Transcription Start Site Choice Regulates HIV-1 RNA Conformation and Function

Abstract

HIV-1, the causative agent of AIDS, is a retrovirus that packages two copies of unspliced viral RNA as a dimer into newly budding virions. The unspliced viral RNA also serves as an mRNA template for translation of two polyproteins. Recent studies suggest that the fate of the viral RNA (genome or mRNA) is determined at the level of transcription. RNA polymerase II uses heterogeneous transcription start sites to generate major transcripts that differ in only two guanosines at the 5' end. Remarkably, this two-nucleotide difference is sufficient to alter the structure of the 5'-untranslated region and generate two pools of RNA with distinct functions. The presence of both RNA species is needed for optimal viral replication and fitness.

RNA conformational heterogeneity controlled by HIV-1 transcription start site choice ensures two critical functions for viral replication

The unspliced retroviral RNA must perform two completely distinct roles to generate virions. It is the mRNA for synthesis of Gag, a polyprotein consisting of matrix, capsid, nucleocapsid (NC) and p6 domains, which functions as the major building-block of the assembling virion, and Gag-Pol, a polyprotein in which Gag is extended by the enzymes protease, reverse transcriptase, and integrase. In order to perform its second major role, this RNA must be packaged into the assembling particles, to transmit genetic information to the next viral generation [1]. Nearly 50 years ago, it was suggested that two distinct pools of this RNA may exist, one for each of these functions [2,3]. Indeed, it is possible that the RNA molecules in these two pools might somehow differ from each other.

Ongoing studies have progressively analyzed the virus-specific RNAs in HIV-1-producing cells in great detail. As in all retrovirus particles, the packaged genomic RNA is a dimer, in which two viral RNA molecules are initially held together by a small number of base-pairs during the virus assembly process [4]. Many observations indicate that dimerization is necessary for the selective packaging of HIV-1 RNA [5-7]. In HIV-1, the initial “joint” between the two RNA monomers is at a 6-base palindrome at the apical end of a stem-loop, often called the dimerization-initiation site (DIS) [8-11]. One important advance was the realization that the 5'-untranslated region (5'UTR) of the RNA can adopt at least two distinct conformations: in one, the DIS is exposed and available for pairing with the DIS of another monomer, while in the other, it is protected by intramolecular base-pairing [12,13]. Conversely, in the molecules with the exposed DIS, the translational initiation site for Gag synthesis is sequestered in intramolecular base-pairs, while it is exposed in the form with the blocked DIS. Thus, the existence of these two alternative conformations of the RNA seems an elegant solution to the problem of the two functions, as each conformation is apparently well-suited for one of the functions but not the other.

Recent studies have shown that the population of full-length viral RNAs in HIV-1-infected cells is far more complex than was originally suspected. It had been presumed that all of these RNA molecules were identical in primary sequence, but that they could adopt alternative secondary and tertiary structures [12-14]. However, as originally reported by Masuda et al. [15], because one of several neighboring nucleotides (nt) can be used to initiate transcription, the sequence at the 5' end of the HIV-1 RNA can vary [15-19]. This review summarizes our current knowledge regarding the structural and functional consequences of this unexpected diversity in primary sequence.

Transcription start sites and RNA functions

The HIV-1 provirus (the DNA copy of the viral genome) is integrated into the host DNA and is transcribed by host RNA polymerase II (pol II) to generate viral RNAs, which are then modified by host machinery including addition of a 5' modified guanosine cap and 3'-end polyadenylation. Pol II uses several neighboring nt as transcription start sites (TSSs) including three consecutive guanosines, generating RNAs with three, two and one guanosine at the 5' end referred to as 3G, 2G, and 1G RNA, respectively. Intriguingly, the distributions of HIV-1 unspliced RNA (HIV-1 RNA) species are different in cells and viral particles produced from these cells [15-18]. For example, in cells infected with NL4-3, an HIV-1 molecular clone, 3G RNA is the most abundant HIV-1 RNA species. However, in the particles produced from these cells, 1G RNA is the major species. Therefore, 1G RNA is selected over 3G RNA as the virion genome during the RNA genome packaging process, thereby enriching the 1G RNA in the viral particles. In other words, the 3G and 1G RNA are transcribed from the same proviral DNA template and are 99.9% identical but are nevertheless functionally distinct and HIV-1 can distinguish between these two RNA species.

Distinct 5' UTR structures and regulation of HIV-1 RNA packaging

The main secondary structural features of the 356-nt HIV-1 5'UTR are shown in Figure 1. From 5' to 3', these elements are termed transactivation response element (TAR) hairpin, polyadenylation (polyA) signal hairpin (so named as it is identical to the polyA signal in the 3'UTR), primer-binding site (PBS) domain, and three dynamic stem-loop (SL) elements (SL1, SL2, and SL3). The DIS is located at the apical loop of SL1. Multiple preferred NC/Gag binding sites [20-25] are located within these three SLs and are critical for genomic RNA packaging [21,26].

Fig. 1. The impact of transcription start site heterogeneity on HIV-1 RNA structures.

HIV-1 long terminal repeat DNA consists of U3, R, and U5 regions, and transcription start sites are located at the junction of U3 and R. HIV-1 5'UTR contains the following structural elements: TAR hairpin, polyA hairpin (so called because it is repeated at the 3' end of the RNA where it serves as a signal for polyadenylation), PBS region, and SL1, SL2, and SL3. Sequences near the junction of TAR and polyA harpins in predicted secondary structures are shown in inset. Two classes of HIV-1 mutants, addNN and AC/GU, that abrogate selective 1G packaging are shown; inserted NN and altered AC or GU are shown in red. Guanosine(s) at the 5' end of the transcripts are shown in blue.

As mentioned earlier, the HIV-1 5'UTR can fold into two distinct conformations: one displays all three SLs along with an exposed DIS and another does not. The exposure of the DIS enables the former conformation to dimerize [12]. The influence of TSS heterogeneity on 5'UTR dimerization propensity was explored using native gel electrophoresis [16,17]. Although all RNAs existed as equilibrium mixtures of monomeric and dimeric species, the 1G RNA adopted predominantly dimeric structures, whereas the 3G RNA existed predominantly as a monomer. The dimerization behavior of the in vitro transcribed and capped RNAs was similar to an RNA with one extra 5' guanosine [16]; for example, ^cap1G RNA behaved like 2G RNA. Based on mutagenesis experiments designed to stabilize the monomeric or dimeric conformations, a model for transcription start site-dependent control of dimerization was proposed [16]. In this model, additional base pairing in the ^cap3G transcripts partially disrupts the polyA hairpin stem allowing a guanosine at the base of the hairpin to base-pair with a cytosine in the DIS palindrome, thus stabilizing the monomeric conformer. Subsequent structural studies using nuclear magnetic resonance (NMR) spectroscopy [27] and RNA chemical probing [17,28] have all supported the critical role of polyA hairpin stability in altering the monomer-dimer equilibrium; however, these studies revealed the surprising impact of two additional guanosines, which distinguish the 1G from the 3G RNA, on global RNA structure, as described below.

NMR spectroscopy studies were used to probe the structures of ^Cap1G, ^Cap2G, and ^Cap3G HIV-1 5'UTRs derived from the HIV-1 MAL strain (which is somewhat different in sequence from NL4-3) [27]. Studies of the dimeric HIV-1 ^Cap1G RNA confirmed the presence in each monomer of the U5 helix that pairs upstream sequences with those flanking the gag start codon. Studies with smaller model RNAs were consistent with cap:C base pairing at the base of TAR and end-to-end stacking of TAR and polyA helices, as previously observed by small angle X-ray scattering (SAXS) [29]. NMR studies of monomeric ^cap2G and ^cap3G 5'UTR RNAs showed that structures of TAR and a portion of PBS were similar to those observed for the cap1G dimer, but the lower portion of the polyA hairpin was remodeled and signals diagnostic of the SL1 hairpin were missing [27]. Instead, residues in the downstream (3') portion of the polyA hairpin exhibited NMR signals indicative of base pairing with residues of DIS in SL1. A single additional G:C base pair formed at the base of TAR in these RNAs in comparison to the ^cap1G structure and the cap was exposed in these longer RNAs, rather than sequestered.

Chemical-probing studies of 5'UTR RNAs derived from HIV-1_NL4-3 also supported global structural differences between 1G and 3G RNAs [17,28]. 5'UTR RNAs (2G and 4G mimics of ^cap1G and ^cap3G, respectively) were analyzed using in-gel SHAPE (selective 2'-hydroxyl acylation followed by primer extension) probing [30,31] and found to fold into distinct structural ensembles [17]. The structures differed in several significant features including polyA hairpin stem structure and the exposure of DIS and NC-binding sites. The 1G RNAs maintained all these features, whereas the 3G RNAs had globally refolded structures that eliminated them. Importantly, several sets of mutants with changes at the base of the polyA stem, or nt insertions between TAR and polyA, were identified that eliminated 1G selective packaging into HIV-1 virions [17] (Figure 1); the structural differences between 1G and 3G RNAs in these mutants were also abolished [17]. These observations show that polyA stem stability is a key determinant of RNA packaging selectivity and that ^cap1G RNA is preferentially packaged due to exposure of structural elements that promote dimerization and Gag binding.

These basic conclusions were substantiated later using other chemical probing techniques. Functional analysis of RNA structure-sequencing (FARS-seq) was used to identify sequences and structures in the HIV-1_NL4-3 5'UTR that regulate RNA dimerization [28]. The authors investigated the effect of TSS heterogeneity on the dimerization propensity and structure of capped and uncapped 1G, 2G, and 3G 5'UTRs. Similar to previous studies, the RNAs had different dimerization propensities with 1G and 2G transcripts folding into dimer-promoting conformations more readily than 3G RNA. Capped and uncapped RNAs had similar functional profiles. Comparison of the 1G/2G dimer versus 3G monomer dimethyl sulfate reactivity profiles revealed statistically significant structural changes that mapped primarily to the polyA and SL1 domains. The 1G/2G dimer structure closely resembled a “canonical” HIV-1 5'UTR with exposed DIS and exposed NC-binding sites as observed in the NMR and SHAPE-probing studies described above. Also consistent with the earlier studies, the 3G monomer sample was dramatically restructured, with the polyA and SL1 hairpins re-folding into alternative structures, in this case involving long-range interactions between the DIS loop of SL1 and the apical portion of the polyA stem. Although all studies agree with a global rearrangement of polyA hairpin and SL1 in the 3G monomer, the details of the alternative structures formed differ between the three studies [17,27,28], likely due to differences in HIV-1 isolates and/or methods used.

Findings from Obayashi et al also support the idea that the number of G residues at the 5'-terminus of the HIV-1 5'UTR function in conformational regulation [32]. In this work, small 36- to 38-nt model RNAs that mimic the junction between TAR and polyA hairpins were used (Figure 1). The short (ranging from 5-8 bp each) stem regions were closed with stable tetraloops to resemble a dumbbell shape. The ^cap1G RNA (where a standard G was used in place of the cap G) migrated faster on a native gel than the ^cap2G or ^cap3G RNAs, suggesting a more compact structure for the 1G RNA. Thermodynamic melting profiles revealed that the ^cap1G RNA melted with a single transition, suggesting both stems were part of a single structural element. In contrast, both ^cap2G and ^cap3G RNAs melted in two steps, with one melting temperature (T_m) significantly increased and the other decreased relative to the single T_m measured for the ^cap1G RNA. The higher T_m was assigned to TAR and the lower one assigned to polyA. NMR data were consistent with the TAR and polyA domains forming a stable coaxially-stacked stem in the case of the ^cap1G RNA, as previously observed by SAXS analyses of the full TAR-polyA hairpins [29]. In contrast, for the ^cap2G or ^cap3G RNAs, the extra G nt(s) prevent the formation of an additional GC base pair at the end of polyA, making the two stems structurally independent.

Distinct 5'UTR structures and regulation of HIV-1 RNA translation

Whether transcription start site heterogeneity regulates HIV-1 RNA translation has been less extensively studied than its impact on RNA packaging. Using sucrose gradient step-fractionation, ^cap2G/3G HIV-1 RNAs were reported to be enriched on polysomes in HIV-1 infected cells, whereas ^cap1G RNAs were dominant in virion fractions [16]. Cap-binding proteins important for mRNA translation and processing, i.e., eukaryotic translation initiation factor 4E (eIF4E) and the human decapping enzyme hDcp2, respectively, were found to bind efficiently to monomeric ^cap2G/3G 5'UTRs, which were shown by NMR to be cap-exposed conformations of the leader, but not to the cap-sequestered ^Cap1G dimer [27]. In fact, cap sequestration was proposed to be an essential determinant of genome packaging [33,34]. Insertion of residues after G1 in a 5'UTR ^cap1G construct was expected to expose the 5' cap, but did not prevent dimerization or nucleocapsid protein binding and still failed to promote packaging [33]. Whether this cap-exposed RNA served as an mRNA for translation was not examined. In addition to cap sequestration, various structural features present only in ^cap1G RNAs and not ^cap3G RNAs were speculated to stabilize the 5' end of 5'UTR transcripts and inhibit their translation, although no experiments were carried out to test this idea [28].

A recent study directly probed whether a 2-nt difference affects the translation of HIV-1 RNA by performing in vitro translation assays using reporter constructs and polysome profiling of HIV-1 infected cells (Z Cheng et al., unpublished). These studies showed that ^cap1G and ^cap3G RNAs are both translated; however, ^cap3G RNA is translated more efficiently than ^cap1G RNA due to differences in 5'UTR RNA structures. These data strongly support the hypothesis that the 2-nt difference in two 9-kb RNAs can alter their global 5'UTR structures and make one RNA highly packageable and the other RNA more efficiently translated, thereby fulfilling both functions of HIV-1 unspliced RNA.

Biological importance of having different RNA species

Herein, we described the structures of the 5'UTR of HIV-1 RNAs that regulate their functions. Describing long (>100 nt) RNAs as a single structure or conformation is almost never warranted [35,36]. RNAs are more accurately described as a conformational ensemble of structures dictated by statistical thermodynamics. Because each HIV-1 RNA species can form an ensemble of structures [17,28], within a given RNA species, some RNA molecules will have the structures that enable packaging while the rest do not (Figure 2). Consequently, the division of function is not absolute among the RNA species; rather, each RNA species can serve both functions but with different efficiencies. For example, a larger portion of 1G RNA, relative to 3G RNA, folds into structures that expose DIS and the major Gag binding sites, allowing for the selective packaging of 1G RNA (Figure 3). This mechanism provides a basis for HIV-1 to use only a subpopulation of the unspliced RNA for packaging, leaving the other RNA species to carry out translation, thereby fulfilling both functions required for viral replication. Indeed, we have shown that HIV-1 replication can take place without the expression of either 1G RNA or 3G RNA [18]. Thus, neither RNA species is absolutely required for viral replication [18,37]. However, these mutants of HIV-1 have replication fitness defects compared with wild-type HIV-1 that expresses both major transcripts [18]. These findings indicate that both 1G and 3G RNA species are needed for optimal viral replication [18].

Fig. 2. Illustration showing how structural ensemble differences between RNAs may dictate their fate.

In these hypothetical principal component graphs, each dot represents an RNA conformation. Open dots represent conformations not favorable for packaging but favorable for translation and vice versa for closed dots. The ensemble of RNA structures adopted by 3G RNAs (left) and 1G RNAs (right) is different, which in turn dictates their different fates. RNA secondary structure schematics are drawn according to Nikolaitchik et al. [17].

Fig. 3. Structure-based mechanism for selective HIV-1 1G RNA packaging.

The HIV-1 5'UTR folds into structural ensembles wherein the 1G RNA is shifted toward packageable conformations with an intact polyA hairpin (green), exposed DIS sequence (blue), and exposed NC/Gag binding sites (yellow stars). In contrast, the 3G RNA preferentially adopts nonpackageable structures with a destabilized polyA hairpin that participates in long-range interactions with downstream sequences, and the DIS (blue) and NC/Gag binding sites (yellow dots) are sequestered. RNA secondary structure schematics are drawn according to Nikolaitchik et al. [17]. NC/Gag binding sites are illustrative and are not meant to be comprehensive.

Future Studies

Significant progress has been made in understanding the role of transcription start site heterogeneity in HIV-1 5'UTR structure and function; however, numerous open questions remain. How do the structural ensemble differences between 1G and 3G RNAs affect Gag binding affinity, stoichiometry, and conformation? What about differential interactions with other factors such as host cell RNA helicase A, the tRNA^Lys,3 reverse transcription primer, and human lysyl-tRNA synthetase, which all recognize different elements within the primer-binding site region of the 5'UTR [38]. Do differences in RNA nucleoside modifications regulate structural/functional differences between 1G and 3G RNAs? Does cap hypermethylation observed in some HIV-1 RNA transcripts [39,40] depend on the number of 5' guanosines? What is the mechanism of the observed differences in translational efficiency between 1G and 3G RNAs? Are the 1G and 3G RNAs differentially localized in the cell? These are among the questions that will be important for future studies to address.