In eukaryotes, the specific cotranslational insertion of selenocysteine at UGA codons requires the presence of a secondary structural motif in the 3′ untranslated region of the selenoprotein mRNA. This selenocysteine insertion sequence (SECIS) element is predicted to form a hairpin and contains three regions of sequence invariance that are thought to interact with a specific protein or proteins. Specificity of RNA-binding protein recognition of cognate RNAs is usually characterized by the ability of the protein to recognize and distinguish between a consensus binding site and sequences containing mutations to highly conserved positions in the consensus sequence. Using a functional assay for the ability of wild-type and mutant SECIS elements to direct cotranslational selenocysteine incorporation, we have investigated the relative contributions of individual invariant nucleotides to SECIS element function. We report the novel finding that, for this consensus RNA motif, mutations at the invariant nucleotides are tolerated to different degrees in different elements, depending on the identity of a single nonconserved nucleotide. Further, we demonstrate that the sequences adjacent to the minimal element, although not required for function, can affect function through their propensity to base pair. These findings shed light on the specific structure these conserved sequences may form within the element. This information is crucial to the design of strategies for the identification of SECIS-binding proteins, and hence the elucidation of the mechanism of selenocysteine incorporation in eukaryotes.

The eukaryotic small nucleolar RNAs (snoRNAs) are involved in processing of pre-rRNA and modification of rRNA nucleotides. Some snoRNAs are derived from mono- or polycistronic transcription units, whereas others are encoded in introns of protein genes. The present study addresses the role of the RNA lariat-debranching enzyme (Dbr1p) in the synthesis and function of intronic snoRNAs in the yeast Saccharomyces cerevisiae. Intronic snoRNA production was determined to depend on Dbr1p. Accumulation of mature intronic snoRNAs is reduced in a dbr1 mutant; instead, intronic snoRNAs are “trapped” within host intron lariats. Interestingly, the extent of intronic snoRNA accumulation in the form of lariats in dbr1 cells varied among different intronic snoRNAs. Intronic snoRNAs encoded within shorter introns, such as U24 and snR38, accumulate more unprocessed lariat precursors than those encoded within longer introns, e.g., U18 and snR39. This correlation was corroborated by experiments conducted with model intron:U24 snoRNA constructs. These results support a splicing-dependent exonucleolytic pathway for the biosynthesis of intronic snoRNAs. Curiously, U24 in a lariat may be functional in directing methylation of ribosomal RNA.

The Saccharomyces cerevisiae TIF3 gene encodes a 436-amino acid (aa) protein that is the yeast homologue of mammalian translation initiation factor eIF4B. Tif3p can be divided into three parts, the N-terminal region with an RNA recognition motif (RRM) (aa 1–182), followed in the middle part by a sevenfold repeat of 26 amino acids rich in basic and acidic residues (aa 183–350), and a C-terminal region without homology to any known sequence (aa 351–436). We have analyzed several Tif3 proteins with deletions at their N and C termini for their ability (1) to complement a tif3Δ strain in vivo, (2) to stimulate Tif3-dependent translation extracts, (3) to bind to single-stranded RNA, and (4) to catalyze RNA strand-exchange in vitro. Here we report that yeast Tif3/eIF4B contains at least two RNA binding domains able to bind to single-stranded RNA. One is located in the N-terminal region of the protein carrying the RRM, the other in the C-terminal two-thirds region of Tif3p. The RRM-containing domain and three of the seven repeat motifs are essential for RNA strand-exchange activity of Tif3p and translation in vitro and for complementation of a tif3Δ strain, suggesting an important role for RNA strand-exchange activity in translation.

A number of heuristic descriptors have been developed previously in conjunction with the mfold package that describe the propensity of individual bases to participate in base pairs and whether or not a predicted helix is “well-determined.” They were developed for the “energy dot plot” output of mfold. Two descriptors, P-num and H-num, are used to measure the level of promiscuity in the association of any given nucleotide or helix with alternative complementary pairs. The third descriptor, S-num, measures the propensity of bases to be single-stranded. In the current work, we describe a series of programs that were developed in order to annotate individual structures with “well-definedness” information. We use color annotation to present the information. The programs can annotate PostScript files that are created by the mfold package or the PostScript secondary structure plots produced by the Weiser and Noller program XRNA (Weiser B, Noller HF, 1995, XRNA: Auto-interactive program for modeling RNA, The Center for Molecular Biology of RNA, Santa Cruz, California: University of California; Internet: ftp://fangio.ucsc.edu/pub/XRNA). In addition, these programs can annotate ss files that serve as input to XRNA. The annotation package can also handle structure comparison with a reference structure. This feature can be used to compare predicted structure with a phylogenetically deduced model, to compare two different predicted foldings, and to identify conformational changes that are predicted between wild-type and mutant RNAs.

We provide several examples of application. Predicted structures of two RNase P RNAs were colored with P-num information and further annotated with comparative information. The comparative model of a 16S rRNA was annotated with P-num information from mfold and with base pair probabilities obtained from the Vienna RNA folding package. Further annotation adds comparisons with the optimal foldings obtained from mfold and the Vienna package, respectively. The results of all of these analyses are discussed in the context of the reliability of structure prediction.

Translation termination in eukaryotes requires a codon-specific (class-I) release factor, eRF1, and a GTP/GDP-dependent (class-II) release factor, eRF3. The model of “molecular mimicry between release factors and tRNA” predicts that eRF1 mimics tRNA to read the stop codon and that eRF3 mimics elongation factor EF-Tu to bring eRF1 to the A site of the ribosome for termination of protein synthesis. In this study, we set up three systems, in vitro affinity binding, a yeast two-hybrid system, and in vitro competition assay, to determine the eRF3-binding site of eRF1 using the fission yeast Schizosaccharomyces pombe proteins and creating systematic deletions in eRF1. The in vitro affinity binding experiments demonstrated that the predicted tRNA-mimicry truncation of eRF1 (Sup45) forms a stable complex with eRF3 (Sup35). All three test systems revealed that the most critical binding site is located at the C-terminal region of eRF1, which is conserved among eukaryotic eRF1s and rich in acidic amino acids. To our surprise, however, the C-terminal deletion eRF1 seems to be sufficient for cell viability in spite of the severe defect in eRF3 binding when expressed in a temperature-sensitive sup45 mutant of the budding yeast, Saccharomyces cerevisiae. These results cannot be accounted for by the simple “eRF3–EF-Tu mimicry” model, but may provide new insight into the eRF3 function for translation termination in eukaryotes.

The eukaryotic nucleolus contains a diverse population of small nucleolar RNAs (snoRNAs) that have been categorized into two major families based on evolutionarily conserved sequence elements. U14 snoRNA is a member of the larger, box C/D snoRNA family and possesses nucleotide box C and D consensus sequences. In previous studies, we have defined a U14 box C/D core motif that is essential for intronic U14 snoRNA processing. These studies also revealed that nuclear proteins that recognize boxes C/D are required. We have now established an in vitro U14 snoRNP assembly system to characterize protein binding. Electrophoretic mobility-shift analysis demonstrated that all the sequences and structures of the box C/D core motif required for U14 processing are also necessary for protein binding and snoRNP assembly. These required elements include a base paired 5′,3′ terminal stem and the phylogenetically conserved nucleotides of boxes C and D. The ability of other box C/D snoRNAs to compete for protein binding demonstrated that the box C/D core motif-binding proteins are common to this family of snoRNAs. UV crosslinking of nuclear proteins bound to the U14 core motif identified a 65-kDa mouse snoRNP protein that requires boxes C and D for binding. Two additional core motif proteins of 55 and 50 kDa were also identified by biochemical fractionation of the in vitro-assembled U14 snoRNP complex. Thus, the U14 snoRNP core complex is a multiprotein particle whose assembly requires nucleotide boxes C and D.

hnRNP A1 regulates alternative splicing by antagonizing SR proteins. It consists of two closely related, tandem RNA-recognition motifs (RRMs), followed by a glycine-rich domain. Analysis of variant proteins with duplications, deletions, or swaps of the RRMs showed that although both RRMs are required for alternative splicing function, each RRM plays distinct roles, and their relative position is important. Surprisingly, RRM2 but not RRM1 could support this function when duplicated, despite their very similar structure. Specific RNA binding and annealing are not sufficient for hnRNP A1 alternative splicing function. These observations, together with phylogenetic and structural data, suggest that the two RRMs are quasi-symmetric but functionally nonequivalent modules that evolved as components of a single bipartite domain.

The nonsense-mediated mRNA decay pathway decreases the abundance of mRNAs that contain premature termination codons and prevents suppression of nonsense alleles. The UPF1 gene in the yeast Saccharomyces cerevisiae was shown to be a trans-acting factor in this decay pathway. The Upf1p demonstrates RNA-dependent ATPase, RNA helicase, and RNA binding activities. The results presented here investigate the binding affinity of the Upf1p for ATP and the consequences of ATP binding on its affinity for RNA. The results demonstrate that the Upf1p binds ATP in the absence of RNA. Consistent with this result, the TR800AA mutant form of the Upf1p still bound ATP, although it does not bind RNA. ATP binding also modulates the affinity of Upf1p for RNA. The RNA binding activity of the DE572AA mutant form of the Upf1p, which lacks ATPase activity, still bound ATP as efficiently as the wild-type Upf1p and destabilized the Upf1p–RNA complex. Similarly, ATPγS, a nonhydrolyzable analogue of ATP, interacted with Upf1p and promoted disassociation of the Upf1p–RNA complex. The conserved lysine residue (K436) in the helicase motif Ia in the Upf1p was shown to be critical for ATP binding. Taken together, these findings formally prove that ATP can bind Upf1p in the absence of RNA and that this interaction has consequences on the formation of the Upf1p–RNA complex. Further, the results support the genetic evidence indicating that ATP binding is important for the Upf1p to increase the translation termination efficiency at a nonsense codon. Based on these findings, a model describing how the Upf1p functions in modulating translation and turnover and the potential insights into the mechanism of the Upf1p helicase will be discussed.

Escherichia coli rRNA contains 10 pseudouridines of unknown function. They are made by synthases, each of which is specific for one or more pseudouridines. Here we show that the sfhB (yfiI) ORF of E. coli is a pseudouridine synthase gene by cloning, protein overexpression, and reaction in vitro with rRNA transcripts. Gene disruption by miniTn10(cam) insertion revealed that this synthase gene, here renamed rluD, codes for a synthase which is solely responsible in vivo for synthesis of the three pseudouridines clustered in a stem-loop at positions 1911, 1915, and 1917 of 23S RNA. The absence of RluD results in severe growth inhibition. Both the absence of pseudouridine and the growth defect could be reversed by insertion of a plasmid carrying the rluD gene into the mutant cell, clearly linking both effects to the absence of RluD. This is the first report of a major physiological defect due to the deletion of any pseudouridine synthase. Growth inhibition may be due to the lack of one or more of the 23S RNA pseudouridines made by this synthase since pseudouridines 1915 and 1917 are universally conserved and are located in proximity to the decoding center of the ribosome where they could be involved in modulating codon recognition.

Hepatitis delta virus (HDV) is a human pathogen that can greatly increase the severity of liver damage caused by an hepatitis B infection. HDV contains a circular, single-stranded RNA genome that encodes a unique protein, the delta antigen. Two forms of the delta antigen, δAg-S and δAg-L, are derived from a single open reading frame by RNA editing. Here we analyze the subcellular distribution of HDV RNA and its spatial relationship to known intranuclear structures. The human hepatoma cell line Huh7 was stably transfected with wild-type HDV cDNA and the viral RNAs were localized by in situ hybridization and fluorescence confocal microscopy. HDV RNA is detected throughout the nucleoplasm, with additional concentration in focal structures closely associated with nuclear speckles or clusters of interchromatin granules. Both the smaller form of the delta antigen (δAg-S), which is required for HDV genomic replication, and the larger form of the delta antigen (δAg-L), which represses replication, co-localize with delta RNA throughout the nucleoplasm and in the foci. However, the foci do not incorporate bromo-UTP and do not concentrate either RNA polymerase II or cleavage and polyadenylation factors required for viral RNA synthesis and 3′ end processing, respectively. Thus, it is unlikely that the delta foci represent major sites of viral transcription or replication. In conclusion, the data show that viral RNA–protein complexes accumulate in structures closely associated with interchromatin granules, a subnuclear domain highly enriched in small nuclear ribonucleoproteins, poly(A+) RNA, and RNA splicing protein factors. This implies a specific compartmentalization of ribonucleoproteins in the nucleus.

The SR proteins are essential metazoan pre-mRNA splicing factors that can also influence the selection of alternative 5′ splice sites in a concentration-dependent manner. Their activity in alternative splicing in vitro is antagonized by members of the hnRNP A/B family of proteins. The opposite effects of members of these two families of antagonistic splicing factors in vitro and upon overexpression in vivo suggest that changes in their relative levels may be a natural mechanism for the regulation of alternative splicing in vivo. One prediction of this model is that the ratios of these antagonists should vary in different cell types and in other situations in which cellular or viral transcripts are differentially spliced. We raised monoclonal antibodies specific for SF2/ASF and used them to measure the abundance of SF2/ASF protein and its isoforms, its phosphorylation state in vivo and during splicing in vitro, and its association with the spliceosome. SF2/ASF exists predominantly or exclusively in a highly phosphorylated state in vivo in all cell types examined, and unphosphorylated protein was not detectable. Unphosphorylated recombinant SF2/ASF becomes rapidly phosphorylated under splicing conditions in HeLa cell extracts and associates stably with one or more exons of β-globin pre-mRNA. This interaction appears to persist through the splicing reaction and SF2/ASF remains bound to spliced mRNA. We compared the distribution of SF2/ASF to that of its antagonist, hnRNP A1, in different rat tissues and in immortal and transformed cell lines. We found that the protein levels of these antagonistic splicing factors vary naturally over a very wide range, supporting the notion that changes in the ratio of these proteins can affect alternative splicing of a variety of pre-mRNAs in vivo.

The SECIS element is an RNA hairpin in the 3′UTR of selenoprotein mRNAs required for decoding UGA selenocysteine codons. Our experimentally derived 2D structure model for the SECIS RNA revealed the conservation of four consecutive non-Watson–Crick base pairs, with a central G.A/A.G tandem. The present study was dedicated to gaining insight into the role of this quartet of base pairs. The effects of mutations introduced into the SECIS quartet of the glutathione peroxidase (GPx) cDNA, an enzyme with selenocysteine in its active center, were reported in vivo by the GPx activity. The detrimental consequence of an all-Watson–Crick mutant quartet disclosed the paramount importance of the non-Watson–Crick base pairs for GPx activity. Next, structure probing established that base pair changes in the central G.A/A.G tandem, predicted by the model to be structurally unfavorable, effectively led to local opening of the helix at the quartet. A concomitant abolition of GPx activity was observed, arising from translational impairment of full-length GPx. In contrast, an isosteric base pair replacement in the tandem did not affect base pairing in the quartet, leading to an almost wt GPx activity. Collectively, the data provided conclusive evidence for the functional relevance of these non-Watson–Crick base pairs in vivo, thus identifying a noncanonical RNA motif crucial to SECIS function in mediating selenoprotein translation. Within the quartet, the prominent requirement for the central G.A/A.G tandem is highlighted, our previous structural model and the mutagenesis data presented here strongly arguing in favor of a sheared arrangement for the G.A base pairs. The SECIS RNA is therefore another member to be added to the growing list of RNAs containing building blocks of non-Watson–Crick base pairs, required for structure and/or function.

Spb4p is a putative ATP-dependent RNA helicase that is required for synthesis of 60S ribosomal subunits. Polysome analyses of strains genetically depleted of Spb4p or carrying the cold-sensitive spb4-1 mutation revealed an under-accumulation of 60S ribosomal subunits. Analysis of pre-rRNA processing by pulse-chase labeling, northern hybridization, and primer extension indicated that these strains exhibited a reduced synthesis of the 25S/5.8S rRNAs, due to inhibition of processing of the 27SB pre-rRNAs. At later times of depletion of Spb4p or following transfer of the spb4-1 strain to more restrictive temperatures, the early pre-rRNA processing steps at sites A0, A1, and A2 were also inhibited. Sucrose gradient fractionation showed that the accumulated 27SB pre-rRNAs are associated with a high-molecular-weight complex, most likely the 66S pre-ribosomal particle. An HA epitope-tagged Spb4p is localized to the nucleolus and the adjacent nucleoplasmic area. On sucrose gradients, HA-Spb4p was found almost exclusively in rapidly sedimenting complexes and showed a peak in the fractions containing the 66S pre-ribosomes. We propose that Spb4p is involved directly in a late and essential step during assembly of 60S ribosomal subunits, presumably by acting as an rRNA helicase.

Classical glutathione peroxidase (GPX1) mRNA levels can decrease to less than 10% in selenium (Se)-deficient rat liver. The cis-acting nucleic acid sequence requirements for Se regulation of GPX1 mRNA levels were studied by transfecting Chinese hamster ovary (CHO) cells with GPX1 DNA constructs in which specific regions of the GPX1 gene were mutated, deleted, or replaced by comparable regions from unregulated genes such as phospholipid hydroperoxide glutathione peroxidase (GPX4). For each construct, stable transfectants were pooled two weeks after transfection, divided into Se-deficient (2 nM Se) or Se-adequate (200 nM Se) medium, and grown for an additional four days. On day of harvest, Se-deficient GPX1 and GPX4 activities averaged 13 ± 2% and 15 ± 2% of Se adequate levels, confirming that cellular Se status was dramatically altered by Se supplementation. RNA was isolated from replicate plates of cells and transfected mRNA levels were specifically determined by RNase protection assay. Analysis of chimeric GPX1/GPX4 constructs showed that the GPX4 3′-UTR can completely replace the GPX13′-UTR in Se regulation of GPX1 mRNA. We did not find any GPX1 coding regions that could be replaced by the corresponding GPX4 coding regions without diminishing or eliminating Se regulation of the transfected GPX1 mRNA. Further analysis of the GPX1 coding region demonstrated that the GPX1 Sec codon (UGA) and the GPX1 intron sequences are required for full Se regulation of transfected GPX1 mRNA levels. Mutations that moved the GPX1 Sec codon to three different positions within the GPX1 coding region suggest that the mechanism for Se regulation of GPX1 mRNA requires a Sec codon within exon 1. Lastly, we found that addition of the GPX1 3′-UTR to β-globin mRNA can convey significant Se regulation to β-globin mRNA levels when a UGA codon is placed within exon 1. We conclude that Se regulation of GPX1 mRNA requires a functional selenocysteine insertion sequence (SECIS) in the 3′-UTR and a Sec codon followed by an intron.

The 5′ untranslated region (UTR) of the long-lived Escherichia coli ompA transcript functions as an mRNA stabilizer that can prolong the cytoplasmic lifetimes of a variety of messages to which it is fused. Previous studies have identified two domains of this 5′ UTR that together are responsible for its stabilizing effect. One is a 5′-terminal stem-loop. The other is a single-stranded RNA segment (ss2) that contains a ribosome binding site highly complementary to 16S ribosomal RNA. Here we report a detailed investigation of the function of these two stabilizing elements. Our data indicate that mRNA protection by a 5′ stem-loop requires no sequence features or thermodynamic stability beyond the minimum necessary for stem-loop formation. Stabilization by ss2 appears to result not from a high frequency of translation initiation, but rather from a high degree of occupancy of this 5′ UTR segment by bound ribosomes. Although close spacing of translating ribosomes is not critical for message stabilization by the ompA 5′ UTR, mRNA longevity does require the periodic passage of ribosomes through the protein-coding region. Unlike bound ribosomes, which hinder mRNA cleavage by RNase E, the 5′ stem-loop appears to impede degradation of ompA mRNA via a distinct pathway that is RNase E-independent. These findings imply that the ompA 5′ UTR prolongs mRNA longevity by impeding multiple pathways for mRNA degradation.

The eukaryotic nucleolus contains a large number of small nucleolar RNAs (snoRNAs) that are involved in pre-ribosomal RNA (pre-rRNA) processing. The H box/ACA-motif (H/ACA) class of snoRNAs has recently been demonstrated to function as guide RNAs targeting specific uridines in the pre-rRNA for pseudouridine (Ψ) synthesis. To characterize the protein components of this class of snoRNPs, we have purified the snR42 and snR30 snoRNP complexes by anti-m3G-immunoaffinity and Mono-Q chromatography of Saccharomyces cerevisiae extracts. Sequence analysis of the individual polypeptides demonstrated that the three proteins Gar1p, Nhp2p, and Cbf5p are common to both the snR30 and snR42 complexes. Nhp2p is a highly basic protein that belongs to a family of putative RNA-binding proteins. Cbf5p has recently been demonstrated to be involved in ribosome biogenesis and also shows striking homology with known prokaryotic Ψ synthases. The presence of Cbf5p, a putative Ψ synthase in each H/ACA snoRNP suggests that this class of RNPs functions as individual modification enzymes. Immunoprecipitation studies using either anti-Cbf5p antibodies or a hemagglutinin-tagged Nhp2p demonstrated that both proteins are associated with all H/ACA-motif snoRNPs. In vivo depletion of Nhp2p results in a reduction in the steady-state levels of all H/ACA snoRNAs. Electron microscopy of purified snR42 and snR30 particles revealed that these two snoRNPs possess a similar bipartite structure that we propose to be a major structural determining principle for all H/ACA snoRNPs.

The human malaria parasite, Plasmodium falciparum, maintains at least two distinct types, A and S, of developmentally controlled ribosomal RNAs. To investigate specific functions associated with these rRNAs, we replaced the Saccharomyces cerevisiae GTPase domain of the 25S rRNA with GTPase domains corresponding to the Plasmodium A- and S-type 28S rRNAs. The A-type rRNA differs in a single nonconserved base pair from the yeast GTPase domain. The S-type rRNA GTPase domain has three additional changes in highly conserved residues, making it unique among all known rRNA sequences. The expression of either A- or S-type chimeric rRNA in yeast increased translational accuracy. Yeast containing only A-type chimeric rRNA and no wild-type yeast rRNA grew at the wild-type level. In contrast, S-type chimeric rRNA severely inhibited growth in the presence of wild-type yeast rRNA, and caused lethality in the absence of the wild-type yeast rRNA. We show what before could only be hypothesized, that the changes in the GTPase center of ribosomes present during different developmental stages of Plasmodium species can result in fundamental changes in the biology of the organism.

While exons were originally defined as coding regions of split eukaryotic genes, introns have long been considered as mainly noncoding “genetic junk.” However, recognition that a large number of small nucleolar RNAs (snoRNAs) are processed from introns of pre-mRNAs demonstrated that introns may also code for functional RNAs. Moreover, recent characterization of the mammalian UHG gene that encodes eight box C/D intronic snoRNAs suggested that some genes generate functional RNA products exclusively from their intron regions. In this study, we show that the human U19 box H/ACA snoRNA, which is encoded within the second intron of the U19H gene, represents the only functional RNA product generated from the long U19H primary transcript. Splicing of the U19H transcript, instead of giving rise to a defined RNA, produces a population of diverse U19H RNA molecules. Although the first three exons of the U19H gene are preserved in each processed U19H RNA, the 3′ half of the RNA is generated by a series of apparently random splicing events. Because the U19H RNA possesses limited potential for protein coding and shows a predominant nucleoplasmic localization, we suggest that the sole function of the U19H gene is to express the U19 intronic snoRNA. This suggests that, in marked contrast to our previous dogmatic view, genes generating functionally important RNAs exclusively from their intron regions are probably more frequent than has been anticipated.

Splicing of exons 2 and 3 of α-tropomyosin (TM) involves mutually exclusive selection of either exon 3, which occurs in most cells, or of exon 2 in smooth muscle (SM) cells. The SM-specific selection of exon 2 results from the inhibition of exon 3. At least two essential cis-acting elements are required for exon 3 inhibition, the upstream and downstream regulatory elements (URE and DRE). These elements are essential for repression of TM exon 3 in SM cells, and also mediate a low level of repression of exon 3 in an in vitro 5′ splice site competition assay in HeLa extracts. Here, we show that the DRE consists of at least two discrete components, a short region containing a number of UGC motifs, and an essential pyrimidine-rich tract (DY). We show that the specific sequence of the DY element is important and that DY is able to bind to factors in HeLa nuclear extracts that mediate a low background level of exon 3 skipping. Deletion of a sequence within DY identified as an optimal binding site for PTB impairs (1) regulation of splicing in vivo, (2) skipping of exon 3 in an in vitro 5′ splice site competition, (3) the ability of DY competitors to affect the 5′ splice site competition in vitro, and (4) binding of PTB to DY. Addition of recombinant PTB to in vitro splicing reactions is able to partially reverse the effects of the DY competitor RNA. The data are consistent with a model for regulation of TM splicing that involves the participation of both tissue-specific and general inhibitory factors and in which PTB plays a role in repressing both splice sites of exon 3.

Cis-acting RNA signals are required for replication of positive-strand viruses such as the picornaviruses. Although these generally have been mapped to the 5′ and/or 3′ termini of the viral genome, RNAs derived from human rhinovirus type 14 are unable to replicate unless they contain an internal cis-acting replication element (cre) located within the genome segment encoding the capsid proteins. Here, we show that the essential cre sequence is 83–96 nt in length and located between nt 2318–2413 of the genome. Using dicistronic RNAs in which translation of the P1 and P2–P3 segments of the polyprotein were functionally dissociated, we further demonstrate that translation of the cre sequence is not required for RNA replication. Thus, although it is located within a protein-coding segment of the genome, the cre functions as an RNA entity. Computer folds suggested that cre sequences could form a stable structure in either positive- or minus-strand RNA. However, an analysis of mutant RNAs containing multiple covariant and non-covariant nucleotide substitutions within these putative structures demonstrated that only the predicted positive-strand structure is essential for efficient RNA replication. The absence of detectable minus-strand synthesis from RNAs that lack the cre suggests that the cre is required for initiation of minus-strand RNA synthesis. Since a lethal 3′ noncoding region mutation could be partially rescued by a compensating mutation within the cre, the cre appears to participate in a long-range RNA–RNA interaction required for this process. These data provide novel insight into the mechanisms of replication of a positive-strand RNA virus, as they define the involvement of an internally located RNA structure in the recognition of viral RNA by the viral replicase complex. Since internally located RNA replication signals have been shown to exist in several other positive-strand RNA virus families, these observations are potentially relevant to a wide array of related viruses.