A Heterologous, High-Affinity RNA Ligand for Human Immunodeficiency Virus Gag Protein Has RNA Packaging Activity

Abstract

Retroviral RNA encapsidation depends on the specific binding of Gag proteins to packaging (ψ) signals in genomic RNA. We investigated whether an in vitro-selected, high-affinity RNA ligand for the nucleocapsid (NC) portion of the Gag protein from human immunodeficiency virus type 1 (HIV-1) could mediate packaging into HIV-1 virions. We find that this ligand can functionally substitute for one of the Gag-binding elements (termed SL3) in the HIV-1 ψ locus to support packaging and viral infectivity in cis. By contrast, this ligand, which fails to dimerize spontaneously in vitro, is unable to replace a different ψ element (termed SL1) which is required for both Gag binding and dimerization of the HIV-1 genome. A single point mutation within the ligand that eliminates high-affinity in vitro Gag binding also abolishes its packaging activity at the SL3 position. These results demonstrate that specific binding of Gag or NC protein is a critical determinant of genomic RNA packaging.

The process by which retroviral genomic RNAs are targeted into virions during particle assembly is known as RNA packaging, or encapsidation. It is of interest not only because of its essential role in the retroviral life cycle and its utility for designing gene transfer vectors but also as a useful model for studying RNA recognition and trafficking within cells. Packaging is highly specific, in that genomic RNA accounts for a much greater proportion of total RNA within virions than in the cytoplasm where those virions formed; this implies that viral genomic RNA is packaged in preference to cellular RNAs and spliced viral RNAs, which tend to be excluded (for a review, see references 4 and 12).

The specificity of packaging is dictated in part by sequences in the viral polyprotein Gag and particularly in its C-terminal proteolytic derivative, the nucleocapsid (NC) protein. Zinc finger-like domains, found in nearly all known retroviral NC proteins, appear especially critical in this regard. The NC protein of human immunodeficiency virus type 1 (HIV-1), for example, includes two such zinc fingers, and mutations in or around those domains can eliminate specific packaging (1, 7, 15, 18, 35, 38), whereas genetically replacing the entire NC sequence of HIV-1 with that of the Moloney murine leukemia virus, or vice versa, yields a chimera that preferentially packages RNA from the virus that contributed its NC domain (7, 38).

The viral genomic RNAs, in turn, are recognized by virtue of cis-acting elements called packaging signals, or ψ sites, which are defined experimentally as sequences necessary and/or sufficient to target RNAs into the virions of a given virus (reviewed in reference 4). The best-characterized ψ sites have been found to coincide with one or more short RNA stem-loops formed near the 5′ end of the genome, typically at locations immediately upstream or downstream of the major splice donor, which are believed to serve as binding sites for NC proteins (1, 5, 6, 8–10, 13, 14, 17, 19–21, 25, 27, 29, 30, 36). For example, the portion of the HIV-1 ψ locus depicted in Fig. 1A includes at least three such stem-loops, designated SL1, SL3, and SL4; these each bind directly and specifically to HIV-1 Gag or NC protein in vitro, with apparent dissociation constants (K_d) of roughly 200 nM individually or 50 nM as a group, and are each required for optimal packaging in vivo (8–10, 14, 29, 30). The overall fidelity and efficiency of HIV-1 packaging appear to reflect the combined effects of multiple discrete ψ stem-loops binding to multiple Gag subunits within each capsid particle as it forms.

FIG. 1.

(A) (Top) Sequence and putative secondary structures of the SELEX ligand (SW8.4) and a point mutant (SW8.4pt.mut.) used in this study. SW8.4 comprises the 37-base Gag-binding core of a 97-base ligand originally derived through SELEX (26). A seven-base region of identity between SW8.4 and another group's optimal SELEX ligand is indicated with open lettering (3, 26). The transposed C residue in the point mutant is indicated with an asterisk. Arrows indicate the positions where these ligands were inserted into the HIV-1 genome, exactly replacing the entire stem and loop of SL1 or SL3 without altering adjacent residues. (Bottom) Secondary structure in a portion of the HIV-1 ψ locus from residues 689 to 813 of strain HXB2 (8). Locations of the major 5′ splice donor (SD) and the gag start codon (open lettering) are indicated. (B) Nitrocellulose filter-binding assay with radiolabelled RNAs and glutamine S-transferase-Gag fusion protein. The binding of SW8.4 RNA (squares) was compared to SW8.4pt.mut. RNA (diamonds) and to the original, unselected (round 0) RNA pool used for SELEX (circles) in the study reported previously (26). Labelling of the RNA and conditions for filtering binding were exactly as previously described (26).

The available evidence thus strongly supports the view that the capacity to bind Gag is necessary and sufficient for ψ activity. Formal proof of this hypothesis, however, would require that a heterologous Gag-binding RNA, unrelated to any known ψ locus, could be shown to support retroviral packaging. Recently, we and others have utilized the SELEX technique to isolate, from a large, random starting population of RNA oligomers, those RNA species that are bound preferentially by HIV-1 Gag in vitro (2, 3, 26). Among the ligands we identified was a 50-base RNA, termed SW8.4, that has no evident sequence similarity to any part of HIV-1 but which specifically binds the NC region of HIV-1 Gag in vitro with an affinity comparable to that of the authentic HIV-1 ψ (26). In the present study, we tested whether this heterologous Gag ligand could exhibit packaging activity in cis.

The sequence and predicted secondary structure of a slightly truncated form of the original SW8.4 ligand (26), shown in Fig. 1A, are notable for the presence of a stable stem structure and an adjacent purine-rich single-stranded region—features that appear to promote recognition by Gag (26). Remarkably, sequence comparison of SW8.4 with an HIV-1 Gag-NC SELEX ligand isolated independently by Berglund et al. (SelPsi) (3) reveals a seven-base region of identity between both ligands (GUGGUGC; shown in open lettering in Fig. 1A). Nitrocellulose filter binding studies (Fig. 1B), conducted in the presence of excess nonspecific RNA, revealed that SW8.4 specifically bound HIV-1 Gag with an apparent K_d of approximately 1 to 10 nM, whereas the starting pool of random RNAs from which SW8.4 was originally derived showed no detectable binding. Moreover, a mutant form of SW8.4 (termed SW8.4pt.mut.) in which the sequence and structure of the seven-base conserved region were disrupted by transposing a single C residue showed markedly diminished binding, suggesting that this region is relevant for Gag recognition. Further evidence of the importance of this seven-base sequence comes from Berglund et al. (3), who showed that altering it reduces the in vitro affinity for Gag-NC by 400-fold, and from Fisher et al. (16), who found that NC binds preferentially to deoxyoligonucleotides containing oligo(dTG) repeats.

To test the biological activity of SW8.4, we utilized the proviral clone HIV-gpt, in which a portion of the env sequence has been deleted and replaced by a selectable drug resistance marker. When transiently transfected into 293T cells along with an amphotropic murine leukemia virus env vector, HIV-gpt expresses genomic RNA that is efficiently encapsidated and released from the cells within pseudotyped virions (22, 33). For this study, we compared the properties of parental HIV-gpt to those of seven mutants. Three constructs (ΔSL1, dS.SL1, and dS.SL3) are previously described, packaging-defective viruses which harbor mutations deleting the entire SL1 element or disrupting the base pairing in the SL1 or SL3 stems, respectively (10). Four additional constructs were prepared by replacing either SL1 or SL3 with the SW8.4 or SW8.4pt.mut. sequences, creating the proviral constructs SELEX.SL1, Mut-SELEX.SL1, SELEX.SL3, and Mut-SELEX.SL3. We prepared virus with each of these constructs and quantified the viral RNA content of the virions and the transfected cells that produced them with a previously described RNase protection assay (9, 10). In this assay, a given viral RNA population yields three major protected fragments upon polyacrylamide gel electrophoresis: the largest band corresponds to genomic RNA, an intermediately sized band corresponds to the spliced viral RNAs, and the smallest band corresponds to the 3′ repeat sequences which are present in all viral transcripts, thus providing an internal control for the total amount of virion-derived RNA.

As shown in Fig. 2A and Table 1, the cytoplasm of cells transfected with parental HIV-gpt contained roughly equimolar amounts of spliced and unspliced viral RNAs. Comparable ratios were also observed in cells transfected with dS.SL3, Mut-SELEX.SL3, and SELEX.SL3, indicating that these mutations did not greatly perturb viral RNA expression in the cytoplasm (Fig. 2A; Table 1). Similarly, levels of total or particle-associated p24 capsid antigen released into the cytoplasm and of exogenous reverse transcriptase activity were approximately the same for each mutant as for the parental virus (Table 1).

FIG. 2.

Quantitative RNase protection assays. Cytoplasmic (cyto.) (A) or virion-derived RNAs (B and C) were annealed to an excess of radiolabelled riboprobe and treated with single-strand-specific RNases, and the resulting protected fragments were separated on denaturing polyacrylamide gels. All RNAs containing mutations within SL1 (including SELEX.SL1 and Mut-SELEX.SL1) were annealed to mutant-specific riboprobes, whereas all SL3-altered virions and HIV-gpt RNAs were annealed to a wild-type riboprobe. For each construct, the largest protected fragment corresponds to genomic sequences, the second major fragment corresponds to spliced sequences, and the smallest fragment (of invariant size) corresponds to the 3′ viral repeat sequences. All riboprobes were also mixed with 2 μg of Escherichia coli tRNA and subjected to the assay with (−) or without (+) RNase treatment. Represented in each panel is an aliquot (1/20) of the wild-type probe minus RNase. Assays were performed exactly as previously described (9, 10). MW, molecular weight markers (indicated in nucleotides at the left).

TABLE 1.

Summary of mutant phenotypes

Construct (parental virus and mutant)	Mean p24 concn (ng/ml)a	% Particle-associated p24b	RT/p24 ratioc	% Genomic RNA contentd	Ratio of genomic RNA/spliced RNA in:		Infectivity (CFU/ng of p24)g
					Virione	Cytoplasmf
HIV-gpt	297 ± 84	59 ± 15	100	100	15.2 ± 3.1	ND	1,975 ± 99
ΔSL1	267 ± 57	48 ± 6	94	38 ± 16	0.9 ± 0.1	ND	128 ± 7
dS.SL1	266 ± 85	43 ± 7	104	34 ± 12	0.9 ± 0.1	ND	138 ± 11
SELEX.SL1	288 ± 44	51 ± 4	88	142 ± 78	2.0 ± 0.0	ND	453 ± 84
Mut-SELEX.SL1	211 ± 46	42 ± 6	78	53 ± 4	1.2 ± 0.1	ND	187 ± 7
HIV-gpt	371 ± 110	52 ± 9	100	100	14.7 ± 2.5	1.2 ± 0.2	784 ± 21
dS.SL3	215 ± 34	69 ± 9	102	52 ± 10	1.1 ± 0.1	0.9 ± 0.1	145 ± 11
SELEX.SL3	257 ± 25	46 ± 10	67	112 ± 23	13.5 ± 2.5	0.7 ± 0.0	943 ± 121
Mut-SELEX.SL3	172 ± 41	56 ± 13	105	110 ± 49	2.9 ± 0.9	0.9 ± 0.1	441 ± 79

^aAverage p24 concentrations in viral stocks from at least three independent transfections ± standard errors. 

^bPercentages of p24 which were recovered from viral stocks after pelleting through a 20% sucrose cushion ± standard errors. 

^cReverse transcriptase (RT) activity in viral stocks relative to the concentration of p24, with the ratio from HIV-gpt as 100; data from one representative experiment is shown. 

^dPercentages of genomic RNA in virions relative to the parental HIV-gpt virus ± standard errors. 

^eRatios of genomic to spliced viral RNAs extracted from pelleted virions [defined as (arbitrary Phosphorimager units in genomic RNA/arbitrary Phosphorimager units in the spliced RNA band)] ± standard errors. 

^fRatios of genomic to spliced viral RNAs extracted from the cytoplasm of transfected cells (defined as above) ± standard errors. ND, not determined. 

^gInfectivity, based on the number of CFU per nanogram of p24 on HOS cells ± standard errors. 

Measurement of the total amount of genomic RNA packaged by the various mutants confirmed (8) that deleting or disrupting either SL1 or SL3 alone decreased packaging by 50 to 65%; inserting the SELEX sequence in place of either of these elements appeared to increase total packaging, but the variability among experiments was too large to support strong conclusions (Table 1). Significant differences were apparent, however, when we compared the amounts of genomic and subgenomic RNA species within the virions (Fig. 2B; Table 1). HIV-gpt virions contained almost exclusively genomic RNA, indicating a high degree of selectivity. By contrast, dS.SL3, the mutant in which the SL3 stem was disrupted, packaged both genomic and spliced RNAs in equimolar ratio, indicating a complete loss of selectivity (Fig. 2B; Table 1). Virions in which SL3 was replaced by the non-Gag-binding SW8.4-mut1 (mut-SELEX.SL3) were likewise indiscriminate in packaging. However, the construct SELEX.SL3 containing the optimal SELEX ligand SW8.4 in place of SL3 packaged its genome with the same fidelity as parental HIV-gpt, incorporating only minimal amounts of spliced viral RNA into its virions. These results, summarized graphically in Fig. 3A, indicate that SW8.4 can substitute functionally for the native ψ component SL3 to support specific HIV-1 packaging and that this activity correlates with its ability to bind HIV-1 Gag and NC proteins. It should be emphasized that this activity is most evident in the discrimination between genomic and subgenomic transcripts, rather than in total genomic RNA content, and that our results do not directly address the discrimination between viral and nonviral RNAs.

FIG. 3.

Quantitation of RNA packaging specificities and viral infectivities. (A) Ratio of genomic RNA to spliced RNA packaged in each mutant compared to that of the parental virus (HIV-gpt), based on quantitation of the genomic and spliced bands with the RNase protection assay and Phosphorimager analysis. (B) Infectivities of the constructs on HOS cells, expressed as the number of gpt⁺ CFU per nanogram of viral p24 antigen. Assays of viral stocks from at least three transfection and infection assays yielded similar results. Colony formation assays were performed exactly as previously described (9, 10). Error bars, standard error.

As a further test of SW8.4 function, we tested the capacity of these same viral supernatants to infect human HOS target cells in culture (Fig. 3B; Table 1). Owing to their disrupted env gene, pseudotyped virions of HIV-gpt can support only a single round of infection but confer on each infected cell the ability to form a colony in selective medium; enumeration of these colonies thus provides a quantitative measure of viral infectivity. Using this approach, we found that disruption of the SL3 stem reduced infectivity by roughly 90%, as previously reported (10). Inserting SW8.4 in place of SL3 restored infectivity to fully wild-type levels, whereas SW8.4pt.mut. was significantly less infectious. Thus, the heterologous Gag-binding SELEX ligand SW8.4 can functionally replace SL3 in the HIV-1 ψ locus to support both RNA packaging and viral infectivity.

When we initially characterized the SW8.4 ligand, we noted the presence of a palindromic sequence, GGUGCAUC, near its apex (Fig. 1A). This sequence is reminiscent of the palindrome found in the SL1 loop, which is critical for the process of genomic RNA dimerization (11, 23, 24, 28, 31, 32, 34, 37). SL1 has been shown to facilitate HIV-1 dimerization through a kissing-loop interaction in which the palindrome forms the initial contact point between two RNAs (11, 23, 24, 28, 31, 32, 34, 37). In particular, we noted the central GC dinucleotide which we had previously shown to be critical for in vitro HIV-1 RNA dimerization (11), as well as the alternating purines and pyrimidines which occur in the SL1 loop palindromes of all replication-competent strains of HIV-1. We therefore asked whether SELEX ligand SW8.4 could functionally replace the SL1 element. We first undertook studies to determine if our SELEX ligand could, like SL1, mediate spontaneous RNA dimerization in vitro. RNA transcripts were prepared that spanned HIV-1 residues 637 to 829, including either the wild-type SL1 element at its normal location or either SW8.4 or SW8.4pt.mut in its place. These radiolabelled RNAs were then incubated under conditions which either did or did not favor dimerization (11). As we had observed previously (11), wild-type HIV-1 RNA readily dimerized under these conditions in the present study (Fig. 4). However, HIV-1 RNA sequences harboring the SELEX ligands at the SL1 position showed no detectable dimerization under these same conditions (Fig. 4). These results indicate that the SELEX ligands cannot mediate dimer initiation in vitro. We then tested whether these ligands could substitute for SL1 in virions. As previously reported (10), deletion or disruption of SL1 in HIV-gpt (ΔSL1 and dS.SL1, respectively) produced defects in both packaging (Fig. 2C and 3A; Table 1) and infectivity (Fig. 3B; Table 1) that were similar to those produced by the SL3 disruption (dS.SL3); these were not appreciably corrected by inserting the SELEX mutant (mut-SELEX.SL1). Inserting the wild-type SELEX sequence in place of SL1, however, produced only modest increases in packaging specificity (Fig. 2C and 3A; Table 1) and infectivity (Fig. 3B; Table 1) that remained well below the levels achieved by this sequence in the SL3 position. The heterologous RNA thus can substitute fully for SL3 but only partially for SL1. These results are consistent with the idea that SL1's contributions to packaging extend beyond its ability to interact with Gag and NC proteins and further support the idea that in vitro dimerization activity reflects an important biological function of the SL1 locus.

FIG. 4.

In vitro dimerization of HIV-1 RNAs containing altered sequences at the SL1 position. Radiolabelled, synthetic RNAs representing HIV-1 residues 637 to 829 and harboring a wild-type, SW8.4, or SW8.4pt.mut. sequence at the SL1 position were tested for their ability to dimerize spontaneously under conditions which favor monomer (low-salt) or dimer (high-salt) formation, exactly as previously described (11). To achieve full denaturation, these constructs were also assayed immediately after a 2-min low-salt incubation at 90°C.

Our group and two others have reported using SELEX to derive high-affinity RNA ligands for the HIV-1 Gag and/or NC proteins (2, 3, 26). As noted above, the optimal ligands evolved by our group (26) and by Berglund et al. (3) share a number of properties in common, whereas we could not detect any sequence or structure similarity between our ligands and those of the third group (2). This disparity may reflect differences in the conditions under which SELEX was performed: whereas all three groups performed their Gag-RNA binding experiments with comparable salt concentrations, we (26) and Berglund et al. (3) utilized a pH of 7.4 to 7.5, whereas the other group (2) used a pH of 5.3.

To our knowledge, this report is the first to show that an in vitro-evolved HIV-1 RNA ligand with affinity for Gag has biological packaging activity in the context of a provirus. This finding suggests that it may be possible to create an entirely artificial retroviral packaging element that could function nearly as well as a wild-type ψ signal. Such an element would be of particular value in the design of retrovirus-based gene therapy vectors, in that they would not be expected to readily undergo homologous recombination with wild-type viral sequences and so would help minimize the potential for generating replication-competent recombinants.