Molecular Cell, Vol. 13, 157–168, January 30, 2004, Copyright 2004 by Cell Press

Reprogrammed Genetic Decoding in Cellular Gene Expression Olivier Namy,1,* Jean-Pierre Rousset,2 Sawsan Napthine,1 and Ian Brierley1 1 Division of Virology Department of Pathology University of Cambridge Tennis Court Road Cambridge, CB2 1QP United Kingdom 2 Institut de Genetique et Microbiologie Laboratoire de Genetique Moleculaire de la Traduction Universite Paris-Sud Baˆt 400, 91405 ORSAY cedex France

Reprogrammed genetic decoding signals in mRNAs productively overwrite the normal decoding rules of translation. These “recoding” signals are associated with sites of programmed ribosomal frameshifting, hopping, termination codon suppression, and the incorporation of the unusual amino acids selenocysteine and pyrrolysine. This review summarizes current knowledge of the structure and function of recoding signals in cellular genes, the biological importance of recoding in gene regulation, and ways to identify new recoded genes. Introduction Postgenomic studies constitute one of the current challenges in biology due to the burgeoning quantity of genome sequence available and the difficulties in utilizing the data to understand how an organism functions. A major problem is to couple the DNA sequences with the battery of proteins a given organism can express. One step relates to the decoding of genetic information into polypeptides by way of translation. The genetic code, deciphered almost 50 years ago, provides the fundamental clues for such decoding. However, deducing the protein(s) encoded by a given DNA segment remains a difficult task, even in simple microorganisms with few or no introns. This is due, in part, to our imperfect knowledge of the signals embedded in the genome that are involved in the translation of the genetic information. The latter is spectacularly illustrated for some genes, where the standard rules of decoding are subverted by “recoding” signals (Gesteland et al., 1992) that promote alternative decoding events like programmed ribosomal frameshifting, hopping, and termination codon reassignment (Atkins et al., 2001; Brierley and Pennell, 2001; Stahl et al., 2001). Several recoding events have been characterized in transposable elements and viruses, but the recent advances in genomics opens avenues for the discovery of related and novel recoding signals in cellular genes. Comprehensive analysis of the genome of Saccharomyces cerevisiae, for example, indicates that frameshifting and stop codon readthrough could *Correspondence: [email protected]

Review

be far more frequent events than previously suspected (Harrison et al., 2002; Namy et al., 2002, 2003; Shah et al., 2002). This is supported by the description of biologically important recoding events in the genes of several other eukaryotes and also in gram-positive, gram-negative, and methanogenic bacteria. To date, most of these have been identified serendipitously and a systematic search for recoding sites in sequenced genomes is urgently needed. Here, recoding in conventional cellular genes is reviewed, focusing on the biological importance of recoding in gene regulation. Approaches to identifying novel recoded genes are also discussed. Translational Recoding The dogma of the universal genetic code has been compromised by the finding that in certain organisms and organelles, the meaning of selected codons has been changed; for example, the specification of serine by CUG in Candida albicans (Santos et al., 1997) or of tryptophan by UGA in mitochondria (Barrell et al., 1979). In addition to such reassignment, which affects all genes of a given organism, alternative ways of reading the genetic code have been described that are programmed by signals present in specific mRNAs at defined locations in the messenger. Such recoding events occur generally in competition with standard decoding and allow the synthesis of two or more polypeptides, at a defined ratio, from a single mRNA molecule. Recoding can occur during translational elongation (⫹1 or –1 frameshifting, hopping) or at the termination step (stop codon readthrough/redefinition) (Gesteland and Atkins, 1996; Baranov et al., 2002). In frameshifting, the ribosome is induced to shift to an alternative, overlapping reading frame, whereas in hopping, the ribosome effectively detaches from the mRNA at a take off site until a landing site is reached where peptidyl tRNA re-pairs functionally to the mRNA and translation continues. In readthrough, the specific context of the termination codon is such that occasionally it is decoded by a tRNA rather than a release factor, allowing ribosomes to synthesize an extended polypeptide. Sometimes, this tRNA specifies an unusual amino acid, selenocysteine (Hatfield and Gladyshev, 2002) or pyrrolysine (Hao et al., 2002; Srinivasan et al., 2002). Recoding events also affect cell physiology in numerous ways and regulate a range of cellular processes (Table 1). Recoding during Translational Elongation Natural frameshift errors occur very rarely, but programmed ribosomal frameshifting signals increase the probability of tRNA slippage enormously, occasionally to such an extent that up to 50% of ribosomes change frame. The general principles of ⫹1 (3⬘-wards) and –1 (5⬘-wards) programmed frameshifting have largely been elucidated from studies of viruses, retrotransposons, and insertional elements (Chandler and Fayet, 1993; Atkins et al., 2001; Brierley and Pennell, 2001; Stahl et al., 2001; Plant et al., 2003). In essence, frameshifting is

Molecular Cell 158

Table 1. Recoding Sites in Cellular Genes Translational Step

Recoding Type

Elongation

⫹1 frameshifting

⫺1 frameshifting

Hopping Termination Readthrough

Selenocysteine Pyrrolysine

Gene(s)

Function

Species

prfB oaz1, 2, 3 ABP140 EST3 IL-10 PKAR EoNDR2 p43 TERT1, 2, 3 dnaX cdd HpfucT2 fucA1 Edr plaA CS3 topo-3 oaf kelch hdc PDE2 ⬎15 ⬎40 MtmB1/B2

release factor antizyme actin reorganization telomerase subunit Immune response modulator cAMP protein kinase nuclear protein kinase telomerase subunit telomerase subunit DNA polymerase III cytidine deaminase ␣(1,2) fucosyltransferase ␣-fucosidase Unknown Adhesin Pili formation DNA topoisomerase 3 Neuronal development Ring canal actin organizer tracheal branching inhibitor cAMP regulation Redox reaction → prok Antioxidant → euk methylamine methyltransferase

prok euk euk euk euk eup eup eup eup prok prok prok prok euk prok prok prok euk euk euk euk prok euk prok

Cellular genes employing a recoding strategy are detailed. The host species is identified by prok (prokaryote), euk (eukaryote), and eup (euplotes). Further details and relevant references are provided in the text. Data can be found at http://recode.genetics.utah.edu/

triggered by two elements, a slippery sequence in the mRNA, where tRNA movement or misalignment is favored, and a stimulator that enhances the process, probably by induction of a ribosomal pause. Often the stimulator is a slowly decoded region in the mRNA, like a codon for a low-abundance tRNA or a stop codon. Alternatively, a local mRNA secondary structure is present, a stem-loop or pseudoknot that may perturb normal decoding beyond simple induction of a pause (Kontos et al., 2001). Programmed ⫹1 Frameshifting in Prokaryotes The prfB gene of E. coli, encoding release factor 2 (RF2), harbors the paradigm ⫹1 ribosomal frameshift signal, yet remains the only example of this class in a prokaryotic gene (Craigen and Caskey, 1986). RF2 recognizes and promotes translation termination at UAA and UGA. The prfB open reading frame (ORF) is interrupted by a UGA stop codon some 26 codons downstream of the initiation codon, yet the coding sequence continues immediately, albeit in the ⫹1 reading frame, for a further 340 codons. The frameshift, typically by some 30%–50% of ribosomes, occurs at the slippery sequence CUU UGA C (the zero frame is indicated by triplets) present at the junction of the zero and ⫹1 ORFs, and generates active RF2 protein (Figure 1). Mechanistically, the process requires tRNALeu decoding zero frame CUU to move a single base 3⬘-wards onto UUU in the ⫹1 frame, a process enhanced in three ways; the similar anticodon: codon contacts formed postslippage, the relatively poor termination context of the stop codon (UGAC) and the stimulatory effect of base pairing of a Shine-Delgarno (SD)-like sequence, situated 3 nt upstream of the slippery site, with the anti-SD sequence of 16S rRNA (Weiss et al., 1988). The distance between this SD-like sequence

and the start of the RF2 slippery sequence is shorter than that typically seen between the SD and the initiator AUG (5 nt; Chen et al., 1994) and this may lead to some “tension” in the mRNA, promoting mRNA realignment in the 3⬘ direction (a push forward). This recoding event represents an elegant autoregulatory mechanism controlling the abundance of RF2 (Adamski et al., 1993). At high RF2 levels, the competition between termination and frameshifting is shifted in favor of termination, leading to a decrease of the RF2 concentration in the cell. At this point, frameshifting begins to predominate, raising RF2 levels. The fine-tuning facilitated by regulating RF2 levels by frameshifting may be required to maintain a constant termination efficiency at UGA codons, particularly for genes in which this codon is used to specify the incorporation of selenocysteine into proteins (see readthrough section below). The selection of tRNA(Ser)Sec is in competition with RF2 (Mansell et al., 2001) and if the concentration of the release factor became too high it could interfere with selenocysteine incorporation and prevent selenoproteins from playing a key role in activities such as the catalysis of redox reactions. Programmed ⫹1 Frameshifting in Eukaryotes Several ⫹1 frameshift signals have been described in eukaryotic mRNAs. Perhaps the most intricate is present in the mRNA encoding ornithine decarboxylase antizyme (Matsufuji et al., 1995). Antizyme 1 (AZ1) regulates the stability of ornithine decarboxylase (ODC), an enzyme responsible for converting ornithine to putrescine (Figure 2A), by targeting it to the proteasome in an unusual ubiquitin-independent mechanism, after which it gets recycled (Murakami et al., 1992). The AZ1 frameshift site comprises a slippery sequence that includes and requires the ORF1 stop codon (5⬘ UCC UGA U 3⬘), and

Recoding in Cellular Gene Expression 159

Figure 1. Autoregulation of RF2 Synthesis by ⫹1 Frameshifting The frameshift signal of the prfB gene is depicted, with the internal termination codon (UGA) highlighted (STOP) and the number of amino acids to the termination codon in the ⫹1 frame shown. As RF2 becomes limiting (blue arrow) due to termination at this site, ⫹1 slippage of tRNALeu occurs from the zeroframe codon CUU (underlined) to the overlapping ⫹1 frame codon (UUU), enhanced by the upstream SD-like sequence (purple). However, as RF2 accumulates (red arrow), translation termination begins to predominate.

two stimulatory elements; a core sequence of unknown action (5⬘ CCG GGG CCU CGG 3⬘) upstream of the slippery sequence and an RNA pseudoknot beginning 3 nt downstream of the ORF1 stop (Figure 2B). AZ1 activity resides only in the frameshift product, but the precise mechanism of frameshifting is not known. Like RF2,

AZ1 frameshifting is linked to a feedback mechanism, although in this case as a sensor to regulate polyamine levels in mammalian cells. At high concentrations, ⫹1 frameshifting is increased, promoting the synthesis of AZ1, which in turn degrades ODC, leading to a reduction in cellular polyamine concentrations (Ivanov et al., 2000).

Figure 2. Regulation of Cellular Polyamine Levels Using Antizyme ⫹1 Frameshifting as a Sensor (A) High polyamine levels stimulate ⫹1 frameshifting required for the synthesis of functional antizyme 1 (AZ1). AZ1 binds ornithine decarboxylase (ODC) and triggers its degradation by the 26S proteasome, being itself recycled. As ODC catalyzes the first step of the polyamine biosynthesis pathway, its degradation leads to a decrease in polyamine levels, which in turn reduces frameshifting efficiency. (B) Key features of the AZ1 frameshift signal are shown. The site of frameshifting is a “shifty stop” (5⬘ UCC UGA 3⬘, in blue and red, respectively) and the process is enhanced both by a stimulatory RNA pseudoknot just 3⬘ of the slippery sequence and a poorly defined 5⬘ stimulatory sequence (?). The mechanism of frameshifting is not known. Currently an occlusion model is favored, with the pseudoknot and/or upstream stimulator modifying the structure of the A site (when occupied by the termination codon UGA) such that the U is occluded and the ⫹1 frame GAU codon can be decoded out-of-frame by tRNAAsp (Atkins et al., 2001). How polyamines stimulate the frameshift event remains to be determined.

Molecular Cell 160

Figure 3. Conservation of EST3 Organization among Saccharomyces Species EST3 expression requires a ⫹1 frameshift to fuse the products of ORF1 (93 amino acids) and ORF 2 (92 amino acids). The percentage amino acid identity between the sequences identified in S. cerevisiae and related Saccharomyces species is shown. The “missing” portion of ORF2 in S. bayanus (*) reflects the absence of sequence information (contig. end). The slippery sequences identified in the EST3 homologs, which are identical, are indicated with the stop codon of ORF1 in red.

Frameshifting is employed in the expression of all known antizymes in many different species and in all cases to date, the process is used as a sensor of free polyamines. The conservation of this mechanism throughout evolution highlights a crucial role for frameshifting in the regulation of polyamine levels. A requirement for ⫹1 frameshifting in telomerase activity has been shown in the last few years. The synthesis of telomeres in S. cerevisiae, as in many organisms, depends upon telomerase, a reverse transcriptase that uses an internal RNA as a template. The yeast enzyme is a ribonucleoprotein composed of at least four proteins (Est1p, Est2p, Est3p, Cdc13p) and the template RNA (TLC1). Est3p is a stable component of the telomerase holoenzyme and essential for the maintenance of telomeres in vivo. The ⫹1 frameshifting mechanism employed in the expression of the protein involves an essential, short slippery sequence, CUU AGU U (Morris and Lundblad, 1997). As the AGU codon in this stretch is decoded by a low abundance tRNA, a ribosomal pause is thought to occur, promoting ⫹1 slippage of peptidyl tRNALeu from the CUU to the overlapping UUA codon. Unsurprisingly, this slippery sequence is somewhat underrepresented in the S. cerevisiae genome (34 of 115 expected; Shah et al., 2002) but it is not especially rare, raising the possibility that other flanking stimulators may yet be identified, or that other functional sites exist. That frameshift efficiencies promoted by the slippery sequence alone (Shah et al., 2002) are lower than those seen with the complete EST3 sequence (Morris and Lundblad, 1997) supports the existence of such elements. The organization of EST3 genes and the utilization of ⫹1 frameshifting are conserved among different Saccharomyces species suggesting a key role for frameshifting in telomere maintenance (Figure 3). Additional support for this belief comes from an analysis of the genome of the ciliate Euplotes. Ciliates have

long been known to employ unconventional decoding in their reassignment of termination codons, with Euplotes decoding the conventional stop codon UGA as cysteine (Meyer et al., 1991). However, their capacity to use alternative genetic decoding can be extended to the utilization of ⫹1 frameshifting to express nuclear proteins (Klobutcher and Farabaugh, 2002). It has been estimated that ⬎5% of genes require ⫹1 frameshifting for their expression and although this value is based upon only the available 67 sequences, it suggests that programmed frameshifting is very frequent in Euplotes. So far, six genes have been confirmed to use ⫹1 frameshifting and of these, four are involved in telomerase activity. The p43 protein identified in E. aediculatus is associated with the active telomerase complex and probably facilitates the binding of the catalytic subunit to the RNA. The ⫹1 frameshift is suspected to take place at an AAA UAA A stretch by a mechanism dependent upon slow decoding of the stop codon (UAA) in the ribosomal A site, allowing tRNALys decoding the zero frame AAA codon in the P site to slip ⫹1 onto the overlapping AAU codon and subsequently, another lysyl-tRNA to decode the incoming ⫹1 frame codon (Aigner et al., 2000). In E. crassus, three TERT genes encoding the telomerase catalytic subunit have been identified. EcTERT2 expression requires a single ⫹1 frameshift event, probably at a AAA UAA C stretch, whereas expression of EcTERT1 and 3 needs two sequential ⫹1 frameshifts, at the sequences AAA UAA C and AAA UAA G (Karamysheva et al., 2003). Similar to the telomerase-related examples, the other two documented ⫹1 frameshift signals in Euplotes contain a “shifty” UAA stop codon (Weiss et al., 1987), namely AAA UAA A (Tan et al., 2001a, 2001b). PKAR of E. octocarinatus encodes a regulatory subunit of a cAMPdependent kinase and the EoNdr2 gene is a member of the Ndr serine/threonine protein kinase family. In mechanistic terms, the presumed slow decoding of the UAA stop codon may be a consequence of the use of UGA as a cysteine codon. The changes within RF1 that allowed UGA to escape recognition (as a stop codon) could well have compromised the ability of the release factor to recognize UAA efficiently (Klobutcher and Farabaugh, 2002). Future work will determine if these ⫹1 frameshifting events have any regulatory function and whether other mRNA elements are involved. Two other examples of ⫹ 1 frameshifting in eukaryotes warrant mention. The ABP140 gene of S. cerevisiae (Asakura et al., 1998) encodes an actin binding protein whose expression requires ⫹1 frameshifting at the slippery sequence CUU AGG C. This sequence is one of the most underrepresented heptameric stretches in the S. cerevisiae genome (11 found for 54 expected; Shah et al., 2002) and promotes highly efficient frameshifting, with about one in three ribosomes changing frame at this site. It is closely related to that documented for EST3 above and similarly, the potential involvement of other mRNA elements needs to be addressed. Another interesting example of ⫹1 programmed frameshifting comes from the human IL-10 gene, encoding an immunosuppressive, antiinflammatory cytokine (Saulquin et al., 2002). It has been shown that a cytotoxic T cell epitope generated from the IL-10 gene by ⫹1 frameshifting could activate autoagressive T cells leading to the elimination of the subset of cells producing this cytokine.

Recoding in Cellular Gene Expression 161

However, that the epitope is generated by translational frameshifting needs to be confirmed, as the proposed slippery sequence is unconventional (CUU CCC U), with no obvious flanking stimulatory elements. Programmed -1 Frameshifting in Prokaryotes Compared to ⫹1 frameshifting, few examples of –1 frameshifting in cellular genes have been described. Indeed, the sole example with obvious biological relevance in prokaryotes is present in the dnaX gene, encoding the ␥ and ␶ subunits of DNA polymerase III holoenzyme. The holoenzyme is composed of three components; a DNA polymerase, a processivity factor (␤ sliding clamp), and a clamp loader. Their association is required to obtain high processivity during genomic DNA replication. In E. coli, the ␶ subunit coordinates the synthesis of the leading and lagging strands by dimerization of the core polymerase. In addition, it mediates rapid replication fork movement by its interaction with the DnaB helicase, and affects the processivity during lagging strand synthesis (Figure 4A). The ␥ subunits present in the clamp loader play a crucial role in driving the ATP-induced conformational changes to the other subunits resulting in the opening of the sliding clamp that then binds primed DNA with high affinity. The synthesis of the ␥ protein results from a -1 frameshifting event that directs ribosomes to a premature stop codon while the longest form (␶) is translated by continued standard decoding (Blinkowa and Walker, 1990; Flower and McHenry, 1990). The last 213 amino acids of the ␶ protein, missing from the ␥ protein, confer the ability to bind directly the core polymerase and DNA. The frameshift occurs on the slippery sequence A AAA AAG, by simultaneous slippage of both P and A site tRNALys species from the zero (A AAA AAG) to the –1 frame (AAA AAA) (Jacks et al., 1988; Tsuchihashi and Brown, 1992). This process requires two stimulatory signals in the mRNA, an SDlike sequence 10 nt upstream of the slippery sequence and a stem loop structure 5 nt downstream of it (Larsen et al., 1994; see Figure 4B). In contrast to the RF2 signal discussed earlier, the distance between this SD-like sequence and the start of the dnaX slippery sequence is longer than that typically seen between the SD and the AUG during initiation of translation and this likely leads to some tension in the rRNA, promoting realignment in the 5⬘ direction (pull-back). Furthermore, the A site slip from AAG to AAA is favored by the stronger interaction of tRNALys with the –1 frame AAA. Thus multiple features are involved in this frameshift strategy. The conservation of the frameshift signal in many Enterobacteriacea, the related bacterium Vibrio cholerae, and in the more distant Neisseria genus might indicate an important function in the regulation of the ratio between the subunits ␶ and ␥ (Baranov et al., 2002). Remarkably, Thermus thermophilus uses a transcriptional slippage mechanism rather than frameshifting to produce ␶ and ␥, although the ratio of the two proteins probably remains the same (Larsen et al., 2000). Not all bacteria express a ␥ subunit however, with dnaX encoding solely the ␶ subunit. The absence of ␥ is believed to generate a polymerase whose action is less dedicated to the replication fork (Bruck et al., 2002). An unusual –1 frameshift signal is present in the Bacil-

lus subtilis cytidine deaminase (CDA) gene (cdd) (Mejlhede et al., 1999; Licznar et al., 2003). CDAs are zinccontaining enzymes involved in the pyrimidine salvage pathway and catalyze the formation of uridine and deoxyuridine from cytidine and deoxycytidine, respectively. The physiological relevance of the cdd frameshift event is uncertain, since the C-terminal extension has no apparent effect on CDA activity. The mechanism of the frameshift however, is very unusual, involving –1 slippage of a single A site tRNALys at the slippery sequence CGA AAG (from AAG to AAA) but with the P site tRNAArg appearing to remain bound in the zero phase, with the wobble base of its anticodon being displaced by the incoming lysyl-tRNA. An SD-like sequence 14 nt upstream of the CGA AAG motif stimulates frameshifting in a manner similar to that seen with dnaX. It has been speculated that the cdd frameshift allows translational regulation of the following gene, bex, (encoding a homolog of the E. coli membrane-bound G protein), since the 3⬘ end of cdd overlaps the 5⬘ end of bex. The SD-like sequence present upstream of the cdd slippery sequence may be used to initiate translation at the bex gene (Figure 5C). Under these conditions, a ribosomal pause at the SD-like sequence during translation of the cdd gene could prevent initiation at the bex gene (Mejlhede et al., 1999). Two further examples of –1 frameshifting in prokaryotic cellular genes have been described, in the HpfucT2 gene of certain Helicobacter pylori strains (Wang et al., 1999) and the fucA1 gene of Sulfolobus solfataricus (Cobucci-Ponzaro et al., 2003). To date, neither signal has been characterized regarding the role of potential cisacting stimulatory elements and the frameshift events remain to be confirmed in vivo. Programmed -1 Frameshifting in Eukaryotes Although many eukaryotic RNA viruses employ –1 frameshifting, there is only one example to date in a conventional cellular gene, mouse Edr (embryonal carcinoma differentiation regulated), and this signal appears to be a relic of retroviral origin (Shigemoto et al., 2001). The frameshift signal of Edr is present between two long overlapping ORFs and resembles a typical retrovirus frameshift signal (Brierley and Pennell, 2001), with a slippery sequence G GGA AAC and a potential pseudoknotforming region five bases downstream. Pausing at the pseudoknot is believed to promote simultaneous –1 slippage of both ribosome-bound tRNAs in manner similar to that described for the dnaX frameshift signal. The resemblance to viruses is increased when the coding potential of the overlapping ORFs is taken into account. ORF1 contains a putative zinc binding domain of the CCHC subclass with a high content of basic amino acids commonly found in retroviral Gag proteins, whereas the protein encoded by the downstream ORF contains a consensus motif for an aspartyl protease catalytic site. Thus the organization is reminiscent of a retroviral gag/ pro overlap. The function of the gene is unknown but the conservation of the frameshifting site in mouse and human and the expression pattern during development and in adult tissues argues for an important role for frameshifting. Indeed, a more widespread involvement of –1 frameshifting in eukaryotic gene expression seems

Molecular Cell 162

Figure 4. Examples of –1 Frameshifting in Prokaryotic Genes (A and B) Frameshifting in the E. coli dnaX gene. (A) Shows the structural organization of DNA polymerase III during the synthesis of the leading and lagging strands. The ␶ (normal translation) and ␥ subunits (–1 frameshifting) are derived from the dnaX gene. (B) Organization of the dnaX frameshift signal. Frameshifting occurs at the slippery sequence A AAA AAG (blue), enhanced by an upstream SD-like element (purple) and a downstream stem-loop structure. The stop codon of the ␥ subunit, in the –1 frame, is shown in red. Panel A of Figure 4 was reprinted from Leu, F., Georgescu, R. and O’Donnell, M. (2003). Mechanism of E. coli ␶ processivity. Switch during lagging strand synthesis. Molecular Cell 11, 315-327. Copyright (2003) with permission from Elsevier. (C) Ribosomal frameshifting in the B. subtilis cdd gene. Frameshifting occurs at the slippery sequence CGA AAG by an unusual single-tRNALys slippage event in the ribosomal A-site, from AAG to the overlapping AAA codon in the –1 frame. An SD-like sequence (purple) can stimulate the frameshift process or translation initiation at the start codon of the overlapping bex gene (orange). The cdd stop codon is indicated in red.

likely as judged from computer-assisted database screens (Hammell et al., 1999).

Hopping Hopping or translational bypassing during elongation appears to be a rare event in natural mRNAs. The only well-studied example is in gene 60 of bacteriophage T4, where complex cis- and trans-acting signals allow a 50 nt stretch of mRNA to be bypassed at high efficiency (Herr et al., 2000). There are no unquestionable examples of bypassing in cellular genes, but the expression of the Prevotella loescheii adhesin gene appears to require bypassing of a 29 nt coding gap in the generation of the SO34 protein (Manch-Citron et al., 1999). As yet, the elements that promote the ribosomal gymnastics have not been identified.

Recoding during Termination Translation termination is an efficient process, essential for the correct expression of proteins. The process is achieved by release factors, three in prokaryotes (RF1, RF2, RF3), and two in eukaryotes (eRF1 and eRF3). When a stop codon is presented in the ribosomal A site, competition occurs between natural suppressor tRNAs and class 1 release factors (RF1, RF2, eRF1) that bind the stop codon and trigger release of the polypeptide chain. RF1 recognizes UAA and UAG codons; RF2, UGA, and UAA, whereas eRF1 recognizes all three stop codons. Termination efficiency can be influenced by a number of factors, including the nucleotide context of the stop codon, particularly the base immediately 3⬘ of the stop, the identity of the last two amino acids incorporated into the polypeptide chain, and the presence of stimulatory elements downstream (Bertram et al., 2001). These ele-

Recoding in Cellular Gene Expression 163

Figure 5. Incorporation of Unusual Amino Acids at Stop Codons (A) Insertion of selenocysteine at UGA codons requires a selenocysteine insertion element (SECIS). In eubacteria, a specialized translational elongation factor SelB binds both the bSECIS element, located immediately downstream of the recoded UGA codon (in red), and the selenocysteine-incorporating tRNA(Ser)Sec. In eukaryotes, the SECIS is located in the 3-untranslated region. Association of the eukaryotic SelB ortholog mSelB (also known as eEFsec) requires an adaptor protein SBP2, the SECIS binding factor. (B) Provides examples of SECIS elements from eubacteria (E. coli fdnG) and eukaryotes (H. sapiens selZ). Unconventional base pairs within the selZ SECIS known to be essential for function are indicated in purple and the recoded UGA in red. Alongside the SECIS elements are putative PYLIS elements (pyrrolysine insertion elements) located downstream of the recoded UAG codons (in red) of the mtmB1 and B2 genes of various Methanosarcini strains. As mfold-derived structures (http://www.bioinfo.rpi.edu/applications/mfold/old/rna/form1.cgi), unconventional base pairs of the kind seen in SECIS elements would not be identified.

ments can greatly increase the probability that a stop codon will be decoded by a tRNA rather than a release factor and so allow ribosomes to synthesize an elongated protein with potentially different biochemical properties. The most common programmed readthrough event corresponds to the insertion of the amino acid selenocysteine at UGA codons in both prokaryotes and eukaryotes. In others cases, the flanking regions direct the incorporation of a natural amino acid.

Incorporation of Unusual Amino Acids: the 21st and 22nd Amino Acids Selenium is an essential micronutrient for many organisms, including humans. The major biological form of selenium is selenocysteine, which is obtained by biosynthesis from serine on a special tRNA, tRNA[Ser]Sec (Hatfield and Gladyshev, 2002). In prokaryotes, the incorporation of selenocysteine at a specific UGA site depends upon tRNA[Ser]SecUCA, a specialized translation elongation factor

Molecular Cell 164

SelB and a structured RNA, bSECIS, a bacterial selenocysteine insertion sequence located immediately after the UGA (Figure 5). In eukaryotes, perhaps as an evolutionary advantage gained from the uncoupling of transcription and translation, the SECIS elements are located in the 3⬘UTR of the mRNA. One or two SECIS elements are present and remarkably these can direct the incorporation of selenocysteine at multiple UGA codons within the same mRNA. In the zebrafish Danio rerio, for example, expression of selenoprotein P requires the reassignment of 17 UGA codons, indicating that selenocysteine insertion ought to be highly efficient (Tujebajeva et al., 2000). However, the precise efficiency of selenocysteine incorporation at various sites is an area of debate (Hatfield and Gladyshev, 2002; Krol, 2002; Driscoll and Copeland, 2003). Some signals are genuinely inefficient (eg: E. coli fdfF, 4%–5%, Suppmann et al., 1999) and as premature termination is the consequence of failure to insert selenocysteine, the cell must bear certain metabolic costs in order to accommodate this amino acid into proteins. In prokaryotes, selenoproteins are primarily involved in catabolic processes and use selenium to catalyze redox reactions, whereas in eukaryotes, selenoproteins participate in antioxidant and anabolic processes (reviewed in Hatfield and Gladyshev, 2002). Although within the eukaryotic kingdom, sequenced yeasts and higher plants do not appear to possess the machinery for inserting selenocysteine into proteins, selenoproteins are essential for mammalian development, as evidenced by the embryonic lethality observed in knockout mice lacking tRNA[Ser]Sec (Bosl et al., 1997). The UAG codon can trigger incorporation of pyrrolysine, the 22nd amino acid, which has recently been identified and found naturally in monomethylamine methyltransferases in Methanosarcina barkeri (Hao et al., 2002; Srinivasan et al., 2002). This archaebacterium is a member of the methanogen group able to thrive on a wide range of methanogenic substrates, including mono-, di-, and trimethylamines. Each substrate requires activation by a methyltransferase to generate methane. All known methanogen methylamine methyltransferase genes contain an in frame UAG stop codon that we now know to be read as pyrrolysine. The readthrough efficiency of the UAG present in the mtmB1 gene of M. barkeri has been estimated at more than 97%. The presence in this organism of a tRNAPyl with anticodon 3⬘AUC5⬘ and studies on the activation of this tRNA (Srinivasan et al., 2002; Polycarpo et al., 2003) argue for cotranslational incorporation of pyrrolysine, but experimental support is lacking; for example, mRNA signals to subvert the meaning of this particular UAG codon such as a PYLIS (pyrrolysine insertion sequence) element. However, putative secondary structures with conserved features can be predicted 5 or 6 nt downstream of the appropriate termination codon in the mtmB genes of three Methanosarcina strains (Figure 5). Whether these elements are required for pyrrolysine insertion requires investigation. Stop Codon Readthrough in Prokaryotes Contrary to popular belief, authentic examples of readthrough in prokaryotic genes are very scarce. The synthesis of the pathologically-important, adherent CS3 pilus of enterotoxigenic E. coli CFA/II strains requires the

expression of a 104 kDa protein (p104) by readthrough and the presence of a suppressor tRNAGln is necessary for efficient readthrough (Jalajakumari et al., 1989). In the absence of the suppressor tRNA, p104 is still produced but at insufficient levels to support pilus formation. Another potential example of prokaryotic readthrough comes from the topo-3 gene of Bacillus firmus OF4, encoding a protein with similarities to E. coli DNA topoisomerase III (Ivey et al., 1992). This gene is interrupted by an in frame UGA, but suppression has not been experimentally demonstrated. In fact, the stop codon is not conserved among other species, and at present this readthrough event must be viewed as speculative. Stop Codon Readthrough in Eukaryotes Recent studies of Drosophila developmental pathways have revealed the occurrence of readthrough in three genes, out at first (oaf) (Bergstrom et al., 1995), kelch (Robinson and Cooley, 1997), and headcase (hdc) (Steneberg et al., 1998). The proteins encoded are involved in diverse developmental processes. Oaf is expressed during oogenesis in nurse cells and is widely distributed throughout embryo development. It is found in embryonic, larval, and adult gonads of both sexes and its inactivation can cause early death during larval development. Similarly, kelch is necessary for the production of viable eggs in Drosophila ovaries; it is a structural component of the ring canals that provide the intercellular conduits through which cytoplasm is transported from the nurse cells to the oocyte in the egg chamber. Headcase is involved in the regulation of tracheal branching; it is expressed in a subset of tracheal fusion cells and inhibits, in a dose-dependent manner, the terminal branching of neighboring cells. The signals that promote readthrough in the three genes are uncertain and the amino acid(s) inserted at the readthrough sites have not been determined. The mechanism of readthrough is of great interest however, since in at least two of the cases, the efficiency of readthrough appears to vary during development. In ovarian cells, the oaf readthrough product is minor, but in larvae and pupae, it is expressed to levels similar to the non-readthrough protein. Similarly, a spatial and temporal regulation of readthrough in kelch has been observed, with the ratio of the shorter and longer forms of kelch approaching 1:1 during metamorphosis, with the most efficient suppression in the imaginal discs and the testis. In oaf and kelch, UGA is suppressed, raising the possibility that selenocysteine insertion could be occurring, perhaps supported by the observation that it was not possible to detect the kelch readthrough product when the UGA codon was replaced by UAA. However, the 3⬘ UTRs of the two genes do not possess an obvious SECIS element, and at least in the case of kelch, it was not possible to show incorporation of 75Se into the protein (Robinson and Cooley, 1997). In hdc, the stop codon being suppressed is unusually UAA, commonly associated with the highest termination efficiencies, yet the readthrough is highly efficient, with about 20% of ribosomes continuing translation (Steneberg and Samakovlis, 2001). Computer-assisted secondary structure analysis predicts the formation of stem-loop structures downstream of

Recoding in Cellular Gene Expression 165

the oaf, kelch, and hdc termination codons, especially stable in the case of hdc, and these may play a role in stimulating readthrough. In kelch, it is noticeable that the triplets flanking the UGA are identical (AUG) and a ribosomal hop over the stop codon could be taking place rather than readthrough. This may simply be a coincidence, but flanking sequences CCC AUG UGA AUG GAA are conserved across Drosophila species. Clearly, more work is needed to elucidate the molecular details. Readthrough has also been described in the yeast PDE2 gene, encoding a phosphodiesterase known to bind to and degrade cAMP with high affinity. Readthrough extends the Pde2 polypeptide with a motif that targets it for 26S proteasome degradation (Namy et al., 2002). The efficiency of the suppression event is influenced by the environment; under normal circumstances it is low (2%) but is enhanced about 10-fold in the absence of glucose and the presence of the yeast [PSI⫹] factor. The degradation of a proportion of translated Pde2 leads to an increase in the intracellular concentration of cAMP and a measured consequence of this is reduced thermoresistance (Namy et al., 2002). As cAMP is a major secondary messenger that controls, in response to environmental modifications, several biological processes including filamentation, stress response, adaptation to carbon source, and cell cycle progression, it seems likely that PDE2 readthrough will have a plethora of effects on yeast physiology. For example, it could account for some of the numerous phenotypes associated with the appearance of the yeast prion [PSI⫹] (Eaglestone et al., 1999; True and Lindquist, 2000). Conclusions Translational recoding events are associated with a variety of biological processes spanning the breadth of cellular expression strategies, from genomic replication and maintenance through protein synthesis and degradation. In some cases, recoding is an autoregulatory process, a tool for controlling gene expression in response to environmental factors. This is potentially of evolutionary benefit; the regulation of recoding could unveil silent genetic information by releasing the selective pressure on sequences downstream of the recoding site, allowing it to evolve quickly. Subsequent stimulation of recoding by environmental factors would allow the expression of proteins with potentially novel biochemical properties. This reversible adaptive process could allow organisms to occupy a new ecological niche without losing their capacity to occupy the old one, as has been described for [PSI⫹] in yeast (True and Lindquist, 2000) and Hsp90 in Arabidopsis (Queitsch et al., 2002). Recoding is clearly not limited to a subset of organisms, so one would envisage the identification, by genomic and proteomic studies, of numerous novel recoded cellular genes within the next few years. However, the prediction of recoding sites from genomic databases is currently a difficult task. Two independent bioinformatics strategies are being developed to facilitate the identification of recoded genes in whole genomes. In modelbased approaches, genomic sequences are searched for regions exhibiting a well-characterized recoding sig-

nal. Such analyses have already allowed the identification of several candidate recoded genes (Hammel et al., 1999; Namy et al., 2002). The difficulty here is specificity, with an imprecise model leading to the identification of false positive candidates and too rigid a model failing to identify truly positive candidates. Further, the identification of combinations of motifs involving primary and secondary structural elements in the RNA is best approached using stochastic context-free grammars (SCFGs), but few practical SCGF programs are available and are computationally complex, although efforts to reduce computational memory requirements are in progress (Eddy, 2002). Of course, the parameters of the model critically influence the outcome. For example, a model based on our current knowledge of a viral recoding signal would probably miss distinct cellular counterparts. For this reason, several groups are trying to develop bioinformatics approaches that do not rely on models of recoding sites and can be performed without a priori knowledge of the mechanism involved. These similarity based approaches seek genomic configurations compatible with recoding, such as ORFs separated by a unique stop codon or overlapping along an intermediate shared region. Several candidate recoded genes have already been identified in yeast (Harrison et al., 2002; Namy, et al., 2003) and in Drosophila (Sato et al., 2003) in this way. The main difficulty is to discriminate between active recoding sites, which play a physiological role, and pseudogenes, undetected introns, or sequencing errors. Naturally, the presence of known elements (stimulatory RNA structural motifs, SD-like sequences, nucleotide contexts) simplifies the identification of sites that are likely to have biological significance. In the case of selenoproteins, for example, the systematic presence of the SECIS element has been used to identify 25 novel selenoprotein genes in the human genome (Kryukov et al., 2003). Similarly, comparative genome sequence analysis should be of benefit in the identification of bona fide candidates. There are clearly similarities between cellular recoding events and those that take place in viruses and transposable elements, hinting at evolutionary links. For example, the –1 frameshift signals dnaX and Edr have counterparts in prokaryotic insertional elements and mammalian retroviruses, respectively (Baranov et al., 2002); the ⫹1 frameshifting signals of the Euplotes TERT genes are closely related to that used by the yeast retrotransposon Ty1 (Stahl et al., 2001), and the stop codon readthrough signal of Drosophila hdc bears similarity to that of the Dictostelium retrotransposon skipper (Bertram et al., 2001). Whether any of the cellular recoding signals have been acquired from transposable elements and viruses is unknown, but it is a possibility, especially in the case of Edr (Shigemoto et al., 2001). There is, paradoxically, good evidence of transmission the other way, namely the acquisition of a human glutathione peroxidase selenoprotein homolog by the poxvirus Molluscum contagiosum (Shisler et al., 1998). Whatever their evolutionary origins, a common feature of almost all recoding signals appears to be a requirement for a ribosomal pause, brought about by the use of slowly decoded sense codons, termination codons, mRNA secondary structures, or mRNA-rRNA interactions. In mechanistic terms, pausing would increase the time at

Molecular Cell 166

which ribosomes are held at a recoding site, promoting alternative events that would normally be unfavorable kinetically. Although the individual elements that comprise a recoding signal undoubtedly have functions beyond simple induction of a pause, pausing is probably a mechanistic requirement at most signals. Over the next few years, it will be interesting to see whether new recoding mechanisms will emerge. One possibility is ribosomal skipping, described in the picornavirus foot and mouth disease virus. Here a termination event occurs at a sense codon (without release factor involvement), followed by release of the nascent polypeptide prior to ribosomal translocation to the next inframe codon (Donnelly et al., 2001). While the mechanism needs further analysis, the process seems to be influenced by local sequences and could be used to regulate the ratio between the synthesized polypeptides. Another possible recoding event would be the regulation of amino acid misincorporation at sense codons leading to the modification of the activity, localization, or stability of the protein. Recoding, especially readthrough, has gained biomedical attention recently in the use of aminoglycosides to suppress disease-causing nonsense mutations (Howard et al., 2000; Sleat et al., 2001; Keeling and Bedwell, 2002). Although a promising new therapy for a large number of genetic diseases, the drugs lack specificity at present and their efficacy is influenced by the nucleotide sequence identity flanking the termination codon in question. A better understanding of the readthrough mechanism coupled with specific drug design should circumvent these problems allowing the use of this alternative treatment in numerous genetic diseases and cancers. Acknowledgments We would like to express our gratitude to Alain Krol and Guillaume Stahl for helpful discussions and to the maintainers of and contributors to the RECODE database (http://recode.genetics.utah.edu/cgibin/WebObjects/RecodeWeb.woa). References Adamski, F.M., Donly, B.C., and Tate, W.P. (1993). Competition between frameshifting, termination and suppression at the frameshift site in the Escherichia coli release factor-2 mRNA. Nucleic Acids Res. 21, 5074–5078.

R.K. (1995). Regulatory autonomy and molecular characterization of the Drosophila out at first gene. Genetics 139, 1331–1346. Bertram, G., Innes, S., Minella, O., Richardson, J., and Stansfield, I. (2001). Endless possibilities: translation termination and stop codon recognition. Microbiol. 147, 255–269. Blinkowa, A.L., and Walker, J.R. (1990). Programmed ribosomal frameshifting generates the Escherichia coli DNA polymerase III ␥ subunit from within the ␶ subunit reading frame. Nucleic Acids Res. 18, 1725–1729. Bosl, M.R., Takaku, K., Oshima, M., Nishimura, S., and Taketo, M.M. (1997). Early embryonic lethality caused by targeted disruption of the mouse selenocysteine tRNA gene (Trsp). Proc. Natl. Acad. Sci. USA 94, 5531–5534. Brierley, I., and Pennell, S. (2001). Structure and function of the stimulatory RNAs involved in programmed eukaryotic –1 ribosomal frameshifting. Cold Spring Harbor Symp. Quant. Biol., LXVI, 233–248. Bruck, I., Yuzhakov, A., Yurieva, O., Jeruzalmi, D., Skangalis, M., Kuriyan, J., and O’Donnell, M. (2002). Analysis of a multicomponent thermostable DNA polymerase III from an extreme thermophile. J. Biol. Chem. 277, 17334–17348. Chandler, M., and Fayet, O. (1993). Translational frameshifting in the control of transposition in bacteria. Mol. Microbiol. 7, 497–503. Chen, H., Bjerknes, M., Kumar, R., and Jay, E. (1994). Determination of the optimal aligned spacing between the Shine-Dalgarno sequence and the translation initiation codon of Escherichia coli mRNAs. Nucleic Acids Res. 22, 4953–4957. Cobucci-Ponzano, B., Trincone, A., Giordano, A., Rossi, M., and Moracci, M. (2003). Identification of an archaeal alpha-L-fucosidase encoded by an interrupted gene. Production of a functional enzyme by mutations mimicking programmed -1 frameshifting. J. Biol. Chem. 278, 14622–14631. Craigen, W.J., and Caskey, C.T. (1986). Expression of peptide chain release factor 2 requires high-efficiency frameshift. Nature 322, 273–275. Donnelly, M.L., Luke, G., Mehrotra, A., Li, X., Hughes, L.E., Gani, D., and Ryan, M.D. (2001). Analysis of the aphthovirus 2A/2B polyprotein “cleavage” mechanism indicates not a proteolytic reaction, but a novel translational effect: a putative ribosomal “skip.” J. Gen. Virol. 82, 1013–1025. Driscoll, D.M., and Copeland, P.R. (2003). Mechanism and regulation of selenoprotein synthesis. Annu. Rev. Nutr. 23, 17–40. Eaglestone, S.S., Cox, B.S., and Tuite, M.F. (1999). Translation termination efficiency can be regulated in Saccharomyces cerevisiae by environmental stress through a prion-mediated mechanism. EMBO J. 18, 1974–1981. Eddy, S.R. (2002). A memory-efficient dynamic programming algorithm for optimal alignment of a sequence to an RNA secondary structure. BMC Bioinformatics 3, 18. Flower, A.M., and McHenry, C.S. (1990). The gamma subunit of DNA Polymerase III holoenzyme of Escherichia coli is produced by ribosomal frameshifting. Proc. Natl. Acad. Sci. USA 87, 3713–3717.

Aigner, S., Lingner, J., Goodrich, K.J., Grosshans, C.A., Shevchenko, A., Mann, M., and Cech, T.R. (2000). Euplotes telomerase contains an La motif protein produced by apparent translational frameshifting. EMBO J. 19, 6230–6239.

Gesteland, R.F., and Atkins, J.F. (1996). Recoding: dynamic reprogramming of translation. Annu. Rev. Biochem. 65, 741–768.

Asakura, T., Sasaki, T., Nagano, F., Satoh, A., Obaishi, H., Nishioka, H., Imamura, H., Hotta, K., Tanaka, K., Nakanishi, H., et al. (1998). Isolation and characterization of a novel actin filament-binding protein from Saccharomyces cerevisiae. Oncogene 16, 121–130.

Hao, B., Gong, W., Ferguson, T.K., James, C. M, Krzycki, J.A., and Chan, M.K. (2002). A new UAG-encoded residue in the structure of a methanogen methyltransferase. Science 296, 1462–1466.

Atkins, J.F., Baranov, P.V., Fayet, O., Herr, A.J., Howard, M.T., Ivanov, I.P., Matsufuji, S., Miller, W.A., Moore, B., Pre`re, M.F., et al. (2001). Overriding standard decoding: implications of recoding for ribosome function and enrichment of gene expression. Cold Spring Harbor Symp.Quant. Biol. LXVI, 217–232. Baranov, P.V., Gesteland, R.F., and Atkins, J.F. (2002). Recoding: translational bifurcations in gene expression. Gene 286, 187–201.

Gesteland, R.F., Weiss, R.B., and Atkins, J.F. (1992). Recoding: reprogrammed genetic decoding. Science 257, 1640–1641.

Hammell, A.B., Taylor, R.C., Peltz, S.W., and Dinman, J.D. (1999). Identification of putative programmed -1 ribosomal frameshift signals in large DNA databases. Genome Res. 9, 417–427. Harrison, P., Kumar, A., Lan, N., Echols, N., Snyder, M., and Gerstein, M. (2002). A small reservoir of disabled ORFs in the yeast genome and its implications for the dynamics of proteome evolution. J. Mol. Biol. 316, 409–419.

Barrell, B.G., Bankier, A.T., and Drouin, J. (1979). A different genetic code in human mitochondria. Nature 282, 189–194.

Hatfield, D.L., and Gladyshev, V.N. (2002). How selenium has altered our understanding of the genetic code. Mol. Cell. Biol. 22, 3565– 3576.

Bergstrom, D.E., Merli, C.A., Cygan, J.A., Shelby, R., and Blackman,

Herr, A.J., Gesteland, R.F., and Atkins, J.F. (2000). One protein from

Recoding in Cellular Gene Expression 167

two open reading frames: mechanism of a 50 nt translational bypass. EMBO J. 19, 2671–2680. Howard, M.T., Shirts, B.H., Petros, L.M., Flanigan, K.M., Gesteland, R.F., and Atkins, J.F. (2000). Sequence specificity of aminoglycoside-induced stop condon readthrough: potential implications for treatment of Duchenne muscular dystrophy. Ann. Neurol. 48, 164–169. Ivanov, I.P., Gesteland, R.F., and Atkins, J.F. (2000). Antizyme expression: a subversion of triplet decoding, which is remarkably conserved by evolution, is a sensor for an autoregulatory circuit. Nucleic Acids Res. 28, 3185–3196. Ivey, D.M., Cheng, J., and Krulwich, T.A. (1992). A 1.6 kb region of Bacillus firmus OF4 DNA encodes a homolog of Escherichia coli and yeast DNA topoisomerases and may contain a translational readthrough of UGA. Nucleic Acids Res. 20, 4928. Jacks, T., Madhani, H.D., Masiarz, F.R., and Varmus, H.E. (1988). Signals for ribosomal frameshifting in the Rous sarcoma virus gagpol region. Cell 55, 447–458. Jalajakumari, M.B., Thomas, C.J., Halter, R., and Manning, P.A. (1989). Genes for biosynthesis and assembly of CS3 pili of CFA/II enterotoxigenic Escherichia coli: novel regulation of pilus production by bypassing an amber codon. Mol. Microbiol. 3, 1685–1695. Karamysheva, Z., Wang, L., Shrode, T., Bednenko, J., Hurley, L.A., and Shippen, D.E. (2003). Developmentally programmed gene elimination in Euplotes crassus facilitates a switch in the telomerase catalytic subunit. Cell 113, 565–576. Keeling, K.M., and Bedwell, D.M. (2002). Clinically relevant aminoglycosides can suppress disease-associated premature stop mutations in the IDUA and p53 cDNAs in a mammalian translation system. J. Mol. Med. 80, 367–376. Klobutcher, L.A., and Farabaugh, P.J. (2002). Shifty ciliates: frequent programmed translational frameshifting in euplotids. Cell 111, 763–766. Kontos, H., Napthine, S., and Brierley, I. (2001). Ribosomal pausing at a frameshifter RNA pseudoknot is sensitive to reading phase but shows little correlation with frameshift efficiency. Mol. Cell. Biol. 21, 8657–8670. Krol, A. (2002). Evolutionarily different RNA motifs and RNA-protein complexes to achieve selenoprotein synthesis. Biochimie 84, 765–774. Kryukov, G.V., Castellano, S., Novoselov, S.V., Lobanov, A.V., Zehtab, O., Guigo, R., and Gladyshev, V.N. (2003). Characterization of mammalian selenoproteomes. Science 300, 1439–1443. Larsen, B., Wills, N.M., Gesteland, R.F., and Atkins, J.F. (1994). rRNA-mRNA base-pairing stimulates a programmed –1 ribosomal frameshift. J. Bacteriol. 176, 6842–6851. Larsen, B., Wills, N.M., Nelson, C., Atkins, J.F., and Gesteland, R.F. (2000). Nonlinearity in genetic decoding: homologous DNA replicase genes use alternatives of transcriptional slippage or translational frameshifting. Proc. Natl. Acad. Sci. USA 97, 1683–1688. Licznar, P., Mejlhede, N., Prere, M.F., Wills, N., Gesteland, R.F., Atkins, J.F., and Fayet, O. (2003). Programmed translational -1 frameshifting on hexanucleotide motifs and the wobble properties of tRNAs. EMBO J. 22, 4770–4778. Manch-Citron, J.N., Dey, A., Schneider, R., and Nguyen, N.Y. (1999). The translational hop junction and the 5⬘ transcriptional start site for the Prevotella loescheii adhesin encoded by plaA. Curr. Microbiol. 38, 22–26. Mansell, J.B., Guevremont, D., Poole, E.S., and Tate, W.P. (2001). A dynamic competition between release factor 2 and the tRNA(Sec) decoding UGA at the recoding site of Escherichia coli formate dehydrogenase H. EMBO J. 20, 7284–7293. Matsufuji, S., Matsufuji, T., Miyazaki, Y., Murakami, Y., Atkins, J.F., Gesteland, R.F., and Hayashi, S. (1995). Autoregulatory frameshifting in decoding mammalian ornithine decarboxylase antizyme. Cell 80, 51–60. Mejlhede, N., Atkins, J.F., and Neuhard, J. (1999). Ribosomal –1 frameshifting during decoding of Bacillus subtilis cdd occurs at the sequence CGA AAG. J. Bacteriol. 181, 2930–2937. Meyer, F., Schmidt, H.J., Plumper, E., Hasilik, A., Mersmann, G.,

Meyer, H.E., Engstrom, A., and Heckmann, K. (1991). UGA is translated as cysteine in pheromone 3 of Euplotes octocarinatus. Proc. Natl. Acad. Sci. USA 88, 3758–3761. Morris, D.K., and Lundblad, V. (1997). Programmed translational frameshifting in a gene required for yeast telomere replication. Curr. Biol. 7, 969–976. Murakami, Y., Matsufuji, S., Kameji, T., Hayashi, S., Igarashi, K., Tamura, T., Tanaka, K., and Ichihara, A. (1992). Ornithine decarboxylase is degraded by the 26S proteasome without ubiquitination. Nature 360, 597–599. Namy, O., Duchateau-Nguyen, G., and Rousset, J.P. (2002). Translational readthrough of the PDE2 stop codon modulates cAMP levels in Saccharomyces cerevisiae. Mol. Microbiol. 43, 641–652. Namy, O., Duchateau-Nguyen, G., Hatin, I., Hermann-Le Denmat, S., Termier, M., and Rousset, J.P. (2003). Identification of stop codon readthrough genes in Saccharomyces cerevisiae. Nucleic Acids Res. 31, 2289–2296. Plant, E.P., Jacobs, K.L., Harger, J.W., Meskauskas, A., Jacobs, J.L., Baxter, J.L., Petrov, A.N., and Dinman, J.D. (2003). The 9A˚ solution: How mRNA pseudoknots promote efficient programmed -1 ribosomal frameshifting. RNA 9, 168–174. Polycarpo, C., Ambrogelly, A., Ruan, B., Tumbula-Hansen, D., Ataide, S.F., Ishitani, R., Yokoyama, S., Nureki, O., Ibba, M., and Soll, D. (2003). Activation of the pyrrolysine suppressor tRNA requires formation of a ternary complex with class I and class II lysyl-tRNA synthetases. Mol. Cell 12, 287–294. Queitsch, C., Sangster, T.A., and Lindquist, S. (2002). Hsp90 as a capacitor of phenotypic variation. Nature 417, 618–624. Robinson, D.N., and Cooley, L. (1997). Examination of the function of two kelch proteins generated by stop codon suppression. Development 124, 1405–1417. Santos, M.A., Ueda, T., Watanabe, K., and Tuite, M.F. (1997). The non-standard genetic code of Candida spp.: an evolving genetic code or a novel mechanism for adaptation? Mol. Microbiol. 26, 423–431. Sato, M., Umeki, H., Saito, R., Kanai, A., and Tomita, M. (2003). Computational analysis of stop codon readthrough in D. melanogaster. Bioinformatics 19, 1371–1380. Saulquin, X., Scotet, E., Trautmann, L., Peyrat, M.A., Halary, F., Bonneville, M., and Houssaint, E. (2002). ⫹1 Frameshifting as a novel mechanism to generate a cryptic cytotoxic T lymphocyte epitope derived from human interleukin 10. J. Exp. Med. 195, 353–358. Shah, A.A., Giddings, M.C., Parvaz, J.B., Gesteland, R.F., Atkins, J.F., and Ivanov, I.P. (2002). Computational identification of putative programmed translational frameshift sites. Bioinformatics 18, 1046– 1053. Shigemoto, K., Brennan, J., Walls, E., Watson, C.J., Stott, D., Rigby, P.W., and Reith, A.D. (2001). Identification and characterisation of a developmentally regulated mammalian gene that utilises -1 programmed ribosomal frameshifting. Nucleic Acids Res. 29, 4079– 4088. Shisler, J.L., Senkevich, T.G., Berry, M.J., and Moss, B. (1998). Ultraviolet-induced cell death is blocked by a selenoprotein from a human dermatropic poxvirus. Science 279, 102–105. Sleat, D.E., Sohar, I., Gin, R.M., and Lobel P. (2001). Aminoglycosidemediated suppression of nonsense mutations in late infantile neuronal ceroid lipofuscinosis. Eur. J. Paediatr. Neurol. 5 Suppl A, 57–62. Srinivasan, G., James, C.M., and Krzycki, J.A. (2002). Pyrrolysine encoded by UAG in Archaea: charging of a UAG-decoding specialized tRNA. Science 296, 1459–1462. Stahl, G., Ben Salem, S., Li, Z., McCarty, G., Raman, A., Shah, M., and Farabaugh, P.J. (2001). Programmed ⫹1 translational frameshifting in the yeast Saccharomyces cerevisiae results from disruption of translational error correction. Cold Spring Harb. Symp. Quant. Biol., LXVI, 2249–258. Steneberg, P., and Samakovlis, C. (2001). A novel stop codon readthrough mechanism produces functional Headcase protein in Drosophila trachea. EMBO Rep. 2, 593–597.

Molecular Cell 168

Steneberg, P., Englund, C., Kronhamn, J., Weaver, T.A., and Samakovlis, C. (1998). Translational readthrough in the hdc mRNA generates a novel branching inhibitor in the Drosophila trachea. Genes Dev. 12, 956–967. Suppmann, S., Persson, B.C., and Bock, A. (1999). Dynamics and efficiency in vivo of UGA-directed selenocysteine insertion at the ribosome. EMBO J. 18, 2284–2293. Tan, M., Heckmann, K., and Brunen-Nieweler, C. (2001a). Analysis of micronuclear, macronuclear and cDNA sequences encoding the regulatory subunit of cAMP-dependent protein kinase of Euplotes octocarinatus: evidence for a ribosomal frameshift. J. Eukaryot. Microbiol. 48, 80–87. Tan, M., Liang, A., Brunen-Nieweler, C., and Heckmann, K. (2001b). Programmed translational frameshifting is likely required for expressions of genes encoding putative nuclear protein kinases of the ciliate Euplotes octocarinatus. J. Eukaryot. Microbiol. 48, 575–582. True, H.L., and Lindquist, S.L. (2000). A yeast prion provides a mechanism for genetic variation and phenotypic diversity. Nature 407, 477–483. Tsuchihashi, Z., and Brown, P.O. (1992). Sequence requirements for efficient translational frameshifting in the Escherichia coli dnaX gene and the role of an unstable interaction between tRNALys and an AAG lysine codon. Genes Dev. 6, 511–519. Tujebajeva, R.M., Ransom, D.G., Harney, J.W., and Berry, M.J. (2000). Expression and characterisation of nonmammalian selenoprotein P in the zebrafish, Danio rerio. Genes Cells 5, 897–903. Wang, G., Rasko, D.A., Sherburne, R., and Taylor, D.E. (1999). Molecular genetic basis for the variable expression of Lewis Y antigen in Helicobacter pylori: analysis of the alpha (1,2) fucosyltransferase gene. Mol. Microbiol. 31, 1265–1274. Weiss, R.B., Dunn, D.M., Atkins, J.F., and Gesteland, R.F. (1987). Slippery runs, shifty stops, backward steps, and forward hops: –2, –1, ⫹1, ⫹2, ⫹5, and ⫹6 ribosomal frameshifting. Cold Spring Harb. Symp. Quant. Biol. 52, 687–693. Weiss, R.B., Dunn, D.M., Dahlberg, A.E., Atkins, J.F., and Gesteland, R.F. (1988). Reading frame switch caused by base-pair formation between the 3⬘ end of 16S rRNA and the mRNA during elongation of protein synthesis in Escherichia coli. EMBO J. 7, 1503–1507.