:33164–33170+ ment of the 23 S and 5 S rRNA in the Escherichia coli 50 S Cate JH, Yusupov MM, Yusupova GZh, Earnest TN, Noller HF+ 1999+ ribosomal subunit based on a cryo-electron microscopic recon- X-ray crystal structure of 70S ribosomal functional complexes+ struction at 7+5 Å resolution+ J Mol Biol 298:35–59+ Science 285:2095–2104+ Murgola E, Arkov AL, Chernyaeva NS, Hedenstierna KOF, Pagel FT+ Diamond A, Dudock B, Hatfield D+ 1981+ Structure and properties of 2000+ rRNA functional sites and structures for peptide chain ter- a bovine liver UGA suppressor serine tRNA with a tryptophan mination+ In: Garret RA, Douthwait SR, Liljas A, Matheson AT, anticodon+ Cell 25:497–506+ Moor PB, Noller HF, eds+ The ribosome: Structure, function, anti- Fuller W, Hodgson A+ 1967+ Conformation of the anti-codon loop in biotics, and cellular interactions. Washington, DC: ASM Press+ tRNA+ Nature 215:817–821+ pp 509–518+ Gabashvili IS, Agrawal RK, Spahn ChMT, Grassucci RA, Svergun DI, Nakamura Y, Ito K, Ehrenberg M+ 2000+ Mimicry grasps reality in Frank J, Penszek P+ 2000+ Solution structure of the E. coli 70S translation termination+ Cell 101:349–352+ ribosome at 11+5 Å resolution+ Cell 100:537–549+ O’Connor M, Dahlberg AE+ 1995+ The involvement of two distinct Geerdes HAM, van Boom JH, Hilbers CW+ 1980+ Nuclear magnetic regions of 23 S ribosomal RNA in tRNA selection+ J Mol Biol resonance studies of codon-anticodon interaction in tRNA Phe+ I+ 254:838–847+ Effect of binding complementary tetra and pentanucleotides to Osawa S, Jukes TH, Watanabe K, Muto A+ 1992+ Recent evidence for the anticodon+ J Mol Biol 142:195–218+ evolution of the genetic code+ Microbiol Rev 56:229–264+ Hatfield D, Diamond A+ 1993+ UGA: A split personality of the universal Parker J+ 1989+ Errors and alternatives in reading the universal ge- genetic code+ Trends Genet 9:69–70+ netic code+ Microbiol Rev 53:273–298+ Hirsh D, Gold L+ 1971+ Translation of the UGA triplet in vitro by Porse BT, Garrett RA+ 1999+ Ribosomal mechanics, antibiotics, and tryptophan transfer RNAs+ J Mol Biol 58:459–468+ GTP hydrolysis+ Cell 97:423–426+ Ito K, Uno M, Nakamura Y+ 2000+ A tripeptide ‘anticodon’ deciphers Saenger W+ 1984+ In: Cantor CR, ed+ Principles of nucleic acid struc- stop codons in messenger RNA+ Nature 403:680–684+ ture+ New York: Springer-Verlag+ pp 331–349+ Ivanov VI, Bondarenko SA, Zdobnov EM, Beniaminov AD, Minyat Scolnick EM, Tompkins R, Caskey CT, Nirenberg M+ 1968+ Release EE, Ulyanov NB+ 1999+ A pseudoknot-compatible universal site is factor differing in specificity for terminator codons+ Proc Natl Acad located in the large ribosomal RNA in the peptidyltransferase Sci USA 61:768–774+ center+ FEBS Lett 446:60–64+ Spirin AS+ 1985+ Ribosomal translocation: Facts and models+ Prog Jang SB, Hung LW, Chi YI, Holbrook EL, Carter RJ, Holbrook SR+ Nucleic Acids Res Mol Biol 32:75–114+ 1998+ Structure of an RNA internal loop consisting of tandem Tate WP, Brown ChM+ 1992+ Translation termination: “Stop” for pro- C-Aϩ base pairs+ Biochemistry 37:11726–11731+ tein synthesis or “pause” for regulation of gene expression+ Bio- Janke A, Arnason U+ 1997+ The complete mitochondrial genome of chemistry 31:2443–2450+ Alligator mississippiensis and the separation between recent ar- Yusupov MM, Yusupova GZh, Baucom A, Liberman K, Earnest ThH, chosauria (birds and crocodiles)+ Mol Biol Evol 14:1266–l272+ Cate JH, Noller HF+ 2001+ Crystal structure of the ribosome at Kisselev LL, Buckingham RH+ 2000+ Translational termination comes 5+5Å+Science 292:883–896+ to age+ Trends Biochem Sci 25:561–566+ Zimmer C, Wahnert U+ 1986+ Non-intercalating DNA binding ligands: Lee CC, Timms KM, Trotman CN, Tate W+ 1987+ Isolation of a rat Specificity of the interaction and their use as tools in biophysical, mitochondrial release factor+ Accommodation of the changed ge- biochemical and biological investigation of the genetic material+ netic code for termination+ J Biol Chem 262:3548–3552+ Progr Biophys Mol Biol 47:31–112+ Merouch M, Gronar PJ, Qiu J, SantaLucia J Jr, Scaringe SA, Chiu Downloaded from rnajournal.cshlp.org on October 2, 2021 - Published by Cold Spring Harbor Laboratory Press

A mechanism for stop codon recognition by the ribosome: a bioinformatic approach.

V Ivanov, A Beniaminov, A Mikheyev, et al.

RNA 2001 7: 1683-1692

License

Email Alerting Receive free email alerts when new articles cite this article - sign up in the box at the Service top right corner of the article or click here.

To subscribe to RNA go to: http://rnajournal.cshlp.org/subscriptions