Unpaired Guanosines in the 5′ Untranslated Region of HIV-1 RNA Act Synergistically To Mediate Genome Packaging

HIV-1 must package its RNA genome during virus assembly to generate infectious viruses. To better understand how HIV-1 packages its RNA genome, we examined the roles of RNA elements identified as binding sites for NC, a Gag-derived RNA-binding protein. Our results demonstrate that binding sites within stem-loop 1 of the 5′ untranslated region play important roles in genome packaging. Although mutating one or two NC-binding sites caused only mild defects in packaging, mutating multiple sites resulted in severe defects in genome encapsidation, indicating that unpaired guanosines act synergistically to promote packaging. Our results suggest that Gag-RNA interactions occur at multiple RNA sites during genome packaging; furthermore, there are functionally redundant binding sites in viral RNA.

ABSTRACT

The viral protein Gag selects full-length HIV-1 RNA from a large pool of mRNAs as virion genome during virus assembly. Currently, the precise mechanism that mediates the genome selection is not understood. Previous studies have identified several sites in the 5′ untranslated region (5′ UTR) of HIV-1 RNA that are bound by nucleocapsid (NC) protein, which is derived from Gag during virus maturation. However, whether these NC binding sites direct HIV-1 RNA genome packaging has not been fully investigated. In this report, we examined the roles of single-stranded exposed guanosines at NC binding sites in RNA genome packaging using stable cell lines expressing competing wild-type and mutant HIV-1 RNAs. Mutant RNA packaging efficiencies were determined by comparing their prevalences in cytoplasmic RNA and in virion RNA. We observed that multiple NC binding sites affected RNA packaging; of the sites tested, those located within stem-loop 1 of the 5′ UTR had the most significant effects. These sites were previously reported as the primary NC binding sites by using a chemical probe reverse-footprinting assay and as the major Gag binding sites by using an in vitro assay. Of the mutants tested in this report, substituting 3 to 4 guanosines resulted in <2-fold defects in packaging. However, when mutations at different NC binding sites were combined, severe defects were observed. Furthermore, combining the mutations resulted in synergistic defects in RNA packaging, suggesting redundancy in Gag-RNA interactions and a requirement for multiple Gag binding on viral RNA during HIV-1 genome encapsidation.

IMPORTANCE HIV-1 must package its RNA genome during virus assembly to generate infectious viruses. To better understand how HIV-1 packages its RNA genome, we examined the roles of RNA elements identified as binding sites for NC, a Gag-derived RNA-binding protein. Our results demonstrate that binding sites within stem-loop 1 of the 5′ untranslated region play important roles in genome packaging. Although mutating one or two NC-binding sites caused only mild defects in packaging, mutating multiple sites resulted in severe defects in genome encapsidation, indicating that unpaired guanosines act synergistically to promote packaging. Our results suggest that Gag-RNA interactions occur at multiple RNA sites during genome packaging; furthermore, there are functionally redundant binding sites in viral RNA.

INTRODUCTION

HIV-1 packages two copies of full-length viral RNA into their particles as genetic materials during virus assembly. Although viral RNA only occupies a minor portion of cytoplasmic mRNA, most viral particles contain two copies of full-length HIV-1 RNAs (1). Therefore, HIV-1 genome packaging is a highly regulated and efficient process. It is known that the viral structural protein Gag and the 5′ portion of the full-length viral RNA mediate genome packaging (2, 3). However, the molecular interactions between Gag and viral RNA that lead to efficient genome packaging are not well understood.

The HIV-1 Gag is translated as a polyprotein (4). During or soon after particle assembly, viral protease cleaves the Gag polyprotein into mature proteins, including matrix (MA), capsid (CA), nucleocapsid (NC), and p6, along with two spacer peptides, SP1 and SP2 (4). Mutation studies have shown that the NC domain in Gag is important for genome encapsidation (5–7). HIV-1 NC contains two conserved and zinc-chelating CCHC motifs, which are often referred to as zinc fingers or zinc knuckles. Mutations that alter the conserved CCHC motifs can cause severe defects in genome packaging (8–10). Furthermore, replacing NC with a leucine zipper interaction domain results in a mutant virus that can produce particles efficiently; however, these particles do not contain detectable levels of viral RNA (11, 12).

HIV-1 transcripts are driven from a single promoter in the U3 region of the 5′ long terminal repeat (LTR). HIV-1 transcripts can be fully spliced, partially spliced, or unspliced (full length) (13). Spliced transcripts express various viral proteins, including Env, Vif, Vpr, Vpu, Tat, Rev, and Nef. The unspliced full-length transcript has two functions: it serves as a template for translation to generate viral structural proteins and enzymes, Gag/Gag-Pol, and it is packaged into particles to serve as the viral genome. For simplicity, the full-length unspliced HIV-1 RNA is referred to as HIV-1 RNA here.

HIV-1 RNA has a long 5′ untranslated region (UTR) that is folded into complex structures; multiple studies have shown that elements in the 5′ UTR are important for RNA packaging (10, 14–23). Additionally, it has been shown that the sequence necessary and sufficient to mediate heterologous RNA to be packaged into HIV-1 particles resides in the 5′ UTR and extends into a portion of the gag gene (24). There are multiple proposed structures for the 5′ UTR of HIV-1 RNA (2, 18, 25–33). Although the precise base pairing may vary in different models, most models suggest the presence of multiple stem-loop structures along with regions of long-distance interactions. The general proposed structure of the 5′ UTR (shown in Fig. 1A) includes a stem-loop containing the transactivation response element (TAR), a stem-loop containing the poly(A) signal, primer binding site (PBS) with adjacent regions, stem-loop 1 (SL1) containing the dimerization initiation signal (DIS), stem-loop 2 (SL2) containing the major splice donor site (SD), stem-loop 3 (SL3), and a region near the Gag translation start site that engages in a long-distance interaction with a U5 region. It is generally accepted that the RNA structure of the 5′ UTR plays an important role in HIV-1 RNA packaging (2, 18, 25–30). Although most structural models depict a monomer of the HIV-1 RNA 5′ UTR, HIV-1 packages a dimeric RNA consisting of two copies of viral RNA. HIV-1 RNA dimerization occurs at the plasma membrane prior to genome packaging (1, 34, 35).

FIG 1.

Structure models of the 5′ untranslated region (5′ UTR) in the full-length HIV-1 RNA. (A) Schematic representation of the structural elements of the 5′ UTR. (B) NC binding sites identified using the SHAPE/reverse-footprinting approach (37). NC binding sites are outlined by red dotted lines. (C) Structures of the tandem three-way junctions identified using the NMR approach (41). For comparisons, in both panels B and C, the guanosines identified using the SHAPE approach are shown in red, whereas guanosines mutated in the 3way1 and 3way2 mutants are shown with yellow highlights. TAR, Transactivation response element; polyA, stem-loop containing polyadenylation signal; U5, unique 5′ region; PBS, primer binding site; SL1, 2, and 3, stem-loops 1, 2, and 3, respectively; DIS, dimerization initiation signal; SD and a black triangle, major splice donor site; AUG boxed in red, Gag translation start site. Nucleotide numbering is according to NL4-3 RNA.

Interactions between the HIV-1 RNA 5′ UTR and Gag or NC protein have been studied using various methods. Using selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE), the structure of the HIV-1 genome in virions has been described (36). Furthermore, the NC binding sites in the first ∼1 kb of the RNA genome including the 5′ UTR were mapped by comparing the virion RNA structures with and without the treatment of Aldrithiol-2 (AT-2) (37), a zinc ejecting agent that severely compromises the ability of NC to bind to RNA (38, 39). AT-2 treatment caused significant changes in several localized regions of RNA and altered their SHAPE reactivities to resemble those from the in vitro transcribed RNA without viral protein (37). Thus, this “reverse-footprinting” procedure identified sites that interact with HIV-1 NC. Most of the identified sites are clustered in the 5′ UTR sequence between PBS and the Gag AUG translational start site (nucleotide position 224 to 334 in NL4-3 RNA numbering). Interestingly, regions significantly altered by AT-2 treatment contain a similar structural motif that includes a single-stranded guanosine(s) flanked by stable helices. Based on the magnitude of change in the SHAPE reactivity upon AT-2 treatment, these sites were classified into two primary sites, two secondary sites, and three tertiary sites (Fig. 1B). The two primary sites exhibited the strongest effects upon AT-2 treatment and were located in the single-stranded bulges of SL1 (positions 239 to 244 and 271 to 275). Two secondary sites were located between PBS and SL1 (position 224 to 228) and immediately upstream of SL3 (position 308 to 312). Additionally, three tertiary sites were located in the loop of SL2, the loop of SL3, and immediately downstream of SL3 (positions 289 to 292, 318 to 320, and 326 to 329, respectively). In a separate report, a biochemical footprinting assay was performed using in vitro transcribed RNA and purified Gag protein (40); this study showed that Gag provided strong protection for the exposed guanosines in SL1, the same primary sites identified by the SHAPE studies.

Nuclear magnetic resonance (NMR) studies of in vitro-generated short transcripts suggested that dimerized HIV-1 5′ UTR formed a distinct tandem three-way junction structure (Fig. 1C) (41). These transcripts included a portion of U5 and sequences from downstream of the PBS to 10 nucleotides (nt) downstream of the gag AUG translational start site. In this proposed structure, SL2 does not form a hairpin; instead, SL2 sequences interact with the base of SL1, sequences upstream of SL1, and U5, to create a three-way junction (3way junction 2 in Fig. 1C). Additionally, the stem of SL3 is extended and the flanking nucleotides interact with U5 to form another three-way junction (3way junction 1 in Fig. 1C). It was proposed that the exposed guanosines near the two three-way junctions play important roles in RNA packaging. To examine this hypothesis, G-to-A substitutions were introduced to the exposed guanosines near the three-way junctions to abolish putative NC binding sites, and these mutants exhibited significant packaging defects. Mutating four guanosines near three-way junction 2, seven guanosines near three-way junction 1, and all 11 guanosines caused the RNA packaging efficiencies to reduce to 17%, 10%, and 5%, respectively, of the wild-type level (41). In a separate study using cross-linking-immunoprecipitation (CLIP) sequencing, the interactions between Gag and HIV-1 RNA were examined in cells and viral particles (42). Results from this strategy confirmed that the 5′ UTR is one of the regions Gag preferentially binds and showed that, in cells, most of the Gag binding sites were guanosine rich (42).

These studies provide strong evidence that single-stranded exposed guanosines in the HIV-1 5′ UTR play important roles in RNA packaging, likely by serving as Gag binding sites. However, it is unclear which guanosines and binding sites are important. There are overlaps between the exposed guanosines identified by the SHAPE/reverse-footprinting study and those in the tandem three-way junction structures. However, two primary NC binding sites identified in the SHAPE studies are located outside the tandem three-way junction structures identified by the NMR studies. Furthermore, the NC binding sites identified using the SHAPE/reverse-footprinting strategy have not been evaluated for their roles in genome packaging. Thus, the impact of these NC binding sites on HIV-1 RNA packaging has not been defined.

We sought to gain a better understanding of the interactions between HIV-1 RNA and Gag that lead to genome packaging. To identify RNA elements important for Gag recognition, we substituted the single-stranded guanosines identified by the SHAPE/reverse-footprinting studies with adenosines. We also generated G-to-A substitution mutants in which the proposed NC binding sites in the three-way junctions were abolished. We then examined the RNA packaging efficiencies using cell lines that harbored two HIV-1 proviruses, one with a mutated 5′ UTR and the other with the wild-type 5′ UTR. We developed quantitative allele-specific reverse transcription-PCR (RT-PCR) capable of distinguishing between wild-type and mutant RNAs and quantified the efficiencies of mutant RNA packaging. Our findings indicate that multiple exposed guanosines participate in mediating HIV-1 RNA packaging, and the primary NC binding sites identified using the SHAPE strategy play a major role in genome encapsidation.

RESULTS

System to study the roles of NC binding sites in HIV-1 RNA packaging.

To examine RNA elements important to HIV-1 packaging, we generated cell lines that harbored two HIV-1 proviruses, one with a wild-type 5′ UTR and another with a mutant 5′ UTR, and determined the mutant virus RNA packaging efficiency in competition with the wild-type virus. We used two previously described structurally similar NL4-3-based constructs, T6-Spe and H0 (Fig. 2A) (43–45). Both constructs contain all the cis-acting elements important for virus replication, including LTRs and the 5′ UTR. Additionally, both constructs contain the entire gag-pol gene and express Tat and Rev. The gag-pol gene in the H0 construct encodes wild-type functional proteins, whereas the gag-pol gene in T6-Spe contains a 4-nt frameshift insertion (shown as a hashtag in Fig. 2A) thereby abolishing the expression of functional Gag/Gag-Pol protein. We have previously shown that when supplemented with functional Gag protein in trans, T6-Spe RNA was efficiently packaged into viral particles (43, 45, 46). Thus, the 4-nt insertion in gag does not affect the ability of the T6-Spe RNA to be packaged. To identify cells harboring T6-Spe or H0 proviruses, a marker gene was inserted into the nef gene of each construct; T6-Spe harbors the mouse thy1.2 gene, whereas H0 harbors the mouse heat stable antigen (hsa) gene. Both Thy and HSA are cell surface proteins, and their expression can be detected by antibody staining using flow cytometry (44).

FIG 2.

Experimental approach used to study the role of NC binding sites in RNA packaging. (A) General structures of HIV-1 constructs used to study the roles of unpaired guanosines in 5′ UTR on genome packaging. #, inactivating 4-nt insertion in gag; *, mutations in 5′ UTR. (B) Experimental protocol. (C) Validation of the quantitative allele-specific RT-PCR method. x axis, the proportion of H0 template input into the experimental mixes; y axis, the percentage of H0 template detected using allele-specific RT-PCR method.

To study the effects of abolishing NC binding sites in the HIV-1 5′ UTR, we introduced various mutations into the 5′ UTR of the H0 construct (shown as H0-mutant in Fig. 2A). We then generated viruses containing RNA genomes derived from T6-Spe, infected human 293T cells at a low multiplicity of infection (MOI) (<0.1), and enriched the infected cells by cell sorting (Fig. 2B). We then infected these cells with a second virus, H0 containing mutated 5′ UTR, at a low MOI (<0.1) and enriched the doubly infected cells by repeated sorting until >95% of the cells expressed both proviruses (Thy⁺ HSA⁺). We isolated the cytoplasmic RNA from the cells to examine the expression of the two proviruses and harvested viruses from these cells for virion RNA isolation to determine the packaging efficiencies. The gag gene of T6-Spe, but not H0, contains a 4-nt insertion. Using this polymorphic site, we established a quantitative allele-specific RT-PCR assay using two sets of primers, one set detected the gag sequence with the 4-nt insertion (T6-Spe) and the other set detected wild-type gag sequence (H0 or H0-mutant). We generated a series of mixtures containing various ratios of T6-Spe and H0 plasmids and performed quantitative allele-specific RT-PCR; our results showed that this assay can reliably detect various amounts of H0 sample in the mixture as low as 1% (Fig. 2C).

Using this assay, we quantified the cytoplasmic and virion HIV-1 RNAs to examine the expression of the two proviruses and the amounts of RNAs packaged into particles. Based on these values, we determined the proportion of H0 RNA in total cytoplasmic or viral HIV-1 RNA. By comparing the ratios of H0 RNA in the virion and in the cytoplasm, we determined the RNA packaging efficiency of the mutant virus. If the H0 RNA was packaged at the same efficiency as the T6-Spe RNA, the RNA packaging efficiency would be 1.

Examining the roles of NC binding sites in the HIV-1 5′ UTR on RNA genome packaging.

Using the SHAPE/reverse-footprinting approach, multiple NC binding sites have been identified in the HIV-1 5′ UTR (37). To study the importance of these NC binding sites on mediating RNA packaging, we introduced mutations in the 5′ UTR sequence of the H0 construct. We replaced the guanosines in both primary sites (nucleotide positions 240, 241, 272, and 273) with adenosines to generate construct H0-1site (Fig. 3). Similarly, construct H0-2site contains G-to-A substitutions in both secondary sites (nucleotide positions 224, 226, and 310). One of the tertiary sites overlaps the HIV-1 major splice donor site, and the guanosines cannot be mutated without altering the function of this important cis-acting element. As a result, we generated H0-3*site, in which we introduced G-to-A mutations in two of the tertiary sites (positions 318, 320, 328, and 329) but left the major splice donor site intact. We have performed RNA folding studies to examine the effects of the G-to-A substitutions on predicted RNA structures (47, 48); our analyses showed that the general structures of these mutant 5′ UTRs have not changed from that of wild-type RNA (see Materials and Methods). We then generated viruses containing these mutant genomes, infected cells, and generated cell lines containing T6-Spe and one of the mutants.

FIG 3.

RNA packaging efficiency of mutants with G-to-A substitutions in the NC binding sites determined by SHAPE/reverse-footprinting. (A) Locations of G-to-A substitutions in the HIV-1 5′ UTR. Red asterisks denote substitutions. (B) RNA packaging efficiencies of NC-binding site mutants. Average packaging efficiency of HIV-1 RNA derived from H0 constructs containing the wild-type 5′ UTR was set as 100%. Averages and standard deviations (error bars) from at least three experiments are shown. Black asterisks denote statistical significance by unpaired one-way ANOVA with Bonferroni adjustment: *, P < 0.01; **, P < 0.001.

To quantify RNA derived from the two proviruses, we first performed allele-specific RT-PCR using a control cell line harboring two proviruses, both containing the wild-type 5′ UTR, H0 and T6-Spe. Our results showed that H0 RNA occupied 69.6% ± 6.3% of the total HIV-1 RNA in the cytoplasm and 77.5% ± 1.9% of the HIV-1 RNA in the virion. Thus, H0 RNA was packaged at a similar efficiency as the T6-Spe RNA; these results are consistent with our published findings that the 4-nt insertion in gag does not affect genome packaging based on an RT-PCR–sequencing method (43, 46). The packaging efficiency of the H0 RNA was calculated as the proportion of H0 RNA in the virion divided by the proportion of H0 RNA in the cytoplasm. For simplicity, the average packaging efficiency of wild-type H0 RNA was set as 100% (Fig. 3B).

We then examined the effects of various mutations in the 5′ UTR on viral RNA expression and found RNAs from H0-1site, H0-2site, and H0-3*site proviruses were expressed equally well in the cytoplasm (Fig. 4A). However, H0-1site RNA and H0-3*site RNA both exhibited significantly reduced packaging efficiencies, 60.0% ± 2.8% and 74.5% ± 6.3% of the H0 wild-type (H0) level, respectively (Fig. 3B) (P < 0.0001 in both cases). All statistical analyses were performed using an unpaired one-way analysis of variance (ANOVA) with Bonferroni correction for multiple comparisons. In contrast, mutating the secondary sites had less effect on HIV-1 RNA packaging; H0-2site RNA was packaged at 87.7% ± 7% of the wild-type level (P < 0.001). Thus, although mutating only the primary, secondary, or tertiary NC binding sites reduced RNA packaging efficiencies, these effects were mild and within 2-fold of the wild-type level.

FIG 4.

Characterization of viral RNA and protein expression in cell lines that harbor two proviruses. (A) Relative levels of cytoplasmic RNAs expressed by two proviruses. Viral RNAs expressed from H0 or H0-derived proviruses are shown in dark gray, whereas RNAs expressed from T6-Spe proviruses are shown as light gray. Averages and standard deviations (error bars) from at least three experiments are shown. (B) Representative images of a Western blot of cellular lysates probed with anti-p24^CA and anti-GAPDH antibodies (top and bottom, respectively). (C) Representative image of anti-p24^CA Western blot of viral particles released from cell lines.

Of these mutants, H0-1site had the strongest effect. Therefore, we examined the effects of combining primary site mutations with either secondary or tertiary site mutations and generated H0-1,2sites and H0-1,3*sites, respectively. Quantitative allele-specific RT-PCR showed that these mutations did not have a negative impact on cytoplasmic RNA expression (Fig. 4A) but affected RNA packaging efficiencies: RNAs from H0-1,2sites and H0-1,3*sites were packaged at 15.1% ± 1.9% and 7.7% ± 3.1% of the wild-type level, respectively (P < 0.0001 in both cases). Even though mutations of secondary sites had little effect on RNA packaging efficiencies by themselves, they exerted negative effects on RNA packaging when combined with primary sites mutations. Similarly, combining mutations in primary and tertiary sites appeared to have more-than-additive effects. If these effects were additive, the H0-1,2sites mutant RNA should be packaged at ∼52.6% of the wild-type level (60.0% × 87.7%); however, H0-1,2*sites RNA was packaged at 15.1% of the wild-type level. Similarly, with an additive effect, the H0-1,3*sites mutant RNA should be packaged at ∼44.7% of the wild-type level (60.0% × 74.5%); however, H0-1,3*sites RNA was packaged at 7.7% of the wild-type level. Thus, combining these mutations resulted in synergistic defects in RNA packaging.

To further investigate the effects of combining mutations, we constructed H0-2,3*sites and H0-1,2,3*sites and then analyzed the cytoplasmic and virion RNA samples from cell lines harboring T6-Spe and one of the mutant H0 proviruses. We found that secondary site mutations also had a negative effect on RNA packaging when combined with 3*site mutations; H0-2,3*sites RNA was packaged at 40.5% ± 4.8% of the wild-type level, significantly lower than that for H0-3*site (P < 0.0001). Combining primary, secondary, and tertiary site mutations generated a mutant with the most severe defect; H0-1,2,3*sites RNA was packaged at 2.9% ± 0.9% of the wild-type level, a nearly 40-fold decrease from the virus with a wild-type 5′ UTR. The effects of combining these mutations were also more than additive; if the effects were additive, RNA from H0-2,3*sites would have been packaged at 65.4% of the wild-type level (87.7% × 74.5%) and RNA from H0-1,2,3*sites would have been packaged at 39.2% of the wild-type level (60.0% × 87.7% × 74.5%). Therefore, the defect in RNA packaging was synergistic when combining these mutations. Furthermore, among the primary, secondary, and tertiary NC binding sites we tested, mutations in the primary sites exert a more severe defect, indicating that the single-stranded guanosines in the primary sites located in the bulges of SL1 play an important role for RNA packaging. In the above-described tested mutants, the single-stranded guanosines were changed to adenosines, and these mutations were not expected to affect the overall RNA structures. To verify this, we have performed free energy calculations to predict RNA structures and found that the 11 G-to-A substitutions in the most severe mutant, H0-1,2,3*sites, should not alter the secondary structure of the RNA (47, 48).

As shown in Fig. 4A, RNAs derived from mutants containing substitutions in sites 1, 2, and/or 3 were expressed at similar levels in the cytoplasm. In addition, we examined the protein expression from these mutants. Because T6-Spe contains a frameshift mutation in gag, Gag/Gag-Pol proteins are expressed from the H0-derived proviruses in these cell lines. We performed Western analyses of cellular Gag expression and virus production from these cell lines (Fig. 4B and C). Our analyses showed that these mutants express Gag and produce viral particles at a level similar to that of H0 virus containing wild-type 5′ UTR. Thus, the 5′ UTR mutations do not affect RNA or protein expression by these mutants.

Examining the effects of mutating exposed guanosines of the tandem three-way junctions in the HIV-1 5′ UTR.

Our studies indicate that replacing the single-stranded guanosines in the NC binding sites identified using the SHAPE strategy has significant effects on HIV-1 RNA packaging efficiencies. Furthermore, the primary sites play a more important role in mediating RNA packaging. The guanosines in primary sites are located outside the tandem three-way junction structures identified by NMR studies (41). To directly compare the effects of mutating the guanosines near the three-way junctions, we constructed H0 mutants that contained G-to-A substitutions near the three-way junction 1 (H0-3way1), three-way junction 2 (H0-3way2), and both three-way junctions (H0-3way1,2). These mutations were previously predicted to not affect RNA secondary structure (41). We generated cell lines harboring one T6-Spe provirus and one mutant provirus and then examined RNA expression and genome packaging using these cells. Similar to that in the mutants described in Fig. 3, G-to-A substitutions did not have negative effects on the cytoplasmic RNA expression of the 3way mutants (Fig. 4A). Our analyses of the virion RNA showed that H0-3way1 and H0-3way2 mutant RNAs were packaged at 33.8% ± 5.3% and 96.2% ± 5.4% of the wild-type H0 level, respectively (P < 0.0001 and P = 0.12, respectively) (Fig. 5). Thus, RNA packaging was significantly reduced by mutations abolishing the NC binding site near three-way junction 1 but not by mutations abolishing the binding site near three-way junction 2.

FIG 5.

RNA packaging efficiency of mutants with G-to-A substitutions in the proposed tandem three-way junctions determined by NMR. (A) Locations of G-to-A substitutions in the 3way1 and 3way2 mutants. Red asterisks denote substitutions. (B) RNA packaging efficiencies of 3-way junction mutants. Average packaging efficiency of HIV-1 RNA derived from the H0 construct containing the wild-type 5′ UTR was set as 100%. Averages and standard deviations (error bars) from at least three experiments are shown. Black asterisks denote statistical significance by unpaired one-way ANOVA with Bonferroni adjustment: **, P < 0.001; n.s., not statistically significant.

We then analyzed the RNA from virions generated by cells harboring H0-3way1,2 provirus and T6-Spe provirus. Despite being well expressed in the cytoplasm (Fig. 4A), our analyses showed that H0-3way1,2 RNA was packaged at 17.4% ± 4% of the wild-type level. Thus, although mutations in 3way2 did not cause a significant defect in RNA packaging by themselves, they led to a further decrease in RNA packaging when combined with 3way1 mutations (P < 0.001 comparing 3way1,2 to 3way1). Furthermore, combining these mutations also had a synergistic effect on RNA packaging efficiency (expected RNA packaging efficiency of a combination mutant would be 33.8% × 96.2% = 32.5%).

Thus far, the packaging efficiencies of the mutant genomes were measured in competition with an HIV-1 RNA containing the wild-type 5′ UTR. To examine the packaging efficiency of a mutant RNA without a competing wild-type HIV-1 RNA, we generated three cell lines, each containing only a single provirus, H0, H0-1,2,3*site, or H0-3way1,2. H0 contains the wild-type 5′ UTR and is used as a control for RNA packaging, whereas both H0-1,2,3*site and H0-3way1,2 exhibit severe packaging defects in our competition assay. We harvested viruses produced from these cell lines, determined virus production by Western analyses, measured the amounts of virion RNA using quantitative RT-PCR, normalized the RNA values to the amount of viral capsid, and calculated the mutant RNA packaging efficiencies. Our results show that H0-1,2,3*site and H0-3way1,2 mutant RNAs were packaged at 1.8% ± 0.2% and 8.1% ± 1.4% of the wild-type H0 levels, which are close to the RNA packaging efficiencies measured using the competition assay, 2.9% and 17.4%, respectively (Fig. 3B and 5B). Therefore, these two mutants displayed severe packaging defects regardless of whether a wild-type RNA was present.

Taken together, these studies showed that replacing as few as four exposed guanosines with adenosines in HIV-1 5′ UTR can cause significant defects in RNA genome packaging. However, mutating some guanosines had much more pronounced effects than others. When comparing mutants with similar numbers of G-to-A substitutions, mutating the primary sites on the SL1 stem caused the most severe defects in RNA packaging, either by themselves or when combined with other mutations (see Discussion). These results confirm that although many guanosines in the 5′ UTR are important, the exposed guanosines in SL1 play a major role in HIV-1 RNA packaging. Of all the sites we tested, combining mutations had synergistic effects regardless of whether the binding sites were identified using SHAPE or NMR. These findings indicate that there is functional redundancy in the Gag-RNA interactions leading to RNA packaging (see Discussion).

DISCUSSION

HIV-1 must package its genome during virus assembly to generate infectious particles. To carry out this step, Gag interacts with viral RNA to mediate genome packaging. In the absence of HIV-1 RNA, Gag packages cellular mRNA in a near-random manner (49). However, in the presence of HIV-1 RNA, a vast majority of the HIV-1 particles contain viral RNA (1). HIV-1 RNA is a minor RNA species among the abundant cellular mRNAs; thus, specific Gag-RNA interactions must occur to allow efficient packaging of the viral genome. Using various methods, Gag or NC binding sites in the 5′ UTR of the HIV-1 RNA have been mapped; however, for many sites, their roles in mediating RNA packaging have not been examined. In this report, using G-to-A substitution to alter the exposed guanosines, we mutated previously mapped NC binding sites and determined the RNA packaging efficiencies of these mutants. These studies reveal the role of previously identified NC binding sites in RNA packaging and provide insights into the mechanisms of viral RNA genome encapsidation.

We have examined two series of NC binding site mutants: those identified using SHAPE and those identified using NMR. To assess the relative importance of these sites, we compared RNA packaging efficiencies of mutants with the same number of guanosine substitutions. Mutants H0-1site, H0-3*site, and H0-3way2 each contained four guanosine substitutions; these RNAs were packaged at 60%, 74.5%, and 96.2% of the wild-type level, respectively (Table 1). There are three mutants, H0-1,2sites, H0-2,3*sites, and H0-3way1, that contained seven guanosine substitutions; their RNAs were packaged at 15.1%, 40.5%, and 33.8% of the wild-type level, respectively. There are two mutants with 11 guanosine substitutions, H0-1,2,3*sites and H0-3way1,2; their RNAs were packaged at 2.9% and 17.4% of the wild-type level, respectively. In all three groups, the most defective mutant within the group contains the H0-1site mutations. Thus, mutants with H0-1site mutations either alone or when combined with other mutations, caused significantly more-severe defects (Table 1). From these results, we conclude that the primary sites identified by the SHAPE/reverse-footprinting approach play a major role in RNA packaging.

TABLE 1.

Comparisons of RNA packaging efficiencies from mutants with the same number of G-to-A substitutions

Mutant	No. of Gs mutated	RNA packaging efficiency (% of WT)	P value compared to the most defective mutant in the groupa
3way2	4	96.2 ± 5.4	<0.0001
3*site	4	74.5 ± 6.3	<0.0001
1site	4	60.0 ±2.8
2,3*site	7	40.5 ± 4.8	<0.0001
3way1	7	33.8 ± 5.3	<0.0001
1,2site	7	15.1 ± 1.9
3way1,2	11	17.4 ± 4.0	<0.0001
1,2,3*site	11	2.9 ± 0.9

^aP values were determined using unpaired one-way ANOVA with Bonferroni adjustment.

Among the G-to-A substitution mutants we tested, altering three to four exposed guanosines resulted in mild defects in RNA packaging, including in H0-1site in which we changed the primary NC binding sites in SL1. Combining mutations resulted in synergistic RNA packaging defects (Fig. 3 and 5). These results indicate that there are multiple Gag binding sites in HIV-1 5′ UTR; additionally, functional redundancy exists in the Gag-RNA interactions leading to genome packaging. We hypothesize that to package the viral genome, multiple Gag proteins, each binding to a different site, interact with HIV-1 RNA 5′ UTR. Furthermore, there are more binding sites than required to mediate genome packaging. Thus, when removing one or two binding sites in the 5′ UTR, multiple Gag proteins can still bind to the remaining sites to mediate packaging, thereby resulting in mild defects. However, when more sites are removed, the process becomes far less efficient, resulting in synergistic defects in RNA packaging. Such functional redundancy has been observed in other retroviruses. It was shown that in murine leukemia virus (MLV), the preferred NC binding site is in a UCUG-UR-UCUG context, and there were two sets of such UCUG motifs. Mutating the guanosines in one set of UCUG motif resulted in a 4- or 12-fold defect, whereas mutating both sets of motifs resulted in a 200-fold effect (50). Thus, it is possible that such functional redundancy is a common feature of retroviral RNA packaging. Although exposed guanosines play important roles in both MLV and HIV-1 NC binding sites, the requirements for surrounding sequences are different for the NC proteins of these two viruses. As described above, the MLV NC binding site has a conserved sequence motif (50). Sequence analyses of the HIV-1 NC binding sites revealed by SHAPE/reverse-footprinting did not show strict sequence conservation; however, there may be conservation at the structural level, as the nucleotides downstream of the exposed guanosines tend to be double stranded (37). Further studies are required to understand how HIV-1 Gag recognizes its binding sites on the viral RNA.

In this study, we have examined the NC binding sites identified using either the SHAPE or the NMR approach. These two studies proposed distinct RNA structures of the HIV-1 5′ UTR; however, there are significant overlaps among the exposed guanosines in the NC binding sites mapped by the two studies (Fig. 1B and C). Although we found that the exposed guanosines in SL1 are important to RNA packaging, it is difficult to deduce RNA structures from our studies. We have studied the G-to-A substitutions in the binding sites proposed by the tandem three-way junction structure. Most of our results agree with the published data with some differences. Both studies showed that G-to-A substitutions in the three-way junction 1 caused a more severe defect than those in the three-way junction 2, and combining both sets of mutations caused a more severe packaging defect than each set of mutations alone. In the previous report, a mutant that contained a G-to-A substitution of the three-way junction 2 had a significant defect in RNA packaging (17% of the wild-type level). However, our study showed that the RNA from H0-3way2 mutant was packaged at 96% of the wild-type level. It is unclear why we observed different effects when mutating three-way junction 2. One possibility is that the observed differences may be due to the distinct measuring methods used. In our system, we analyzed RNA derived from proviruses in dually infected cell lines, whereas the previous studies expressed RNAs from transiently transfected plasmids, in which transfection efficiencies could affect results. Additionally, we used quantitative allele-specific RT-PCR to measure RNA levels, whereas the previous studies used an RNase protection assay to quantify RNA.

The HIV-1 genome is packaged as a dimer into viral particles (51). The dimerization initiation signal (DIS) is an important element that mediates the selection of a copackaged genome and the initiation of RNA dimerization (1, 35). The DIS element was not altered in the mutants we tested in this report. It was previously reported that substitutions of the guanosines near three-way junction 1, 2, or both were predicted to not affect RNA secondary structure (41). Similarly, our RNA folding studies also showed that G-to-A substitutions in the NC binding sites described in the SHAPE/reverse-footprint studies did not change the general structures of these mutant 5′ UTRs. Despite the maintenance of the DIS and the 5′ UTR structure, the loss of Gag binding sites could potentially affect RNA dimer stability. Using a live-cell imaging approach, we have previously shown that HIV-1 RNA dimerizes on the plasma membrane and Gag binding stabilizes RNA dimers (34). Thus, the loss of Gag binding sites may impact the stability of the RNA dimer. In future experiments, it will be interesting to examine the relationship between the stability of the dimer and the ability of the RNA to be packaged.

In this report, we determined the effects of mutating the exposed guanosines in NC binding sites at the HIV-1 5′ UTR and identified the guanosines that are most important for Gag-RNA interactions. Additionally, these studies demonstrate that multiple guanosines act synergistically to mediate RNA packaging, an important mechanistic insight into the HIV-1 genome encapsidation process.

MATERIALS AND METHODS

Plasmid construction.

HIV-1 constructs ON-H0 and T6-RRE-Spe were previously described (44, 45); for simplicity, they are referred to as H0 and T6-Spe, respectively, in this report. Briefly, these constructs were derived from molecular clone NL4-3; they contain all cis-acting elements required for viral replication and express Tat and Rev. Both H0 and T6-Spe encode the gag-pol gene; T6-Spe contains a 4-nt inactivating insertion in gag, whereas H0 expresses functional Gag/Gag-Pol. H0 harbors a mouse hsa gene and T6-Spe harbors a mouse thy1.2 gene in nef. Both H0 and T6-Spe also contain an internal ribosomal entry site from encephalomyocarditis virus and a mutated green fluorescent protein gene in the nef gene; for simplicity, these sequences are not illustrated in Fig. 2. Mutations in the 5′ UTR were introduced by cloning a synthesized DNA fragment (gblocks; IDT) into the H0 construct using either a NEBuilder HiFi DNA Assembly kit (NEB) or standard molecular cloning techniques. The structure of all cloned plasmids was verified by restriction digest, and the regions that underwent PCR amplification were verified by DNA sequencing to avoid introduction of inadvertent mutations.

Cell culture, generation of cell lines, flow cytometry, and Western blotting.

Human embryonic kidney 293T cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 5% calf serum, 5% fetal bovine serum (FBS), 50 U/ml penicillin, and 50 μg/ml of streptomycin (Life Technologies). Cells were maintained in humidified 37°C incubators with 5% CO₂. To produce viruses for generating transduced cell lines, 293T cells were transfected with an HIV-1 construct and two helper plasmids: pCMVΔ8.2 that expresses HIV-1 Gag/Gag-Pol along with all of the accessory proteins, and pHCMV-G that expresses envelope G protein from vesicular stomatitis virus (VSV G) (52, 53). Viruses were collected 24 to 48 h after transfection, serially diluted, and used to infect fresh 293T cells. Human 293T cells were first infected with T6-Spe virus at an MOI of <0.1, and infected cells were enriched by cell sorting. These singly infected cells were then infected with an H0-derivative virus at an MOI of <0.1. Dually infected cells were enriched by repeated sorting until >95% of cells expressed both Thy and HSA markers (44). To generate cell lines containing a single provirus, 293T cells were infected with VSV G-pseudotyped virus at a low MOI (<0.1), and infected cells were enriched by repeatedly sorting until >97% of cells expressed HSA marker. Each cell line contained at least 42,000 independent infection events.

The proportions of cells infected by HIV-1 constructs were determined by flow cytometry by detecting the expression of cell surface markers HSA and Thy. Cells were collected, washed twice with Dulbecco’s phosphate-buffered saline (DPBS) containing 2.5% FBS, and then stained with phycoerythrin-conjugated anti-HSA (BioLegend) and/or allophycocyanin-conjugated anti-Thy1.2 antibodies (eBioscience) for 30 min at 4°C, followed by two additional washes with DPBS containing 2.5% FBS. Flow cytometry analyses were performed using an LSR II system (BD Biosciences), and cell sorting was performed on a FACSAria II system (BD Biosciences). Flow cytometry data were analyzed using FlowJo software (FlowJo, LLC).

To examine Gag expression, provirus-containing cells were washed with phosphate-buffered saline and lysed on plates by using CelLytic solution (Sigma) containing cOmplete ULTRA protease inhibitor cocktail (Roche). To examine viral particles released from the provirus-containing cells, supernatants were harvested, clarified through a 0.45 μm-pore size filter, centrifuged at 17,000 × g for 1 h at 4°C, and lysed in CelLytic solution (Sigma). Immunoblots were probed for HIV-1 Gag with mouse anti-p24^CA antibody, a generous gift from Michael Malim, obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH (54), followed by secondary goat anti-mouse antibody (IRDye-680RD; LI-COR). In some experiments, rabbit antibody against host protein glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used (ab128915; Abcam) followed by a secondary goat anti-rabbit antibody (IRDye-800CW; LI-COR). Western blots were imaged and quantitated using the Odyssey infrared imaging system and Image Studio Lite software (LI-COR).

RNA isolation and quantification, RNA structure modeling, and statistics.

Viruses produced by transduced cell lines were harvested and cleared through a 0.45-μm filter (Millex); virion RNA was isolated using a Qiagen Mini Viral RNA kit according to the manufacturer’s instructions. Cytoplasmic RNA from producer cells was extracted using a PARIS kit (Invitrogen). Both virion and cytoplasmic RNAs were treated with DNase (Turbo DNase-free kit; Ambion) to remove potential DNA contamination. RNA was used for quantitative allele-specific RT-PCR immediately or stored at −80°C.

Quantitative allele-specific RT-PCR was performed with iTaq SYBR green one-step RT-PCR kit from Bio-Rad. Allele-specific primers Spe-WT (5′-ATAGCAGGAACTACTAGTACC-3′) and Spe-Mut (5′-ATAGCAGGAACTACTAGCTAG-3′) were designed to anneal to the sequence around the intact SpeI restriction site in gag gene of H0 and the mutated SpeI site in T6-Spe, respectively. Primer Spe-R (5′-CTACTGGGATAGGTGGATTA-3′) annealed to both templates and was used as a primer for reverse transcription as well as for quantitative PCR. Viral or cytoplasmic RNAs were serially diluted for quantitative allele-specific RT-PCR, and only the data points within the linear range were used for further calculations. Quantitative RT-PCR was performed using the following conditions: 10 min at 50°C, 1 min at 95°C, 45 cycles of 15 s at 95°C, 30 s at 62°C, and acquisition at 70°C. At the end of each quantitative PCR run, a melting curve (60°C for 1 min with 0.11°C ramping to 95°C) was obtained to confirm the correct product size. Reactions using the Spe-Mut primer were specific to T6-Spe and did not detect H0 template; however, reactions using Spe-WT primer detected T6-Spe template at ∼0.5% to 3.6% efficiencies. Therefore, each quantitative RT-PCR run included appropriate controls to measure cross-reactivity levels, and these numbers were used to adjust the template copy numbers. Using this approach and templates generated from mixing H0 and T6-Spe plasmids, we found that the assay can reliably and reproducibly detect H0 template when it is ≥1% of the mixture.

To calculate the proportion of H0 RNA in the total HIV-1 RNA, the copy number of the H0 RNA was divided by the copy number of H0 RNA plus T6-Spe RNA [H0/(H0 + T6-Spe)]. RNA packaging efficiency was calculated by dividing the proportion of the H0 RNA in the virion by the proportion of H0 RNA in the cytoplasm (% H0 in virion/% H0 in cytoplasm). At least three independent measurements of RNA packaging efficiencies were performed for each mutant.

Statistical analyses were performed in GraphPad Prism v8.3.1. RNA packaging efficiencies were compared using an unpaired one-way ANOVA with Bonferroni correction for multiple comparisons. Comparisons with P values less than 0.05 were considered statistically significant. RNA secondary structures of wild-type (NL4-3) and mutant 5′ UTR sequences (RNA nucleotides 1 to 350) were predicted using the RNAstructure v6.0.1 Fold web server (47).