The A-nucleotide preference of HIV-1 in the context of its structured RNA genome

Mapping the structural landscape of the yeast Ty3 retrotransposon RNA genome

Long terminal repeat (LTR)-retrotransposons are significant contributors to the evolution and diversity of eukaryotic genomes. Their RNA genomes (gRNA) serve as a template for protein synthesis and reverse transcription to a DNA copy, which can integrate into the host genome. Here, we used the SHAPE-MaP strategy to explore Ty3 retrotransposon gRNA structure in yeast and under cell-free conditions. Our study reveals the structural dynamics of Ty3 gRNA and the well-folded core, formed independently of the cellular environment. Based on the detailed map of Ty3 gRNA structure, we characterized the structural context of cis-acting sequences involved in reverse transcription and frameshifting. We also identified a novel functional sequence as a potential initiator for Ty3 gRNA dimerization. Our data indicate that the dimer is maintained by direct interaction between short palindromic sequences at the 5' ends of the two Ty3 gRNAs, resembling the model characteristic for other retroelements like HIV-1 and Ty1. This work points out a range of cell-dependent and -independent Ty3 gRNA structural changes that provide a solid background for studies on RNA structure-function relationships important for retroelement biology.

Structural basis of sequence-specific RNA recognition by the antiviral factor APOBEC3G

An essential step in restricting HIV infectivity by the antiviral factor APOBEC3G is its incorporation into progeny virions via binding to HIV RNA. However, the mechanism of APOBEC3G capturing viral RNA is unknown. Here, we report crystal structures of a primate APOBEC3G bound to different types of RNAs, revealing that APOBEC3G specifically recognizes unpaired 5'-AA-3' dinucleotides, and to a lesser extent, 5'-GA-3' dinucleotides. APOBEC3G binds to the common 3'A in the AA/GA motifs using an aromatic/hydrophobic pocket in the non-catalytic domain. It binds to the 5'A or 5'G in the AA/GA motifs using an aromatic/hydrophobic groove conformed between the non-catalytic and catalytic domains. APOBEC3G RNA binding property is distinct from that of the HIV nucleocapsid protein recognizing unpaired guanosines. Our findings suggest that the sequence-specific RNA recognition is critical for APOBEC3G virion packaging and restricting HIV infectivity.

Silent codon positions in the A-rich HIV RNA genome that do not easily become A: Restrictions imposed by the RNA sequence and structure

There is a strong evolutionary tendency of the human immunodeficiency virus (HIV) to accumulate A nucleotides in its RNA genome, resulting in a mere 40 per cent A count. This A bias is especially dominant for the so-called silent codon positions where any nucleotide can be present without changing the encoded protein. However, particular silent codon positions in HIV RNA refrain from becoming A, which became apparent upon genome analysis of many virus isolates. We analyzed these 'noA' genome positions to reveal the underlying reason for their inability to facilitate the A nucleotide. We propose that local RNA structure requirements can explain the absence of A at these sites. Thus, noA sites may be prominently involved in the correct folding of the viral RNA. Turning things around, the presence of multiple clustered noA sites may reveal the presence of important sequence and/or structural elements in the HIV RNA genome.

Biased Mutation and Selection in RNA Viruses

RNA viruses are responsible for some of the worst pandemics known to mankind, including outbreaks of Influenza, Ebola, and COVID-19. One major challenge in tackling RNA viruses is the fact they are extremely genetically diverse. Nevertheless, they share common features that include their dependence on host cells for replication, and high mutation rates. We set out to search for shared evolutionary characteristics that may aid in gaining a broader understanding of RNA virus evolution, and constructed a phylogeny-based data set spanning thousands of sequences from diverse single-stranded RNA viruses of animals. Strikingly, we found that the vast majority of these viruses have a skewed nucleotide composition, manifested as adenine rich (A-rich) coding sequences. In order to test whether A-richness is driven by selection or by biased mutation processes, we harnessed the effects of incomplete purifying selection at the tips of virus phylogenies. Our results revealed consistent mutational biases toward U rather than A in genomes of all viruses. In +ssRNA viruses, we found that this bias is compensated by selection against U and selection for A, which leads to A-rich genomes. In -ssRNA viruses, the genomic mutational bias toward U on the negative strand manifests as A-rich coding sequences, on the positive strand. We investigated possible reasons for the advantage of A-rich sequences including weakened RNA secondary structures, codon usage bias, and selection for a particular amino acid composition, and conclude that host immune pressures may have led to similar biases in coding sequence composition across very divergent RNA viruses.

A comprehensive analysis of genome composition and codon usage patterns of emerging coronaviruses

An outbreak of atypical pneumonia caused by a novel Betacoronavirus (βCoV), named SARS-CoV-2 has been declared a public health emergency of international concern by the World Health Organization. In order to gain insight into the emergence, evolution and adaptation of SARS-CoV-2 viruses, a comprehensive analysis of genome composition and codon usage of βCoV circulating in China was performed. A biased nucleotide composition was found for SARS-CoV-2 genome. This bias in genomic composition is reflected in its codon and amino acid usage patterns. The overall codon usage in SARS-CoV-2 is similar among themselves and slightly biased. Most of the highly frequent codons are A- and U-ending, which strongly suggests that mutational bias is the main force shaping codon usage in this virus. Significant differences in relative synonymous codon usage frequencies among SARS-CoV-2 and human cells were found. These differences are due to codon usage preferences.

Codon usage similarity between viral and some host genes suggests a codon-specific translational regulation

The codon usage pattern is a specific characteristic of each species; however, the codon usage of all of the genes in a genome is not uniform. Intriguingly, most viruses have codon usage patterns that are vastly different from the optimal codon usage of their hosts. How viral genes with different codon usage patterns are efficiently expressed during a viral infection is unclear. An analysis of the similarity between viral codon usage and the codon usage of the individual genes of a host genome has never been performed. In this study, we demonstrated that the codon usage of human RNA viruses is similar to that of some human genes, especially those involved in the cell cycle. This finding was substantiated by its concordance with previous reports of an upregulation at the protein level of some of these biological processes. It therefore suggests that some suboptimal viral codon usage patterns may actually be compatible with cellular translational machineries in infected conditions.

Evolution of Subgenomic RNA Shapes Dengue Virus Adaptation and Epidemiological Fitness

Changes in dengue virus (DENV) genome affect viral fitness both clinically and epidemiologically. Even in the 3′ untranslated region (3′ UTR), mutations could affect subgenomic flaviviral RNA (sfRNA) production and its affinity for host proteins, which are necessary for successful viral replication. Indeed, we recently showed that mutations in DENV2 3′ UTR of epidemic strains increased sfRNA ability to bind host proteins and reduce interferon expression. However, whether 3′ UTR differences shape the overall DENV evolution remains incompletely understood. Herein, we combined RNA phylogeny with phylogenetics to gain insights on sfRNA evolution. We found that sfRNA structures are under purifying selection and highly conserved despite sequence divergence. Only the second flaviviral nuclease-resistant RNA (fNR2) structure of DENV2 sfRNA has undergone strong positive selection. Epidemiological reports suggest that substitutions in fNR2 may drive DENV2 epidemiological fitness, possibly through sfRNA-protein interactions. Collectively, our findings indicate that 3′ UTRs are important determinants of DENV fitness in human-mosquito cycles.

•

Dengue viruses (DENVs) preserve RNA elements in their 3′ untranslated region (UTR).

•

Quantification of natural selection revealed positive selection on DENV2 sfRNA

•

Flaviviral nuclease-resistant RNAs (fNR) in the 3′ UTRs contribute to DENV speciation

•

A highly evolving fNR structure appears to increase DENV2 epidemiological fitness

Euclidean Distance Analysis Enables Nucleotide Skew Analysis in Viral Genomes

Nucleotide skew analysis is a versatile method to study the nucleotide composition of RNA/DNA molecules, in particular to reveal characteristic sequence signatures. For instance, skew analysis of the nucleotide bias of several viral RNA genomes indicated that it is enriched in the unpaired, single-stranded genome regions, thus creating an even more striking virus-specific signature. The comparison of skew graphs for many virus isolates or families is difficult, time-consuming, and nonquantitative. Here, we present a procedure for a more simple identification of similarities and dissimilarities between nucleotide skew data of coronavirus, flavivirus, picornavirus, and HIV-1 RNA genomes. Window and step sizes were normalized to correct for differences in length of the viral genome. Cumulative skew data are converted into pairwise Euclidean distance matrices, which can be presented as neighbor-joining trees. We present skew value trees for the four virus families and show that closely related viruses are placed in small clusters. Importantly, the skew value trees are similar to the trees constructed by a “classical” model of evolutionary nucleotide substitution. Thus, we conclude that the simple calculation of Euclidean distances between nucleotide skew data allows an easy and quantitative comparison of characteristic sequence signatures of virus genomes. These results indicate that the Euclidean distance analysis of nucleotide skew data forms a nice addition to the virology toolbox.

RNA Structure-A Neglected Puppet Master for the Evolution of Virus and Host Immunity

The central dogma of molecular biology describes the flow of genetic information from DNA to protein via an RNA intermediate. For many years, RNA has been considered simply as a messenger relaying information between DNA and proteins. Recent advances in next generation sequencing technology, bioinformatics, and non-coding RNA biology have highlighted the many important roles of RNA in virtually every biological process. Our understanding of RNA biology has been further enriched by a number of significant advances in probing RNA structures. It is now appreciated that many cellular and viral biological processes are highly dependent on specific RNA structures and/or sequences, and such reliance will undoubtedly impact on the evolution of both hosts and viruses. As a contribution to this special issue on host immunity and virus evolution, it is timely to consider how RNA sequences and structures could directly influence the co-evolution between hosts and viruses. In this manuscript, we begin by stating some of the basic principles of RNA structures, followed by describing some of the critical RNA structures in both viruses and hosts. More importantly, we highlight a number of available new tools to predict and to evaluate novel RNA structures, pointing out some of the limitations readers should be aware of in their own analyses.

Within-patient mutation frequencies reveal fitness costs of CpG dinucleotides and drastic amino acid changes in HIV

HIV has a high mutation rate, which contributes to its ability to evolve quickly. However, we know little about the fitness costs of individual HIV mutations in vivo, their distribution and the different factors shaping the viral fitness landscape. We calculated the mean frequency of transition mutations at 870 sites of the pol gene in 160 patients, allowing us to determine the cost of these mutations. As expected, we found high costs for non-synonymous and nonsense mutations as compared to synonymous mutations. In addition, we found that non-synonymous mutations that lead to drastic amino acid changes are twice as costly as those that do not and mutations that create new CpG dinucleotides are also twice as costly as those that do not. We also found that G→A and C→T mutations are more costly than A→G mutations. We anticipate that our new in vivo frequency-based approach will provide insights into the fitness landscape and evolvability of not only HIV, but a variety of microbes.

The evolution of subtype B HIV-1 tat in the Netherlands during 1985-2012

For the production of viral genomic RNA, HIV-1 is dependent on an early viral protein, Tat, which is required for high-level transcription. The quantity of viral RNA detectable in blood of HIV-1 infected individuals varies dramatically, and a factor involved could be the efficiency of Tat protein variants to stimulate RNA transcription. HIV-1 virulence, measured by set-point viral load, has been observed to increase over time in the Netherlands and elsewhere. Investigation of tat gene evolution in clinical isolates could discover a role of Tat in this changing virulence. A dataset of 291 Dutch HIV-1 subtype B tat genes, derived from full-length HIV-1 genome sequences from samples obtained between 1985-2012, was used to analyse the evolution of Tat. Twenty-two patient-derived tat genes, and the control Tat_HXB2 were analysed for their capacity to stimulate expression of an LTR-luciferase reporter gene construct in diverse cell lines, as well as for their ability to complement a tat-defective HIV-1_LAI clone. Analysis of 291 historical tat sequences from the Netherlands showed ample amino acid (aa) variation between isolates, although no specific mutations were selected for over time. Of note, however, the encoded protein varied its length over the years through the loss or gain of stop codons in the second exon. In transmission clusters, a selection against the shorter Tat86 ORF was apparent in favour of the more common Tat101 version, likely due to negative selection against Tat86 itself, although random drift, transmission bottlenecks, or linkage to other variants could also explain the observation. There was no correlation between Tat length and set-point viral load; however, the number of non-intermediate variants in our study was small. In addition, variation in the length of Tat did not significantly change its capacity to stimulate transcription. From 1985 till 2012, variation in the length of the HIV-1 subtype B tat gene is increasingly found in the Dutch epidemic. However, as Tat proteins did not differ significantly in their capacity to stimulate transcription elongation in vitro, the increased HIV-1 virulence seen in recent years could not be linked to an evolving viral Tat protein.

Impact of the biased nucleotide composition of viral RNA genomes on RNA structure and codon usage

We are interested in the influence of nucleotide composition on the fundamental characteristics of the virus RNA genome. Most RNA viruses have genomes with a distinct nucleotide composition, e.g. ranging from minimally 12.9 % to maximally 40.3 % (C- and U-count, respectively, in coronavirus HKU). We present a global analysis of diverse virus types, including plus-strand, minus-strand and double-strand RNA viruses, for the impact of this nucleotide preference on the predicted structure of the RNA genome that is packaged in virion particles and on the codon usage in the viral open reading frames. Several virus-specific features will be described, but also some general conclusions were drawn. Without exception, the virus-specific nucleotide bias was enriched in the unpaired, single-stranded regions of the RNA genome, thus creating an even more striking virus-specific signature. We present a simple mechanism that is based on elementary aspects of RNA structure folding to explain this general trend. In general, the nucleotide bias was the major determinant of the virus-specific codon usages, thus limiting a role for codon selection and translational control. We will discuss molecular and evolutionary scenarios that may be responsible for the diverse nucleotide biases of RNA viruses.

Nucleotide composition of the Zika virus RNA genome and its codon usage

BACKGROUND

RNA viruses have genomes with a distinct nucleotide composition and codon usage. We present the global characteristics of the RNA genome of Zika virus (ZIKV), an emerging pathogen within the Flavivirus genus. ZIKV was first isolated in 1947 in Uganda, caused a widespread epidemic in South and Central America and the Caribbean in 2015 and has recently been associated with microcephaly in newborns.

METHODS

The nearly 11 kb positive-stranded RNA genome of ZIKV was analyzed for its nucleotide composition, also in the context of the folded RNA molecule. Nucleotide trends were investigated along the genome length by skew analyses and we analyzed the codons used for translation of the ZIKV proteins.

RESULTS

ZIKV RNA has a biased nucleotide composition in being purine-rich and pyrimidine-poor. This preference for purines is a general characteristic of the mosquito-borne and tick-borne flaviviruses. The virus-specific nucleotide bias is further enriched in the unpaired, single-stranded regions of the structured ZIKV RNA genome, thus further imposing this ZIKV-specific signature. The codons used for translation of the ZIKV proteins is also unusual, but we show that it is the underlying bias in nucleotide composition of the viral RNA that largely dictates these codon preferences.

CONCLUSIONS

The ZIKV RNA genome has a biased nucleotide composition that dictates the codon usage of this flavivirus. We discuss the evolutionary scenarios and molecular mechanisms that may be responsible for these distinctive ZIKV RNA genome features.

Full-length RNA structure prediction of the HIV-1 genome reveals a conserved core domain

A distance constrained secondary structural model of the ≈10 kb RNA genome of the HIV-1 has been predicted but higher-order structures, involving long distance interactions, are currently unknown. We present the first global RNA secondary structure model for the HIV-1 genome, which integrates both comparative structure analysis and information from experimental data in a full-length prediction without distance constraints. Besides recovering known structural elements, we predict several novel structural elements that are conserved in HIV-1 evolution. Our results also indicate that the structure of the HIV-1 genome is highly variable in most regions, with a limited number of stable and conserved RNA secondary structures. Most interesting, a set of long distance interactions form a core organizing structure (COS) that organize the genome into three major structural domains. Despite overlapping protein-coding regions the COS is supported by a particular high frequency of compensatory base changes, suggesting functional importance for this element. This new structural element potentially organizes the whole genome into three major domains protruding from a conserved core structure with potential roles in replication and evolution for the virus.

Sequence-specific activation of the DNA sensor cGAS by Y-form DNA structures as found in primary HIV-1 cDNA

Cytosolic DNA that emerges during infection with a retrovirus or DNA virus triggers antiviral type I interferon responses. So far, only double-stranded DNA (dsDNA) over 40 base pairs (bp) in length has been considered immunostimulatory. Here we found that unpaired DNA nucleotides flanking short base-paired DNA stretches, as in stem-loop structures of single-stranded DNA (ssDNA) derived from human immunodeficiency virus type 1 (HIV-1), activated the type I interferon-inducing DNA sensor cGAS in a sequence-dependent manner. DNA structures containing unpaired guanosines flanking short (12- to 20-bp) dsDNA (Y-form DNA) were highly stimulatory and specifically enhanced the enzymatic activity of cGAS. Furthermore, we found that primary HIV-1 reverse transcripts represented the predominant viral cytosolic DNA species during early infection of macrophages and that these ssDNAs were highly immunostimulatory. Collectively, our study identifies unpaired guanosines in Y-form DNA as a highly active, minimal cGAS recognition motif that enables detection of HIV-1 ssDNA.

RNA structure. Structure of the HIV-1 RNA packaging signal

The 5' leader of the HIV-1 genome contains conserved elements that direct selective packaging of the unspliced, dimeric viral RNA into assembling particles. By using a (2)H-edited nuclear magnetic resonance (NMR) approach, we determined the structure of a 155-nucleotide region of the leader that is independently capable of directing packaging (core encapsidation signal; Ψ(CES)). The RNA adopts an unexpected tandem three-way junction structure, in which residues of the major splice donor and translation initiation sites are sequestered by long-range base pairing and guanosines essential for both packaging and high-affinity binding to the cognate Gag protein are exposed in helical junctions. The structure reveals how translation is attenuated, Gag binding promoted, and unspliced dimeric genomes selected, by the RNA conformer that directs packaging.

On the biased nucleotide composition of the human coronavirus RNA genome

•

The nucleotide composition of a coronaviral RNA genome is biased (high U, low C).

•

This bias is a relatively stable property along the viral genome, but less prominent in the last 1/3 of the genome.

•

This bias is even more pronounced in the single-stranded, unpaired RNA domains.

•

The bias dictates the atypical codon usage of the coronaviruses.

•

The RNA genome of the zoonotic viruses MERS and SARS is extremely biased.

We investigated the nucleotide composition of the RNA genome of the six human coronaviruses. Some general coronavirus characteristics were apparent (e.g. high U, low C count), but we also detected species-specific signatures. Most strikingly, the high U and low C proportions are quite variable and act like communicating vessels, C goes down when U goes up and vice versa. U ranges among virus isolates from 30.7% to 40.3%, and C makes the opposite movement from 20.0% to 12.9%, respectively. The nucleotide biases are more pronounced in the unpaired regions of the structured RNA genome, which may suggest a certain biological function for these distinctive sequence signatures. Coronaviruses have an atypical codon usage that has been linked to mutational events operating on the viral RNA genome on an evolutionary time scale. We suggest that the atypical nucleotide bias may serve a distinct biological function and that it is the direct cause of the characteristic codon usage in these viruses. The relevance for evolution of the novel human pathogens MERS and SARS is discussed.

On the nucleotide composition and structure of retroviral RNA genomes

Retroviral RNA genomes display a rich variety in their nucleotide composition. For instance, the single-stranded RNA genome of human T cell leukemia virus (HTLV-1) is C-rich and G-poor and that of the human immunodeficiency virus (HIV-1) is A-rich and C-poor. Animal retroviruses add further variation to this unexplained, but many times remarkable virus-specific property. We previously described that the nucleotide bias is even more extreme in the unpaired regions of the structured HIV-1 RNA genome, which has been probed by SHAPE technology. We now document that the same trend is apparent for the MFold-predicted RNA structure of HIV-1 RNA and subsequently investigated the predicted structures of the RNA genomes of other retroviruses. We conclude that all virus-specific signatures are enhanced for the unpaired nucleotides in the RNA genome. Consequently, the differences in nucleotide count between the diverse human and animal retroviruses are further exposed in the single stranded genome regions. We used a skew analysis to visualize these striking differences in nucleotide usage. Evolutionary events responsible for these nucleotide signatures will be discussed.

Large-scale nucleotide optimization of simian immunodeficiency virus reduces its capacity to stimulate type I interferon in vitro

Lentiviral RNA genomes present a strong bias in their nucleotide composition with extremely high frequencies of A nucleotide in human immunodeficiency virus type 1 (HIV-1) and simian immunodeficiency virus (SIV). Based on the observation that human optimization of RNA virus gene fragments may abolish their ability to stimulate the type I interferon (IFN-I) response, we identified the most biased sequences along the SIV genome and showed that they are the most potent IFN-I stimulators. With the aim of designing an attenuated SIV genome based on a reduced capacity to activate the IFN-I response, we synthesized artificial SIV genomes whose biased sequences were optimized toward macaque average nucleotide composition without altering their regulatory elements or amino acid sequences. A synthetic SIV optimized with 169 synonymous mutations in gag and pol genes showed a 100-fold decrease in replicative capacity. Interestingly, a synthetic SIV optimized with 70 synonymous mutations in pol had a normal replicative capacity. Its ability to stimulate IFN-I was reduced when infected cells were cocultured with reporter cells. IFN regulatory factor 3 (IRF3) transcription factor was required for IFN-I stimulation, implicating cytosolic sensors in the detection of SIV-biased RNA in infected cells. No reversion of introduced mutations was observed for either of the optimized viruses after 10 serial passages. In conclusion, we have designed large-scale nucleotide-modified SIVs that may display attenuated pathogenic potential.

IMPORTANCE

In this study, we synthesized artificial SIV genomes in which the most hyperbiased sequences were optimized to bring them closer to the nucleotide composition of the macaque SIV host. Interestingly, we generated a stable synthetic SIV optimized with 70 synonymous mutations in pol gene, which had a normal replicative capacity but a reduced ability to stimulate type I IFN. This demonstrates the possibility to rationally change viral nucleotide composition to design replicative and genetically stable lentiviruses with attenuated pathogenic potentials.

In vivo SELEX of single-stranded domains in the HIV-1 leader RNA

The 5' untranslated leader region of the human immunodeficiency virus type 1 (HIV-1) RNA genome is a strongly conserved sequence that encodes several regulatory motifs important for viral replication. Most of these motifs are exposed as hairpin structures, including the dimerization initiation signal (DIS), the major splice donor site (SD), and the packaging signal (Ψ), which are connected by short single-stranded regions. Mutational analysis revealed many functions of these hairpins, but only a few studies have focused on the single-stranded purine-rich sequences. Using the in vivo SELEX (systematic evolution of ligands by exponential enrichment) approach, we probed the sequence space in these regions that is compatible with efficient HIV-1 replication and analyzed the impact on the RNA secondary structure of the leader RNA. Our results show a strong sequence requirement for the DIS hairpin flanking regions. We postulate that these sequences are important for the binding of specific protein factors that support leader RNA-mediated functions. The sequence between the SD and Ψ hairpins seems to have a less prominent role, despite the strong conservation of the stretch of 5 A residues in natural isolates. We hypothesize that this may reflect the subtle evolutionary pressure on HIV-1 to acquire an A-rich RNA genome. In silico analyses indicate that sequences are avoided in all 3 single-stranded domains that affect the local or overall leader RNA folding. IMPORTANCE Many regulatory RNA sequences are clustered in the untranslated leader domain of the HIV-1 RNA genome. Several RNA hairpin structures in this domain have been proposed to fulfill specific roles, e.g., mediating RNA dimer formation to facilitate HIV-1 recombination. We now focus on the importance of a few well-conserved single-stranded sequences that connect these hairpins. We created libraries of HIV-1 variants in which these segments were randomized and selected the best-replicating variants. For two segments we document the selection of the (nearly) wild-type sequence, thus demonstrating the importance of these primary nucleotide sequences and the power of the in vivo SELEX approach. However, for the third segment a large variety of sequences is compatible with efficient HIV-1 replication. Interestingly, the A-rich sequence of this segment is highly conserved among HIV-1 isolates, which likely reflects the evolutionary tendency of HIV-1 to adopt A-rich sequences.

Comparison of SIV and HIV-1 genomic RNA structures reveals impact of sequence evolution on conserved and non-conserved structural motifs

RNA secondary structure plays a central role in the replication and metabolism of all RNA viruses, including retroviruses like HIV-1. However, structures with known function represent only a fraction of the secondary structure reported for HIV-1(NL4-3). One tool to assess the importance of RNA structures is to examine their conservation over evolutionary time. To this end, we used SHAPE to model the secondary structure of a second primate lentiviral genome, SIVmac239, which shares only 50% sequence identity at the nucleotide level with HIV-1NL4-3. Only about half of the paired nucleotides are paired in both genomic RNAs and, across the genome, just 71 base pairs form with the same pairing partner in both genomes. On average the RNA secondary structure is thus evolving at a much faster rate than the sequence. Structure at the Gag-Pro-Pol frameshift site is maintained but in a significantly altered form, while the impact of selection for maintaining a protein binding interaction can be seen in the conservation of pairing partners in the small RRE stems where Rev binds. Structures that are conserved between SIVmac239 and HIV-1(NL4-3) also occur at the 5' polyadenylation sequence, in the plus strand primer sites, PPT and cPPT, and in the stem-loop structure that includes the first splice acceptor site. The two genomes are adenosine-rich and cytidine-poor. The structured regions are enriched in guanosines, while unpaired regions are enriched in adenosines, and functionaly important structures have stronger base pairing than nonconserved structures. We conclude that much of the secondary structure is the result of fortuitous pairing in a metastable state that reforms during sequence evolution. However, secondary structure elements with important function are stabilized by higher guanosine content that allows regions of structure to persist as sequence evolution proceeds, and, within the confines of selective pressure, allows structures to evolve.