Vertebrate genomes exhibit marked CG suppression-that is, lower than expected numbers of 5'-CG-3' dinucleotides. This feature is likely to be due to C-to-T mutations that have accumulated over hundreds of millions of years, driven by CG-specific DNA methyl transferases and spontaneous methyl-cytosine deamination. Many RNA viruses of vertebrates that are not substrates for DNA methyl transferases mimic the CG suppression of their hosts. This property of viral genomes is unexplained. Here we show, using synonymous mutagenesis, that CG suppression is essential for HIV-1 replication. The deleterious effect of CG dinucleotides on HIV-1 replication was cumulative, associated with cytoplasmic RNA depletion, and was exerted by CG dinucleotides in both translated and non-translated exonic RNA sequences. A focused screen using small inhibitory RNAs revealed that zinc-finger antiviral protein (ZAP) inhibited virion production by cells infected with CG-enriched HIV-1. Crucially, HIV-1 mutants containing segments whose CG content mimicked random nucleotide sequence were defective in unmanipulated cells, but replicated normally in ZAP-deficient cells. Crosslinking-immunoprecipitation-sequencing assays demonstrated that ZAP binds directly and selectively to RNA sequences containing CG dinucleotides. These findings suggest that ZAP exploits host CG suppression to identify non-self RNA. The dinucleotide composition of HIV-1, and perhaps other RNA viruses, appears to have adapted to evade this host defence.

Infection of animal cells by numerous viruses is detected and countered by a variety of means, including recognition of nonself nucleic acids. The zinc finger antiviral protein (ZAP) depletes cytoplasmic RNA that is recognized as foreign in mammalian cells by virtue of its elevated CG dinucleotide content compared with endogenous mRNAs. Here, we determined a crystal structure of a protein-RNA complex containing the N-terminal, 4-zinc finger human (h) ZAP RNA-binding domain (RBD) and a CG dinucleotide-containing RNA target. The structure reveals in molecular detail how hZAP is able to bind selectively to CG-rich RNA. Specifically, the 4 zinc fingers create a basic patch on the hZAP RBD surface. The highly basic second zinc finger contains a pocket that selectively accommodates CG dinucleotide bases. Structure guided mutagenesis, cross-linking immunoprecipitation sequencing assays, and RNA affinity assays show that the structurally defined CG-binding pocket is not required for RNA binding per se in human cells. However, the pocket is a crucial determinant of high-affinity, specific binding to CG dinucleotide-containing RNA. Moreover, variations in RNA-binding specificity among a panel of CG-binding pocket mutants quantitatively predict their selective antiviral activity against a CG-enriched HIV-1 strain. Overall, the hZAP RBD RNA structure provides an atomic-level explanation for how ZAP selectively targets foreign, CG-rich RNA.

HIV-1 accesses the nuclear DNA of interphase cells via a poorly defined process involving functional interactions between the capsid protein (CA) and nucleoporins (Nups). Here, we show that HIV-1 CA can bind multiple Nups, and that both natural and manipulated variation in Nup levels impacts HIV-1 infection in a manner that is strikingly dependent on cell-type, cell-cycle, and cyclophilin A (CypA). We also show that Nups mediate the function of the antiviral protein MX2, and that MX2 can variably inhibit non-viral NLS function. Remarkably, both enhancing and inhibiting effects of cyclophilin A and MX2 on various HIV-1 CA mutants could be induced or abolished by manipulating levels of the Nup93 subcomplex, the Nup62 subcomplex, NUP88, NUP214, RANBP2, or NUP153. Our findings suggest that several Nup-dependent ‘pathways’ are variably exploited by HIV-1 to target host DNA in a cell-type, cell-cycle, CypA and CA-sequence dependent manner, and are differentially inhibited by MX2.

The human zinc finger antiviral protein (ZAP) recognizes RNA by binding to CpG dinucleotides. Mammalian transcriptomes are CpG-poor, and ZAP may have evolved to exploit this feature to specifically target non-self viral RNA. Phylogenetic analyses reveal that ZAP and its paralogue PARP12 share an ancestral gene that arose prior to extensive eukaryote divergence, and the ZAP lineage diverged from the PARP12 lineage in tetrapods. Notably, the CpG content of modern eukaryote genomes varies widely, and ZAP-like genes arose subsequent to the emergence of CpG-suppression in vertebrates. Human PARP12 exhibited no antiviral activity against wild type and CpG-enriched HIV-1, but ZAP proteins from several tetrapods had antiviral activity when expressed in human cells. In some cases, ZAP antiviral activity required a TRIM25 protein from the same or related species, suggesting functional co-evolution of these genes. Indeed, a hypervariable sequence in the N-terminal domain of ZAP contributed to species-specific TRIM25 dependence in antiviral activity assays. Crosslinking immunoprecipitation coupled with RNA sequencing revealed that ZAP proteins from human, mouse, bat and alligator exhibit a high degree of CpG-specificity, while some avian ZAP proteins appear more promiscuous. Together, these data suggest that the CpG- rich RNA directed antiviral activity of ZAP-related proteins arose in tetrapods, subsequent to the onset of CpG suppression in certain eukaryote lineages, with subsequent species-specific adaptation of cofactor requirements and RNA target specificity.

The ~9.5 kilobase HIV-1 genome contains RNA sequences and structures that control many aspects of viral replication, including transcription, splicing, nuclear export, translation, packaging and reverse transcription. Nonetheless, chemical probing and other approaches suggest that the HIV-1 genome may contain many more RNA secondary structures of unknown importance and function. To determine whether there are additional, undiscovered cis-acting RNA elements in the HIV-1 genome that are important for viral replication, we undertook a global silent mutagenesis experiment. Sixteen mutant proviruses containing clusters of ~50 to ~200 synonymous mutations covering nearly the entire HIV-1 protein coding sequence were designed and synthesized. Analyses of these mutant viruses resulted in their division into three phenotypic groups. Group 1 mutants exhibited near wild-type replication, Group 2 mutants exhibited replication defects accompanied by perturbed RNA splicing, and Group 3 mutants had replication defects in the absence of obvious splicing perturbation. The three phenotypes were caused by mutations that exhibited a clear regional bias in their distribution along the viral genome, and those that caused replication defects all caused reductions in the level of unspliced RNA. We characterized in detail the underlying defects for Group 2 mutants. Second-site revertants that enabled viral replication could be derived for Group 2 mutants, and generally contained point mutations that reduced the utilization of proximal splice sites. Mapping of the changes responsible for splicing perturbations in Group 2 viruses revealed the presence of several RNA sequences that apparently suppressed the use of cryptic or canonical splice sites. Some sequences that affected splicing were diffusely distributed, while others could be mapped to discrete elements, proximal or distal to the affected splice site(s). Overall, our data indicate complex negative regulation of HIV-1 splicing by RNA elements in various regions of the HIV-1 genome that enable balanced splicing and viral replication.